Sergei G. Kazarian

Macro attenuated total reflection (ATR)-Fourier transform infrared (FT-IR) spectroscopic imaging is a powerful analytical tool, but it is still underexploited in the spectroscopic imaging community. This article presents an overview of current macro ATR-FT-IR imaging approaches and opportunities for studying dynamic processes such as diffusion, sorption, crystallization, and dissolution. 

	Fourier transform infrared (FT-IR) spectroscopic imaging combines the chemical specificity of IR spectroscopy with the spatially resolved nature of a mercury cadmium telluride (MCT) focal plane array (FPA) detector. These detectors typically contain 4096 (64 × 64) or 16,384 (128 × 128) infrared sensitive elements or pixels. Unlike using single-element detectors, where the resulting spectrum is averaging information from different parts of the sample, each detector element in an FPA is focused on a different spatial region of the sample. For heterogeneous samples, this increases sensitivity because with a single-element detector, a small particle may be "averaged out" by the presence of other materials in a single spectrum, but with imaging, such a particle would occupy a few pixels in the array, whose spectra would then be that of only the particle and not the surrounding material (1). The spatially resolved nature of imaging also allows for location-based spectral processing from simple masking to improve average spectra, to clustering or classification techniques to identify regions of interest (2). It is important to note that the use of the term 

 indicates that spectra are obtained from all pixels simultaneously in contrast to 

, where a single-element detector is scanned over the desired area. The main advantage of imaging compared to mapping is the ability to study dynamic systems and samples as spectra can be obtained within a few minutes or less (3). Advances in computational power and optimization of scanning parameters have enabled images to be obtained within 50 ms, fast enough to study moving droplets within a microfluidic device (4).

	Imaging is commonly achieved in transmission, in which the IR beam passes through a sample that is on one or between two IR-transparent windows. The principal disadvantages of transmission are that the sample must be thin (<20 μm) and aqueous samples are particularly challenging to study because of the strong absorption of water. Attenuated total reflection (ATR) mode uses a suitably shaped crystal of high refractive index material, in which incident infrared light undergoes total internal reflection at the interface between the crystal and the sample. At this interface, an evanescent wave penetrates into the sample, whose depth of penetration (

) depends on the angle of incidence (θ), refractive indexes of the crystal (

), and wavelength (λ) of infrared light. The 

, where the intensity of the electric field of incident infrared light falls below e

For mid-IR, the depth of penetration is typically 1–5 μm and the crystal material used depends on the type of sample and material properties required of the crystal. Germanium and zinc selenide can be damaged easily with hard samples, with the latter also having a limited pH range for samples. The refractive indexes of these materials are 4 and 2.4, respectively, with germanium commonly used in microscope ATR objectives because of its high numerical aperture. Silicon and diamond are also commonly used materials, with diamond being particularly robust. They have refractive indexes of 3.4 and 2.4, respectively.

	The FPA detector is often mounted in a separate compartment or macrochamber attached to an FT-IR spectrometer in which different ATR accessories can be positioned. The type of accessory can vary depending on the field of view required and the angle of incidence of the incoming beam (which affects depth of penetration). The most suitable accessory depends on the experiment and different types are demonstrated in the following sections. The FT-IR spectrometer sends the IR beam through one of its side-ports, through the macrochamber and ATR accessory, to the FPA where the IR beam is focused on the surface of the crystal and FPA to obtain spatially resolved IR spectra. It should be noted that because of the exponential decay of the evanescent field, particles can appear smaller than their true size (5), but for many applications, the interest is in chemical and spatial changes at the surface. To obtain two-dimensional (2D) color images, univariate analysis, which involves selecting spectral bands unique to components of interest, can be performed on the spectral data cube. Plotting the integrated absorbance of spectral bands of different components for all pixels against a color scale produces images with respect to each component, where regions of differing concentration can be identified by different colors. A summary of this process is shown in Figure 1. Multivariate approaches are often used when many bands overlap such as in complex pharmaceutical systems (6,7). This versatile approach can also be applied in micro mode with an IR microscope (8), though the focus of this article is on macro-mode imaging (9). Macro-mode imaging has many applications (9,10), and this article provides examples of macro-ATR accessories and case studies where their different properties were used to reveal new information about the samples under investigation.

Overview of ATR Spectroscopic Accessories

Large field of view accessories use no magnifying optics and, thus, have a field of view between 2.5 mm × 3.6 mm (11) and 15.4 mm × 21.4 mm

 (12), depending on the pixel size and number of pixels in the FPA (64 × 64 up to 128 × 128), the angle of incidence, and the accessory used. The Specac Oil Analyser (13), FastIR (14), and Pike Veemax 2 (15) accessories (the latter of which has an adjustable angle of incidence) are examples of large field of view accessories. Schematics of different accessories are shown in Figures 2 and 3. In some accessories, an undesirable gradient of absorbance can be observed along the crystal as the optical configuration permits a spread of different angles of incidence, affecting the depth of penetration (14). In other accessories with movable mirrors, this gradient can be minimized or removed completely (10). Crystals used with such accessories are often zinc selenide, silicon, or germanium. Large field of view accessories are used when spatial resolution is secondary to field of view and, as such, are suitable for pharmaceutical studies (16), microfluidics (17), fingerprint analysis (18), depth profiling (15), protein crystallization (19), and other high-throughput approaches. Because the crystals are not as robust as diamond, they cannot be used for high-pressure experiments or in cases where materials need to be compacted against the crystal surface for good contact. The use of acidic or strong bases can also damage zinc selenide, leaving silicon as the only suitable option. However, silicon's higher refractive index than zinc selenide reduces the depth of penetration, resulting in a weaker probing depth, and its higher cut-off precludes the study of spectral bands below ~1000 cm

. The use of depth profiling by changing the angle of incidence is particularly useful for chemically imaging thin polymer films (15) and has potential in the future for depth profiling other materials, overcoming one of the limitations of transmission. The Oil Analyser has a field of view of 2.5 mm × 3.6 mm with a 64 × 64 pixel detector (11). The Pike VeeMax has a field of view of 7.0 mm × 9.8 mm with a silicon crystal (19) or 5.3 mm × 7.9 mm with ZnSe crystal (20), depending on the angle of incidence, when using a 96 × 96 pixel detector. An expanded field of view accessory as shown in Figure 4 is particularly useful for high-throughput studies in which a larger area (15.4 mm × 21.4 mm) of the ATR crystal can be imaged (12). For an angle of incidence of 45°, images will have an aspect ratio of 1:√2 unless correcting optics are used, and other angles will produce images of different aspect ratios as shown in Figure 3.

The imaging Golden Gate accessory, initially developed by Specac for nonimaging applications and, thereafter, containing aspect-ratio correcting optics for imaging, contains a diamond ATR crystal mounted in tungsten carbide (Figure 5). The diamond is cut to form a 45° inverted prism with a top surface of 2 mm × 2 mm. Lenses, normally zinc selenide, condense the incident beam into the diamond, and a similar lens on the opposite side expands the beam, resulting in an imaging area of approximately 656 μm × 579 μm for a 64 × 64 FPA detector, a magnification of approximately 4×. It is important to note that with noncorrecting optics, the imaging area is 820 μm × 1140 μm (21). This accessory allows for in situ compaction and dissolution of pharmaceutical tablets (22), high-pressure and high-temperature experiments (23), and, more recently, the use of apertures for variable angles of incidence to eliminate dispersion artefacts because of high refractive index samples (24). The use of apertures in this system allows for depth profiling in a similar manner to that of the large field of view accessories (25).

Dynamic Applications with Macro ATR-FT-IR Spectroscopic Imaging

Drug release from solid oral dosage forms is one of the most common methods by which drugs are administered to patients, and the release mechanisms that occur after a tablet is ingested determine the bioavailability of the drug. Drugs or active pharmaceutical ingredients (APIs) are often complex molecules with low aqueous solubilities, so one important role of other components or excipients within the tablet is to aid solubilization of the drug as it is released from the tablet. Conventional dissolution studies use United States Pharmacopeia (USP) dissolution tests, in which tablets are dissolved under well-defined flow conditions and the amount of dissolved drug is measured using ultraviolet–visible (UV–vis) spectroscopy. This approach has drawbacks such as not giving information about mechanisms within the tablet as the drug is released, and it only shows the amount of drug dissolved over time, not necessarily released in other solid forms such as particles or precipitates. This distinction is important as the drug can be released from the tablet, but in a form which is not bioavailable (that is, a solid or undissolved form). Other approaches such as UV imaging (26), Raman (27), and magnetic resonance imaging (MRI) (28,29) can also be used noninvasively to obtain information from a tablet as it dissolves. MRI can give three-dimensional (3D) images of the whole tablet under USP IV conditions, but its acquisition times for each 3D image are long, up to 30 min (28). FT-IR spectroscopic imaging of tablets can be done in transmission mode, but the tablets must be very thin to reduce the effects of strong water absorbance (although the use of D

O can aid with this) (30) and formulated to reduce scattering of the infrared light, limiting the applicability of this approach in studying pharmaceutically relevant tablets.

Macro ATR-FT-IR spectroscopic imaging is particularly suited to studying drug release from microparticles (31) and tablets (32) because they do not need to be thin and components within the layer in contact with the crystal can be imaged over time (33). Importantly, spectroscopic imaging can reveal drug crystallization because of changes in molecular interactions (16). Using a diamond ATR accessory, the tablet can be compacted in situ, guaranteeing good contact with the crystal and with a suitable flow cell (34), dissolution can also be carried out without removing the tablet from the crystal. A UV–vis detector connected in-line downstream can give the amount of dissolved drug over time in a comparable way to USP and using a transparent flow cell (with ex situ compaction), visible images can also be captured (32). With the latter approach, differences have been demonstrated between visibly observed dissolution fronts (that is, water penetration into the tablet, drug dissolution and diffusion, and so on) and those produced from FT-IR imaging. This result was particularly important because relying on changes in refractive index or color changes is not a reliable method to study dissolution and it would only be applicable to certain formulation types. It is also possible to use a large field of view ZnSe variable angle accessory for an overall view of the tablet as it dissolves. This approach has been combined with computer modeling of drug release, where information from FT-IR imaging was used for model validation and predictive capability testing (35). The combination of computer modeling and FT-IR imaging opens up the possibility of formulation optimization for tablets of different shapes and composition (20).

	The function of a protein is governed by its tertiary or 3D structure, and X-ray crystallography is usually used to obtain structural information. However, crystallization is a major bottleneck in protein screening because the conditions under which crystals will form and grow are narrow. Crystallization normally occurs as the solution is brought to super-saturation and nucleation followed by crystal growth. However, excessive nucleation can cause crystals to be too small and have insufficient space to grow into long-range ordered structures, or the protein may precipitate into an amorphous solid. Thus, it is necessary to optimize the crysallization conditions based on temperature, pH, and the presence of other chemicals or "leads" whose function is to initiate crystallization or stabilize growing crystals such that they become large enough for X-ray crystallography. One problem is that other species crystallize in addition to the protein, and determining if a particular crystal is proteinaceous is often chemically or physically destructive.

	Consequently, it is desirable to have a chemically specific, nondestructive method that can determine if a crystal is a protein and if possible, obtain structural information at the same time. UV fluorescence imaging can be used, but with wide absorption bands, it is not as chemically specific as IR spectroscopy. In addition, IR spectroscopy allows for the determination of secondary structure. The crystallization of proteins directly onto an ATR crystal, combined with imaging, allows for determining which crystals are indeed proteins and how quickly they grow over time. Unlike Raman spectroscopy, where individual crystals would have to be located and scanned, macro ATR-FT-IR imaging obtains spectra from all protein crystals simultaneously. The crystallization of lysozyme has been demonstrated on a ZnSe ATR crystal, and the use of this large field of view also opens up the possibility for high-throughput approaches, for example, different concentrations of NaCl and protein solution can be imaged over time in microwells (11). However, the spatial resolution suffers with a large field of view, and smaller protein crystals may not be detected. The use of a diamond ATR crystal (11) demonstrated protein crystallization at a higher spatial resolution, and combining this crystal with a microfluidic approach (36) would enable high-throughput screening of different conditions, considerably speeding up the discovery of the optimum set of parameters.

	Another example of a high-throughput approach is the surface modification of the ATR crystal itself (19). In this example, a gradient of hydrophobicity was created using a modified vapor deposition approach, in which octyltrichlorosilane is deposited on a silicon ATR crystal in one direction. It was shown that lysozyme adsorption and crystallization decreased as the hydrophobicity decreased along the crystal. The use of an FPA detector allowed for pixel binning in the direction normal to the gradient on the crystal surface, significantly improving the signal-to-noise ratio. Surface modification with nanoparticles for surface-enhanced infrared absorption spectroscopy (SEIRAS) can increase sensitivity by two orders of magnitude and enables the study of monolayers (37). Nanoparticle layer thickness and uneven deposition can present a problem with loss of spectral intensity and uneven enhancement. However, there exists great potential for imaging in this area and though this has been applied to polymers (10), it has yet to be applied to proteins.

	Another exciting application of macro ATR-FT-IR imaging is for the analysis of materials processed with supercritical fluids. Supercritical CO

 is a widely used supercritical solvent because it is nontoxic, environmentally benign, and relatively inexpensive. It can also be removed from systems at ambient temperature and pressure unlike many organic solvents because it is a gas under these conditions. An important effect is the plasticization of polymers in which the supercritical CO

 is absorbed into the polymer, causing it to swell and the glass transition temperature (

) to be lowered. This aids the diffusion of materials into the polymer and the processing of the polymer itself via extrusion or blending (38,39). Using macro ATR-FT-IR imaging with a high-pressure cell presents a unique opportunity to study polymer behavior under high pressures of CO

. In an article demonstrating this approach (40), two polymer systems were studied. The first was a polystyrene–polyvinylmethylether blend, which has been observed to have a lower critical solution temperature, thus increases in temperature result in phase separation. In this system, phase separation was observed after pressurizing the polymer blend with CO

 to 60 bar and back to atmospheric; the CO

 plasticized the polymers allowing phase separation to occur. The imaging approach allowed in situ observation of the formation of polystyrene- or polyvinylmethylether-rich regions, demonstrating the effect of supercritical CO

 in a similar way to raising the temperature. The second was a study into the sorption of supercritical CO

 into polymethylmethacrylate (PMMA) and polyethylene oxide (PEO) films simultaneously at different temperatures. It was observed that CO

 solubility decreases as a function of increasing temperature with PMMA. For PEO, little sorption was observed at 30 °C, but at 50 °C after the polymer had melted, the absorbed CO

 concentration increased significantly. This demonstrated a high-throughput approach for analyzing different polymers under identical environmental conditions and with imaging, it was also possible to observe polymer interdiffusion.

	Heating of crude oil for distillation accounts for a large fraction of energy use during oil refining, and fouling caused by precipitating species is a major source of inefficiency (41). Fouling materials are complex, containing insoluble impurities, inorganic species, and 

, which are polycyclic aromatic hydrocarbons, one of the heaviest fractions of crude oil. Asphaltenes are a major contribution to fouling and their composition and tendency to form depend on the type of crude oil, temperature, and pressure of the system. There are many analytical approaches for studying asphaltenes, each with their own advantages and drawbacks. FT-IR single-element spectroscopy has been used to study petroleum asphaltenes (42); however, with spectroscopic imaging it is possible to analyze crude oil behavior under different or industrially relevant conditions to observe asphaltene formation and obtain chemical information from their deposits in situ. Using a diamond ATR accessory, solid crude oil deposits were initially studied and the use of custom apertures in this accessory removed anomalous dispersion, enabling different chemical clusters to be observed within the deposit (24). Applying this method to study deposition in situ with crude oil was achieved by Gabrienko and colleagues (43), where 

-heptane was used to precipitate out asphaltenes. The approach demonstrated the viability of using in-situ macro ATR-FT-IR imaging to study these deposits as they formed over time, and it was subsequently used to study asphaltene deposition from different geographical sources of crude oil (44). From this, it appeared that the total asphaltene concentration increased with increasing sulfur content (45) with sulfoxides observed as the dominant sulfur species, but other sulfur species that may be present have spectral bands beyond the lower limit of the FPA detector (<900 cm

) (46). Further investigations into the stability of oils under different conditions such as temperature and pressure are possible with an imaging diamond ATR accessory (43,44), attesting to the versatility of this approach to studying these samples. Thus, in the future, the information gained can be used to find optimal physical or chemical conditions to prevent asphaltene deposition.

	Microfluidic technology continues to expand as an analytical tool because of the ability to manipulate very small volumes (up to microliters) of fluids under conditions where laminar flow and diffusive heat and mass transfer tend to dominate. The importance of label-free methodologies such as FT-IR imaging as a detection tool has recently been discussed by us in this publication (47). This miniaturization and control of flow allow for so-called "lab on a chip" devices to be manufactured. A popular material for manufacturing microfluidic devices is polydimethylsiloxane (PDMS), in which channels can be created using a variety of processes such as soft lithography (48). Direct printing using paraffin waxes has also been proven to be particularly effective for rapid prototyping of devices. With this method, it is possible to print microfluidic channels directly onto infrared transparent windows (49), enabling analysis of reactions by FT-IR spectroscopic imaging (50). However, there is the problem of high water absorption in these transmission devices. Printing wax directly onto an ATR crystal (49), or using a suitable PDMS device, removes this caveat and has allowed mixing of fluids such as PEG and water, or H

O to be observed (17). In these instances, a "Y"-shaped device was used with two fluids entering at the top of the "Y" and mixing, establishing steady-state conditions along the channels. Imaging allows the progression of chemical mixing (in the case of PEG and water) and reaction (in the case of H

O) along the channel to be monitored, with a gradient being established in the former case, or the reaction products (HDO) in the latter case formed at the interface between the two input fluids. Under steady-state conditions, the ATR crystal can be moved so as to map a larger area of the microfluidic device (17). Fast imaging in transmission has been achieved using "kinetic mode," where interferograms are obtained and saved without coaddition, enabling spectral images to be acquired in as little as 50 ms (4). The speed of acquisition is affected by spectral resolution (which determines the distance the moving mirror has to move in the interferometer), the spectral range, and the refresh rate of the FPA. The latter depends on the FPA window size, hence smaller window sizes (64 × 64) will scan faster than larger sizes (128 × 128). This fast scanning is important for imaging segmented flows, in which one phase is separated by another. This allows each droplet to become a "micro-batch reactor" and eliminates axial flow within the channel. In ATR mode, the droplet phase can become detached from the ATR crystal surface, although this can be mitigated by reducing the flow velocity. Monitoring these droplets as they progress through the device allows the reaction progress to be analyzed, and changing the input chemistry of each droplet as it is generated facilitates a high-throughput approach for measuring and optimizing reaction chemistry (47,50).

	In this article, several case studies were presented that demonstrate the versatility of macro ATR-FT-IR spectroscopic imaging. The selection of the appropriate accessory, crystal, and angle of incidence depends on the system being studied and also on the required spatial resolution, field of view, sample type, and desired experimental conditions. As ATR-FT-IR spectroscopy relies on the evanescent field, it does not suffer from strong absorption of water, or other strongly absorbing samples, which can obscure weaker bands and enables a variety of aqueous or other systems to be quantitatively studied. The FPA detector increases sensitivity for heterogeneous samples as each pixel corresponds to a unique location on the sample surface. This feature is particularly powerful when combined with a macro-ATR approach for analyzing pharmaceutical drug or protein crystallization, asphaltene deposition, phase separation, and fluid analysis in microfluidic devices. This approach also facilitates high-throughput analysis, where many different samples can be analyzed under identical conditions, increasing experimental reliability and reducing parameter space exploration times. The time-resolved nature enables dynamic systems to be imaged in-situ, where depending on the spectral quality and range required, spectral data cubes can be obtained within 50 ms. As macro ATR-FT-IR imaging can analyze a variety of dynamic samples and systems with high resolution and chemical specificity, its applications in emerging areas such as biotechnology, supercritical fluids, microfluidics, and novel formulation and material design are far-reaching.

	S.G. Kazarian acknowledges research funding from the European Research Council under the European Community's Seventh Framework Programme (FP7/2007-2013)/ERC advanced grant agreement no. (227950).

	(2) P. Lasch, W. Haensch, D. Naumann, and M. Diem, 

	(4) K.L.A. Chan, X. Niu, A.J. deMello, and S.G. Kazarian, 

	(5) N.J. Everall, I.M. Priestnall, F. Clarke, L. Jayes, G. Poulter, D. Coombs, and M.W. George, 

	(6) J. Tetteh, K.T. Mader, J.M. Andanson, W.J. McAuley, M.E. Lane, J. Hadgraft, S.G. Kazarian, and J.C. Mitchell, 

	(7) J. van der Weerd and S.G. Kazarian, 

, (John Wiley & Sons, Ltd, 2014), pp. 1–26, DOI: 10.1002/9780470027318.a5609.pub2.

	(11) K.L.A. Chan, L. Govada, R.M. Bill, N.E. Chayen, and S.G. Kazarian, 

	(13) J. van der Weerd and S.G. Kazarian, 

	(14) E. Wessel, G. Heinsohn, H. Schmidt-Lewerkuehne, K.-P. Wittern, C. Rapp, and H.W. Siesler, 

	(17) K.L.A. Chan, S. Gulati, J.B. Edel, A.J. de Mello, and S.G. Kazarian,

	(18) C. Ricci, P. Phiriyavityopas, N. Curum, K.L.A. Chan, S. Jickells, and S.G. Kazarian, 

	(19) S. Glassford, K.L.A. Chan, B. Byrne, and S.G. Kazarian, 

	(20) J.A. Kimber, S.G. Kazarian, and F. Štepánek, 

	(22) J. van der Weerd, K.L.A. Chan, and S.G. Kazarian, 

	(23) O.S. Fleming and S.G. Kazarian, in 

 (Wiley-VCH Verlag GmbH & Co. KGaA, 2006), pp. 205–238.

	(25) T. Frosch, K.L.A. Chan, H.C. Wong, J.T. Cabral, and S.G. Kazarian, 

	(26) J. Pajander, S. Baldursdottir, J. Rantanen, and J. Østergaard, 

	(27) J. Østergaard, J.X. Wu, K. Naelapää, J.P. Boetker, H. Jensen, and J. Rantanen, 

	(28) C.A. Fyfe, H. Grondey, A.I. Blazek-Welsh, S.K. Chopra, and B.J. Fahie, 

	(29) P. Kulinowski, A. Mlynarczyk, P. Dorozynski, K. Jasinski, M.H. Gruwel, B. Tomanek, and W. Weglarz, 

	(30) C.A. Coutts-Lendon, N.A. Wright, E.V. Mieso, and J.L. Koenig, 

	(31) H. Keles, A. Naylor, F. Clegg, and C. Sammon, 

	(34) J. van der Weerd, K.L. Andrew Chan, and S.G. Kazarian, 

	(35) J.A. Kimber, S.G. Kazarian, and F. Štepánek, 

	(36) B. Zheng, L.S. Roach, and R.F. Ismagilov,

	(42) J.B. Cooper, K.L. Wise, W.T. Welch, M.B. Sumner, B.K. Wilt, and R.R. Bledsoe, 

	(43) A.A. Gabrienko, C.H. Lai, and S.G. Kazarian, 

	(44) A. Gabrienko, V. Subramani, O. Martyanov, and S. Kazarian, 

	(45) A.E. Pomerantz, D.J. Seifert, K.D. Bake, P.R. Craddock, O.C. Mullins, B.G. Kodalen, S. Mitra-Kirtley, and T.B. Bolin, 

	(46) T. Tveter, P. Klaeboe, and C.J. Nielsen, 

	(49) K.L.A. Chan, X. Niu, A.J. de Mello, and S.G. Kazarian, 

 is a professor of physical chemistry with the Department of Chemical Engineering at Imperial College London in London, England. Direct correspondence to:  s.kazarian@imperial.ac.uk

Articles

Macro attenuated total reflection FT-IR spectroscopic imaging is a powerful and underutilized tool. This article presents an overview of approaches and opportunities for using this method to study dynamic processes such as diffusion, sorption, crystallization, and dissolution.

Macro ATR-FT-IR Spectroscopic Imaging of Dynamic Processes

Fourier transform infrared (FT-IR) spectroscopic imaging is a highly versatile technique that can be applied to a wide range of systems. This article summarizes some of the recent efforts developing applications of FT-IR imaging for microfluidics. The main advantage of FT-IR imaging compared to traditional imaging methods is that it is a label-free imaging technique. There is no need to develop tags or labels, multiple components are simultaneously traced, and images can be taken without disturbing the sample. All of these advantages are accompanied with a near-video frame rate acquisition speed. Different approaches to obtain FT-IR images (transmission and attenuated total reflection mode) of microfluidic devices are discussed including novel ways to create microfluidic devices.

	Microfluidic technology is a powerful tool that has a wide range of applications in chemical and biological analysis (1). The improved heat and mass transfer in microfabricated systems in comparison to traditional processes provides the opportunity to increase control over the yield, the speed of the turnover of experiments for high-throughput studies, and reduce the amount of precious reagents used (2). Knowing the chemical composition at a specific point in the microfabricated device can aid in the design and optimization of these devices (3,4).

	Detection in microfabricated devices often relies on additional tracers or tags to visualize the existence and the distribution of particular components. One of the most widely used methods is confocal fluorescence microscopy. The advantage of fluorescence is that it is highly sensitive (5) and can often achieve sufficient spatial resolution. However, finding a suitable fluorescence agent and photobleaching are some of the remaining challenges to overcome. The tracer also has to be inert enough that it will not decompose or interfere with the process of interest (for example, chemical reactions and diffusion). Raman, surface-enhanced Raman spectroscopy (SERS), and coherent anti-Stokes Raman scattering (CARS) have been shown as promising label-free detection methods in microfluidic systems (6–10). In this article, we discuss some recent studies that have demonstrated that Fourier transform infrared (FT-IR) spectroscopic imaging can be a powerful detection tool, as was proposed earlier (11), to extract spatially resolved rich chemical information from microfluidic devices in a label-free manner.

	FT-IR imaging has been used as a highly versatile analytical method, providing spatially resolved, chemically specific information for studying multicomponent systems (12). Recently, conventional FT-IR microscopy using a single-element detector was applied to analyze fast reactions in microfluidic flows where spectral information from a specific location in a channel was obtained (13). FT-IR spectroscopic imaging combines the benefits of imaging and spectroscopy providing chemical maps of studied samples. The chemical specificity comes from the intrinsic molecular vibrations, revealed by spectral bands, while spatial information is collected from the focal plane array (FPA) detectors. An FPA detector comprises thousands of detector pixels, each collecting an infrared spectrum so that thousands of infrared spectra are collected simultaneously in a single imaging measurement. This approach reduces the time required to collect all the spectral data when compared to point-to-point mapping using a single element detector with apertures (14).

	Characteristic spectral bands can then be used as markers for specific components, which allows their distribution within the imaged area to be revealed. Multivariate methods are also available to separate components with similar spectral features to generate representative maps for each component. A full mid-IR spectrum (4000–900 cm

) is collected from each detector pixel, and multiple components can be simultaneously tracked in one single imaging measurement either using univariate or multivariate approaches. This multicomponent imaging feature is a significant advantage compared to ordinary fluorescence measurements in which only one component is traced at a time. The absorbance of a spectral band is directly correlated to the concentration of the component so that results obtained can be quantified.

	Because all spectra are collected simultaneously, FT-IR imaging using FPA detectors is suitable for studying dynamic systems such as diffusion and dissolution (15–20) and has great potential for high-throughput applications, especially when combined with the attenuated total reflection (ATR) sampling method (11,21).

	In the ATR approach, the infrared light is totally internally reflected at the interface between the high refractive index infrared-transparent element and the lower refractive index medium (the sample). Common elements used for ATR measurement are diamond (

≈ 4). In this measurement mode, the infrared light probes into the sample as an evanescence wave, the field strength of which decays exponentially into the sample. The depth of penetration is on the order of a few micrometers and the resultant pathlength (the equivalent pathlength) is also in the order of several micrometers depending on the refractive indices of the sample, the ATR element, and the incident angle. The ATR element is selected such that it will provide a relatively small, but known, pathlength that will produce an appropriate on-scale absorbance. Other considerations include the physical and chemical properties of the ATR element for specific applications. The generally small pathlength from ATR mode measurement offers the opportunity to measure materials that have strong mid-IR bands such as water and reduces sample preparation because no microtoming or polishing of the sample is required. The possibility of acquiring spectra using ATR mode increases the applicability to study samples in situ and to monitor processes on line. The simplicity of measurement allows rapid analysis of an array of deposited samples in the imaging area (11,21–23).

	Because the evanescent wave probes a relatively shallow layer, the sample must be in close contact with the ATR element to be exposed to and interact with this evanescent wave. This sampling requirement for ATR measurement is automatically satisfied when the sample being measured is a fluid. FT-IR imaging in ATR mode is therefore suitable to study microfluidic systems. Microfluidic devices are often made of polydimethylsiloxane (PDMS) using templates in which channels are created when the PDMS stamp is bonded to a flat surface such as glass. With ATR FT-IR imaging, instead of bonding the PDMS device to glass, the PDMS device is bonded to the imaging surface of the ATR element so that the measuring surface forms one of the walls of the microfluidic channel. The fluid flowing inside the channel that comes to close proximity with the surface of the ATR element is measured.

ATR elements are often expensive and cannot be used as disposable units; therefore, the bonding of PDMS onto the ATR surface must be reversible. PDMS will bond weakly with a clean, oil-free surface. However, most microfluidic experiments cannot rely on this weak bonding as the seal for the channels because pressure from the fluid will cause liquid to escape by overcoming this weak adhesion force. Fortunately, this issue can be easily rectified. A rigid poly(methyl methacrylate) (PMMA) sheet (or any rigid, ideally transparent material for visible inspection) can be used as a press to provide a stronger bonding between the PDMS and the ATR element. One of the inherent advantages of ATR imaging measurements is that the presence of foreign particles (dust) that may affect contact quality between the PDMS chip and the ATR surface can be easily detected (24). Leakage of fluids through the gap between the ATR surface and the PDMS chip can also be detected from the image as the escaped fluid will be in contact with the ATR element. Figure 1 shows the typical "Y-junction" PDMS microfluidic device that we have used as the first test of the principle (25). An image of the device can be generated based on the band specific to PDMS. When two different fluids are injected from the two entrances of the microfluidic device, the mixing between the two fluids can be studied. As an illustration, we injected water (75 μL/h) through the right entrance and PEG 200 (25 μL/h) through the left channel. Because both fluids are transparent in visible range, it was not possible to visualize the flow pattern using visible light without adding dye to the solution. Using ATR FT-IR imaging, the two fluids are easily distinguished. Water can be traced using the ν(O-H) band at 3700–3100 cm

, and PEG 200 can be traced using the characteristic ν(C-H) bands at 3000–2800 cm

. A red-green-blue (RGB) figure (Figure 1) has been created to show the distribution of PDMS (red), water (blue), and PEG (green). The image allows us to immediately identify the effect of viscosity to the flow pattern. The water stream, despite having a higher flow rate, was constrained to the right while most of the channel was occupied by the PEG 200 because of the difference between their viscosities.

	There are both advantages and disadvantages to working in transmission mode. One of the main advantages is that the whole volume of the fluid is analyzed as compared to the first few micrometers from the surface in ATR mode. However, for the same reason, one of the main disadvantages is that the fluid thickness is often limited to ~10 μm to ensure that most of the spectral bands are within the linear detection limit (around 0.8 absorbance units). For the stronger spectral bands such as the band from water ν(O-H) mode, the thickness may need to be further reduced to 2–3 μm. In general, therefore, certain parts of the spectral region will not be accessible in transmission with a workable pathlength when the fluid studied has strong spectral bands.

	To study in transmission mode, the top and bottom layer of the microfluidic device has to be transparent in the infrared region. Common infrared-transparent materials used for making spectroscopic windows include BaF

, and ZnSe. Hygroscopic materials such as NaCl and KBr are not suitable for this purpose because water will dissolve the window. Silicon is also transparent in the mid-IR region and has great potential for use as a material to construct microfluidic devices for FT-IR imaging because etching techniques for silicon wafers are well established and its surface properties can be tuned with self-assembled monolayers. Other infrared transparent windows are often not malleable and specific etching techniques may first need to be developed (26). Furthermore, these crystals are often too expensive to be used as a one-off device (a pair of large windows would cost as much as $320). The ability to create reusable microfluidic devices on these windows would be an important step forward for the wider application of this approach.

	One of the possibilities for producing microfludic devices is direct wax printing. Machines that can print molten wax droplets at precise locations are commercially available. Previous studies used wax printers in one of two ways, either to create the master, which was then used to cast PDMS devices to obtain the negative design of the printed features (27,28), or printing the device on paper to create paper microfluidics (29,30). Our wax printing approach manufactures microfluidic devices with free-flowing channels directly without the need for casting PDMS over the print. We have applied this method to create various microfluidic devices on a CaF

 substrate. The procedure is relatively straightforward. Microfluidic devices are first designed and programmed into the computer that operates the microdrop machine. Molten wax droplets, each of which has a controllable diameter of 30–70 μm, are printed directly onto the CaF

 window can be precoated with a thin layer of polymer to create a hydrophobic surface when necessary. The adjacent molten wax droplets join together and solidify to form a wax wall. Figure 2 shows a photograph of a T-junction microfluidic device created by the wax printing method.

A spacer is added that determines the depth of the channel, which is also the pathlength of the measurement. The microfludic channel is created when another CaF

 window is pressed on top and held together by screws, for example. Experiments can be conducted immediately after the printing process, and the infrared-transparent windows can be reused because the wax channels can be removed easily. Depending on the complexity of the design, the printing time is typically less than 10 min. New microfluidic devices can be modified and created in a matter of minutes. The size of the wax droplet is on the order of 50 μm and the machine has an accuracy of 40 μm, which defines the resolution of the microfluidic device. So far, we have only experimented using wax that is vulnerable to many organic solvents and oil. However, waxes that are highly chemically resistant are also available; we plan to test these and present the results in the future.

	In transflection mode, the sample is placed on an infrared-reflective substrate surface where the IR radiation passes through the sample, is reflected off the surface of the substrate, and passes through the sample again before being collected with the sample objective. For microfluidic applications, the advantage of this approach is that only the top surface needs to be infrared transparent. When an open channel is used (that is, when the top surface is not covered and is open to air), there is no need to construct the microfluidic device in an infrared-transparent material. The main requirement is that the bottom of the channel is infrared reflective, which can be easily achieved. However, the thickness constraint that applies in transmission mode measurement is more severe because the pathlength is twice the thickness of the fluid. Nevertheless, such an approach was used in the study of biofilms and live cells in an open channel setting (31).

	In comparison to laminar flow, segmented flow and droplet flow offer the possibility to perform high-throughput studies without increasing the size or complexity of the microfluidic device (32). Although each droplet or segment is compartmentalized by the immiscible inert phase, each segment can be considered as individual experiments.

FT-IR imaging of droplets or segments can be captured easily when they are stationary. The applications of segmented flows include protein crystallization, live cells in droplets, and other slow processes. Many of these research ideas are yet to be realized in combination with FT-IR imaging. One of the main challenges in imaging segmented flows is when one wants to capture the droplet or segments while they are moving. The acquisition time for a typical FT-IR imaging measurement is on the order of seconds. Although this imaging speed is already a great improvement from the early FT-IR imaging systems; it is often not fast enough to capture the moving droplets. Fortunately, through optimization of the data acquisition process, FT-IR imaging of moving segments or droplets in microfluidic channels has been demonstrated (33). Traditionally, the acquired data in an FT-IR imaging measurement are the average of several scans to improve on the signal-to-noise ratio. (In theory, the signal-to-noise ratio is improved by the square root of the number of scans taken.) However, to capture droplet flow, only one scan can be collected. The time taken for one scan is governed by the spectral range and spectral resolution set in the experiment as well as the number of detector pixels used. For an imaging experiment with a 64 × 64 pixel array, 8 cm

 spectral range, the scanning time is approximately 600–700 ms. With current technology, one needs to reduce the amount of data collected to improve this speed. For example, most of the spectral information is contained in the so-called fingerprint region of the infrared spectrum (between 1800 and 500 cm

). Because the detection limit for the FPA detector is around 900 cm

 is used as the substrate material, there is an opportunity to reduce the amount of data collected without losing significant chemical information by reducing the spectral range to 1800–1000 cm

. This can increase the speed of the scan fourfold. By reducing the spectral resolution from 8 cm

, the signal-to-noise ratio will be improved and the spectral bands will be broader while the speed of the scan can be doubled, quadrupled, or even octupled. Together, this brought the scanning time down to well below 100 ms (33).

First, we attempted imaging of droplet flow using the ATR approach. A microfluidic chip with a "T-junction" design was mounted on the measuring surface of the ATR element (Si was used in this case). The surface of the Si was pretreated to increase the hydrophobicity. The oil (FC-40) and water were introduced at flow rates of 2 μL/min and 1 μL/min, respectively. The result is shown in Figure 3. The images of oil have shown that the water droplets flowing in the oil have been captured. In the region in which a water droplet is present, the oil absorbance is lowered. However, spectra extracted from the water droplet region have shown a high absorbance of oil while the water absorbance was very low (see spectra in Figure 3). The reason for this observation is that when water droplets are moving at high speed inside the channel, there is a thin layer of oil in between the water droplet and the ATR element (see Figure 3). Because in ATR measurements only the layer of several micrometers from the surface of the ATR element is detected, the resultant spectral bands were contributed mostly by the oil layer on the surface of the ATR element. On the other hand, the spectrum of pure water can be obtained from the imaging data when the water droplet is stationary. Comparing the absorbance of the oil in the water droplet region to the bulk oil, the thickness of this oil layer can be calculated using the following equation (34,35):

) is the absorbance of the oil at the water droplet region, 

(∞) is the absorbance of the oil at the region without the water droplet, and 

 is the depth of penetration. The thickness of the oil layer is calculated as 790 nm.

To capture the spectrum of water more clearly in the segment, imaging in transmission mode needs to be used. Figure 4 shows the image, the schematic diagram, and the extracted spectra of a water droplet flowing in oil in a wax-printed microfluidic device with a pathlength of 25 μm. The absorbance of water is far more pronounced compared to the spectrum extracted from the ATR measurement shown in Figure 3. However, as discussed in the transmission mode section, the large pathlength could lead to some parts of the spectral region becoming inaccessible. In this case, the spectral region in which oil absorbs IR light strongly (1100–1350 cm

 ) is not accessible in the oil-rich region of the image. This is in contrast to the spectra extracted from the ATR imaging measurement, in which the absorbance of all spectral bands was less than 0.8, the value below which analysis can be quantitative. On the other hand, a large amount of the spectral range remains accessible in the water droplet region (1600–1000 cm

). We have demonstrated previously that, despite the use of only one scan per image, the presence of protein can be detected at the ~1 mM level in the water segment in transmission mode by observing the amide II band at 1560 cm

	One of the potential benefits of combining FT-IR imaging with microfluidics is the in situ study of chemical reactions inside the channel. This allows for optimization of reaction conditions to obtain better yields, or understanding the diffusion and reaction kinetics inside the microfliudic device. This technique can be developed into a tool for chemical detection in microreactors for studying the effect of changing parameters to the overall efficiency. One of the examples (25) that demonstrated this potential was the observation of isotopic exchange between water and D

O forming HDO (semiheavy water). The spectral bands of H

O, and HDO are readily distinguishable. Because FT-IR measurements can be quantified, the concentrations of reactants and products at different points of the devices can be obtained. Another possible experiment is a reaction in a two-phase system in which one of the reactants is dissolved in the oil phase while the other reactant is dissolved in aqueous phase (37). The resultant effect is that the reaction can only happen at the interface between the two fluids. Diffusion profiles and the formation of product in such systems can be monitored simultaneously using the FT-IR imaging approach (37). This detection approach is set to benefit automation of reactions and other processes in microreactors (38).

	In this article, we have discussed a methodology that permits the in situ chemical imaging of flows in microfluidic channels using FT-IR spectroscopic imaging that does not require added labels or dyes. This inherent chemical specificity of FT-IR imaging significantly adds to the detection capabilities of flows in microfluidic devices because it obtains quantitative chemical information as a function of space and time. This chemical imaging methodology has wide applicability to the study of dynamic systems ranging from the analysis and modeling of mixing in laminar flows to studies of reactions in segmented flows and separating live cells in moving droplets. We hope that this article can stimulate further applications of FT-IR spectroscopic imaging to study processes and reactions in microfabricated devices and microreactors because this methodology has great potential for in situ fast chemical analysis of microfluidic flows.

	SGK acknowledges research funding from the European Research Council under the European Community's Seventh Framework Programme (FP7/2007-2013)/ERC advanced grant agreement no. [227950].

is a Professor of Physical Chemistry with the Department of Chemical Engineering at Imperial College London in London, UK. Please direct correspondence to: s.kazarian@imperial.ac.uk.

	(1) A. Huebner, S. Sharma, M. Srisa-Art, F. Hollfelder, J.B. Edel, and A.J. Demello, 

	(3) J. Aubin, M. Ferrando, and V. Jiricny, 

Chemical Engineering and Processing: Process Intensification 

	(6) M. Lee, J.P. Lee, H. Rhee, J. Choo, Y.G. Chai, and E.K. Lee, 

	(7) J.B. Salmon, A. Ajdari, P. Tabeling, L. Servant, D. Talaga, and M. Joanicot, 

	(8) K.R. Strehle, D. Cialla, P. Rosch, T. Henkel, M. Kohler, and J. Popp, 

	(9) A. Marz, T. Henkel, D. Cialla, M. Schmitt, and J. Popp, 

	(10) G. Bergner, C.R. Albert, M. Schiller, G. Bringmann, T. Schirmeister, B. Dietzek, S. Niebling, S. Schlucker, and J. Popp, 

	(13) C. Wagner, W. Buchegger, M. Vellekoop, M. Kraft, and B. Lendl, 

	(14) E.N. Lewis, P.J. Treado, R.C. Reeder, G.M. Story, A.E. Dowrey, C. Marcott, and I.W. Levin, 

	(20) D. Velasco, C.B. Danoux, J.A. Redondo, C. Elvira, J. San Roman, P.S. Wray, and S.G. Kazarian, 

	(24) C. Ricci, K.L.A. Chan, and S.G. Kazarian, 

	(25) K.L.A. Chan, S. Gulati, J.B. Edel, A.J. de Mello, and S.G. Kazarian, 

	(26) T. Pan, R.T. Kelly, M.C. Asplund, and A.T. Woolley, in 

26th International Symposium on Capillary Chromatography and Electrophoresis 

(Elsevier Science Bv, Las Vegas, Nevada, 2003), p. 231.

	(27) J.C. McDonald, M.L. Chabinyc, S.J. Metallo, J.R. Anderson, A.D. Stroock, and G.M. Whitesides, 

	(28) G.V. Kaigala, S. Ho, R. Penterman, and C.J. Backhouse, 

	(29) Y. Lu, W.W. Shi, L. Jiang, J.H. Qin, and B.C. Lin, 

	(30) E. Carrilho, A.W. Martinez, and G.M. Whitesides, 

	(31) H.Y.N. Holman, R. Miles, Z. Hao, E. Wozei, L.M. Anderson, and H. Yang, 

	(32) S.Y. Teh, R. Lin, L.H. Hung, and A.P. Lee, 

	(33) K.L.A. Chan, X. Niu, A.J. deMello, and S.G. Kazarian, 

	(36) K.L.A. Chan, X.Z. Niu, A.J. de Mello, and S.G. Kazarian, 

Annual Review of Analytical Chemistry, Vol 3

, E.S. Yeung and R.N. Zare, Eds. (Annual Reviews, Palo Alto, California, 2010), p. 19.

Fourier transform infrared (FT-IR) spectroscopic imaging is a highly versatile technique that can be applied to a wide range of systems. This article summarizes some of the recent efforts developing applications of FT-IR imaging for microfluidics. The main advantage of FT-IR imaging compared to traditional imaging methods is that it is a label-free imaging technique.

Sergei G. Kazarian

Articles

Macro ATR-FT-IR Spectroscopic Imaging of Dynamic Processes

Label-Free Chemical Detection in Micro-Fabricated Devices Using FT-IR Spectroscopic Imaging

Label-Free Chemical Detection in Micro-Fabricated Devices Using FT-IR Spectroscopic Imaging