www.pharmaguideline.

com

HPLC Troubleshooting
1. Column Life-Time

27. Optimal Flow Rates

2. Variable Retention Times

28. Carbon Load

3. Drifting Retention Times

29. pH Control

4. Column-to-Column and Batch-to-Batch

30. Mobile Phase Composition

Reproducibility

31. Column Contamination

5. Sample Preparation Problems

32. Method Verification

6. Sources of Peak Tailing

33. Double Peaks in Sugar Separations

7. Normal-Phase Chromatography

34. Method Control

8. System Volume, Dead Volume, Dwell Volume

35. Mobile Phase pH

9. Transfer of Gradient Methods

36. Signal-to-Noise Improvements

10. Clogged System

37. Overload

11. Column Plate-Count

38. Column Durability

12. Column Backpressure

39. Fast Analysis and Column Backpressure

13. Peak Area Fluctuations

40. Column Backflushing

14. Ghost Peaks

41. Selectivity Shift

15. Dependence of Retention Times on pH

42. New Method

16. Column Equilibration

43. Negative Peaks

17. Column Conditioning

44. Tricky, Tedious, Time Consuming

18. Complex Sample Matrices

45. Fast Separations

19. Hydrophobic Collapse

46. Buffers for LC/MS

20. Baseline Noise

47. Alkaline Buffers for RPLC

21. Narrow-bore Columns

48. Post-Column Derivatization

22. Sample Solvent

49. Gradient Dwell Volume

23. Gradient Scaling

50. Buffer Capacity

24. Column Storage

51. Flow Rate Changes and Quantitation

25. Paired-Ion Chromatography

52. Analysis of polar compounds

26. Hydrolytic Stability of Reversed-Phase

Packings

1. Column Life-Time
Q.: My column lasted only for about 100 injections. After
that time, the peaks became distorted and the plate-counts
were very low. Whats wrong?
A.: 100 injections is indeed a short life-time. Under normal
circumstances, one can expect a column to be in service
much longer. In order to determine what is wrong, we need
to establish first, if short column life is the rule for your
application or not.
There are two fundamental cases:
1. previously columns used for the same assay
lasted much longer.
2. all columns used for this application die
after about the same amount of use.
In the first case, one would explore if the assay has
remained truly constant. Has the sample composition
changed? Strongly adsorbed contaminants in your sample
can destroy column performance. Are the seals in the fluid
path of your instrument in a good condition? Shedding seals
can clog column filters and the top layers of the packing and
thus effect the distribution of the sample.
If one can be reasonably assured that there are no changes
in the chromatographic conditions, one can safely assume
that the cause of the problem is a mechanical weakness of the
packed bed. This can be induced by rough handling of the
column in your lab (did you drop it?) or during shipment, or
it could be a manufacturing defect. Such a defect can not be
detected by standard column QC and could show up only
after some use of the column. In this case, column
manufacturers will replace your column free of charge.
Q.: That is nice of these manufacturers, but this is not my
problem. My columns always last only a short time.
Sometimes its 100 injections, sometimes 200. I could live
with 200 injections, but only 100 is not good enough. This
really is getting expensive. What can I do?
A.: I agree with you 100%. What we need to do together is
to find the cause of your problem and then see, what we can
do about it.
The most likely cause of your problem is adsorption of
sample constituents on the top of the column. They may
either precipitate because of a low solubility in the mobile
phase or they may be strongly adsorbed. As you inject more
and more samples, these contaminants build up on the top of
the column and prevent the sample to properly adsorb and
distribute. This results in a distortion of the peak profile.
Often this problem is accompanied by an increase in back
pressure.
Q.: OK, that could be it. How do I get around this problem?
A.: There are several ways to prevent this from happening.
One is to clean up the sample with a suitable sample
preparation technique. Solid phase extraction using a SPE

cartridge with a similar chemistry as the separation column

works well for this problem.
Another and more powerful approach is to use a guard
column. The precolumn serves as a sacrificial column top
that is replaced when the problem occurs. For best
performance, you should use a guard column that contains
exactly the same packing as the analytical column and is
packed with the same high performance packing technique as
the analytical column. If you use precolumns made with a
different brand of packing, you will not get the optimal
performance both in separation capability and in protection
of your analytical column. Also, do not use a larger particle
size. Larger particles or badly packed precolumns can result
in a deterioration of the separation due to band- broadening
in the precolumn.
Q.: To use a guard column sounds ok. Do you have any
other solutions?
A.: Well, not really. There are a few other possibilities, but
they all have their drawbacks.
I am not an advocate of column "washing" with solvents
that are supposed to dissolve the contaminants on the top of
the column. In many cases, this process simply does not
work. For example, if the contaminants are proteins that have
precipitated on the column top, by the time you try to wash
them off they have aged a lot by denaturation and maybe
even cross-linking that it may be impossible to solubilize
them again. Furthermore, every washing will also remove
hydrolyzed bonded phase, which otherwise remains in a local
equilibrium at the site where the hydrolysis occurred.
Consequently, a repetitive washing can actually result in an
accelerated aging of the column. Also, after this washing you
have to re-equilibrate your column with the mobile phase,
which in some cases like in ion-pair chromatography may be
quite time consuming.
Another approach that is often tried is column
backflushing. If you do it with a different solvent than your
mobile phase, the same objections hold as for column
washing. If you do it just with mobile phase, it will take a lot
of time until the contaminants are removed or it may not
work at all. Also, any backflushing weakens the column.
Although todays columns are packed well enough that they
can withstand backflushing, I would still recommend to not
make this a standard practice.
Q.: So your best recommendation is to use a precolumn?
A.: Absolutely. They dont cost that much - depending on the
brand and type between $ 10 and $ 50 - and they protect a
column that usually costs about ten times as much.
Furthermore, they protect your column also from other
sources of contaminants that may be more difficult to trace.
The source of your column problem could for example be
dust in the mobile phase or debris from shedding pump seals.
Or it could be the adsorption of contaminants from the
mobile phase. Some of these may be very difficult to

www.pharmaguideline.com

troubleshoot, but the guard column will simply take care of

them.
Q.: Are there any other causes of short column life?
A.: Yes, but they are less likely if you follow the
manufacturers recommendations.
One possibility is that the column is collapsing due to a
mobile phase pH outside the recommended range. This can
also be caused by a sample dissolved in a strongly acidic or
alkaline solution.
Furthermore, there are a few items that are specific to
certain columns.
Amino columns for example react with aldehydes and
ketones. Amino columns in an unbuffered aqueous solution
generate a strongly alkaline pH that leads to a partial
dissolution of the silica.
Columns can also collapse when exposed to the wrong
solvents. The reason for this is that columns are
fundamentally very loose structures. They are partially held
together by the adhesion of the particles to each other. When
you put them into mobile phases that break this adhesion,
there is an increased chance of a collapse of the bed. This
happens occasionally to CN columns in solvents of
intermediate polarity or to Phenyl columns in THF.
One more reason to stay away from column "washing".

2. Variable Retention Times

Q.: The retention times of my peaks are inconsistent. What is
the problem?
A.: First, we need to find what the pattern of this variation is.
This will tell us a lot about the potential source of the
problem. If the retention times change randomly from one
run to the next, then I would first check out the pump(s) and
the solvent mixing devices. To verify that the pumps are
working, measure the flow-rate with a graduated cylinder and
a stopwatch. To verify that the mobile phase composition is
not changing, you can add a tracer to one of the solvents and
observe the baseline. If the mobile phase composition is
constant, the baseline will be steady. If there is a shift in the
composition, you will observe corresponding shifts in the
baseline. For example, if you use a reversed phase method
with UV-detection, you can add 0.1% acetone to the organic
solvent and monitor the baseline at 254 nm. Alternatively,
you can prepare the mobile phase manually and bypass the
solvent mixing devices. If the retention time fluctuations go
away, the source of your problem was the mixing device.
Q.: The retention times are pretty consistent from run to run,
but they vary from day to day.
A.: In this case, the instrument itself is less likely to be the
culprit. The most likely source of variation is the

composition of the mobile phase. In reversed phase

chromatography, there is an exponential relationship between
the retention factor k and the volume fraction of the organic
solvent in the mobile phase. As a rule of thumb, if you make
an error of 1% in the amount of organic solvent, the retention
time can change by between 5% and 15%, typically by about
10%. This means that you have to measure the amount of
solvent very carefully. The best approach is to prepare the
mobile phase gravimetrically rather than volumetrically.
Also, how you degas your mobile phase may contribute to
variability. The best degassing method is the simultaneous
application of vacuum and ultrasound for about one minute.
This results in good degassing with a minimal amount of
evaporation of the solvent. An alternative good method is the
helium sparging method. After the initial equilibration of the
mobile phase with helium, the flow of helium should be
turned down. Otherwise the helium will carry solvent vapors
with it and the solvent composition can change due to
evaporation.
If your sample constituents are ionic or ionizable, then the
control of the pH of the mobile phase is very important. A
change of as little as 0.1 pH units can result in a retention
time shift of 10%. So it is very important to measure the pH
accurately and to keep the pH meter well calibrated. In
reversed phase, the retention of acids decreases and the
retention of bases increases with increasing pH.
By the way, my chemistry teacher always said: "A buffer is
called a buffer because its supposed to buffer the pH. If it
doesnt do that, it isnt a buffer." For example a solution of
ammonium acetate is just that, a solution of ammonium
acetate. An acetate buffer contains the acetate ion and acetic
acid. If they are present in equal amounts, the resulting pH is
the pK of the buffer, 4.75 in the case of acetate. A buffer
always has its highest buffering capacity at its pK.
Q.: I did not realize how closely one needs to control the
mobile phase composition to get reproducible results. Are
there any other variables that one must pay attention to?
A.: Another important factor is temperature. You would
suspect that temperature is the cause of the fluctuations, if
the retention of all peaks is moving in the same direction.
The rule of thumb is that retention times change by about 1%
to 2% per 1 Celsius. That isn't that much under normal
circumstances, but in many places the heat or air
conditioning are shut down over the weekend. And then you
wonder why the analyses run automatically over the weekend
show different retention times than the ones run during the
week. This can obviously be avoided when a column
thermostat is used.
In reversed phase chromatography using ionic or ionizable
sample compounds, the retention is also influenced by the
ionic strength of the buffer, but the influence is so small that it
is usually negligible. Typical changes are under 1% for a
20% change in molarity of the buffer. Since the buffer
constituents are always weighed, there is no way that such a
large error could happen.

Q.: Are there any special cases where additional parameters

need to be considered?
A.: Yes. In ion-pair chromatography, the concentration of
the ion-pairing agent influences the retention of ionic sample
constituents. The retention of analytes that are oppositely
charged to the ion-pairing agents increases, the retention of
analytes of equal charge decreases. Neutral compounds are
practically unaffected. At low concentration, i.e. below 5
mM/L the retention of the sample constituents that interact
with the agent changes in proportion to the concentration of
the ion-pair reagent. At high concentration of the ion-pairing
reagent, i.e. around 10 mM/L, the surface of the adsorbent is
saturated with the reagent, and a change in concentration of
the reagent does not result in an appreciable change in the
retention time of the analytes. This is something to think
about during methods development. If you can make the
method insensitive to one of the variables, you should do so.
In normal phase chromatography the retention times are
very sensitive to the concentration of polar constituents in the
mobile phase. One of these constituents that is always there
whether one wants it or not is water. So variable amounts of
water can lead to variable retention. One trick that has been
used to get around this problem is to use non-polar solvents
like hexane or methylene chloride that are half saturated with
water. To accomplish this you take a volume of the solvent
and saturate it with water by adding some water to it and stir
it for a while, then you mix this water-saturated solvent and
mix it with an equal volume of "dry" solvent. This procedure
results in reasonably good reproducibility of the watercontent of the non-polar solvents. It also helps to prevent
drifting retention times, but this is the subject of a future
section.

3. Drifting Retention Times

Q.: The retention times of the peaks in my chromatogram are
drifting. What is the problem?
A.: This is an interesting problem. There could be many
different causes for this. Let us go through them one by one.
People most commonly assume that drifting retention times
are an equilibration problem. If you are doing normal phase
chromatography on unmodified silica columns this is also the
most likely problem. The retention times in normal phase
chromatography are very susceptible to the amount of water
adsorbed on the silica surface, which in turn is a function of
the water dissolved in the mobile phase. Since the solubility
of water in solvents like hexane or methylene chloride is
extremely low, it takes a long time for the columns to
equilibrate. I have seen cases were the retention times in very
dry hexane were still shifting after a week of equilibration. I
therefore recommend to avoid very dry solvents. A common
solution to the equilibration problem of silica with water is to
use solvents that are "half-saturated" with water. They are
prepared by saturating a given volume of hydrophobic

solvent with water and then mixing it 1:1 with "dry" solvent.
This approach speeds equilibration times up tremendously.
In reversed-phase chromatography, equilibration is usually
very fast. A few (5 to 10) column volumes of mobile phase
are usually sufficient for equilibration. This is however not
always the case. A notable exception is the equilibration of a
column with ion-pairing reagent in ion-pair chromatography.
The ion-pairing reagents are typically used at a concentration
of about 2 to 5 mmol/L or less. They adsorb onto the surface
of the reversed-phase packing at a surface concentration of
between 0.5 to 2 mol/m2. A 4.6mm x 250mm column
contains about 3 g of packing with a surface area of 330
m2/g. That means that there are 1000 m2 of surface in the
column. At a surface concentration of 2 mol/m2 2 mmol of
reagent need to be pumped into the column for complete
equilibration. At a concentration of 2 mmol/L this takes a
liter of mobile phase. This is clearly an extreme case, but
don't be surprised if it takes a few hundred milliliters of
mobile phase to equilibrate the column. It is therefore unwise
to convert columns used in ion-pair chromatography to an
organic solvent for overnight storage, because during this
process the ion-pairing reagent is washed off and the next
day you have to go through a lengthy reequilibration
procedure.
Q.: Both of these phenomena have a common factor: low
mobile phase concentrations of an agent that is strongly
adsorbed. Is this the most common cause of drifting retention
times?
A.: I would say yes. Other phenomena are not as common.
But the strongly adsorbed agent could also come from the
sample. In this case it would slowly build up during repeated
injections and thus change the chemical properties of the
column. An example of this would be the adsorption of
excipients from a drug sample.
One can tell whether a contamination is coming from the
mobile phase or from the sample by looking at the rate at
which retention times are changing.
An experiment could be as follows:
1. Inject the sample several times, e.g., four times for a
total run-time of 1 hr.
2. Pump mobile phase for the same amount of time without
injecting sample.
3. Repeat the first step.
4. Plot the retention times
a, versus time
b, versus number of injections.
If the first plot gives you a smooth curve, then the mobile
phase is the carrier of the contamination. If the second plot
yields a smooth curve, the sample is the source of the
contamination.
Q.: Are there other causes of shifting retention times?
A.: Yes. It is possible that the mobile phase composition is
changing with time. If you are not using the on-line mixing

capabilities of todays instruments, you may be looking at a

slow evaporation of a component in the mobile phase. This is
especially true when you are sparging the mobile phase with
helium to avoid air-bubbles in the pump. One should keep
the rate of sparging to a minimum.
An often underestimated cause of shifts in retention is the
hydrolysis of the bonded phase. The manufacturers usually
specify a pH-range, outside which the bonded phase is
"unstable". This range is typically from pH 2 to 8 or 9.
However, one has to treat these limits with a lot of caution.
There is not a sharp boundary, hydrolysis depends also on
other factors like temperature and organic solvent, and slow
hydrolysis occurs inside these limits as well.
The hydrolytic stability of a bonded phase is best at
intermediate pH-values, around pH 3 to 5 and at low
temperature. Isocratic chromatographic conditions are better
than gradients. While hydrolysis does occur in isocratic
conditions as well, the bonded phase often adsorbs to itself
and is in a local equilibrium. However, when a higher
concentration of organic solvent is used - as in a gradient or
during column cleaning procedures - this local equilibrium is
interrupted and the hydrolyzed bonded phase is washed out
of the column.
Q.: I will keep this in mind. Can temperature changes cause
drifts of retention times?
A.: Yes, temperature is always suspect. If you run your
samples in unattended operation overnight or over the
weekend, you may get drifting retention times according to
the shifts in lab temperature. In many places, the ambient
temperature is maintained at a different setting during the
night or on the weekend to conserve energy. As a rule of
thumb, the retention times shift by about 1% to 2% per 1 C.
Related to the last phenomenon are shifts in retention times
that are caused by an increase of back-pressure in the
column. Increasing back-pressure may indicate a
contamination of the column, but even a clogged frit can
affect retention times. The pressure needed to push the
mobile phase through the frit warms up the mobile phase by
friction, and this increase in temperature can affect the
retention times.
There are other causes of drifting retention times, but they
are exceedingly rare.
A slow equilibration phenomenon encountered in
reversed-phase chromatography is the equilibration with
mobile phases that contain less than a few % of organic
solvent. "Fully end-capped" packings with a high coverage
of C18 are not wetted well by these mobile phases. This
leads to several phenomena from loss of contact of the
mobile phase with the stationary phase to "hydrophobic
collapse", which is the reduction of the surface area available
for interaction with the sample due to self-adsorption of the
bonded phase onto itself. The columns can be regenerated
rapidly using a few column volumes of organic solvent,
preferentially the organic modifier in your mobile phase.
This phenomenon is less pronounced or even not

encountered at all with bonded phases that are not endcapped.

4. Column-to-Column
Reproducibility

and

Batch-to-Batch

Q.: I am about to start the development of a new HPLC

method. After validation, the method will be transferred to
the QC-lab, where it will be used for many years. I am
concerned about the long-term reproducibility of the method,
especially about the long-term reproducibility of the column.
What can I do to assure good reproducibility?
A.: First of all, let me compliment you for thinking about this
aspect of your method before you start working on it. If it
can be anticipated that a new method will be used for several
years, this knowledge should be part of the column selection
process. You may want to assure yourself that the column
will be available for the anticipated life-time of the method.
This means that you want to select a manufacturer that has a
large operation and is likely to stay in business for the
foreseeable future. In addition, you want to select a
"standard" surface chemistry - like C18 or C8 - over novel
and/or less used surface chemistries. In other words, your
column selection should be conservative.
The next criterion should then be the reproducibility of the
column. If you or your colleagues have had good experience
with the reproducibility of a particular column in the past
that is a good starting point. You should also consider the
information that is available from the manufacturer. Recently
some column manufacturers have started to publish their
specifications and the results of their batch-to-batch
reproducibility studies. You can obtain this information from
the manufacturers and review it critically.
Q.: What am I looking for in this information?
A.: Let me back up a little and explain the various aspects of
column-to-column and batch-to-batch reproducibility. In the
following, I will talk about reversed-phase packings, since
they are the most commonly used.
If you are looking at different columns that are made from
the same batch of packing, you may get differences in
column plate-count (peak-width), back-pressure and
retention times. The retention times change in proportion to
the packing density and the volume of the column, the latter
depending mostly on the reproducibility of the internal
diameter of the column. This variability should not be
detrimental to your method, since the retention times of all
peaks change proportionally to each other and resolution is
preserved. However, you need to be aware of this so that you
specify the elution time windows for your method correctly.
Variation of retention time due to this effect is typically 3%
RSD, but can be as little as 1% RSD, if the column
manufacturer has good control over the column hardware.

www.pharmaguideline.com

Variation in column plate-count can affect your method,

especially when you are dealing with peak-pairs that are
barely baseline resolved. A typical reproducibility of platecount is +/- 10%, although standard deviations of as little as
2 to 3% have been achieved. Remember that plate-count is a
squared number; resolution depends only on the square root
of plate-count. Therefore a variation of 2% in plate count
means that the peak-width varies by only 1%. You should
check if the column manufacturer has an upper and lower
limit for plate-count or only a minimum specification. Twosided specifications are more desirable.
Column back-pressure is very reproducible for any given
packing procedure. The variation within a given batch of
packing should be under +/- 5%.
Q.: How about batch-to-batch reproducibility?
A.: If you are looking at the reproducibility of
chromatographic parameters over different batches of
packing, the same parameters as above can vary. But more
importantly, the selectivity of the separation can change with
the batch of packing. This means that the relative retention of
peaks can vary, i.e. their relative position in the
chromatogram. This can be detrimental to the specificity of
your method.
This variation is the most important issue that you want to
guard yourself against. Therefore what you want to obtain
from the column manufacturer is information about how
batch-to-batch reproducibility is ensured. This usually entails
some measurements of the physical, chemical and
chromatographic properties of a packing. Among the
physical measurements, the most important one is probably
the specific surface area of the packing, since a variation of
this parameter directly translates into variation of retention
times. Therefore you would like to understand what
variability of the specific surface area the manufacturer finds
acceptable. One may commonly find a range of +/- 10%, but
+/- 5% is achievable.
As the most important chemical parameter, you would like
to understand the variability of the surface coverage for the
bonded phase. This is customarily expressed as mol/m2 of
bonded phase. Many manufacturers don't give that parameter
directly, but have specifications on the carbon content of the
packing. What one would like to see are once again twosided specifications and a tight range. Better than +/- 10%
may be quite common, while less than +/- 4% is achievable.
Finally, you would like to understand the chromatographic
reproducibility test that the manufacturer of the packing uses.
Frequently, columns are accompanied with a chromatogram
of a mixture of simple neutral compounds, like toluene,
naphthalene and anthracene. This is not a good batch-tobatch reproducibility test. The relative retention between
neutral hydrophobic compounds is extremely reproducible
and very insensitive to variations in the surface coverage of
the packing. Better batch-to-batch reproducibility tests
incorporate strongly basic compounds in the test mix. Wellselected basic compounds in a well-designed test interact

predominantly with the "residual" silanols on a packing,

while neutral compounds interact predominantly with the
bonded phase. Thus the relative retention between wellselected basic compounds and neutral compounds represents
a stringent test of batch-to-batch reproducibility of a
reversed-phase packing. If a manufacturer uses such a test for
batch-to-batch reproducibility testing and has reasonably
tight specifications, your comfort level should increase.
Q.: I am not sure how the manufacturer's tests relate to my
compounds in my assay. What can I do to assure good batchto-batch reproducibility for my assay?
A.: You are absolutely correct: the information given above
is a starting point that allows you to select a column that has
a good chance of being reproducible. But the ultimate test is
a test of the reproducibility for your assay itself. Some
manufacturers provide kits for this purpose that comprise
columns from different batches of packing. Other
manufacturers supply columns prepared from different
batches on demand. Make sure you understand what you are
getting. Different column lots containing the same batch of
packing are not useful, you need to get columns packed with
packing from different bonding reactions. There can be some
confusion if the manufacturer considers columns packed on
different days with the same batch of packing to comprise
different lots. So it is best to ask twice, if the manufacturer is
really supplying you with what you want.
You probably want to obtain three different batches of
packing for testing the reproducibility of your method. If
your assay proves to be unaffected by the batch-to-batch
variability, you can rest reasonably assured that your method
will be reproducible for years to come.

5. Sample Preparation Problems

Q.: I am encountering some problems with my sample
preparation. I am using solid-phase extraction with a
reversed-phase sorbent. Usually the recovery is good, around
80% or better. But sometimes it drops down to 50% or even
less. What could be the problem?
A.: The problem that you have is not unusual. I have found
that most of the time low recovery can be attributed to poor
methods development. Although it may be adequate for the
analytical purpose of your method, 80% recovery is not a
good sign from the standpoint of method ruggedness. It is
highly likely that there is a single reason for the incomplete
recovery that sometimes results in a loss of 20% and
sometimes in a loss of 50%. What we need to do together is
to find out, where the missing 20% to 50% ends up.
Q.: O.K., how do we do that?
A.: There are four possible scenarios:

1. The analyte is lost before the solid-phase

extraction.
2. The analyte is not adsorbed completely during
the adsorption step.
3. The analyte is partially washed out in a washing
step.
4. Part of the analyte remains on the solid-phase
extraction cartridge after the elution step.
To find out whether the analyte breaks through the first
SPE cartridge during the adsorption step, you can simply put
a second SPE cartridge behind the first one and then treat it
the same way as the primary cartridge. If the fraction
obtained from the second cartridge contains a fair amount of
analyte, then the adsorption step is incomplete. There are
several possible reasons that this may occur. First, the
solvent in which the sample is dissolved may be too strong a
solvent. If this is the case, you may want to dilute your
sample with water, or - if your analyte is ionizable - with
buffer. Second, you may verify that you are activating the
SPE cartridge before you load on the sample. Reversedphase SPE cartridges need to be activated with an organic
solvent like methanol or acetonitrile, followed by an
equilibration with water or buffer. Only then the sample
should be applied. Many people try to avoid these tedious
additional steps, but they are necessary for a correct
performance of the method.
Q.: I do activate the cartridge according to the
manufacturers recommendation. So I dont think that this is
the problem. I do like your proposal to use a second cartridge
to verify the completeness of the adsorption step. What do I
do to check the other steps of the method?
A.: To check the washing steps, you should collect all the
fractions and analyze them by HPLC. This may not be easy if
the washing steps are removing compounds that interfere
with the quantitation of the analyte. Then the quantitation of
the analyte in these fractions is by definition difficult. You
may try the following technique to work around this
problem. Evaporate the questionable fractions to dryness,
and reconstitute them in the same solvent composition as the
original sample. Then take this sample and process it on the
solid-phase extraction cartridge in the same way as your
original sample. If you now find another fraction of your
analyte in the elution step, you know that a portion of the
analyte is washed off in one of the washing steps.
Another way to do this is to use standards and run them
through the sample preparation process. This is less rigorous,
since the presence of the matrix may affect the behavior of
the analytes.
Q.: These are some good suggestions. How do I analyze for
analyte that remains on the cartridge after the elution step?
A.: You need to elute the cartridge again with more eluent
after your original elution step. This may elute additional
analyte. Often, you need to increase the elution strength of

the eluent used in the elution step. Think about how a portion
of the analyte may be retained stronger. Could it be due to
interaction with residual silanols? Then you may want to
change the buffer pH or buffer strength in your elution step.
Is it due to hydrophobic interaction? Then increase the
concentration of organic solvent in the elution step or go to a
stronger organic solvent (e.g. replace methanol with
acetonitrile). Is it possible that the sample interacts with
residual silanols via hydrogen bonding? Then the addition of
methanol to acetonitrile or tetrahydrofuran might help.
Q.: O.K. If I dont find the missing fraction in all of these
steps, can I exclude the solid phase extraction step as the
source of the problem?
A.: Yes. Now we need to explore, if the analyte gets lost in
another step in the sample preparation. Could it be that it is
strongly adsorbed to a sample vessel. A strongly
hydrophobic analyte may adsorb to the walls of a
polyethylene vial. Strongly basic compounds may bind to the
silanols on the surface of glass vials. It is also possible that
the analyte could adsorb to solids in the matrix or bind to
other constituents in the matrix (e.g. proteins). These
problems may be more difficult to isolate.
Fundamentally, try to get the recovery of the analyte as
close to 100% as possible. This is a good assurance that your
method is rugged from the start. Even then, circumstances
may arise that results in irreproducible recovery, but they are
less likely than if you start with a method that is already
compromised.
If you can achieve good recovery from the start, then
neither a normal variability of the eluent composition or of
the packing material itself is likely to affect your method.
The only remaining cause of variable recovery could be the
quality of the packed bed of the SPE cartridge. If a void is
formed in the bed, then the flow is non-uniform, and an early
breakthrough of the analyte is possible. You can guard
yourself against this problem by a cursory inspection of the
device. A void that would affect the method is quite obvious.

6. Sources of Peak Tailing

Q.: What can I do to get rid of peak tailing?
A.: First, we will have to find out, where the peak tailing
comes from. There are many sources of peak-tailing, ranging
from column problems to chemistry problems to instrument
problems. The most common reasons for peak tailing are
extra-column band broadening, deterioration of the packed
bed, and interaction of the analytes with active sites on the
packing. Obviously, what needs to be done depends on the
cause of tailing.
Q.: I accept that. Now, how do I determine the cause of
tailing?

A.: A quick first step is a careful examination of the

chromatogram. There is a lot of information in a
chromatogram that can give you clues about the nature of the
problem without any knowledge about the samples or the
chromatographic conditions. You can then use the additional
information to test the hypotheses that you have formed
based on the examination of the chromatogram.
One of the first things to examine is the height of the
peaks. If a UV detector has been used and the peak-heights
are in the order of 1 AUFS, a good guess is that the column
is overloaded and that peak-tailing is due to overload. This
judgement supposes that the extinction coefficients for the
compounds is in the order of 1000, which is a reasonable
rule of thumb. To confirm mass overload, you would then
ask the question of how much mass has been injected onto
the column. For a normal, fully porous packing with a poresize of about 100 , overload starts distorting peaks at a load
of about 100 g.
These are all rules of thumb, but they can be used as
reasonable guidelines for sample overload. You can test for
overload by injecting about 10 x less mass on the column and
see, if the peak-shape improves.
Q.: Ok. But what if tailing occurs at much lower amounts
injected?
A.: Once again, a peek at the peaks in the chromatogram is
helpful. If there are many peaks in the chromatogram,
determine whether the peak-shape remains roughly constant
or if there is a consistent change of peak-shape throughout
the chromatogram. If the peaks are tailing more in the early
part of the chromatogram than in the later part, one would
suspect that extra-column effects are responsible. The
influence of extra-column effects decreases as the peaks
become wider, which is why they distort early peaks more
than later peaks. Two extra-column effects should be
considered:
1. extra-column band broadening
2. detector time-constant.
Band spreading in connection tubing, injectors and
detectors results in tailing in the early peaks in the
chromatogram. You may encounter this, if you put a column
with a small diameter on an instrument that was not
configured for the use of small volume columns, or if you
have recently re-plumbed your system. If the latter is the
case, examine the type of tubing that you have used (it
should be 9/1000" i.d. throughout from injector to detector).
Also inspect all the connections, if they were made properly.
A common cause of extra-column band spreading is the fact
that different column manufacturers use different distances
between the tip of the ferrule and the end of the tubing in
their column end-fittings. An extra-column band spreading
with a standard deviation of only 15 L significantly
influences peak shape up to an elution volume of about 3 mL
on a 5 m column.
A large detector time constant has the same influence.
More tailing is observed on early peaks than on late eluting

peaks. Often, the default setting of the time constant is 1 sec.

If a high-performance 5 m column is used, a distortion of
the peaks can be observed up to 2 to 3 minutes into the
chromatogram. Larger time constants obviously would give
larger effects.
If the peaks are all tailing pretty much to the same degree
for all sample compounds in the chromatogram, there are two
possibilities:
1. the packed bed is damaged
2. all sample components are chemically similar,
and we are dealing with chemical effects.
In the first case, a plate-count test under standard
conditions, especially under conditions recommended by the
manufacturer, will reveal whether the column has
deteriorated or not. The column could have been damaged by
adsorption of contaminants or particles. Sometimes, it is
possible to remove these contaminants with an appropriate
solvent (for example THF on a reversed-phase column), but
if the bed itself has shifted, there is nothing that one can do
to repair the column.
In the case that all peaks are chemically similar, chemical
effects are potential candidates as causes of tailing. If only
some peaks in the chromatogram are tailing and other peaks
give a good peak shape, chemical effects are the prime
candidates as the causes of peak tailing.
Q.: What are these "chemical effects"? Please explain!
A.: There could be several effects here as well, but the most
common cause is the interaction of the analytes with an
energetically non-uniform surface. A typical example is the
tailing of strong bases on some reversed-phase packings.
These kind of compounds interact strongly with residual
silanol groups on the surface of the packing as well as with
the bonded phase. If the silanol groups on the surface form a
non-homogeneous population, tailing can result.
Q.: What can I do to eliminate tailing due to chemical
influences?
A.: First, you should consider using a column that does not
exhibit this phenomenon. Some of the newer reversed-phase
packings have been designed to minimize the tailing of
bases. Second, this tailing can be mediated by either the pH
of the mobile phase or by using organic bases in the mobile
phase that compete with the analytes for active sites. This is
due to the fact that silanophilic interactions are often ionexchange interactions. At acidic pH, fewer silanols are
negatively charged. Therefore, less tailing is observed for
positively charged analytes. As a competing base,
triethylamine is often used. But among competing bases,
those with a larger hydrophobicity often work better.
Examples are octylamine or tetrabutylammonium salts.
In the case that the silanophilic interaction of the analyte
are not ion exchange, but hydrogen bonding, you can
improve peak-shape by changing the hydrogen bonding
character of your solvent. For example, methanol often

www.pharmaguideline.com

results in an improved peak-shape compared to acetonitrile

as organic modifier of the mobile phase.
Similar phenomena are encountered in normal phase
chromatography, and similar reasoning can be applied there
to suppress tailing. Significant changes in tailing can be
observed depending on whether alcohols or acetonitrile are
used as polar modifier of the mobile phase.
There are still some other causes of peak-tailing, but they
are comparatively rare. Adsorption of sample constituents to
column frits or injector parts has been observed. It is also
possible that the analyte is subject to a chemical change
during the chromatographic process. Examples of this can be a
degradation or a slow equilibrium between different
molecular forms of the analyte.

7. Normal-Phase Chromatography
Q.: I have always worked with reversed-phase chromatography and avoided normal-phase chromatography. I did this
on the advice of my colleagues that normal-phase
chromatography is much more difficult than reversed-phase
chromatography. Is there any truth to this?
A.: Unfortunately, there is some truth to this assessment.
However, equipped with the right knowledge, you can
ameliorate some of the difficulties.
The most common problem with normal-phase
chromatography is retention time variability. The reason for
this variability is the strong dependence of the retention on
the concentration of small amounts of very polar constituents
in the mobile phase. This holds especially true for the water
content of the mobile phase. But it is also true for the small
amounts of polar modifiers like methanol or acetonitrile that
are added to the mobile phase to control retention and
selectivity.
Water is present in all organic solvents to some degree.
Concentrations are typically in the ppm range. Table 1 gives
the solubilities of water in some non-polar solvents (1,2) that
are used in normal-phase chromatography. As a
consequence, the water content of the mobile phase can vary
widely, if one does not take special precautions to control it.
It is not usually desirable to work with dry solvents, since it
may take a long time (days have been reported) before a
column is in equilibrium with a dry mobile phase. Mobile
phases that are saturated with water are not desirable either,
since under these circumstances the water accumulates in the
pores of the packing and one obtains a partitioning column.
One would like to use mobile phases with an intermediate,
but controlled water content. A good approach to obtain a
reproducible water content has been the use of mobile phases
that are half-saturated with water. The preparation is
relatively straightforward. One divides the mobile phase into
two equal portions, then saturates one portion with water.
This is accomplished by adding 1 or 2 mL of water to 500
mL of mobile phase and stirring for approximately half an

hour. The excess water is removed, and the two portions are
recombined. This results in a mobile phase that is "halfsaturated" with water provided that the original mobile phase
was reasonably dry.
Table 1
Solvent Solubility of Water in Solvent (Temperature)
% w/w (C)
Hexane
0.0111 (20)
Heptane
0.0091 (25)
Isooctane
0.0055 (20)
Toluene
0.0334 (25)
Dichloromethane
0.1980 (25)
Chloroform
0.0720 (23)
Carbon tetrachloride
0.0100 (24)
o-Dichlorobenzene
0.3090 (25)
Ethyl acetate
2.9400 (25)
Butyl acetate
1.8600 (20)
Diethyl ether
1.4680 (25)
Methyl-t-butyl ether
1.5000
Equilibration with a "half-saturated" mobile phase is much
faster than with dry mobile phases, but it still may take hours.
Therefore it is desirable to dedicate a column to a particular
mobile phase. Now, equilibration may only take minutes
instead of hours.
Obviously, the concentration of other polar modifiers to
the mobile phase needs to be well controlled as well. Far
example, if you use methanol as a modifier, often only very
small amounts (<0.5%) may be needed, and a small change
can result in large changes in retention. In such a case it is
desirable to use a less polar modifier, such as isopropanol or
another higher alcohol. Or you may want to stay away from
alcohols altogether and use the less polar ethers or esters.
This of course is likely to significantly influence the
selectivity of the separation.
Q.: I see that you have to pay much more attention to the
mobile phase composition in normal phase chromatography
than in reversed-phase chromatography. How about the
stationary phase? Are some stationary phases better than
others?
A.: Indeed there are differences. Silica and alumina are very
hygroscopic and are therefore very sensitive to the water
content in the mobile phase. Polar bonded phases such as
cyano-, diol- or amino-packings, are less sensitive. From the
standpoint of retention time reproducibility, they are
therefore the preferred packings when a new normal-phase
method is to be developed. However, there are some things
that one should know about the behavior of these phases.
For example, the primary amino groups on the surface of an
amino-packing can easily form Schiff-bases (imines) with
aldehydes and ketones under typical normal-phase
chromatographic conditions. Therefore, amino-columns
should not be used with samples containing aldehydes or
ketones.

10

Cyano columns are quite stable in non-polar solvents, but

often exhibit the curious phenomenon of bed collapse when
exposed to solvents of intermediate polarity like neat
acetonitrile, THF or methanol. This condition may be
encountered during attempts to "wash" a column that has
deteriorated due to build-up of sample debris. In our
experience, it is best to maintain a constant high flow
through the column during such washing procedures. The
bed collapse is more likely to occur at low flow rates or
when the column is stored in these solvents.
Amino columns may change quite drastically, if they are
exposed to aqueous eluents. The high concentration of
amino-groups in the pores of the packing forms a strongly
basic environment, which causes hydrolysis of the bonded
phase. After returning to the normal-phase mobile phase, one
should not be surprised to find a significant difference in the
behavior of the packing.
These polar bonded phases are quite universally useful,
just as silica itself or alumina. Therefore, one can not say that
one phase is "better" than others.
They will exhibit unique selectivities that are different
from silica or alumina, and they are quite different from each
other. Just as silica columns show strong retention for basic
analytes, amino columns will retain acids more strongly.
Cyano and diol columns are neutral. Diol columns are more
polar and therefore generally more retentive then cyano
columns.
As in the case of reversed-phase packings, there are
significant differences between different brands of the same
type of bonded phase. Leaving aside the differences in the
base silica, there are differences in the bonding process and
in the end-capping procedure. Monofunctional, difunctional
or trifunctional silanes can be used. Also, if multifunctional
silanes are used, significantly different coating procedures
can be used that result in "monolayer" coating or
"polymeric" coating. Cyano-columns may or may not be
endcapped, while amino- and diol-packings are usually not
end-capped. So just like with reversed-phase packings, one
cannot assume that a cyano-column from brand A will
perform the same separation as a cyano-column from brand
B.
References:
1. Techniques of Chemistry, Vol. II, Organic Solvents,
Physical Properties and Methods of Purification, Third
Edition (1970), John A. Riddick and William B. Bunger,
Wiley-Interscience
2. Merck Index, Eleventh Edition (1989), Merck & Co., Inc.

11

8. System
Volume

Volume,

Dead-Volume,

Dwell

Q.: I often encounter terms like system band spreading,

system volume, dead-volume, dwell volume. Could you
explain these terms?
A.: Gladly! When we talk about these parameters, we need
to clearly define them. As we will see, they are quite
different and affect different elements of a chromatographic
separation.
Let us start with the dead-volume. It is also called the
extra-column volume, which is somewhat clearer. It
comprises the volume of an HPLC-system between the point
of injection and the point of detection, but excluding the part
of the column that contains the packing. Therefore it includes
the injection volume, the volume of the injector, the volume
of the connection tubing before and after the column, the
volume in the endfittings of the column, including the frits,
and the detector volume. Actually, to be precise, it includes
half the injection volume and half of the detector volume.
We are concerned about the extra-column volume, because
it causes extra-column bandspreading. Bandspreading means
that the peaks become broader as they flow through the
extra-column volume. This is undesirable since it may
destroy some of the separation achieved in the column. We
would like to keep the extra-column bandspreading as small
as possible.
Q.: How can we measure the extra-column volume and the
extra-column bandspreading?
A.: The total extra-column volume and extra-column
bandspreading can only be measured using special
equipment. This is due to the fact that it includes the column
end-fittings and the frits in the column, which are not
accessible by the user. So we just have to trust that the
column manufacturers have done a good job and minimized
that part of the extra-column volume. What we can do easily
though is measure the extra-column volume associated with
the HPLC system. To that purpose we simply disconnect the
column and replace it with a "zero dead-volume" union.
Then we inject a small volume of sample and record the
detector response. An example of the response is shown in
figure 1.
From this "chromatogram" we can determine the extracolumn volume and the extra-column bandspreading. The
extra-column volume is strictly speaking the distance from
the point of injection to the center of the gravity of the peak.
To determine it we need to compute the first moment of the
peak. However, we do not have to be overly precise and can
for simplicity use the distance between the point of injection
and the peak maximum as our measure of system deadvolume.

Figure 1

w
Figure 1: Measurement of Extra-Column Volume and
Bandspreading. V is the extra-column volume, w is the
peak-width at 4.4% of the peak height, corresponding to 5
standard deviations.
We can use the standard deviation of the peak as a
measure of the system band-spreading. The standard
deviation of the peak can be computed from the second
moment of the peak. An easier way is to measure the width
of the peak. Since this peak is highly asymmetric, it is best to
measure the width as close to the baseline as possible to get a
good measure of extra-column bandspreading. For example
the width of the peak at 4.4 % of the height of the peak
corresponds to a width of 5 standard deviations.
Q.: OK, but what sample and what mobile phase do I use for
this measurement?

A.: I do not like the term "system volume", because it is

ambiguous. You may have a system dead-volume or a system
dwell volume. We should obsolete the term system volume
and rather use the words extra column volume and gradient
dwell volume. The latter can also be called gradient delay
volume. It refers to the volume of a gradient HPLC-system
between the point of mixing of the gradient and the top of the
column. (It exists also in isocratic chromatography, but there
it is unimportant).
The gradient dwell volume comprises the volume of the
gradient mixer, the connection tubing to the pump (if lowpressure mixing is used), the volume of the pump heads and
check-valves, the tubing between pump and injector, the
volume of the injector and the connection tubing between
injector and column. As one can see, there could be a lot of
volume in all these parts.
When you start a gradient, it will take some time until this
volume is purged and the gradient enters the column. During
this time, the peaks are subject to isocratic migration in the
starting mobile phase. If there are large differences in the
gradient dwell volume between different systems, this
isocratic migration can be different enough to affect the
chromatogram, especially the early portion of the
chromatogram.
Another annoying effect is that your actual separation may
be delayed significantly. If you have a system with a total
delay volume of 2 mL and you run a gradient at 200 L/min, it
will take 10 minutes until the gradient reaches the top of
your column. Therefore the start of your separation is
delayed by 10 minutes.
Q.: How do I measure the dwell volume?

A.: You can simply use the mobile phase used in your
analysis. As sample you may just utilize the standards that
you use in chromatographic analysis. However, most of the
time the standard is likely to be too concentrated, and you
need to dilute it to make it compatible with the small extracolumn bandspreading. If you do this, you get a direct idea
of extra-column bandspreading under your actual
chromatographic conditions. On the other hand, you may
standardize the measurement in the form of a general system
check that is independent of the chromatographic test that
you are running. The disadvantage is that you may miss
something that is specifically related to your particular
analysis. I encountered a case a few years ago, where the
sample was interacting strongly with a part in the injector.
Although everything that we did pointed to excessive extracolumn bandspreading, we could not see it in our
standardized test, which used a different sample and a
different mobile phase. Only when we used the actual sample
under actual mobile phase conditions did we find the
problem.
Q.: This covers extra-column volume and extra-column
bandspreading. What about system volume and dwell
volume?

A.: Once again, you disconnect the column from your

system. Then you run a step-gradient from methanol to
methanol with 10 mg/L propyl paraben, using a UV detector.
This will create an S-shaped detector trace. You then
measure the time delay from the point at which you started
the gradient to the point when half the height of the step is
reached. Multiplying this time with the flow-rate gives you
the gradient delay or dwell volume.
Q.: Is there a literature reference that contains these
definitions?
A.: Many textbooks have sections that define the terms used
in chromatography. There is however one authoritative body,
the Commission on Analytical Nomenclature of the
Analytical Chemistry Division of the International Union of
Pure and Applied Chemistry that defines many of the terms
used in HPLC and other chromatographic techniques. I
believe that their latest publication containing the currently
valid nomenclature has been published in Pure and Appl.
Chem, Vol. 65, No. 4, pp. 819-872, 1993. Unfortunately, not
all terms used are explained, for example the dwell volume is
missing, and the definition of some terms still remains
ambiguous.

12

www.pharmaguideline.com

9. Transfer of Gradient Methods

Q.: I have developed a gradient method that is very
reproducible when I run it on my system, but I cant get the
same results on another system. Whats wrong?
A.: Occasionally, some difficulties arise when gradient
methods are transferred from one HPLC system to another.
Unless the systems are identical, one can usually expect
some shifts in retention times. Most of the times these shifts
in retention do not affect the resolution to the extent that a
method becomes useless, and one can usually proceed by
ignoring the differences. On the other hand, with a proper
understanding of the underlying causes, one may be able to
adjust the gradient to get equal performance from dissimilar
HPLC systems.
In the last column we touched on one aspect of the
problem: the gradient dwell volume. It is the volume between
the point of mixing of the gradient and the top of the column.
After you start your analysis by injecting the sample, the
gradient will not reach the top of the column until the
gradient dwell volume is purged. This means that your
sample is subjected initially to a period of isocratic migration
until the gradient catches up. Since the gradient dwell
volume may be different from system to system, this isocratic
migration time will be different and may result in retention
time differences or even affect resolution.
Another possibility is the gradient itself. There could be
compositional differences from system to system. With most
HPLC systems manufactured today, this should be only of
secondary concern. But generally, any gradient system
delivers a composition with the highest accuracy in the midrange of the composition, i.e. at a mixture of 50% A and
50% B. The accuracy suffers, when very disproportionate
amounts of A and B are mixed, e.g. 5% A or 95% A.
Q.: What can I do to sort out, which of these possibilities
causes my method transfer problem?
A.: The simplest thing to do is to compare your gradient on
both systems. In order to do this, you disconnect the column
and add an UV-absorber to the B-solvent of your gradient. If
you are using a reversed-phase system, you can for example
add 10 mg/L propyl paraben to your B-solvent. Then you run
the gradient on both systems and record the baseline. You
then compare the two plots to each other. You want to find
the point where the gradient starts, and you also want to
measure the gradient profile. If your gradient is linear, then
you only need to check the slope of the gradient.
Most likely you will find that the onset of the gradient is
different between the two instruments, while the profile is
very similar, just off-set by some amount of time. In this
case, you have a difference in the gradient dwell volume.
Q.: If this is the case, is there a simple way to compensate
for the difference in dwell volume?

13

A.: There is a solution that works most of the time. If the

gradient dwell volume is smaller on the system that you are
transferring your method to, you may be able to compensate
for the lack in dwell volume by programming an isocratic
portion at the beginning of your gradient that compensates
for the volume difference. The remainder of the gradient
simply remains constant. If on the other hand the gradient
dwell volume on the second system is larger than the one on
the first system, the situation is more difficult. In principle,
you can start the gradient, and then inject the sample after a
delay time that accounts for the difference in the gradient
dwell volume between both systems. But this may not be
possible on an automatic system, where the injection triggers
the start of the gradient. In this case, you may need to go
back to ground zero and redevelop the method.
Q.: This would not be a very pleasant situation after all the
time that I spent developing the method. What can I do to
prevent this from happening in the future?
A.: You can avoid this situation by developing the method
for the HPLC systems that will ultimately use the method.
This requires some foresight and some planning, but usually
is not impossible. What you need to do up front is to
characterize the systems that are likely to be used for your
method. There are two basic things that you need to know for
each system: the gradient dwell volume and the
compositional accuracy. You can get both pieces of
information in a single experiment: as described above, you
add an UV-absorber to your B-solvent. Then you program a
multiple step gradient in increments of 5% from 0% B to
100% B. The flow-rate should be the flow-rate typically
used, so most likely you will use a flow-rate of 1 mL/min.
The intervals between the steps should be a few minutes,
maybe 5 minutes. Now run this gradient without a column in
place on the different systems and record the detector
response. The time delay between the time programmed for
each step and the actual occurrence of the step gives you the
gradient delay time. The height of the step gives you a
measure of the composition. The steps will be smeared out a
little, which is a function of the mixing volume in your
system.
Now that you have characterized the systems, you can
design you method around the characteristics of the systems.
As I mentioned before, the largest issue is usually the
gradient dwell volume. If you know that the systems that you
need to transfer your method to have a larger dwell volume
than the system on which you are developing your method,
you should automatically add an isocratic step in your
methods development that will compensate for this
difference. If the dwell volumes of the target systems are
smaller than the one of your development system, you should
be able to simply compensate for this by adding a gradient
delay time to the beginning of your method when transferring
the method.

If there are compositional differences in the middle of the

gradient run, you could conceivably compensate for those as
well by adjusting the gradient profile, but I have never
encountered a situation where this was necessary.
This discussion assumes that the column is in complete
equilibrium with the starting mobile phase. Occasionally I
have encountered a situation, where in routine analysis the
gradients follow each other so quickly that the column never
returns to equilibrium in the starting mobile phase. You will
see that you have this situation if your first gradient always
gives results that are different from subsequent gradients.
This may be advantageous for speeding up a method, but it
could be a cause of difficulties when this method is
transferred to another system with a different dwell volume.

10. Clogged System

Q.: My system pressure is much higher than it should be.
What is the problem?
A.: First, let us check your premise. How do you know, what
the pressure should be?
Q.: Previously, the same column with the same mobile phase
conditions at the same temperature gave me a backpressure
of about 2000 PSI. Now it has doubled. I have set the highpressure limit of the pump at 4000 PSI, and the pump shuts
off at this pressure.
A.: It is good that you have a previous reference point, we
can start our troubleshooting from there. There are many
different things that can lead to an increase in back-pressure.
First, the viscosity of your mobile phase could not be correct.
Second, a part of your system may be clogged. The latter is
the more common case, but before we disassemble your
system, let us check your premises.
How do you know that you are pumping the correct mobile
phase? If you use automatic blending, did you put the correct
solvent on the correct inlet line of the pump? For instance a
mix-up of the methanol and the acetonitrile line (if your
instrument is set up this way could easily give a factor of two
difference in backpressure due to the viscosity difference
between the mixtures of each solvent with water.
If you work at elevated temperature using a column heater,
you should check that your column heater is on and that it is
working. The viscosity of a solvent typically changes by 25%
for 10 C. Therefore if your solvent is supposed to be at 60
C, but you heater does not work, you could get double the
backpressure.
Q.: O.K. This is easy to check. What else can I check?
A.: I often ask the question, whether the high-pressure shutdown happened in the middle of a long series of analyses or
upon start-up.

If the high pressure developed upon start-up, it is likely

that some kind of an error occurred. Let us check a few
things, starting with the simple things first: Is it the correct
column with the correct particle size? Column backpressure
changes with the square of the particle size. So if by accident
you grabbed the 5 m version of the 10 m column that you
used to run, that would explain the increased pressure.
If this is not the case, then something is clogged. First, let
us disconnect the column and replace it with a union, leaving
guard columns and precolumn filters in place. One
possibility that we will check is an incompatibility of the
mobile phase. If we would leave the analytical column in
place, we could do damage to it in the subsequent operations.
After disconnecting the column, we check the backpressure
of the system again. If it is substantial, i.e. exceeding 1500
PSI, then one of the remaining parts is clogged and the
column is most likely still O.K. We can then disconnect the
remaining elements of the fluid path one by one and
determine, which of the elements contributes the most to the
back-pressure. Usually, it is most efficient to replace the
clogged part, rather than trying to clean it. But in many
cases, a replacement part is not readily available, and we
may consider cleaning procedures.
It is easier to develop a cleaning procedure, if we know
what we are trying to remove. In this hypothetical case, we
know that the clog occurred upon start-up. We then should
ask, why the clog occurred. It is possible, for example, that
the mobile phase that was last used in the system was not
completely flushed out and was incompatible with the new
mobile phase. A buffer may have precipitated. If this is the
case, you may simply wash or flush the clogged part with a
solvent that will redissolve the precipitate, and you are up
and running in a short time. The clog actually may even
disappear, while you are still checking which part was
clogged.
If it turns out that the analytical column or the guard
column clogged during start-up, you may check in what
solvent the column or guard column was stored. The storage
solvent may be incompatible with the mobile phase and
cause precipitation of a buffer. Or, the column or guard
column was stored with a mobile phase that contained a salt
and dried (partially) out due to loose caps. In this case, you
may be able to resurrect the column or guard column by
flushing it at low flow-rate with a mobile phase that will
redissolve the salt.
Q.: In my case, the high-pressure shutdown happened in the
middle of a series of runs. What could be the problem in this
case?
A.: In this case it is most likely that a part of your system
clogged due to debris that built up. The debris may come
from the seals in your instrument or from the sample. It could
be constituents of your sample that are insoluble in the
mobile phase. It could be a component of the sample that is
strongly adsorbed on the guard column, or if you do not use
a guard column, on the column itself. Often, these are high-

14

molecular weight constituents of your sample that have not

or have only partly been removed. These could be proteins in a
serum sample, excipients in a pharmaceutical formulation or
high-molecular weight constituents in a food sample.
If you use precolumn filters and guard columns, let us
disconnect them one by one and measure the backpressure of
the system without this part. This way, we should be able to
quickly identify the clogged part. If it is a guard column or a
precolumn filter, it is best to just replace the part. It has done
the job that it was supposed to do and protected your
analytical column. To attempt to resurrect that part by
cleaning it is false economy: it will not work as well the next
time, and there is consequently an increased likelihood that
the analytical column might get damaged.
If it is the analytical column itself, it is worthwhile to do
some work to attempt to clean it. However, the best remedy
is prevention, and you may consider using a precolumn filter
or better, a guard column in the future to protect the
analytical column. If the origin of the problem is the sample
itself, you may consider filtering the sample or using solidphase extraction to remove the contaminant.
If you have a spare filter for your column, I would first
attempt to replace the inlet filter of the column. If you do not
have a filter, you may remove the filter and attempt to clean
it in an ultrasonic bath. This is only feasible with silica-based
columns. With polymeric columns, the packing is under
stress and will ooze out of the column from the moment you
remove the filter. If you have a spare filter handy, you can
quickly replace it and close the column again. If you do not
have a spare filter, you should not open a polymeric column.
If a replacement or cleaning of the inlet filter does not
result in a reduction in backpressure, then most likely
something is adsorbed on the surface of the packing or has
precipitated in the column. In this case, the column should be
flushed at slow flow-rates with something that redissolves or
desorbs the suspected contaminant. Column manufacturers
typically recommend a sequence of solvents of increasing
solvent strength to accomplish this task. For reversed-phase
columns, one might consider the following sequence: water,
methanol, THF, methylene chloride, methanol and back to
water. For normal-phase columns, the sequence of polarities
is inverted: methylene chloride, THF, water, methanol,
methylene chloride.
If this does not help, you can still attempt to back-flush the
column. But at this point in time it just might be best to pick
up the phone and order a new column.

11. Column Plate-Count

Q.: When I run my method and measure the plate count, I am
getting about 8000 plates for my column. But the
manufacturers literature specifies 13000 plates. Can you
explain this discrepancy?
A.: Columns do not have fixed plate counts. For a given
column, plate counts depend on the flow rate, the viscosity of

15

the solvent and the molecular weight of the analyte. The

instrument may affect the measurement as well, but we will
ignore this for the time being.
The column manufacturer has set up standardized
conditions by which the quality of the column is measured.
In reversed-phase chromatography, the measurement is
usually done with a simple hydrophobic analyte, like toluene,
naphthalene or acenaphthene. The mobile phases used with
these compounds contain a high amount of organic solvent.
Therefore they have a low viscosity. On the other hand, most
HPLC users are dealing with more polar molecules, which
require mobile phases with a higher water content. These
mobile phases have a higher viscosity. Since under normal
HPLC conditions and normal HPLC flow rates, the plate
count decreases when the viscosity increases,the plate count
is lower under practical use conditions than under column
testing conditions.
Q.: If the plate counts under normal use conditions are lower
than the plate counts reported by the manufacturer, arent the
manufacturers misleading the users as to the capabilities of
the column?
A.: Not really. This is not the purpose of it. The column
testing procedures used by manufacturers are designed to
check the quality of the column, and this is best done at or
close to the maximum of the plate-count capability of the
column. This is the point where the column packing
technique has the largest influence. Therefore, if I want to
check the quality of the packing technique, this is the point at
which I want to measure it.
Q.: Please explain a little bit more, why and how column
plate-count varies with flow-rate, column packing etc.!
A.: Underlying all this dependence of plate count on a
variety of parameters is the reduced plate height, and its
dependence on the reduced velocity. I will sketch this out a
little bit, but for a deeper understanding we need to look at a
textbook of chromatography.
The plate count of the column is the column length L
divided by the reduced plate height h and the particle size
dp:
L
N=
(1)
hd p
The column length and the particle size are fixed (and
known). The reduced plate height depends on the reduced
velocity, which is defined as follows:
ud p
=
(2)
D m
This equation contains the linear velocity u, the diffusion
coefficient DM of the sample in the mobile phase and once
again the particle size. Most HPLC work is carried out at
reduced velocities between 3 and 20.

www.pharmaguideline.com

There are several equations in the literature that describe

the dependence of the reduced plate height on the reduced
velocity. Within the velocity range discussed here they give
identical results. Therefore, let me use the simplest one,
which is based on the van-Deemter equation:
1
h=1.5+ +
(3)
6
The coefficients are empirical and have been derived from a
broad data base for reversed-phase operating conditions.
The curve described by this relationship has a minimum,
which is reached at a reduced velocity of around 2.5. The
first coefficient is a measure of the uniformity of the packing.
It therefore depends on the packing process.
The column plate count depends on the reduced velocity as
follows:
L
N=
(4)

1
+
d
p
1.5+
6

Since the dependence of the HETP on the velocity has a

minimum, the dependence of the plate count on velocity has
a maximum. A little calculus shows that the maximum plate
count is approximately L/2.3 dp. So we can conclude that a
15 cm column packed with 5 m particles that has a plate
count maximum of 13000 is a good column.
Bad column packing largely affects the first coefficient of
the equation 3, while the other ones remain fairly much the
same. The effect of this coefficient is largest, when the
influence of the other ones is minimal. This is the reason why
one should test columns at or close to the maximum plate
count.
Let us now tackle the second part of your question, the
dependence of the plate-count on flow-rate. I glossed over
this a little bit by moving to the reduced velocity. First, the
reduced velocity is proportional to the linear velocity, which
is proportional to the flow rate. Therefore, the plate count
depends on flow rate. It is low at very low flow rate and low
at high flow rate. The maximum plate-count is somewhere in
between. But where? As we have seen from the definition of
the reduced velocity in equation 2, the reduced velocity
includes the diffusion coefficient of the analyte in the mobile
phase. We know from the results above, that the maximum
plate-count occurs at a reduced velocity somewhere around 2
to 3. For the hydrophobic analytes typically used as column
test samples with a mobile phase of 70/30 v/v
acetonitrile/water, this translates to a linear velocity of 1 to
1.5 mm/sec for a 5 m column. The flow-rate that
corresponds to this linear velocity is about 0.5 to 0.7 mL/min
for a column i.d. of 3.9 mm and 0.7 to 1 mL/min for a 4.6
mm column.
Since the location of the minimum of the curve depends on
the diffcusion coefficient, we need to know the dependence
of the diffusion coefficient as a function of the mobile phase.
There are equations in the literature, which allow us to
estimate the diffusion coefficient of an analyte in a particular
solvent (e.g. Wilke-Chang equation). However, for the
current discussion we only need to know that the diffusion

coefficient is inversely proportional to the viscosity of the

mobile phase:
1
DM
(5)

This means that the velocity (or flow rate) at which the
maximum plate count occurs decreases with increasing
viscosity. A mobile phase of 70/30 v/v acetonitrile/water has
a viscosity of 0.7 cP, while the viscosity of a mobile phase of
50/50 v/v methanol/water is 1.8 cP. When we use the
methanol/water moblie phase, the optimal flow rate
decreases therefore to between 0.2-0.3 mL/min for the 3.9
mm column and to between 0.3 to 0.45 mL/min for the 4.6
mm column packed with 5 m particles. This is something
worth keeping in mind during methods development.
There is an interesting side aspect of the dependence of the
diffusion coefficient on the solvent viscosity. One can use the
following rule of thumb for estimating column performance,
when the mobile phase composition is changed: the same
column will give about the same plate count at the same
back-pressure in different mobile phases. A useful rule to
keep in mind!

12. Column Backpressure

Q.: I am using a 4.6 mm x 150 mm 5 m C18 column. My
mobile phase is 50/50 methanol/phosphate buffer pH 7. I am
getting 3000 psi at 1.5 mL/min. Is this normal?
A.: It is a little high, but not outrageous. I would have
expected a backpressure for the column alone of about 2400
psi, plus you have to add maybe another 200 psi for the
backpressure of the connecting tubing. This would get us to
2600 psi. I think, your column may be partially clogged.
Q.: This is quite possible. I suspected this too, and this was
the reason for my question. How did you arrive at the
estimate of 2400 psi as the "expected" backpressure?
A.: I calculated it from the Kozeny-Carman equation. It
works very well for all HPLC columns that are packed with
incompressible packings like silica-based packings.
The Kozeny-Carman equation relates the pressure to the
flow-rate F, viscosity , column length L, column radius r
and particle diameter dp:
FL
p= f
(1)
2
2
r d p
The pressure increases with increasing flow-rate, viscosity
and column length. It decreases with the square of the
column radius and the square of the particle diameter.
The proportionality factor f depends on the interstitial
fraction (the "space" between particles) and therefore the
packing density. Incompressible packings pack to fairly

16

much the same packing density, i.e. to an interstitial fraction

of about 40 %. For such a column the factor f is about 1000.

Q.: This helps a lot. Now, let us go back to the column. It

seems to be partially clogged. What shall I do about it?

Q.: O.K. Let me calculate the pressure for my example

column. I have the flow-rate, column length and diameter,
and particle size. How about the viscosity?

A.: The simplest thing to do is to replace the inlet frit.

Hopefully, this will reduce the pressure. You actually may
also be able to clean the old frit in an ultrasonic bath. If the
replacement of the frit does not result in a reduction in
backpressure, the packing itself might be partially clogged.
You can go through some washing cycles to remove,
whatever is clogging the packing, but this is a fair amount of
work. Since the pressure increase is only small, I would
continue to use the column until it reaches the pressure limit
of your system. I also would try to figure out, what the
source of the pressure increase is and try to prevent it as
much as possible. Often, the use of a guard column between
injector and analytical column solves the problem or at least
slows down any pressure increase.

A.: For neat solvents, you can easily find the viscosity in
handbooks. For solvent mixtures of non-associating solvents,
you can estimate the viscosity by assuming a linear
relationship with the volume fraction. For aqueous mixtures,
this does not work. Most mixtures of organic solvents with
water have a viscosity maximum. The viscosity of the most
commonly used aqueous mixtures is shown in figure 1.
The viscosity of a 50/50 mixture of methanol and water is
about 1.8 cP (=0.018 P) at room temperature. There is not a
lot of difference between water and a dilute buffer, and I can
use the viscosity of the water/methanol mixture from the
chart.
Viscosities of Aqueous Mixtures
3

MeOH
MeCN
EtOH
Acetone
THF

0
0

10 20 30 40 50 60 70 80 90100
Vol % Water

Figure 1:
Viscosity of mixtures of organic solvents with water as a
function of the water content (% v/v). Appreviations: MeOH =
methanol, MeCN = acetonitrile, EtOH = ethanol, THF =
tetrahydrofuran
Therefore the calculation of the pressure is carried
out as follows:
1.5
0.01815
6
p[atm]=1000 60 10
3.14150.23 2 0.0005 2
The last factor (10-6) is the conversion factor to
atmospheres. Do not forget to divide the flow rate by 60 to
obtain the flow in mL/sec. Then we have the same
dimensions throughout.
To convert from atmospheres to psi, multiply by 14.7. You
should get 2388 psi, i.e about 2400 psi.

17

Q.: O.K. I will try to replace the filter. In the case that his
does not help and I decide to clean the column, how should I
go about this?
A.: You first need to remove the buffer from the column, so
you want to wash it with 5 to 10 column volumes of water.
That is between 10 and 20 mL for this column. Then you
switch to 100% methanol and let the column purge for a
while, maybe some 15 minutes. You can do all this at high
flow rate, since your column has not yet clogged completely.
So I would do this at 1.5 mL/min. Let us check the pressure
in methanol. Methanol has a viscosity of 0.6 cP. Therefore
we would expect a pressure of about 800 psi at 1.5 mL/min
in methanol. If the pressure is still high, flush the column
with tetrahydrofuran (viscosity 0.5 cP) or methylene chloride
(viscosity 0.4 cP) or both. The volume for each solvent
should be around 20 mL. Whatever is not removed within
this volume, is not likely to be easily removed in this solvent
at all. After this process, you can go back to the original
mobile phase. You must make sure that the mobile phases in
subsequent steps are miscible. Therefore you need to go from
methylene chloride to methanol to water (or 50/50
methanol/water) to 50/50 methanol/buffer. Now you need to
reequilibrate the column. In your case, with this simple
mobile phase, the reequilibration is fast. If you were to run a
separation requiring an ion-pair reagent, you would now
need to purge your column with the mobile phase for quite a
while, before you are back in equilibrium.

13. Peak Area Fluctuations

Q.: What are the reasons for fluctuations in peak area, when
the same sample is injected multiple times?
A.: Unfortunately, there are many different reasons, too
many to count. Nearly every part of the HPLC instrument

can conceivably contribute to changes in peak area. Let us

review them one by one.
The injector comes to mind first. What problems you
might encounter depends to some degree on the type of
injector. But in all cases, the formation of air bubbles during
the measurement of the injection volume is a major problem
source. If you are injecting manually, make sure that there
are no air bubbles in your syringe. With fixed-loop injectors,
make sure that no air is siphoned into the injector loop. In
autoinjectors, air bubbles can be formed if the cap of the vial
seals well around the needle, which can give rise to a vacuum
when a large portion of the sample is injected. Any air
bubble formation results in a random variation of the
injection volume from injection to injection.
If the concentration of the analyte varies widely from
sample to sample and the peak area obtained from a single
sample becomes constant only after two or three injections,
sample carry-over is a likely source of the problem. Clean
the syringe carefully after each injection, if you are injecting
manually, or check the needle wash in an autoinjector. Make
sure that the solvent used for the needle wash is a good
solvent for the analyte(s).
Decreases in peak area from run to run can also be caused
by temperature changes, but the effect is small, under 2%. If
you take a sample out of the refrigerator and put it into an
autoinjector, it will slowly warm up to the injector
temperature and the solvent will expand. Therefore you will
inject initially a larger mass of sample than later when the
sample has reached the temperature of the sample
compartment.
Random variations of the flow rate can also cause
fluctuations in peak area. They are typically caused by
malfunctioning check valves or air or cavitation in the pump
head. They are usually accompanied by fluctuations in
retention times as well. You can check if this is the problem
by monitoring the backpressure. What fluctuations can be
tolerated depends on the width of the peak. Assume that
there is a flow difference of 10% between the two heads of
your pump. If your peak width is about 20 pump strokes, the
fluctuation in peak area caused by the pulsation of the pump
is under 1%. But it will be 10% if the peak width is about
one pump stroke.
Another potential cause of flow-rate changes are leaks on
the high-pressure side of the system. They should be easy to
find.
Integration errors are more difficult to troubleshoot. Gross
changes in the way the software integrates the peaks can be
found by simple visual examination, but more subtle changes
are difficult to spot. Significant integration errors often occur
with tailing peaks at a signal-to-noise ratio of 100 or less.
Generally, the base-line noise limits the precision of the
integration, but the problem becomes worse with fronting or
tailing peaks. You can try to reprocess the data with different
settings of the integration parameters and see if this reduces
the problem.
In complex samples which contain peaks that are only
marginally resolved from the peaks of interest, the software

can have difficulties determining where the baseline is. In

such a case, a manual integration or an automatic integration
with a forced baseline can improve the reproducibility of the
integration.
The sample itself can be the source of the problem. In
reversed-phase chromatography it has been observed that
some proteins do not elute completely during the initial
gradient. An additional quantity is eluted in subsequent blank
gradients. This ghosting is specific to the protein and the
elution protocol. If this occurs, blank gradients need to be
run between analyses.
Also, proteins can be irreversibly adsorbed on some
columns. The peak area obtained with a brandnew column
may increase slowly with repetitive injections. Presumably,
some protein binds to active sites on the packing. Once these
sites are saturated, a reproducible peak area is obtained. It is
not necessary to use the protein that you need to analyze. It
has been reported that other proteins can be use as well to
saturate the active sites.
The detector contributes to peak area fluctuations in an
indirect way. How well the peak can be integrated depends
on its signal-to-noise ratio. You can not expect a
reproducibility of the peak area that is any better than the
ratio of the baseline noise to the signal at the peak maximum.
The mobile phase is rarely a contributor to peak area
fluctuations. If there are any shifts in mobile phase
composition, they will affect retention times long before they
affect peak areas. Ghost peaks or negative peaks that may be
generated by the mobile phase affect the integration in the
same way as any other interference. They can often be
avoided by dissolving the sample in the mobile phase.
Q.: What reproducibility of the peak area can I reasonably
expect?
A.: Unless you are limited by the signal-to-noise ratio, the
relative standard deviation of the peak area for repetitive
injections should be definitely under 1%. Values of 0.2 to
0.5% are achievable. The reproducibility can be further
increased by using an internal standard and calculating the
ratio of the peak area of the analyte(s) to the peak area of the
internal standard, but even with this method you are not
likely to be able to drive the r.s.d. much under 0.1%.

14. Ghost Peaks

Q.: A peak appears when I am making a blank injection.
Where does it come from?
A.: Ghost peaks can occur in several different situations. In
the following, we will discuss the various causes of these
ghost peaks. First, different phenomena can occur in isocratic
chromatography and in gradient chromatography. Therefore
we will divide the subsequent discussion into these subjects.

18

www.pharmaguideline.com

1. Isocratic Chromatography

2. Gradient Chromatography

Let us define what you mean by a blank injection. If you

are injecting mobile phase, then indeed we do not expect to
see a peak. But if you are injecting something else, for
example the solvent in which you commonly dissolve the
sample, you should not be surprised to see peaks. The
injection of anything but the mobile phase disturbs the
equilibrium between the mobile phase and the stationary
phase, and peaks can be observed.

All of the previous comments applied to isocratic

chromatography. If you are running gradients, then there are
several other possible sources of peaks in blank runs.

Recirculated Mobile Phase

If a constituent of the mobile phase has a moderate
retention, it will show up as a retained peak. These peaks can
be positive or negative. If you recirculate the mobile phase,
and if this mobile phase has been used for awhile, you can
see a negative peak profile that corresponds to the sample
that you usually inject. The previously injected samples have
accumulated in the recirculated mobile phase, and the
stationary phase is partially coated with the analytes. When
you inject a freshly prepared mobile phase or just some blank
solvent, there is a concentration deficit of the sample
constituents. This concentration deficit moves down the
column at the same speed as the corresponding peaks, and a
negative image of the normal chromatogram results. The
phenomenon is called vacancy chromatography and the
peaks are called vacancy peaks. This is one of the reasons
why a recirculation of the mobile phase is generally not a
good idea.
Carryover and Precipitation
If you inject a sample that is identical to the mobile phase
and still observe a peak, we need to look at sources outside
the column and the mobile phase. For example, there could
be some carryover from previous injections. You should look
if the needle wash in your autoinjector works well. If you use a
manual injector, make sure that your syringe is properly
cleaned. This includes the outside of the needle.
Is it possible that previously injected sample constituents
have precipitated in the injector or are adsorbed to a part of
the injector? The area of the peaks that you are observing
should then decrease with every injection. Make sure that all
sample constituents are soluble in the mobile phase. The best
solution is to dissolve the sample in mobile phase.
Fluid Path Issues
An injection is always accompanied by a pressure pulse. If
there are dead corners in the fluid path, such as a Tconnection to a pressure gauge etc., the pressure pulse can
cause some of the solvent in these dead corners to enter the
fluid path and create a peak. Usually, these peaks are broader
than peaks normally observed in the chromatogram. Inspect
the fluid path and eliminate these dead corners wherever
possible.

19

Solvent Impurities
All minor constituents and impurities in the mobile phase
will enrich on the column during equilibration of the column
with the starting composition. As you increase the strength of
the mobile phase during the gradient, these minor
constituents and impurities will be desorbed from the column
just like a sample. This can result in a complex
chromatograms with many peaks, if your mobile phase
contains many impurities. A typical source of problems is the
quality of the water used in reversed-phase chromatography.
Water can contain impurities from many sources. These
sources include the water purification system itself, the
containers in which the water is stored, and bacterial growth.
The performance of a water purification system should be
monitored on a regular basis using a blank reversed-phase
gradient. Otherwise, the purchase of HPLC-grade water is
recommended.
Even with high-quality solvents and reagents, it is possible
to observe peaks due to the reagents themselves. This
depends on the detector that you are using. If you are using a
UV-detector, it depends on the wavelength and how sensitive
the detector is to changes in the refractive index of the
mobile phase. With high-quality reagents, the gradient
background can appear as a hump rather than a collection of
peaks.
Protein-Containing Samples
When protein-containing samples are analyzed by
reversed-phase chromatography, it is also possible to observe
peaks with blank gradients that stem from a previous
injection of the sample. The peak area decreases rapidly with
subsequent blank gradients. The phenomenon is reminiscent
of sample carryover in the injector, except that no sample is
injected. The injector can even be taken off-line. What is
happening is that the elution of some proteins is incomplete
in the first gradient. For example, under a specific set of
conditions only 2/3 of the injected amount of ovalbumin is
recovered in the first gradient. In every subsequent gradient
approximately the same amount of the remaining ovalbumin
is recovered. After a few runs the amount of ovalbumin
becomes negligibly small. This phenomenon seems to be
specific for proteins. To my knowledge, it has never been
observed with small molecules.
Summary
Being aware of all these causes of ghost peaks will allow
you to take the proper precautions to avoid these situations.

15. Dependence of Retention Times on pH

Q.: I have a separation that seems to be very sensitive to
variations in pH. It is a simple reversed-phase method, and I
am not using any ion-pairing reagents. I experienced a lot of
changes in retention time from day to day, which I was able
to trace to small variations in pH. What is the problem?
A.: From your description of the problem it appears that you
are dealing with ionogenic analytes with a pKa close to the
pH of the mobile phase. Therefore, you are dealing with two
forms of your analytes which have significantly different
retention times. This could be for example an acid in its
protonated and unprotonated form. The charged form of the
analyte has a much lower retention factor than the uncharged
form. Both are in an equilibrium with each other that
depends on the pH. Small variations in the pH of the mobile
phase change the ratio of both forms of the analyte and
therefore the retention time.
As a consequence of this, it is necessary to control the pH
tightly. Since you have made some experiments already, you
may have obtained enough information to know
quantitatively the dependence of retention time on pH. If not,
make a few controlled experiments to obtain this knowledge.
Also, you may check the history of the method. Such
information should have been generated during method
robustness testing. Once you the dependence, you need to
decide, what are the retention time fluctuations that you can
live with. Then you calculate, what the accuracy of the pH
adjustment should be. Even in difficult cases, a precision of
+/- 0.02 pH units should be sufficient.
Q.: It is difficult to adjust the pH that accurately. If I add
only a little bit of acid, the pH changes drastically.
A.: Unfortunately, this indicates that you are not using a
buffer. The definition of a buffer is that it buffers the pH
from small additions of acid or base, i.e. the pH should
change very little upon the addition of acid or base. This
ability to maintain the pH is called the buffer capacity. It
depends on the concentration of the buffer and on the pKa of
the buffering ions. Maximum buffer capacity is obtained at
the pKa of the buffering ion. Table 1 gives you the pKas of
some buffers that are commonly used in HPLC. Note that
they are measured in water, without the addition of an
organic solvent. The pKas will shift upon addition of an
organic solvent. But generally, all pKas of all compounds
including your analyte will experience the same shift.
For the buffer concentrations commonly used in HPLC,
the pH of your buffer should remain within 1.5 pH units
around the pKa of the buffering ion. For very dilute buffers,
under 5 mM, you may want to narrow the range even further
to +/- 1 pH units around the pKa.
Unfortunately, I frequently encounter methods that
completely ignore the buffer capacity. An example is a
solution of potassium dihydrogen phosphate at pH 4.5. This
is not a buffer, this is a salt solution. As you can see from the

table, the pKas of a phosphate are about 2 and 7. 4.5 is right

in the middle, and phosphate has no buffering capacity
whatsoever at this pH. To buffer the pH at 4.5, you need to
use a different buffer, for example acetic acid.
Similarly, a solution of ammonium acetate at pH 7 is not a
buffer, but a salt solution. The use of ammonium acetate
"buffers" has become popular, since ammonium acetate is
volatile and can be used with mass spectrometry detectors.
But at pH 7 it is not a buffer, and you might as well leave it
out of the mobile phase, unless you need a salt for a different
reason.
Table: pKa Values of Commonly Used Buffer Ions
(at 25 C)
Buffer
pKa
Acetate
4.75
Ammonium
9.24
Borate 1
9.24
Borate 2
12.74
Borate 3
13.80
Citrate 1
3.13
Citrate 2
4.76
Citrate 3
6.40
2-(N-morpholino)ethanesulfonic acid (MES)
6.15
Oxalate 1
1.27
Oxalate 2
4.27
Phosphate 1
2.15
Phosphate 2
7.20
Phosphate 3
12.38
Trichloroacetic acid
0.52
Triethanolamine
7.76
Triethylamine
10.72
Trifluoroacetic acid
0.50
Tris(hydroxymethyl)aminomethane (Tris)
8.08
Q.: What can I do to improve the method?
A.: If you are using the wrong buffer for the intended pH,
you should simply select a more suitable one from the table.
Often, the substitution does not affect the separation,
otherwise you may need to make additional adjustment.
Sometimes, a substitution of the buffer is not possible due
to constraints imposed by the detector. We mentioned
already the need to use volatile buffers with mass
spectrometers. Buffers based on carboxylic acids are less
suitable for low-UV detection due to a high background
absorption, which results in excessive noise. If you can not
find a detection compatible buffer other than the one that you
are using, you are in trouble. You can either live with the
retention variability as it is, or you should redevelop the
method at or near the pKa of the buffer that you are using.
This will clearly change the chromatogram drastically, and
you need to reoptimize the method from scratch. This may
not be good news, but you will be better off in the long run.

20

I generally recommend for methods development to select

the buffer and therefore the general pH range first, then use
solvent selectivity to fine-tune the separation. This methods
development strategy is fast and highly efficient. It also
prevents you from ending up in the wrong place with respect
to buffer pH. You might want to consider this approach in
your next methods development.

16. Column Equilibration

Q.: How long do I need to equilibrate my column?
A.: Column equilibration is rather simple, once the basic
principles are understood. How long you need to equilibrate
depends primarily on the state of the column before
equilibration, the concentration of the ingredients of the
mobile phase and the retention factor of these ingredients.
You have to realize that the kinetics of equilibration are
rather fast, and the local equilibrium in the column is reached
practically instantaneously. Therefore, equilibration is
primarily limited by the speed with which mobile phase
constituents are transported into the column or can be
removed from the column. Let us first discuss the cases
where the column needs to be equilibrated with new
constituents of the mobile phase.
In equilibrium, the concentration of mobile phase
constituents that are adsorbed on the surface may be between
1 and 20 mol/m2. A typical packing has a surface area of
300 m2/g, and a typical HPLC column contains around 0.5 g
of packing per mL of column volume. Consequently, the
surface area per column volume is around 150 m2/mL and
the concentration of adsorbed ingredients is 150 to 3000
mol/mL.
Solvents usually have a high surface concentration. Let us
take methanol with a molecular weight of 32 as an example.
Using a stationary phase concentration of 3 mmol/mL, we
need approximately 100 mg of methanol per mL of column
volume to completely saturate the surface. If we start with a
surface that contains no methanol whatsoever and if the
concentration of methanol in the mobile phase is 10% or
higher, one mL of mobile phase is sufficient to deliver the
necessary amount of methanol to the column. For complete
equilibration, add a factor of 2 to 3 or so, and your column is
equilibrated. This means that for mobile phase ingredients
that are present at high concentration, equilibration is fast.
This is one of the reasons why gradient chromatography
works as well as it does. On the other hand, if the
concentration of mobile phase additives is low, long
equilibration times result. For example, if the column has
never seen methanol and the mobile phase concentration is
only 1% methanol, then we need at least 10 column volumes
to deliver the correct amount of methanol to the column. In
such a case, other mobile phase equilibria may actually
complicate things. Assume that the column was originally
equilibrated with acetonitrile, and we want to equilibrate it

21

now with a mobile phase containing 1% methanol. Methanol

needs to substitute for the acetonitrile adsorbed on the
surface. This equilibration is not favorable to methanol,
therefore a longer equilibration time may be anticipated. But
we can speed up the process by washing the column first
with a mobile phase containing a high methanol
concentration and then reducing the concentration to the 1%
level.
Even larger equilibration volumes may be needed, if the
concentration of an ingredient of the mobile phase is still
lower. Such an example are ion-pairing reagents. They are
typically used at concentrations of 5 mmol/L (5 mol/mL).
Luckily, their surface concentration is also usually rather
low, between 1 and 3 mol/m2. We can go through the same
calculation as before and find that we need between 150 to
450 mol of ion-pairing reagent per mL of column volume.
At the given concentration of ion-pairing reagent in the
mobile phase, we need between 30 and 150 column volumes
for delivering the necessary amount of ion-pairing reagent to
the column. Complete equilibration will necessitate still
larger volumes of mobile phase.
Consequently, in the first case under discussion, the factor
limiting the equilibration time is the concentration of the
relevant ingredient in the mobile phase and its final surface
concentration. If we want to speed up equilibration, it might
be advantageous to increase the concentration of the critical
mobile-phase constituent.
Q.: Let me summarize what you said. If we need to
equilibrate the column with a new constituent of the mobile
phase, we need to take into consideration how much is
needed to equilibrate the column and how much is dissolved
in the mobile phase. Then we can estimate, how long it takes
to deliver the needed amount of this constituent to the
column. But what about the opposite case, when we need to
remove an ingredient from the column?
A.: Correct. In the second case, the factor determining
equilibration is the retention factor of the compound that is
adsorbed on the surface in the new mobile phase. The
organic solvents typically used in reversed-phase
chromatography have a low retention factor in all typical
mobile phases. This is not the case for polar solvents in
normal-phase chromatography. For example, the retention
factor of methanol on silica with hexane as mobile phase is
very large. To remove methanol from the surface of a
normal-phase sorbent, we are better off to wash the column
first with a solvent of intermediate elution strength. It may be
best to choose the solvent that is used to modify the elution
strength of hexane. If ethyl acetate is the polar constituent of
our hexane-based mobile phase, we should first wash the
column with ethyl acetate to remove the methanol. Only then
should we equilibrate it with the final hexane-ethyl acetate
mobile phase.
If we understand what we need to do to equilibrate the
column fast, we can design efficient equilibration protocols.

www.pharmaguideline.com

Therefore it is important that we know the history of the

column.
Another example of strongly adsorbed mobile-phase
constituents are ion-pairing reagents. It is generally
recommended to dedicate columns used with ion-pairing
reagents to ion-pairing applications because these reagents
are difficult to remove from the surface. These reagents are
strongly retained on reversed-phase packings for two
reasons: one is hydrophobic interaction, the other is polar
interaction. If we had to deal with hydrophobic interaction
alone, we could wash our packing with a strong organic
solvent such as THF. But ion-pairing reagents also interact
strongly with surface silanols. Cationic ion-pairing reagents,
such as tetrabutylammonium salts, are particularly difficult to
remove. Therefore, any washing procedure must take this
complex interaction into account. Without special washing
procedures, a complete removal of ion-pairing reagents is
practically impossible.
One of the most difficult equilibration problems is the
equilibration with water in normal-phase chromatography.
Water is usually present only in very small amounts, and it is
strongly retained on normal-phase packings, especially silica
and alumina. Due to the strong retention of water,
equilibration with a dry hydrocarbon mobile phase may take
several days. To remove water from silica, it is preferred to
use a protocol with sequentially weaker solvents. A possible
sequence is methanol, ethylacetate, methylene chloride,
hexane. This is a faster way to obtain a "dry" column then to
attempt to wash a "wet" column with hexane.
Q.: Therefore, if the mobile phase does not contain strongly
retained components at low concentrations, I should observe
fast equilibration. What is the problem, if retention times
nevertheless drift slowly?
A.: In such a case I would suspect that the cause of your
retention time drift is not related to equilibration per se. It is
possible that the cause of the retention time drift is something
other than a column phenomenon. Is the mobile phase
composition slowly changing due to evaporation? Is the
temperature in the lab changing slowly? Maybe some
component of your sample is accumulating on the column.
Maybe we are simply looking at column aging, for example
through hydrolysis of the stationary phase. The latter two
phenomena are usually put into the category of column
conditioning rather than equilibration. We will discuss
column conditioning in a separate HPLC Troubleshooting
column.

17. Column Conditioning

Q.: What is column conditioning?
A.: In the last troubleshooting column we talked about
column equilibration. Column equilibration comprises all

phenomena that are reversible, while column conditioning

changes the column irreversibly. By column conditioning
you are changing the product that the manufacturer has
delivered to you, and the reproducibility of this step is your
responsibility. I generally advice against conditioning, but
there are some circumstances where column conditioning is
unavoidable.
Q.: Please give some examples of column conditioning!
A.: My first example is a commonly used procedure.
However, I want to point out that I do not recommend to use
this procedure. If you are using a strongly acidic mobile
phase, pH 2 or less, with a fully endcapped C18 column, you
will fairly rapidly hydrolyze the endcapping groups.
Consequently, the column will have a much larger silanol
activity than what has been delivered from the manufacturer.
This may influence the selectivity of a separation. It is
possible that somebody has developed a method using a
strongly acidic mobile phase, and by the time that methods
development was complete the column had already changed.
When a brand-new column is later used for the same assay,
the selectivity of the separation may be different. But the
separation comes back, when the column is "conditioned" for
a day or two with the acidic mobile phase. As you can see,
this is a permanent change of the column outside the
manufacturers specifications. I recommend against such a
procedure, but it is practiced in some labs, and detailed
column conditioning protocols have been set up for new
columns. Instead of changing the properties of a column with
such a conditioning process, I recommend to explore the
possibility to redevelop the method on a non-endcapped
packing. This avoids such a conditioning step in the future.
Another example of column conditioning occurs when
aminopropyl bonded phases are used in aqueous solvents.
The most common application is the use of this column for
the separation of carbohydrates by hydrophilic interaction
chromatography.
Typical
mobile
phases
are
water/acetonitrile mixtures with 60 to 90% acetonitrile.
When an aminopropyl column is exposed for the first time to
an aqueous eluent, the high concentration of the amino
groups in the pores of the packing creates a basic pH, which
results in a slow hydrolysis of the silica and the bonded
phase. The amount of bonded phase that is washed off
decreases exponentially with time, and soon nearly stable
retention times are achieved. However, the column has
changed significantly from its original properties and should
not be used for normal phase separations any more without
specifying the exposure to the aqueous eluents as part of the
column history.
If you want to use aminopropyl bonded phases for the
separation of carbohydrates, there is just no way around this
conditioning problem. Some manufacturers offer columns
that are dedicated for this application. In this case, the
manufacturer has performed the conditioning step for you.
Since this is performed with a fixed protocol to fixed
specifications, you are better off to purchase a

22

preconditioned column instead of doing the conditioning

yourself.
Another unpleasant, but apparently unavoidable
conditioning phenomenon happens with columns used in the
separation of proteins, very specifically with diol bonded
phases used for the aqueous size-exclusion chromatography
of proteins. It has been observed that for some proteins the
initial injections give smaller peaks than later injections. This
has been attributed to non-specific binding of protein to
adsorptive sites on the packing. To avoid this, it has been
proposed to inject first a large amount of a protein, for
example bovine serum albumin, to condition the column and
saturate the active sites. In general, I do not recommend this
conditioning procedure and suggest that you should observe
for yourself, whether or not your sample exhibits such a
phenomenon. When you observe an increase in peak height
with subsequent injections, you should make sure that this is
not caused by carry-over in the injector.
The last example that I want to discuss does not fit my
definition of conditioning, but is considered a conditioning
step by many people. The example involves dry reversedphase columns. It is possible that a column has dried out
during storage, because the fittings were not tightened well.
Also, radial compression cartridges are shipped dry. In these
cases, the column needs to be wetted first with an organic
solvent, such as methanol or acetonitrile. This drives the air
out of the pores and wets the surface. Now you can exchange
the methanol or acetonitrile for your mobile phase and obtain
reproducible retention times. In the case of a column that has
dried out accidentally, a problem may arise if the column has
been stored in a mobile phase containing buffer or salt. The
precipitated buffer may give rise to high backpressure during
the reconditioning step, and you may need to re-equilibrate
the column at low flow rates.
A related phenomenon is "hydrophobic collapse". Very
hydrophobic, well endcapped reversed-phase packings may
lose retention in highly aqueous mobile phases. It has been
observed that this can happen suddenly, for example when
the flow through a column is stopped. In other cases a
gradual decrease in retention has been observed over the
course of a few days. In both cases, retention can be restored
by washing the column with an organic solvent for a few
column volumes and re-equilibrating it with the mobile
phase.

18. Complex Sample Matrices

Q.: I am studying the metabolism of a newly developed drug.
The sample medium is blood plasma. Despite the fact that I
am using a sample preparation technique, I still get a
significant amount of interferences eluting in the
chromatogram. In addition to this problem, the drug recovery
from my spiked samples varies more than I would like. What
can I do?

23

A.: You are dealing with one of the most difficult separation
problems. Blood contains an innumerable number of
compounds that can interfere with the detection of your
analytes. In addition, the number of interferences increases
as the analyte concentration decreases. Therefore, the sample
preparation technique is an essential part of the
chromatographic method and needs to be optimized together
with the chromatographic method itself. The detection
method for the compound(s) of interest is, of course, an
important part of the method. More selective detection
methods such as derivatization and mass spectrometry
simplify the separation problem significantly, but these
techniques are not available to everybody.
Let us discuss therefore some of the options that you have
in sample preparation. Plasma samples contain a significant
amount of salt and proteins that can precipitate or adsorb on
reversed-phase packings. The adsorbed protein can easily
foul the column, resulting in changes in the separation and
ultimately clogging the column. Although packings have
been designed for the direct injection of plasma samples,
most analyses are performed using classical reversed-phase
packings. If one desires a reasonable column life, sample
preparation is unavoidable.
There are several sample preparation techniques available
for the pretreatment of plasma samples. The simplest one is a
protein precipitation. This technique entails the addition of
an organic solvent, for example acetonitrile, to the plasma
sample. At least two milliliter of acetonitrile should be added
for every milliliter of plasma. A significant amount of the
proteins present in the serum are precipitated in the presence
of the organic solvent. However, a large fraction of the
proteins remains soluble and can cause interferences with the
analytes. Also, all low molecular weight compounds, for
example lipids, stay in the sample.
Another sample preparation option is liquid-liquid
extraction. A separation using this technique requires two
immiscible solvents. Polar, water miscible solvents can not
be used for the extraction. Due to this fact, liquid-liquid
extraction works best for more non-polar analytes; it is not
suitable for very polar analytes. Since many samples also
have very polar metabolites, this method has limitations.
A third option - and the one which is most frequently used
- is solid phase extraction. It is a very convenient technique,
and due to its similarity to chromatography, solid phase
extraction
is
a very popular technique with
chromatographers. One can use several extraction techniques
to clean the sample; reversed-phase chromatography and ionexchange are the techniques that are most commonly
employed. As a result of its versatility and simplicity, solid
phase extraction is the sample preparation method that has
the broadest range of applicability. Therefore, it is the
sample preparation method that I commonly recommend.
Q.: I am using solid phase extraction for sample preparation.
The SPE cartridge is a reversed-phase cartridge. I am
experiencing low, variable analyte recoveries.

A.: Let us examine the SPE procedure in more detail. The

general procedure of a reversed-phase solid-phase extraction
method is as follows:
Step 1: load the sample
Step 2: remove polar interferences
Step 3: elute analytes, leaving behind the more non-polar
interferences
Before loading the sample onto the SPE cartridge, we must
first activate the cartridge. This is typically done by initially
washing the reversed-phase cartridge with an organic
solvent, usually methanol. Methanol is then replaced by a
polar solvent, usually water or buffer. The buffer should be
at the same pH as the plasma sample. This washing
procedure preconditions the cartridge to the sample. The
sample is then loaded onto the preconditioned cartridge. It is
important that the cartridge does not dry between the
conditioning step and the sample loading step. Cartridge
drying may be one of the causes of low and variable
recoveries on C18-type sorbents.
After the sample is loaded onto the cartridge, removal of
the more polar interferences is necessary. Table 1 gives the
constituents of a serum or plasma sample. Typical polar
interferences are salts, carbohydrates, and a significant
amount of the proteins. Salts and most of the carbohydrates
do not adhere to a reversed-phase sorbent and are removed
without difficulty. The elution solvent for most of these very
polar interferences is water or a buffer solution at the pH of
your choice. Some proteins are also eluted using these
conditions.
To remove all the interferences that are more polar than
your analytes, it is best to wash the cartridge with a solvent
that is more polar than your mobile phase. In general, most
protein interferences are removed using a wash of around 5%
organic solvent. You can also change the retention your
analytes relative to the interferences by changing the pH of
the wash solution. You have to examine this step carefully to
make sure that you are not loosing a portion of your analytes
in this step.
Elution of the analytes requires a stronger eluent than your
mobile phase. Many people use a straight organic solvent,
most commonly methanol, for the elution. Methanol is easy
to evaporate, thereby resulting in a sample that can be easily
reconstituted in mobile phase. However, an elution step
employing methanol is neither specific nor selective. In your
case, a potential problem can arise from strong interactions
of your analytes with surface silanols, which are not broken
by methanol. If this is indeed the case and the cause of your
variable recovery, you need to consider a sorbent that does
not contain silanols. Reversed-phase sorbents that are not
based on silica are commercially available.
Alternatively, more specific extraction procedures can be
designed. A manipulation of the pH in conjunction with an
increase in the organic modifier content is something to
consider. You should compare the recovery of your analytes
at neutral pH and at acidic pH. Of course, such an approach
will work only for compounds with ionizable functional
groups. However, the description of your analytical problem

strongly suggests that an analyte-silanol interaction is a

possible cause of your recovery problems. In such a case, an
alternative to silica-based sorbents might be the best
solution.

19. Hydrophobic Collapse

Q.: I have two reversed-phase C18 columns of the same type
from the same manufacturer that show a significant
difference (40%) in retention. What is the problem?
A.: This is obviously a serious problem. The retention time
reproducibility from column to column should be
significantly better, in the range of +/- 5%. For higher quality
packings, the retention time reproducibility is on the order of
+/- 2%, but achieving this level of reproducibility requires
good temperature control as well. Knowing this, the columnto-column reproducibility observed in your case is obviously
unacceptable. Let us investigate, what the problem could be.
First, let me ask the question if one of the columns has been
used extensively. It is not uncommon to find differences in
retention times between old columns and new columns. This
can be especially noted, if the old column has been used
close to the limits of the bonded phase pH stability. Also, it
is not unusual to observe a retention time change if sample
constituents have accumulated on the column. Accumulation
of sample constituents can cause a drastic change in
retention. In such instances, it is necessary to examine a
suitable washing protocol to remove the sample constituents
from the column. So, which of the two columns has been
extensively used?
Q.: Neither! Both columns are new, fresh out of the box. I
would have expected to achieve much better reproducibility
from brand-new columns.
A.: Indeed, column-to-column reproducibility should be
much better. Even batch-to-batch reproducibility of the
packing should be better than what you are observing. Batchto-batch reproducibility should be better than +/- 10%, even
better than +/-5% if you purchase the column from a
reputable manufacturer. Have you investigated whether both
columns were prepared from the same batch of packing?
Q.: Yes, I have. The manufacturer declared that both
columns were prepared from the same batch of packing.
A.: This makes the retention time differences even less
understandable. It appears that the only difference between
the two columns is the packing process. Column packing
processes are sufficiently reproducible such that packing
densities do not vary by more than +/-2%. Therefore,
retention times should not vary by more than +/- 2%. Even if
uncertainties in the mobile phase makeup were added to the

24

www.pharmaguideline.com

packing process uncertainty, it still would not explain the

large retention time discrepancies that you are observing.
Typically, errors in the mobile phase composition could be
an explanation of your observation. If the organic content of
the mobile phase varies by 1%, you may observe retention
time differences of 10%. On the other hand, differences in
the concentration of a mobile phase buffer of up to 20%
result in retention time differences of only 1% in many cases.
Another possibility would be some error in the mobile phase
pH. If the pKa of your analytes is close to the pH of the
mobile phase buffer, a 0.1 unit change in the pH can result in
retention time differences of 10%. Therefore a careful check
of the mobile phase pH is necessary.
Other external influences are less pronounced. Changes in
temperature can affect retention by roughly 10% per 5 C.
Therefore, changes in temperature are not likely to be the
cause of your problem. If you are dealing with an ion-pair
separation, it is not unusual to unknowingly use columns that
have not been completely equilibrated with the mobile phase.
It may often take several 100 mL of mobile phase to
equilibrate the column with an ion-pairing reagent, especially
if the mobile phase concentration of the ion-pairing reagent
is low.
Q.: I am not using an ion-pairing reagent. Also, let me add
that both columns were tested with the identical mobile
phase, and the results are reproducible. This excludes the
influences that you just described, doesnt it?
A.: This makes it even more difficult to explain your results.
How were the columns treated before you tested them?
Q.: Both were treated in exactly the same way. I took them
out of the box and I equilibrated them with mobile phase.
A.: Hmm... This is getting difficult now. To summarize, we
have two new columns, prepared from the same batch of
packing
material,
giving
significantly
different
chromatographic results. The only rational conclusion is that
there are some differences in the column equilibration
causing these changes in retention time. What is the
concentration of organic solvent in the mobile phase?
Q.: The mobile phase is nearly 100% aqueous. It consists of
98% buffer and 2% methanol. I removed the columns from
the box and equilibrated them with the mobile phase for one
hour before I started to inject my samples.
A.: Now we are getting closer. The low concentration of
organic solvent in the mobile phase raises the possibility that
the differences in the retention times observed by you are
due to differences in the wetting of the columns. Very
hydrophobic, well endcapped C18 columns can exhibit
significant differences in retention depending on their prior
history. If one of the two columns has inadvertently dried or
partially dried during the shipment, it can exhibit
significantly lower retention times than a fully wetted C18

25

column. This is due to the fact that water and mobile phases
with a very high water content do not wet the hydrophobic
C18 surface very well. When you use a mobile phase with a
high water content, only part of the C18 surface is therefore
available for retention. Consequently, the retention factors
are significantly lower than the ones observed on a well
wetted column. You can eliminate the problem by
reconditioning the column with 100% methanol, then reequilibrating it with mobile phase. This "rewetting" step
ensures that the entire surface of the packing is available for
interaction with the analytes. After you have done this for
both columns, the retention times should be very
reproducible, perhaps as close as 1%. Based on the
description of the problem and the fact that we have
eliminated all other possibilities, I am very certain that this is
the cause of your retention time problem.

20. Baseline Noise

Q.: I set up a method a while ago, and it was working fine
for a long time. Recently, the baseline noise has increased
significantly, about 5-fold compared to where it was earlier.
The method is a reversed-phase method, and I am using a
UV detector. What could be the problem?
A.: The very first things that come to mind in your case are
either a dirty cell window or a dirty mobile phase. Let us
discuss the dirty mobile phase first. Although this may be
more difficult to troubleshoot, it does not require any
disassembly of the detector cell for cleaning the window.
If you are using a reversed-phase method, you are likely to
use an aqueous mobile phase with methanol, acetonitrile or
tetrahydrofuran (THF) as the organic modifier. For any one
of the organic solvents you need to make sure that they are
HPLC-grade. If other grades are used, they often contain
small amounts of miscellaneous chemicals that can influence
the UV-background of the solvent. Especially when using
THF, you need to make sure that you are using an HPLC
grade. Other grades of THF contain antioxidants that can be
detected by UV detectors and therefore affect the UV
background. Since we are talking about THF: use of old THF
may also result in a high UV background due to the
formation of peroxides.
The other solvents usually do not deteriorate with time, but
methanol is less suitable than acetonitrile for detection at low
UV wavelengths. However, this is an intrinsic property of the
solvent and not something that would change with time and
show up after long use. But there are other ingredients of
reversed-phase mobile phases that can create background
problems. In many reversed-phase methods, buffers are used
to control the pH of the mobile phase. Most inorganic buffer
salts like phosphate are of sufficient purity that an increase in
the UV background is not very likely. But it is generally

recommended to use high-quality reagents, for example p. a..

If you are using triethylamine or related amines as a buffer
ingredient, then it is possible that the UV background of your
buffer varies with the purity of the amine. The quality of
amines also deteriorates with age, largely due to oxidation.
Under some circumstances, water can also be the source of
the problem that you have described. Usually, water is
subjected to a purification scheme such as ion exchange or
reverse osmosis. Problems with impurities can occur, when
the cartridges in the purification system are saturated and
need to be replaced. This can increase the background noise
of the chromatographic method. A good quality control
scheme for the water generated by the purification device
should warn the user when the useful life of the purification
cartridges has been reached. This is best accomplished by
running a reversed-phase gradient from water to 100%
organic solvent using a dedicated reversed-phase column.
The UV-trace obtained from this gradient is then compared
to a standard good UV-trace.
If you are using a dual-pump gradient system, you can use
an old column between the pump for the more aqueous
component of your mobile phase and the point where the
gradient is mixed. Impurities in your more aqueous mobile
phase are adsorbed on the precolumn, and you get a cleaner
baseline. Of course, it is necessary to clean the precolumn on
a regular basis. Unfortunately, this elegant solution to the
problem of removing mobile phase impurities is only
available to users of HPLC-systems with dual-pump
gradients.
Generally, you will see many more solvent impurities
when your detection method uses the low UV range between
200 and 215 nm than when you are using the standard
wavelength of 254 nm. This can make a significant
difference in the UV background.
Q.: In this assay, I am also using 254 nm as the standard
wavelength. I have already considered some of the issues
discussed above, but have not yet been able to solve the
problem. How about instrument issues like a dirty cell
window?
A.: This is exactly the point that I want to discuss next. If we
cannot find a problem with the mobile phase, the cause of the
increased noise is most likely related to the detector. The
most common cause of increased baseline noise is dirt in the
detector cell or a dirty detector window. Of course, an air
bubble in the detector cell will look exactly like dirt, but it
can be removed easily by briefly pressurizing the detector
cell or flushing the cell at higher flow rate. Often, a short
piece (4" long) of 9/1000 tubing can be used in the waste line
to prevent the formation of air bubbles in the detector cell.
However, if dirt has accumulated in the detector or on the
cell window, this simple trick will not change the
phenomenon at all.

The only way to clean the detector cell and the detector
window is to disassemble the detector cell and clean the
window manually. The disassembly of the detector cell will
also give you access to the body of the detector cell, and any
material that has accumulated in this area can be removed
without difficulty. To clean the cell window, one needs to be
careful to use gentle techniques. Often, the window can be
treated in an ultrasonic bath with a solvent that will remove
any film on its surface. You can try a soap solution or
tetrahydrofuran (HPLC grade). This is a rather mild
treatment and should always be tried first. If such a simple
washing does not remove the film, than one can try to clean
the window by rubbing it with a suitable piece of wet paper
or cloth. You can get suitable tools by going to a camera
store and ask for lens cleaning tissue and fluid. This should
solve the problem.
Since this procedure is involved, it is better to prevent a
contamination of the detector window. Unfortunately, there
can be many different causes for dirty detector cell windows,
and in many cases it is not at all clear, what causes the
contamination of the window. But generally, one should
prevent precipitation by using the same precautions as for a
column.
If the detector cell is clean and the problem still persists,
then the likelihood is high that the detector lamp has aged.
You should consult the detector manual to determine, if you
can replace the lamp by yourself or if you need technical
service. In most cases, the replacement of a UV detector
lamp is something that can be carried out by the HPLC
operator without difficulty.
If the problem persists after cleaning of the cell and
replacement of the detector lamp, the increased noise could
be due to issues with other parts of the HPLC equipment
such as the integrator or data system. In such a case, it is best
to have the system checked by trained service personnel.

21. Narrow-bore Columns

Q.: I understand that solvent consumption is dependent on
the column diameter. I would like to reduce the solvent
consumption in the laboratory and therefore would like to
use 2 mm columns. Unfortunately, the first 2 mm columns
that I purchased are not performing as well as the standard
columns. What is the problem?
A.: Unfortunately, your experience is not unique. I get
similar comments from many people who are trying small
volume columns for the first time. However, we should not
assume that the problem is necessarily a result of poor
column performance. More frequently, the lower-thanexpected column performance is due to instrument issues,

26

more specifically the instrument bandspreading of a standard

wt = wc +w

2
o

(1)

The peak width contribution from the column alone

increases with retention. The peak width contribution outside
the column remains constant. This system contribution is also
called extra-column bandspreading or extra-column effect.
The measurement of the extra-column bandspreading is
very straightforward: disconnect the column and inject
sample directly into the detector, using the same mobile
phase and flow rate as you use for the HPLC analysis. You
may need to dilute the sample, if the peak height exceeds the
dynamic range of the detector.
In order to go further, we need to specify the contribution
from the extra-column effect. We will assume that the peak
width contributions are all measured the same way,
specifically by either the tangent or the 4-sigma method. For
this method, the contribution of the column to the total peak
width is:
w 2 =16c

V 2R
N C

(3)
C

If this ratio approaches 1, then the apparent column

performance is similar to the true column performance. This
is realized, when the extra-column bandspreading wo
becomes negligible.
Q.: Can you show me with some concrete example
calculations
how the extra-column bandspreading
deteriorates the column performance?
A.: Yes. Let us look at figure 1. In this figure, I have plotted
the relationship between the plate-count ratio (equation 3)
and the retention volume. I have assumed for this calculation
that the column plate-count is 10,000, and that the extracolumn bandspreading (measured as the peak width by the
tangent method) is 80 L. One can see that the true platecount of the column is never reached, even at an elution
volume of 10 mL. An elution volume of 10 mL translates to
roughly a k' of 4 for a 4.6 mm x 150 mm column. The graph
also shows tha at an elution volume of about 2 mL, which is
roughly the point of elution of an unretained peak on this
column, the real plate-count is only half of the maximally
achievable plate-count.

Influence of Extra-Column Bandspreading

Realized Fraction of Column

Performance

HPLC instrument.
Before I explain this in detail, I would like to point out a
quick way in which you can determine if system
bandspreading plays a role in your standard assays. I
guarantee that you will be surprised about the magnitude of
the instrument effects.
Let us assume that you are running an isocratic method,
and that your assay contains several analytes that elute at
different retention factors. Check the plate-count that your
instrument measures for the different analytes. In many cases
you will observe that the plate-count increases with increased
retention. Now note that there is no rule in HPLC theory that
predicts a higher plate-count at increased retention.
Therefore, the observed increase in column performance for
the peaks with higher retention values is most often due to a
decrease of the influence of extra-column bandspreading.
Since this effect is usually present to some degree with
columns of standard diameters (4.6 mm, 4.0 mm and 3.9
mm), it will play a larger role with smaller diameter columns.
In the following, I will discuss this effect in more detail and
give you the equations that allow you to calculate the
deterioration of column performance due to instrument
influence.
In isocratic chromatography, the square of the peak width
observed by the detector w2t is the sum of the square of the
peak width inside the column w2c and the square of the peak
width of all effects outside the column w2o:

w R
16 VR
Na =
=
N
w 2 + w2 16 V 2 + w2 N

1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0

10

Elution Volume [mL]

Figure 1: Influence of extra-column bandspreading on

column performance

(2)

In this equation, NC is the column plate count and VR is the

retention volume of the peak of interest. What we are
interested in is the apparent plate count Na that the
combination of instrument and column delivers. This can be
derived from the previous two equations:

27

Let us consider now a reduction of the column diameter to

2 mm. The column dead volume of such a column is only
about 0.33 mL. This is the elution volume at which our first
peak is eluting. At such a low elution volume, the extracolumn effects are dominating the peak width, and less than
1/30th of the true column plate-count can be realized. The

www.pharmaguideline.com

elution volume of about 2 mL discussed above now

corresponds to a k of about 5. As we have seen above, one
obtains approximately only half the plates of which the
column is capable at this k. Only at a retention factor of
about 10 or higher, equivalent to elution volumes in excess
of 5 mL, are we achieving good column performance again.
For most chromatographic applications, a retention factor of
5 or lower is normal, and under these circumstances the
performance of the column is reduced drastically by the
extra-column effects. In summary, this is the fundamental
reason for the low performance of 2 mm i.d. columns on
normal HPLC instruments. Therefore, it is usually not the
column that is the problem, but the bandspreading of the
standard HPLC instrument.
Q.: This is fairly much in line with my experience. What can I
do to improve the situation?
A.: We have to reduce the extra-column bandspreading to a
significant level. To achieve roughly the same performance
level you have experienced with the large diameter column,
you need to reduce the extra-column bandspreading by the
same factor as the change in column volume. In the case
where you go from a 4.6 mm column to a 2 mm column, you
should reduce the extra-column bandspreading effects
roughly by a factor of 5. This means that if you had good
performance on a standard instrument with a bandspreading
of around 80 L, you should strive for a system
bandspreading of around 16 L for use with a 2 mm column.
To do this, you usually have to reduce the length of the
tubing from the injector to the column and from the column
to the detector. Also, one usually needs to reduce the
diameter of the connection tubing from the standard 9/1000"
tubing to a 5/1000" tubing. Naturally, you should reduce the
injection volume in proportion to the column volume. If your
injection volume for the standard column was around 25 L,
it should be only around 5 L for the 2 mm column. The next
issue is the reduction of the bandspreading in the detector
cell. You should use a detector cell that is designed for the
smaller diameter column. This cell should have a
significantly smaller volume than your standard detector cell.
Commonly, 3 L cells are readily available. Of course, that
does not get you quite to the factor of 5 in volume reduction,
but it is a start. The other thing that you have to realize is that
a reduction in the detector cell volume usually increases the
detector noise. Therefore if you need high sensitivity and
have enough sample to inject, the step to a smaller diameter
column is actually ill advised. Of course, your intentions
were to reduce solvent consumption, and this goal can indeed
be accomplished with smaller diameter columns.

22. Sample Solvent

Q.: Despite sample preparation, the concentration of my
sample analyte is very low, and I need to achieve high

sensitivity. When I inject 100 L of the sample, I see already

an unacceptable deterioration of the separation. What can I
do to improve the sensitivity of the assay?
A.: There are several different techniques you can employ to
increase the sensitivity of an assay. One of the first things to
consider is the choice of the detector or the detection
wavelength, if a UV detector is used. If several detection
principles are possible, and if the different detectors are
available in your laboratory, this might be a first approach.
For example, fluorescence detectors are several orders of
magnitude more sensitive than UV detectors, if fluorescence
is an option. Also, detection in the low UV range can be
significantly more sensitive than detection at higher
wavelength. Of course, one also gets more signal from
interferences in the low UV range, which can be
counterproductive. Electrochemical detection can be very
specific and highly sensitive, but electrochemical detectors
are not readily available in many laboratories. Post-column
and precolumn derivatization techniques can improve the
detection limits of an analyte quite significantly, but the
complexity of the procedures and/or the instrumentation
makes this approach less desirable.
Q.: I have already explored the possibility of other detection
methods. The analyte does not fluoresce, and the best UVwavelength has already been chosen. Other detectors are not
available in my lab.
A.: The next approach would be to preconcentrate the
sample using either a liquid-liquid extraction or a solid-phase
extraction technique. Solid-phase extraction is generally
simpler and more predictable than liquid-liquid extraction,
especially if you are using the same type of sorbent as you
use for the HPLC assay. Therefore, if your HPLC assay uses
a reversed-phase method, it is convenient to use a reversedphase solid-phase extraction technique to enrich your
analyte.
Q.: This is exactly what I am doing. The sample is
concentrated on a reversed-phase sorbent. Then interferences
are removed during the wash step, and finally the analyte is
eluted from the sample preparation device. It is then injected
into the HPLC instrument.
A.: May I ask you, which solvent you are using to elute the
sample from the SPE device and what the HPLC mobile
phase composition is?
Q.: I am using 1 mL of methanol to elute the analyte from
the SPE device. The HPLC mobile phase is 50 : 50
acetonitrile : phosphate buffer, pH 7.2.
A.: This sheds some light on your problem. You said earlier
that you experience a deterioration of the separation when
you inject 100 L of sample. Since the sample is dissolved in
methanol, this is not a case of column overload, but a case of

28

inappropriate sample solvent. Methanol is a stronger eluent

than the mobile phase. Therefore, the analyte will move more
rapidly down the column as long as there is a high
concentration of methanol around the analyte band. At low
injection volumes, the sample gets diluted with mobile phase
in the injector, the connection tubing, the column frit and the
column top. The analyte then enriches rapidly on the column
top and is separated from the excess methanol due to the fact
that the analyte is moving down the column much more
slowly than the methanol. At large injection volumes, the
initial dilution of the sample with mobile phase becomes
ineffective, and the sample band moves with the methanol at
a higher velocity down the column. This leads to a distortion
of the peak shapes, which in turn forces you to inject only
100 L or less.
Q.: This is indeed the case. What can I do to solve this
problem?
A.: There are two possibilities. The first suggestion is to
evaporate the sample to dryness and then redissolve it in the
mobile phase or a solvent composition whose elution
strength is weaker than that of the mobile phase, for example
30% acetonitrile : 70% buffer. This is best done in the
presence of an internal standard to monitor the concentration
of the analyte, but you probably use an internal standard
anyway for the sample preparation step. This is the classical
approach used to eliminate strong solvents that interfere with
the HPLC assay.
Another option is to dilute the sample with water (or
buffer) and then inject more (1). For example, if the sample
is diluted 1:1 or 2:1 with water, you can probably inject the
entire sample (2 or 3 mL) without peak distortion. Assuming
that you want to keep some of the sample for reanalysis, you
should dilute the sample 2:1 with water and then inject 1.5
mL of the sample. Under these circumstances, the sample
enriches on the top of the column and is then eluted as a
sharp band with roughly 5-fold increased peak height
compared to the current situation, without any distortion of
the peaks. If your injection system lets you do this, this is
clearly the more convenient approach.
Reference:
1. Uwe D. Neue and Ed Serowik, "Sample Dilution Increases
Sensitivity and Resolution", Waters Column VI, 2 (1996),
pp. 8-11

23. Gradient Scaling

Q.: I have developed a separation on a 150 mm x 4.6 mm
column. There is one pair that is not well resolved, and I
hoped to get better results from a longer 250 mm column.
Unfortunately, this did not turn out to be true: the resolution
of the critical peak pair is worse, and also other peaks have
shifted in the chromatogram. The manufacturer says that both

29

columns are from the same batch of packing. What is the

problem?
A.: If the packing in the 15 cm column and in the 25 cm
column is indeed identical, you should get the same
separation but with improved resolution using the longer
column. Your observation that the position of peaks is
shifting indicates that the packing is not the same in both
columns. Since the manufacturer insists that the packing in
both columns comes from the same batch and since you have
used the 15 cm column already for a while, one would
suspect that the 15 cm column has aged during your method
development and is not representative of the packing
material anymore. This situation is not uncommon and is
often the cause of significant frustration on the part of the
chromatographer. For this reason I always recommend to
verify and validate a new method after methods development
using a new column.
Q.: I indeed have another 15 cm column of the same packing
around. When I ran the same gradient on this column, I got
practically identical results as on the other 15 cm column.
This should prove that the packing has not aged, doesn't it?
A.: This is indeed a good possibility. But you said that you
are using a gradient method. This complicates the discussion
somewhat. How did you scale the gradient from the shorter
column to the longer column?
Q.: I did not scale the gradient. I ran the same gradient on
the short column and on the longer column. I used the same
flow rate and the same gradient time.
A.: This is most likely the cause of the problem. When you
use a gradient method on different columns, the gradient
volume should be scaled in proportion to the column volume
to give you results with identical elution patterns. Of course,
this increases the analysis time for the longer column, just as
in isocratic chromatography. There are a few additional
complications that can occur due to the system delay volume.
In the following, I will discuss the scaling of gradient
methods in a little more detail.
To obtain the identical gradient profile between different
columns, the gradient volume should be changed in direct
proportion to the column volume. You started with a 15 cm
column and wanted to obtain the same results on a 25 cm
column. Both columns had the same internal diameter.
Therefore, your gradient volume should be 25/15 = 5/3
larger for the longer column. If you keep the same flow rate
on both columns, the gradient duration and other gradient
event times should increase by 5/3.
The rule to change the gradient volume in proportion to
the column volume also holds, when you change column
diameter. If you want to scale a gradient from a 150 mm x
4.6 mm column to a 150 mm x 3 mm column you should
reduce the gradient volume by a factor of 2.35. Generally,
this will be automatic, since you are reducing the flow rate in

direct proportion to the column volume anyway. Under those

circumstances, you can keep the gradient profile constant and
get the same results.
All of the calculations until now assumed that the gradient
delay volume of the instrument that you are using is
negligible and/or has no influence on the separation. This is
in general true for well retained compounds that elute late in
the gradient. However, this is not necessarily the case for
early eluting compounds. In the commonly used single-pump
gradient systems, the gradient is generated on the lowpressure side of the pump, and on older systems there is a
significant delay until the gradient reaches the top of the
column. The elution pattern of early eluting compounds can
be affected by this gradient delay volume. The ratio of the
gradient delay volume to the column volume should be held
constant, when we change column dimensions.
Unfortunately, this can be a significant problem, when one
wants to scale a gradient from a larger volume column to a
smaller volume column. Fortunately, you want to scale the
gradient from a smaller volume column to a larger volume
column, which simplifies the situation.
In order to adjust the gradient delay volume, you first have
to measure it for the system that you are using. This is best
done by running a step gradient at 1 mL/min from methanol
to methanol with a small amount of a UV absorber, for
example 1 % acetone in methanol, without a column in place.
One measures the time from the start of the program to the
time that the step has reached half the height of the final
concentration. Multiplying this time with the flow rate results
in the delay volume of the instrument. Since the start of the
programmed gradient at the column top will be delayed by
this volume on the small column, we need to delay the
gradient on the larger column by the same ratio of delay
volume to column volume to obtain exactly the same
gradient profile on both columns. For example, if the system
delay volume was 1 mL when the 150 mm x 4.6 mm column
was used, it should be 1.66 mL for the 250 mm x 4.6 mm
column. Therefore we need to add an isocratic delay of 0.66
mL to the beginning of the gradient program for the 250 mm
column to fully account for the differences in the gradient
profile between the actual gradient obtained on the 150 mm
column and the 250 mm column. After this is done, identical
elution profiles are obtained on both columns.

24. Column Storage

Q.: I asked my colleagues how I should store my HPLC
columns. Unfortunately, everybody gave me a different
answer. What should I do?
A.: In most cases, there are many different ways in which an
HPLC column can be stored that have little effect on column
longevity. Therefore, it is quite possible that many of the
answers that you got were correct. However, the best way of
storing columns depends on the type of the stationary phase.
Also, some column storage conditions are more convenient

than others. Let me first discuss this in general terms, then I

will discuss specific and special considerations that depend
on the nature of the packing.
The most convenient way to store a column is in the
mobile phase in which it is commonly used. The biggest
advantage of this approach is that reequilibration of the
column with the mobile phase is very fast. Therefore, one
can get reproducible results within a short time after start-up.
This approach is especially recommended for normal-phase
chromatography, where a change to a storage solvent
different from the mobile phase can result in lengthy
reequilibration times. However, one needs to think this
approach through very carefully. Bonded-phase columns
often change slowly in the commonly used mobile phases.
Therefore, the convenience of storing the column in the
mobile phase needs to be balanced against the reduction in
column life.
In the following, I will discuss storage conditions for
various HPLC packings based on the nature of the packing.
Let me first talk about silica and alumina. Both of these
packings are very stable in the mobile phases in which they
are commonly used. These mobile phases commonly
comprise organic solvents with small amounts of polar
modifiers, including water. Due to the low concentration of
the polar modifiers, it often takes a considerable time to
equilibrate the columns with mobile phase. Therefore it is
best and most convenient to store the columns in mobile
phase.
The situation is similar for polar bonded phases used in
normal phase chromatography. Equilibration with mobile
phase is somewhat more rapid than with silica or alumina;
nevertheless, it is still fastest to store the columns in mobile
phase. Some mobile phase ingredients are not suitable for
certain columns. Therefore, they also should not be
considered for column cleaning or column storage. An
example is the incompatibility of amino columns with
acetone.
Polydivinylbenzene-based packings such as SEC packings
and ion-exchangers are chemically very stable, but they swell
and shrink in different solvents. The manufacturers usually
supply you with information about which solvents are
compatible with a particular column. General statements can
not be made, since the stability of the columns depends on
the packing conditions. However, the best storage solvent for
these columns is the solvent in which the columns are used.
For silica-based bonded-phase columns used in reversedphase chromatography, the situation is more complicated.
Water attacks the bonded phase, but the process is very slow
in the commonly used pH range from pH 2 to pH 8. In
addition, it depends on the nature of the bonded phase. The
commonly used C18 and C8 columns are sufficiently stable
to be used for several months with little change in their
hydrophobicity. However, bonded phases based on shorter
chains can hydrolyze measurably within a few weeks.
Therefore, the best storage conditions for reversed-phase
columns depend on the frequency of their use. If they are
used every day, or even every few days, it is most convenient

30

www.pharmaguideline.com

to store them in the commonly used mobile phase. On the

other hand, if they will not be used for an extended period of
time, it is best to store them in a solvent that prevents
hydrolysis, commonly acetonitrile or methanol.
However, some reversed-phase columns are not stable in
these solvents. Some cyano columns can void when stored in
organic solvents. For these columns it is best to use the
mobile phase for storage, despite the danger of faster
hydrolysis. Read the manufacturers recommendations!
Amino columns are often used in acetonitrile/water mobile
phases for the analysis of carbohydrates. Unfortunately, the
amino group creates a basic pH in the pores of the packing,
which leads to a slow loss of the functional group. Therefore,
amino columns used for this application are best stored in
acetonitrile instead of the mobile phase, at least for long term
storage.
Additional considerations need to be made for the longterm storage of columns in highly aqueous mobile phases
that allow the growth of algae or bacteria. Often, the addition
of sodium azide to the storage buffer is recommended. If
feasible, organic solvents are better solvents for long-term
storage.
As a final remark, it should be emphasized that it is always a
good idea to record the storage solvent in a permanent file.
This can be done conveniently by recording the column type
and the serial number. When you create such a file, it is also
worthwhile to record the date and time of the last use of the
column, as well as the number of analyses run since the last
storage. Such a column history file gives you a permanent
record that allows you to go back and check column use and
column life time. An alternative and less efficient way is to
mark the column with your choice of the storage solvent. Of
course, this approach works best if you always use the same
storage solvent for this particular column.

25. Paired-Ion Chromatography

Q.: I am using a chromatographic method on a C18 column
that employs an ion-pairing reagent, octyl sulfonic acid. The
mobile phase consists of 20% methanol and 80% of the
aqueous buffer solution. The aqueous buffer consists of 5
mM of the ion pairing reagent and 50 mM acetate buffer; the
pH of the aqueous solution is adjusted to pH 4.0 with acetic
acid. What bothers me is the lengthy equilibration time
necessary to get consistent retention times. Whats wrong?
A.: Most likely nothing is wrong. Long equilibration times
are typical when ion-pairing reagents are used. Let me
discuss this in detail. Once we understand what the issues
are, we also should be able to find a faster protocol.
Ion-pairing reagents are used with reversed-phase columns
to add ion-exchange properties to the stationary phase.
Purely hydrophobic interaction is influenced very little as the
ion-pairing reagent is added to the mobile phase. Therefore,

31

the ion-pairing reagent opens a new dimension to the

separation.
The columns used in ion-pair chromatography are mostly
C18 columns, just as in your case, but C8 columns often
work as well. The ion-pairing reagent is adsorbed on the
surface of the reversed-phase packing material. The
concentration of the ion-pairing reagent on the surface of the
packing depends on its concentration in the mobile phase.
The retention times of compounds that interact with the ionpairing reagent depend on its surface concentration. If the
mobile phase concentration is low, the retention of analytes
with a charge opposite to the charge of the pairing reagent
increases in direct proportion to the concentration of the ionpairing reagent. If the concentration is high, around 10
mmol/L, the retention of such analytes often becomes
independent of the concentration of the ion-pairing reagent.
The retention of non-ionic compounds is nearly unaffected
by the concentration of the ion-pairing reagent. Therefore,
one can use the concentration of the ion-pairing reagent to
influence the retention of ionic compounds relative to noninteracting compounds. This makes it possible to affect the
selectivity of a separation.
These observations are only true for columns that are fully
equilibrated with the ion-pairing reagent. Since the ionpairing reagent is adsorbed onto the surface of the packing, it
may take some time until the column is equilibrated with the
reagent. The surface concentration of the ion-pairing reagent
depends on its mobile phase concentration as well as on the
organic content in the mobile phase. But we can make an
estimation on the amount of reagent that is adsorbed on the
column by assuming that the surface concentration is around
1 mol/m2. A column may contain around 2 g of packing
material, with a specific surface area of 300 m2/g. This
means that it contains around 0.6 mmol of ion-pairing
reagent at full equilibration. If the mobile phase
concentration of the ion-pairing reagent is 5 mmol/L, you
need about 120 mL of mobile phase to send enough ionpairing reagent to the column to achieve a surface
concentration of 1 mol/m2. In reality, a complete
equilibration may actually take somewhat longer, maybe 150
mL of mobile phase. If you equilibrate the column at 1.5
mL/min, this means that you need to wait for 100 minutes nearly two hours - before the column is equilibrated and
ready for analyses.
Q.: This appears to be the problem that I am running into.
What is the solution?
A.: What you are probably doing is to convert the column to
an organic solvent for storage. While this is a good approach
for normal reversed-phase separations, it is a problem when
you are using ion-pairing reagents. The best solution to the
problem is to store the column in mobile phase, at least for
overnight storage and for storage over weekends. If you do
this, the column should be completely equilibrated with the
mobile phase after only a short purge of 10 or less column
volumes. Of course, since you are using a buffered mobile

phase, the fittings and end-caps of the column should be well

tightened to prevent the column from drying out during
storage.
Usually, I recommend storing a column in an organic
solvent, if the column is not used for a period of time longer
than a weekend. However, in the case of ion pairing reagents, I
generally recommend to store the column in mobile phase
due to the lengthy equilibration times. Only if you intend to
not use the column for an extended period of time, maybe
around a month or so, should you consider storing the
column to an organic solvent.
Q.: Does this apply also to reagents like triethylamine, which
are often used to suppress the tailing of basic compounds?
A.: No. These reagents are primarily constituents of mobile
phase buffers. They are adsorbed on the packing, but due to
the high concentration in which they are typically used, there
is little concern about lengthy equilibration. Buffers are
commonly used at concentrations around 50 mmol/L, and the
concentration of the buffering reagents on the surface of the
packing are much lower than the common concentrations of
ion-pairing reagents. Therefore, the equilibration of the
column with the buffering reagent is much faster than the
equilibration with the ion-pairing reagents. Since the issue of
long equilibration times does not exist with these simple
buffering reagents, columns should be stored in an organic
solvent if they will not be used for several days or longer.

26. Hydrolytic Stability of Reversed-Phase

Packings
Q.: Some manufacturers claim that their silica-based
reversed-phase columns can be used to pH 9 or 10, while
others recommend not to use a pH above 8. I recently was
forced to use a reversed-phase column at pH 9, since this was
the only condition which separated all the compounds of
interest without interferences. While this was outside the
recommended use range of the manufacturer, column life
time was quite acceptable. I now wonder how seriously one
should take the manufacturers recommendation about the pH
range over which their columns should be used.
A.: This is indeed a good question. It is probably best
answered by your experiments: if you are happy with the
lifetime that you are able to achieve than there is nothing
wrong with the use of a higher pH than the manufacturer is
recommending. However, I would make sure that the same
separation can be obtained on a brand-new column as on a
column that has been used for a while under your conditions.
If this is indeed the case, then there is little reason to expect
trouble.
The pH stability of packings is a more complex issue than
can be stated in a simple rule. Let me explain this in a little
more detail to provide a better understanding.

At alkaline pH, hydroxyl ions (OH-) can attack and

dissolve the silica. The speed of the process depends on the
concentration of the hydroxyl ions in the mobile phase, their
access to the surface of the silica, and the solubility of the
dissolved silica in the mobile phase. As you can see, the
concentration of the hydroxyl ions, which is determined by
the pH of the mobile phase, is only part of the story. In
addition, the speed of all these processes is a function of the
temperature. What works well at room temperature may
represent an unacceptably short column life at 60 C.
The access of the hydroxyl ions to the silica plays a crucial
role in the stability of a packing. A dense coverage of the
surface of the silica with a C18 or C8 ligand improves the
stability substantially. Also, the protection of the surface
through a good endcapping process is important. The total
hydrophobic ligand density is probably a reasonable measure
of the protection of the silica surface from the attack of the
hydrophilic hydroxyl ions. Consequently, one can expect that
modern packings with a high surface coverage are more
stable than packings with a low surface coverage. In
addition, the quality of the endcapping process may play an
important role.
At alkaline pH, the silica itself dissolves. Therefore, the
nature of the ligand plays a secondary role only. The stability
of a bonded phase based on a monofunctional silane is not
dissimilar to the stability of a bonded phase based on a
trifunctional silane, at equal coating level. However, the
access of the hydroxyl ions to the silica surface plays a
crucial role. Therefore, the stability of a monofunctional
bonded phase with a bulky isopropyl side chain is inferior to a
standard bonded phase, simply because the maximally
achievable surface coverage is lower.
If the column is run continuously under the same mobile
phase conditions and is never washed with an organic
solvent, the desorption and dissolution of the bonded ligand
can be very slow. Therefore, retention time changes can be
very small. Nevertheless, the underlying silica is dissolving
slowly. As a consequence, a sudden collapse of the column
can be experienced, with little indication before this event as
to the slow deterioration of the column.
Of course, the silica density is a crucial factor in such a
case. Silicas with a high specific pore volume are less stable
than silicas with a low specific pore volume, simply due to
the fact that they have a weaker skeleton. The porosity of a
silica may range from 40% to 70%, but the change in
strength may easily be 10-fold. Therefore, one can expect
substantial differences in the properties of a bonded phase
simply based on the density of the parent silica. In addition,
the specific surface area decreases as the pore size of the
packing increases. Therefore, packings with a larger pore
size are more stable than packings with a small pore size, all
other parameters kept constant (1).
The nature of the buffer ingredients is also crucial to the
stability of the packing. At equal pH, organic buffers such as
a Tris, citrate or HEPES buffer are much less aggressive than
the commonly used phosphate buffers (2). Also, borate and

32

www.pharmaguideline.com

References:
1. T. Walter, B. Alden, P. Casellini, paper presented at
HPLC 97, Birmingham, UK
2. H. A. Claessens, M. A. van Straten, J. J. Kirkland, J.
Chromatogr. A. 728 (1996), 259-270

the curve is not a good place to be in, since increased

analysis time is accompanied by decreased resolution. When
you use a 5 micron 4.6 mm column at 1 mL/min with a low
viscosity mobile phase, you may be fairly close to this point
of decreasing resolution with increasing analysis time, but
the exact location of this point depends on your analytes and
the mobile phase composition. For common reversed-phase
mobile phases and common reversed-phase columns, you
may be far away from the point of maximal performance. Let
me examine this in more detail.

Resolution

glycine buffers have been shown to be less aggressive even

at pH 10.
It should be pointed out that in the literature studies
available the stability of a packing is commonly tested under
isocratic operating conditions. If you are forced to switch
occasionally to an organic solvent to clean the column from
contaminants, you may also wash off unbonded but adsorbed
ligand. Therefore, cleaning cycles may impact the stability of a
column quite drastically.
All of these observations hold for the commonly used C18
and C8 ligands. More polar ligands such as those used for
the preparation of CN packings exhibit significantly lower
stability, even under normal operating conditions. At pH 7,
the hydrolysis of a CN packing can be by a factor of 1000
faster than that of a C18 or C8 packing.
As you can see, you can get reasonable column life at
higher pH than commonly recommended, if you are using the
correct combination of operating conditions. The most stable
columns are based on a high-density silica, have a dense
coverage of a C18 or C8 bonded phase and are endcapped.
The nature of the buffer ingredients can have a drastic effect
on column life and should be chosen carefully. However, if
your separation demands it, and if you achieve a column life
that is acceptable to you, there is nothing wrong with an
exploration of the frontiers of column stability.

0.5

1.5

Flow Rate [mL/min]

Figure 1
Resolution as a function of flow rate for a 5 m 15 cm
column for the conditions assumed here.

27. Optimal Flow Rates

Q.: I am always using my reversed-phase column (5 m, 4.6
mm x 150 mm) at a flow rate of 1 mL/min. A colleague told
me that I can get better resolution at 0.5 mL/min. Upon
reducing my flow rate according to his suggestion, I do not
see much of an improvement in my separation. Additionally,
the runtime has been increased 2 fold. Nevertheless, I would
like to understand the concept and would appreciate, if you
would discuss the influence of flow rate on resolution.
A.: Gladly! Hidden behind your question is the dependence
of the height-equivalent to a theoretical plate on the linear
velocity. Both are terms of chromatographic theory that most
practitioners forget the moment they complete the
introductory course to chromatography. What I will try to do
in the following is explain the phenomena in simpler terms.
In isocratic chromatography, resolution usually decreases as
the flow rate is increased and increases as the flow rate
decreases (see figure 1). However, at a particular flow rate,
which is analyte and mobile phase specific, the resolution
reaches a maximum. If I decrease the flow rate any further,
the resolution will decrease again. Obviously, this region of

33

First, let us explore the rules that determine the position of

the maximum column performance. It can be calculated by
using the classical van Deemter equation, given here in its
reduced form:
h = A+

+C

(1)

h is the reduced plate height, is the reduced velocity, and

A, B, and C are coefficients describing the packed bed
quality, the diffusion of the sample in the packed bed and the
mass transfer of the sample into the packing, respectively.
We really do not need to spend a lot of time on this equation.
What we need to do is find the minimum of this equation,
which is simply derived by calculating the first derivative
and equating it to 0. When we do this we obtain for the
reduced velocity at the minimum:
B
(2)
min =

The value of B is typically about 1.5, and C is typically 1 /

6, making the coefficient under the square root about 9 (1).
Using this information and converting from the reduced
velocity to the linear velocity, we obtain

u min =3

theory. The last step is really a simple conversion to the

(3)

dp

umin is the linear velocity, at which maximum column

performance is reached. From the last equation, one can see
that the linear velocity for achieving maximum performance
depends on the diffusion coefficient of the analyte, Dm, and
the particle size of the column, dp. The smaller the particle
size, the higher is the linear velocity at which maximum
column performance is reached. Also, one can see that this
velocity depends on the diffusion coefficient of the sample,
which in turn depends on the mobile phase composition.
Q.: Thank you for your nice theoretical exploration! So,
what am I going to do with this now? I dont know anything
about the diffusion coefficient of my analytes.
A.: Luckily, you dont need to know anything about it. There
is a way around this issue that is very practical. Just bear
with me one more moment. The theories of diffusion in
liquids relate the diffusion coefficient of an analyte to the
viscosity of the solvent. With a few assumptions, one can
relate the diffusion coefficient in an aqueous or polar solvent at
room temperature to the viscosity of the mobile phase
using the following equation:
6

Dm 1.710

1
V

(4)

0.6

is the viscosity of the mobile phase in Poise, and V is the

molar volume of the solute, in mL. Most analytes have a
molecular weight between 200 and 500 Dalton, which results
in diffusion coefficients ranging from Dm 0.4*10-7 * 1/ to
Dm ~ 0.7*10-7 * 1/. Let us just assume that a typical
diffusion coefficient is about Dm ~ 0.6*10-7 * 1/. Therefore,
the optimum linear velocity is
u min =1.810

(5)

dp

(6)

1000 t L

Substituting this into equation 5 and assuming that the

typical total column porosity is about t = 0.7, we obtain
p =1.2510

p[psi]= 200

L
dp

(8)
3

For example, if I want to know the pressure at the optimum

performance point for a 150 mm column packed with 5 m
particles, I get 240 psi.
Q.: This looks like a good rule of thumb. Should I use the
column at this pressure?
A.: These considerations represent a simple way to get you
calibrated for the point at which maximum column
performance is expected for normal analytes. But you have
to realize the limitations of this estimation. If your analytes
are smaller than about 200 Dalton or larger than about 500
Dalton, this estimate is not going to work. For smaller
analytes, the maximum plate count is reached at a higher
pressure, and for larger analytes at lower pressure.
In addition, I generally would not run the column at the
maximum plate-count, but at a linear velocity that is about a
factor of 2 higher than this estimate. This is generally the
point of the best compromise between column performance
and analysis time. Therefore, I would select the flow rate that
gives me a backpressure of about 500 psi. This is the best
choice from the standpoint of overall column performance.
Of course, many separations do not need the best column
performance. However, if you have optimized your
separation around this point, you can then select other
conditions that might speed up your separation. You can
increase the flow rate or use a shorter column, or a
combination of both.

dp

This means that the optimum linear velocity is inversely

proportional to the viscosity of the mobile phase. The higher
the viscosity, the lower the velocity! This is worth knowing,
but is still not yet anything that we can deal with easily in
practice. However, the column backpressure depends on the
viscosity of the mobile phase, and this is something that we
can read off the instrument. The relationship between the
linear velocity and the backpressure can be obtained from the
Kozeny-Carman equation:
u=

dimensions commonly used in the US. I also rounded the

numbers to get a simple formula that is easy to remember. If
you use the particle size in m, the column length in mm and
you want to get the pressure at the optimal linear velocity in
psi, you should use the following simple formula:

L
dp

3[

dyn / cm2]

(7)

The column backpressure at the point of optimum column

performance depends only on the column length and the
particle size. This is a beautifully simple result after this tourde-force through three independent branches of column

Reference:
1. U. D. Neue, in "HPLC-Columns, Theory, Technology and
Practice", Wiley-VCH, 1997

28. Carbon Load

Q.: I always thought that the carbon load of a reversed-phase
packing determines the retention of a compound - at least for
compounds that interact with the packing mainly by
hydrophobic interaction. A colleague told me that this is an
oversimplified view. I would appreciate a good explanation
of the dependence of retention on carbon load.
A.: Your colleague is correct: the commonly quoted "carbon
load" of a packing is not the factor that determines retention.
The story is a little bit more complicated. Let us discuss the
simple case of purely hydrophobic interaction first!
The question of retention can be easily analyzed, if we
view reversed-phase chromatography simply as a partitioning

34

mechanism. In this mechanism, the retention factor k of a

compound is determined by its distribution coefficient K
between the stationary phase and the mobile phase and the
volume of the stationary phase VS and the mobile phase VM:
S
(1)
k = K V
VM

The ratio of the volume of the stationary phase to the

volume of the mobile phase is called the phase ratio. The
volume of the mobile phase in the column is just the
retention volume of an unretained peak. Therefore the
adjusted retention volume VR of a peak is:
VR=VR VM = VSK
(2)
This means that the retention volume of a compound is
simply proportional to the volume of the stationary phase in
the column. The volume of the stationary phase in the
column is however not proportional to the commonly quoted
carbon load. This is due to the fact that the specific pore
volume of different silicas is different. A silica with a small
specific pore volume is denser than a silica with a high
specific pore volume. Therefore, more silica can be packed
into a column, and more stationary phase is packed into a
column as well. The amount of stationary phase in the
column can be calculated as follows:
VS=V (1 ) f
(3)
C

where VC is the column volume, and I is the interstitial

fraction in the column, which is constant for all practical
purposes. The factor f contains all the components that vary
with the type of packing:
V st g
(4)
f =V

pore

g +V silica

g is the number of grams in a column, Vst is the amount of

volume of the stationary phase in the column, proportional to
the % carbon given in the literature, Vpore/g is the specific
pore volume, and the last factor Vsilica/g is the inverse of the
skeleton density of silica, which is 2.2 g/mL. This factor f
determines the retention of the packing. All the components
of this factor are commonly given in the manufacturers
literature, so it can be used without difficulty to estimate the
retentivity of one packing compared to another.
Let me show this in a simple example! Let us compare
three different C18 packings: Nova-Pak C18, Spherisorb
ODS2 and Lichrosorb RP18. The carbon content of the
three packings is significantly different (data from Phase
Separations HPLC Columns and Supplies Catalog): 7.3% for
Nova-Pak C18, 11.5% for Spherisorb ODS2 and 16.2%
for Lichrosorb RP18. Thus if retention was determined by
the carbon content of the packing, one would expect 2.5
times more retention for Lichrosorb RP18 than for NovaPak C18, with Spherisorb ODS2 somewhere in between.
However, the packing density of the different packings is
significantly different, because of the differences in the
specific pore volume between the three packings (table 1).
Consequently, if one calculates the factor given in equation
4, one will discover that this factor is very similar for the

35

three packings, and that therefore the hydrophobic retention

for the three packings is very similar as well.
Table 1
Example calculation
Nova-Pak C18
Spherisorb ODS2
Lichrosorb RP18

% Carbon
7.3%
11.5%
16.2%

Vpore/g Factor f
0.3
0.10
0.5
0.12
1
0.11

In summary: the carbon content alone cannot be used to

compare the retention of packings with significantly different
porosities. However, all the information that one needs for a
good estimate of the hydrophobic retention of a packing is
readily available in the manufacturers literature.
Q.: This is enlightening. It would be nice if the column
manufacturers would use this factor in their literature. If
would make the comparison of packings easier. You
mentioned that this applies to hydrophobic retention only. I
assume that the other factor to consider is the silanol activity
of a packing?
A.: Yes, you are correct. The silanol activity of a packing is
the second important factor determining retention on
reversed-phase packings, especially for basic analytes. Of
course, the influence of silanols on the retention of an analyte
depends on the properties of an analyte. For neutral
hydrophobic analytes, the activity of silanols on the surface
of a packing is unimportant. For basic compounds, the
retention factor can increase 10-fold due to the activity of
silanols. For other polar compounds, more subtle differences
in selectivity can be observed. Therefore, a knowledge of the
silanol activity of a packing plays a role in our judgement of
the usefulness of this packing for a separation. One can get a
first impression of the silanol activity of a packing by
checking whether a packing is endcapped or not. Nonendcapped packings have a higher silanol activity than
endcapped packings.
Most of the modern reversed-phase packings are well
endcapped. Thorough endcapping results in good peak shape
for basic analytes without the need of mobile phase
modifiers. Nevertheless, packings with a high silanol activity
can be used even for basic analytes to obtain different
selectivities than those available with the fully endcapped
packings. Of course, modifiers such as triethylamine or
octylamine often must be added to the mobile phase to
improve peak shapes for basic compounds. This is a nuisance
to many chromatographers, and the reason why fully
endcapped packings are preferable for most applications.
Whether a packing is endcapped or not unfortunately gives
only a very rough impression on the activity of the silanol
groups of different packings. Therefore, the catalogues of
some column suppliers contain additional, more detailed
information on the activity of silanols (1, 2). This
information can then be used to select packings that are
significantly different from each other for your next

www.pharmaguideline.com

applications development. After all, the differences between

different columns are an advantage in methods development.
References:
1. Phase Separations HPLC Columns and Supplies Catalog
2. Alltech Chromatography Catalogue 400

29. pH Control
Q.: I have recently read an article that recommends the use
of mobile phases with an acidic pH for acidic compounds. It
was suggested that using acidic mobile phases reduces the
ionization of the acidic analytes and the silanols. This is
supposed to improve the peak shape and reduce tailing. Can
you explain the reason for this phenomenon?
A.: In general, tailing peaks are observed quite commonly on
reversed-phase packings with basic analytes at neutral pH.
This phenomenon is due to the interaction of the bases with
the silanols on reversed-phase packings. Tailing of acidic
analytes on the other hand is quite rare, at least under
circumstances where complicating factors such as
complexation or size-exclusion phenomena can be excluded.
There is no reason to expect an interaction between the
negatively charged acidic analytes and the negatively
charged surface silanols could cause such a phenomenon.
Therefore, the recommendation to exclusively use acidic
mobile phases for the separation of acidic analytes seems to
be quite limiting. Can you give me an example of this
improvement in peak shape?
Q.: Yes. In the article that I read, the example analyte was
ibuprofen, which is a simple hydrophobic analyte with an
acidic functional group. A C18 column was used, and the
mobile phase was a mixture of acetonitrile and buffer. The
buffer was a 5 mM phosphate buffer at pH 4.4. As you can
see from the chromatogram (Figure 1a), significant tailing
was observed. In the article, the pH of the mobile phase was
then changed from pH 4.4 to pH 3.0, using the same 5 mM
phosphate buffer, and a significant improvement in the peak
shape was observed. Additional improvements were shown
at pH 2.5 with 0.1% trifluoroacetic acid. This demonstrates
quite clearly the improvement in peak shape for ibuprofen at
acidic pH, doesnt it?
A.: I strongly disagree. This set of experiments does not
demonstrate at all, that you need an acidic mobile phase to
suppress the tailing of acidic analytes. The only thing that
this demonstrates is the fact that one should use a buffer in
the mobile phase. Phosphate has two pKa values, one is
around 2, the other around 7. Therefore, phosphate buffers
have their optimal buffering capacity around pH 2 and
around pH 7. At pH 4.4, phosphate has no buffering capacity
whatsoever. Therefore, the tailing of the ibuprofen peak with
the phosphate buffer at pH 4.4 is due to the lack of pH
control under these conditions.

Figure 1: Chromatogram of ibuprofen at pH 4.4 using 60%

acetonitrile and 40% of
a. a 5 mM phosphate solution and
b. a 5 mM acetate buffer.
This can easily be demonstrated in a simple experiment. In
this experiment, we keep the pH at the same value as in your
example, at pH 4.4, but we use a true buffer with good
buffering capacity at this pH. This can readily be
accomplished utilizing an acetate buffer. Figure 1b shows the
resulting chromatogram of ibuprofen at pH 4.4 using a 5 mM
acetate buffer. As you can see, good peak symmetry has been
achieved, and no tailing is noticed. This demonstrates clearly
that the original cause of the tailing is not due to the mobile
phase pH, but rather due to the fact that the mobile phase is
not buffered. Consequently, good peak shapes and good
results are achievable with acidic analytes at any pH value,
provided that the mobile phase is properly buffered. This
gives you a wider choice of options in the optimization of a
separation.
If you are dealing with ionizable compounds, the
manipulation of the pH value is a very powerful tool in the
development of a separation (1). Frequently, as the pH is
changed, the elution order of peaks changes. The effects that

36

pH changes have on the selectivity of a separation are

commonly much more powerful than the variation of the
organic solvent. Therefore, the suggestion that you should
not use the tool of pH manipulation in your methods
development is counterproductive. For many compounds, but
especially for acidic analytes, an interaction with surface
silanols is unlikely to cause a problem. If you use one of the
newer reversed-phase stationary phases based on high-purity
silicas, then you often wont encounter tailing peaks even at
neutral pH using basic analytes.
As we have seen in this example, the proper control of
mobile phase pH is not only important for the peak shape,
but it is also vital for the ruggedness of the separation. How
well the pH is controlled, depends on the buffering capacity
of a buffer. In turn, the buffering capacity is a function of the
buffer concentration and the difference between the pH and
the pKa of the buffering ions. In most HPLC separations, the
concentration of a buffer is around 50 mM. At this
concentration, the buffer can be used to control the mobile
phase pH in a range of +/- 1.5 pH units around the pKa of the
buffer, but a range of +/- 1 pH unit is preferred. If you need
to work at lower buffer concentrations, it is definitely better
to stay within the smaller range. This is due to the fact that
the buffering capacity of a buffer is a function of both the
concentration of the buffer and the difference between the
pH and the pKa of the buffer. Therefore, lower buffer
concentrations correspond to a narrower buffering range.
In summary: there is no reason whatsoever to limit
yourself to acidic mobile phases when running acidic
analytes. A good control of the mobile phase pH by using the
correct buffers at the correct concentration is important for
the reproducibility of your method and the peak shape of
your analytes. And dont forget: A buffer is a buffer if it
buffers the pH. If it doesnt, it isnt.
Reference:
1. M. Zoubair El Fallah, HPLC Methods Development, in
Uwe D. Neue, HPLC Columns - Theory, Technology and
Practice, Wiley-VCH (1997)

30. Mobile Phase Composition

Q.: We recently converted an isocratic method from an older
two-pump gradient system to a single pump HPLC system.
We were surprised to find significant differences in the
retention times between both systems. The differences are
consistent and reproducible. What is the reason?
A.: The most likely reason for your observations is the
contraction (or expansion) of the solvents upon mixing. This
is an old issue in the preparation of reversed-phase mobile
phases. It is due to the significant volumetric contraction of
the commonly used mixtures of water or buffer with
methanol, acetonitrile and tetrahydrofuran. I rarely get any
questions about this subject today, possibly due to the fact

37

that the HPLC systems in use today commonly use similar

principles for solvent mixing.
Q.: Our method is indeed a reversed-phase method. We use a
mobile phase of 50% methanol and 50% water.
A.: Methanol-water mixtures represent the worst case of
solvent contraction in HPLC. You can observe significant
differences in retention, depending on how you mix the
solvents. You need to understand that if you mix 500 mL of
water with 500 mL of methanol or if you take 500 mL of one
solvent and fill the graduate with the other solvent you do not
obtain the same composition. Since you do not get the same
composition, the elution strength of both mixtures is
different. Since the elution strength is not the same, you
observe different retention times. Examples of this effect
have been published in the literature (1), so let me use the
literature example to demonstrate the effect.
The separation used to demonstrate the mixing effect is a
separation of explosives dissolved in acetone. It was carried
out using a 4.6 mm x 250 mm C18 column using a mobile
phase of 40% water and 60% methanol. The details of the
method are given in the legend to the graph that show the
results of the experiment.
Four different ways to prepare the mobile phase of 40%
water and 60% methanol were used. In the first case, 400
mL of water were put into a 1 L volumetric flask, and the
flask was then topped off with methanol. Due to the
contraction of the mixture during the combination of the tow
solvents, more than 600 mL of methanol are added to the
flask. Consequently, the elution strength of this mobile phase
preparation is the highest, and the retention times are shorter
than with the other preparation techniques.
In the second case, 400 mL of methanol and 600 mL of
water were measured out separately and combined in a flask.
The retention times were longer than in the previous case. As
we will discuss below in more detail, this is the most
preferred approach. It is also the way that mobile phases are
prepared in most laboratories.
In the third case, the gradient was formed by two pumps
delivering the correct quantities of either methanol or water.
The mixing of the gradient was done at high pressure. In this
case, the solvent composition is identical to the solvent
composition of the second case. However, the contraction of
the solvent mixture occurs behind the pump, resulting in an
actual flow rate of less than 1 mL/min. Therefore, the slower
flow rate results in longer retention times, while the solvent
composition is accurate.
The fourth case is the inverse of the first case: 600 mL of
methanol were added to the volumetric flask, which was then
topped off with water. Due to the contraction of the mixture,
more than 400 mL of water were needed to fill the flask.
Consequently, this mobile phase contains the highest water
concentration, and the retention times are longer than in the
other cases.

the difference is the flow rate delivered by the system. Is that

correct?
A.: Yes, it is. The mobile phase composition is identical in
these two cases. Therefore, the selectivity of the separation is
not influenced at all. The only difference between the second
and the third method is the actual flow rate in the HPLC
system. Therefore, the retention factors and the selectivity of
the separation are absolutely identical. From this standpoint,
the transfer of the method from one system to the other is
successful. The minor differences in the actual flow rate
between both systems is nothing to worry about. You have
successfully transferred the method from your two-pump
system to your single pump system.
Reference:
1. Veronika R. Meyer, Pitfalls and Errors of HPLC in
Pictures, 1997, Hthig, Heidelberg - Oxford, CT, page 51

31. Column Contamination

Q.: My column lifetime is not very long. After only about
500 injections, the peaks broaden and begin to tail.
Previously, the column lifetime was approximately 1000
injections. What is wrong with the column?

Figure 1: Separation of explosives using different methods

of mobile phase preparation.
Sample: explosives dissolved in acetone (octogen, hexogen,
tetryl, trinitrotoluene, nitropenta); column: Grom-Sil 80
ODS-7 PH, 4 m, 4.6 mm x 250 mm; detection: UV at 220
nm; mobile phase: 40% water and 60% methanol; flow
rate: 1 mL/min; (reprinted from reference 1 with permission
of the publisher and the author)
The second approach is the most preferred way of mobile
phase preparation. On one hand, it is the most similar to the
commonly used low-pressure gradient mixing systems. On
the other hand, it does not require a volumetric measurement
after mixing. Therefore, this approach avoids problems
which arise from the heating or cooling of the solvent
mixture upon mixing. It represents the practice that is
commonly recommended in HPLC training courses. I would
therefore believe that this is the method that you used as
well.
Q.: This is indeed the case. From your description of the
problem, the differences that we observe in the retention
times on our different instruments are related to the different
ways of preparing the mobile phases due to the fact that we
are comparing a two-pump system to a single-pump system.
This is identical to the second and third case in your
examples. If I understand you correctly, the composition of
the mobile phase is identical in case two and case three, and

A.: It is quite possible that absolutely nothing is wrong with

the column. Before we discuss this in detail, let me ask you a
question. Are you using a guard column?
Q.: No, I am not using a guard column. But I use a
precolumn filter. This should protect the column, shouldnt
it?
A.: Unfortunately, precolumn filters are only a partial
protection for the column. They serve to remove particulates
from the mobile phase stream. Particulate derive either from
your sample or from the moving parts of the HPLC
instrument. Precolumn filters can not protect your column
from material that is smaller than the filter and may adsorb
on the surface of the packing. Most of the time, but not all of
the time, the origin of such material is the sample. What is
your sample?
Q.: It is a plasma extract from a solid-phase extraction
procedure. Admittedly, plasma components can reduce a
columns lifetime fairly rapidly, but the chromatogram is
fairly clean, with few background interferences. More
importantly, the column lasted previously for over 1000
injections, and only now do we observe the more rapid
deterioration.
A.: In many cases, the sample preparation procedure is not
entirely perfect. After all, some background interferences are
still present in the chromatogram. Similarly, some protein

38

www.pharmaguideline.com

may be left in your sample despite your best efforts. Proteins

are often strongly adsorbed on the packing material and can
slowly accumulate on the top of the column. In addition, the
concentration of these background contaminants may vary
slightly with the samples. Therefore, a decrease in the
column lifetime by a factor of two does not shock me. My
recommendation would be to use a guard column to protect
your analytical column from this contamination.
Q.: I have tried guard columns before. The results were not
good. The peaks were distorted right from the beginning.
Therefore, I dont think that employing a guard column is a
good solution.
A.: A lot depends on your choice of a guard column. I
generally recommend to use a guard column that is packed
with the identical packing material as the analytical column. If
you do this, the guard column works as an extension of
your analytical column and retains the contaminants that
would normally adsorb to your main column. If this guard
column is reasonably well packed, the deterioration of the
separation performance can be minimized. In some cases,
you may actually gain a little bit in overall performance.
However, this is not exciting; the important part is the
protection of your analytical column.
If you choose a guard column that does not contain the
same packing as your analytical column, you can get
significant peak distortion. In addition, you can not predict
how well this guard column is protecting your analytical
column. Both of these effects are due to the fact that the
retention properties of different packing materials are
different. Therefore, it is always better to match the packing
material in the guard column to the packing material in the
analytical column. This allows you both the best possible
column protection and the least amount of peak distortion.
Q.: The problem with guard columns is that they are very
expensive. How about column clean-up and washing
procedures?
A.: I always look at clean-up procedure as a last resort, if I
dont have any other options. The problem with clean-up
procedures is that they work best, if you know what is
contaminating the column. In most cases, we do not have a
clue, and can only make some guesses as to the nature of the
contamination. Additionally, in some cases it may be very
difficult to remove the contaminants. Your problem is most
likely an example for this case. I suspect that the
contamination consists mostly of small amounts of proteins.
To find a good column clean-up protocol is not easy. There
are a few generalized cleaning protocols available in the
manufacturers literature, but they are quite involved and
may or may not help you. In all of these protocols, you need
to wash the column with a series of solvents that are meant to
remove the contaminants. This may take a significant amount
of time, and you may or may not be successful.

39

Therefore, I favor the use of guard columns. They are

guaranteed to capture the contaminants that would otherwise
foul up your column. They are not that expensive; the cost is
only a fraction of the analytical column that you are trying to
protect. It takes only a few minutes to replace a guard
column, while a thorough column washing protocol including
reequilibration of the column may take hours. Your savings
in time and aggravation are significant. If you adhere to the
principles that we discussed above, they are nearly
guaranteed to work and protect you from many headaches.

32. Method Verification

Q.: A colleague recently passed a new QC method on to me.
I bought a new column to verify his method. Unfortunately,
the retention times obtained in my lab were different than the
retention times that he had reported. He sent me his old
column, and indeed I was able to obtain with his column the
separation that he had reported. I then changed a few mobile
phase parameters and was able to obtain the separation on
my column too. I suspect that I am looking at batch-to-batch
differences from the manufacturer, since my colleagues
column is over one year old. How can I deal with such a
situation in a QC department? I cannot write into the QC
procedure that the method should be reoptimized every time
we buy a new column!
A.: Indeed, you should not have to do that. There are
possible solutions to the problem, but before I discuss them
with you I would like to verify an important point: are the
two columns that you have mentioned the only columns on
which the method has been run?
Q.: Yes, this is the case. My colleague developed the method
on his column, and the only other column on which the
method has been tried is my new column.
A.: In such a case, the very first thing to do is to verify that
you are indeed looking at batch-to-batch differences. It is not
impossible that the differences that you are observing are due
to a comparison between an old column and a new column,
and not due to batch-to-batch differences. After all, I would
suspect that your colleague needed to try a lot of different
mobile phases before he established the method. During the
method development process his column might have aged,
maybe lost some bonded phase or strongly adsorbed a
mobile phase additive that was used during the development
of the method but was not needed for the final method. Since
events like this can happen inadvertently, it is always a good
idea to check a new method with a brand-new column from
the same batch and verify that consistent results are obtained
on both columns.
I suggest the first thing to do is to contact the manufacturer
and ask him to pack you a new column from the old batch of
packing material. Most manufacturers maintain a small

amount of material from every batch of packing for exactly

this purpose. Since your colleagues column is only about a
year old, you should be able to get a fresh column from the
old batch of packing material. Depending on the
manufacturer, it is even possible that the batch of packing
has not changed over this period of time. This would be the
simplest case, since this would prove immediately that the
differences observed are not due to batch-to-batch
differences.
Run the assay on the new column packed with the old
batch, using your colleagues mobile phase. If you get the
same results as your colleague obtained, then you are indeed
looking at batch-to-batch differences, and you have your
work cut out to create a rugged method. If you get the same
or similar results that you obtained on your column, then the
issue is that your colleagues column had aged during
method development. I would then recommend that you
design a column lifetime study with your new mobile phase
conditions to establish how long the column is expected to
last when treated only with the mobile phase used in this
assay.

purpose. They consist of columns from different batches.

Make sure that at least one of the columns in the set is from
the same batch as the column that you used to establish the
method. This way, you can check the column-to-column
reproducibility of your method as well as the batch-to-batch
reproducibility. This is something that I generally
recommend to everyone developing a new method that is
supposed to be reproducible over an extended time period.
In all of the discussion above I have assumed that the type
of column that you are employing is a mainstream column,
such as a C18 or C8 column. If the original method was
developed using a cyano or amino column, it is highly likely
that the reproducibility problem is due to the limitations of
the stability of the bonded phase. If this is the case, my
recommendation would be to see if the method can be
redeveloped using a C18 or C8 column. Of course, method
redevelopment is a significant amount of work, but it may be
the best solution in this case.

Q.: This is a reasonable suggestion, and Ill see that I can get a
new column from the same batch that my colleague has
used. But what do I do if the problem turns out to be due to a
batch-to-batch difference?

Q.: I am working with a column designed for sugar

separations. It is a specially prepared amino column. The
mobile phase is a mixture of acetonitrile and water. For the
standard separation we use a mobile phase of 75%
acetonitrile and 25% water. Recently, I am getting double
peaks for all compounds. I thought the column is voiding,
but a new column gives the same results. What is the
problem?

A.: This, indeed, would be the worst case. First, I would try
to make a judgement as to how long the assay will be
needed. Sometimes, only a limited number of columns is
required, and the project will discontinue at some point in the
near future. Under these circumstances, the simplest thing to
do would be to ask the manufacturer to reserve sufficient
packing material from this batch for your project and ask
them to prepare columns for you upon request. Sometimes,
the manufacturer may ask you for an up-front payment for
the amount of packing that they put aside for you. However,
if you are in a situation where the new test will be used for an
indefinite amount of time into the far future, you need to
carefully think through what can be done to make the method
rugged.
Having demonstrated to yourself that the packing material
that your colleague has chosen does not give you
reproducible results, it might be worthwhile to redevelop the
method on a different packing. This is a significant amount
of work, but it might very well be worth your while. There
are significant differences in the reproducibility of different
packings. You may discuss the problem with the
manufacturer and see if he can demonstrate to you that
another packing could give significant improvements in
reproducibility. Or you may consider a column with a good
reputation of reproducibility from another manufacturer. In
all of these cases, the development of the method starts again
at the bottom. You should inquire if the manufacturer can
make several batches of the packing available to you, so you
can verify the ruggedness of the method for yourself. Some
manufacturers offer column sets specially prepared for this

33. Double Peaks in Sugar Separations

A.: Unfortunately, the principle behind this separation is

more complex than the principle of other separations. We
need to discuss it in detail to get to the root of the problem.
There is a simple way in which you can rapidly check
whether the columns have voided or if something else is
going on. Use glycerol as your test compound! It will give
lower retention than most of your sugars, but that is O.K. If
you get the same double peak from glycerol as from the other
carbohydrates in your chromatogram, then the column has
voided. If you get a single, undistorted peak from glycerol,
while you get double peaks from the other carbohydrates,
then the column is O.K., and the problem is not with the
column, but with the mobile phase.
Q.: Glycerol is one of the compounds in my sample mixture. It
elutes early in the chromatogram. I do not see any double
peak for glycerol. It actually gives a very nice sharp peak,
sharper than the other peaks in the chromatogram.
A.: This demonstrates that the column is in good shape.
Therefore, we have to look somewhere else for the
explanation for the double peaks of your other analytes.
Sugars exist in two anomeric forms. In solution, both
anomeric forms are in equilibrium with each other. These
anomers are two distinct molecules that can be separated
without difficulty by HPLC. Therefore, one should get two

40

www.pharmaguideline.com

peaks for every sugar, and indeed, this is what you are
getting.
Q.: But under normal circumstances, I am getting only one
peak for every sugar. Only recently did I get these double
peaks.
A.: Yes, amino columns usually give only single peaks for
every sugar. The reason for this is simply the pH in the pores
of the packing. The rate of the conversion from one anomeric
form into the other is a function of the pH. At alkaline pH,
the conversion is very rapid, and one gets only a single,
reasonably sharp peak. At acidic pH, the conversion is much
slower, and one can get two distinct peaks for every
carbohydrate. This is the prime reason for the use of amino
columns for this separation. The amino groups create an
alkaline environment in the pores, which speeds up the
interconversion, and one gets a single peak for all
carbohydrates. If for whatever reason the pH in the pores of
the packing is neutral or acidic, you will get broad peaks or
double peaks or even two very distinct and well separated
peaks for every sugar.
Now that we understand that this is the most likely cause
of your problem, we need to determine why there is a pH
shift. Commonly, the reason for this behavior is the aging of
the stationary phase; as the column is used, it looses some of
the basic bonded phase and forms more acidic silanols.
However, your description indicates that you observe the
same behavior on a new column. This would speak for a shift
in the pH of the mobile phase. Since the mobile phase
contains only two components, acetonitrile and water, you
could simply check the pH of the water alone and then of the
water/acetonitrile mixture. Hopefully, this will pinpoint the
source of the problem.
The next step is to regenerate the columns. Since most
likely the columns have been neutralized by an acidic
ingredient in the mobile phase, it should be possible to
remove this ingredient by a wash of the column at an alkaline
pH. The best approach is probably to wash the column with a
dilute solution of an ethylenediamine oligomer, such as
tetraethylene pentamine. This wash can regenerate the
column, but avoiding the acidic contamination to begin with
is still the best solution.

34. Method Control

Q.: I am working on a new HPLC procedure that will be
used in our QC department. I have worked out the details of
the procedure; I know the linearity, the detection limits, and
the influence of mobile phase pH and organic concentration
on retention and selectivity. I am now discussing with my
colleagues, how to set the specifications for the method: what
should be the precision with which the mobile phase is made
up, what variations of the mobile phase pH are tolerable etc.
Can you help us with some advice?

41

A.: It is good to know that you have done your homework.

This will help significantly in setting up the necessary
controls over the mobile phase composition and other
parameters. The primary question that you have to answer is
at what point of variation of every parameter the method will
start to fail. Sometimes this requires a very tight control over
the method parameters, sometimes the specification of the
method can be varied widely without a negative effect on the
quality of the results. Unfortunately, it is not possible to give
you good general guidelines. You have to study all the
relevant process parameters of your method. Once you
understand the influence of the method parameters on the
goals of the assay, you can set the limiting values for every
parameter.
Let me give you some examples: a complex case that
commonly requires fairly tight control of the method
parameters is an impurity profile of a pharmaceutical
compound. Frequently, many impurities are possible; often
they are closely related to the parent drug and some peak
pairs are difficult to separate. Under these circumstances, a
tight control of the method parameters is required. An
entirely different situation is given for a dissolution test or a
compositional analysis. Under these circumstances, the
interest is simply to quantitate the parent compound, which
needs to be separated from the internal standard. Such an
analysis is not very demanding, and one can often allow for
significant variations in mobile phase composition or
temperature without affecting the results. In such a case, the
best approach is to specify method control parameters that
are readily achievable by an average technician and leave it
at that.
However, if your analysis is complex and requires tight
controls, you should have been able to demonstrate the limits
in the preliminary experiments. How much of a variation of
the pH can you tolerate before the resolution between peaks
becomes inadequate? How tightly do you need to control the
organic content of the mobile phase? There are some rules of
thumb that can tell you, how much variation in retention time
one can expect if a given parameter is varied, but the best
approach is the experiment. Since you have studied several
of the important method parameters and know their effects,
you should be able to set reasonable specifications. Only
your experiments can provide the answer to the question of
the tolerances of the method guidelines.
Yet it appears that you have studied only some of the
relevant parameters. While some of the variables that you
have not mentioned may have only a smaller effect on the
method, it is nevertheless important to know their influence.
The influence of temperature is often small, but it is worth
knowing. Another commonly small effect is the buffer
concentration. Nevertheless, both should be included in the
method validation studies.
Another parameter that is often overlooked and can cause
problems in the future is the column variability. I always
recommend that columns are purchased from several batches
of the packing material when a new method is established.
Many manufacturers provide such a service. Of course, this

costs some money, since a few columns need to be

purchased, but this is money well spent. An early check of
the batch-to-batch variability of a packing can save you lots
of headaches later in the process. If I were your QC manager,
I would require these data from you before I would accept
your method.
Another thing worth incorporating in a method is a
troubleshooting procedure. Since you know all or most of the
variables affecting the method, it is good to write them down
in a way that allows a future investigator to find a problem
rapidly. Such a procedure should also specify the criteria for
replacing the column with a new one. This can be a
specification of the selectivity of the separation, or criteria on
the peak shape, or both. Changes in peak shape or tailing
make a quantitation more difficult and are most often used as
criteria for the end of the useful life of a column.

35. Mobile Phase pH

Q.: Recently, we have had discussions in my group about the
adjustment of the pH of the mobile phase. The simplest and
most reliable way to do this appears to be in the final mobile
phase, after the organic solvent has been added. Others have
advocated to measure the pH always in the pure aqueous
component of the mobile phase. When comparing the two
approaches, we found differences in the final pH as well as in
the chromatography. What is the reason for this, and what is
the best way of adjusting the pH?
A.: This is a very good question, because there is often
confusion between different labs using different practices for
mobile phase preparation. Let me first describe some of the
problems that one encounters. This will then lead us to the
best solution.
The adjustment of the pH is accomplished using a buffer
solution. The capability of a buffer to maintain and control
the pH is called the buffer capacity. The buffer capacity is a
function of the difference between the pH and the pKa of the
buffer and of the concentration of the buffer. For example,
acetic acid has a pKa of 4.75 (in water), therefore acetate
buffers have the best buffer capacity at this pH. Typical
buffer concentrations used in HPLC vary from 20 to 50 mM,
therefore good buffering capacities are achieved within +/- 1
pH units around the pKa of a buffer. To achieve reasonable
buffering capacities, the pH of the buffer should never depart
more than 1.5 pH units from the pKa of the buffer.
If an organic solvent is added to the buffer, the measured
pH shifts to higher values. This is due to two separate effects.
One is a true shift in the pH due to the shift of the
concentration of the hydrogen ion in the presence of the
organic solvent. The second effect is related to the pH
measurement itself: since the pH meter has been calibrated
and adjusted to give accurate results in water, it will exhibit a
departure from the true value if the measurement is made in
the presence of a large concentration of an organic solvent.
However, both effects are irrelevant for the actual

composition of the buffer and its buffering capacity: a buffer

solution prepared from a given ratio of the acidic form and
the basic form of the buffer components will still have the
identical ratio of both in the presence of an organic solvent
and therefore the identical buffer capacity as in pure water.
Therefore, if we measure the buffer composition and its pH
in water, we have a fixed constant reference point, and
unquestionable tools to measure the pH value. For example,
the acetate buffer has its maximum buffer capacity at pH
4.75 in water. An acetate buffer whose pH is adjusted to this
value in water will be at its maximum buffering capacity
independent of the pH value that is read by the pH meter in
the presence of an organic solvent. For all commonly used
buffers, I can look up the pKa values in water in a table.
Then, I can make the adjustment of the buffer pH in water,
using the pKa as the reference point. Then, as the last step, I
add the organic solvent.
However, there are occasional unavoidable exceptions to
this rule. If you use an older procedure developed using an
older column, it may specify a long-chain amine as a
component of the buffer. Sometimes, the solubility of the
amine in neat water is limited, and the simplest way to
prepare the buffer solution is by adding the organic solvent
prior to the adjustment of the pH. There is nothing wrong
with this procedure, as long as you know the titration curve
and the buffering capacity of your buffer in the presence of
the organic solvent. The best way of getting this information
is to obtain a titration curve in the presence of the targeted
concentration of the organic solvent and to construct the
buffer capacity curve from this titration curve. A titration
curve is a plot of the pH of the solution as a function of the
amount of base added to the solution. The buffer capacity
curve is a plot of the (negative) change in pH per amount of
base added versus the pH. This curve has a maximum at the
pKa of the buffer in the presence of the organic solvent.
Since the buffer capacity is at its maximum at the pKa of the
buffer, you can then use this information to prepare a buffer
for your assay with an optimal capacity.
The preferred practice in the industry is to prepare the
buffer in water and then to mix the pH adjusted buffer with
the organic solvent. This gives unequivocal results and
allows for a good and universal judgement on the capacity of
the buffer. It should be noted that this can be indicated
clearly in the statement of the mobile phase composition. To
avoid confusion, the composition should be stated as the
aqueous component including the buffer and its pH mixed
with the organic solvent. Here is an example:
50 mM
KH2PO4/K2HPO4 buffer, pH 7.0 / methanol 40/60, or 40%
50 mM KH2PO4/K2HPO4 buffer, pH 7.0 and 60% methanol.
The inverse statement is somewhat ambiguous: methanol /
KH2PO4/K2HPO4 buffer, pH 7.0, 60/40 and should not be
used.
Q.: Your proposal is different from what I had originally in
my mind, but it makes sense.
What is a typical shift in pH with solvent composition?

42

www.pharmaguideline.com
A.: It is relevant enough to cause measurable retention
changes if the buffer is prepared before or after the addition
of the organic solvent. In the literature (1), upward shifts of
the pKa of acids of about 1 unit have been reported between
100% water and 50% water/methanol. For the bases
reported, the pKa shifted by approximately 0.5 units between
100% water and 50% water methanol. If the analytes are not
completely ionized or non-ionized, this difference causes
substantial shifts in their elution times. Even changes in the
elution order are possible. Therefore, from a practical
standpoint, the current practice of standardizing the way the
mobile phase pH is specified is extremely important. I am
happy that I managed to convince you to follow the same
practice and to measure and specify the pH in the aqueous
component of your mobile phase.
Reference:
1. E. Bosch, P. Bou, H. Alleman, M. Ross, Anal. Chem. 68
(1996), 3651

36. Signal-to-Noise Improvements

Q.: I would like to improve the sensitivity of my HPLC
method. Would you please give me some advice on what I
can do?
A.: Gladly! There are many different components to the story
(1). Let us first see what we can do to reduce the detector
noise. First of all, we need to know that the detector is in
good working condition. This means, for example, that the
lamp is working at full capacity, and that the cell window of
the detector is clean. The detector performance is best
verified using a set of standard testing conditions.
Q.: I know that the detector is in good working order. We
monitor the detector linearity, its noise and its response on a
regular basis to verify that it works for routine analysis. My
question is more directed towards what one can do with a
fully functioning detector rather than towards detector
troubleshooting.
A.: O.K. The first thing to do is probably to examine the
detector response and the detector noise as a function of the
wavelength. Then select the wavelength that results in the
best signal-to-noise ratio. This varies with the detector, the
mobile phase and the sample that you are trying to analyze.
In principle, the noise of a variable wavelength detector is
inversely proportional to the amount of light the photodiode
receives. This is the reason why the detector noise increases
as the light source ages or the detector window becomes
dirty. It is also the reason for increased noise, as the mobile
phase becomes more adsorptive. For example, alcohols have
a significant absorbance at 210 nm, while the absorbance of
acetonitrile is very low. This is the reason for reduced noise
and consequently higher sensitivity using acetonitrile in the

43

low UV. Around 250 nm both acetonitrile and methanol

work equally well.
If you need to use a mobile phase that has a background at
the wavelengths where your sample has a good response,
both noise and response must be measured as a function of
the wavelength. With this data, you can then calculate which
wavelength is best suited achieving the maximum signal-tonoise ratio. If you have a photodiode array detector, this
exercise is very straightforward and quick. You simply
obtain a spectrum at the peak maximum and compare it to
the mobile phase background spectrum
Then you choose the wavelength that gives you the best
signal-to-noise ratio. Now you have selected the best
wavelength for a given mobile phase composition. If this still
does not give you satisfactory results, you have your work
cut out.
The simplest thing to do next is to examine an increase in
the detector time constant, or the equivalent noise reduction
mechanisms built into modern detectors. If you increase the
time constant or its equivalent, the peaks will become
broader, and the noise is reduced. If your chromatogram is
fairly empty, this approach can improve the signal-to-noise
ratio significantly. However, if there are many peaks in the
chromatogram and the resolution between some of them is
small, this approach has its limitations.
Another relatively simple thing to consider is the amount
of sample that you inject. If you see disturbances in the
separation long before you run out of sample, you should
think about the solvent in which the sample is dissolved (2).
Often, the sample can be dissolved in a solvent composition
that has a much lower elution strength than the mobile phase
(3), and you can inject very large sample volumes before you
see any distortions of the peak shape. If your sample is
dissolved in an organic solvent and your analysis is done by
reversed-phase chromatography (i.e. the mobile phase is a
mixture of water and methanol or acetonitrile), you will be
able to inject only a small volume before the peaks become
misshapen. On the other hand, if you make up the sample in
a solvent composition that contains 20% more water than the
mobile phase, you can inject a milliliter of sample without
appreciable peak distortion. The sample simply enriches on
the top of the column at the beginning of the
chromatography.
Q.: This is a neat trick worth remembering. What other
things should one consider?
A.: Another approach is a reduction in column volume.
However, there are several things necessary to consider
before implementing this option. First of all, you need to
determine, if indeed the amount of sample available to you is
limited, and that an injection of the total amount of sample
on the standard column does not reveal any other limitations.
The simplest approach towards a reduction in column
volume is to reduce the column diameter, keeping the
column length and the particle size constant. For example, a
reduction of the column diameter from 4.6 mm to 2.0 mm

increases the mass sensitivity by about a factor of 5. As a

side benefit, it reduces the flow rate and thus the solvent
consumption by about the same factor. As a tradeoff, the
effects of instrument bandspreading become more
significant. The instrument bandspreading is likely to reduce
the resolution in the chromatogram, but you can address at
least part of it.
There are two components to instrument bandspreading.
The first one is the pre-column bandspreading, caused by the
injection volume, the bandspreading in the injector and in the
tubing leading to the column. You can reduce its influence
by using the trick described above: dissolve the sample in a
solvent with a lower elution strength than the mobile phase
or dilute the sample with such a solvent and inject more. The
second component is more bothersome. It is the post-column
bandspreading, caused by the tubing leading from the
column to the detector and by the detector cell itself. The
length of tubing leading from the column to the detector
should be kept to an absolute minimum. Since you can
address the influence of the pre-column tubing volume by
other means, it is best to keep the post column tubing to an
absolute minimum. The question of the detector cell volume
is more tricky. On one hand, you want to minimize the
volume to reduce bandspreading. On the other hand, a
reduction in the detector cell volume increases the detector
noise. In the worst case, the whole exercise of reducing the
column volume and the detector volume may result in only a
marginal improvement in sensitivity.
Q.: I guess this needs to be thought through carefully. How
else can I improve the signal-to-noise ratio of my assay?
A.: Another relatively simple thing might be to change the
detector. Maybe a fluorescence detector or an
electrochemical detector are a better choice than the standard
UV detector. Of course, the purchase of a new detector is a
significant expense.
There are a few more drastic changes to your assay that
can improve the signal-to-noise ratio. Often, noise is caused
by components of the mobile phase. Earlier I had mentioned
the reduced absorption of acetonitrile compared to methanol
in the low UV. Such a change requires the redevelopment of
the separation. Another frequent cause of excessive detector
noise are additives in the mobile phase, for example amines
used to suppress peak tailing. You can eliminate the need for
such mobile phase additives by using a better column. In a
similar vein, if the low sensitivity is caused by peak tailing, a
reduction of the tailing by using a better column can increase
the sensitivity by a factor of about three. All of these
suggestions represent a significant amount of work, and the
improvements that you are getting are comparatively small. It
is best to make these choices at the beginning of the
development of a new assay.
References:
1. J. B. Li, Signal-to-Noise Optimization in HPLC UV
Detection, LC-GC 10 (1992), 11, pages 856-864

2. U. D. Neue, HPLC Columns, Theory, Technology and

Practice, Wiley-VCH (1997)
3. U. D. Neue, E. Serowik, Sample Dilution Increases
Sensitivity and Resolution, Waters Column VI, 2 (1996),
pages 8-11

37. Overload
Q.: When I increase the amount of sample that I am injecting
the peaks start to tail and become broader. I am working with
fairly low concentrations and, therefore, would not expect
that the column is overloaded. I would appreciate if you
would discuss the phenomena that could cause such an event.
A.: Mass overload of the column is only one of the
phenomena that can cause a peak distortion upon increasing
the injection size. It is also possible that the mobile phase is
overloaded. Something like this can occur, if the sample is
not dissolved in the mobile phase. However, let me discuss
mass overload of the column first.
Mass overload of the column can be diagnosed very
rapidly. For common analytes, mass overload starts at an
injection of about 0.1 mg to 1 mg of sample per milliliter of
column volume. If you are using a UV detector, you can use
another rule. For most analytes possessing a strong UV
chromophore, mass overload starts if the peak height reaches
about 0.1 AUFS. Of course, these rules are rough rules of
thumb, but if you are injecting significantly less sample,
mass overload is not likely to be the problem.
Q.: Indeed, the amount of sample injected is much smaller.
Therefore, I do not think that mass overload is the cause of
the peak distortion. What else can be the problem?
A.: A common issue is the solvent in which the sample is
dissolved. If the sample diluent is a stronger eluent than the
mobile phase, peak distortion can occur. There can be
several reasons for this. It may be that the sample originated
in a solid phase extraction procedure that required a stronger
solvent than the mobile phase for the elution of the sample.
Under these circumstances, a simple dilution of the sample in
the weaker eluent may avoid the problem. Alternatively, the
sample can be evaporated to dryness and then reconstituted
in mobile phase or a weaker eluent. A similar problem is
caused by samples generated from dissolution tests. The acid
in the sample may overload the buffer of your mobile phase
and cause peak distortion. This can be fixed by either
adjusting the pH of the sample or using a stronger buffer in
the mobile phase. To prepare a stronger buffer, increase its
concentration and adjust the pH to a value close to the pKa
of the buffer. This requires a little bit of homework, but is
relatively straightforward. In most dissolution tests, the
number of analytes in a sample is very small, often just the
compound of interest plus the internal standard.

44

www.pharmaguideline.com

If the sample pH is the cause of the peak distortion, a

simple increase in the buffer concentration is often the
solution, even if the chromatogram is fairly crowded.
Typically, changes in retention are very small for even
significant changes in buffer concentration. Of course, you
should check at what concentration the buffer starts to
precipitate.
Another reason for peak distortion problems can be low
sample solubility in the mobile phase. The effects are similar
to the case where the sample is dissolved in a different
solvent. After all, the sample was dissolved in a different
solvent than the mobile phase to get around the problem of
low solubility. Fortunately, cases where this is the true
problem are very rare. In such a case, the best approach may
be to live with the problem, which means that the
concentration range over which the sample concentration can
be determined is rather narrow. Alternatively, the choice of a
completely different chromatographic mode may offer a
solution.
Q.: None of these issues seem to apply. What is left?
A.: One of the items that we have not yet discussed is a
simple sample volume overload. If the sample is dissolved in
mobile phase, and a large volume is injected, the peaks are
becoming broader. This phenomenon has more of an effect
on early eluting peaks, and becomes less of a problem for
late eluting peaks. One can use this as a criterion to
distinguish this problem from the other items that we have
discussed. The solution is fairly simple: dissolve the sample
in a weaker eluting solvent than the mobile phase. You can
check very rapidly, if this is the problem. Dilute your sample
2x with the weaker eluent in your mobile phase, and then
inject 2 times the sample volume. This will concentrate the
sample at the column top, and the total amount of sample
remains the same. If the peak distortion disappears, a simple
volume overload is the problem. Simply dissolve your
sample in the future in a solvent composition that has a lower
elution strength, and the problem will disappear. Thinking
carefully through what the impact of the sample solvent on
the analytical results might be can solve many problems.

38. Column Durability

Q.: What can I do to make an HPLC column last longer?
A.: There are several options available to you to increase
column lifetime. An acceptable column lifetime is
approximately 1,000 injections. However, I have seen
columns last for over 10,000 injections without any
performance deterioration. The only requirement was a
reasonable column care.
Q.: 10,000 injections? I have never achieved that many
injections on a column. I find this difficult to believe.

45

A.: Today, columns prepared by the leading manufacturers

are very rugged. In general, the column performance
deterioration of is due to column contamination. If this is
prevented, good column lifetimes will result. My coworkers
and I have investigated column lifetime issues in detail. Let
me explain some of our experiments, and then we can discuss
other issues that affect column life.
We examined column lifetime by running the columns
practically continuously using a single isocratic assay. The
columns were Symmetry C18 columns from Waters
Corporation. The sample was a mixture of pharmaceutical
standards, and the mobile phase contained 79 % water, 20%
methanol, and 1% acetic acid. We performed two sets of
experiments. In one of them, the sample was injected directly
onto the column. In the second set, the columns were
protected by a guard column.
The experiments that were performed without use of a
guard column resulted in an irreversible deterioration of
column performance somewhere around 1,500 to 2,000
injections. While the retention times remained constant, the
plate count dropped shortly before the end of the column life,
the backpressure increased and, finally, double peaks were
observed. Even with a clean sample, column life without
column protection appears to be limited.
We then repeated the same experiments with a guard
column in place. We used high-performance guard columns
prepared from the same packing material as the analytical
columns. Under these circumstances, the guard columns do
not alter the separation and function as just an extension of
the analytical column. Thus the analytical separation is not
affected.
We found soon that the combination of the analytical
column and the guard column exhibited a deterioration of the
separation after somewhere around 800 injections at the
earliest. When we replaced the guard column, the separation
was restored to the same performance as we had observed
initially. To avoid deterioration of column performance, we
decided to regularly replace the guard column at a
predetermined number of injections that was lower than the
observed failure rate. When we did this, we did not
experience any deterioration of the separation until over
10,000 injections. Then we stopped the experiment. The
column was still in excellent condition. Neither an
appreciable shift in retention nor a substantial change in plate
count were observed. We also did not find a difference in
performance, when we ran the same test using columns
packed with different particle sizes: 3.5 m packings
performed equally well to 5 m packings in these
experiments.
This set of experiments clearly demonstrates that the
durability of well-packed columns is quite excellent. The
limitations in column lifetime that we all experience are not
due to the column, but are caused by other factors. Some of
these factors are extraneous materials from the sample, the
injector, pump seals, etc., that normally accumulate on the
top of the column and cause an irreversible deterioration of

the column performance. Many people believe that filters

will prevent this mechanical deterioration of the columns.
We used the standard filters in our HPLC system, and they
simply did not prevent column deterioration. Only the
sacrifice of the guard columns preserved the performance of
the analytical column. The reason for this is simply that
whatever is causing the deterioration of the analytical column
will be retained on the guard column, independent of the
nature of the cause.
We also performed similar experiments with spiked serum
samples. In this case, we did a simple protein precipitation
for sample cleanup. As in the experiment described above, a
regular replacement of the guard column preserved the
analytical column. In this case, the guard column needed to
be replaced more frequently due to the nature of the sample
matrix and the simple sample preparation procedure.
Unfortunately, it is not possible to give general advice as
to when a guard column should be replaced. It largely
depends on the sample and on the sample clean-up procedure
that is used. For samples with a low amount of extraneous
impurities, a guard column may be a reasonable substitute for
a sample preparation and cleanup procedure. Conversely,
preparation procedures for plasma or milk samples still leave
a sufficient amount of endogenous material behind that the
use of a guard column is absolutely necessary. However, you
may be able to reduce the complexity of the sample
preparation procedure knowing that the column is protected
by a guard column. If the life of the guard column is less than
100 injections, I would consider incorporating a sample
cleanup procedure or revising the sample preparation
process. If the guard column protects your analytical column
for 100 injections or more, the cost is under 50 cents per
analysis. In the long run, it is cheaper to replace the guard
column then to continue to buy new analytical columns.
All of the previous discussion assumed that the column is
used for a single assay or in a single procedure. Under these
circumstances, the column contamination is fairly
predictable. If a lab uses many different HPLC procedures, I
always recommend to dedicate a single column, including the
guard column, to each procedure. This eliminates cross
contamination of the column from one assay to another and
makes the column life predictable and controllable. If your
lab does not have any standard assays, this is a mute point.
Even under these circumstances, I would keep the guard
column and the analytical column together until a
deterioration of the performance requires a replacement of
the guard column.
Q.: This is good advice, but I am sure that the analytical
column does not last forever even if it is protected by a guard
column.
A.: This is indeed correct. There are other elements that limit
column life, even when a guard column is used. Reversedphase packings are very stable between pH 2 and 8 at room
temperature. However, if you increase the temperature,
especially when you are working close to the pH limits of the

packing, column life may deteriorate rapidly even when a

guard column is used. From a column stability standpoint,
the best procedures use an intermediate pH. Also, the choice
of the buffer ions plays a significant role in column
degradation (1). The commonly used phosphate buffers are
not the best choices. Significantly better results have been
obtained with organic based buffers like TRIS or citrate
buffers at pH 7-8. Also, the buffer concentration plays a role:
a lower buffer concentration results in improved column life.
This needs to be balanced against the ruggedness of the
assay, which is improved at higher buffer concentrations.
Therefore, even under more stressful conditions, good
column lifetimes can be achieved.
Reference:
1. H. A. Claessens, M. A. van Straten and J. J. Kirkland, J.
Chromatogr. A, 728 (1996), 259

39. Fast Analysis and Column Backpressure

Q.: I am currently trying out a new column. It is a new 3 m
packing. It is supposed to give higher resolution and
therefore faster analysis than 5 m packings. However, I
cant see how this can be faster, since the backpressure is
higher on the 3 m packing, and I actually cant run it at the
same high flow rates as I used to run my 5 m column. Isnt
it wrong to think of smaller particle size columns to be
faster?
A.: Hmm This is more complicated. It all depends how the
comparison is made. We need to discuss this step-by-step to
see under what circumstances each advantages can best be
achieved. A lot of this may seem quite counterintuitive, and
we need to look at the details more closely.
Let us start with the simplest case: same column length
packed with different particle sizes. I believe that this is the
comparison that you are referring to anyway.
Q.: Yes, indeed. Same column length and column diameter,
and different particle sizes.
A.: Under these circumstances, you will get a higher plate
count, i.e. more resolution, for the 3 m packing in most
situations. However, you will also get a higher backpressure.
If you have been driving your analysis time to the fastest
possible with a 5 m column, and if the limitation was the
column backpressure, than it makes no sense whatsoever to
use a 3 m column of the same length. The higher
backpressure of the 3 m column will slow down the
analysis.
What you really need to do to get the speed advantage of
the smaller particle size, is to use a shorter column with the 3
m packing. The best thing to do is to reduce the particle
size and the column length in the same proportion. If you

46

www.pharmaguideline.com

have used a 5 cm long column with the 5 m packing, you

want to use a 3 cm long column with the 3 m packing.
If you now use the same flow rate on both columns, the
backpressure on the 3 m column will still be higher, but
also, since you are using a shorter column, the run time will
be shorter. In addition, if you reduce the flow rate in
proportion to the reduction in column length, your
backpressure on the 5 cm 5 m column and on the 3 cm 3
m column will be about the same. If you do this, then your
analysis time will also be about the same.

long column, your run time is reduced to about 4 minutes,

and the peaks are still well separated. But your backpressure
is only about 350 PSI now. You can take advantage of this
by increasing the flow rate. 3 mL/min will give you about the
same backpressure as you had on the 15 cm column, but now
your run time is under 1.5 minutes. Your peaks will not be as
well separated as they were in the initial assay, but you are
still able to resolve them cleanly at these fast conditions.
Following this logic, the time needed for many simple assays
can be drastically reduced.

Q.: This is a bit complicated. Let me repeat! I have started

with a 5 cm column packed with 5 m particles at a flow rate
of 1 mL/min. What you recommend is to use a 3 cm column
packed with 3 m particles at a flow rate of 3/5 of the
previous flow rate, i.e., at a flow rate of 0.6 mL/min. Then
the analysis time and the backpressure of the 3 m column
and the 5 m column are the same. Then why would I
consider using the 3 m column, if there is no benefit
whatsoever?
A.: What you have not yet considered is the fact that under
these conditions the 3 m column will give you a better
performance than the 5 m column. So, if the speed of your
separation was limited by the backpressure of the column,
you will still have a superior separation on the 3 m column
under conditions that give equivalent backpressure and
analysis time.
Let us assume for the moment that you can tolerate a
somewhat higher backpressure and can run at 1 mL/min on
the 3 m column. Then, if the speed of your separation is
limited by the resolution of a pair of peaks, the 3 m column
will allow you to run the separation faster than it was
possible with the 5 m column. The shorter run time allows
you to increase throughput. This is the primary benefit of the
smaller particle size. In todays world, where there is a lot of
pressure on everybody to increase the output of a lab, a gain
in analysis time by 50% or so is a large improvement. For
this reason, more and more people are using columns packed
with smaller particles.
If there is plenty of resolution in your chromatogram, then
the first thing to do is to explore how fast you can go with
your 5 m column, or even better, how short a 5 m column
you can use to do the analysis. In many simple QC
applications, the column length is much larger than needed
for the assay. Often, all that is needed is adequate separation
of the analyte from the internal standard, and there is a giant
gap between these peaks. Under these circumstances, you
should simply explore a shorter 5 m column at higher flow
rates.
For example, an assay for content uniformity is run at 1
mL/min on a 15 cm 5 m column, and the backpressure is
about 1000 PSI. The assay takes about 12 minutes. There are
two peaks in the assay, the compound of interest and the
internal standard, and they are well separated from each
other. The first thing to do is to decrease the column length,
maybe to about 5 cm. Using the same flow rate as on the

47

a. 5 m 0.7 mL/min

10.00

20.00

30.00

b. 3.5 m 1.4 mL/min

10.00
Minutes

20.00

Figure 1: Separation of chlordiazepoxide degradation

products using a 150 mm x 3.9 mm 5 m Symmetry C18
column (a) or a 100 mm x 4.6 mm 3.5 m Symmetry C18
column. The plate count of peak 2 is practically identical for
both chromatograms.
Of course, many application examples do not tolerate such
a radical reduction in run time. In many cases, there are
several peak pairs in the chromatogram that do not tolerate a
significant change in plate count. Under such circumstances,
a switch to a shorter column packed with a smaller particle
size results primarily in an improvement in the resolution
between the peak pairs. Scaling the column and the flow rate
according to the rules mentioned above can get you into a
spot of equal resolution but reduced run time. An example
for such a case is shown in figure 1. A 150 mm x 3.9 mm
column packed with 5 m particles is compared to a 100 mm
4.6 mm column packed with 3.5 m particles for the analysis
of degradation products of chlordiazepoxide. The shorter 3.5
m column was run at a higher flow rate, allowing a
reduction in analysis time by a factor of 2 at equal resolution
of the critical peak pairs. An important consideration in this
comparison is the fact that both columns have the identical

column volume, which makes extra-column effects and

sensitivity independent of the choice of the column.
This is an example of the scaling of a separation to the
same resolving power. In such a case the shorter column with
the smaller particle results in a shorter run time. The
backpressure will be higher, but as long as it is below the
instrument limitations, this is not a problem. In many typical
analyses, the pressure limitation of the instrumentation is still
far away from the actual running conditions of an assay.
As you can see, the best solution to faster analysis depends
on the assay. If there is a large separation between the
different peaks and the chromatogram contains only a few
peaks, the easiest way to reduce the run time is to use a
shorter column with the same particle size. If you have a
complicated chromatogram, the better choice is to reduce the
column length and the particle size at the same time. If you
only reduce the particle size without changing the column
length, you will get a larger plate count at a higher
backpressure, but you will miss out on the real benefits of
smaller particles: shorter analysis time at equal resolution.

40. Column Backflushing

Q.: Column backflushing is a common practice in our lab.
When a column shows a deterioration in peak shape, we
connect it to an HPLC that is not used at the time and flow
through it in the direction opposite to the normal flow
direction. Sometimes, we use different solvents for this
backflushing. While I have seen some improvements, it does
not work reliably; actually, most of the time the
improvements are small, and after about another hundred
injections we throw the column away anyway. What can we
do to improve this procedure?
A.: Actually, I am not an advocate of column backflushing.
Most of the time it does not work, and when indeed it does
work, I think that the improvement can be accomplished in
more rational ways. In addition, I believe that there are more
effective solutions to improve column life. Let us discuss the
issues in detail!
What would be an event that would get us to try a
backflushing of the column? There could be two reasons; one
is an increase in backpressure, and the other one a significant
deterioration of column performance. Both are believed to be
caused by the accumulation of sample ingredients or debris
from the mobile phase or HPLC instrumentation at the head
of the column.
The first question is, where is this debris accumulating? If
it is large enough, it will collect on the external face of the
frit. Under these circumstances, a replacement of the frit
would do the same good as backflushing. A replacement frit
costs maybe $ 10. If you have a replacement frit on hand,
you can substitute the old frit with a new one and you are up
and running again. Then you can check, if the old frit can be
rejuvenated, maybe by putting it into an ultrasonic bath for a

while or by just plainly backflushing the frit. You can do this

using the empty barrel of an old column. This is a fairly rapid
procedure.
However, replacing the frit will only work if the frit is the
problem. In my experience, this is rarely the case. Most of
the time, something has accumulated on the head of the
column. The mere fact that the accumulation occurs indicates
that the material is strongly adsorbed at the column top, in a
relatively narrow band. If there is a simultaneous increase in
backpressure, this indicates that the contaminant has a
reasonably large molecular weight or low solubility and is
clogging the channels between the particles. As the
backpressure increases, this means that a significant force is
applied to the first layers of particles in the column, and that
these layers will rearrange and ultimately cause a
deterioration of the peak shape. To repair this kind of
damage to the packed bed structure is very difficult or even
impossible. This is the reason why column backflushing is a
gamble at best.
Q.: At least one can get a little bit more life out of the
column. Backflushing may not be a very good solution, but it
does help.
A.: Of course it helps, but I would prefer a more reliable and
more permanent solution. Have you ever used guard
columns? They protect the column from all different types of
debris, from the sample or from the instrumention. If your
sample is relatively clean, then a guard column can last for a
thousand injections. The analytical column is often as good
as new afterwards. If your sample is fairly dirty, then a guard
column will last only for maybe 50 injections. You will know
what to expect, since the lifetime of a guard column is about
the same as the lifetime of your analytical column today.
Of course, guard columns are not free. However, the price
that you pay for the guard column is relatively low,
compared to the cost of an analytical column. Often, a holder
is needed, since most guard columns are in the form of a
cartridge column. But this one-time cost is soon recovered by
the longer lifetime that you get from your analytical columns.
What guard column should you use? The entire discussion
assumed that you use the same packing in your guard column
as in your analytical column. Most of the time, this is the best
solution, for several reasons. The most important one is that
the addition of the guard column does not effect your
separation to any appreciable degree if you use the identical
packing as in the main column. At the same time, this
approach is generally also the best way to protect the
analytical column. The use of different packings from
different manufacturers is generally not recommended. The
bonded phases from different manufacturers are quite
different, and a guard column made from a different packing
can not guarantee as good a protection as a guard column
made from the same packing. In addition, the differences
between the packings may cause some additional
bandspreading of your analytes. As you can see, there are

48

www.pharmaguideline.com

several good reasons to keep the packing in the guard

column and in the analytical column identical.
There are a few rare occasions, where the contaminant that
is causing reduced column life is known and the use of a
different guard column would capture this contaminant more
effectively. You need to think this through quite carefully.
An example might be the case where the samples, which are
analyzed using a C18 packing, are contaminated with a
polyamine. This polyamine can be adsorbed much better on a
cation exchanger than on a C18 guard column. If you are
using a different packing in the guard column, you need to
make sure that it does not interfere with your analysis. In the
example given here, all the analytes of interest are acidic or
neutral compounds. Therefore, the addition of a cation
exchanger to capture the polyamine is quite feasible. If the
analytes of interest would contain basic compounds, the use
of a cation exchanger as a precolumn is more problematic
and requires a thorough investigation. However, cases as the
one described here are fairly rare, and most of the time the
best guard column contains the identical packing as the
analytical column.
Q.: Many different guard column designs are commercially
available. Which one should I use?
A.: The important thing is to use a guard column that
contains the identical packing as your analytical column.
Therefore, you should talk to the supplier of your analytical
column, as to which guard column he recommends.
Sometimes, several types are available, with different prices
and different functions. Practically all guard columns that I
am familiar with are of the cartridge column type. This
reduces the cost of the guard column as the endfittings can be
reused. Some are designed to be connected to your analytical
column with an additional piece of tubing, some connect
directly to your analytical column. While some guard column
designs may be better than others, the important point is to
use a guard column with the identical packing as in the
packing in your analytical column. With this approach,
column life can be increased much more effectively than
with column backflushing.

41. Selectivity Shift

Q.: We are determining the impurity profile of a drug. For
the parent drug and all impurity peaks but one the retention
time is constant. For one of the impurity peaks the retention
is decreasing slowly, day by day. The key difference between
the compound with the shifting retention and all the other
compounds is the fact that it contains a carboxylic acid
group. The column used is a reversed-phase column with an
embedded polar group. The mobile phase consists of
methanol and an ammonium acetate buffer adjusted to pH
4.9. What is going on? Why is this peak shifting, while all
the other ones have a constant retention?

49

A.: The observation that the retention remains rock solid for
most of the peaks tells us that the overall properties of the
column and the mobile phase are not changing. General
shifts in retention could be explained for example by
impurities collecting on the column. Since this is not the
case, we need to look specifically into the properties of the
analyte, for which this shift occurs.
You describe the analyte as the only one that contains a
carboxylic acid group. This indicates that the reason for the
shift in retention is specifically related to this group. This is
not impossible as you have described that you are using a
reversed-phase column with an embedded polar group. This
bonded ligand could be the explanation of the phenomenon
observed.
There exist several types of bonded phases with an
incorporated polar group. They contain different embedded
polar groups, and the preparation procedure can be different.
The SymmetryShield and XTerra RP columns (1,2,3)
manufactured by Waters Corp. contain a carbamate group,
the Prism and Spectrum columns (4) from Keystone an
urea function, and the Discovery RPAmideC16 from
Supelco an amide group (Fig. 1). All of these columns are
prepared in a single-step bonding reaction: the ligands are
assembled first, and then attached to the surface in a single
step, or at least there is no evidence for a multi-step surface
reaction.
O
e
C
SM
Me

R
N
H

SymmetryShield RP (Waters)
XTerra RP (Waters)

O
Si

Si

Discovery RP Amide 16

N C R
H
O
N
H

N
H

Prism (Keystone)

Figure 1. Commercial bonded phases with incorporated

carbamate group
Some of the older packings (5) with an incorporated amide
group, and unfortunately also some of the newer ones (6), are
prepared in a multi-step surface reaction. In this case, an
aminosilane is attached to the surface first. Then the amino
bonded phase is reacted with a long-chain acid chloride or
something similar to create the reversed-phase character of
the phase. Unfortunately, due to steric effects, such a surface
reaction is always incomplete and leaves many residual
amines on the surface. Attempts are made then to remove the
residual amines with something like acetic anhydride or
some other activated short-chain acid, but even such attempts
have not been successful. The surface of such a packing is
shown in figure 2. As it turns out, it is as difficult to remove
the residual amino groups on such a packing as it is to
remove residual silanols on a classical bonded phase.

OH
Si

HN
R

O
OH

Si

O
NH 2

hydrophobic collapse, but this is not a problem for bonded

phases with embedded polar groups. The retention
mechanism is no different than the standard reversed-phase
mechanism, and the bothersome interactions with residual
silanols are largely eliminated. More and more people are
discovering the advantages of these phases, and they are used
in more and more applications. Good phases are available
from more than one manufacturer, but you should inquire
about the bonding process. Single step processes are
preferred; they are free from the complications that you
encountered in your application.

Si
O
OH

N
H

CH 3

Figure 2. Amide phase prepared via a two-step reaction.

The phase contains measureable amounts of residual
amines.
Q.: Indeed, the phase that I am using is an amide phase. But
why does the retention of my acidic analyte change with
time?
A.: If your phase has such residual amino groups, one would
expect the acidic analyte to interact with these groups via
ion-exchange. This would result in increased retention, just
as the interaction of basic analytes with silanol groups results
in increased retention on classical packings. It is also known
from the preparation of amino phases that the stability of the
amino group in an aqueous mobile phase is not very good. It
appears as if your column is slowly loosing the residual
amino groups, and that this is the cause of the reduced
retention of your acidic analyte from one day to the next. All
the other analytes do not interact with the amino group,
therefore their retention times remain stable.
Q.: This is a possible explanation of the problem, but I do
not know anything about the preparation procedure of this
particular packing. How can I solve my problem? I have to
say that I do like these phases with the embedded polar
groups. They give good peak shapes for practically
everything.
A.: First, I would call up the manufacturer and discuss the
problem with him. If a manufacturer is aware of a problem,
he can address it in a next generation product. Or, you can
choose one of the products that is free from this problem.
Generally, I have to agree with you: the bonded phases
with embedded polar groups indeed give a superior peak
shape for even the most difficult samples. Unless one makes
an error in the choice of the mobile phase, no tailing is
observed for practically all analytes. Another good feature of
these packings is the fact that they can be used without
difficulty with mobile phases containing 100% water. This
means that even fairly polar compounds can be analyzed
without difficulty using packings with an embedded polar
group. Many of the better reversed-phase packings do not
tolerate highly aqueous mobile phases well due to

References:
1. J. OGara, B. Alden, T. H. Walter, C. Niederlnder, U. D.
Neue, Simple preparation of a C8 HPLC stationary phase
with an internal polar functional group, Anal. Chem. 67,
3809 (1995)
2. J. E. OGara, D. P. Walsh, B. A. Alden, P. Casellini, T. H.
Walter, Systematic study of chromatographic behavior vs.
alkyl chain length for HPLC bonded phases containing an
embedded polar group, Anal. Chem. 71, 2992 (1999)
3. U. D. Neue, T. H. Walter, B. A. Alden, Z. Jiang, R. P.
Fisk, J. T. Cook, K. H. Glose, J. L. Carmody, J. M. Grassi,
Y.-F. Cheng, Z. Lu, R. J. Crowley,
Use of highperformance LC packings from pH 1 to pH 12, American
Laboratory, November 1999, 36
4. K. J. Duff,
Improved HPLC separations using
intrinsically
base-deactivated columns, American
Laboratory, February 1998, 32AA
5. T. L. Askah, K. M. R. Kallury, C. A. Szafranski, S. D.
Corman, F. Liu, Chracterization and HPLC evaluation of
a new amide-functionalized reversed-phase column, J.
Liq. Chrom. Rel. Technol., 19, 3049 (1996)
6. J. J. Kirkland, J. W. Henderson, J. D. Martosella, B. A.
Bidlingmeyer, J. Vasta-Russell, J. B. Adams, Jr., A highly
Stable Alkyl-Amide Silica-Based Column Packing for
Reversed-Phase HPLC of Polar and Ionizable
Compounds, LC-GC 17 (1999), 634

42. New Method

Q.: In my department, we are constantly developing new
HPLC methods, versions of which will be used for multiple
purposes. Most of the time, the primary goal is to establish
an impurity profile for new drug substances. Versions of
these methods will then be used for dissolution testing or
stability testing, and ultimately simplified versions will be
used to test content uniformity. Currently, the methods often
get reinvented as they move from department to department.
Is there a way to streamline this process?
A.: This is a difficult question which has both a technical
component as well as a managerial component. Let us just
discuss the technical component here!

50

The tasks you describe suggest to me that you are working

in the pharmaceutical industry. Some of the answers may be
specific to your industry, but I am sure that similar issues
exist in other industries as well. Nevertheless, Ill address
your question in the context of the pharmaceutical industry.
An impurity profile a method often requires the resolution of
a large number of compounds. Analysis time is often not a
factor, since such an analysis is only done once in a while. In
dissolution testing, on the other hand, you are only interested
in a single peak, and you want to have a fast method with
good column lifetime. The differences between the needs of
the two types of analysis are great, and in principle there are
good reasons for a reinvention of the method, as you called
it.
Some components of this process can be streamlined, and
actually optimized. First of all, you need to decide on a
column family where the same packing is available in many
different column configurations. In principle, this should not
be difficult, since nearly all column suppliers have a large
number of column configurations. For example, the
XTerra family of columns (1) from Waters Corp.
comprises over 270 part numbers. A large number of column
configurations should allow you to select the correct column
for the different needs of the different departments.
Next, you should select a column configuration that is
suitable for generating an impurity profile. This is often is
the most complex type of analysis. Fortunately, it is also the
analysis which is needed earliest in a project. Therefore,
other methods can be derived from the method developed for
the impurity profile.
Method development for an impurity profile is frequently a
most demanding task. Often, columns with the best resolving
power (long columns packed with small particles) are used.
Gradient elution is most of the time the only way a
reasonable chromatogram of all possible impurities can be
obtained. A method with a complex elution profile also
needs to be rugged: you need to demonstrate that the results
from the method can be achieved using different columns
and different instruments in different departments. The good
news is that the person who has developed the method for the
impurity profile has done a large amount of homework useful
for the development of subsequent methods. This knowledge
can be used for the simpler methods either in the same
department or in other departments.
How does one go from this complex impurity profiling
method to a simpler one? In some cases, people just use the
same column and develop an isocratic separation for the
other needs, for example, for content uniformity testing. On
one hand, the use of the same packing material has its merits.
After all, the studies for the impurity profile just established
that the method is reproducible using this packing. On the
other hand, for a simple two-peak-chromatogram one does
not need a high-resolution column. One can use the same
packing in a shorter column to speed up the separation. Or
one can use a shorter column packed with a larger particle
size of the same packing to improve the ruggedness of the
method.

51

The path from a complex gradient method to a simple

isocratic method can be very rapid and straightforward, if
you have the right information. You need to determine, at
which solvent composition the parent peak elutes in the
gradient method. In other words, you need to know the exact
solvent composition at the column exit at the point of elution
of the parent peak. This requires a good knowledge of your
instrument, since there is a delay between the time that the
gradient is formed and the time it reaches the column exit
(2). You need to know the gradient delay volume of the
instrument, and the dead volume of the column. The dead
volume of the column is easy to obtain: you just inject an
unretained peak, such as dihydroxyacetone for a reversedphase column. The gradient delay volume of the instrument
is obtained by running a gradient with a UV absorber. The
delay volume is calculated from the difference between the
programmed start of the gradient and the time that the UV
absorber shows up in the detector.
The development of the isocratic method should start with
the same packing material and the solvent composition at
which the compound elutes in the gradient. If you do this,
you should get a retention factor around 2 and no less than 1.
You now have the start of a simple and fast isocratic method
for the other assays. It may be entirely sufficient to have such
a low retention factor. On the other hand, people prefer a
slightly higher retention factor, maybe between 2 and 5. To
obtain this, you simply need to adjust the solvent
composition slightly.
For content uniformity, stability testing and dissolution
testing, you do not need such a high-powered column as for
the impurity profile. You should select a shorter column.
Often, even a 5 cm 5 micron column gives satisfactory
results for these tests, the run time is much shorter and the
backpressure is much lower than with a 25 cm 5 micron
column. For the dissolution testing, you may consider a
guard column to protect the analytical column from the
acidic sample. However, no major redevelopment of the
methods is needed. As a matter of fact, the same simple
isocratic method can be used for content uniformity, stability
testing and dissolution testing. This should also streamline
the method validation process (3).
Of course, in many places the different tasks are executed
in different departments. Therefore, this requires a
coordination of the efforts in the different departments. In
todays world, with the large workload for everybody, such a
streamlining can only be beneficial.
References
1. U. D. Neue, T. H. Walter, B. A. Alden, R. P. Fisk, J. L.
Carmody, J. M. Grassi, Y.-F. Cheng, Z. Lu, R. Crowley, Z.
Jiang, Family of novel high-performance LC-packings
can be used from pH 1 to pH 12, American Laboratory
31, 22 (1999), 36 - 39
2. U.D. Neue, HPLC Troubleshooting, American Laboratory,
October 1997

www.pharmaguideline.com

3. M. E. Swartz, I. S. Krull, Analytical method development

and validation, Marcel Dekker Inc., New York, Basel,
Hong Kong, 1997

43. Negative Peaks

Q.: I am running a reversed-phase separation with a
phosphate buffer at pH 2.5 and 20% THF on a C18 column.
The detector is a PDA, and the wavelength is 225 nm. I am
getting a negative peak in the middle of the chromatogram,
between the second and the third peak. I am accustomed to
seeing negative peaks at the beginning of the chromatogram,
around V0, but I have never seen one in the middle of the
chromatogram. Can you explain what is happening?
A.: Maybe the best thing to do is to start with a simple
philosophy: a peak indicates a difference in the composition
of the sample and the mobile phase. It matters not, whether
the peak is positive or negative or whether it is retained or
not. Therefore, what we need to do is to figure out, what the
source of the extra peak in the chromatogram is. The fact that it
is negative is additional helpful information.
You state that you are used to negative peaks early in the
chromatogram. These peaks are mostly due to differences in
the composition of the sample matrix and the mobile phase.
This can be the solvent composition used to make up the
sample, or the pH. Unless one is extremely careful in the
preparation of the sample in mobile phase, extra peaks can
stem from this difference. For most of us, this is not a
concern. We realize that there are differences between the
sample solvent and the mobile phase. Sometimes, we also
create such differences on purpose, for example for an
enrichment of the sample on the column.
Q.: This is indeed correct. Often, my samples contain
extraneous ingredients such as excipients. However, they
typically elute with the solvent front. This is the case here as
well. Also, the sample is not exactly dissolved in the mobile
phase, but is it diluted with mobile phase.
A.: To sort out, if the extra peak is coming from excipients
or from somewhere else, all you need to do is inject
standards. If this is an excipient peak, it should disappear.
Q.: I thought of this already. The extra peak is still present,
when I inject standards or even mobile phase.
A.: OK. Now, this is an important piece of information. If
you get the same phenomenon independent of the nature of
the sample, we can exclude the sample as the cause of the
problem. Therefore, we should focus on instrument
components or the mobile phase itself.
If your extra peak were positive, one could consider a
contamination in the injector, the syringe used, seals in the
sample vials or related things associated with the injection

process. Since this is not the case, I would reject the injection
process as a source of the problem.
The mobile phase itself and all the components that it is in
touch with are the most likely source of the problem. Imagine
an ingredient in the mobile phase that is present at very low
concentration and has some retention on the column. If you
inject mobile phase that does not contain this ingredient, you
will get a negative peak with the same retention factor as the
ingredient would have, if you would inject it as a sample.
This is called vacancy chromatography. To solve your
problem, all we need to do is to find this ingredient. The
unfortunate part is that you are working at a very low
wavelength. You will see more compounds at a higher
sensitivity in the low UV region. I am sure that this is the
reason why you selected this wavelength to start with.
Q.: This sounds like a good description of what my problem
might be. What can I do to narrow it down further so that I
can eliminate it?
The first thing to do is to check the mobile phase itself.
What is the quality of the THF? What is the quality of the
water? You can simply inject a sample of THF and a sample
of water and see if you encounter a peak at the same
retention time as your negative peak. If you can create a
peak, you have found the source of the problem, and then
you can go to the next step. If it is the THF, you can
investigate the quality and purity of the THF. THF is prone
to form peroxides, and some forms of the THF contain
components that are designed to inhibit the formation of
peroxides. You should be able to get this information from
the label on your solvent bottle. THF is also an excellent
solvent for plastic parts and for extracting additives from
these plastics. Anything in touch with your THF solvent line
is suspect: filters, the tubing, seals, ferrules, anything
The water is a possible source as well. In most places,
water purification devices such as a Milli-Q system from
Millipore Corp are used. The filters in these devices have a
limited capacity and should be replaced after some period of
time.
Another possible source of contamination is the buffer or
the buffer preparation itself. What did you use to stir the
flask when preparing the buffer? Was the flask clean? If you
measured the pH with a pH meter, did you clean the
electrode before you used it?
After you checked all the components of the mobile phase
and the devices in touch with the mobile phase and the
problem still has not yet been resolved, I would check the
HPLC instrument in detail. Are there any components that
are incompatible with your mobile phase, especially with
THF? Are there any pockets in the instrument that have not
been purged properly? Is there a possibility for bacterial
growth in your solvent path? I would place my bets on the
solvent, the water and the buffer preparation and look very
closely at these steps.
Q.: This sounds good, but it also is a large amount of work.
Is there a way to narrow it further?

52

A.: On first glance, I do not see how. On the other hand, you
may want to ask yourself the question, why do you want to
eliminate the negative peak. From the description of the
problem it appears that the negative peak does not interfere
with the analysis. Therefore, you may just declare that this is
a fact of life and ignore this. Of course, if you are in a
regulated environment, this should be thought through
carefully.

44. Tricky, Tedious, Time Consuming

Q.: We are doing a very large number of routine assays of
samples from clinical trials. The samples are plasma samples
from patients, and we are determining the metabolic fate of
the parent drug and its metabolites. This is very important
work, and the samples are precious, but the work is also very
laborious. The sample preparation is the major problem. We
are using C18 cartridges to remove plasma constituents, and
the entire sample preparation procedure has to be carried out
with a high precision. If we dont do that, the results vary a
lot. The actual analytical method is a HPLC method. Is there a
better way to do the sample preparation?
A.: Indeed, this type of work is tricky, tedious and very time
consuming. This applies both to the development of the
methods and to the handling of the samples themselves.
However, in recent years a range of developments has
occurred that makes this task easier (1).
The first problem encountered with classical C18-type
packings is the difficulty encountered with the wettability of
such packings with aqueous samples. The typical
preconditioning of the SPE cartridge requires a wetting with
methanol, then a brief conditioning with water or buffer
before the loading of the sample. If the SPE cartridge dries
between the preconditioning step and the loading of the
sample, low recoveries of the analytes of interest are
encountered. This requires to pay close attention so that this
does not happen. This is not such a large problem, when one
is dealing with individual samples. But it does create
difficulties when one needs to deal with many samples at the
same time such as in a 96-well plate. Therefore,
manufacturers have introduced improved SPE products that
do not suffer from this problem. An example is the Oasis
HLB SPE product line from Waters Corp. This packing
material contains a mixture of both hydrophobic and
hydrophilic groups. The hydrophobic groups impart
exceptional retention, higher than most C18 packings, while
the hydrophilic groups provide good wetting with aqueous
samples. The consequence is that these packing materials can
dry out during the sample preparation step without loosing
capacity. This makes the sample preparation step much less
tedious. Now you dont have to sit there any more and
carefully watch the fluid level in your SPE device to make
sure that it does not dry out.

53

Q.: This sounds interesting. What is the secret behind this?

A.: It is very simple. The reason for the loss in capacity
when a C18 cartridge dries out is the fact that C18 is not
wetted by water. What prevents the water from penetrating
the pores is the poor wetting angle between water and the
hydrophobic surface of the C18. If one provides hydrophilic
groups on the surface, the wetting is improved and water
stays in the pores or penetrates the pores easily. For example,
the Oasis HLB packing is prepared from a mixture of
divinylbenzene as the hydrophobic component that provides
retention and N-vinyl pyrrolidone, which provides the
wettability with water. The control of the amount of both
components provides the right balance between
hydrophobicity for solute retention and hydrophilicity or
water wettability.
Q.: OK. I understand the improvement in wettability.
However, how does it behave as an adsorbent? I like the way
the reversed-phase packings are acting. It is easy to
understand, and not difficult to make them work.
A.: There are a lot of similarities between these hydrophobic
sorbents and a C18 packing. Generally, the retention
mechanism is similar: whatever is retained strongly on a C18
packing due to a reversed-phase mechanism is retained
strongly on these packings as well. On the other hand, they
are not based on silica, which means that the retention
mechanism is not complicated by silanols. Plus, for the same
reason, they can be used without difficulty with alkaline
solutions, which opens new doors in the sample preparation
process.
Q.: The fact that there are no silanols is a mixed blessing. I
can see how this can make the elution protocol simpler. But
we also use the silanol interaction to improve retention, if we
need it.
A.: On classical C18 packings, residual silanols are a side
product. The control of their activity is generally not as good
as the control of the hydrophobic activity of these packings.
While residual silanols indeed can help in retaining basic
analytes, the reproducibility of such a procedure is more
tricky than that of a packing that exhibits only reversedphase character.
On the other hand, these polymeric packings are stable to
higher pH-values than silica-based packings. This allows you
to use pH as a more powerful tool in SPE than was possible
with silica-based packings. You can selectively improve the
retention of basic analytes by carrying out the washing
protocol at basic pH values. Or you can achieve selective
elution by changes of the pH. Generic procedures for this
have been worked out (2,3). The manipulation of the ionic
interaction is now entirely under your control, and does not
depend any more on the concentration of residual silanol
groups.

www.pharmaguideline.com
The good wettability of these polymeric packings also
opens a new avenue: they can be used without difficulty in
parallel processing schemes such as 96-well plates (2). The
fact that you do not have to worry about the drying of every
single well just makes the whole sample preparation process
simpler. In addition, the simultaneous processing of many
plates eliminates many sources of error. For example,
standards can be processed together with the samples to
control the process or to correct the results. Also, significant
time savings are realized by parallel processing of many
samples.
To summarize:
The favorable wettability of the second generation SPE
sorbents makes sample preparation less tedious. The absence
of strong secondary interactions such as the silanol
interactions of C18 packings makes sample preparation less
tricky. The use of 96-well plates for parallel processing of
samples makes sample preparation much less time
consuming than it used to be. There is progress
References
1. Y.-F. Cheng, D. J. Phillips, U. D. Neue, M. Capparella, L.
L. Bean, Simple extraction methods for the determination
of drugs in serum, American Biotechnology Laboratory
December 1997
2. Y.-F. Cheng, U. D. Neue, L. L. Bean, Straightforward
solid-phase extraction method for the determination of
verapamil and its metabolites in plasma in a 96-well
extraction plate, J. Chromatogr. A 828 (1998), 273-281
3. Y.-F. Cheng, U. D. Neue, L. L. Woods, Novel highperformance liquid chromatographic and solid-phase
extraction methods for quantitating methadone and its
metabolite in spiked human urine, J. Chromatogr. B 729
(1999), 19-31

best approach is therefore a combination of LC with MS. If

you can synthesize 1000 new entities per day, your analysis
time can not be much longer than 1 minute to keep up with
the demands.
Until a short time ago, the difficulty has been that HPLC
was considered to be too slow to get such a short analysis
time (1, 2). However, with some rethinking of the details of
the process, faster and faster analyses have become possible
(3).
There are two elements that have contributed to this. One
is a reconsideration of the parameters underlying the gradient
separation. The other element is the commercial availability
of very short columns packed with very small particle sizes
(4). The combination of both elements is the secret to ultrafast separations.
In order to maximize the number of peaks that can be
separated in a gradient, one needs to consider two separate
contributions. The first one is the width of the peaks
stemming from the manipulation of flow rate, or better linear
velocity, the second one is the interplay between the gradient
volume and the width of the peaks as the gradient volume
expands. Both parameters together determine, how many
peaks can be crammed into a limited analysis time.
The maximum number of the peaks that can be generated in
a gradient is called the peak capacity. Since the peaks in
gradients are for the most part of the same width, it can be
calculated easily by dividing the gradient duration tg by the
peak width w:
P =1+

tg

(1)

The theory of chromatography allows us to calculate the

peak capacity as a function of the operating conditions (3).
N
Bc
(2)
P =1+

Bc

+1

tg

45. Fast Separations

Q.: I recently heard a lot about ultra-fast separations, with a
complete analysis within less than 5 minutes. Another subject
that interests me is the subject of ballistic gradients. I
would be interested to hear more about this, and what one
can do today.
A.: This is a very interesting subject indeed. While it is not a
troubleshooting subject, I think that it is worth a discussion.
The development of these very fast separation techniques is
driven by the ability to generate new chemical entities very
rapidly through combinatorial chemistry. One would like to
get good information on the success of these fast synthetic
techniques. Mass spectrometry or NMR can help you to
determined whether you have synthesized the chemical entity
that you wanted to create, but only a separation technique
such as HPLC can get you information on the quality or the
yield of your synthesis and the possible side products. The

N is the plate count of the column under the operating

conditions, B . c is a parameter that relates to the type of
samples to be separated and the solvent span c over which
the gradient is executed, t0 is the column dead time, and tg is
the gradient duration, as above. We can take the factor B . c
to be a constant that depends only on the type of analysis that
is run. There are two reasons for this: one is the fact that the
types of compounds to be analyzed in a combinatorial library
are all similar to each other with a similar structure and a
similar molecular weight. The other is the fact that we are
always running the same gradient, mostly from 5% organic to
95% organic. The column plate count is a function of the
linear velocity, which in turn depends on the column dead
time t0, and the gradient duration determines the total
analysis time.
From equation 2, we can generate a three-dimensional
graph showing peak capacity as a function of the linear
velocity - or better flow rate - and the gradient duration
(Figure 1). It can be seen as a graph which measures the
resolution capability of a column. For every column length

54

and particle size, different graphs can be generated. The

example in Figure 1 shows the performance of a 5 cm 5 m
column with an internal diameter of 4.6 mm. For every
gradient duration, the graph exhibits a performance optimum at
a different flow rate. Also, longer gradients give a higher
peak capacity than shorter gradients. The maximum flow rate
that can be used with every column depends on the pressure
limitation of the instrument.

5 cm 5 m Column, 4.6 mm i.d.

Peak Capacity

>

200

150

100

50

7.45

1.0
10.53

3.72

F [mL/min] >

5.26

1.86

2.63

0.93

2.0
1.32

0.47

0.66

0.23

8.0
4.0
0.33

0.12

0.16

Gradient
duration
[min]

32.0
16.0

Figure 1. Peak capacity as a function of the flow rate and

the gradient duration for a 5 cm 5 m column
How does the picture change if we use another column?
Figure 2 shows the same graph for a 2 cm 2.5 m column
with the same internal diameter.
2 cm 2.5 m Column, 4.6 mm I.D.

of the interplay between the expansion of the gradient as

expressed by the ratio t0/tg in equation 2 and the dependence
of the plate count on the linear velocity or flow rate. For the
column shown in figure 1, the best flow rate is around 7
mL/min for a 1 minute gradient. This is a higher flow rate
than what most people consider to be best.
We see immediately that the optimum performance is
higher at fast gradient times. Also we see that the flow rate at
which the optimum performance occurs is similar to the 5 cm
5 m column. One can see that the same one-minute
ballistic gradient can still be executed on these very short
and very fast columns with better resolving power than with
the more traditional 5 cm 5 m column shown in Figure 1.
The faster mass transfer of the 2.5 m particles shifts the
optimum of the curve up to a higher resolution value. This
means that quite powerful separations can be carried out in
timeframes compatible with the needs of high-throughput
combinatorial chemistry - at least for the moment. If
combinatorial chemistry output speeds up still further, the
analytical chemists will need to think again how to speed up
chromatography still further.
References
1. J. N. Kyranos, J. C. Hogan, Jr., Anal. Chem. News &
Features, June 1 1998, 389A
2. W. K. Goetzinger, J. N. Kyranos, American Laboratory,
April 1998,
3. U. D. Neue, J. L. Carmody, Y.-F. Cheng, Z. Lu, C. H.
Phoebe, T. E. Wheat, Design of Rapid Gradient Methods
for the Analysis of Combinatorial Chemistry Libraries and
the Preparation of Pure Compounds, Advances in
Chromatography, in print
4. Y.-F. Cheng, T. H. Walter, Z. Lu, P. Iraneta, B. A. Alden,
C. Gendreau, U. D. Neue, J. M. Grassi, J. L. Carmody, J.
E. OGara, R. P. Fisk, Hybrid Organic/Inorganic Particle
Technology: Breaking Through Traditional Barriers of
HPLC Separations, LC-GC, November 2000

200
180

46. Buffers for LC/MS

160

Peak Capacity

140

Q.: What mobile phase additives can be used with reversedphase columns and LC/MS detection?

120
100
80
60
40
23

8.42

4.21

1
5.96

1.49

2.11

0.74

2.98

Flow Rate [mL/min]

1.05

0.37

3
0.53

0.19

0.26

0.07

0.09

0.13

8
0.05

Gradient
Duration
[min]

20

Figure 2. Peak capacity as a function of the flow rate and

the gradient duration for a 2 cm 2.5 m column
It can be seen that the performance maximum occurs at a
fairly high flow rate for very fast gradients. This is the result

55

A.: Since hyphenated LC/MS instrumentation has become

more and more popular during the last few years, the
standard buffers of former times, for example phosphate,
have dropped out of favor. The issue is that one would like to
use volatile mobile phase additives. Phosphates are not
volatile, and - with time - will clog the LC/MS interface.
This results in significant downtime, and more importantly,
in a large amount of work to clean up the interface again.
Some new source designs can accommodate now up to 10
mM of a phosphate buffer for an extended period of time, but
regular cleaning is still required.
Therefore, standard HPLC methods used with MS
detection use mobile phase additives that are volatile. The

www.pharmaguideline.com
simplest additives are simple acids, such as formic acid or
acetic acid. They are often used in a concentration of 0.1%
up to 1%. This provides an acidic environment, enough to
fully protonate basic analytes. In many cases, this is
sufficient to provide good control over the retention of both
acidic and basic analytes.
However, there are exceptions, where such an approach
has its limitations. One of the disadvantages of a strongly
acidic mobile phase environment is the fact that the
ionization of acidic analytes is suppressed. Therefore, a
better condition for MS detection is to use a slightly higher
pH in conjunction with detection in negative ion mode. In
such a case, ammonium acetate and ammonium formate
buffers are used, and pH control is important to control
retention. The pKa of formic acid is 3.75, and the pKa of
acetic acid is 4.75. Consequently, these are the optimal pH
values for these buffers. The typical concentrations are
around 20 mM or lower, even as low as 5 mM are possible.
At such a low buffer concentration, it is definitely best to use
the buffers at a pH very close to the pKa of the buffering ion.
A reasonable rule of thumb for such low concentrations is to
use a buffer at +/- 1 pH units around the pKa of the buffer.
Another difficulty that can be avoided with a correct
choice of the pH is the suppression of the ionization of
analytes due to matrix interferences. This is a common
problem in the analysis of plasma samples by LC/MS.
However, most matrix interferences are ionic or ionizable in
nature, and the analytes of interest often are ionic as well.
Therefore the coelution pattern can be changed by changing
the pH of the mobile phase. With an appropriate change in
the elution pattern of the analytes and the interferences, ion
suppression can be avoided.
Occasionally, it might be desirable to run the
chromatography at alkaline pH. The reason for this can be
the avoidance of ion suppression, or a desire to improve
sensitivity by improving the ionization of the analytes of
interest. Acidic analytes give the best MS response in the
alkaline pH range. On the other hand, we found that even
basic analytes still respond surprisingly well under alkaline
conditions. The mobile phase additive that we use for weakly
alkaline mobile phases is ammonium bicarbonate. The pKa
values of the buffer constituents are 9.2 and 10.2. Therefore,
ammonium bicarbonate can be used over a broad alkaline pH
range, with the most preferred value at pH 10. This buffer is
completely MS compatible. Upon heating, ammonium
bicarbonate decomposes into only volatile components:
ammonia, water, and carbon dioxide.
To summarize: we have some good solutions for MS
compatible buffers in the acidic pH range from pH 3 to pH
5.5 and in the alkaline pH range from pH 8.5 to pH 10.
Q.: How about the salt ammonium acetate? I have seen it
being used at neutral pH.
A.: Yes, I have seen this as well. However, we must realize
that it is not a buffer at all. At pH 7, ammonium acetate has
no buffering capacity whatsoever. One should consider it to

be just a salt additive. It may help in the MS ionization

process, but it has little function in the separation process.
For the neutral pH range, the best solution is to start with
ammonium bicarbonate and add formic acid. The actual
buffering is provided by the first dissociation of the
carbonate ion. At high concentrations, carbonic acid
dissociates to form carbon dioxide and water. But at the low
concentrations of the buffers used in LC-MS, this is a
functional approach and no degassing has been found.

47. Alkaline Buffers for RPLC

Q.: In the last few years, reversed-phase packings have
become available that are stable in the alkaline pH range. I
would like to explore some of these new capabilities. What
buffers can be used with reversed-phase columns in the
alkaline pH range?
A.: Indeed, the exploration of a broader pH range opens up
new capabilities that were not accessible in the past. Some of
the newer packings are stable to pH 10, and others even to
pH 12. Even classical packings can often be used in the
alkaline pH range, if less aggressive buffers are chosen.
Q.: What are the more aggressive buffers that should be
avoided?
A.: Clearly, phosphate has been found to be more aggressive
in the alkaline pH range than other buffers. Therefore, I do
not recommend to use alkaline phosphate buffers with silicabased packings or related packings based on inorganicorganic hybrids. Phosphate has a limited pH range anyway.
Ammonia is also fairly aggressive, but it can be substituted
without difficulty with organic amines. In addition, organic
amines cover a broad pH range, with pKa values ranging
from 9 to 11.5.
Q.: OK, then which buffers are recommended?
A.: We have assembled three tables (1) of buffers suitable
for the alkaline pH range. Table 1 contains inorganic buffers,
table 2 organic buffers, and table 3 zwitterionic buffers, as
they are commonly used in biochemical applications. Some
of the buffers can be used as mobile phase additives as well,
just as formic acid and acetic acid are used in the acidic pH
range. Mobile phase additives may be useful for the control
of the ionization of analytes outside +/- 2 pH units around
the pKa of the analyte. However, if the pKa of the analyte and
the pKa of the mobile phase additive are in the same pH
range, a control of retention or peak shape may not work as
well as with a true buffer.
Among the inorganic buffers, ammonium bicarbonate is a
very good choice for a multitude of reasons. For one, it
covers a broad pH range due to the combined buffering
capabilities of the carbonate ion and the ammonium ion. In

56

addition, it is compatible with MS detection, since it

decomposes into volatile components. While it has a small
background in the low UV-range, the absorbance is about the
same as the background of formic or acetic acid or TFA in
the acidic range. This is not perfect, but it can be used even
in gradients in the low UV with a good match in the
composition of the mobile phase.
Table 1, Inorganic Buffers
Buffer
pKa-Value
Phosphate 2
7.21
Borate 1
9.14
Ammonia
9.25
Carbonate 2
10.25
Phosphate 3
12.33
Ammonium Bicarbonate 9.25 and 10.25

pH-Range
6.2 - 8.2
8.1 - 10.1
8.2 - 10.3
9.2 - 11.3
11.3 - 13.3
8.2 - 10.5

Table 2, Organic Buffers

Buffer
Glycine
Trimethylamine
1-Methylpiperidine
Triethylamine
Piperidine
Pyrrolidine

pH-Range
8.8 - 10.8
8.7 - 10.7
9.3 - 11.3
9.7 - 11.8
10.2 - 12.2
10.3 - 12.3

pKa-Value
9.8
9.74
10.3
10.75
11.2
11.3

References:
1. Data assembled by Charles H. Phoebe, Waters
Corporation
2. U. D. Neue. T. H. Walter, B. A. Alden, Z. Jiang, R. P.
Fisk, J. T. Cook, K. H. Glose, J. L. Carmody, J. M. Grassi,
Y. F. Cheng, Z. Lu, R. Crowley, American Laboratory 31,
22 (1999), 36 - 39

48. Post-Column Derivatization

Q.: I am considering post-column derivatization to improve
the specificity and the detection limits of a method. What do I
need to take into account to be successful?

Table 3, Zwitterionic Buffers

Buffer
MES
MOPSO
MOPS
TAPS
CAPSO
CAPS

range from 6 to 10.5. There is a suitable buffer for all the pH

ranges of interest in this discussion.

pKa-Value
pH-Range
6.2
5.2 - 7.2
7.0
6.0 - 8.0
7.3
6.3 - 8.3
8.5
7.5 - 9.5
9.7
8.7 - 10.7
10.5
9.5 - 11.5

Among organic bases (table 2), several options are

available that cover a broad range of pKa-values, from 9.7
for trimethylamine to 11.3 for pyrrolidine. Trimethylamine
has a very low boiling point and is compatible with MS
detection. Of course, you need to use a MS compatible
counterion such as formate or you can use trimethylamine
just as a mobile phase additive, as you would use formic acid
or acetic acid in the acidic pH range. We have used
pyrrolidine buffers at pH 11.5 successfully for extended
periods of time (2) with XTerra reversed-phase columns.
Triethylamine is the traditional additive to reversed-phase
buffers used to suppress tailing of basic analytes. People try
to avoid using it due to its unpleasant odor.
The odor is a general disadvantage of the bases. However,
there is a solution to this as well: zwitterionic buffers, shown
in table 3. They are not volatile, therefore there is no smell.
On the other hand, this makes them not compatible with MS
detection any more. However, if your method will never see
a mass spectrometer, zwitterionic buffers may be a very good
choice. They are used in many applications in biochemistry,
and are available in good purity. The pKa-values in table 3

57

A.: A lot of things! First of all, you need to find a fast

reaction scheme for the compound(s) of interest. To be
compatible with an HPLC method, the reaction needs to give
you a good response in less than 5 minutes, preferentially in
less than 2 minutes. Second, the reaction needs to be
compatible with the HPLC mobile phase. Specifically, if you
are using a reversed-phase method for your separation, the
reaction needs to be compatible with an aqueous medium.
This may not always be the case. Third, the reaction
product(s) need to be easily detectable with common HPLC
detectors. To achieve low detection limits, the reaction
product(s) should be detectable under conditions where there
is little native interference from other compounds. Examples
of this condition is absorbance detection in the visible wavelength range or fluorescence detection. Fourth, the detection
should be sufficiently specific for the goals of your analysis.
For example, a generic reaction scheme for amines will
enhance the detection of all amines in the world, but if your
sample contains only those amines that you want to detect, it
may be the best solution to your problem.
Q.: Indeed, there are a lot of requirements. I would like to
understand them a little bit better. Why do I need such a
short reaction time?
A.: I just gave you a rule of thumb. This means that there is
some flexibility around this. Let us consider, what is
required! The HPLC separation requires a constant stream of
mobile phase. This stream of mobile phase is mixed postcolumn with the derivatization reagent. Then the combined
streams need to be stored somewhere until the reaction has
developed the desired signals. The common way for doing
this is in a flowing stream. The mobile phase may be flowing
at 1 mL/min. To this, you add reagent at the same flow rate.
If your reaction takes 5 minutes, you need a storage volume
of 10 mL between the column and the detector. This is about
250 m of standard 9/1000 capillary tubing. Even if you are
using tubing with an i.d. of 0.5 mm, you still need about 50

m of tubing. This is a lot of tubing. You need to worry about

the bandspreading in the tubing as well. If you would use a
run-of-the-mill tubing, the bandspreading in it would be in
the order of a milliliter or more. The peak width of a peak
eluting at a retention factor of 2 from a 4.6 mm x 150 mm
column is only about 0.2 mL. Peaks that are well resolved in
the column would be diluted and mixed with each other.
Therefore the extra-column bandspreading in our example
would destroy the separation that you have achieved in your
high-performance HPLC column. Now you see, why one
wants to have short reaction times.
Q.: OK, I see. You said that this applies when one is using a
run-of-the-milltubing. What does this mean, and what
other options do I have?
A.: What I was referring to was a straight piece of tubing or
simple coils. If you are using teflon tubing, there is
something that you can do about the bandspreading. You can
knit it.
Q.: Whats that? Knitting a piece of teflon tubing???
A.: You heard correctly. This refers to a technique of
applying a controlled deformation pattern to the tubing (1, 2,
3). This controlled pattern creates a secondary flow inside
the tubing that provides radial mixing. The secondary flow
reduces the bandspreading in the tubing quite drastically.
The HETP may be 100 or even 1000 fold smaller in a well
designed geometrically deformed tubing than in a straight
tubing. The effect costs a bit in pressure, but this is not a
problem under typical operating conditions.
The important part about the knitting technique is the
pattern of the deformation of the flow path (1). One wants to
achieve a constant shift of the direction of the centrifugal
force that acts on the flow in the tubing and creates the radial
mixing. Something like a three-dimensional roller coaster.
The best configurations have an HETP around 0.5 cm,
resulting in about 10 000 plates in 50 m of tubing. Such highperformance configurations make post-column derivatization
compatible with HPLC separations and largely eliminate the
bandspreading problem described above.
Teflon tubing is also a good choice for other reasons (3). It is
inert to practically all reaction media and mobile phases that
one can think of, and it has a good temperature stability. In
order to speed up a reaction, you may want to work at
elevated temperature. Other plastic tubing may become soft
under the conditions that you may want to use.
Steel or copper tubing are other choices for your postcolumn reactor. They may be sufficiently inert to the reaction
medium of your derivatization reaction. However, to deform
them into the optimum forms of the knitted pieces of
tubing is not a simple task. You will compromise
bandspreading performance compared to what can be done
with teflon tubing.

Q.: Let us discuss the reaction conditions! What do I need to

think about?
A.: First of all, the reaction should satisfy your needs. Is it
specific enough, and is the detection sensitive enough. A
large number of reactions have been explored already, and a
book has been compiled on the subject (4). This should help
you to get the basic information that is needed. You need to
explore the differences between the analytes of interest and
the matrix. The more specific your reaction and your
detection schemes are, the more matrix interferences you can
tolerate. Also the response of the compounds of inerest
should be significantly different from the non-reacting
background. This is the reason why most published reaction
schemes result in the formation of color or fluorescence.
Next, you want to have a reasonably short reaction time.
Slow reactions and short analysis times are not compatible.
For most practical purposes, the reaction time should not
exceed 5 minutes. You can speed up most reactions by
increasing the temperature or by increasing the concentration
of the reagent. The reaction does not need to go to
completion. Due to the combined effect of the progress of
the reaction and the bandspreading in the tubing, you will get
the maximum detector response before completion of the
reaction. However, you need to find conditions, where small
variations in the reagent concentration or the temperature do
not affect the response to an appreciable degree. It is usually
not difficult to find a plateau in the response pattern. Of
course, this can be analyzed mathematically as well (1), but
for most practical purposes a few experiments will provide
the correct answer.
You are planning to get involved in an exciting
technology, and I wish you much success.
References:
1. U. D. Neue, Ph.D. Thesis, Universitt des Saarlandes,
Saarbrcken (1976)
2. H. Engelhardt, U. D. Neue, "Reaction Detector with
Three-Dimensional Coiled Open Tubes in HPLC",
Chromatographia Vol. 15 (1982) No. 7, 403-408
3. I. Krull, Reaction Detection in Liquid Chromatography,
Marcel Dekker, New York (1986)
4. G. Lunn, L. C. Hellwig, Handbook of Derivatization
Reactions for HPLC, Wiley & Sons, Inc., New York
(1998)

49. Gradient Dwell Volume

Q.: What is the gradient dwell volume of an HPLC system,
and how does it affect the analysis?
A.: The gradient dwell volume is the volume between the
point where the gradient is mixed and the column inlet. This
volume delays the onset of the gradient, and is therefore also
called the gradient delay volume.

58

Most modern gradient HPLC systems are single-pump

low-pressure gradient systems. This means that the gradient
is mixed upstream of the pump. The first component of the
gradient delay volume is the volume of the gradient mixer.
To this you need to add the connections between the gradient
mixer and the pump heads. Next is the volume of the pump
heads. What follows is the connection tubing to the injector.
Often, this volume is increased to provide a mixing of the
gradient components to smoothen the gradient. The volume
of the injector and the connection between the injector and
the column represent the next part of the gradient dwell
volume. As you can see, with a typical low-pressure gradient
system, there is a lot of volume where the gradient can
dwell
It is also possible to generate the gradient on the
high-pressure side. This requires at least two pumps, but
enables us to eliminate all of the gradient delay volume.
However, high-pressure gradient systems are not commonly
set up this way. Normally, the connection between both
pumps is made upstream of the injector. This has the
advantage that the sample is always transported to the
column, and does not depend on the flow in either pump. A
better approach, albeit one where a bit more care is needed, is
to inject the sample into the flow of the pump that delivers all
or most of the starting mobile phase. This means that the
gradient is mixed behind the injector, and the gradient delay
volume can be made completely negligible. A small
drawback of this technique is that the sample is diluted a bit
with the second component of the mobile phase. This is only a
minor issue if the gradient starts with 90% of the Acomponent of the gradient. Then the sample is diluted only by
10%. The other thing is that in general most peaks in a
gradient are focussed on the column. Therefore this small
drawback may affect to a small degree only a few peaks that
elute in the isocratic portion of the chromatogram before the
gradient starts. Under these circumstances, we need to make
sure that the volume between the injector and the column is
negligibly small. High-pressure gradient systems allow us to
completely avoid any gradient delay.
Now let us discuss how the gradient delay volume affects
the separation! Since it takes time for the gradient to reach
the column, the onset of the gradient is delayed. Therefore
there is a time period where the column is operated
isocratically in the mobile phase used at the beginning of the
gradient. The first effect may be that early eluting peaks are
affected by the gradient delay volume and may elute
differently on another instrument with a different gradient
delay volume. If the gradient delay volume is a substantial
part of the chromatographic run, one can observe quite
significant differences in the elution profile early in the run.
Late in the gradient, the elution pattern remains the same
independent of the delay volume. However, the elution time
will shift in direct proportion to the change in the gradient
delay volume. Both of these conditions are rather annoying.
They are the reasons why gradient methods are often avoided
in a QC laboratory. After all, analytes are commonly
identified using the retention time, and if the retention times

59

change from instrument to instrument, one has trouble with

the identification. Of course, there are ways around this. For
example, one can specify the retention pattern in the form of
differences from the retention times of a standard. However,
this complicates things
Q.: I agree, this is a bit more complicated, but it is not an
insurmountable obstacle. Are there any other issues to
consider?
A.: If there is a substantial mixing volume in a low-pressure
gradient system, one will also observe that the gradient is not
sharp. This usually does not play a big role in standard linear
gradients, but it may affect the elution pattern if steps are
used in the gradient. Generally, it is preferred to use gradient
systems with a small delay volume. Many modern HPLC
systems are designed for a minimal gradient delay.
For very rapid gradients with minimal cycle time, even a
very small gradient delay volume may not be sufficient.
Therefore other solutions have been implemented. For very
rapid routine analyses, the delayed injection technique can be
used. In principle, the gradient itself is executed just like
under normal operating conditions. However, the first
injection is delayed until the gradient reaches the injector.
Right at that moment, the sample is injected. The sample is
now essentially separated by the gradient as if no gradient
delay volume were present. Of course, there is a small delay
volume comprising the volume of the injection loop and the
tubing between the injector and the column. But this can be
made negligibly small.
After the gradient has been executed, the column is
reequilibrated with the starting mobile phase. Of course, the
reequilibration will be delayed by the same time as the
gradient itself. Therefore we dont have to wait until the
column is reequilibrated to start the next gradient run. We
just have to remember that what we program into the
instrument and what the pump is executing is different from
what is actually happening at the column inlet. Gradient
execution and column reequilibration are just offset by the
time needed to purge the gradient delay volume. This
modern solution to the problem of the gradient delay allows
us to execute fast gradients without delay on single-pump
instruments. In addition, if we execute the same program
with different instruments, the possible differences in the
gradient delay volume can be made negligibly small. This
should enable us to execute true gradients reproducibly on
different instruments. Of course, all of this assumes that the
gradient generator works reliably and reproducibly and
indeed generates the gradient that we would like to see.
Q.: OK, this sounds good. How can I measure the gradient
delay volume and how can I make sure that the system
accurately delivers the gradient that I want?
A.: There is a standard test that is quite useful. You can run
your gradient without a column in place with a small amount
of an UV absorber in the mobile phase B. This allows you to
observe the actual execution of the gradient. From the

difference between the program and the actual execution of

the gradient, you can determine the gradient delay volume of
your system. This is something that you may want to know
anyway, if you want to take full advantage of the trick with
the delayed injection. The more you know about your
system, the more successful you will be eliminating the
wasted time in gradient separations.

50. Buffer Capacity

Q.: What is buffer capacity, and why is it important?
A.: Buffer capacity is a measure of the strength of the buffer.
Mathematically, it is the reciprocal value of the slope of the
titration curve of a buffer. It specifies the amount of
hydronium ions or hydroxy ions that are needed to change
the pH of the buffer by a certain value. The larger the buffer
capacity, the larger is the amount of acid or base that can be
added to the solution without a change in pH.
This also shows why the buffer capacity is important in
chromatography. We know that the retention of ionic or
ionizable analytes may depend on the pH of the mobile
phase. If we are in a situation like this, it is important to have
good control over the pH of the mobile phase. If we do not
have good control, the retention of the analytes and even the
selectivity of the separation may vary from day to day with
the mobile phase preparation. This of course is not desirable.
Let me give you a few examples to clarify this. Let us assume
that we are separating two acids with a similar pKa between 4
and 5 using a reversed-phase column. We have prepared the
mobile phase with KH2PO4, which gives us a pH of around
4.5. This is about the worst situation that one can encounter.
In reversed-phase chromatography, the retention of both of
our sample acids will depend strongly on the degree, with
which the acids are ionized. The ionic form of an analyte is
always much less retained than the non-ionic form. A good
rule of thumb is that the retention of both forms is different
by about a factor of 30. Therefore the retention of both of
our analytes can change a lot, if the pH varies just a little. In
addition, the change in retention may be different for both
acids, especially considering that both have a similar pKa,
but not the same pKa. Consequently, the resolution between
both compounds will change, as there are small changes in
the pH of the mobile phase. Of course, if the peaks are
separated by a mile, this is not relevant, but if they elute
close to each other, a small shift in pH may mean that
sometimes we obtain perfect resolution and another time the
peaks overlap. The consequence of this is that we need good
control over the pH of the mobile phase to obtain a good
reproducibility of retention and resolution.
However, this is not achievable with the mobile phase that
we are using. We have prepared the mobile phase with
KH2PO4, which has no buffering capacity whatsoever at pH
4.5. The pKa values of phosphate are at 2 and 7 If only a
small amount of acid or base are added to this mobile phase,

the pH changes drastically. Since the pH changes, the

retention of both of my analytes will change and the
resolution will change as well. With this set-up, we have
maximized the irreproducibility of the separation. Of course,
other problems are possible as well. It is highly likely that the
peak shape of one or even both analytes suffers as well,
especially if the amount of sample injected onto the column
is high.
Q.: OK, I agree that we have a problem. What can we do to
solve it?
A.: The simplest thing is to use a buffer with a good
buffering capacity. If the pH can be controlled such that it is
the same from day to day with every preparation of the
mobile phase, we have a much better chance of achieving
reproducible chromatography. We want to maintain the pH
to maintain the peak spacing. Therefore we need a buffer that
has a good buffer capacity around pH 4.5. A buffer has
always the best buffer capacity around its pKa. Acetic acid
has a pKa of 4.75. It is ideally suited for the separation
problem under discussion. It has a very good buffer capacity
at pH 4.5. If indeed the best separation and the best retention
is obtained at pH 4.5, we can use an acetate buffer at pH 4.5.
On the other hand, if the separation is still satisfactory at pH
4.75, it is even better to select this pH since the buffer
capacity of a buffer is always highest at the pKa of the buffer.
Q.: This is interesting. Can you provide a bit more
background on the buffer capacity?
A.: Yes, let me describe the theory behind this! The buffer
capacity is defined as follows:
=

2.303C

[H ]
(K a + [H + ])2
K a

(1)

C is the total concentration of the buffer, Ka is the

dissociation constant of the buffer, and [H+] is the hydrogen
ion concentration. In our common nomenclature of pKa and
pH, this can be rewritten to read:
10

pKapH

=2.303C

(2)

(10 +10 )
With this equation, we can calculate the buffer capacity of
any buffer. The buffer capacity curve for an acetate buffer is
shown in figure 1. The curve looks similar to a
chromatographic peak. One can see that the buffer capacity
is highest around the pKa of the buffer, at pH 4.75. It slowly
drops in both directions, and at pH 3.15 and 6.35 it is only at
1/10 of the value it had at the maximum. This is the reason
for the rule of thumb that says that a buffer should never be
used outside 1.5 pH units around its pKa, since it looses its
buffering capacity outside this range. We are commonly
using buffers with a concentration of 50 mM, and the rule
has been created for buffers of this concentration. However,
in LC/MS applications, the buffer concentration is often only
around 10 mM. This means that at such a low buffer
pKa

60

pH

concentration one may want to limit the rule of buffer

usefulness to approximately +/- 1 pH unit around the pKa of
the buffer.
You can use the equations shown here to calculate the
buffer capacity of other buffers, such as phosphate buffers or
citrate buffers. For buffers with multiple dissociation
constants, the buffer capacities of the different species can
simply be added up. However, the important point of this
discussion is the fact that we should choose buffers with a
good buffering capacity at the pH that we need to use to
optimize a separation.

detector. If I let a sample sit in the detector cell of a

concentration sensitive detector, the signal will remain
constant, at least within reasonable times.
Peak area is determined by integration of the signal over
time. This means in simple terms that if I let the same signal
be output by the detector for double the time, I get also
double the peak area. This is exactly what is happening when
you change the flow rate from 1 mL/min to 0.5 mL/min.
Here is the math. The peak area A is the signal S multiplied
by the time t:

Buffer Capacity of an Acetate Buffer

A=St

(1)

0.07

The signal S is proportional to the concentration c of the

analyte:

Buffer Capacity

0.06
0.05
0.04

S= pc

(2)

A= pct

(3)

0.03
0.02

The concentration is mass m divided by volume V:

0.01
0
0

10

12

14

A= p

pH

m
V

(4)

Figure 1: Buffer capacity of a 0.1 M acetate buffer.

And finally, volume divided by time is nothing but the flow
rate F:

51. Flow Rate Changes and Quantitation

A= pm1

Uwe D. Neue and Tony Gilby

(5)
F

Q.: My injector is doing something very peculiar that I do

not understand. It injects double the amount of sample at 0.5
mL/min than at 1 mL/min, and it does it reproducibly.
Whats happening?
A.: How do you come to the conclusion that this is the case?
Q.: Well, I put into the sample injection table an injection
volume of 5 L of sample. I run the method at flow rates of 1
mL/min and 0.5 mL/min, using a UV detector. The peak
areas that I obtain at 0.5 mL/min are nearly exactly double
the peak areas that I get at 1 mL/min.
A.: OK, now I understand. Here is what is happening: your
injector is working just fine, but the integration of the peak
area depends on the flow rate. Let me explain!
Nearly all classical HPLC detectors are designed to give an
output signal which measures the amount of sample in the
flowcell at a given moment in time - the number of analyte
molecules if you like. Since the cell volume is constant, this
is equivalent to measuring the average concentration in the
cell. A typical example of this is the UV absorbance

61

As we can see, the peak area is proportional to the mass

injected divided by the flow rate. This is the reason why you
get double the integrated signal at half the flow rate.
Q.: OK. I understand. You said that this holds for
concentration sensitive detectors. What does this mean?
Arent all detectors concentration sensitive?
A.: No, not all detectors are. In the case discussed above, the
signal remains the same, even if I change the flow rate. All
the classical LC detectors are of this type: photometric
detectors such as UV and UV-Vis absorbance or
photodiodearray
detectors,
fluorescence
detectors,
refractometers or detectors measuring the dielectric constant,
or conductivity detectors as used in ion chromatography.
Other rarely used detectors fall into the same category.
Examples are detectors that measure the optical rotation of a
sample.
A simple test can be carried out to show if one is dealing
with a true concentration sensitive detector or not: one
pumps an analyte carrying liquid into the detector, and then

stops the flow. If the signal remains constant, one is dealing

with a true concentration sensitive detector.

remains the same mass, and my balance gives me the same

signal.

Q.: I once made use of a radioactivity detector, and noticed

that if the flow stopped, the detector output kept increasing.
Is this different from the concentration sensitive detectors
you describe above?

Other mass proportional detectors are some GC detectors

such as the flame ionization detector (FID), which results in
complete destruction of the analyte molecules. The FID has
been used in LC through the use of specific interfaces, such
as the moving belt or moving wire interface. The use of GC
detectors in LC has remained rare though.

A.: No, this is still a concentration sensitive detector. The

counts per second are a measure of the number of radioactive
isotope atoms in the cell. What is confusing is that the
detector output, total counts, is actually the integral of the
count rate over time. The detector output is giving you
directly the integrated signal over the peak. At slower flow
rates, you will get a higher signal from the same peak, and
therefore a higher sensitivity. You can optimize the signal to
the needs of the analysis by stopping the flow at the time the
sample enters the detector.
Detection via post-column derivatization can be complicated
as well. The actual detectors used are concentration sensitive
detectors such as absorbance detectors or fluorescence
detectors. However, only for very fast reactions, the reaction
is driven to the endpoint. Most of the time, the reaction is
incomplete. Under these circumstances, the yield will depend
on the flow rate, with slower flow rates resulting in a longer
reaction time and consequently a higher yield and a higher
sensitivity.
An interesting case is the use of electrochemical detectors
such as amperometric detectors and coulometric detectors
(1). For amperometric detectors, the electrolysis of the
detected species is not complete. Only 1% to 10% of the
analyte are consumed. Therefore, the amperometric detector
works as a concentration sensitive detector. In coulometric
detection, all of the analyte is consumed in the detection
process. Thus, the total integrated signal is independent of
the flow rate. This is the opposite of what we had discussed
above for a UV detector. Coulometric detection therefore is a
mass-proportional detection technique.
The main characteristic of a true mass-proportional detector
is the fact that the integrated signal does not depend on the
time it took to acquire the signal. This is the most important
thing to remember about mass-proportional detection. Here
is an example. A simple mass proportional detector has been
reported in the early times of HPLC (2). The column effluent
was collected on a balance, the solvent was evaporated and
the residue was weighed. While this was interesting, it did
not prove to be practical. However, it clearly demonstrates
how a mass proportional detector works. In this case, the
height of the signal is the mass that has accumulated on the
balance. Thus, the signal is always an integrated signal, and
its height does not depend, if I transport the sample to the
detector quickly or slowly. If I inject the same mass into the
HPLC system at high flow rate or slow flow rate, it always

Today, mass spectrometers are often used in combination

with LC. Whether a mass spectrometer is a mass or
concentration proportional detector depends on the interface.
Current electrospray interfaces behave like concentration
sensitive detectors over the flow-rate range commonly used.
One might expect that the number of analyte ions measured
by the mass spectrometer per second (its signal, an ion
current) would increase if the number of analyte molecules
reaching the electrospray tip per second were increased - i.e.
by increasing the flow rate. This is not found in practice,
because of competition in the interface to ionize solvent
molecules, which are also arriving at a faster rate. Without
frequent injection of mass standards, mass spectrometers and
the associated HPLC interface, make unreliable quantitative
detectors.
I hope that this small excursion into detector response has
clarified the issues in quantification. The most important
thing to remember for the user of liquid chromatography is
the fact that most LC detectors are concentration sensitive
detectors, and that the integrated signal of concentration
sensitive detectors depends on the flow rate.
References:
1. C. F. Poole and S. K. Poole, Chromatography Today,
Elsevier, 1994, pp. 586 ff.
2. W.W. Schulz, W. H. King, Jr., J. Chromatogr. Sci. 11
(1973), 343

52. Analysis of polar compounds

Q.: I have a few very polar compounds that I need to
analyze. I have trouble with retention on standard C18
columns. What can I do?
A.: You are not alone. This question is appearing in ever
increasing frequency on bulletin boards discussing HPLC
problems. The good news is that there are a few solutions
available to this problem, and one or the other is likely to
work for your specific problem.
Q.: Oh good. I am glad to hear this. What are these
solutions?

62

A.: Overall, there are three possibilities that can be used to

get good retention for very polar compounds. The first one is
to switch from reversed-phase columns to HILIC

does not wet the surface, it is driven out of the all or part of
the pores of the packing. The surface becomes unwetted,
and retention is lost.

Q.: HILIC? What is that?

There are two ways in which a reversed-phase packing can

be made more compatible with an aqueous mobile phase.
The first one is the incorporation of a polar group into the
stationary phase, either built into the ligand or as a second
reaction. Typical functional groups for the phases with
incorporated polar groups are amide or carbamate groups
(5). These polar functional groups prevent the hydrophobic
collapse, but they also reduce somewhat the hydrophobicity
of the stationary phase as well. Therefore they give an
improvement of the retention properties of a packing, but
there is a still better solution.

A.:
HILIC stands for hydrophilic
interaction
chromatography, a term first used by Andy Alpert (1). It
works with polar stationary phases such as silica and mobile
phases with a high content of organic solvent and a smaller
content of water. The retention mechanism is the opposite of
reversed-phase. Polar compounds are retained more strongly,
and retention decreases when the water content of the mobile
phase increases. The best known example is the separation of
sugars on amino columns, a technique first published by Fred
Rabel and Art Caputo (2). In this case, the mobile phase
consists of about 75% acetonitrile and 25% water. The
underlying principle of this technique is the partitioning of
the analyte into a surface layer highly enriched with water
(3). Therefore, the more soluble the analyte is in water, the
more retention is observed. Conversely, the less soluble the
analyte is in mobile phase, the more retention is observed.
This technique can be enhanced further via other retention
mechanisms, such as ionic interactions or ion exchange (4).
One of the difficulties of this technique is the low solubility
of very polar samples in a mobile phase with a high
concentration of acetonitrile. However, it is not as bad you
would think. Among other things, you can dissolve the
sample in a solvent composition with a higher water content
than the mobile phase, or even in pure water. If you do that,
you just need to reduce the injection volume so that you do
not get peak distortions.

This second solution is simply a reduction of the

hydrophobic effect that causes the hydrophobic collapse. The
hydrophobic collapse is a wetting phenomenon. If I reduce
the ligand density of the C18 ligand, I can make a reversedphase packing that is still very hydrophobic, but is also water
wettable. This requires a good understanding of the
underlying properties of the packing, and the influence of the
different parameters on wettability and retention. Packings
have been prepared that deliberately balance hydrophobicity
with water wettability (6). A carefully designed reversedphase packing like this still gets good retention in fully
aqueous mobile phases without undergoing the hydrophobic
collapse.
Q.: OK, and what is the third option?

A.: The second possibility is the use of reversed-phase

columns that were specially designed for the retention of
very polar compounds. The issue with some of the best
standard reversed-phase columns is that they cannot be used
with mobile phases containing 100% water.

A.: The last option depends on the type of analytes that you
are using. If they are ionic compounds such as amines or
acids, they can be converted into a neutral form by changing
the pH of the mobile phase. Amines specifically can be
converted to a neutral uncharged form in the alkaline pH,
around pH 10. Under these circumstances, the
chromatographic retention increases by a large factor. Often
a 10- to 30-fold increase in retention is found by changing
the pH and changing the analyte from the ionic form to the
non-ionic form. Of course, you need a packing that has been
designed for use at high pH. Fortunately, packings of this
type are available today (7), and this procedure can be used
very effectively and without difficulty.

Q.: Exactly. When I tried to use a mobile phase with a very

high water content, I could not get reproducible retention,
and sometimes I got less retention than I thought I should
get.

Q.: Thank you for your description of the different options!

My compounds are very polar compounds with acidic and
basic groups. It appears that either one of these solutions
might work for me. Which one do you recommend?

A.: Yes, this is the problem that I was referring to. It is

sometimes called hydrophobic collapse, but a better
expression is dewetting of the pores. The modern very
hydrophobic and very deactivated phases loose retention in
highly aqueous mobile phases. The underlying issue is the
wettability of the stationary phase with water. Since water

A.: Without experimental details, this may be hard to predict.

I mentioned already the difficulty with the solubility of the
analyte in the mobile phase with HILIC. Among the
reversed-phase solutions, I would go with the solution that
has been optimized specifically for the purpose of achieving
high retention for the type of polar compounds that you need

In general, the important thing to remember is that HILIC

works best for very polar samples with a good solubility in
water.
Q.: OK, sounds good. What are the other options?

63

to separate. However, you also need to consider that any one

of these solutions can give you the improvement that you
need. In addition, the selectivity of the separation is likely to
be different between the different solutions. In most cases, it
might therefore be worthwhile to explore more than one tool.
References:
1. A. J. Alpert, J. Chromatogr. 499 (1990), 177-196
2. F. M. Rable, A. G. Caputo, E. T. Butts, J. Chromatogr.
126 (1976), 731-740
3. U. D. Neue, HPLC Columns, Wiley-VCH (1997)
4. M. A. Strege, S. Stevenson, S. M. Lawrence, Anal.
Chem 72 (2000), 4629-4633
5. U. D. Neue, Y.-F. Cheng, Z. Lu, B. A. Alden, P. C.
Iraneta, C. H. Phoebe, K. Tran, Chromatographia 54
(3/4) (2001), 169-177
6. D. M. Wagrowski-Diehl, E. S. Grumbach, P. C. Iraneta,
Development of New HPLC Columns for the Retention and
Separation
of
Highly
Polar
Compounds,
presentation at Pittsburgh Conference 2002
7. U. D. Neue, C. H. Phoebe, K. Tran, Y.-F. Cheng, Z. Lu,
J. Chromatogr. A 925 (2001), 49-67

64

Index
Subject
Alumina
Air bubble
Amino columns
Backflushing
Backpressure
Bandspreading
Baseline
Blood
Bonded phases
Buffer

Buffer capacity
Carbohydrates
Carbon load
Carryover
Clogged system
Column heater
Conditioning
Contamination
Cyano columns
Dead volume
Degassing
Delay volume
Detector
Diffusion coefficient
Double peaks
Drift
Dwell volume
Equilibration
Evaporation
Extra-column effects
Fast analysis
Filter
Flow rate

Ghost peaks
Gradient

Scaling
Ballistic gradients
Guard column
HETP
HILIC

Page
10, 11, 22, 30
6, 18, 26
4, 10, 11, 22, 30, 31, 40,
41
3, 48
14, 15, 16, 17, 18, 23,
34, 45, 46, 47, 48, 51
11, 12, 27, 28, 44, 48, 58
4, 7, 12, 13, 18, 25, 26
23
3, 6-11, 22-24, 30-33,
39-41, 48-50
4, 8, 14-17, 20, 23-26,
28, 29, 31-33, 36, 37,
41-44, 46, 49, 52, 53,
55, 56, 60, 61
20, 42, 60
22, 24, 31, 40, 41
34, 35
19
14
14
22-25, 53
5, 6, 26,38, 39, 41,45,
46, 53
10, 11, 31, 40
11, 27, 51
4, 56
12, 29, 30, 51, 58, 59
9, 11-13, 18-20, 25-28,
43, 44, 51, 57, 58, 61, 62
15, 16, 34
40, 41, 45
5, 6, 22
10, 12-14, 58, 59
4-6, 8, 10, 19, 21-22, 25,
30-32, 39, 59
4, 6, 22
9, 27, 28, 48
46
3, 14, 15, 17, 38, 46, 52
11, 15-18, 23, 26-27, 2930, 33, 37-38, 44, 47,
54-55, 57
18,19
6, 12-14, 18, 19, 21, 26,
29, 30, 37, 38, 51, 54,
55, 57-59
29-30
57-59
3, 14, 15, 17, 38, 39, 46,
48, 49, 51
16, 57, 58
63

65

Subject
Hydrolytic stability
Hydrolysis
Hydrophobic collapse
Impurities
Injection volume
Integration
Interference
Ion exchanger
Ion pairing
Knitted tubes
Lifetime
Liquid-liquid extraction
Method control
Method development
Method verification
MS detection
Narrow-bore columns
Noise
Normal Phase Chromatography
Overload
pH

pKa
Paired-ion chromatography
Peak area
Peak shape
Plasma
Plate count
Polar compounds
Post-column derivatization
Precipitation
Precolumn
Preconcentration
Protein
Recovery
Reproducibility
Batch-to-batch
Column-to-column
Retention
Variable
Drifting
Retention and pH
Sample preparation
Sample matrix
Sample solvent
Seals

Page
6, 31
3, 6, 22, 31, 33
6, 23, 24, 50, 63
19, 26, 41, 46, 49, 51
11, 18, 28, 29, 44
18
18, 23, 24, 28, 32, 38,
56, 58
49
31, 32
57, 58
32, 38-40, 45, 46, 48, 51
23, 28
41
29, 39, 40, 51, 52
39
55, 56
26
18, 20, 25, 26, 28, 43, 44
5, 10, 22, 30
9, 28, 44, 45
4, 6, 8, 20-22, 24, 25,
30-33, 36, 37, 41-43, 46,
52, 53, 55, 56
20, 25, 36, 37, 42-44,
56, 57, 60
31, 32
17-19
9, 29, 35-37, 42, 43, 48,
50, 56, 60
23, 24, 38, 46, 53, 54, 56
7, 15, 16, 27, 34, 45-48,
54, 55
62
57, 58
14, 19, 23, 26, 46
3, 14, 15, 26, 28, 38, 49
28
3, 8, 15, 18, 19, 23, 24,
38, 39, 46
7, 8, 23, 24
5, 6, 7, 10, 22, 24, 40
7, 24, 40
5, 6, 24
4, 22
4
5
20
3, 7, 8, 23, 28, 29, 38,
46, 53, 54
46, 52
28, 29, 45, 52
3, 14, 18, 45, 52

Subject
Selectivity
Sensitivity
Silanols
Silica
Size-exclusion chromatography
Solid-phase extraction
Solvent consumption
Solubility
Specifications
Storage (column)
Sugar separation
System volume
Tailing
Temperature
Time constant
van Deemter equation
Viscosity
Water saturated solvents

Page
7, 10, 21, 22, 37, 38, 41,
42, 49, 60, 64
28, 29, 43, 44, 51, 56, 62
6, 8, 9, 22, 24, 34-37,
49, 50, 53
4, 5, 10, 11, 16, 21, 22,
30, 32, 33, 35, 37, 50,
53, 56. 63
23
7, 8, 15, 24, 28, 38
26, 28, 44
3, 5, 32, 42, 45, 48, 63
6, 7, 22
5, 14, 23, 30-32, 57
40, 41
11, 12
8-10, 18, 32, 36, 37, 44,
50, 57
4, 6, 14, 17, 18, 22, 24,
25, 32, 41, 46, 58
9, 43
16, 33
14-17, 33, 34
5, 10

Copyright www.pharmaguideline.com