Journal of Molecular Catalysis A: Chemical

Download as pdf or txt
Download as pdf or txt
You are on page 1of 27

Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

Contents lists available at ScienceDirect

Journal of Molecular Catalysis A: Chemical


journal homepage: www.elsevier.com/locate/molcata

Temporal analysis of products (TAP)—Recent advances in technology for kinetic


analysis of multi-component catalysts
John T. Gleaves a , Gregory Yablonsky a,b,∗ , Xiaolin Zheng a , Rebecca Fushimi a , Patrick L. Mills c
a
Department of Energy, Environmental, and Chemical Engineering, Washington University in St. Louis, 1 Brookings Drive, St. Louis, MO 63130, USA
b
Parks College of Engineering, Department of Chemistry, Saint Louis University, 3450 Lindell Blvd, St. Louis, MO 63103, USA
c
Department of Chemical and Natural Gas Engineering, Frank H. Dotterweich College of Engineering, Texas A&M University-Kingsville,
700 University Blvd, MSC 188, Kingsville, TX 78363-8202, USA

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents an overview of the evolution, advancements, and capabilities of the temporal anal-
Available online 26 June 2009 ysis of products (TAP) reactor system as a unique catalyst characterization tool. The origination of the
TAP reactor based on molecular beam scattering experiments is briefly mentioned. The advancement in
Keywords: TAP reactor design from the TAP-1 system to the TAP-3 system is introduced to highlight its relevance
Temporal analysis of products (TAP) as a valuable tool for elucidating mechanistic and kinetic aspects of adsorption, diffusion, and reaction
Heterogeneous catalysis
in gas–solid systems. Since the invention of the TAP reactor system, a series of TAP microreactor con-
Kinetics
figurations has been introduced with different amounts of catalyst packing starting from the one-zone
Transient methods
Gas–solid reactions
microreactor to the most recent introduction, the single particle microreactor in which a single Pt parti-
Knudsen diffusion cle is packed among 100,000 inert quartz particles. An advantage to decreasing the catalyst zone inside
Time-of-flight mass spectrometer the microreactor is to eliminate non-uniformity in the active zone while still achieving high conversions
Atomic beam deposition (95%). Experimental designs and results coupling the TAP reactor to other experimental systems such
Single particle experiments as a time-of-flight mass spectrometer and atomic beam deposition system is also presented. Key results
Pressure gap from recent TAP experiments are presented to show how the TAP reactor is used to answer fundamental
questions in catalysis such as bridging the pressure gap between industrial catalysis and surface science,
understanding the surface lifetimes of reactive adspecies in TAP pump-probe experiments, finding kinetic
rate constants related to changes in catalyst composition and its performance.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction and chemicals when oil is no longer readily available is one of the
most challenging and important problems now facing humanity. An
Over the next several decades, a new generation of catalytic essential piece of the puzzle is the development of new approaches
materials and processes will be needed to meet the increasing to unravel the tangle of processes that occur on complex industrial
demands for fuel, food, chemicals, pharmaceuticals, and a whole catalysts.
host of materials (e.g., plastics and fibers). New catalysts will also Despite advances in surface science and computational chem-
play a pivotal role in environmental remediation and halting global istry that provide molecular level insight into surface physics and
warming. Currently, the petroleum-based process industry gen- chemistry, catalyst development for realistic processes still relies
erates products that have an annual worldwide production value extensively on empirical methods and trial and error processes.
exceeding 4 trillion dollars. Heterogeneous catalysis and multi- Generally the method used to develop practical catalysts follows
phase reaction engineering are the technology engines that power the catalyst development cycle (CDC) illustrated in Fig. 1. In the first
the chemical and fuel industries since nearly 75% of all chemicals, step of the CDC, candidate materials are selected to test as catalysts.
polymers, and advanced materials are produced with the aid of cat- In step two, the materials are synthesized, and in steps three and
alysts. In addition, over 90% of newly developed processes involve four, the materials are tested for catalytic activity, and characterized
catalysis, and over 95% of industrial reactors are based on heteroge- structurally.
neous catalytic processes. How to supply the vast quantities of fuels, The CDC includes a decision branch, which is reached through
steps three and four. This branch gives rise to a number of cyclic
paths. The path depicted by the red arrows represents a key cycle
in the CDC, and provides crucial performance information used to
∗ Corresponding author at: Department of Energy, Environmental, and Chemical
determine if the cycle can be exited. In practice, however, infor-
Engineering, Washington University in St. Louis, 1 Brookings Drive, St. Louis, MO
63130, USA. mation gained in a single cycle of the CDC often does not provide
E-mail address: gregoryyablonsky@yahoo.com (G. Yablonsky). guidance for the next step. In many cases, a cycle only provides

1381-1169/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.molcata.2009.06.017
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 109

Nomenclature

as surface concentration of active sites (mol/cm2 of cat-


alyst)
A cross-sectional area of the microreactor (cm2 )
˛ ratio of total number of active centers and number
of molecules of gas A in the pulse
CA concentration of gas A (mol/cm3 )
CA dimensionless concentration of gas A
De effective Knudsen diffusivity (cm2 /s)
DeA effective Knudsen diffusivity of gas A (cm2 /s)
ız delta function with respect to axial coordinate z
ı delta function with respect to dimensionless axial
coordinate 
Fig. 1. Catalyst development cycle.
εb fractional voidage of the packed bed inside microre-
actor
FA flow of gas A at microreactor outlet (mol/s) placed on rapid screening of many different catalyst compositions
FA dimensionless flow of gas A at microreactor outlet to determine promising candidates that can be further developed.
Hp peak height of the normalized gas exit flow The combinatorial catalyst or so-called high-throughput approach
ka adsorption rate constant (cm3 /mol s) represents the most advanced example of this strategy [1–4].
k a adsorption rate constant (s−1 ) The second strategy is focused on making the CDC more efficient
ka dimensionless adsorption rate constant by increasing the quantity and quality of information obtained in
kads apparent adsorption constant (s−1 ) each cycle. Information that links changes in kinetic properties to
kd desorption rate constant (s−1 ) changes in the structure of a catalyst sample is very valuable. An
kd dimensionless desorption rate constant example of the second strategy is the use of surface science tech-
ka dimensionless modified kinetic constant of adsorp- niques to characterize different catalyst samples after reaction to
tion determine if differences in surface composition and catalyst per-
L length of the microreactor (cm) formance can be related.
LII length of diffusion zone II (cm) Understanding how the catalytic activity and selectivity of a
L thickness of the catalytic zone (cm) substance is related from the composition and structure of the sub-
M molecular mass stance is a fundamental problem in the field of catalysis. Solving
NpA number of moles or molecules of gas A in the input the problem involves finding relationships between kinetic data
pulse and structural data. Relationships that provide a link between the
R universal gas constant reaction kinetics and the structure of the “active catalytic site” are
Sv surface area of catalyst per volume of catalyst (cm-1 ) particularly important because they provide essential information
t time (s) for explaining how catalytic systems work, and for creating new or
tp time at which the gas exit flow is at a maximum (s) improved catalysts.
 dimensionless time To develop an activity–structure relationship for a gas–solid cat-
p dimensionless time at which gas exit flow is at a alytic system, it is necessary to obtain kinetic data that can be
maximum directly related to the structure/composition of the solid catalyst,
conv
res residence time for convective flow in CSTR (s) especially the catalyst’s surface. For reactions involving technical
dif
res,cat residence time for diffusion through the catalytic catalysts (e.g., supported metals or mixed-metal oxides), acquiring
zone in thin-zone TAP microreactor (s) such kinetic data presents a formidable challenge since the reaction
T temperature (K, ◦ C) kinetics and the catalysts are typically very complex. Moreover, in
A fractional surface coverage many cases, and particularly in the case of mixed-metal oxides, the
A pulse normalized surface concentration structure/composition of the catalyst is influenced by the reaction
Vvoid void volume inside microreactor mixture and can undergo changes as a function of time-on-stream.
(cm3 gas/cm3 catalyst) This has obvious implications during extended studies involving
X conversion investigations on kinetics and mechanism.
z axial coordinate (cm) Kinetic testing of technical catalysts is normally performed
 dimensionless axial coordinate at industrial conditions. It is common to prepare a series of
catalyst samples with different compositions and/or structures,
and then test the samples at the same reaction conditions to
determine differences in catalytic activity. The samples are also
negative information such as a particular catalyst synthesis leads characterized using one or more structural techniques. To obtain
to an inactive or non-selective candidate. Generally, a large number an activity–structure relationship, the differences in activity are
of iterations, often involving the preparation of perhaps thousands matched against the structural information. Sometimes, a well-
or tens-of-thousands of catalyst samples, precede the finalization characterized standard catalyst is used for comparison purposes.
of a new catalyst formulation. The goal of this procedure is to discover a correspondence between
Two general strategies have been used in an effort to acceler- the measured activity and some structural characteristic. One
ate the rate of catalyst invention when fundamental guidance or obstacle to developing an activity–structure relationship using this
previous data is either absent or has failed to produce a viable cata- methodology is that the measured kinetics generally reflects the
lyst candidate. One strategy is focused on decreasing the CDC cycle global characteristics of the catalytic process rather than the local
time by decreasing the time required to synthesize and evaluate characteristics of the active catalytic sites. Also, during reaction, a
the performance of new catalyst samples. In this case, emphasis is catalyst can change so that the observed kinetics is not represen-
110 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

tative of one specific catalyst state. Such changes can occur during micron size range. In addition to the crystalline phase or phases,
kinetic testing, and it is difficult to distinguish different catalyst catalysts may contain one or more amorphous phases deposited
states with most existing kinetic testing procedures. on the surface of the crystalline phases. The surfaces of the crystal-
According to the methodology of surface science, reactions are lites are composed of different crystal planes containing a variety of
carried out over a well-defined surface (e.g., a specific plane of a defects such as kinks, edges, steps, and vacancies. Defects may range
metal single crystal), so that the observed kinetics can be directly in size from microns to Angstroms (Å). The chemical composition
associated with the structure of the surface [5–8]. Surface science of a surface may be different from the underlying crystal bulk, and
techniques are used to define the structure of the crystal surface, can change under reaction conditions. In the case of metal oxides,
and to determine if a surface is changed by reaction. Information there is a lack of knowledge regarding surface reactions even on
pertaining to reaction-induced changes can be obtained by charac- well-defined surfaces [10]. In the case of “real” oxides, surface reac-
terizing a surface before and after reaction. Kinetics are obtained by tions at defect sites may dominate the chemistry, and these types
thermal desorption techniques or from scattering experiments. The of reactions are particularly difficult to study by surface science
kinetics often reflects the intrinsic reactivity of the specific crystal techniques.
plane being studied. By comparing the reactivities of different crys- Under reaction conditions, the kinetic state of a catalyst can
tal planes of the same metal, inferences can be drawn regarding the change. For example, the absolute number of active sites can change
nature of the active catalytic site. as a result of a change in the catalyst surface area, a change in the
Surface science techniques [5–8] can be used to characterize crystal morphology, or a change in surface oxidation state. During
the structure or composition of technical catalysts; however, direct reaction, the available number of active sites can change as a result
kinetic testing of technical catalysts in surface science systems is of adsorption. The specific reactivity of sites can change because
usually impractical. Structural and compositional data from surface the composition or structure of the sites changes. A process that
science experiments may be matched with data from atmospheric causes a change in the number of active sites may also cause a
pressure kinetics, but the heterogeneous surface structure of tech- change in structure of the active sites. In this case, the changes may
nical materials prevents the observed kinetics from being directly even compensate one another. Such effects complicate the prob-
associated with a specific crystal plane or surface structure. Also, lem of determining the number of active sites on a catalyst and
the structure of a technical catalyst may change when it is moved their specific reactivity. Consequently, to simplify the problem of
from atmospheric pressure conditions (esp. when the catalyst has determining the number of active sites and the kinetic characteris-
been exposed to a reactive mixture) to high vacuum conditions. tics of individual sites, the experimental strategy that is used should
In this case, the structural data may not be directly related to the provide a means of maintaining the catalyst in a constant state or
actual active site. should involve a change that is “insignificant”.
Experiments commonly used to obtain kinetic data at industrial Since Bennett and Kobayashi–Kobayashi times, it is well
reaction conditions are usually based on a steady-state approach. accepted in the heterogeneous catalysis literature that tran-
An important advantage of this approach is that it can be readily sient experiments can provide more mechanistic information
applied to technical catalysts. The steady-state approach has disad- about the reaction intermediates and pathways of the various
vantages, however, when it comes to developing activity–structure elementary steps when compared to steady-state experiments
relationships. First, steady-state kinetic data is usually related [11–13,155,164–167]. Steady-state kinetic experiments can reveal
to only the slow steps of a complex kinetic process, and does the performance of a catalyst after it has evolved into a stable
not provide detailed information. Second, the observed kinetics structure. Non-steady-state kinetic experiments reflect changes in
is a complex function of the gas-phase composition, the struc- performance that occur during the evolution process from one
ture/composition of the solid, and other process variables. As state to another state. The changes in kinetic properties can be
a result of these factors, details of the gas–solid reaction, and related to changes in catalyst composition and structure. However,
detailed activity–structure relationships, cannot be readily estab- non-steady-state experiments utilizing continuously stirred tank
lished using steady-state kinetic data. reactors have been plagued by a number of obstacles, e.g., macro-
The surface science approach provides a procedure for relating scale hydrodynamic non-uniformity in catalyst bed. In the Bennett
the surface structure of a solid to the reaction kinetics that occurs on review (1999) [167], it was concluded that “it is no longer advanta-
the solid surface. Consequently, surface science studies can provide geous to do these experiments in an ideal mixed-flow reactor”. See
activity–structure relationships. On the other hand, surface science a detailed analysis of this problem in [148].
experiments are performed at ultrahigh vacuum conditions using A new approach for characterizing the catalytic activity of tech-
well-defined model surfaces (e.g., metal single crystals), and are not nical and model catalysts called “interrogative kinetics” (IK) [14]
well suited for investigating kinetics on technical catalysts. combines vacuum pulse-response experiments, commonly known
A well-known problem that arises when attempting to compare as TAP experiments [14–16], with atmospheric pressure steady-
data from more practical studies with data from surface science state and transient experiments, and allows a single catalyst sample
studies is the so-called “pressure gap” problem first delineated [17,18] to be tested over a wide domain of pressures (105 –10−6 Pa)
by Bonzel [9]. He noted that the large difference in pressures and relaxation times (103 –10−4 s) [14,16]. The specific transport
(often greater than 9–10 orders-of-magnitude), and the difference domain in which kinetic data are obtained in TAP experiments
between single crystals versus technical catalysts makes it very dif- (Knudsen diffusion domain) is located on the boundary of the sur-
ficult to compare results from the two experimental regimes. In fact, face science domain. The approach is applicable to both model and
despite enormous strides in understanding how reactions occur on realistic catalysts, and provides a procedure for bridging the pres-
well-defined surfaces, large gaps exist in our fundamental under- sure gap [9].
standing of how reactions occur on “real” surfaces. Such surfaces The IK approach uses two types of experiments, called “state-
exhibit a complex multi-scale structure with critical dimensions defining” (SD) and “state-altering” (SA) experiments to probe
ranging from the microscopic to the atomic scale, and are not well different states of a catalyst sample. In a SD experiment, the catalyst
suited for study using ultrahigh vacuum surface science techniques. composition and structure change insignificantly during a kinetic
Many industrial catalyst particles are composed of metals test. In a SA experiment, the catalyst composition is changed in a
deposited on metal oxides or mixed-metal oxide microcrystallites controlled manner. The IK approach involves kinetic testing of a
bound together in a random orientation. A typical catalyst particle catalyst, and systematically altering its composition or structure
will contain a complex pore structure with pore dimensions in the in a well-defined process. For example, the oxygen composition of
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 111

some metal oxides can be altered in a precisely measurable fashion


by heating the oxide in an oxygen atmosphere. After the catalyst
composition is incrementally changed, its kinetic characteristics
are measured and matched with the composition or structural
change. The kinetic measurement is performed under non-steady-
state conditions in such a way so as to not significantly alter the
kinetic properties of the sample, and to give intrinsic kinetic infor-
mation. The observed kinetics can be directly correlated with the
observed composition change. The approach provides an experi-
mental and theoretical methodology for determining the number
of active sites on technical metal oxide catalysts, and a method for
measuring the intrinsic kinetic characteristics of the active sites.
The IK approach uses a combination of TAP vacuum pulse response
and steady-state and transient experiments carried out at atmo-
spheric pressures.
The primary objective of this paper is to review the experi-
mental and theoretical basis of the TAP reactor system along with
recent and emerging aspects of the technology. Emphasis is placed
upon the physical basis for the TAP experiment and its relation- Fig. 2. Conceptional comparison of key components of (a) molecular beam scatter-
ing experiment and (b) TAP experiment.
ship with molecular beam experiments, the state-of-the-art for TAP
reactor systems and recent advances in a next-generation system,
developments in coupling a TAP reactor with a time-of-flight mass of the scattered molecules are measured. Molecular beam exper-
spectrometer and other analytical systems, mathematical models iments are performed under single collision conditions so that
that can be used for interpretation of the transient responses for reactant molecules collide once with the target surface but not
extraction of adsorption and kinetic rate parameters, and selected with each other. The observed distributions contain dynamic infor-
recent applications. A recent special issue of Catalysis Today [15] mation, which describes the process of energy exchange between
was dedicated to the TAP reactor with an emphasis upon applica- gas and surface atoms, as well as kinetic and mechanistic informa-
tions of the technology over the past 10–15 years to various catalytic tion. Scattered molecules from a collimated beam may also exhibit
systems to which the reader is referred. diffraction effects, which provide information on surface structure
and bond lengths [21–31].

2. The TAP experiment 2.2. TAP reactor setup

2.1. Overview A basic TAP setup, depicted in Fig. 2(b), has some character-
istics common to an MBS experiment, and others common to
The Temporal Analysis of Products, or TAP experiment was con- conventional microreactor experiments. The key components are
ceived in the late 1970s to study catalytic reaction mechanisms on a “reaction zone” or microreactor, a fast-pulse gas feed system, a
industrial catalysts [14,16,19,20]. The initial thought was to devise a mass spectrometer detector, and a high-throughput ultrahigh vac-
simplified “molecular beam” experiment for multi-component cat- uum system. The reaction zone (Fig. 2), which holds the catalyst
alysts (e.g., mixed-metal oxides and supported metals) that have sample, is a temperature controlled cylindrical tube usually made
high surface areas and complex pore structures. Molecular beam of stainless steel, inconel, or quartz. One end of the tube receives
scattering (MBS) experiments [21–31] can provide fundamental input from the feed system, and the opposite end is open to vac-
information on surface structure, reaction dynamics, the elemen- uum. During an experiment the reaction zone and catalyst sample
tary steps of a reaction, and the kinetic parameters of individual are continuously evacuated. The reaction zone can hold practical
steps. When used in conjunction with surface characterization catalysts, which are commonly studied in conventional microreac-
techniques, data from a MBS experiment can establish the link tor experiments, as well as single particles, single crystals, or model
between kinetic properties and surface structure. On the other catalysts, which are studied in MBS experiments.
hand, MBS experiments use planar or decorated planar targets to MBS experiments obtain essential kinetic and mechanistic infor-
take advantage of the spatial characteristics of the beam. Industrial mation by modulating the reactant flux, and measuring the shift
catalysts have complex surfaces, are composed of multi-component in arrival times at the detector between scattered reactant and
mixtures of different metal oxides or metals combined with metal product molecules. The difference in arrival times can be used
oxides, and are generally not suitable for MBS experiments. The goal to determine reaction sequences, surface lifetimes of adspecies,
of early TAP designs was to retain the time-dependent features of a and rates of surface reactions. TAP pulse-response experiments
molecular beam experiment, minimize gas-phase interactions, and [14,16,19,20,32,33] extract kinetic information in a similar manner.
provide a way to extract intrinsic kinetic information from reac- Injection of a narrow gas pulse into the reaction zone initiates an
tions on bulk catalysts. The TAP experiment can be viewed as a experiment. The gas molecules travel through the reaction zone
bridge between MBS experiments and conventional microreactor where they encounter the catalyst and can react to form product
experiments. molecules. Molecules that exit the reaction zone are monitored
A variety of MBS setups have been described in the litera- by the mass spectrometer positioned at the outlet. An experi-
ture, and they generally fall into two different classes: (1) systems ment ends when the flow of reactant and product molecules is
designed to study gas-surface dynamics, and (2) systems designed no longer detected by the mass spectrometer. The observed char-
to study the kinetics and mechanism of surface reactions [6]. acteristic feature in a TAP experiment is the time-dependent gas
Fig. 2(a) presents a simplified diagram showing key compo- flow F(t) [moles/s] or [molecules/s] that escapes from the exit of
nents of a MBS apparatus. In a typical experiment a well-collimated the microreactor. The flow dependencies also have integral char-
beam of molecules having a known translational energy is directed acteristics that are related to the moments of the “flow-time”
toward a target surface, and the velocity and angle distributions dependencies [14,16,34]. Kinetic and mechanistic information is
112 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

obtained by analyzing these dependencies, and gas transport is At sufficiently small pulse intensities, a one-pulse TAP experi-
used as a “measuring stick” to determine the rates of chemical ment can be considered a state-defining experiment. The number
transformations. of molecules in a reactant pulse is typically much smaller (102 –105
In addition to kinetic and mechanistic information, TAP exper- times smaller) than the number of surface atoms in the catalyst
iments have been used to study transport processes in porous sample being probed [14]. As a result, the reactant pulse does not
materials, such as zeolites, and oxygen diffusion in metal and significantly perturb the catalyst surface.
metal oxide catalysts. The temperature dependence of TAP tran-
sient response curves provides intrinsic rate parameters that can be
2.4. TAP reactor applications
associated with the surface composition of a catalyst. The number
of active sites can be determined in TAP titration experiments.
Table 1 provides an overview of TAP reactor applications, and
the types of catalytic materials that have been studied including
2.3. Types of TAP experiments supported metals, mixed-metal oxides, zeolites, metal particles,
metals deposited on screens, catalytic monoliths, and nanoparti-
In both TAP and MBS experiments, the sample is maintained
cles or atoms deposited on microparticles, single crystals, and other
under vacuum conditions, which strongly promote desorption of
model catalysts.
adspecies. Prior to performing a TAP experiment, it is common to
heat the catalyst sample and monitor the desorption spectrum. Dur-
ing the heating process, adspecies (e.g., water, CO, CO2 , etc.), which 3. The TAP reactor system
cover the surface at ambient pressures, vacate the surface leav-
ing the coverage to more closely resemble that of an MBS target. The temporal analysis of products (TAP) reactor system was
Conversely, pulsing a specific species into the reaction zone can first patented by the Monsanto Company in 1986 as a novel
increase its coverage so that it resembles coverage at ambient pres- device for studying the kinetics and mechanisms of heterogeneous
sures. Pulsing different molecules in an alternating sequence can catalyzed gas-phase reactions by using a transient response tech-
adjust the surface coverage of two or more species. As a result, cov- nique with submillisecond time resolution [19]. Two years later
erage in a TAP experiment can be manipulated to resemble coverage in 1988, open literature publications by Monsanto catalyst scien-
in an MBS experiment or coverage in a conventional microreactor tists described how the TAP reactor system was used to elucidate
experiment. the mechanisms for several heterogeneous catalytic reactions hav-
The input pulse in a TAP experiment typically contains ≈1014 ing commercial significance [16,20]. Particular reactions that were
molecules, and the local pressure in an empty reaction zone may studied included n-butane oxidation to maleic anhydride over
reach ≈10−1 Pa during a pulse. If unimpeded, an oxygen molecule vanadium–phosphorus oxide (VPO) catalysts, propylene oxidation
can traverse the reaction zone in under 100 ␮s. If the reaction zone to acrolein over bismuth–molybdates, methanol ammoxidation to
is filled with particles the pressure may reach ≈1.33 Pa in the void HCN over MnPOx and FeMoOx , and ethylene epoxidation over sil-
spaces if no adsorption occurs. The mean free path in an empty ver metal [95,96]. In 1989, Monsanto granted an exclusive license
reaction zone is about half the length of the reactor. In an empty to manufacture and sell a commercial version of the TAP reactor
reactor, molecules move in beam-like fashion and suffer relatively to Autoclave Engineers of Erie, PA. Later in that same year, DuPont
few collisions. In a packed reactor, the mean free path is ≈4000 ␮m, became the first chemical company in the United States to purchase
which is significantly larger than the space between particles. As a a TAP reactor from Autoclave Engineers for the newly established
result, in a packed-bed reactor molecules collide with particles, but Corporate Catalysis Center in Central Research and Development
seldom with one another (Fig. 3). [20]. Later, the TAP-1 system was redesigned and simplified and

Fig. 3. Scale drawing of a TAP “reaction zone” and packed-bed microreactor containing a loading of nonporous spherical particles ≈250 ␮m in diameter.
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 113

Table 1
Applications and theory of the TAP reactor including the corresponding catalytic materials tested.

Research area Catalytic materials References

Reactor transport (adsorption and diffusion) Zeolites, ␥-Al2 O3 , Rh/Al2 O3 , sulfated ZrO2 [35–42]

Conversion of simple hydrocarbons


Ethane Pt/Al2 O3 , Na/CaO, Sm2 O3 /CaO, Sm2 O3 [43,44]
Methane MgO, Na/CaO, Sm2 O3 , Ni/Al2 O3 [45–51]
Conversion to syngas Rh/␥-Al2 O3 , Pt/MgO, Pt gauzes, Ru/Al2 O3 , Pt/ZrO2 , Pt/Al2 O3 [52–55]
n-Butane VOPO4 , (VO)2 P2 O7 [56–60]
n-Pentane (VO)2 P2 O7 [57]
Propane VCrMnWOx , Fe-ZSM-5, mixed vanadia-based oxides, VOx /MgO, Sm2 O3 , [61–67]
VOx /␥-Al2 O3 , Al-Sb-V-W-oxide catalyst
Cracking/reforming of hydrocarbons Zeolite-based catalysts, Pt-supported catalysts, MgO-Ru/C [68–73]

Environmental catalysis
N2 O abatement Pt, Pt-Rh mixed catalysts, Fe-MFI [74–78]
NOx storage Pt/BaO/Al2 O3 [79]
SCR of NOx Ag/Al2 O3 , Pt/ZSM-5, mordenite [80–82]
VOC U3 O8 [83]

Hydrogenation reactions
Acrolein hydrogenation Ag/SiO2 [84]

Oxidation reactions
Ammonia oxidation Pt, Pt–Rh mixed catalysts [85–89]
CO oxidation Au/Fe2 O3 , Au/Ti(OH)4 , La1−x Srx Fe(Pd)O3 , Pt, Pd/SiO2 [17,90–94,154,161,169]
Ethylene oxidation Ag powder [95,96]
o-Xylene oxidation VOx /TiO2 [97]
Propene oxidation Co10 Mo12 FeBiOx , ␥-bismuth molybdate [98–100]
Soot oxidation La3+ -doped CeO2 , CeO2 [101–103]
Toluene ammoxidation ␣-(NH4 )2 [(VO)3 (P2 O7 )]2 [104]
Toluene oxidation VOx /TiO2 , VOx /SiO2 [105–107]

TAP theory and modeling [14,16,18,32,34,96,100,108–124,168]

eventually developed into the TAP-2 reactor system [14], which was The microreactor is positioned directly above the ionizer of the
commercialized by Mithra Technologies. The TAP-3 system retains mass spectrometer, and is attached to a movable stainless steel bel-
the basic design of the TAP-2, but is a fully automated instrument lows. A unique rotary-valve assembly is located between the ionizer
that can be operated either locally or remotely via the Internet. and the microreactor. It permits the microreactor to be operated at
vacuum conditions or atmospheric pressures, and allows the reac-
3.1. TAP apparatus design tor to be removed from the system without venting the vacuum
chambers. The reactor can be easily and rapidly switched from one
Fig. 4 presents a simplified schematic of a TAP-3 reactor sys- kinetic regime to another without exposing the catalyst sample to
tem, and Fig. 5 shows a photograph of a system. The TAP-3 reactor the atmosphere.
is comprised of (1) a pulse-valve manifold assembly that supplies When the rotary-valve is closed, the TAP microreactor can be
gas reactants for pulsed and flow experiments at user defined tem- operated as a continuous plug flow-type reactor at atmospheric
peratures and pressures, (2) a microreactor assembly that can be or higher pressures. In the “high-pressure” mode, the bulk of the
operated isothermally or in a temperature programmed mode, (3) a reactor effluent exits through an external vent. A small amount
mass spectrometer detector contained in a high-throughput ultra- can be leaked into the vacuum chamber through a variable leak
high vacuum system, and (4) a computer based control and data valve. The leaked material can be monitored with the mass spec-
acquisition system. trometer. In high-pressure experiments, the mass signal usually
The pulse-valve manifold contains one continuous flow valve changes relatively slowly, and can be collected in a scan mode.
and four high-speed pulse valves that are arranged to minimize In this mode, a complete mass spectrum can be acquired about
the dead volume between the valves and the microreactor. The once every 0.25 s. After performing an experiment at atmospheric
four pulse valves can be triggered simultaneously or in a pro- pressures, the microreactor can be quickly switched to vacuum con-
grammed alternating sequence. Pulse intervals can be varied from ditions by opening the rotary-valve.
microseconds to minutes. Switching between feeds is accom- With the rotary-valve in the open position, all of the reactor
plished electronically and can occur almost instantaneously. The effluent vents into the vacuum chamber. The movable bellows
continuous flow valve is connected to a separate manifold contain- allows the reactor outlet to be positioned within 2 mm of the ionizer
ing four flow controllers. The flow controllers and pulse valves can so that most of the effluent passes through the ionizer. As a result,
be operated simultaneously. very small inputs of gas can be detected, with a very high signal to
The TAP-3 vacuum system is comprised of two chambers sepa- noise (S/N) ratio (Fig. 6). The high S/N ratio allows accurate calcula-
rated by a pneumatically operated gate-valve, which is closed in the tion of the zeroth, first, and second moments of the response curves,
standby position. The lower chamber is evacuated by a 40 cm diam- and direct determination of many important kinetic characteris-
eter diffusion pump and associated mechanical pump. The chamber tics.
contains a cylindrical liquid nitrogen trap and adjustable baffle that For example, reactant conversion for a series of pulses can be
closes when data is not being acquired. The upper chamber holds determined from the zeroth moment. This change in conversion
the mass spectrometer and is pumped by a turbo molecular pump. can be related to a change in the concentration of the active cat-
The background pressure in the mass spectrometer chamber is typ- alytic species. The following quantities can be directly calculated
ically ≈10−7 Pa. from 0th, 1st, and 2nd moments: Conversion, selectivity, product
114 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

Fig. 4. Schematic of TAP-3 apparatus showing two-chamber vacuum system (base pressure = 10−7 Pa) with gate-valve, adjustable baffle, liquid nitrogen trap, rotary-valve for
switching between vacuum and pressure experiments, and quadrupole mass spectrometer; microreactor carousel with autoload microreactors and catalyst sample input
lines; reactant feed manifold with high speed pulse valves and continuous flow valve; gas feed system with heatable mass flow controllers, heated blend tanks for preparing
gas mixtures, and vacuum pump; computer control and data acquisition system indicating main control lines.

yield, residence time, apparent equilibrium and rate constants, and tions. Reactors are heated electrically and the reactor temperature
apparent time delay. controller provides an array of control options, which include con-
The TAP-3 system can accommodate microreactors of differ- stant temperature operation, programmed heating or cooling at
ent lengths (up to 25 cm) and diameters (up to 2 cm), which are a user defined rate, and the ability to perform multistep ramp-
useful for handling a variety of catalyst forms (see Fig. 7). A stan- soak programs. The standard microreactor can be heated to 800 ◦ C.
dard TAP microreactor is constructed of type 316 stainless steel Microreactors have also been fabricated from inconel and quartz
and can be operated at atmospheric pressures or at vacuum condi- which allow operation to temperatures as high as 1000 ◦ C.
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 115

Fig. 5. TAP-3 system.

The TAP-3 menu of experiments includes, but is not limited


to, high-speed vacuum pulse-response experiments (TAP Knudsen
pulse-response experiments, TAP pump-probe experiments, and
TAP multipulse experiments, pulse experiments with a change of Fig. 6. TAP pulse response curve with inset showing the noise magnitude in the tail
of the response. Signal to noise ratio is >30,000 to 1.
time within a pulse and between pulses), atmospheric pressure
steady-state, step-transient and SSITKA experiments, tempera-
ture programmed desorption (TPD), and temperature programmed 4. TAP reactor: advanced analytical systems
reaction (TPR). In addition, newly developed software allows the
user to create programmed experimental sequences, which can be 4.1. Overview and motivation
stored in memory, and then performed automatically. Sequences
may include complex temperature treatments, switching back and The study of reaction mechanisms and kinetics for gas-phase
forth between atmospheric pressure and vacuum experiments, heterogeneous catalyzed systems requires both qualitative identi-
switching from continuous flow to transient response experiments, fication and quantitative analysis of various gas-phase and surface
or combinations of step-transient, pulse transient, steady-flow, and species as a precursor to development of robust kinetic mod-
temperature programmed experiments. els. However, most catalytic reactions, especially those that occur
When the TAP-3 system is operated in the high-pressure mode, in many practical industrial gas-phase catalytic processes, often
it can automatically perform a sequence of atmospheric pressure generate a complex distribution of primary and secondary gas-
experiments, including temperature programmed (TPD, TPR and phase reaction products through a sophisticated sequence of
TPO) experiments, step response experiments, and steady-state iso- surface-catalyzed reactions whose surface processes and mecha-
topic switching (SSITKA) experiments. Some examples of these are nisms are generally not well-understood. For this reason, practicing
provided in the references cited earlier in Table 1. engineers that must either design new catalytic processes, or

Fig. 7. Examples of TAP packed-bed microreactor configurations and catalyst forms. See Fig. 22 for schematic of TAP system with pulse valves, slide valve, and mass
spectrometer.
116 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

optimize existing ones, often develop kinetic models based upon Reaction intermediates and products that have been reported
empirical power-law, classical Langmuir–Hinshelwood (LH), or in the catalyst literature when either n-butane, various butylenes,
Hougen–Watson (HW) rate expressions using kinetic data gen- or butadiene are used as the carbon source over various metal
erated from steady-state flow-type catalytic reactors, such as oxide catalysts include 1-butene, 2-butene, butadiene, acetone,
packed-bed microreactors [125]. Approaches based on mecha- methyl ethyl ketone, methyl vinyl ketone, furan, dihydrofuran,
nisms involve sequences of elementary steps where the precise maleic anhydride, acetic acid, crotonic acid, acrylic acid, glyoxylic
nature of catalyst active sites is not identified. These rate acid, methacrylic acid, formaldehyde, acetaldehyde, propionalde-
forms provide a well-known approach for fitting reaction kinetic hyde, crotonaldehyde, methacrolein, carbon oxides, and water
data, but do not provide a fundamental basis for design of [131]. Identification and quantification of the products gener-
new multi-functional catalyst materials using advanced princi- ated from these C4 hydrocarbons using steady-state and transient
ples of inorganic, solution-phase and computational chemistry response experiments poses a significant challenge for any on-
along with supporting advanced surface characterization meth- line analytical system because of the variety of reaction products
ods. Experimental reaction studies involving model compounds on that can exist under various reaction conditions and the common
well-characterized catalyst surfaces provide one alternate approach molecular functionalities between the various species. In the case
for developing insight in reaction mechanisms for these complex of packed-bed microreactor systems used for catalyst screening
systems. However, the absence of other key species in the reaction and kinetic evaluation for n-butane oxidation, such as the multi-
mixture can lead to various degrees of uncertainty in the applica- ple automated reactor system (MARS) [132–134], special-purpose
bility of model compound studies with idealized catalyst surfaces multi-dimensional gas chromatographic methods have been devel-
to the actual system. Transient response methods where all surface oped to resolve and identify all of the C1 to C4 hydrocarbons,
and gas-phase species are simultaneously monitored in real-time oxygen-containing organic species, and inorganic products [131].
provide a more robust approach for deciphering the kinetics and This suggests that special methods may be needed when perform-
mechanisms of these complex systems [126–130]. ing single ion monitoring of selected species to obtain reliable
Many examples can be cited from the literature where the transient response data for reacting systems involving a complex
products produced from gas-phase catalyzed reactions result in distribution of reaction products.
complex, multi-component mixtures containing a spectrum of both Fig. 9 is a matrix that shows typical molecules that have
organic and inorganic products. Two examples will be provided been identified in reaction mixtures for the partial oxidation of
here to illustrate some of the key challenges associated with mon- n-butane to maleic anhydride over vanadium–phosphorus oxide
itoring the transient responses of the gas-phase species using catalysts under oxygen-rich conditions as encountered in fixed-
quadrupole mass spectrometry, which is the primary analytical bed processes [135,141]. The organic products include acetic acid,
detector in the TAP reactor system. furan, dihydrofuran, acrylic acid, and maleic anhydride along with
Consider first the partial oxidation of various C4 hydrocarbons, various C4 hydrocarbons, COx , and H2 O. Selected ions produced
such as n-butane, iso-butane, iso-butylene, 1-butene, 2-butene, or from electron-impact ionization of each species are also listed
1,3-butadiene to various organic ethers, ketones, aldehydes, and to illustrate complexities that can occur if single-ion monitoring
carboxylic acids over various metal oxide catalysts. Fig. 8 shows is performed. Other ions are also produced but are not listed for
some of the reaction pathways and the associated products that can brevity. For a given hydrocarbon or organic oxygen-containing
be produced through the selective abstraction of hydrogen or the specie, several common ions occur so that the response for a
addition of oxygen to various starting reactants and reactive inter- particular ion would represent the combined response from
mediates. It is evident that a variety of isomers and other derivatives several species. Data handling methods for quantifying multiple
can be produced depending upon the dominant reaction pathway. ion responses that account for these effects are well-known in

Fig. 8. Reaction pathways and products that can be generated from the selective oxidation of various C4 hydrocarbons over metal oxide catalysts.
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 117

Fig. 9. List of species identified from GC/MS analysis of reaction products from the partial oxidation of n-butane over a vanadium–phosphorus oxide catalyst [134]. The
indicated ions for a given species are a subset of all the ions that can be generated to illustrate the overlap that can occur. The first row in the table indicates the mass to
charge ratio (m/z) of the species.

the mass spectrometry literature, which are often based upon these systems, and to briefly demonstrate their utility for studying
the implementation of convolution methods [136]. However, one heterogeneous catalysis behavior using the TAP reactor system.
complication that can occur when applying these to TAP data is that
transient experiments must be repeated at different ions before a 4.2. TAP reactor—time-of-flight system
composite set of ion response versus time data can be assembled.
This requires that the initial state of the catalyst can be reproduced This section summarizes the key results from a prototype sys-
within a tolerable level of error so that the resulting responses are tem that was developed to demonstrate the feasibility of coupling
representative of the same surface reactions that occurred in each a TAP reactor to a time-of-flight (TOF) mass spectrometer system
of the preceding set of experiments where the responses of differ- for simultaneous monitoring of multiple ions from a single pulse
ent ions were measured. This is particularly important for many or other type of input. A quadrupole mass spectrometer (QMS) was
catalysts where the surface coverage, oxidation state, active sites, also included in the system to provide an independent source of
etc. can undergo significant changes when exposed to small doses data for single ion monitoring.
of one or more species. Unless the composition of the reaction
products can be determined a priori through some independent 4.2.1. System description
means, such as GC/MS or from previous experience with the same A schematic of the prototype system that was developed to
catalyst system, caution must be exercised to ensure that the reso- test coupling of a TAP microreactor to a TOF mass spectrometer
lution and quantification of the primary reactants and products, i.e., is shown in Fig. 10. A simplified vacuum system was developed
n-butane, oxygen, maleic anhydride, carbon oxides, and water, is that allowed the TOF analytical package to be readily coupled to
not corrupted by the presence of one or more unidentified species. a TAP microreactor that was fitted with a set of high-speed pulse
Simultaneous measurement of both gas-phase and surface valves. Referring to Fig. 10, the system contains a single 20 cm vac-
species during the course of various TAP reactor transient response uum chamber in which the TAP fixed-bed microreactor and pulse
experiments for complex, multi-component catalytic reactions is valves are mounted in a vertical orientation in the center of the top
one of the outstanding opportunities associated with strengthen- plate. The microreactor and pulse-valve manifold are located on
ing, advancing, and broadening the utility of the technique. This the top of the vacuum chamber for ease of operation and mainte-
section summarizes some recent advances in coupling analytical nance. The reactor can be isolated from the primary chamber using
techniques to the TAP reactor as part of a larger effort to per- a custom-designed flange valve. A quadrupole mass spectrometer
form qualitative identification and quantitative analysis for the (UTI, Model 100C) is mounted on a 12 cm flange so the ionization
gas-phase species that are produced during the course of a single cage is perpendicular to the flight of the reactor product gas. This
transient response experiment. The particular methods described provides a secondary source of ion monitoring in addition to the
here include: (1) time-of-flight (TOF) mass spectrometry with TOF multiple ion monitoring comparison of the reactor responses.
electron-impact (EI) ionization, and (2) on-line gas chromatography A linear time-of-flight (TOF) mass spectrometer (Jordan TOF Prod-
with mass spectrometry (GC/MS) where gas samples are introduced ucts, Grass Valley, CA) is connected to the main vacuum chamber in
to the GC column by capturing a small (250 ␮l) gas sample of the a horizontal orientation using a six-inch flange. The flight tube has
transient response at various times along the response trajectory. an overall length of ca. 91 cm that terminates with a microchan-
The primary objective here is to illustrate the basic features for nel plate (MCP) detector. The electron gun for the TOF system is
118 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

Fig. 10. Schematic of a TAP reactor coupled to a time-of-flight mass spectrometer.

mounted on a 7 cm flange directly across from the flight tube grid electron-impact ionization, is illustrated in Fig. 11. The basic princi-
assembly. To monitor the pressure in main vacuum chamber and ples and operation of the TOF are generally described by Cotter [137]
the flight tube, full-range ion gauges are mounted on two 2 cm so they are omitted here. The molecular beam generated from the
flanges. Vacuum in the primary chamber is provided by a 520 l/s TAP is introduced to the ionization zone which initiates the various
turbo molecular pump, while additional vacuum for the TOF flight processes needed to create a product signal from the microchannel
tube is provided by a 210 l/s turbo molecular pump. With this com- plate detector (MCP). The design shown in Fig. 11 is a linear TOF,
bination of pumps, a background pressure of ca. 10−6 Pa is achieved although the current version in our lab employs an angular reflec-
in the flight tube, while 10−5 Pa is obtained in the primary chamber. tron (AREF) TOF for higher mass resolution. Because the TOF system
The fixed-bed TAP microreactor is constructed of type 316 stain- was being evaluated for the first time as a new technique for collec-
less steel and has an inner diameter of 5.5 mm with an overall tion of time-resolved transient kinetic data from a heterogeneous
length of 41.7 mm. It seals against the heated valve manifold using a catalyzed reaction, a special-purpose control and data acquisition
Kalrez® O-ring (DuPont Performance Elastomers, Wilmington, DE). system was developed using commercially available components.
A custom-designed splitter valve for directing a continuous con- The TOF spectrometer is typically calibrated by injecting a gas
trolled leak of the reactor product gas to the vacuum chamber can mixture containing He, Ne, N2 , Ar, and CO2 , Xe, and other compo-
also be attached to the reactor exit. The reactor is packed by sand- nents using the TAP reactor pulse-valve manifold and measuring
wiching the catalyst particles between two sections of 250–425 ␮m the time-of-flight of the detected ions. These gases are selected
quartz beads. The front inert section serves to preheat the inlet to span the range of molecular weights that are typically encoun-
gas, while the post-inert section minimizes the dead volume and tered in TAP reactor applications, such as those involving the partial
broadening of the response once the product gas exits the catalytic oxidation of light hydrocarbons. The flight time for CO2 is about
section. 10 ␮s so that a complete product gas mass spectrum for typical gas-
phase heterogeneous reactions can be typically collected in about
4.2.2. TOF operation and data collection 30–40 ␮s. By contrast, the time required for the UTI quadrupole to
The basic design concept of the TOF spectrometer, when used scan over the same range, assuming 4 points per amu, would be a
as a detector for TAP reactor transient response experiment with few seconds. The TOF clearly permits rapid detection of multiple

Fig. 11. Linear time-of-flight mass spectrometer. VA1 = repeller plate voltage, VA2 = extraction grid voltage, VFOC = focusing voltage, VX1 = beam steering voltage and
MCP = microchannel plate detector.
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 119

Fig. 12. Comparison of experimental and model-predicted responses obtained when the TAP–TOF system is operated using the QMS and TOF detectors. (a) QMS detector;
pulsed gas = He; packing = SiC granules; particle density p = 3.2 g/cc; particle diameter dp = 175 ␮m; reactor inner diameter dr = 5.5 mm; T = 25 ◦ C; packing length L = 10 mm
(case 1) and 40 mm (case 2). (b) TOF detector; pulsed gas = CO2 ; packing = quartz granules; particle density p = 2.6 g/cc; particle diameter dp = 338 ␮m; reactor inner diameter
dr = 5.5 mm; T = 25 ◦ C; packing length L = 40 mm.

ions from a single pulse input at a rate that is several orders-of- sion is the primary mode of gas transport and instrument effects
magnitude faster when compared to the QMS. are negligible.

4.2.3. Example application: simultaneous collection of multiple


4.3. TAP reactor with on-line GC/MS system
ion spectra
Fig. 12 compares the experimental and model-predicted
TAP reactor transient response experiments that produce reac-
responses obtained when a Knudsen diffusion model discussed in
tion products with complex mass spectrometer fragmentation
a later section is used to fit TAP vacuum response data measured
patterns from the quadrupole mass spectrometer detector require
using the QMS and TOF detectors in separate experiments using
development of special experimental methods and signal process-
inert gases and inert packing. The experimental details are provided
ing techniques before the response of a particular species can be
in the figure caption. The results show that both the QMS and TOF
extracted from the ion response versus time matrix. Analysis of
experimental responses are in good agreement with the predictions
the product gas composition along the transient response trajec-
of the single parameter Knudsen diffusion model. This provides ver-
tory represents key information that is needed for modeling of the
ification that the responses are not affected by the vacuum system
transport-kinetic interactions that describe the TAP reactor perfor-
design or other related instrument-related parameters. Additional
mance. When a time-of-flight mass spectrometer is used instead
details on the time-of-flight system design in connection with the
of a quadrupole mass spectrometer, the ion response versus time
TAP reactor and its application are omitted here for brevity but will
matrix can be assembled from the data generated by the TOF
be provided in future publications from our group.
microchannel plate detector during the passage of a single pulse
The validity of the TAP reactor responses measured by the TOF
input or other user-defined sequence of pulse inputs, e.g., multi-
in Fig. 12, as well as the response data measured by the QMS, can
pulse or pump-probe inputs. Conversion of the raw ion response
be evaluated by comparing ratios of the Knudsen diffusion coef-
versus time matrix to the species concentration versus time matrix
ficients to the ratio of the inverse square root of their molecular
first requires qualitative identification of all species in the mixture.
weights in accordance with the theoretical expression for the Knud-
Once this is accomplished, response factors for the individual ions
sen diffusion coefficient based upon the kinetic theory of gases. The
of a given species can be assigned so that the measured response
Knudsen diffusion coefficients were obtained by fitting the exper-
of a given ion can be used to compute the instantaneous gas com-
imental responses to the model-predicted experimental response.
position using standard data modeling methods, such as AMDIS
The results of this analysis are shown in Fig. 13. The data for He
[138].
was used as the reference gas, although the use of other species for
Special-purpose hardware interfaces have been developed to
the reference had a negligible effect. The data generate a response
generate step-up or step-down concentration inputs to a TAP-1
whose slope is unity, which should be the case if Knudsen diffu-
microreactor system [16,19,20] operating with a continuous gas
flow at atmospheric pressure. The reactor product gases can be
routed through a low dead volume heated gas leak valve and trans-
fer line so that the transient responses can be sampled on-line and
analyzed using a GC/MS system. The time at which a gas sample
is captured is user-defined and can be any point along the tra-
jectory of the transient response. In addition, a small (<0.5 sccm),
controlled continuous leak of the product gas can be directed to the
TAP analytical chamber that houses the quadrupole, thereby allow-
ing for continuous monitoring of a single ion. The technique could
be readily extended to the case when a TOF is used, thereby provid-
ing simultaneous monitoring of all key ions and their subsequent
specie assignments using data from the GC/MS system.
Although the experimental methodology was initially demon-
strated using a TAP-1 system, the approach used here for capturing
the on-line gas sample for subsequent GC/MS analysis could be
readily adapted to the TAP-2 [14] and future system designs. It
Fig. 13. Test of the experimental response data fitted to the single parameter Knud- could also be adapted for product analysis from TAP pulsed high-
sen diffusion model using data generated from both the TOF and QMS detectors. vacuum experiments for product identification and quantification,
120 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

Fig. 14. Chromatogram from FID analysis of a product mixture from a typical steady-
state catalyst test for the partial oxidation of 1,3-butadiene to furan. Reactor feed
composition: 8.8% 1,3-butadiene, 10.3% oxygen, 9.0% nitrogen and balance: helium;
reaction temperature: 380 ◦ C. GC column: 0.53 mm i.d. × 30 m 1701 phase, film thick-
ness = 5 ␮m.

which represents a departure from dedicated step-response reac-


tor systems designed for dedicated operation at normal pressures. Fig. 15. Structures of the organic products identified in the product gas for the partial
Addition of a GC/MS is particularly useful since the various species oxidation of 1,3-butadiene to furan. Reactor feed composition: 8.8% 1,3-butadiene,
10.3% oxygen, 9.0% nitrogen and balance: helium; reaction temperature: 380 ◦ C.
present in the product gas mixture are first resolved into com-
ponent peaks using an appropriate arrangement of GC columns
using either a flame ionization or thermal conductivity detector for deconvolution of the product spectrum, which must be incor-
with a small continuous split gas flow being directed to the mass porated into future TAP reactor systems that have multi-functional
spectrometer. The GC peaks are assigned to a unique species by capabilities.
matching the ionization fragmentation patterns using standard
mass spectrum library software, such as the Wiley NIST/EPA/NIH
library [158]. 4.3.2. Closing comments
Catalytic reactions that produce a distribution of products
require identification of the major species that are present
4.3.1. Example application: oxidation of 1,3-butadiene over a
before any detailed study of the transient kinetics can be con-
CuBib Pbc Mod Ox catalyst
ducted. Advanced analytical methods, such as time-of-flight mass
Fig. 14 shows the chromatogram produced from analysis of
spectrometry and GC/MS, are both useful for both qualitative identi-
the product mixture from steady-state TAP reactor continuous-
fication and quantitative analysis when coupled to the existing TAP
flow experiments at atmospheric pressure for the partial oxidation
quadrupole mass spectrometer system. Future TAP applications will
of 1,3-butadiene to furan over a Cu0.2 Bi1.95 Pb0.05 MoO6 catalyst at
require the use of more sophisticated analytical methods to obtain
380 ◦ C. This particular catalyst is a member of a family of modified
quantitative information on the interaction between gas-phase and
Pb–Bi–Mox Oy catalysts containing vanadium, copper, or gold, and is
surface-catalyzed processes.
useful for the gas-phase oxidation of an unsaturated acyclic hydro-
carbon with selectivity to the corresponding furan product [99,159].
Other experimental particulars are provided in the figure caption. 5. Theory of TAP packed-bed microreactor configurations
This example was selected to illustrate the analytical complexity
that can be encountered during the catalyst discovery stage, and to The analysis of TAP pulse-response experiments are based on
illustrate the utility of an on-line GC/MS for species identification specific microreactor models that describe the catalyst zone. Cur-
when integrated into the TAP system configuration. The product rent models apply to packed-bed reactors, starting with a one-zone
gas sample was captured on-line using a multi-port gas sampling reactor, and progressing to a reactor containing a single particle
valve, and was transferred directly to the inlet of an Agilent 6890 [17,18]. When the input pulse intensity is ≈1014 molecules/pulse
GC equipped with both a FID and an Agilent 5973 mass selective transport in the reactor is dominated by Knudsen diffusion, which
detector through microcapillary tubing. This same product distri- is mainly driven by a gas concentration gradient. The magnitude
bution was also independently reproduced in a MARS reactor [132] of the gradient is dependent on the reactor length and the vac-
for validation of the TAP system setup. uum boundary condition at the reactor outlet. Under the conditions
The organic products that were positively identified in the prod- of catalytic reaction, the gas concentration gradient causes a sur-
uct gas matrix are shown in Fig. 15, which accounts for more face concentration gradient in the catalyst bed. When the length
than 80% of the total number of peaks. Inspection of the indicated of the catalyst bed is a significant fraction of the reactor length,
compounds provides initial insight into the complexities of the cat- the reactant can cause non-uniformity in the catalyst composition
alytic mechanisms and reaction pathways that must be operative [109–112].
for some of these to be produced, particularly the functionalized TAP microreactor development has focused on minimizing the
aromatics, cyclic ethers, and organic acids. A more in-depth inter- effects of temperature and concentration gradients. An axial tem-
pretation of the catalytic chemistry is not included here, but is a perature gradient exists because of the close proximity of the
topic of ongoing research. However, this product distribution points microreactor outlet to the vacuum chamber. Typically the tem-
to the complexities that will occur when single or multiple ion perature near the center of a microreactor will be higher than the
monitoring is used to detect the transient behavior of a particular temperature close to the microreactor outlet. The evolution of TAP
reactant or product. This suggests that identification of the transient microreactor configurations and their theoretical description will
responses of various species will require data processing methods be presented in the following sections.
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 121

5.1. Basic concepts of reaction. This process has been demonstrated using the selective
oxidation of hydrocarbons over a transition metal oxide catalyst
There are three basic concepts that form the basis for extracting (e.g., vanadyl pyrophosphate or VPO) [33,140]. An oxidized VPO
kinetic information from TAP Knudsen pulse-response experi- sample was exposed to a series of hydrocarbon pulses and the
ments: change in kinetic properties were determined as a function of the
oxidation degree.
1. Well-defined Knudsen diffusion is a “measuring stick” for mea-
suring chemical reaction rates and extracting kinetic parameters. 5.1.4. TAP studies and model-free kinetic analysis
2. During a single pulse experiment the solid catalyst changes The goal of model-free kinetic analysis is to obtain the rate
insignificantly and controlled change occurs in a multi-pulse of chemical transformation without assuming a specific kinetic
experiment. model. Temkin and Denbigh applied a model-free approach to the
3. The solid catalyst surface composition will remain uniform if the analysis of steady-state kinetics for over 50 years. In their approach,
active zone is a sufficiently small fraction of the total bed length. the rate of chemical substance transformation is equal to the dif-
ference between the inlet and outlet molar flow rates divided by
The first and second concepts were reported implicitly in the the catalyst volume or surface area. In CSTR steady-state experi-
first TAP paper in 1988 [16]. In the 1997 TAP paper [14], the above ments no assumptions regarding the type of kinetic dependence,
concepts were expressed explicitly and presented mathematically. the reaction mechanism, or the corresponding model are needed to
The concept of uniformity was stressed in the paper devoted to the determine the rate of chemical transformation. Model-free kinetic
thin-zone TAP reactor in 1999 [109]. analysis of non-steady-state reactions [148] is a recent develop-
ment that began with the TZTR microreactor configuration.
5.1.1. Knudsen diffusion as a “measuring stick” A model-free kinetic method known as the “Y-procedure” [148]
Since Temkin’s [143–145] and Denbigh’s [146,147] times, a com- has been used to extract the non-steady-state rate of chemical
mon approach for extracting kinetic information is to measure the transformation from reaction-diffusion data with no assumptions
rate of chemical reaction using the rate of mass transport as a regarding the kinetic model. This method will be briefly described
“measuring stick.” In traditional steady-state experiments with per- in a separate section of this paper. The Y-procedure consists of the
fect mixing, convectional transport is the “measuring stick” and following steps:
diffusional transport is neglected. In TAP Knudsen pulse-response
experiments there is no convective flow and Knudsen diffusion is 1. Exact solution in the Laplace domain.
the only gas transport process. In the absence of reaction, the gas 2. Switching to the Fourier domain to allow sufficient computation.
exit flow from the microreactor is described by a standard diffusion 3. Introduction of discretization and filtering in the Fourier domain
curve. In the “reaction-diffusion” case, the exit flow response curve to deal with real data (in the time domain) subject to noise.
is different from the standard diffusion curve, and this difference is
attributed to reaction. Overall, the methodology used in TAP Knudsen pulse-
response experiments combines experimental conditions that
5.1.2. Insignificant change in the solid catalyst during single pulse simplify the physico-chemical system with basic mathematical
experiment and controlled change during multi-pulse experiments ideas—especially with the idea of infinitesimal change.
When the number of reactant gas molecules in a single pulse is
significantly smaller than the number of catalytically active sites on 5.2. One-zone TAP microreactor
the catalyst, the catalytic system remains in essentially the same
state after the measurement. This type of experiment is called a In the one-zone microreactor model the total reactor volume is
“state-defining” experiment [14]. A series of state-defining exper- uniformly packed with catalyst particles. The one-zone model and
iments can cause the catalyst state to change, and the change its mathematical framework were introduced in the first TAP paper
is characterized by the amount of consumed/released gaseous published in 1988 [16]. The mathematical model is based on the
substances. A series of state-defining experiments is termed a following assumptions:
“state-altering” experiment [14].
1. The fractional voidage of the catalyst bed is constant.
5.1.3. Uniformity of the catalyst surface composition across the 2. There is no radial concentration gradient in the catalyst zone.
active zone 3. There is no radial or axial temperature gradient in the catalyst
When a heterogeneous catalyst is exposed to a reactant gas the zone.
composition and kinetic properties of the catalyst can change as a 4. There is no intra-particle or surface diffusion.
result of reaction with the gas. In a packed-bed reactor, the gas flow 5. The diffusivity of each gas is constant and independent of the
can cause the change to occur non-uniformly. The inlet of the bed composition of the mixture as a whole.
will see the highest reactant concentration, and will change by the
largest amount. In a non-steady-state experiment the bed compo- The last assumption results from using an evacuated microreac-
sition can also change in time. An important and unique feature of tor and small pulse intensities, which guarantees the validity of the
a TAP pulse-response experiment is that the catalyst composition Knudsen diffusion regime. The mass balance equations for a num-
remains essentially uniform when the thickness of the catalyst zone ber of important gas transport and transport-kinetics interactions
is small compared to the total length of the packed bed. In practice a in the one-zone reactor are presented below.
small amount of catalyst can be packed in a “thin zone” between two
beds of inert particles or in some cases a single particle can be used. 5.2.1. Diffusion only
The advantage and properties of a thin-zone TAP reactor (TZTR) In a packed-bed reactor, the mass balance equation for Knudsen
have been discussed in detail in the literature [33,109–112,148]. The diffusion of a non-reacting gas A (e.g., neon, argon and krypton) is
properties of a single particle TAP reactor [17,18] have recently been given by
described.
Using a TZTR, a catalyst sample can be characterized “state-by- ∂CA ∂2 CA
εb = DeA , (1)
state” to determine how its catalytic properties change as a result ∂t ∂z 2
122 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

with initial condition:


NpA
0 ≤ z ≤ L, t = 0, C A = ız , (2)
εb A
and boundary conditions:
∂CA
z = 0, = 0, (3)
∂z
z = L, CA = 0. (4)

The initial condition, Eq. (2), specifies that at t = 0, the gas con-
centration at the reactor inlet can be represented by the delta
function. Boundary condition one, Eq. (3), specifies that the input
flux is zero at the microreactor entrance when the pulse-valve is
closed. Boundary condition two, Eq. (4), specifies that the gas con-
centration at the microreactor outlet is very close to zero. This
condition results from the continuous evacuation of the microre-
actor outlet by the vacuum system. The diffusivity term in the
Knudsen diffusion regime is determined by the following equation
[32]:

ε d 8RT 4εb
De = b i , di = rp . (5)
 3 M 3(1 − εb )

The flow rate, FA , at the microreactor outlet is described by the


following equation:
∂CA
FA = −ADeA |, (6)
∂z
and the gas flux by
FA
FluxA = . (7)
A
In order to solve for the gas exit flow rate, it is useful to express
Eq. (1) in dimensionless form:

∞ Fig. 16. (a) Standard diffusion curve showing key time characteristics and the
F̄A = (−1)n (2n + 1) exp(−(n + 0.5)2 2 ). (8) criterion for Knudsen diffusion. (b) Comparison of standard diffusion curve with
experimental inert gas curve over inert packed bed [14].
n=0

Eq. (8) describes the dimensionless exit flow rate as a function and
of dimensionless time. The resulting curve (Fig. 16) is known as
DeA
the standard diffusion curve. For any TAP vacuum pulse-response Hp = 1.85 . (13)
experiment that involves only gas transport, the standard diffusion εb L2
curve should be the same regardless of the gas, microreactor bed By multiplying Eq. (10) with (11) and Eq. (12) with (13) gives a
length, catalyst particle size, or reactor temperature. The area under relationship between the time at which the peak maximum occurs
the standard diffusion curve is to equal unity. In dimensional form, on a standard diffusion curve and its corresponding peak height.
Eq. (8) can be rewritten as This calculation can be used to verify that gas transport through
the reactor is Knudsen diffusion:
DeA 

FA tDeA
= (−1)n (2n + 1) exp(−(n + 0.5)2 2 ). (9) F̄A,p p = tp Hp ≈ 0.31. (14)
NpA εb L2 εb L2
n=0
5.2.2. Diffusion + irreversible adsorption/reaction
Eq. (9) indicates that the pulse shape of the curve generated in
If adsorption or reaction is first order in gas concentration, and
a diffusion only experiment should be independent of the pulse
surface coverage is negligible (the result of small pulse intensity)
intensity if the gas molecules are transported through the microre-
compared to the total amount of active catalytic material, then the
actor in the Knudsen diffusion regime.
mass balance for the gas-phase component A is given by Eq. (15):
A unique feature of the standard diffusion curve is the time, p ,
at which the curve maximum occurs is given by ∂CA ∂2 CA
εb = DeA − as Sv (1 − εb )ka CA , (15)
1 ∂ ∂z 2
p = , (10)
6 which can also be written in dimensionless form:
and the corresponding height of the peak maximum is given by ∂C̄A ∂2 C̄A as Sv (1 − εb )ka L2
= − k̄a C̄A , k̄a = . (16)
∂ ∂ 2 DeA
F̄A,p = 1.85. (11)
The dimensionless exit flow rate for irreversible adsorption or
Eqs. (10) and (11) can be rewritten in dimensional form in terms reaction combined with Knudsen diffusion process is given by
of time and height as


1 εb L2 F̄A = exp(−k̄a ) (−1)n (2n + 1) exp(−(n + 0.5)2 2 ), (17)


tp = , (12)
6 DeA n=0
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 123

Fig. 18. Comparison of standard diffusion and diffusion + reversible adsorption exit-
Fig. 17. Comparison of diffusion + irreversible adsorption/reaction exit-flow curves
flow curves. (a) Standard diffusion exit-flow curve, k̄a = 0, (b) diffusion + reversible
with the standard diffusion curve. (a) Standard diffusion exit-flow curve, k̄a =
adsorption curve, k̄a = 20, k̄d = 20 and (c) diffusion + reversible adsorption curve,
0, (b) diffusion + irreversible adsorption/reaction curve, k̄a = 3 and (c) diffu-
k̄a = 20, k̄d = 5[14].
sion + irreversible adsorption/reaction curve, k̄a = 10.

Eq. (17) can also be written in dimensional form as ∂¯ A


= k̄a C̄A − k̄d ¯ A , (24)
∂
FA DeA 

= exp(−ka t) (−1)n (2n + 1) where
NpA εb L2
n=a
¯ A = ˛A , (25)
 
DeA
× exp −(n + 0.5)2 2 , (18) and
εb L2
as Sv
˛ = (1 − εb )AL . (26)
where NpA
as Sv (1 − εb )ka
ka = . (19) Eqs. (23) and (24) can be solved using the dimensionless initial
εb and boundary conditions for the diffusion only process with the
A comparison between the dimensional forms of the exit flow for additional initial condition for the coverage of adsorbed component
the irreversible adsorption/reaction case, Eq. (18), and the standard A:
diffusion curve, Eq. (9), shows that the values of the normalized
t = 0, ¯ A = 0. (27)
exit flow for irreversible adsorption/reaction is less than the values
obtained in the diffusion only case by a factor of exp(−ka t). There- The complete solutions for the dimensionless concentration and
fore, the normalized exit-flow curve versus time for irreversible exit flow can be found in the literature [14].
adsorption/reaction is always smaller than the standard diffusion In contrast to the normalized exit-flow curve in the process of
curve (Fig. 17). Accurate extraction of kinetic constants based on diffusion + irreversible adsorption/reaction, the exit-flow curve for
comparison with the standard diffusion curve is possible only in a the diffusion + reversible adsorption crosses the standard diffusion
special domain of parameters. Time of reaction cannot be too fast exit-flow curve (Fig. 18).
or too slow in comparison with transport residence time. A special The point at which the curves intersect depends on the
analysis of this domain is in the literature [156,157]. adsorption and desorption rate constants. The diffusion + reversible
adsorption exit-flow curve is wider and crosses the standard dif-
5.2.3. Diffusion + reversible adsorption fusion curve because of the time delay in molecular transport
When diffusion + reversible adsorption occurs, and the number resulting from the interaction of gas molecules with the cata-
of gas molecules is small compared to the total number of catalyt- lyst surface. The shape and magnitude of the diffusion + reversible
ically active sites, the mass balances of component A in the gas adsorption exit-flow curve is also strongly influenced by the
phase, and on the catalyst surface are described respectively by the adsorption–desorption parameters. For example, when the adsorp-
following two equations: tion rate constant is large, and the desorption rate constant is small;
the exit-flow curve has two peaks [14]. The first peak resembles
∂CA ∂2 CA
εb = DeA − as Sv (1 − εb )(ka CA − kd A ), (20) an irreversible adsorption exit-flow curve, and is governed by the
∂t ∂z 2 interaction between diffusion and adsorption. The second peak is
∂A the result of slow desorption, and its shape is dependent on the
= ka CA − kd A . (21)
∂t parameters of adsorption, desorption, and diffusivity. Mathematical
The dimensionless desorption rate constant is defined as modeling of the two peaks shows that the catalyst surface cover-
age initially increases with time due to fast gas adsorption, then
εb L2 decreases at which time desorption becomes more significant [14].
k̄d = kd . (22)
DeA
Eqs. (20) and (21) can be rewritten in dimensionless form as 5.3. Three-zone TAP microreactor

∂C̄A ∂2 C̄A In the three-zone microreactor, the catalyst zone is sandwiched


= − k̄a C̄A + k̄d ¯ A , (23)
∂ ∂ 2 between two inert zones, and all three zones are of equal length
124 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

[14]. In the three-zone configuration the catalyst can be more easily


maintained under isothermal conditions. However, the gas concen-
tration gradient across the catalyst zone can create non-uniform
surface coverage. Also, it is difficult to analyze the transcendental
functions that contain the kinetic parameters that are solutions of
the three-zone TAP model.
The three-zone mathematical model is the same used in the one-
zone model with two additional boundary conditions between the
inert and catalyst zones. The additional boundary conditions are
given by

CA,zone1 |z1 = CA,zone2 |z1 , (28)

CA,zone2 |z2 = CA,zone3 |z2 , (29)


∂CA,zone1 ∂CA,zone2
−DeA,zone1 |z1 = −DeA,zone2 |z1 , (30)
∂z ∂z
∂CA,zone2 ∂CA,zone3
−DeA,zone2 |z2 = −DeA,zone3 |z2 , (31)
∂z ∂z
where z1 is the axial coordinate at the end of zone 1, and z2 is
the axial coordinate at the end of zone 2. Eqs. (28)–(31) describe
the continuity of the gaseous concentrations and flows inside the
microreactor.

5.4. Thin-zone TAP microreactor (TZTR)

The thin-zone TAP microreactor (TZTR) model [33,109–112] is


a three-zone configuration in which the thickness of the catalyst
zone is made very small in comparison to the whole length of Fig. 19. Schematic representation of the thin-zone TAP microreactor and the depen-
the microreactor. The advantage of the TZTR configuration is that dencies of ka and  on the axial coordinate x [32].
any change in gas concentrations across the catalyst bed can be
conv is the convectional
where kCSTR is the apparent rate constant, res
neglected when compared to their average values. Diffusional gas
transport can be explicitly separated from the chemical reaction average residence time, and
rate. conv Vcat
The TZTR mathematical model determines the active zone reac- res = . (38)
Fv
tion rate as the difference between two diffusional flow rates at the
The difference between the conversion expressions for a CSTR
boundaries of the thin catalyst (active) zone: dif
and TZTR is that the residence time found in the TZTR, res,cat , is
Rate = Flowleft (t) − Flowright (t). (32) proportional to the position of the catalyst zone inside the microre-
actor. When the catalyst zone position along the reactor axis is
Eq. (32) is analogous to a steady-state CSTR where the reaction
changed the conversion changes. For example, if it is moved closer
rate is determined by the difference between two convectional flow
to the microreactor inlet, conversion increases and closer to the
rates. For first-order irreversible adsorption/reaction, conversion in
outlet, conversion decreases. Fig. 19 shows a schematic of the TZTR
the TZTR can be found using:
and the dependence of the dimensionless adsorption constant, ka ,
dif
kads res,cat and the dimensionless surface coverage, , on the axial coordinate
X= , (33) x.
dif
1 + kads res,cat
Although concentration and temperature gradients in the cat-
where alyst zone are very small in the TZTR, some non-uniformity will
(L)LII still be present. Non-uniformity can be attributed to two fac-
dif
res,cat = εb . (34) tors: the applied concentration gradient, which drives diffusion
DeA
and is present even when no reaction occurs, and chemical reac-
The apparent adsorption/reaction rate constant, kads , can be tion, which changes the concentration profile in the catalyst zone
found from the relationship: [109–112]. These factors are taken into account in the following
DeA equation:
kads = Ka , (35)
LLεb Cin − Cout Lc X
≈2 + , (39)
where the dimensionless parameter Ka can be calculated from the Cin Lr 1 + (1 − X)Lr /Lc
zeroth moment: The first term on the right-hand side of Eq. (39) relates only to
1 LII the geometric configuration of the reactor, and the second relates
= 1 + Ka . (36)
M0 L to the influence of chemical reaction and geometry. As with a differ-
ential PFR, non-uniformity in a TZTR is proportional to conversion,
The zeroth moment is obtained by measuring the area of a pulse
but the proportionality is not linear. Fig. 20 shows a comparison of
response curve. Moment based analysis of TAP pulse response data
non-uniformity in the TZTR and a PFR as a function of increasing
is described in the literature [14,34]. Eq. (33) is analogous to the
conversion.
conversion expression for a first-order reaction in a CSTR:
Over a wide range of conversions (up to 75%), the level of non-
conv
kCSTR res uniformity in a TZTR is not higher than 20%, and only becomes
XCSTR = conv , (37)
1 + kCSTR res significant at conversions greater than 80%.
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 125

cle. If the residence time of gas molecules inside the microreactor is


long, then that molecule has a high probability of spending a longer
time in proximity to the active catalyst particle. Therefore, conver-
sion of gas molecule A is likely to occur. From another standpoint, if
the gas residence time in the microreactor is long, the gas molecule
has a higher probability of contacting or come in close proximity
to the active particle through a longer random walk before exiting
the microreactor. A correlation between the numerical model and
the experimental results explains that due to the high number of
random gas collisions prior to exiting the microreactor, there is a
greater probability for the reactant gas molecules to come into con-
tact with the active catalyst particle and convert to the product (i.e.,
higher conversion). Detailed results of the numerical calculations
can be found in the paper [18].

6. TAP experimental results


Fig. 20. Comparison of non-uniformity versus conversion for the differential PFR
6.1. Bridging the pressure gap
and TZTR. Different ratios of entire microreactor length to length of catalyst zone
are given for the TZTR [109–112].
With the TAP reactor system, a catalyst sample can be rapidly
cycled between vacuum and atmospheric pressure without expos-
5.5. Probabilistic theory of single particle TAP experiments
ing it to the atmosphere. The process is illustrated in Fig. 22, which
shows the slide valve in the high-pressure (a) and vacuum positions
The single particle experiment (see Section 6.1 for a detailed
(b).
description) is the most recent addition to TAP microreactor config-
In the high-pressure position, the reactor effluent flows through
urations. With the single particle microreactor configuration, even
the slide valve and out an external vent where it can be ana-
higher conversions (up to 95%) can be achieved compared to the
lyzed using a GC (see Section 4.3). A small portion of the effluent
80% conversion obtained in a TZTR to ensure uniformity in the active
can be diverted to the mass spectrometer chamber, through an
zone.
adjustable needle valve. The mass spectrum of the reaction prod-
A two-dimensional numerical model based on probabilistic the-
ucts is collected in real-time. After running atmospheric pressure
ory and interpretation of conversion based on the principle of
experiments the reactor can be evacuated, and the slide valve
Brownian motion of reactant molecules inside the microreactor is
moved to the vacuum position. During the switch from high pres-
created to understand the trajectory and transport of reactant gas
sure to vacuum the reactor effluent can be monitored and desorbing
molecules in the single particle experiment.
adspecies left on the surface during pressure experiments can be
The numerical experiment is modeled in which a pulse of gas of
measured. TAP pulse-response experiments are performed after
chemical species A is released into the microreactor and the out-
the reactor reaches vacuum, and the desorption of adspecies has
let flow is collected at the right-hand side of Fig. 21. The black disc
stopped. Switching back and forth between high-pressure and vac-
represents the catalyst particle and the light shaded dots repre-
uum operation typically takes less than 30 s.
sent the inert medium used to pack the microreactor. The black
Recently, we reported results from atmospheric pressure and
line inside the microreactor represents a possible trajectory route
vacuum pulse-response experiments on the oxidation of CO [17].
for a reactant gas molecule prior to exiting the microreactor.
The catalyst sample was a single 400 ␮m diameter polycrystalline
The numerical model states that if a reactant gas molecule A
platinum (Pt) particle, which was placed in a microreactor bed with
remains near or comes into contact with an active catalyst particle
≈100,000 inert quartz particles with diameters between 210 and
during the course of its random motion in the packed bed of the
250 ␮m.
microreactor for a certain period of time, then the probability that
Fig. 23 presents a scale drawing of the reactor configuration. The
a reaction will occur during that time period is 1. If a reactant gas
particle occupies less than 0.3% of the cross-sectional area of the
molecule A is far away from the active catalyst particle, then the
microreactor, so the reaction zone can be considered a point source.
probability of conversion is 0 regardless of how long gas molecule
Gas concentration or temperature gradients across the catalyst zone
A stays in that position.
can be assumed to be negligible since the zone is a single particle.
The numerical model can also determine conversion in terms
Vacuum pulse-response experiments were performed using a
of the molecular residence time in or near the active catalyst parti-
“pump-probe” format illustrated in Fig. 24. Oxygen/Ar and CO/Ar
mixtures are injected from different pulse valves in an alternating
sequence, and the CO2 response is measured during each pulse.
Argon is used as an internal standard. CO2 does not appear on the
first oxygen pulse since no CO is present in the reactor. CO2 appears
on the first CO and all subsequent oxygen pulses.
At 170 ◦ C the yield for the individual O2 and CO pulses reaches a
maximum, making the total yield equal to 95% during one pump-
probe cycle indicating that at least 95% of CO molecules pulsed
into the reactor must strike the particle. Above 170 ◦ C the CO2 yield
decreases, more rapidly for the O2 pulse than the CO pulse.
Atmospheric flow experiments were performed after closing the
slide valve. The particle bed was first exposed to a hydrogen flow
Fig. 21. Model of microreactor with catalyst (black dot) and inert quartz particles
(20 cc/min diluted in Ar, H2 /Ar = 1) at 350 ◦ C for 1 h. Hydrogen was
(blue dots). The dashed line at the right indicates the microreactor outlet. (For inter-
pretation of the references to color in this figure legend, the reader is referred to the used to remove any memory of the previous pump-probe experi-
web version of the article.) ments. A total flow of 50 cc/min of O2 , CO, and Ar (O2 /CO/Ar = 1) was
126 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

Fig. 22. Illustration of high-pressure (a), and vacuum (b) configurations of TAP reactor system. The microreactor can be shifted from high-pressure operation to vacuum in
minutes by opening the slide valve.

introduced through the continuous valve, giving a gas residence and then decreased at the same ramp rate to room temperature.
time in the reactor of 1.8 s. A small amount of the reactor effluent was diverted into the mass
The temperature dependence of CO2 production was obtained spectrometer chamber, and its mass spectrum was continuously
by heating or cooling the reactor at a constant rate while maintain- monitored.
ing an input flow of 50 cc/min. The internal reactor temperature Both TAP vacuum and atmospheric flow data exhibit a turning
was ramped from 40 to 430 ◦ C over a 40 min interval. Upon reach- point in CO2 production indicating a transition from reaction con-
ing 430 ◦ C, the reactor temperature was held constant for 5 min trolled by one adsorbed species to one controlled by a different

Fig. 23. (a) Schematic of TAP single particle microreactor configuration. The 400 ␮m diameter Pt particle is packed within a sea of inert quartz particles with diameters
between 210 and 250 ␮m. (b) Image comparing a 400 ␮m Pt particle to a pencil point. (c) SEM image showing the complex surface structure of a polycrystalline Pt particle.
(d) Higher magnification (15,000×) of the particle shown in (c), which shows the surface is non-porous [17].
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 127

Fig. 24. Illustration of a TAP pump-probe experiment in which O2 /Ar and CO2 /Ar are pulsed in an alternating sequence and the CO2 transient response is measured during
each pulse.

adsorbed species at 170 ◦ C. The upper branch in the CO2 curve of ments can be used to describe kinetic behavior in the atmospheric
the atmospheric flow experiment corresponds to an O2 covered sur- pressure domain.
face, and the lower branch corresponds to a CO covered Pt surface Taking the conversion of CO or the CO2 yield at the “turning
(Fig. 25(b)). From the pulse response data, in the region of the CO2 point” from vacuum pulse response and atmospheric flow data, the
maximum, the areas under the CO2 response curves (CO2 yield) cor- apparent kinetic rate constant can be calculated. In combination
responding to the O2 and CO pulses are approximately the same, with the gas residence time () in the catalyst zone, the apparent
indicating nearly equal coverages of O2 and CO (Fig. 25(c)). The kinetic rate constant is given by the following expression:
correspondence in “turning points” indicates that the coverage in
vacuum and atmospheric pressure experiments is approximately X
kapparent = . (40)
the same and the intrinsic kinetic data obtained in vacuum experi- (1 − X)

Fig. 25. Comparison of CO2 produced during TAP vacuum pump-probe experiments and atmospheric flow experiments for CO oxidation over single Pt particle with the
same composition of reactants. (a) A typical set of pump-probe CO2 responses (m/e = 44) for reaction at 140, 170, and 350 ◦ C. There is a shift in the amount of CO2 produced
during both CO and oxygen pulses as temperature increases. (b) CO2 production observed from atmospheric flow experiment. The CO2 produced while increasing reactor
temperature is less than the CO2 produced during reactor temperature decrease as shown by the counter-clockwise hysteresis loop. (c) CO2 production observed from vacuum
pump-probe experiment. The black line represents the total CO2 yield. The red and blue lines represent the CO2 yield on the oxygen pulse and CO pulse, respectively. (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
128 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

Fig. 26. (a) Oxygen uptake over reactor-equilibrated VPO at 105 Pa O2 and (b) O2 desorption spectrum under vacuum from 18 O2 -treated (VO)2 P2 O7 .

Using an approximate conversion of 90% in the vacuum VOPO4 phases [56–60]. The oxidation of n-butane, and other C4
pulse-response experiment, the apparent kinetic rate constant is compounds was investigated in TAP pulse-response experiments
calculated to be 9000 s−1 . In the atmospheric flow experiment, by pulsing a C4 /Ar mixture over VPO and measuring the pulse
using a conversion of 20% the apparent kinetic rate constant is response curves of the reactants and products. Fig. 27 shows typical
calculated to be 9280 s−1 . The two apparent constants differ by response curves and an Arrhenius plot for n-butane oxidation over
approximately 3%. Although there may be some error involved in an oxygen-treated catalyst.
experimentation, the values of the apparent kinetic rate constants The change in kinetic parameters as the surface oxygen con-
are approximately the same. The ability to relate data both qual- centration is altered is shown in Fig. 28, which plots the activation
itatively and quantitatively in the atmospheric pressure domain energy and apparent equilibrium constant for different products
to the data obtained in vacuum pulse-response experiments using at different stages in the reduction of VPO. The mechanistic impli-
a single Pt particle is a step toward bridging across the pressure
gap.

6.2. Tracking the evolution of catalytic properties

Structure–activity correlations can be established using a sur-


face science strategy and measuring reactions on a well-defined
model surface. Another approach is based on the application of
in situ spectroscopic techniques, which try to measure surface
adspecies or identify surface structures that change during reaction.
Surface spectroscopy techniques can also be used to characterize
catalyst samples before and after reaction. In this case a change in
catalyst performance can often be associated with a change in sur-
face composition of structure. Transient kinetic experiments can
also be used to indirectly measure changes in the composition and
structure while simultaneously measuring changes in kinetic prop-
erties. This latter approach is the one adopted in TAP reactor studies,
and is illustrated by the following examples.

6.2.1. C4 oxidation over VPO catalysts


Catalysts based on vanadium oxides are used extensively
in selective oxidation processes [56–60,150–153]. The selective
oxidation of n-butane and other C4 molecules over vanadyl
pyrophosphate (VPO)-based catalysts to maleic anhydride and
intermediate compounds (e.g., furan, butadiene and butene) is
strongly influenced by the feed conditions, especially the oxy-
gen to hydrocarbon ratio. A combination of atmospheric pressure
and TAP pulse-response experiments determined that “reactor-
equilibrated” VPO adsorbs oxygen at elevated oxygen pressures to
form a more active-selective catalyst [56–60]. If the oxygen-treated
catalyst is heated in vacuum the adsorbed oxygen desorbs leaving
a less active-selective catalyst. Fig. 26 shows three oxygen uptake
curves at 430, 450 and 460 ◦ C and an oxygen pressure of 105 Pa,
and the oxygen desorption spectrum when an oxygen-18 treated
VPO sample is heated in vacuum. The O16 to O18 ratio indicates
that oxygen is only taken up in the first few monolayers of the VPO
Fig. 27. (a) n-Butane pulse response curves over oxygen-treated VPO at various
surface.
temperatures. Each curve is obtained for the same initial VPO oxidation state. (b)
Oxygen-treated VPO catalysts were determined to be more Arrhenius plot obtained from the temperature dependence of the n-butane conver-
active and selective provided the oxidation did not lead to less active sion giving an activation energy of 12 kcal/mol.
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 129

spectrum is shown in Fig. 30 for a sample prepared by exposing the


Pd target to 750 Pd laser pulses.
A maximum in CO2 production is observed at 154 ◦ C. After the
maximum, CO2 production drops rapidly as the sample continues
to heat. The initial steep drop in CO2 production is followed by a
gradual decrease with periodic bursts giving the appearance of a
damped oscillation. All freshly deposited samples exhibited this
oscillatory behavior. The trend in CO2 production and the total
amount produced was highly reproducible on separate samples
prepared with identical Pd loadings. Reoxidation of sample fol-
lowed by reduction did not yield oscillations.
Heuristically, every peak in the dependence in Fig. 30 can be
viewed as a typical TPR peak related to a specific form of catalyst
oxygen. CO2 production during the first peak corresponds to the
depletion of accessible surface oxygen. Then, the series of peaks
that follow can be considered as resulting from reaction of CO with
other forms of catalyst oxygen. Thus, abrupt changes in the amount
of oxygen indicate abrupt changes in the Pd/PdO composition. The
observed phenomenon can be interpreted as an example of reactive
self-assembly, the combination of reaction kinetics, diffusion, and
surface phase transitions [94,149].

6.3. Surface lifetimes of reactive species

The surface lifetime of an adspecies under reaction condi-


tions is a function of the rate of reaction, the rate of desorption,
and the rate at which the adspecies diffuses into the catalyst
bulk. The adsorption–desorption characteristics of a species can
be determined in TAP pulse-response experiments by comparing
the exit-flow curve of the species with the standard diffusion curve
(STD) (see Sections 5.2.2 and 5.2.3). If the curve falls inside the STD
then the species is irreversibly adsorbed. The surface concentration
of an active species can also decrease if the species diffuses into the
catalyst bulk or is depleted by reaction with some other surface
species. The reactive lifetime of an adspecies can be measured in
TAP pump-probe experiments by changing the pump-probe inter-
val.
The reactive lifetime of oxygen on a Pt particle was measured
in a series of pump-probe experiments using two separate reactant
Fig. 28. (a) Arrhenius plots for n-butane oxidation over a single VPO sample reduced mixtures of O2 /Ar (70/30 ratio) and CO/Ar (70/30 ratio), which were
by a long series of n-butane pulses. As the surface is reduced the activation energy injected from two separate pulse valves. The mixtures were pulsed
increases and (b) apparent equilibrium constant for various products as a function
in an alternating sequence into a microreactor containing a single
of VPO oxidation state.
400 ␮m Pt particle packed in a bed of inert quartz particles. The
interval separating the oxygen pulses and the CO pulses was varied
cations of the data presented in Figs. 27 and 28 are presented in between 1 and 9 s. In all cases the pump-probe cycle time was 10 s.
Section 7.1.3. Thus when the interval between the oxygen and CO pulse is 5 s,
the interval between the CO and following oxygen pulse is also 5 s.
6.2.2. Reactive formation of metal deposits Fig. 31 shows the CO2 production for different pump-probe intervals
Atomic deposition under ultrahigh vacuum conditions offers a at 150 and 350 ◦ C.
precise method of delivering metal atoms to a solid surface. Fig. 29 CO2 production on the oxygen pulse at 350 ◦ C is significantly
shows an atomic beam deposition (ABD) apparatus coupled to a TAP lower than production at 150 ◦ C. It is independent of the pump-
reactor system. The ABD system produces metal atoms by focusing probe interval. CO2 production on the CO pulse at 150 ◦ C is also
light from a pulsed excimer laser onto the surface of a metal (pal- independent of the pump-probe interval. At 350 ◦ C, CO2 decreases
ladium) target. The energy from the pulse ejects a spray of atoms. with the pump-probe interval. The drop in CO2 production can be
Directly beneath the target is a cylindrical sample holder, which has attributed to a decrease in the amount of active oxygen. The rate
a magnetic diaphragm for vibrating a bed of support particles. Send- of the drop in active oxygen can be calculated from the zeroth
ing a pulsed current through an electromagnetic coil causes the moments of the CO2 and is plotted in Fig. 32.
diaphragm to vibrate, continuously agitating the particles to pro-
duce a uniform coating of metal atoms. After deposition, samples 7. Unraveling complex catalytic mechanisms using TAP
are transferred to a TAP-2 reactor and kinetically characterized. pulse-response experiments
In our experiments, we deposited different loadings of Pd atoms
onto 210–250 ␮m diameter inert quartz (SiO2 ) particles. The parti- 7.1. Qualitative information on mechanisms
cles were then reduced with a long series of CO pulses while heating
the sample in a linear ramp. TAP transient response data provides the identity, amount, and
CO2 production resulting from the oxidation of CO/reduction of time dependence of the flow of different species exiting the TAP
Pd displayed an interesting oscillatory behavior. The resulting TPR microreactor. The response data is a sensitive function of the trans-
130 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

Fig. 29. Schematic of atomic beam deposition system coupled to a TAP-2 reactor system [94].

port process, the temperature, and gas–solid interactions. It is oping structure–activity relationships. Constructing a detailed
standard practice to use an inert gas (typically argon) as an inter- mechanism requires knowledge of the different species (reactants,
nal reference. At reaction temperatures, the total time inert gas surface complexes, catalyst components, desorbing intermediates,
molecules spend interacting with the solid is negligible, and the and products) participating in the overall reaction, and the time
inert gas response can be assumed to be solely a function of the gas
transport process and temperature. Under Knudsen flow conditions
an inert gas response is described by the “standard diffusion curve”
(see Section 5.2.1). The standard diffusion curve for any gas with
a unique molecular weight at any temperature can be constructed
from the inert gas response. Deviation from the standard diffusion
curve is evidence of a gas–solid interaction.
Catalytic reaction mechanisms represent the sequence of
molecular level processes that occur during a single catalytic cycle.
Processes typically included in a mechanism are reactant adsorp-
tion and desorption, surface diffusion, surface reaction, and product
desorption and re-adsorption. The mechanism may also include
structural and compositional changes that occur in the catalyst.
Experimental data on such changes are very useful for devel-

Fig. 31. Pump-probe data showing CO2 production as a function of temperature and
pump-probe interval. At 150 ◦ C the CO2 production is essentially independent of the
Fig. 30. Normalized CO2 production (determined by measuring the zeroth moment pump-probe interval out to 9 s separation. At 350 ◦ C the CO2 production drops as the
of the CO2 responses) over fresh Pd deposits (750 pulses) obtained by pulsing CO pump-probe interval increases indicating a drop in the active oxygen concentration
and ramping the reactor temperature from 32 to 400 ◦ C [94]. with time.
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 131

following reaction mechanism was proposed [32,160]. Methanol


dehydrates to dimethyl ether, which reacts further to olefins. The
production of dimethyl ether proceeds via the dissociative adsorp-
tion of methanol which forms water and a methoxy group. The
methoxy group reacts with methanol to produce adsorbed dimethyl
ether. Propylene and ethylene are formed from dimethyl ether and
involve a common surface intermediate, which can decompose to
form ethylene or react with adsorbed methanol to give propylene
and water. Higher olefins are produced via reaction of lighter olefins
with methoxy groups on the surface.

7.1.2. Pump-probe information


TAP pump-probe experiments provide a means of adding reac-
tants to the surface of a catalyst in sequence, and precisely
controlling the time interval between reactant inputs. Typically,
Fig. 32. Normalized CO2 production on the CO pulse calculated from the zeroth the gas-phase component of one reactant is no longer present in
moment of the pulse response curve. CO2 production is constant at 150 ◦ C, and drops the reactor when the other reactant is introduced. Pump-probe
to ≈0.5 times its value in 9 s at 350 ◦ C.
experiments can be used to study processes that involve rapid
changes in the concentration of a reactive adspecies. The use of
of their appearance in the sequence of steps. TAP experiments
isotopes can help distinguish between mechanistic routes that pro-
provide direct information on reactants, products, and desorbing
duce the same product. The pump-probe format can also be used
intermediates, and indirect information on surface complexes and
to distinguish processes that involve bulk diffusion from surface
catalyst components. Different types of TAP experiments provide
processes, and can provide information on the reactive formation
unique information on the sequence of steps in the overall reac-
of catalytic sites. Pump-probe experiments were used to study the
tion, on the lifetime of surface adspecies, the diffusion of catalyst
activation of silver powder, and subsequent epoxidation of ethy-
components, the influence of different surface concentrations on
lene [95], in a TAP-1 reactor using a one-zone reactor. Experiments
the rate and energetics of various surface processes.
involved injecting an alternating sequence of oxygen and ethylene-
The types and sequence of TAP experiments performed during
d4 pulses into the microreactor using a fixed pump-probe interval
a catalytic study are often motivated by questions that naturally
and monitoring the CO2 and ethylene oxide yield as a function of
arise during the construction of a reaction mechanism. Interac-
the O2 /C2 D4 ratio. Experiments were also performed using fixed
tion between the experiment and the researcher can be viewed
amounts of reactants at different pump-probe time intervals and
as “conversational”, i.e., an experiment, based on a simple question
monitoring C2 D4 O as a function of the pump-probe interval. In
provides data, that raises a new question, which in turn suggests a
experiments using a fixed pump-probe interval, ethylene oxide
new experiment, and so on. The sequence of questions and answers
production was observed during the ethylene pulse, but not dur-
can follow one another rapidly because a single TAP pulse response
ing the oxygen pulse. CO2 production occurred during both pulses.
typically takes only a few seconds.
The amount of both products increased nearly linearly with the
size of the oxygen pulse for a O2 /C2 D4 ratio below 0.6 and was
7.1.1. Single pulse information nearly independent of the O2 pulse size for high O2 /C2 D4 ratios.
Initial data on the adsorption/desorption of pure gases (reac- The results indicate that the rate of ethylene oxidation depends
tants, products and intermediates) over catalytic and inert on the concentration of adsorbed oxygen and that selective oxi-
materials can be obtained by comparing the shape of their pulse dation to ethylene oxide involves the reaction of ethylene with
response curves to a standard diffusion curve. Simple qualitative adsorbed oxygen species. Comparison of the ethylene and ethy-
analysis based on well-defined theoretical patterns [14,32] (see lene oxide pulse shapes indicate no measurable reaction time, as
Sections 5.2.2 and 5.2.3) can be used to compare the adsorption both species appear simultaneously at the reactor exit. The surface
strength of different molecules, to determine if an adsorption pro- lifetime of active oxygen species was investigated by measuring
cess is reversible or irreversible, and to determine if molecules the C2 D4 O product yield as a function of the pump-probe interval.
compete for the same adsorption sites. Changes in the pulse shape The maximum C2 D4 O yield occurred at a large time interval, when
of a single species over a series of pulses can reveal how the surface gas-phase oxygen is close to zero, and the amount of adsorbed oxy-
properties of a catalyst change during reaction. For example, puls- gen is at a maximum. The results were interpreted to mean that
ing CO or a hydrocarbon over a mixed-metal oxide or metal catalyst the formation of the active oxygen species involves a slow step in
pretreated by oxygen will reveal how adsorption properties change which atomically adsorbed oxygen becomes associated with a par-
as the surface oxygen is depleted. tially oxidized silver cluster, and that the active oxygen has a long
Qualitative comparison of reactant and product pulse shapes can surface lifetime. The experiments provided clear evidence that the
be used to construct a multistep reaction sequence. For example, in active oxygen species are atomically adsorbed oxygen species, and
TAP-1 studies using a three-zone reactor, the reaction of methanol not molecularly adsorbed oxygen.
over an H-ZSM-5 zeolite [32,160] produced olefins (especially Experiments involving the oxidation of CO over a Pt catalyst (see
propylene), water, dimethyl ether, formaldehyde and methane. Section 6.3), using different pump-probe intervals indicate that the
Compared to the standard diffusion curve, the methanol and prod- active surface oxygen is rapidly depleted at higher reaction temper-
uct responses were all very broad. From shifts in the pulse maxima, atures immediately after an oxygen pulse. The disappearance can
the pulse widths, and pulse decay curves the methanol adsorption be attributed to oxygen exchange between the catalyst surface and
process and product formation sequence was deduced. The shape bulk.
of the methanol pulse indicates it spends a significant amount of In general, the broadening of the pulse response of a product
time on the catalyst, but is reversibly adsorbed. The water response in comparison to a reactant is evidence of one or more surface
is similar to the methanol response indicating a rapid exchange intermediates participating in a complex surface reaction. The
process and that water adsorption is similar to methanol. Based on dependence of the exit flow on the pump-probe interval indicates
the relative shapes of the remaining product response curves the a parallel process that depletes one of the surface reactants.
132 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

7.1.3. Multi-pulse TAP experiments and discriminated using these expressions. In the case of the
TAP experiments usually involve a series of single or alternat- global-transfer matrix approach, the general theory of the pulse-
ing pulses containing fixed concentrations of reactants and inert response–pulse experiment for the multi-zone configuration has
reference gas. A single reactant can be used to produce a con- been described. The theory offers an efficient means to compute
trolled change in the catalyst surface composition. For example, the actual profiles of gas and surface concentration in the reac-
CO, hydrogen, or a hydrocarbon can be pulsed over a mixed-metal tor as well as the corresponding values at the reactor exit using
oxide to deplete the surface oxygen, and change the oxidation state the fast Fourier transform. For the thin-zone TAP reactor (TZTR) in
of surface metal ions. At the same time the transient responses of which the gas and catalyst composition are nearly uniform, the the-
reactants and products over the course of the reduction can be used oretical expressions are very simplified (see Section 5.4 [109,112]).
to monitor changes in reaction products and in catalytic properties. In this case, every substance can be characterized by three appar-
Reduction of a mixed-metal oxide strongly influences its adsorption ent parameters obtained using three observed moments, i.e., the
properties, and can cause a change in the reaction mechanism. Com- apparent kinetic constant, the time delay, and ‘equilibrium con-
parison of changes in catalytic properties with changes in surface stant’ [33,163]. These expressions have been used for developing
composition provides information to develop activity–structure or a new strategy for kinetic characterization of states in a TAP reac-
activity–composition relationships, and to construct models of the tor with multi-pulse responses that is called “state-by-state-kinetic
active catalytic site. screening”. This strategy was illustrated using furan oxidation
TAP multipulse experiments using a TAP-2 reactor system with over a VPO catalyst [33]. Essential information on the apparent
three-zone and thin-zone microreactor configurations were used parameters as a function of the catalyst oxidation/reduction degree
to investigate the dependence of kinetic parameters for hydrocar- and comparative analysis of these parameters create a basis for
bon oxidation on VPO catalysts as a function of the oxygen surface developing a detailed reaction mechanism. For example, calcula-
concentration [170,171]. Typical experimental results are shown tions demonstrate that apparent kinetic constants for all products
in Figs. 27 and 28 in Section 6.2.1. Catalysts were initially heated (maleic anhydride, acrolein, and CO2 ) are clearly different at all oxi-
in an oxygen atmosphere, and the oxygen uptake was measured dation/reduction degrees. Thus, the reaction routes for all products
as a function of temperature and O2 partial pressure. Oxygen was should be distinguishable. The non-linear dependence of apparent
readily adsorbed at temperatures above 400 ◦ C, in amounts total- kinetic constants on the oxidation degree is explained by assuming
ing 1–3 monolayers (see Fig. 26). The resulting “oxygen-treated” a reaction with participation of catalyst oxygen, particularly sub-
catalyst was then reduced with a hydrocarbon (e.g., n-butane, surface oxygen. Time delays for all mentioned products are clearly
butene, butadiene and furan) in series of anaerobic pulses at dif- distinguished as well, and the number of distinguished time delays
ferent temperatures (see Fig. 27a). Arrhenius plots obtained from indicates the number of surface intermediates. The proposed mech-
the pulse response data for different VPO oxidation states is plotted anism must include at least three independent routes involving at
in Figs. 27b and 28a. The data shows that the number of active sites least four surface intermediates.
or active oxygen species decreases as the catalyst is reduced and Recently, a new procedure, the so-called Y-procedure [148], was
the activation energy for n-butane conversion increases. The con- developed for determining the gas concentration and reaction rate
version increases suddenly without a significant decrease in the in the active zone of the thin-zone TAP reactor (TZTR) without mak-
surface oxygen concentration indicating the reaction shifts to a dif- ing any assumptions on the detailed mechanism and corresponding
ferent active site (active oxygen species). The rapid change can be kinetic dependence. Hence, the method is essentially independent
explained as follows: energy of activation changes linearly with the of the kinetic model. The mathematical basis of this procedure is
change in surface composition and therefore, the kinetic constant a Laplace-domain analysis of two inert zones in a TZTR followed
and conversion change exponentially with the composition as well. by transposition to the Fourier-transform domain. When combined
Similar experiments were performed with other hydrocarbons, and with time discretization and filtering, the Y-procedure leads to an
it was found that both the product spectrum and kinetic parame- efficient and practical method for reconstructing the kinetic depen-
ters (Fig. 28b) were a strong function of the hydrocarbon reactant dences in the catalyst zone. It also can be considered as a basis for
and the VPO oxidation state. During butane oxidation experiments advanced software for non-steady-state kinetic data interpretation
it was found that reaction intermediates such as butene, butadiene, and detail mechanism development.
and furan formed and desorbed at lower surface oxygen concentra- In summary, a variety of mathematical techniques have been
tions, their production at higher surface oxygen concentrations was developed for parameter estimation from theoretical models of the
suppressed. transport-kinetics interactions that occur in the TAP reactor using
pulse response data. Some of these are traditional methods that
7.2. Quantitative information on mechanisms have a well-established history in the transient response literature
for catalytic systems. Newer tools that recognize the unique oper-
Models that describe the transport-kinetic processes that typi- ating features of the TAP reactor, such as the “Y-procedure”, have
cally occur in TAP pulse-response experiments are parabolic partial also been developed and provide a simple yet robust approach for
differential equations (see Section 5). Typically, the model parame- extracting fundamental kinetic and other surface-related parame-
ters are extracted from the pulse response data by using parameter ters. It is expected that the increased development of new software
estimation methods. A review of different methods for parameter tools for solution of partial differential equations as well as user
estimation from TAP pulse response data was recently presented by interfaces will lead to a new generation of methods that allow
Schuurman [162]. Different approaches based on analytical solu- robust analysis to be performed using a large range of TAP pulse
tions have been developed during the last decade and include the response data sets.
moment-based approach [32–34] and the global-transfer matrix
approach [118,119,123,124]. 8. Summary
The moment-based approach is well established and com-
pares the analytical expressions for the zeroth, first and second The TAP reactor system provides the catalyst scientist with a
moments of the model-based pulse response to the moments that unique capability to unravel the chemistry, kinetics, and mecha-
are evaluated from the TAP response data. Parameters such as the nisms of gas–solid catalyzed reactions using a variety of transient
effective diffusivity and kinetic constants can be obtained. Dif- response protocols. These protocols and the associated knowledge
ferent mechanisms and corresponding models can be compared have been developed over a period of 20+ years, and are based
J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134 133

upon several hundred man-years of research effort. The develop- [22] M. Valden, J. Aaltonen, E. Kuusisto, M. Pessa, C.J. Barnes, Surf. Sci. 307–309
ment of various commercial versions of the TAP reactor system (1994) 193–198.
[23] M.P. D’Evelyn, R.J. Madix, Surf. Sci. Rep. 3 (1984) 413.
has allowed a wide cross-section of both academic and industrial [24] J.A. Barker, D.J. Auerbach, Surf. Sci. Rep. 4 (1985) 1.
research groups to use the technology on a variety of challenging [25] M. Asscher, G.A. Somorjai, G. Scoles (Eds.), Atomic and Molecular Beam Meth-
applications. These collective efforts have created new fundamental ods, vol. 2, Oxford University Press, 1988.
[26] M.L. Yu, L.A. Delouise, Surf. Sci. Rep. 19 (1994) 285.
and practical knowledge on complex industrial catalysts for a vari- [27] C.T. Rettner, D.J. Auerbach, J.C. Tully, A.W. Kleyn, J. Phys. Chem. 100 (1996)
ety of important chemistries, some of which have been highlighted 13021.
in the paper. In closing, it is useful to mention some research thrusts [28] A.W. Kleyn, Chem. Soc. Rev. 32 (2003) 87.
[29] A.W. Kleyn, D.P. Woodruff (Eds.), The Chemical Physics of Solid Surfaces (Sur-
in TAP technology that will allow a broader cross-section of catalyst face Dynamics), vol. 11, Elsevier, Amsterdam, 2003.
scientists to address emerging challenges in catalytic science and [30] J. Libuda, H.-J. Freund, J. Phys. Chem. B 106 (2002) 4901.
technology, such as those associated with clean energy systems and [31] J. Libuda, Chem. Phys. Chem. 5 (2004) 625.
[32] G.S. Yablonsky, M. Olea, G.B. Marin, J. Catal. 216 (2003) 120–134.
the environment. In terms of advanced analytical capabilities, cou-
[33] S.O. Shekhtman, G.S. Yablonsky, J.T. Gleaves, R. Fushimi, Chem. Eng. Sci. 58
pling of the TAP system to a broader range of mass spectroscopy (2003) 4843–4859.
and spectroscopic instrumentation will provide simultaneous col- [34] G.S. Yablonskii, S.O. Shekhtman, S. Chen, J.T. Gleaves, Ind. Eng. Chem. Res. 37
lection of real-time transient and steady-state response data on (1998) 2193–2202.
[35] D.Z. Wang, O. Dewaele, G.F. Froment, J. Mol. Catal. A: Chem. 136 (1998) 301.
both gas-phase and surface species. Coupling of the TAP system [36] O.P. Keipert, M. Baerns, Chem. Eng. Sci. 53 (1998) 3623.
to an angular reflectron time-of-flight mass spectrometer has been [37] O. Dewaele, D.Z. Wang, G.F. Froment, J. Mol. Catal. A: Chem. 149 (1999) 263.
demonstrated. The concept of a Coupled Array is also under devel- [38] O. Dewaele, G.F. Froment, Appl. Catal. A 185 (1999) 203.
[39] T.A. Nijhuis, L.J.P. Van Den Broeke, M.J.G. Linders, M. Makkee, F. Kapteijn, J.A.
opment and will allow the TAP system to serve as a type of catalyst Moulijn, Catal. Today 53 (1999) 189.
analytical “engine” that can be directly coupled to other systems [40] A.H.J. Colaris, J. Hoebink, M. de Croon, J.C. Schouten, AIChE J. 48 (2002) 2587.
that either generate catalyst particles with unique properties, or [41] J.A. Delgado, T.A. Nijhuis, F. Kapteijn, J.A. Moulijn, Chem. Eng. Sci. 59 (2004)
2477.
which provide advanced tools for catalyst characterization. Cou- [42] C. Breitkopf, J. Mol. Catal. A: Chem. 226 (2005) 269.
pling will be accomplished using special-purpose particle transport [43] E.V. Kondratenko, O. Buyevskaya, M. Baerns, J. Mol. Catal. A: Chem. 158 (2000)
systems and robotic units that will allow small aliquots of cata- 199.
[44] B. Silberova, R. Burch, A. Goguet, C. Hardacre, A. Holmen, J. Catal. 219 (2003)
lyst particles (on the order of tens of milligrams) to be transferred 206.
back and forth between user-selected array elements and the TAP [45] O. Dewaele, G.F. Froment, J. Catal. 184 (1999) 499.
system engine. Examples of these array elements might include [46] E.V. Kondratenko, O.V. Buyevskaya, M. Soick, M. Baerns, Catal. Lett. 63 (1999)
153.
catalyst surface modification systems, catalyst synthesis worksta-
[47] O.V. Buyevskaya, M. Rothaemel, H.W. Zanthoff, M. Baerns, J. Catal. 146 (1994)
tions, catalyst activity testing systems, catalyst libraries, and even 346.
lab or pilot-scale reactors involving moving solids, such as hot flu- [48] O.V. Buyevskaya, M. Baerns, Catal. Today 21 (1994) 301.
idized beds. A special-purpose automation and control system will [49] G.A. Martin, C. Mirodatos, Fuel Process. Technol. 42 (1995) 179.
[50] D.J. Statman, J.T. Gleaves, D. McNamara, P.L. Mills, G. Fornasari, J.R.H. Ross,
allow catalyst scientists located at a remote location to define and Appl. Catal. 77 (1991) 45.
conduct TAP experiments using one of several strategically located [51] E.P.J. Malens, J.H.B.L. Hoebink, G.B. Marin, Stud. Surf. Sci. Catal. 81 (1994) 205.
Coupled Array systems. A suite of data modeling tools will allow fun- [52] O.V. Buyevskaya, D. Wolf, M. Baerns, Catal. Lett. 29 (1994) 249–260.
[53] O.V. Buyevskaya, K. Walter, D. Wolf, M. Baerns, Catal. Lett. 38 (1996) 81.
damental kinetic and other related parameters to be extracted from [54] M. Soick, O. Buyevskaya, M. Hohenberger, D. Wolf, Catal. Today 32 (1996) 163.
the data. This approach will permit them to focus on experimen- [55] M. Fathi, A. Holmen, F. Monnet, Y. Schuurman, C. Mirodatos, J. Catal. 190 (2000)
tal aspects, and also enhance interactions between other catalyst 439.
[56] U. Rodemerck, B. Kubias, H.W. Zanthoff, M. Baerns, Appl. Catal. A 153 (1997)
scientists in a type of cyber-enabled discovery and innovation plat- 203.
form for catalytic systems. [57] U. Rodemerck, B. Kubias, H.W. Zanthoff, G.U. Wolf, M. Baerns, Appl. Catal. A
153 (1997) 217.
[58] J.T. Gleaves, G. Centi, Catal. Today 16 (1993) 69.
[59] B. Kubias, U. Rodemerck, H.W. Zanthoff, M. Meisel, Catal. Today 32 (1996) 243.
References [60] P.L. Mills, H.T. Randall, J.S. McCracken, Chem. Eng. Sci. 54 (1999) 3709.
[61] A. Pantazidis, S.A. Bucholz, H.W. Zanthoff, Y. Schuurman, C. Mirodatos, Catal.
[1] I.E. Maxwell, P. van den Brink, R.S. Downing, A.H. Sijpkes, S. Gomez, Th. Today 40 (1998) 207.
Maschmeyer, Top. Catal. 24 (2003) 125–135. [62] O.V. Buyevskaya, M. Baerns, Catal. Today 42 (1998) 315.
[2] R.A. Potyrailo, E.J. Amis, High Throughput Analysis: A Tool for Combinatorial [63] Y. Schuurman, T. Decamp, J.C. Jalibert, C. Mirodatos, Stud. Surf. Sci. Catal. 122
Materials Science, Kluwer Academic Publishers, New York, 2003. (1999) 133.
[3] Y.L. Dar, Macromol. Rapid Commun. 25 (2004) 34–47. [64] E.V. Kondratenko, O.V. Buyevskaya, M. Baerns, Top. Catal. 15 (2001) 175.
[4] M.A.R. Meier, U.S. Schubert, J. Mater. Chem. 14 (2004) 3289–3299. [65] E.V. Kondratenko, J. Pérez-Ramírez, Appl. Catal. A 267 (2004) 181.
[5] D.W. Goodman, Chem. Rev. 95 (1995) 523–536. [66] E.V. Kondratenko, M. Cherian, M. Baerns, Catal. Today 99 (2005) 59.
[6] C.R. Henry, Surf. Sci. Rep. 31 (1998) 231–325. [67] M. Olea, M. Florea, I. Sack, R.P. Silvy, E.M. Gaigneaux, G.B. Marin, P. Grange, J.
[7] G.A. Somorjai, B.E. Bent, Phys. Today 48 (1995) 58. Catal. 232 (2005) 152.
[8] G.A. Somorjai, Chem. Rev. 96 (1996) 1223–1235. [68] D.S. Lafyatis, G.F. Froment, A. Pasauclaerbout, E.G. Derouane, J. Catal. 147
[9] H.P. Bonzel, Surf. Sci. 68 (1977) 236. (1994) 552.
[10] C.T. Campbell, Surf. Sci. Rep. 27 (1997) 1–111. [69] Y. Schuurman, C. Marquez-Alvarez, V.C.H. Kroll, C. Mirodatos, Catal. Today 46
[11] H. Happel, Isotopic Assessment of Heterogeneous Catalysis, Academic Press, (1998) 185.
Orlando, 1986. [70] Y. Schuurman, A. Pantazidis, C. Mirodatos, Chem. Eng. Sci. 54 (1999) 3619.
[12] P.L. Silveston, Composition Modulation of Catalytic Reactors, Gordon and [71] Y. Schuurman, C. Mirodatos, P. Ferreira-Aparicio, I. Rodriguez-Ramos, A.
Breach, Ontario, 1998. Guerrero-Ruiz, Catal. Lett. 66 (2000) 33.
[13] C.O. Bennett, Adv. Catal. 44 (2000) 329–416. [72] V. Fierro, Y. Schuurman, C. Mirodatos, J.L. Duplan, J. Verstraete, Chem. Eng. J.
[14] J.T. Gleaves, G.S. Yablonskii, P. Phanawadee, Y. Schuurman, Appl. Catal. A: Gen. 90 (2002) 139.
160 (1997) 55–88. [73] A.M. O’Connor, Y. Schuurman, J.R.H. Ross, C. Mirodatos, Catal. Today 115 (2006)
[15] J. Pérez-Ramírez, E.V. Kondratenko (Eds.), Catal. Today 121 (2007) 1–124. 191.
[16] J.T. Gleaves, J.R. Ebner, T.C. Kuechler, Catal. Rev. -Sci. Eng. 30 (1988) 49–116. [74] V.A. Kondratenko, M. Baerns, J. Catal. 225 (2004) 37.
[17] X. Zheng, J.T. Gleaves, G.S. Yablonsky, T. Brownscombe, A. Gaffney, M. Clark, S. [75] J. Pérez-Ramírez, E.V. Kondratenko, M.N. Debbagh, J. Catal. 233 (2005) 442.
Han, Appl. Catal. A: Gen. 341 (2008) 86–92. [76] E.V. Kondratenko, J. Pérez-Ramírez, Appl. Catal. B 64 (2006) 35.
[18] R. Feres, G.S. Yablonsky, A. Mueller, A. Baerstein, X. Zheng, J.T. Gleaves, Chem. [77] J. Pérez-Ramírez, F. Kapteijn, G. Mul, J.A. Moulijn, J. Catal. 208 (2002) 211.
Eng. Sci. 64 (2009) 568–581. [78] E.V. Kondratenko, J. Pérez-Ramírez, Catal. Lett. 91 (2003) 211.
[19] J.T. Gleaves, J.R. Ebner, U.S. Patent 4,626,412 assigned to Monsanto Company, [79] K.S. Kabin, P. Khanna, R.L. Muncrief, V. Medhekar, M.P. Harold, Catal. Today
December 2, 1986. 1145 (2006) 72.
[20] J.T. Gleaves, J.R. Ebner, P.L. Mills, Studies in Surface Science and Catalysis, vol. [80] C. Rottlander, R. Andorf, C. Plog, B. Krutzsch, M. Baerns, Appl. Catal. B 11 (1996)
38, Elsevier, Amsterdam, 1988. 49.
[21] J. Libuda, H.-J. Freund, Surf. Sci. Rep. 57 (2005) 157–298. [81] T. Gerlach, M. Baerns, Chem. Eng. Sci. 54 (1999) 4379.
134 J.T. Gleaves et al. / Journal of Molecular Catalysis A: Chemical 315 (2010) 108–134

[82] E.V. Kondratenko, V.A. Kondratenko, M. Richter, R. Fricke, J. Catal. 239 (2006) [127] Y.S. Matros (Ed.), Proceedings of the 3rd International Conference on
23. Unsteady-State Processes in Catalysis, VSP Press, Utrecht, The Netherlands,
[83] C.S. Heneghan, G.J. Hutchings, S.R. O’Leary, S.H. Taylor, V.J. Boyd, I.D. Hudson, 1998.
Catal. Today 54 (1999) 3. [128] Y.S. Matros, G.A. Bunimovich, Ind. Eng. Chem. Res. 34 (1995) 1630.
[84] M. Bron, E. Kondratenko, A. Trunschke, P. Claus, Z. Phys. Chem. 218 (2004) 405. [129] P.L. Mills, J.J. Lerou, Transient Response Methods for Assisted Design of Gas
[85] J. Pérez-Ramírez, E.V. Kondratenko, V.A. Kondratenko, M. Baerns, J. Catal. 227 Phase Heterogeneous Catalysts: Experimental Techniques and Mathematical
(2004) 90. Modeling, Rev. in Chem. Eng., vol. 9, Freund Publishing House, Ltd, London,
[86] J. Pérez-Ramírez, E.V. Kondratenko, Chem. Commun. (2004) 376. England, 1993, pp. 3–97.
[87] J. Pérez-Ramírez, E.V. KOndratenko, V.A. Kondratenko, M. Baerns, J. Catal. 229 [130] P.L. Silveston, Composition Modulation in Chemical Reactors, Cambridge Uni-
(2005) 303. versity Press, Cambridge, England, 1997.
[88] M. Baerns, R. Imbihl, V.A. Kondratenko, R. Kraehnert, W.K. Offermans, R.A. van [131] P.L. Mills, W.E. Guise Jr., J. Chromatogr. Sci. 34 (1996) 431–459.
Santen, A. Scheibe, J. Catal. 232 (2005) 226. [132] P.L. Mills, J.F. Nicole, Ind. Eng. Chem. Res. 44 (2005) 6435–6452.
[89] E.V. Kondratenko, J. Pérez-Ramírez, Appl. Catal. A 289 (2005) 97. [133] P.L. Mills, J.F. Nicole, Ind. Eng. Chem. Res. 44 (2005) 6453–6465.
[90] T.A. Nijhuis, M. Makkee, A.D. van Langeveld, J.A. Mouijn, Appl. Catal. A 164 [134] P.L. Mills, J.F. Nicole, Ind. Eng. Chem. Res., in preparation.
(1997) 237. [135] M.J. Lorences, G.S. Patience, F.V. Diez, J. Coca, Ind. Eng. Chem. Res. 42 (2003)
[91] M. Olea, M. Kunitake, T. Shido, Y. Iwasawa, Phys. Chem. Chem. Phys. 3 (2001) 6730–6742.
627. [136] P. Perner, O. Salvetti (Eds.), Proceedings of the International Conference on
[92] S.T. Daniells, A.R. Overweg, M. Makkee, J.A. Moulijn, J. Catal. 230 (2005) 52. Advances in Mass Data Analysis of Signals and Images in Medicine, Biotechnol-
[93] Z.X. Song, H. Nishiguchi, W. Liu, Appl. Catal. A 396 (2006) 175. ogy, and Chemistry, MDA 2006/2007, Springer-Verlag, Leipzig, 2008 (Lecture
[94] R. Fushimi, J.T. Gleaves, G.S. Yablonsky, A. Gaffney, M. Clark, S. Han, Catal. Today Notes in Computer Science).
121 (2007) 170–186. [137] R.J. Cotter, ACS Symposium Series, vol. 549, American Chemical Society, Wash-
[95] J.T. Gleaves, A.G. Sault, R.J. Madix, J.R. Ebner, J. Catal. 121 (1990) 202. ington, DC, 1994.
[96] G.D. Svoboda, J.T. Gleaves, P.L. Mills, Stud. Surf. Sci. Catal. 82 (1994) 481. [138] W.G. Mallard, O.V. Toropov, Automatic Mass Spectral Deconvolution and Iden-
[97] G. Creten, F.D. Kopinke, G.F. Froment, Can. J. Chem. Eng. 75 (1997) 882. tification Software (AMDIS), Agilent Technologies, 2005.
[98] G. Creten, D.S. Lafyatis, G.F. Froment, J. Catal. 154 (1995) 151. [140] G. Centi, F. Cavani, F. Trifiro, Selective Oxidation by Heterogeneous Catalysis,
[99] D.R. Coulson, P.L. Mills, K. Kourtakis, P. Wijnen, J.J. Lerou, L.E. Manzer, Stud. Kluwer Academic and Plenum Publishers, New York, 2001, pp. 363–495.
Surf. Sci. Catal. 75 (1993) 2015. [141] R.M. Contractor, U.S. Patent 5,021,588 assigned to E.I. DuPont de Nemours and
[100] A. Hinz, B. Nilsson, A. Andersson, Chem. Eng. Sci. 55 (2000) 4385. Company, June 4, 1991.
[101] A. Setiabudi, J.L. Chen, G. Mul, M. Makkee, J.A. Moulijn, Appl. Catal. B 51 (2004) [143] M.I. Temkin, Adv. Catal. 28 (1979) 173–291.
9. [144] M.I. Temkin, Int. Chem. Eng. 11 (1971) 709–717.
[102] A. Bueno-Lopez, K. Krishna, M. Makkee, J.A. Moulijn, J. Catal. 230 (2005) [145] M.I. Temkin, Kinet. Katal. 13 (1972) 555–565.
237. [146] K.G. Denbigh, J. Electrochem. Soc. 103 (1956) 137C.
[103] A. Bueno-Lopez, K. Krishna, M. Makkee, J. Moulijn, Catal. Lett. 99 (2005) 203. [147] K.G. Denbigh, Am. J. Phys. 20 (1952) 385.
[104] A. Martin, Y. Zhang, H.W. Zanthoff, M. Meisel, M. Baerns, Appl. Catal. A 139 [148] G.S. Yablonsky, D. Constales, S.O. Shekhtman, J.T. Gleaves, Chem. Eng. Sci. 62
(1996) L11. (2007) 6754–6767.
[105] F. Konietzni, H.W. Zanthoff, W.F. Maier, J. Catal. 188 (1999) 154. [149] R. Fushimi, X. Zheng, J.T. Gleaves, G.S. Yablonsky, A. Gaffney, M. Clark, S. Han,
[106] C. Freitag, S. Besselmann, E. Loffler, W. Grunert, F. Rosowski, M. Muhler, Catal. Top. Catal. 49 (2008) 167–177.
Today 91–92 (2004) 143. [150] P.L. Gai, K. Kourtakis, Science 267 (1995) 661–663.
[107] D.A. Bulushev, E.A. Ivanov, S.I. Reshetnikov, L. Kiwi-Minsker, A. Renken, Chem. [151] B.K. Hodnett, Catal. Rev. 27 (1985) 373–424.
Eng. J 107 (2005) 147. [152] T.P. Moser, G.L. Schrader, J. Catal. 92 (1985) 216–231.
[108] G.S. Yablonsky, S.O. Shekhtman, J.T. Gleaves, P. Phanawadee, Catal. Today 64 [153] J.M.C. Bueno, G.K. Bethke, M.C. Kung, H.H. Kung, Catal. Today 43 (1998)
(2001) 227. 101–110.
[109] S.O. Shekhtman, G.S. Yablonsky, S. Chen, J.T. Gleaves, Chem. Eng. Sci. 54 (1999) [154] S.T. Daniells, A.R. Overweg, M. Makkee, J.A. Moulijn, J. Catal. 230 (2005) 52–65.
4371. [155] R.J. Berger, F. Kapteijn, J.A. Moulijn, G.B. Marin, J. De Wilde, M. Olea, D. Chen,
[110] P. Phanawadee, S.O. Shekhtman, C. Jarungmanorom, G.S. Yablonsky, J.T. A. Holmen, L. Lietti, E. Tronconi, Y. Schuurman, Appl. Catal. A: Gen. 342 (2008)
Gleaves, Chem. Eng. Sci. 58 (2003) 2215. 3–28.
[111] S.O. Shekhtman, G.S. Yablonsky, J.T. Gleaves, R.R. Fushimi, Chem. Eng. Sci. 59 [156] J.H.B.J. Huinink, J. Hoebink, G.B. Marin, Can. J. Chem. Eng. 74 (1996) 580–
(2004) 5493. 585.
[112] S.O. Shekhtman, G.S. Yablonsky, Ind. Eng. Chem. Res. 44 (2005) 6518–6522. [157] W.L.M. Weerts, M.H.J.M. de Croon, G.B. Marin, Surf. Sci. 367 (1996) 321–
[113] B.S. Zou, M.P. Dudukovic, P.L. Mills, J. Catal. 148 (1994) 683. 339.
[114] M. Rothaemel, M. Baerns, Ind. Eng. Chem. Res. 35 (1996) 1556. [158] Anonymous, NIST/EPA/NIH Mass Spectral Library, 2nd edition, 2008,
[115] D.S. Lafyatis, G. Creten, O. Dewaele, G.F. Froment, Can. J. Chem. Eng. 75 (1997) http://www.wiley.com/WileyCDA/WileyTitle/productCd-0470425180.html.
1100. [159] K. Kourtakis, P.L. Mills, C.Z. Cao, US Patent 6,921,831 assigned to E.I. DuPont de
[116] G.S. Yablonskii, P. Phanawadee, J.T. Gleaves, I.N. Katz, Ind. Eng. Chem. Res. 36 Nemours and Company, July 26, 2005.
(1997) 3149. [160] O. Dewaele, V.L. Geers, G.F. Froment, G.B. Marin, Chem. Eng. Sci. 54 (1999)
[117] P. Phanawadee, G.S. Yablonsky, P. Preechasanongkit, K. Somapa, Ind. Eng. 4385.
Chem. Res. 38 (1999) 2877. [161] Y.J. Mergler, J. Hoebink, B.E. Nieuwenhuys, J. Catal. 167 (1997) 305.
[118] D. Constales, G.S. Yablonsky, G.B. Marin, J.T. Gleaves, Chem. Eng. Sci. 56 (2001) [162] Y. Schuurman, Catal. Today 121 (2007) 187–196.
1913. [163] S.O. Shekhtman, N. Maguiere, A. Goguet, R. Burch, C. Hardacre, Catal. Today
[119] D. Constales, G.S. Yablonsky, G.B. Marin, J.T. Gleaves, Chem. Eng. Sci. 56 (2001) 121 (3–4) (2007) 255–260.
133. [164] C.O. Bennett, AIChE J. 13 (1967) 890–895.
[120] J.A. Delgado, T.A. Nijhuis, F. Kapteijn, J.A. Moulijn, Chem. Eng. Sci. 57 (2002) [165] H. Kobayashi, M. Kobayashi, Catal. Rev. -Sci. Eng. 10 (1974) 139–176.
1835. [166] C.O. Bennett, Catalysis Under Transient Conditions, ACS Symposium Series
[121] D.Z. Wang, Chem. Eng. Sci. 56 (2001) 3923. 178, In: A.T. Bell, L.L. Hegedus (Eds.), American Chemical Society, Washington,
[122] D.Z. Wang, J. Chin, Chem. Soc. 50 (2003) 551. DC, (1982) p. 1.
[123] D. Constales, G.S. Yablonsky, J.T. Gleaves, G.B. Marin, Chem. Eng. Sci. 59 (2004) [167] C.O. Bennett, Adv. Catal. 44 (1999) 329–415.
3725. [168] S.C. Van der Linde, T.A. Nijhuis, F.H.M. Dekker, F. Kapteijn, J.A. Moulijn, Appl.
[124] D. Constales, S.O. Shekhtman, G.S. Yablonsky, G.B. Marin, J.T. Gleaves, Chem. Catal. A: Gen. 151 (1997) 27.
Eng. Sci. 61 (2006) 6. [169] M. Olea, M. Kunitake, T. Shido, K. Asakura, Y. Iwasawa, Bull. Chem. Soc. Jpn. 74
[125] G.F. Froment, K.B. Bischoff, Chemical Reactor Analysis and Design, 2nd edition, (2001) 255.
Wiley, New York, 1990. [170] Y. Schuurman, J.T. Gleaves, Catal. Today 33 (1997) 25–37.
[126] Y.S. Matros (Ed.), Unsteady-State Processes in Catalysis, VSP Press, Utrecht, [171] Y. Schuurman, J.T. Gleaves, Ind. Eng. Chem. Res. 33 (1994) 2935–2941.
The Netherlands, 1990.

You might also like