Infrared spectroscopy for chemically specific sensing in silicon-based microreactors

Article abstract of DOI:10.1021/ac049265c

Fourier transform infrared (FT-IR) spectroscopy in a multiple internal reflection (MIR) geometry is integrated with silicon-based microreactors to allow detection of a wide range of chemical species while taking advantage of inexpensive batch fabrication techniques applicable to silicon substrates. The microreactors are fabricated in silicon and glass using standard microfabrication and selective etching techniques. The small (～1 cm side) reactor size provides access to nearly die full mid-IR frequency region with MIR-FT-IR, allowing us to probe both solution-phase and surface-bound chemical transformations. The wide applicability of this approach is demonstrated with two representative test cases: kinetics of acid-catalyzed ethyl acetate hydrolysis and amidization of surface-tethered amine groups.

Full text of DOI:10.1021/ac049265c

Anal. Chem. 2004, 76, 6476-6483
In fra re d S p e c t ro s c o p y fo r Ch e m ic a lly S p e c ific
S e n s in g in S ilic o n -Ba s e d Mic ro re a c t o rs
Ra c he l He rzig-Ma rx† a nd K. T. Que e ne y*
Department of Chemistry, Smith College, Northampton, Massachusetts 01063
Re be c c a J . J a c km a n,§,‡ Ma rtin A. Sc hm idt, a nd Kla vs F. J e ns e n,
⊥
|,§
Department of Chemical Engineering and Microsystems Technology Laboratories, Massachusetts Institute of Technology,
Cambridge, Massachusetts 02139
Fourier transform infrared (FT-IR) spectroscopy in a
multiple internal reflection (MIR) geometry is integrated
with silicon-based microreactors to allow detection of a
wide range of chemical species while taking advantage of
inexpensive batch fabrication techniques applicable to
silicon substrates. The microreactors are fabricated in
silicon and glass using standard microfabrication and
selective etching techniques. The small (∼1 cm side)
reactor size provides access to nearly the full mid-IR
frequency region with MIR-FT-IR, allowing us to probe
both solution-phase and surface-bound chemical trans-
formations. The wide applicability of this approach is
demonstrated with two representative test cases: kinetics
of acid-catalyzed ethyl acetate hydrolysis and amidization
of surface-tethered amine groups.
traditional absorption spectroscopy by optimizing the optical
configuration for the sensitivity demands of µTAS,3 while
alternative approaches have focused on detecting more sensitive
photothermal6 and refractive index changes.7,8 The chemical
specificity of NMR spectroscopy has been exploited in an on-chip
NMR probe, but without the spectral resolution that makes NMR
such a powerful analytical technique for complex macroscopic
samples.
Fourier transform infrared (FT-IR) spectroscopy is in many
ways an ideal detection technique, since virtually all molecules of
interest absorb IR radiation. In contrast to UV/ vis absorption, FT-
IR is a fingerprinting method. Characteristic functional groups are
apparent by quick inspection of most IR spectra, and the unique
IR signature of many molecules allows identification of an
unknown analyte by comparison with spectral libraries. Several
difficulties arise, however, in the effort to integrate FT-IR with
µTAS. First, the strong IR absorption of many solventss
particularly waterslimits the quantity of analyte solution through
which the incident IR radiation can travel. The materials of the
microreactor themselves must also allow coupling of the IR light
-5
9
The detection of chemical species in micro total analysis
systems (µTAS) presents both familiar and new challenges,1,2
including the ability to sample drastically reduced volumes and
the compatibility of detection techniques with materials amenable
to device fabrication. Although fluorescence is the most widely
2
to the sample. Traditional IR-transparent materials (e.g., CaF ,
1
used detection method for µTAS, it suffers from the need to
ZnSe) are not compatible with the usual techniques used for the
fabrication of micro total analysis systems, whereas many tradi-
tional µTAS substrates (e.g. glass, polymers) are strong IR
absorbers.
chemically modify most analytes to induce fluorescence. Electro-
chemical detection, which has gained popularity owing to the ease
of conversion from the detected quantity to an electrical signal,
is limited to electroactive analytes.
Although absorption spectroscopic methods for detection are
attractive because of their near universality, the sensitivity of these
techniques is often a limitation in the short path length configura-
tions dictated by µTAS. Some progress has been made in
These incompatibilities between µTAS substrate materials and
the optical requirements for FT-IR have led researchers in most
previous studies to couple FT-IR to the µTAS via an analysis
1
0-13
chamber that is separate from the microreactor itself.
In one
(
3) Verpoorte, E.; Manz, A.; Luedi, H.; Bruno, A. E.; Maystre, F.; Krattiger, B.;
Widmer, H. M.; Van der Schoot, B. H.; De Rooij, N. F. Sens. Actuators, B
1 9 9 2 , B6, 66-70.
*
To whom correspondence should be addressed. Fax: 413-585-3786.
E-mail: kqueeney@smith.edu.
(4) Pandraud, G.; Koster, T. M.; Gui, C.; Dijkstra, M.; van den Berg, A.; Lambeck,
|
To whom correspondence about fabrication should be addressed. Fax: 617-
P. V. Sens. Actuators, A 2 0 0 0 , A85, 158-162.
2
58-8224. kfjensen@mit.edu.
(5) Nishimoto, T.; Fujiyama, Y.; Abe, H.; Kanai, M.; Nakanishi, H.; Arai, A. In
Proceedings of Micro Total Analysis Systems 2000; Kluwer Academic Publish-
ers: Dordrecht, The Netherlands, 2000; pp 395-398.
(6) Malik, A. K.; Faubel, W. Chem. Soc. Rev. 2 0 0 0 , 29, 275-282.
(7) Burggraf, N.; Krattiger, B.; de Rooij, N. F.; Manz, A.; de Mello, A. J. Analyst
1 9 9 8 , 123, 1443-1447.
(8) Kameoka, J.; Craighead, H. G. Sens. Actuators, B 2 0 0 1 , 77, 632-637.
(9) Massin, C.; Daridon, A.; Vincent, F.; Boero, G.; Besse, P. A.; Verpoorte, E.;
de Rooij, N. F.; Popovic, R. S. In Proceedings of Micro Total Analysis Systems
2001; Kluwer Academic Publishers: Dordrecht, The Netherlands, 2001; pp
438-440.
§
Department of Chemical Engineering, Massachusetts Institute of Technol-
ogy.
⊥
Microsystems Technology Laboratories, Massachusetts Institute of Technol-
ogy.
†
Current address: Lincoln Laboratories, Cambridge MA 02139.
Current address: Commonwealth School, 151 Commonwealth Ave., Boston
‡
MA 02116.
(
(
1) Schwarz, M. A.; Hauser, P. C. Lab Chip 2 0 0 1 , 1, 1-6.
2) Auroux, P.-A.; Iossifidis, D.; Reyes, D. R.; Manz, A. Anal. Chem. 2 0 0 2 , 74,
2
637-2652.
6476 Analytical Chemistry, Vol. 76, No. 21, November 1, 2004
10.1021/ac049265c CCC: $27.50 © 2004 American Chemical Society
Published on Web 10/01/2004

case, on-chip FT-IR detection was achieved by deep reactive ion
etching to form a thin transmission cell within a Si substrate,14
but this approach does not provide the surface sensitivity required
for sensor-type applications in which solution- or gas-phase
analytes react with a substrate-bound reagent. The ideal optical
configuration to combine solution-phase and surface sensitivity
is multiple internal reflection (MIR) FT-IR, which samples both
the substrate surface and any solution in contact with it via IR
light propagated from inside the substrate.
detection, as described briefly in a preliminary publication.22 We
demonstrate the utility of this approach both for analysis of
solution-phase reactions and for detection of the reaction of a
solution-phase analyte with a surface-bound functional group in
situ, the first example of an on-chip FT-IR detection scheme suited
to both of these general applications.
EXPERIMENTAL SECTION
Microreactor Fabrication. Microreactors were fabricated by
anodically bonding MIR crystals produced on silicon wafers and
fluidic channels created in Pyrex glass wafers. Figure 1a shows a
schematic of the fabrication sequence for these devices. To
produce suitably polished optical surfaces by anisotropic etching
of Si, alignment of a mask to the wafer flat is not sufficient. We
therefore used a two-step method described by Ensell21 for
improving alignment. This technique is based on the formation
of square pyramidal etch pits during potassium hydroxide (KOH)
etching of Si(100); these features form with their square cross
sections aligned to the 〈110〉 axes that the wafer flat(s) ap-
proximate. Production of two series of test pits on opposite sides
of one wafer face allows identification of an exact 〈110〉 axis by
connecting the “best-aligned” pits on either side of the wafer. The
reader is referred to ref 21 for a detailed explanation of this
method, summarized as follows.
Silicon (Si) is a natural choice for the fabrication of microre-
actor substrates15 and a material that is compatible with MIR-FT-
IR, albeit subject to the limitations outlined below. The central
role of Si in microelectronics and in microelectromechanical
systems (MEMS) means that methods for processing it are well-
developed, including parallel processing techniques to enable the
fabrication of many microreactor devices on a single silicon wafer.
The use of silicon microfabrication techniques also offers the
possibility of producing structures with fully integrated electrical
circuits. Furthermore, surface chemistry for well-defined func-
tionalization of Si surfaces is also well-established, making Si
substrates an ideal choice for sensor-type applications.
Although the use of MIR-FT-IR with Si substrates is a common
technique for increasing sensitivity to Si surface species,16 silicon
has not found widespread use as a more universal MIR substrate,
A layer of low-pressure chemical vapor deposition (LPCVD)
silicon nitride (∼1 to 2 µm) was first deposited onto the Si wafer
because multiphonon bands effectively absorb all the incident IR
radiation below 1500 cm-1. However, it was recently shown that
(
FZ Si(100), 1901-3680 Ω cm, double-side-polished, ∼500 µm
shortening the length of the Si substrate parallel to the incident
IR beam can extend the frequency cutoff well below 1000 cm-
1 17
thick). A first mask with a series of alignment marks (Figure 1b)
was used to pattern a layer of photoresist (OCG-825, OCG
Microelectronic Materials, Inc.); these alignment marks consisted
of two sets of 41 circles (71 µm diameter). Each circle was created
with its center along an arc of radius 45 mm, and the circles in
each set were arranged relative to one another with an angle of
,
providing access to a large portion of the so-called IR “fingerprint”
region, even in studies of the Si/ liquid interface.18,19 The remaining
obstacle to incorporating Si MIR substrates into µTAS is the
economical fabrication of these substrates, since MIR crystals are
traditionally produced by individually polishing the substrates,
which makes them both expensive to manufacture and difficult
to integrate with other microreactor features. Anisotropic etching
of silicon has been used to create beveled Si surfaces that are
suitably polished for use in optical systems,20,21 providing the
potential for parallel production of multiple Si crystals for MIR-
FT-IR applications.
0
.1° between then. The two sets of marks were positioned on
opposite sides of the wafer and were a mirror image of one
another. This pattern was transferred into the nitride using reactive
ion etching, and the layer of photoresist was removed using
acetone.
To create the test etch pits, the wafer was blown dry, and
the exposed Si was etched for ∼30 min in a 20% KOH bath at
In the present work, we incorporate anisotropic etching into
the parallel fabrication of Si-based microreactors with MIR-FT-IR
∼
80 °C. Although the origin of this etchant anisotropy is very
complicated,23 it is well-known that aqueous bases etch Si(111)
faces much more slowly than other faces. In the case of KOH
etching of Si(100), if an etch mask exposes a square of silicon
whose sides align perfectly to the 〈110〉 axes, the result is a square
pyramidal pit with the faces of the pit defined by {111} planes. In
our case, where the etch mask exposes a circular area of silicon,
the KOH bath will still create the same axis-aligned etch pit, in
this case with its size determined by the diameter of the circle.24
While the circular openings in the nitride layer are evenly
displaced relative to one another, the relative displacements of
the KOH-etched test pits depend on their alignment to the 〈110〉
(
(
10) Kolhed, M.; Lendl, B.; Karlberg, B. Analyst, 128, 2-6.
11) Louden, D.; Handley, A.; Lafont, R.; Taylor, S.; Sinclair, I.; Lenz, E.; Orton,
T.; Wilson, I. D. Anal. Chem. 2 0 0 2 , 74, 288-294.
(
12) Lendl, B.; Schindler, R.; Frank, J.; Kellner, R.; Drott, J.; Laurell, T. Anal.
Chem. 1 9 9 7 , 69, 2877-2881.
(
(
13) Kellner, R.; Lendl, B. Anal. Methods Instrum. 1 9 9 5 , 2, 52-54.
14) Jackman, R. J.; Floyd, T. M.; Schmidt, M. A.; Jensen, K. F. In Proceedings of
Micro Total Analysis Systems 2000; Kluwer Academic Publishers: Dordrecht,
The Netherlands, 2000; pp 155-158.
(
15) Reyes, D. R.; Iossifidis, D.; Auroux, P.-A.; Manz, A. Anal. Chem. 2 0 0 2 , 74,
2
623-2636.
(
(
16) Chabal, Y. J. Surf. Sci. Rep. 1 9 8 8 , 8, 211-357.
17) Weldon, M. K.; Stefanov, B. B.; Raghavachari, K.; Chabal, Y. J. Phys. Rev.
Lett. 1 9 9 7 , 79, 2851-2854.
(
(
18) Queeney, K. T.; Fukidome, H.; Chaban, E. E.; Chabal, Y. J. J. Phys. Chem.
(22) Jackman, R. J.; Queeney, K. T.; Herzig-Marx, R.; Schmidt, M. A.; Jensen,
K. F. In Proceedings of Micro Total Analysis Systems 2001; Kluwer Academic
Publishers: Dordrecht, The Netherlands, 2001; pp 345-346.
(23) Wind, R. A.; Jones, H.; Little, M. J.; Hines, M. A. J. Phys. Chem. B 2 0 0 2 ,
106, 1557-1569.
B 2 0 0 1 , 105, 3903-3907.
19) Fukidome, H.; Pluchery, O.; Queeney, K. T.; Caudano, Y.; Raghavachari,
K.; Weldon, M. K.; Chaban, E. E.; Christman, S. B.; Kobayashi, H.; Chabal,
Y. J. Surf. Sci. 2 0 0 2 , 502-503, 498-502.
(
(
20) Klumpp, A.; Kuehl, K.; Schaber, U.; Kaeufl, H. U.; Lang, W. Sens. Actuators,
A 1 9 9 5 , A51, 77-80.
21) Ensell, G. Sens. Actuators, A 1 9 9 6 , A53, 345-348.
(24) In reality, underetching creates a square slightly larger than the circular
opening; for simplicity, we have shown in Figure 1 the ideal case of a square
circumscribed by the circular opening in the etch mask.
Analytical Chemistry, Vol. 76, No. 21, November 1, 2004 6477

etching of the wafer for several hours etched through the wafer.
Removal of the nitride mask using reactive ion etching produced
a wafer of polished silicon MIR crystals.
The Pyrex glass wafer was coated with a layer of image-reversal
photoresist (AZ-5214E, Hoechst Celanese, Somerville NJ) that was
then patterned using a mask with fluidic channels. After a brief
exposure to an oxygen plasma to remove debris in areas where
the photoresist had been removed, a layer of chromium (∼200
nm) was deposited onto the wafers using an electron-beam
evaporator. Metal deposited onto the photoresist was lifted off as
the photoresist dissolved in an acetone bath, and then the wafer
was cleaned. Channels (∼50 µm deep) were transferred into the
wafer by etching in dilute hydrofluoric acid (HF) using the
chromium metal as an etch stop. The chromium layers were
removed in a standard chromium etch, and then the silicon and
Pyrex wafers were cleaned in piranha solution. The silicon and
Pyrex wafers were anodically bonded25 by placing the wafers
(
aligned to one another) in contact, heating them to ∼450 °C,
applying a DC voltage (∼700 V) across the wafers for ap-
proximately 5 min, then slowly cooling the wafer stack to room
temperature. This anodic bonding step produced sealed devices.
Dicing the wafer stack produced individual microreactors with
integrated MIR crystals.
Ports for the introduction of fluid into the microreactor were
incorporated into the silicon MIR crystal. We attached microtubing
to these ports using black poly(dimethylsiloxane) (PDMS, Sylgard
1
70, Dow Corning, Midland MI) and epoxy.26 The use of black
PDMS effectively blocks out any stray light from the IR beam in
the lateral direction. Maintaining a maximum beam size in the
vertical direction is, in fact, advantageous: because the height of
the microreactor substrate is greater than the beam size, this
approach yields maximum beam/ sample interaction with no
possibility of stray light reaching the detector. Our ability to use
the maximum aperture diameter for the IR beam avoids the
limitation previously put forth12 that the FT-IR spot size must be
reduced significantly for use with µTAS. When mounting the
device in the sample holder, we also used a piece of black PDMS
on the glass side of the device to block out stray light on that
side.
Fig u re 1 . (a) Schematic of microreactor fabrication process (not
drawn to scale). A single path for IR light propagation is shown in
the last picture in the sequence; in fact, IR light will enter the reactor
from multiple positions along the bevel of the MIR substrate. (b) Top
view of a portion of the mask used to etch alignment marks into the
silicon nitride. (c) Top view of KOH-etched alignment marks (square
cross section) in Si exposed by alignment marks (circular) etched in
the nitride.
Ethyl Acetate Hydrolysis. In all experiments, ethyl acetate
(
Fisher, A.C.S. grade) was dissolved in deionized H
2
O (Millipore,
1
8.2 MΩ) to yield a 0.409 M solution. After ∼30 min of vigorous
stirring, the solution was adjusted with 6.05 M HCl to the desired
HCl concentration, with the resulting solutions all falling in the
range 0.3-0.4 M in ethyl acetate. Immediately after addition of
the acid, the first reaction aliquot was removed for analysis in
the microreactor. (This choice of experimental protocols
maintaining vigorous stirring of the reaction mixture during the
reaction progresssis discussed below in Results and Discussion.)
Subsequent aliquots were transferred to the microreactor as soon
as the spectrum of the previous sample was finished, ∼9 min apart.
An excess of the total volume needed to fill the microreactor and
inflow and outflow tubing (∼1 mL) was flowed through the
microreactor each time to ensure optimal mixing; however, as
evidenced by the experiment described above, any reaction
axes (Figure 1b). The smaller the lateral displacement between
two adjacent etch pits, the better aligned to the 〈110〉 axis those
pits and their corresponding partners on the opposite side of the
wafer. Examination of the etch pits under an optical microscope
enabled identification of the set of alignment marks most closely
aligned to the 〈110〉 axes (e.g., those with the smallest displace-
ment relative to their two nearest neighbors); this set of alignment
marks was then used to orient subsequent masks.
After cleaning the wafers twice in a piranha cleaning bath
(
sulfuric acid/ hydrogen peroxide, 3:1;Caution: piranha solution
reacts violently with organics and should be handled with extreme
care), the steps to pattern the nitride layer were repeated using a
second mask that defined the edges of the silicon crystal and
fluidic ports to introduce solutions into the devices. Anisotropic
(
25) For a thorough explanation of anodic bonding (in this reference, called “field
assisted bonding”), see: Madou, M. J. Fundamentals of Microfabrication;
CRC Press: Boca Raton, FL, 2002, pp 384-389.
(26) McDonald, J. C.; Whitesides, G. M. Acc. Chem. Res. 2 0 0 2 , 35, 491-499.
6478 Analytical Chemistry, Vol. 76, No. 21, November 1, 2004

mixture left over from the previous aliquot is not expected to alter
the data from the next sample. Because the spectrum of each
aliquot is an average of the reaction progress over a ∼6-min time
period, as detailed below, the resulting spectral changes represent
the average ethyl acetate concentration over this interval. For
kinetic analysis, the important parameter is ∆t between the first
and subsequent aliquots, which we obtain from the time stamp
recorded at the start of each spectrum. Because the reaction
progresses during transfer of the aliquot from the reaction vessel
to the microreactor, this time measurement is more accurate than
the recorded time interval for aliquot removal.
Reaction of Surface Amine Groups. Amine-functionalized
surfaces were prepared by reaction of oxidized Si with (amino-
propyl)triethoxysilane (APTES, Aldrich, 99%). Prior to reaction
with the APTES, Si surfaces were rigorously cleaned according
to the protocol described by Fadeev and McCarthy27 in a solution
of ∼70% sulfuric acid in which 6 g of sodium dichromate was
dissolved and ∼30% hydrogen peroxide (30% aqueous). This
cleaning step is required both to rid the surface of hydrocarbon
contamination and to fully hydroxylate the overlying silica to allow
siloxane bond formation between APTES molecules and surface
silanol groups. The reactive cleaning solution was incompatible
with organic components of the microreactor (PDMS or epoxy),
however. The in situ amidization experiments were therefore
carried out on beveled Si pieces (FZ Si(100), 10-100 Ω cm,
double-side-polished) with the same length dimension (∼1 cm)
as the substrates for the microreactors. In place of the microre-
actor, the Si was placed in a Teflon-based liquid cell with a liquid/
Si contact area per side equivalent to the contact area for the
trough microreactors. This reusable cell, which is described in
detail elsewhere,18 relies on a spring-loaded seal between the
Teflon reservoir, a polished graphite insert, and the polished Si
sample. Because the design of the liquid cell allows contact with
both sides of the Si substrate, this experiment provides twice the
sensitivity of the microreactors, a difference addressed below in
Results and Discussion.
The cleaned silicon substrates were reacted with 1-2 mmol
APTES in dry toluene for 24 h at room temperature. The samples
were held vertically in the reaction mixture in a custom Teflon
holder to allow reaction with both sides of the polished substrates.
After removal from the reaction solution, the substrates were
thoroughly rinsed sequentially in toluene, ethanol/ toluene (1:1)
and ethanol, then dried with compressed, filtered air.
To identify IR peaks characteristic of amide bond formation
on the surface amine groups, we carried out ex situ reaction and
transmission IR characterization (described below) of analogous,
unbeveled Si substrates. In these experiments, the APTES-
functionalized substrates (made of the same Si as the beveled
substrates) were reacted with 0.1 M acetic anhydride in methylene
chloride overnight at room temperature. After removal from the
reaction solution, samples were rinsed twice in methylene chloride
and dried with compressed air.
into the liquid cell via syringe and was flowed through continu-
ously during spectral acquisition, using a syringe pump at a rate
of 4 mL/ h.
Spectral Acquisition. All FT-IR data were obtained with a
Nicolet Nexus 670 spectrometer operating at 4 cm- resolution.
The incident IR beam was parallel to the faces of the microreactor,
with no additional optics added to the spectrometer sample
compartment, resulting in an incidence angle of 35.3° from the
surface normal of the beveled edge. This geometry gives rise to
∼12 internal reflections before exiting the microreactor, with an
incidence angle of 64.4° from the surface normal at the microre-
actor/ liquid interface. Spectra for in situ experiments were
acquired with a liquid nitrogen-cooled MCT-high D* detector. For
the ethyl acetate hydrolysis experiment, each spectrum consisted
of 500 scans (approximately 6.2-minute collection time); each
spectrum for the in situ amidization experiment consisted of 100
scans (∼1.2-min collection time).
1
Data for the ex situ amidization experiment were acquired in
an external transmission geometry, with the incident IR beam at
6
0° to the surface normal of the Si sample. Spectra for this
experiment were collected with a thermoelectrically cooled DTGS
detector and consisted of 1000 scans each for sample and
background.
RESULTS AND DISCUSSION
Reactor Characteristics. Lithography and anisotropic etching
enable the parallel fabrication of many Si MIR crystals. We
fabricated 16 device substrates from one 100-mm Si wafer, with a
total processing time of 12 h, or <1 h per device. This compares
favorably with the time to polish each substrate individually, a
process that takes at least 3 h per substrate. This advantage to
our approach would be increased dramatically by simultaneous
processing of multiple wafers, since alignment of the mask for
the final KOH etch step is the only nonparallel step in the process.
Furthermore, fabrication of MIR substrates by the technique
detailed above can be combined with other microfabrication
techniques to integrate the simple microreactors used in this study
with more elaborate functionalitysfor instance, a separation
module or electronics for signal conversionsfrom one single Si
substrate.
The MIR prisms we fabricated were successfully integrated
with microfluidic channels to produce microreactors with detection
capability. For any such applications, the material for the channels
must be able to bond to silicon and must be compatible with the
solutions with which it is in contact. We used Pyrex glass for the
channels here, but we have also produced reactors from PDMS
using soft lithographic techniques.14
For most of the experiments performed here, the channel
design was very simple, either a single, trough-like channel ∼0.75
cm wide or a serpentine channel ∼100 µm wide (Figure 2a). These
designs result in a microreactor volume of ∼3 to 4 µL in contact
with the Si substrate. We chose this straightforward design of a
single reaction chamber to maximize surface area in contact with
the MIR crystal and to minimize detection time. Ultimately, the
optimal microchannel design would be a compromise between
the greater detection volume afforded by a single chamber and
the need to mix multiple reactant streams via microfluidic
channels. The design used in this study, however, could be
For the in situ experiments, a beveled APTES-functionalized
substrate was inserted into the Teflon liquid cell. The cell was
first flushed with CCl
in the mid-IR region) for acquisition of background spectra. A
solution of acetic anhydride (0.1 M in CCl ) was then introduced
4 2 2
(used instead of CH Cl to limit absorbance
4
(
27) Fadeev, A. Y.; McCarthy, T. J. Langmuir 2 0 0 0 , 16, 7268-7274.
Analytical Chemistry, Vol. 76, No. 21, November 1, 2004 6479

Fig u re 3 . Infrared spectra (100 scans) of reactants and products
of ethyl acetate hydrolysis, acquired in aqueous solution (5 mL/100
mL H2O) in a microreactor: ethyl acetate (solid line), ethanol (dotted
line), and acetic acid (dashed line). All spectra are ratioed to a
background of pure H2O in the microreactor.
Ta b le 1 . Ra t e Co n s t a n t s fo r t h e HCl-Ca t a lyze d
Hyd ro lys is o f Et h yl Ac e t a t e a
[
ethyl acetate]0, M
0.375
[HCl], M
0.5
k (× 10- min-1
2
)
Fig u re 2 . (a) Photograph of a microreactor fabricated with glass
channels. (b) IR throughput of a microreactor with a glass trough filled
with H2O.
0.23
.23
b
0
0
0
0
.341
.307
.3629
1.0
1.5
1.714
0.40
0.71
1.17
c
d
1.12
modified (by modifying the mask for the Pyrex wafer) to include
more complex channel designs.
0
.274
2.0
2.435
1.21
1.66
c
0.1157
d
1.570
As predicted, the 1-cm length of the Si substrate for
the microreactor allows transmission of significant IR intensity
throughout the mid-IR range, as shown in Figure 2b. In this single-
beam spectrum from a water-filled microreactor, the low-frequency
cutoff (∼850 cm-1) arises from a combination of the intrinsic cutoff
of the MCT detector and Si phonon bands. Absorption is evident
both from components of the microreactor (PDMS and glass, as
labeled in the figure) and from stretching (∼3100-3800 cm-1)
a
Unless otherwise noted, the values were obtained in this study.
Value from experiment carried out in microreactor with no stirring
e.g., data from Figure 4). c Value from ref 29. d Value from ref 30.
b
(
for any of the three species that do not overlap appreciably with
those from the remaining species, the sharp peak from ethyl
acetate at 1379 cm-1 (derived mainly from the symmetric
deformation of the methoxy group)28 is easily distinguished from
the relatively broad band of acetic acid that it overlaps. Monitoring
the depletion of this sharp band concomitant with the growth of
a broad band around it does not require any spectral deconvolu-
tion, only straightforward integration of the depletion peak, as
demonstrated below. It is important to note that this spectral
region is only accessible with a Si substrate when using a short-
sample geometry. Although there are, no doubt, organic reactions
for which finding a unique spectral feature will prove impossible,
the ability to monitor the fingerprint region typically provides
access to a multitude of discrete peaks, thereby increasing the
chances of finding a suitable tag for reactant or product.
The kinetics of ethyl acetate hydrolysis were monitored over
a range of HCl concentrations. First-order rate constants from this
study are presented in Table 1 and compared to literature values
in the same HCl concentration range derived via interferometry29
and titration.30 As outlined briefly above, the reaction was
monitored by following the depletion of the ethyl acetate peak at
1379 cm-1. Because the primary focus of the current study is to
-
1
and bending (∼1640 cm ) bands of H
interaction region between the water and the Si allows significant
IR throughput, even in the regions of H O absorption, a distinct
advantage for analyzing bands, such as ν(CdO) modes that often
overlap with δ(H O) in aqueous systems. It is also important to
2
O. The relatively small
2
2
note that the selective etching used to generate bevels for the IR
prism creates a surface of high enough optical quality to couple
sufficient light into the Si prism without excessive scattering.
Ethyl Acetate Hydrolysis. Although the acid-catalyzed hy-
drolysis of ethyl acetate is an extremely simple reaction, it provides
an excellent test case for the use of IR to detect or characterize
organic reactions. One of the biggest challenges in following the
reactions of organic species is the conservation of characteristic
functional groups in reactants and products. This challenge is
exemplified by ester hydrolysis, since carbonyl stretches for esters
and the corresponding carboxylic acids may well overlap, as they
do in this case for ethyl acetate and acetic acid (Figure 3). The
aqueous medium for this reaction provides a further challenge,
since hydrogen bonding makes it difficult to quantify growth in
ν(OH) modes from the alcohol and acid products.
As the spectra of reactants and products in Figure 3 demon-
strate, the ability to monitor the fingerprint region below 1500
cm- is critical in this case. Although there are no significant peaks
(
(
28) Nolin, B.; Jones, R. N. Can. J. Chem. 1 9 5 6 , 34, 1392-1404.
29) O’Brien, R. N.; Glasel, A. Can. J. Chem. 1 9 6 9 , 47, 223-228.
1
(30) Harned, H. S.; Pfanstiel, R. J. Am. Chem. Soc. 1 9 2 2 , 44, 2193-2205.
6480 Analytical Chemistry, Vol. 76, No. 21, November 1, 2004

shown (second spectrum from the top) corresponds to a change
of 7 mM ethyl acetate. While the 1379 cm- peak was chosen for
optimal selective quantification of ethyl acetate in this particular
reaction mixture, it is obviously (Figure 3) not the most intense
peak available for analysis. For example, ν(CdO) at 1707 cm-
has an integrated absorptivity ∼4.7 times greater than the peak
analyzed in this work. It is also important to note that sensitivity
in this experiment is tied to the sampling interval, since smaller
signals could in some cases be resolved with a greater number
of scans. (Conversely, for monitoring kinetics of a faster reaction
than ethyl acetate hydrolysis, the need for shorter scan times
would result in a lower sensitivity). Taking into account these
considerations, it is reasonable to expect our detection scheme
to yield sensitivity in the 1-10 mM range for a wide range of
reaction systems.
1
1
Quantitative analysis of solution-phase reactions using MIR-
IR may be hindered in the case of preferential adsorption of
reactants or products to the microreactor substrate, in this case
silica. Although the penetration depth in this experiment is ∼0.77
µm at the frequency of our peak analysis,31 the exponential decay
of the evanescent IR wave causes surface-associated species to
be weighted more heavily in sampling than their solution-phase
Fig u re 4 . Sample kinetic data from ethyl acetate hydrolysis in 0.5
M HCl. Spectra in (a) show the growth of the ethyl acetate depletion
peak as a function of time from top to bottom (from 6 to 112 min). All
spectra are ratioed to the first spectrum acquired after introduction
of the reaction mixture into the microreactor. Spectra are offset for
clarity. The plot of integrated peak area vs time in (b) is fit to eq 1 to
yield the rate constant in Table 1.
counterparts. However, although ethyl acetate is known to adsorb
to silica from nonpolar solvents,32,33 it is unlikely that such
adsorption takes place from aqueous solution. Evidence of any
such adsorption taking place in our system would be obscured
by the peak shifts arising from hydrogen bonding between ethyl
acetate and water, which shifts both ν(CdO) and ν(C-C-O)
peaks of the ethyl acetate as much or more as does interaction of
demonstrate the efficacy of IR detection in our microreactors, the
acid-catalyzed hydrolysis of ethyl acetate was in most cases
monitored by carrying out the reaction with vigorous stirring in
a macroscopic sample and removing aliquots at timed intervals
for analysis in the microreactor. This approach was taken to
ensure adequate mixing of the sample as the reaction progressed,
since initial experiments showed some phase separation in
unstirred ethyl acetate solutions in the concentration range used
in this study.
One data set was obtained by directly injecting the initial
reaction solution into the microreactor and following this solution
as a function of time; the spectra acquired from this experiment,
carried out with 0.5 M HCl in a microreactor with a trough-like
channel, are presented in Figure 4a. The resulting plot of peak
area vs time, shown in Figure 4b, is fit to the expression
adsorbed ethyl acetate with surface silanol groups in a nonpolar
solvent.32 In any case, the good agreement between our rate
constants and those derived from more macroscopic sampling
techniques suggests that such adsorption effects do not play a
significant role in analysis of the aqueous ethyl acetate system.
For systems in which such effects are found to dominate (e.g.,
analysis of polar molecules in nonpolar solvents), functionalization
of the Si substrate to yield a more hydrophobic surface may be
advantageous.
Reaction of Surface-Tethered Functional Groups. The
reaction of APTES with hydrated silica to form surface amine
(
NH
2
) groups was chosen for this work to present a high density
for subsequent reaction with solution-phase species. Since
of NH
2
-
kt
solution deposition of APTES is known to deposit polymeric
multilayers on hydrated silica,34 this route was chosen rather than
reaction of a monoalkoxide silane to form a covalently bound
area ) C(1 - e
)
(1)
where C is a constant and t is the time elapsed since analysis of
the first aliquot. The comparison in Table 1 demonstrates that
our results are consistent with the earlier, more detailed studies
of this system that employed macroscopic sampling methods,
demonstrating the efficacy of our microreactor analysis. Further-
more, within our limits of precision, the experiment carried out
entirely within the microreactor (e.g., without vigorous external
mixing) is identical to an analogous experiment at 0.5 M HCl with
more complete mixing. To reliably ensure adequate mixing,
however, would require modification of our simple microreactor
designsin this case, optimized for detection onlysto mix one or
more reactant streams.
monolayer27 with a lower density of surface NH
To identify the IR peaks indicative of the reaction of surface
NH groups with solution-phase acetic anhydride, we carried out
an ex situ reaction characterized with external transmission IR
as described above. As outlined below, analysis of IR spectra of
the dry surface is helpful in deciphering the complex spectra
arising from the presence of both surface-bound and solution-
phase species in the MIR configuration. A transmission spectrum
2
groups.
2
(
31) Penetration depth was calculated40 assuming reflection at the Si/ SiO
interface.
2
(32) Poston, P. E.; Rivera, D.; Uibel, R.; Harris, J. M. Appl. Spectrosc. 1 9 9 8 , 52,
391-1398.
1
(
(
33) Scott, R. P. W.; Kucera, P. J. Chromatogr. 1 9 7 8 , 149, 93-110.
34) Kallury, K. M. R.; Macdonald, P. M.; Thompson, M. Langmuir 1 9 9 4 , 10,
492-499.
The sensitivity of our detection scheme can be estimated from
the data in Figure 4a, in which the smallest quantifiable peak
Analytical Chemistry, Vol. 76, No. 21, November 1, 2004 6481

arises from gradual displacement in the near-surface region of
the CCl solvent that was used to equilibrate the cell and to acquire
4
background spectra. Growth of surface amide species is marked
by increased intensity in the amide I band (centered at 1654 cm-1),
as highlighted in the inset to Figure 6. This assignment is
supported by concomitant growth both of the 1204 cm- peak
1
(
seen as a shoulder on an acetic anhydride peak) and of ν(NH)
-
1
centered at 3300 cm (data not shown). Growth in the amide II
band (at 1559 cm- in the ex situ experiment) is obscured in this
1
experiment by depletion of the δ(NH
2
) mode from the reactant
amine, which gives rise to a negative peak centered at ∼1584
-
1 36
cm
. However, the near-exact correspondence between the in
Fig u re 5 . Transmission infrared spectra of ex situ reaction of acetic
anhydride with surface-tethered amine groups in an APTES-derived
layer. The spectrum in (a) shows the APTES layer prior to acetic
anhydride exposure, whereas the spectrum in (b) shows growth of
characteristic amide bands after reaction of amine groups with acetic
anhydride. Both spectra are ratioed to a clean, native oxide-coated
silicon reference (the surface just prior to functionalization with
APTES).
situ and ex situ bands (within 1 cm-1), as well as the lack of any
reactant or solvent bands at those frequencies, allows definitive
identification of amide formation.
Detecting the reaction of substrate-tethered functional groups
in a microreactor is analogous to monitoring chemistry of surface
layers in larger samples. Infrared spectroscopy has traditionally
been one of a battery of techniques used for characterization of
self-assembled monolayers and of any subsequent chemistry
carried out on terminal groups of those layers. Most of this
analysis is ex situ; for example, reflectance IR is often used to
follow reaction of terminal groups of thiol-based SAMs on gold.
Traditional (long-sample) MIR-IR has been used previously to
characterize the reaction of an amine-terminated siloxane mono-
layer on Si, but only ex situ.37 To the best of our knowledge, in
situ MIR-IR of solution-phase reaction of a siloxane monolayer
has been limited to studies of the protonation of terminal
carboxylate groups.38,39 One of these studies38 employed a tradi-
Fig u re 6 . Infrared spectra acquired in situ during reaction of APTES-
functionalized silicon with acetic anhydride in CCl4. Peaks marked
with an asterisk correspond to acetic anhydride, whereas the peaks
at 1204 and 1657 cm arise from growth of surface-bound amide
species. Growth of the amide I band is highlighted on an expanded
scale in the inset. All spectra are ratioed to a spectrum of the initial,
APTES-functionalized surface in CCl4 prior to acetic anhydride
introduction.
tional MIR (Ge) substrate, while the other used a prism cut from
a Si wafer,39 since the ν(CdO) modes of interest are all above
500 cm-1
1
.
-
1
The complexity of the spectra in Figure 5 highlights the
difficulty of detecting changes in surface species in the presence
of the solution reaction mixture, even for a solution containing
only one reactant and no other major components, since MIR-IR
detects both surface-bound and solution-phase species simulta-
neously. While the (solution-phase) acetic anhydride features that
dominate these spectra could be removed either by allowing more
time for equilibration or by flushing the microreactor with solvent
after the surface groups have reacted, either of these approaches
increases the time between introduction of the analyte and its
detection. The present case, therefore, demonstrates the efficacy
of IR detection of surface-bound reaction of a solution-phase
analyte in a more demanding environment than may be required
for many applications. The ability to access most of the mid-IR
fingerprint region provides a large number of IR peaks to aid in
deconvolution of spectra of even more complex mixtures than the
surface/ solution mixture seen in this case and would be essential
for application of such a system in chemometrics.
of the APTES-functionalized surface is presented in Figure 5a.
The most intense peaks, from 1000 to 1200 cm-1, arise from
ν(Si-O) modes of siloxane bridges between the silica surface and
APTES and between APTES molecules in the polymerized
overlayer. A weaker, broad δ(NH
2
) mode appears at ∼1572 cm-1.
2
Reaction of the surface-bound NH groups with acetic anhy-
dride is marked by the appearance of strong amide I and amide
II bands at 1655 and 1559 cm-1, respectively, as shown in Figure
5
b. Conversion of surface amines to amide groups is also
accompanied by a sharp increase of ν(NH) intensity centered at
-
1
3
293 cm (data not shown).
Monitoring this amidization reaction in situ reveals detectable
reaction progress on a relatively fast time scale. A series of three
spectra, acquired at 2.5-min intervals following the introduction
of acetic anhydride solution into the liquid cell, is presented in
Figure 6. Peaks marked with an asterisk (*) arise from acetic
anhydride in solution, identified by comparison both with literature
spectra35 and with peaks from the single-beam spectrum in our
own experiment. Increased intensity in acetic anhydride peaks
(
2
36) The apparent maximum for the δ(NH ) peak is shifted somewhat from the
ex situ spectrum simply by overlap of this mode with the positive-going
amide II peak that grows in during depletion of this mode.
(
(
37) Lee, M.-T.; Ferguson, G. S. Langmuir 2 0 0 1 , 17, 762-767.
38) Cheng, S. S.; Scherson, D. A.; Sukenik, C. N. Langmuir 1 9 9 5 , 11, 1190-
1195.
(39) Gershevitz, O.; Sukenik, C. N. J. Am. Chem. Soc. 2 0 0 4 , 126, 482-483.
(
40) Harrick, N. J. Internal Reflection Spectroscopy, 3rd ed.; Harrick Scientific
(
35) SDBSWeb: http:/ / www.aist.go.jp/ RIODB/ SDBS/ (accessed March 2004).
482 Analytical Chemistry, Vol. 76, No. 21, November 1, 2004
Corporation: Ossining, New York, 1987.
6

Even taking into account the 2-fold increase in sensitivity
generated by use of a two-sided reaction cell instead of the
prefabricated microreactor for this experiment, the system used
in this study yields an easily detectable level of signal (absorbance
Finally, although the in situ experiment described above
demonstrates the utility of our Si-based system for detecting
reaction of surface-bound organic groups, further development
of this approach is required to incorporate in a well-defined way
varied organic surface functionality into the actual microreactors
via premodification of the beveled Si platforms. As noted briefly
above, the ability to modify the silicon native oxide that serves as
the de facto microreactor substrate may be advantageous for
inhibiting unwanted adsorption of solution-phase species, as well
as for tailoring the microreactors for detection of different (and
possibly multiple) analytes via reaction with surface functional
groups.
-
2
change of ∼10 at the amide I frequency in 5 min for a 1-sided
reactor). Obviously, this sensitivity is highly dependent on the
system under study and includes the surface density (and
accessibility) of reactive sites for the analyte, the oscillator strength
of the resulting chemical species formed, and the uniqueness of
the IR signature of the product. Quantification of sensitivity in
the present experiment as a function of the two-dimensional
concentration of surface amine groups is complicated by the
multilayer structure of the APTES-derived overlayer; this could
potentially be addressed using the more well-defined (but less
amine-dense) monoalkoxy-derived monolayer. The issue of a
unique IR signature for product identification is ameliorated to
some extent by the ability to monitor several frequencies at
ACKNOWLEDGMENT
The authors thank the staff at the Microsystems Technology
Laboratories for technical assistance. Funding for the MIT portion
was provided by DARPA and the Microchemical Systems Tech-
nology Center. K.Q. acknowledges funding from the Camille and
Henry Dreyfus Foundation Startup Award Program.
once: for instance, in the present case monitoring absorbance at
204 and 1657 cm- and in the ν(NH) region. Again, our ability
1
1
to access the lower-frequency fingerprint region for organic
species provides a significant advantage in identifying multiple
spectroscopic handles for a given system.
Received for review May 19, 2004. Accepted August 20,
2004.
AC049265C
Analytical Chemistry, Vol. 76, No. 21, November 1, 2004 6483