AAPS PharmSciTech 2006; 7 (3) Article 59 (http://www.aapspharmscitech.org).

Acoustic-resonance Spectrometry as a Process Analytical Technology for the

Quantification of Active Pharmaceutical Ingredient in Semi-solids
Submitted: January 24, 2006; Accepted: April 27, 2006; Published: July 14, 2006
Joseph Medendorp,1 Robert G. Buice, Jr,2 and Robert A. Lodder1
1
Department of Pharmaceutical Sciences, College of Pharmacy, A123 ASTeCC Building Lexington, KY 40536-0286
2
Chattem Inc, 1715 West 38th Street, Chattanooga, TN 37409

ABSTRACT and lotions.1 These recalls, resulting from mislabeling and

active pharmaceutical ingredient (API) concentration errors,
The purpose of this study was to demonstrate acoustic reso-
caused companies considerable unnecessary expense. A rap-
nance spectrometry (ARS) as an alternative process analyt-
id and accurate in-process assay, capable of testing and vali-
ical technology to near infrared (NIR) spectroscopy for the
dating individual samples, would obviate the need for the
quantification of active pharmaceutical ingredient (API) in
disposal of entire lots of problem products. The FDA pro-
semi-solids such as creams, gels, ointments, and lotions. The
cess analytical technology (PAT) initiative calls for the devel-
ARS used for this research was an inexpensive instrument
opment and implementation of manufacturing processes to
constructed from readily available parts. Acoustic-resonance
guarantee a predefined quality of pharmaceutical materials
spectra were collected with a frequency spectrum from 0 to
as warranted by risk analysis.2 These processes include mul-
22.05 KHz. NIR data were collected from 1100 to 2500 nm.
tivariate data acquisition and analysis tools, and in-process
Using 1-point net analyte signal (NAS) calibration, NIR for
and end point monitoring tools. The development of acoustic
the API (colloidal oatmeal [CO]) gave an r 2 prediction ac-
resonance spectrometry as an in-process and end point moni-
curacy of 0.971, and a standard error of performance (SEP)
toring device is in harmony with the PAT initiative.
of 0.517%CO. ARS for the API resulted in an r 2 of 0.983
and SEP of 0.317%CO. NAS calibration is compared with Current assays for analyzing API in various topical lotions are
principal component regression. This research demonstrates based on high-performance liquid chromatography (HPLC)
that ARS can sometimes outperform NIR spectrometry and with UV/visible detection,3,4 Fourier transform infrared (FTIR)
can be an effective analytical method for the quantification spectroscopy,5 and NIR spectroscopy.6 These conventional
of API in semi-solids. ARS requires no sample preparation, assays require both more sample preparation and more time
provides larger penetration depths into lotions than optical for each scan than ARS. Sometimes lotions do not exhibit
techniques, and measures API concentrations faster and more an optical chromophore,7 so they must be fixed with labeled
accurately. These results suggest that ARS is a useful process isotopes.4 In addition, optical methods provide poor penetra-
analytical technology (PAT). tion through semi-solids in comparison to acoustic methods.
Acoustic resonance spectrometry (ARS) is typically much
more rapid than HPLC and NIR. It is nondestructive and
KEYWORDS: creams, gels, lotions, net analyte signal, requires no sample preparation as the sampling waveguide
ointments, process analytical technologies (PAT), soundR can simply be pushed into the lotion. To date, the AR
spectrometer has successfully differentiated and quantified
sample analytes in various forms: tablets,8 powders,9-11 and
INTRODUCTION liquids.12-14 It has been used to measure and monitor the
Colloidal oatmeal (CO) is used as an ingredient in some phar- progression of chemical reactions, such as the setting and
maceutical lotions, cosmetics, and toiletries. While listed as hardening of concrete from cement paste to solid.15 Acoustic
inert in some lotions but active in others, CO is one of many spectrometry has also been used to measure the volume
examples in cosmetics and other household products that fraction of colloids in a dispersion medium, as well as for
fall into a gray area for the US Food and Drug Administra- the investigation of physical properties of colloidal disper-
tion (FDA). Between November 2002 and March 2004, sions, such as aggregation and particle size distribution.16,17
there were over 20 FDA recalls on household cosmetic prod- Typically, these experiments are performed with sinusoidal
ucts including shampoos, sprays, hand soaps, toothpastes, excitation signals and the experimental observation of sig-
nal attenuation. From a comparison of theoretical attenua-
tion to experimental observation, the particle size distribution
Corresponding Author: Robert A. Lodder, Department of and aggregation phenomena are inferred. In place of a sinu-
Pharmaceutical Sciences, College of Pharmacy, A123 soidal excitation signal sweeping across the desired fre-
ASTeCC Building, Lexington, KY 40536-0286. Tel: (859) quency range, this research makes use of broadband white
257-9232; E-mail: Lodder@uky.edu noise and standing resonance waves. To our knowledge, this
E1
 AAPS PharmSciTech 2006; 7 (3) Article 59 (http://www.aapspharmscitech.org).

research presents the first application of a broadband white

noise excitation signal to a resonant system for quantifica-
tion of colloidal particles in a dispersion medium.
It must be noted that a description of colloid aggregation
and particle-particle interaction is beyond the scope of this
research. Colloidal oatmeal is a lyophilic colloid, and it is
readily hydrated and dispersed evenly through a solution.
Therefore, the resonant acoustic signal received at the de-
tector is taken to be an approximate representation of the
bulk of the sample, regardless of microscopic differences
between individual colloids. As with other acoustic studies,
the individual colloid particles are considered to be spherical
and uniform, thus each particle produces a uniform effect on
the bulk physical properties of the surrounding medium.16 Figure 2. Frequency domain representation of the white noise
excitation signal source delivered from an MP3 player to the
Figure 1 provides a simple schematic of the AR spectrom-
PZT. All frequencies between 0 and 22.05 KHz are excited
eter for the following discussion. The AR spectrometer simultaneously.
used in this research was built in the near-field configura-
tion, where the wavelength of the excitation signal is much are excited simultaneously. However, the wave collected at
larger than the quartz rod or the sample that is applied to the detector is primarily composed of 3 large resonance struc-
the rod. An acoustic signal is applied to one of the piezo- tures as illustrated in Figure 3 (2.35-2.7 KHz, 9.6-11.2 KHz,
electric transducers (PZTs) and received at the other. The and 14.15-17.0 KHz). When a sample is placed in contact
sample, which is in mechanical contact with the vertex of with the vertex of the waveguide, acoustic waves escape and
the quartz rod, constitutes a load on the resonant system. propagate through the lotion/quartz interface and into the
With no sample in mechanical contact with the waveguide sample holder. The added mass effect causes a shift in the
connecting the PZTs, little excitation signal is lost and the resonant frequency of the system, while the frictional or vis-
acoustic signal received at the detector is the sum of an cous drag force causes a reduction in peak amplitude. This
ensemble of standing waves. The typical AR spectrum re- pattern gives rise to the characteristic AR spectrum for any
sults from the pattern of constructive and destructive inter- given analyte. In some cases, a third transmitting transducer
ference between the 2 sound paths; one that travels down beneath the sample may increase the analytical signal,8 al-
the quartz waveguide and through the sample and back on though it was not required in this research.
the way to the collecting PZT, and the other that stays in
the quartz rod and has no sample interaction. The excitation Because the propagation of an acoustic wave is based on
signal is a broadband white noise source as illustrated in longitudinal compressions and rarefactions of the medium
Figure 2, where all frequencies between 0 and 22.05 KHz through which the sound propagates, each analyte responds
differently to the frequency and amplitude of the applied

Figure 3. Three large resonance structures are prevalent for this

design of the ARS despite excitation with a broadband white
noise source. This figure illustrates AR scans of the different
Figure 1. A schematic diagram of the ARS instrument. colloidal oatmeal concentrations in lotions.

E2
 AAPS PharmSciTech 2006; 7 (3) Article 59 (http://www.aapspharmscitech.org).

acoustic signal. Acoustic spectra are the direct manifestation r¼ ⊥

k , and the signal from the analyte of interest r k , which is
of the physical differences between samples such as density, orthogonal to the contribution from the interferences.24 This
viscosity, acoustic velocity, the degree to which samples and orthogonal portion is termed the NAS and is the portion
analytes compress and expand in response to an applied of the signal used for multivariate calibration. A matrix
acoustic excitation, as well as depth of penetration, and the R (J
I) without the analyte of interest must be available,
contact pressure between the resonator and the sample.18-20 where J is the number of wavelengths and I is the number
These effects are seen in the acoustic spectra as a shift in the of samples. A projection matrix Pk⊥ can be calculated ac-
resonant frequency of the system, a reduction or an increase cording to Equation 1.
in the peak amplitude, or a mixture of both effects. Chemi-
cal changes in samples affect the physical properties mea-
Pk⊥ ¼ ðI−R−k Rþ
−k Þ ð1Þ
sured in acoustic spectra. Figure 4 illustrates the shifts in
resonance frequency and changes in peak height between where I is the identity matrix, and the ‘+’ superscript in-
acoustic spectra from 4 different analytes: water, air, metal, dicates the Moore-Penrose pseudo-inverse. Using 1-point
and lotion. The forces acting on the standing acoustic wave calibration with spectrum rcal, the NAS vector r ⊥cal can be
come from the inertial effect as the sample is moved by the calculated with Equation 2.
vibrating quartz rod, and from the dissipative effect owing
to the viscous drag force. The quartz rod was chosen as the
waveguide because of its high characteristic sound velocity r ⊥cal ¼ Pk⊥ rcal ð2Þ
and because quartz is an electrical insulator, which prevents
electromagnetic standing waves in the waveguide. The ve- This vector is then normalized to length 1 with Equation 3.
locity keeps acoustic impedance to a minimum according to
Zac = P/vA, where Z is acoustic impedance, P is sound pres- r ⊥cal
r NAS
k ¼ ð3Þ
sure, v is sound velocity, and A is cross sectional area.21,22 ‖r ⊥cal ‖

The slope of the calibration line is calculated from Equation 4.

Chemometrics
This research compared 2 different calibration methods, net ‖r ⊥cal ‖
analyte signal (NAS) and principal component regression s¼ ð4Þ
ccal
(PCR). PCR has been described previously.23 For NAS
calibration, a vector of instrumental responses rk is the sum where s is the slope and ccal is the analyte concentration of
of 2 independent signals, the signal from all interferences the calibration spectrum. The first step of NAS calibration

Figure 4. A zoomed view of the 3 large resonance peak structures considered in this study.

E3
 AAPS PharmSciTech 2006; 7 (3) Article 59 (http://www.aapspharmscitech.org).

includes using Equations 2 to 4 to find the NAS direction rectly proportional to the analyte concentration, or is a clas-
and determine the length of the NAS vector. When the NAS sification that represents the sample identity. Using ISP, little
direction and magnitude are known, the unknown spectrum or no postcollection data processing is required, making
run can be projected in the NAS direction with Equation 4 analyses significantly faster and perhaps even more ac-
and its magnitude compared with the calibration magnitude. curate. An imminent AR spectrometer, already under con-
struction and validation, employs ISP by encoding a special
excitation signal to resemble the spectral features required
y NAS T NAS
un ¼ r un r k ð5Þ
for the quantification of CO in lotion, so the signal received
at the detector is directly proportional to the analyte concen-
Equation 6 can now be used to calculate cun , the unknown
tration without further processing. ISP-ARS has great po-
analyte concentration.
tential as a PAT for sensing in the pharmaceutical industry.
The excitation can be specifically tailored to meet the needs
1
cun ¼ y NAS ð6Þ of demanding analytical challenges. For example, if the need
s un is to differentiate between 1% and 5% CO creams, an ISP
waveform created for this purpose could be easily down-
The NAS approach allows for the calculation of figures of loaded from an online database in the form of an .MP3 file.
merit from multivariate data sets. In severely overlapping The hardware of the ISP-ARS comprises inexpensive com-
spectra, it has historically been difficult to quantify selec- mercial off-the-shelf (COTS) components, simplifying its
tivity, sensitivity, and signal-to-noise (S/N) because of the manufacture and deployment.
inability to distinguish between interferences and the ana-
lyte of interest.25,26 With the NAS, these quantities can be
measured directly. Selectivity is defined as the scalar degree
MATERIALS AND METHODS
of overlap, α, between the NAS vector and the calibration
spectrum according to Equation 7. Sample Preparation
CO samples were prepared using Gold Bond Sensitive Lo-
‖r ⊥cal ‖ tion (Lot No. 03518, Chattern Inc, Chattanooga, TN). Lotion
α¼ ð7Þ was weighed into an 800-mL beaker. Five concentrations
‖rcal ‖
(2.0%, 2.5%, 3.0%, 4.0%, and 5.0%) of CO were prepared by
gravimetric addition of CO to the lotion samples for a total
The selectivity is a measure from 0 to 1 indicating how
sample mass of 350 g. Samples were heated to 50-C on a
unique the analyte of interest is compared with the inter-
hot plate, while mixing with a paddle blade. When lotions
ferences. The sensitivity is a measure of the analyte variation
reached 50-C, the CO (Vendor lot No. 22915, Chattern Inc,
in response to a change in concentration. This quantity can
Chattanooga, TN) was added and the sample was mixed
be expressed as Equation 8.
(Heidolph RZR 50, Frankfort, Germany) at a sufficient rpm
to create a gentle vortex. The heat was then turned off and
sk ¼ r NAS
k =ck ; ð8Þ samples were mixed until they cooled to 35-C.

where ck is the concentration of the k-th analyte. Sensitivity

should be the same for each concentration and each NAS Acoustic Resonance Spectrometry Data Collection
vector.27 The S/N ratio can be expressed as Equation 9. Rubber grommets were positioned on the rod, so that they
firmly held it in place but dampened the signal minimally.
ck ‖r NAS
k ‖
Previous investigations with a similar ARS model made
S=N ¼ ð9Þ use of a third transducer beneath the vertex of the rod, which
‖ε‖
acted as an interferometer.29 Our investigation indicated that
the interference pattern inherent to the 2 sound paths and its
where ε is the random instrumental error.
effects on the resonance frequency and peak amplitudes was
sufficient to negate the need for an additional transducer. A
Integrated Sensing and Processing Acoustic Resonance white noise signal played through an ordinary MP3 player
Spectrometry for Process Analytical Technology was used as the source of random noise across the range
Integrated sensing and processing (ISP) is a paradigm for from 0 to 22.05 KHz. The ARS instrument was designed to
instruments in which large physical fields of data are re- output a voltage signal to a computer sound card (Realtek
duced to high-level information as the data are sensed, be- AC97, Realtek Semiconductor Japan Corp, Yokohama,
fore the data are passed to a computer.28 In many cases, the Kanagawa, Japan), operated with a graphic user interface
signal transmitted from the detector to the computer is di- in Matlab 7.0.1 (The Mathworks Co, Natick, MA). Lotion
E4
 AAPS PharmSciTech 2006; 7 (3) Article 59 (http://www.aapspharmscitech.org).

samples were placed in a small plastic cup and the vertex of and ARS. NIR for the measurement of CO gave an r 2 pre-
the quartz rod was plunged into the cup, so the 2 were in diction accuracy of 0.971, and a standard error of perfor-
mechanical contact for the duration of the scan. The pene- mance (SEP) of 0.517%CO. ARS for the API resulted in an
tration depth of the rod into the lotion was kept constant at average r 2 = 0.983, calculated by averaging the r 2 from the
5 mm. All data were collected for 3 seconds at a sample rate of 3 resonance peaks, and an SEP of 0.317%CO. Each NIR
44.1 KHz. Data were transformed from the time domain into scan took almost 90 seconds to complete (the long inte-
the frequency domain by a fast Fourier transform (FFT), gration time was selected to get an S/N competitive with
resulting in a signal-to-noise ratio of 110/1. Because the data ARS), while each ARS scan required only a maximum of
set dimensions are a product of the sample rate and duration 3 seconds to achieve similar sample predictive ability.
of the collection, each spectrum collected with these instru-
ment parameters was 132 300 data points long. A 20-point
moving boxcar average was performed on the frequency Acoustic Resonance Spectrometry Detection Limits
domain data as a smoothing function. In its current con- ARS spectra collected at n frequencies can be reduced to
figuration, there are significant resonance peaks at 2.35 to a single point in an n-dimensional hyperspace using the
2.70 KHz, 9.6 to 11.2 KHz, and from 14.5 to 17.0 KHz. All BEST. The resulting points usually form complex Lissajous
data analysis was performed on these 3 frequency regions patterns (or curved portions thereof) in hyperspace as ana-
for the duration of the experiment. lyte concentrations change because of the nonlinear nature
of acoustic interactions.33 BEST is a nonparametric cluster
analysis algorithm based on the premise that spectra from
Near Infrared Data Collection similar samples tend to cluster in the same region of multi-
A layer of lotion ~1-mm thick was spread on a single-well dimensional hyperspace.34 In order to demonstrate clus-
depression microscope slide (Gold Seal Products, Portsmouth, tering quantitatively, multidimensional standard deviations
NH). NIR scans were collected with a scanning monochro- (MSDs) can be used to measure the separation between
mator NIR instrument described previously.30 Scans were clusters of different samples.29,32,35 Using the BEST, inter-
collected from samples inside light-proof chamber in a dark- cluster MSDs greater than 3 indicate distinct cluster popula-
ened room to eliminate stray light interference. Samples were tions, while clusters less than 3 MSDs apart are inseparable
scanned in random order to eliminate the effects of drift over with statistical significance. A multidimensional transla-
time on the results. All data were exported to Matlab 7.0.1 for tion operation can be performed with this algorithm to es-
all processing. timate the theoretical detection limits of the instrument.32
The translation operation is based on the principle that clus-
ters from pure component spectra form distinct and separate
Data Analysis populations, and a translation of one toward the other cor-
The objective of this experiment was to test ARS in lotion responds to a mixture of the 2 components when Beer’s
analysis and compare the results with NIR spectrometry for Law holds.
the measurement of concentration change of API in lotion. To estimate the theoretical detection limits for dynamic range
NIR data were multiplicative scatter-corrected to eliminate calculations, intragroup spectra from the 2 lotion populations
baseline variations caused by path-length differences.31 One- (P1 and P2) were used as m x n matrices.32 The columns
point NAS calibration was compared with PCR23 for the of the matrices were averaged by Equations 10 and 11,
measurement of CO from ARS and NIR spectra. Detec- giving two 1
n vectors.
tion limits were estimated by a bootstrap error-adjusted
single sample technique (BEST) multidimensional popula-
tion translation of 1 cluster toward another.32 The clusters 1 i¼m
P1 ¼ ∑ P ð10Þ
used in the translation were unprocessed AR FFT spectra m i¼1 1i
and PC scores. Other figures of merit such as selectivity,
sensitivity, and signal-to-noise ratio were calculated using
the NAS. All algorithms were written by the authors.
1 i¼m
P2 ¼ ∑ P ð11Þ
m i¼1 2i
RESULTS AND DISCUSSION
A difference spectrum X was calculated from P 2 −P 1 . One
Acoustic Resonance Spectrometry Versus Near
population was spatially translated toward the other,
Infrared Spectrometry
PA ¼ yX þ P2 , where y (defined on the interval [0 G y ≤ 1])
The results from the calibration models are summarized in started at zero and increased in increments of 0.01 until
Table 1. NAS calibration outperformed PCR for both NIR P1 and PA were inseparable by the BEST metric. The
E5
 AAPS PharmSciTech 2006; 7 (3) Article 59 (http://www.aapspharmscitech.org).

Table 1. Three Seconds of Data Collection Gave the Highest r 2 and the Lowest RMSEP for Each Frequency Range and Both
Chemometric Models*
Method NAS PCR
2 2
ARS Frequency (KHz) Analysis Time (s) r RMSEP (% CO) r RMSEP (% CO)
2.35-2.70 1 0.892 1.532 0.926 0.306
2 0.985 0.787 0.943 0.260
3 0.991 0.504 0.946 0.265
9.60-11.20 1 0.962 0.491 0.855 0.427
2 0.973 0.299 0.879 0.391
3 0.983 0.251 0.948 0.255
14.5-17.0 1 0.968 0.398 0.813 0.511
2 0.979 0.262 0.816 0.495
3 0.991 0.180 0.855 0.488

NIR 1100-2500 nm 120 0.971 0.517 0.956 0.314

*RMSEP indicates root mean squared error of prediction; NAS, net analyte signal; PCR, principal component regression; CO, colloidal oatmeal;
ARS, acoustic resonance spectrometry; and NIR, near infrared.

detection limits were estimated by the BEST multidimen- and performance. If the experimenter can spend the time to
sional translation experiment using raw AR spectra in the collect only 1s of data, the process is much faster but the
selected frequency regions and PC scores calculated from predictive ability of the instrument suffers slightly.
the selected frequency regions. The limits were calculated
using translation of 2 clusters, the translation from 2%CO
to 2.5%CO, and the translation from 2%CO to 5%CO; all Selectivity, Signal to Noise Ratio, and Sensitivity
results are summarized in Table 2. The best detection limits of Acoustic Resonance Spectrometry
were seen using PC scores rather than raw AR spectra. This Estimation of figures of merit is a straightforward task when
is an expected result as PC scores are inherently the repre- the measured response is a scalar, such as in zero-order
sentation of the largest difference between spectral groups. calibration methods. The measured response in zero-order
Therefore, separation has been optimized prior to estimation methods is simply the contribution of the analyte plus a
of the detection limits by PC scores. constant instrumental background or constant interference.
However, in first-order data, where the measured response is
a vector of scalars, the total instrument response is not suf-
Speed of Acoustic Resonance Spectrometry Method ficient to determine the figures of merit. The presence of
To assess the maximum speed at which the ARS is still multiple chemical and instrumental interferences varying
capable of quantifying CO, the FFT was calculated for simultaneously requires an extra step for multivariate cali-
progressively shorter blocks of time rather than calculating bration. Therefore, the portion of the signal orthogonal to
the FFT from the time domain data in its entirety. For ex- the interfering chemical species and background, termed the
ample, the FT was calculated for 3 seconds, 2 seconds, and NAS, must be determined prior to calculation of figures
1 second. NAS calibration and a PC regression were per- of merit.24-27 Because the ARS functions by the interference
formed on the FFT. The speed of method results are pre- between the analyte acoustic signal and a standing wave,
sented in Table 1. These results suggest that longer periods there is usually a substantial instrument response even in the
of data collection offer an advantage in quantification, though absence of a sample. For this reason, calculation of the NAS
it is not necessary to collect more than 1 second for a high with acoustic data tends to behave much like an additional
predictive ability. There is a trade-off between scan time smoothing function. The NAS and figures of merit were

Table 2. Translations Were Performed With the 2 closest Concentrations and the 2 Most Distant Concentrations*
Raw AR Spectra PC Scores (4 PCs)
(Detection Limit %CO) (Detection Limit %CO)
Frequency (KHz) 2% → 2.5% 2% → 5.0% 2% → 2.5% 2% → 5.0%
2.35-2.70 0.210 1.26 0.100 0.150
9.0-11.20 0.175 1.23 0.050 0.210
14.5-17.0 0.235 1.41 0.160 0.090
*BEST indicates bootstrap error-adjusted single sample technique; AR, acoustic resonance; PC, principal component; and CO, colloidal oatmeal.

E6
 AAPS PharmSciTech 2006; 7 (3) Article 59 (http://www.aapspharmscitech.org).

calculated according to Equations 7 to 9. Selectivity is a (2) To assess direct RF cross-talk, a PZT was suspended
unit-less scalar ratio, indicative of the degree of overlap 10 cm from the epoxy-fastened receiving PZT. A white noise
between the 2 vectors and was calculated to be 0.5963. signal was generated with the suspended PZT and collected
The S/N was 110/1. The sensitivity of the ARS for CO in with the epoxy-fastened PZT. No signal was recorded with
lotion is 0.230, given in instrumental response per unit the sound card, suggesting that when the PZTs are fastened
concentration change. with epoxy to the quartz, they do not move unless driven by
the excitation signal through mechanical contact. (3) More
than 90 lotion samples have been scanned without an anom-
Versatility and Flexibility alous spectrum owing to manual loading. Because the posi-
tion of the quartz rod is regulated by a vertical translation
In addition to being rapid, effective, and inexpensive, the
stage, the depth of penetration into the lotion is consistent,
ARS has also proven to be an extremely versatile instru-
therefore spectra are reproducible, demonstrating the dura-
ment. The same instrument and chemometrics can be ap-
bility of the ARS.
plied to multiple analytes including tablets,8 powders,9-11
and liquids.12-14 Because the system is set up such that a
standing wave forms in the quartz rod in the absence of a CONCLUSION
sample, the addition of an analyte to the vertex of the rod
acts as a load on the resonance of the system and nec- This research demonstrates that ARS is faster, less expen-
essarily disrupts the path length of sound in the waveguide. sive, and outperforms NIR spectroscopy for the quantification
While the excitation frequencies passing through the rod of CO in lotion. Therefore, it has demonstrated its potential
and the analyte are the same, because of the large differ- as a PAT for the quantification of active pharmaceutical
ences in their velocities, the wavelengths are usually longer ingredient in semi-solids.
in the quartz than in the sample. As a result, regardless of
the colloid size, shape, and concentration, the waveforms ACKNOWLEDGMENTS
from the waveguide and the sample will never be perfectly
in-phase as they recombine at the vertex of the rod. This This research was supported in part by the US FDA through
phase change will result in a characteristic pattern of con- contract number 200406251503, Science and Engineer-
structive and deconstructive interferences for every analyte ing Services, Inc. (SESI) and the National Institutes of
because of differences in physical properties. Health (NIH) through contract number N01AA33003 and
the Kentucky Science and Engineering Foundation (KSEF)
through grant number 148-502-03-61. Special thanks to Pa-
tricia Lee and Christina Colquitt for the colloidal oatmeal
Freedom From Interferences
formulations.
The main sources of interferences for the ARS in this con-
figuration are (1) acoustic waves propagating through the
support structure of the instrument, (2) radiofrequency (RF) REFERENCES
cross-talk directly between PZTs, and (3) inconsistent ana- 1. NIH NLM Specialized Information Services. Food and Drug
lyte and quartz rod alignment because of a manual loading Administration Actions: Recalls and Field Corrections 2002-May 12,
2004. National Institutes of Health Web site. Available at: http://
procedure. Each of these points is addressed in the follow-
householdproducts.nlm.nih.gov/NLM_FDARecalls.htm#rayblock1.
ing discussion. (1) The quartz rod was mounted on a frame Accessed: April 1, 2005.
of wood, which is made up of a cellular network of pores 2. USDA. Process Analytical Technology (PAT) Initiative. US Food and
that convert sound energy into heat by frictional and visco- Drug Administration Web site. Available at: http://www.fda.gov/cder/
elastic resistance.36 The cellular pore network creates high OPS/PAT.htm. Accessed: April 1, 2005.
internal friction, thus wood has more sound-dampening ca- 3. Mariani E, Villa C, Neuhoff C, Dorato S. Derivatization procedure
pacity than most structural materials. Because the mounting and HPLC determination of 2-ethoxyethanol in cosmetic samples.
framework is constructed from wood, there is less sound Int J Cosmet Sci. 1999;21:199Y205.
traveling through the beams, and any sound that may get 4. Westgate E, Sherma J. Determination of the sunscreen oxybenzone
through is dampened owing to the placement of the rubber in lotions by reversed-phase HPTLC with ultraviolet absorption
densitometry. J Liq Chromatogr Relat Technol. 2000;23:609Y615.
grommets holding the quartz rod in place. It should be
noted that the wood construction for the purposes of this 5. Sabo M, Gross J, Rosenberg I. Quantitation of dimethicone in lotions
using Fourier transform infrared spectral subtraction. J Soc Cosmet
experiment was the low cost alternative to other potential Chem. 1984;35:273Y281.
materials. For example, if this instrument were needed in a
6. Grunewald H, Kurowski C, Timm D, Grummisch U, Meyhack U.
current good manufacturing practices (cGMP) environment Rapid non-destructive raw material identification in the cosmetic
for PAT, it could just as easily be constructed from other industry with near-infrared spectroscopy. J Near Infrared Spectrosc.
acoustic dampening materials such as concrete and rubber. 1998;6:215Y222.

E7
 AAPS PharmSciTech 2006; 7 (3) Article 59 (http://www.aapspharmscitech.org).
7. Alltech Associates. Dimethicone. Alltech Associates Inc Web site. 21. Mills TP, Jones A, Lodder RA. Identification of wood species by
Available at: http://www.alltechweb.com/productinfo/technical/app/ acoustic-resonance spectrometry using multivariate subpopulation
0048E.pdf. Accessed: April 15, 2005. analysis. Appl Spectrosc. 1993;47:1880Y1886.
8. Buice R, Pinkston P, Lodder R. Optimization of acoustic-resonance 22. Lai E, Chan B, Chen S. Ultrasonic resonance spectroscopic analysis
spectrometry for analysis of intact tablets and prediction of dissolution of microliters of liquids. Appl Spectrosc. 1988;42:526Y529.
rate. Appl Spectrosc. 1994;48:517Y524. 23. Medendorp J, Yedluri J, Hammell DC, Ji T, Lodder RA, Stinchcomb
9. Serris E, Perier-Camby L, Thomas G, Desfontaines M, Fantozzi G. AL. Near infrared spectrometry for the quantification of dermal
Acoustic emission of pharmaceutical powders during compression. absorption of econazole nitrate and 4-cyanophenol. Pharm Res.
Powder Technol. 2002;128:296Y299. 24. Boelens H, Kok W, Noord O, Smilde A. Performance optimization
10. Reynaud P, Dubois J, Rouby D, Fantozzi G. Acoustic emission of spectroscopic process analyzers. Anal Chem. 2004;76:2656Y2663.
monitoring of uniaxial pressing of ceramic powders. Ceramics Int. 25. Lorber A. Error propagation and figures of merit for quantification
1992;18:391Y397. by solving matrix equations. Anal Chem. 1986;58:1167Y1172.
11. Martin LP, Poret JC, Danon A, Rosen M. Effect of adsorbed water 26. Booksh KS, Kowalski BR. Theory of analytical chemistry.
on the ultrasonic velocity in alumina powder compacts. Mater Sci Anal Chem. 1994;66:782aY791a.
Eng A. 1998;252:27Y35.
27. Lorber A, Faber K, Kowalski BR. Net analyte signal calculation
12. Kaatze U, Wehrmann B, Pottel R. Acoustical absorption spectroscopy in multivariate calibration. Anal Chem. 1997;69:1620Y1626.
of liquids between 0.15 and 3000 MHz. I. High resolution ultrasonic
28. Medendorp JP, Lodder RA. Applications of integrated sensing
resonator method. J Phys [E]. 1987;20:1025Y1030.
and processing in spectroscopic imaging and sensing. J Chemometr.
13. Bolotnikov M, Neruchev Y. Speed of sound of hexane + 2006;19:533Y542.
1-chlorohexane, hexane + 1-iodohexane, and 1-chlorohexane +
29. Buice R, Lodder RA. Determination of cholesterol using a novel
1-iodohexane at saturation condition. J Chem Eng Data.
magnetohydrodynamic acoustic-resonance near-IR (MARNIR)
2003;48:411Y415.
spectrometer. Appl Spectrosc. 1993;47:887Y890.
14. Lévêque G, Ferrandis J, Van Est J, Cros B. An acoustic sensor for
30. Fountain W, Dumstorf K, Lowell AE, Lodder RA, Mumper RJ.
simultaneous density and viscosity measurements in liquids. Rev Sci
Near-infrared spectroscopy for the determination of testosterone in
Instrum. 2000;71:1433Y1440.
thin-film composites. J Pharm Biomed Anal. 2003;33:181Y189.
15. Ferrandis JY, Leveque G. In situ measurement of elastic properties
of cement by an ultrasonic resonant sensor. Cement Concrete Res. 31. Geladi P, MacDougall D, Martens H. Linearization and scatter-correction
2003;33:1183Y1187. for NIR reflectance spectra of meat. Appl Spectrosc. 1985;39:491Y500.

16. Dukhin AS, Goetz PJ. Acoustic spectroscopy for concentrated 32. Lodder R, Hieftje G. Detection of subpopulations in near-infrared
polydisperse colloids with high density contrast. Langmuir. 1996;12: reflectance analysis. Appl Spectrosc. 1988;42:1500Y1512.
4987Y4997. 33. Lodder R. CD/MP3 Acoustic Resonance Spectrometer. Analytical
17. Dukhin AS, Goetz PJ. Characterization of aggregation phenomena Spectroscopy Research Group Web site. Available at: http://www.pharm.
by means of acoustic and electroacoustic spectroscopy. Colloids Surf A: uky.edu. Accessed: April 5, 2004.
Physicochem Eng Aspects. 1998;144:49Y58. 34. Jolliffe IT. Principal Component Analysis. New York, NY: Springer;
18. Ramdani A, Cros B, Sidki M, Ferrandis J. Acoustic near field 2002.
technique for characterization of liquids, bitumen and cement setting. 35. Hamilton S, Lodder R. Hyperspectral imaging technology for
Eur Phys J AP. 2001;15:69Y76. pharmaceutical analysis. Proc Soc Photo-Opt Instrum Eng. 2002;
19. Patois R, Vairac P, Cretin B. Near-field acoustic densimeter and 4626:136Y147.
viscosimeter. Rev Sci Instrum. 2000;71:3860Y3863. 36. Bear Creek Lumber. Western Red Cedar Physical Properties.
20. Prugne Ch, Van Est J, Cros B, Leveque G, Attal J. Measurement Bear Creek Lumber Web site. Available at: http://www.bearcreeklumber.
of the viscosity of liquids by near-field acoustics. Meas Sci Technol. com/generalinfo/onlineliterature/technicalinfohtml/wrcphysicalproperties.
1998;9:1894Y1898. html. Accessed: March 1, 2005.

E8