Langmuir 2006, 22, 4125-4130
4125
Characterization of Metal-Supported Poly(methyl methacrylate) Microstructures by FTIR Imaging Spectroscopy Gerald Steiner, Cordelia Zimmerer, and Reiner Salzer* Institute for Analytical Chemistry, Dresden UniVersity of Technology, 01062 Dresden, Germany ReceiVed NoVember 29, 2005. In Final Form: February 20, 2006 Thin microstructured poly(methyl methacrylate) (PMMA) films may be used as scaffolds for biosensor arrays. Microstructured pores form miniaturized vessels, each constituting an individual reaction vessel or detector element. Arrays of micropores with diameters between 2 and 80 µm were prepared in thin PMMA films on gold by optical lithography. Laterally resolved chemical information for microstructured PMMA films on a gold substrate was obtained by FTIR spectroscopic imaging. The carbonyl band was used to characterize the microstructure. Spectroscopic results indicate small amounts of PMMA residues inside the pores. A downshift of 5 cm-1 compared to the position of the PMMA bulk carbonyl band indicates interactions of the PMMA residue with the gold substrate. Additional small bands are observed which indicate the formation of carboxylate during PMMA microstructuring. Three possible types of strong PMMA-gold interactions are discussed. All strong PMMA-gold interactions involve carbonyl or carboxyl oxygen.
I. Introduction Microstructures generated by optical lithography have been studied intensively during the past decade.1 They find application in biosensors,2 in optical and biomedical devices, as well as in microengineering tools. Microstructured polymer films on optical or metal surfaces facilitate the formation of array layouts of sensor chips.3,4 Poly(methyl methacrylate) (PMMA) is commonly used as high-resolution resist for electron beam lithography as well as for X-ray and UV microlithography. PMMA forms homogeneous films and is known for its good optical homogeneity,5 negligible swelling in water,6 and its good adhesion to oxides and metals.7,8 Due to its biocompatibility, PMMA is widely used as a material for medical implants.9,10 Various microscopic methods provide lateral resolution in the low nanometer range, like atomic force microscopy (AFM),6 scanning electron microscopy (SEM),11 and scanning tunneling microscopy (STM).12 The investigation of much larger sample areas by those methods, as it is necessary for structures of single or double digit micrometer size, may turn out to be beyond practical limits (in terms of time, costs, data volume, etc.). Such areas with micrometer patterns may preferably be investigated by spectroscopic ellipsometry,6 optical microscopy,13 surface * To whom correspondence should be addressed. (1) Sun, X.; Zhuang, L.; Zhang, W.; Chou, S. Y. J. Vac. Sci. Technol. B 1998, 16, 3922-3925. (2) Lee, K.-N.; Shin, D.-S.; Lee, Y.-S.; Kim, Y.-K. J. Micromech. Microeng. 2002, 13, 18-25. (3) Lange, S. A.; Benes, V.; Kern, D. P.; Horber, J. K. H.; Bernhard, A. Anal. Chem. 2004, 76, 1641-1647. (4) Alexander, T. A.; Wickenden, A. E. Proc. SPIE Int. Soc. Opt. Eng. 2004, 5588, 78-86. (5) Tabata, Y.; Mita, I.; Nonogaki, S.; Horie, K.; Tagawa, S. Polymers for Microelectronic; VCH: Weinheim, Germany, 1998. (6) Gilchrist, V. A.; Lu, J. R.; Keddie, J. L. Langmuir 2000, 16, 740-748. (7) De Carlo, F.; Song, J. J.; Mancini, D. C. J. Vac. Sci. Technol. B. 1998, 16, 3539-3542. (8) Gilio, E. D.; Cometa, S.; Sabbatini, L.; Zambonin, P. G.; Spoto, G. Anal. Bioanl. Chem. 2005, 381, 626-633. (9) Jaeger, M.; Wilke, A. J. Biomater. Sci., Polym. Ed. 2003, 14, 1283-1298. (10) Fini, M.; Giavaresi, G.; Aldini, N. N.; Torricelli, P.; Botter, R.; Beruto, D.; Giardino, R. Biomaterials 2002, 23, 4523-4531. (11) Braun, H.-G.; Meyer, E. Thin Solid Films 1999, 345, 222-228. (12) Grevin, B.; Rannou, P.; Payerne, R.; Pron, A.; Travers, J. P. J. Chem. Phys. 2003, 118, 7097-7102. (13) Lee, S.; Lee, K.; Lee, S. J. Micromech. Microeng. 2002, 12, 334-340.
plasmon resonance (SPR),14,15 or fluorescence microscopy.16 However, all mentioned methods do not provide the chemical information needed to assess the molecular structure within microstructured areas or to reveal interactions between molecules. Exactly this kind of information is needed if the microstructuring process does not provide the expected result and needs to be improved. Remaining polymer material inside the microstructure cavities is one of the problems which are encountered. FTIR spectroscopic imaging provides lateral resolution in the low micrometer range, it permits observation of sufficiently large sample areas at once, and, more importantly, it provides detailed information concerning both chemical structure and intermolecular interactions within the microstructured area. FTIR spectroscopic imaging provides access to the chemical mechanism of the microstructuring process; hence, it permits systematic quality improvement. Due to the typical layout of current commercial instruments, FTIR spectroscopic images usually cover an area of 270 × 270 µm2 and typically resolve 64 × 64 pixels within this area. Every pixel in an FTIR spectroscopic image holds a complete spectrum. A FTIR spectroscopic image can be collected within a matter of seconds up to a few minutes, depending upon type of sample and spectral resolution chosen. The spatial resolution of FTIR spectroscopic images is diffraction limited, given that sufficiently much IR radiation reaches the detector. FTIR spectroscopic imaging is based on the development of focal plane array (FPA) detectors.17 FPAs soon found application in various scientific areas.18 Our PMMA microstructures are intended to form a polymer scaffold for a biosensor with SPR detection. SPR detection requires a prism in order to couple light into a gold layer on top of the prism (Figure 1). (14) Steiner, G.; Sablinskas, V.; Hu¨bner, A.; Kuhne, Ch.; Salzer, R. J. Mol. Struct. 1999, 509, 265-273. (15) Weeber, J. C.; Finot, E.; Legay, G.; Cathelat, A.; Lacroute, Y.; Dereux, A. Langmuir 2004, 20, 10179-10185. (16) Dibbern-Brunelli, D.; Atvars, T. D. Z. J. Appl. Polym. Sci. 1995, 58, 779-786. (17) Lewis, N. E.; Kidder, L. H.; Arens, J. F.; Peck, M. C.; Levin, I. W. Appl. Spectrosc. 1997, 51, 563-567. (18) Salzer, R.; Steiner, G.; Mantsch, H. H.; Mansfield, J.; Lewis, N. E. Fresenius J. Anal. Chem. 2000, 366, 712-726.
10.1021/la053221x CCC: $33.50 © 2006 American Chemical Society Published on Web 03/24/2006
4126 Langmuir, Vol. 22, No. 9, 2006
Figure 1. Schematic of a biosensor array using a microstructured polymer film. The microstructured polymer film is attached to the SPR transducer surface, a gold layer on top of a glass prism. The gold layer is 50 nm thick, the polymer film 350 nm thick.
The evanescent wave excited by the surface plasmons extends into the thin PMMA film adsorbed on the gold layer. The PMMA microstructure consists of pores, which form miniaturized reservoirs. Every pore constitutes an independent detector element within the array. SPR imaging detects the slightest concentration changes inside every individual pore. The strength of the evanescent field decays exponentially from the gold surface; that is, the highest detection sensitivity is provided directly above the gold surface. This requires pores perfectly carved out. Left over PMMA at the bottom of a pore prevents the analyte from reaching the area close to the gold layer, where the maximum strength of the evanescent field is located. Here we report on the identification of left-over PMMA at the pore bottom. If left-over PMMA is found at the pore bottom, its chemical properties need to be unraveled in order to develop chemical approaches for removing it and to ensure the quality of the micropatterned PMMA films. II. Materials and Methods Preparation of Microstructured PMMA Films. PMMA microstructures were generated on gold-covered prisms of highly refractive SF6 glass and on gold-covered silicon wafers. The neat prisms have a surface roughness of 1 λ, equivalent to 520 nm maximum deviation from an ideal smooth surface. The S/D coefficient (representing stretch and frequency of dashes) was 80/50. All substrates were rinsed with ethanol, treated in a plasma cleaner, and subsequently heated in a vacuum chamber to a temperature of approximately 200 °C. Afterward, a 50 nm gold layer was vapor deposited on the substrates at a pressure of 8.0 × 10-6 mbar. During deposition, the temperature was kept at 40 °C. Thin polymer films were formed by spin coating of standard PMMA (Allresist GmbH, Berlin, Germany) with a molecular weight of 950 000 dissolved in chlorobenzene. All PMMA layers were baked for approximately 1 h at 180 °C, which is approximately 80 °C above the PMMA glass transition temperature.19 The thickness of PMMA layers was measured by spectroscopic reflection interferometry (FTP advanced, Sentech GmbH, Berlin, Germany) using an optical microscope (Carl Zeiss Jena, Jena, Germany). The layer thickness was calculated by fitting the measured reflection curve using the FTP advanced software. An optimal fit could be achieved by using the refractive index for PMMA n ) 1.49 and a function for the complex refractive index of gold taken from ref 20. The calculated thickness of the PMMA film was 350 ( 10 nm. PMMA microstructures were generated by optical UV lithography. The light of a deuterium halogen lamp (DH-2000, Top Sensor Systems, Erbeek, The Netherlands) with a continuous spectrum from 215 to 1500 nm was coupled into a 400 µm multimode optical fiber with high transmission in the UV range (Optocon GmbH, Dresden, (19) Cholod, M. S.; Parker, H. Y. In Polymeric materials encyclopedia; Salamone, J. C., Ed.; Wiley: New York, 1987. (20) Palik, E. D. Handbook of optical constants of solids; Academic Press Inc.: New York, 1985.
Steiner et al.
Figure 2. SEM image (A) and SPR image (B) of a microstructured PMMA film on gold substrate. Germany). The light from the fiber passed a thin copper mesh of hexagons of 80 µm width (Plano GmbH, Wetzlar, Germany). The copper mesh was placed directly on the PMMA film. The distance between the end of the fiber and the mask was approximately 1 mm, resulting in an illuminated sample area of approximately 500 µm in diameter. After an irradiation time of 60 min, the polymer coated substrates were submersed into developer solution consisting of one part of methylisobutyl ketone and one part of isopropyl alcohol. The development process was performed for 3 min at a temperature of 20 °C. Afterward, the sample was dried under a mild flow of nitrogen. FTIR spectroscopic measurements were carried out to in order to verify that no developer remained inside the micropores. FTIR Spectroscopic Imaging. FTIR spectroscopic images were collected using a Bruker FTIR imaging spectrometer Hyperion (Bruker Optik GmbH, Ettlingen, Germany) coupled to an IR microscope (IRscope, Bruker Optik GmbH, Ettlingen, Germany). The imaging detector was a Santa Barbara focal plane MCT 64 × 64 array detector. The 15-fold Cassegrainian objective with a numerical aperture of 0.4 (resulting in a maximum angle of incidence of 23°) imaged a sample area of approximately 270 × 270 µm2. A low pass filter was inserted into the beam to eliminate radiation of unwanted wavelengths and prevent Fourier fold-over perturbations. Camera gain and offset were optimized automatically. Twenty-one interferograms were co-added for each of the 4096 image pixels at a resolution of 2 cm-1. Spectra were acquired in the reflection mode in the spectral range of 950-3600 cm-1 applying Happ-Genzel apodization and zero filling factor of 1. The frame rate of the camera was 252 Hz, yielding a total measurement time of 3 min per polymer sample. Data preprocessing, image processing and information extraction were carried out using the MatLab Package (Version 5.6, MathWorks Inc. Natric, MA). Spectra of the sample image were ratioed against a neat gold surface and transferred to absorbance values.
III. Results and Discussion Initially, the microstructures were probed by SEM, AFM, and SPR (Figures 2 and 3). The SEM and AFM images indicate a perfect microstructure, whereas the SPR image points to severe failure. The SEM image in Figure 2A shows a dimension of the microstructure as required for a biosensor array. However, the aspect ratio is too large to obtain high-resolved AFM images. To probe the bottom of the pores, a microstructure with pores of 20 µm in diameter was used (Figure 3A). The apparent discrepancy between SEM and AFM images on one hand and the SPR image on the other hand points to a crucial difference between the corresponding techniques: SEM and AFM probe the PMMA microstructure from the top, whereas SPR probes from below. The SPR image could be indicative of pores incompletely carved out. The AFM images in Figure 3 clearly show the topology of the pore bottom, but they do not indicate whether the material at the pore bottom is gold or remaining PMMA. AFM reference investigations were made using goldcovered silicon wafers, whose surface roughness does not exceed
Metal-Supported PMMA Microstructures
Langmuir, Vol. 22, No. 9, 2006 4127
Figure 3. (A) AFM image of a sample with the gold film deposited on a silicon wafer. (B) AFM image of a pore region.
Figure 4. FTIR reflection absorption spectrum of a PMMA film of 350 nm thickness on gold.
5 nm; that is, they are much smoother than glass. Such wafers were coated by PMMA and subsequently microstructured. The AFM image in Figure 3B shows an area of 1 × 1 µm2 taken from a pore center. The measured maximum roughness is 22.8 nm, much higher than the initial roughness of max. 5 nm of the reference wafers. Under the given experimental conditions, this enhanced roughness is a first hint to PMMA residues inside the pore. Plasma etching is a common technique to remove substance from surfaces. Upon treating PMMA with O2 plasma, polar functional groups containing oxygen are introduced into the polymer network.21,22 Several polymer chains break and small molecules such as CO2, H2O, and CO evolve.23 This results in the removal of polymer material.24 With the aim of removing the residue inside the pores, the PMMA coated prisms were placed in a plasma chamber (plasma cleaner/sterilizer PDC-32G, Harrick Scientific Corp., Ossining, NY). A plasma treatment of 20 min was performed followed by chemically etching using the developer solution described above. Polymer material was etched away from the upper surface, as microscopic images revealed, but residue inside the pores was not removed. To find ways of removing PMMA residue out of micropores, we investigated the chemical nature of the residue inside the pores as well as its interactions to the gold layer beneath. PMMA has a number of dominant absorption bands in the IR fingerprint region between 1000 and 1800 cm-1. Figure 4 shows a reflection absorption spectrum taken from a 350 nm thick homogeneous PMMA film on a gold substrate. The detailed assignment of the vibrations is given in Table 1. Bands between 1050 and 1300 cm-1 arise mainly from stretching modes of the -COOCH3 group, partly in combination with the deformation mode of the C-H group.25 The bands between 1350 and 1500 cm-1 are assigned to the δ(C-H) modes.26,27 The band at 1730 cm-1 originates from the ν(CdO) mode of the -COOCH3 group.26 Because of its strength and the excellent signal-to-noise
Table 1. Positions and Assignment of Prominent Absorption Band of PMMA in the Spectral Range from 1000 to 1800 cm-1
(21) Groening, P.; Kuettel, O. M.; Collaud-Coen, M.; Dietler, G.; Schlapbach, L. Appl. Surf. Sci. 1995, 89, 83-91. (22) Hook, T.; Gardella, J., Jr. J. Mater. Res. 1987, 2, 117-131. (23) Kupfer, H.; Ostwald, R. Met. Plast.: Fundam. Appl. Aspects 1998, 5&6, 85-96. (24) Chai, J.; Lu, F.; Li, B.; Kwok, D. Y. Langmuir 2004, 20, 10919-10927. (25) Sondag, A. H. M.; Raas, M. C. Polymer 1991, 32, 2917-2920. (26) Patnaik, A.; Li, C. J. Appl. Phys. 1998, 83, 3049-3056. (27) Liu, Y.; Wu, W.; Guan, Y.; Ying, P.: Li, C. Langmuir 2002, 18, 62296232.
wavenumber (cm-1)
assignment25,39,40
1730 1465 1448 1440 1436 1396 1275 1185 1150
ν(CdO) δa(CH3-O) δ(CH2)/ν(CC) δa(CH3)/δ(CH2) δ(CH3-O) δ(CH3) ν(C-C-O) coupled with ν(C-O) C-O-C C-O-C
ratio, the ν(CdO) band permits detection of PMMA down to a thickness of a few nanometers. Because the spatial resolution of infrared imaging is ∼10 µm, a larger dimension of the pattern was chosen in order to image the web as well as the pores in details. Figure 5A shows the microscopic image of a PMMA microstructure in the visible region. PMMA is highly transparent throughout this region, hence this microscopic image was taken by a phase contrast technique. The light is focused to obtain the strongest contrast at the steep pore walls. The thickness difference between web and pore regions can hardly be deduced from this phase contrast image. Similarly, possible thickness changes in transition areas, e.g., between pore bottom and pore wall, are not indicated in the phase contrast image. A corresponding FTIR image of the high-lighted area of Figure 5A was collected (Figure 5B). This FTIR image was calculated from the integrated intensity of the ν(CdO) band in the spectral range from 1660 to 1760 cm-1 and is color-coded on a black-green scale. Black areas indicate smallest intensities (least absorption) at the pore bottom; green areas indicate strongest absorption in the web regions of the microstructured polymer. The hexagon-microstructure is clearly discernible. A transition zone between black and green areas is now clearly seen. It extends across several pixels. The size of a pixel is 4.2 × 4.2 µm2, calculated from the size of the imaged area of 270 × 270 µm2 and the number of 64 × 64 pixels per image. The pixel size in the image must not be confused with the lateral resolution in the image, controlled by the optical parameters chosen for the measurement. Under the experimental conditions mentioned above the diffraction limit amounts to 8 µm, roughly twice the pixel size in the image. The inferior optical resolution is the main reason the width of the webs in Figure 5B appears to be smaller than 30 µm (Figure 5A). Ordinate values of pixels closer to the
4128 Langmuir, Vol. 22, No. 9, 2006
Steiner et al.
Figure 5. (A) Optical micrograph of a microstructured PMMA film. (B) Black-green scaled FTIR image of the same sample in the spectral region 1660-1760 cm-1. (C) Color-scaled FTIR image based on the integrated intensities of the carbonyl stretching band (green channel, intensity normalized, spectral region as in (B)) and on the integrated intensities of the C-H deformation band (red channel, intensity normalized, spectral region 1310-1510 cm-1. (D) Cross section calculated from the area between the dashed lines in (C). The dotted lines indicate the ideal topology of a perfect microstructure.
rim get increasingly influenced by those of the adjacent lowabsorbing pore region. Nonetheless, the trough-shaped crosssection of the pore extends far beyond the diffraction limit; it cannot solely arise from the aberration or resolution power of the IR microscope. A transition area must have been formed between the pore center and the pore wall. Since the transition zone was formed during the development process, the question about its chemical composition arises. Is this zone depleted in particularly reactive groups? A first test about the chemical composition of the transition zone can be performed by comparing the content of C-H groups and CdO groups across the transition zone and across the web region. We evaluated the δ(C-H) band intensities in the spectral range from 1310 to 1510 cm-1. The results are revealed in a composite color image (Figure 5C). Normalized CdO intensities are again printed in green, normalized C-H intensities in red. In the RGB color coding system, equal saturation of red and green produces yellow. Figure 5C does not show variations in its yellow color, except for the slight noise. This reveals chemical homogeneity between the transition zone and the web region. The topological profile of the transition zone is much more discernible in a cross-section. Dashed white lines in Figure 5C indicate the sample area used for calculating the cross section depicted in Figure 5D. It has to be noted that the profile in Figure 5D is based on chemical properties (ν(CdO) band intensity) rather than on common elevation measurements. These chemical properties reveal two important findings: (i) the band intensity inside the pores never reaches the zero level; (ii) the walls of the pores clearly deviate from the expected vertical shape (dotted line in Figure 5D). The transition zone between pore bottom and pore wall extends at least across 7 pixels (approximately 30 µm). For a more detailed investigation of the chemical properties, a cluster analysis of the FTIR spectra was performed. The MatLab
Figure 6. Cluster analysis of the FTIR spectra in the ν(CdO) region. (A) color-coded image of the cluster assignment and (B) centroid spectra of the clusters.
algorithm was used (cf. experimental) to distribute all 64 × 64 spectra of the image into 10 clusters (Figure 6A). The algorithm chose the size of the carbonyl band as the dominating feature for clustering. All spectra within one cluster are very similar. The carbonyl bands show absorbance maxima 0.5 along the web ridge. Average spectra (centroids) were obtained for every cluster. These centroids are given in Figure 6B. Positions of the maxima of the carbonyl band across the microstructured PMMA provide a first clue to the distribution of chemical properties across the sample. The three experimental spectra, whose ν(CdO) bands are shown in Figure 7A, were taken from the indicated spots of Figure 7B. The measured downshift of the band maximum at the particular pixel is colorcoded. In web regions the typical band maxima are located at 1724 cm-1. Maximum downshift is found inside the pore regions, indicating bound carbonyl groups at the pore bottom. Adhesion prevents the formation of perfectly carved-out pores. Upon interaction, the ν(CdO) band of PMMA shifts toward lower frequencies.28-30 Reported downshifts for the bound carbonyl range between 15 and 35 cm-1.31-33 The band envelope becomes asymmetric in shape due to underlying contributions by free and bound carbonyl. Carbonyl bands in pore regions exhibit a much more distinct asymmetry than spectra in web regions (Figure 7A). Moreover, carbonyl bands in pore regions show well discernible shoulders at their low-frequency edge. (28) Berquier, J.-M.; Fernandes, A.-C.; Chartier, P.; Arribart, H. Proc. SPIE 1989, 1145, 245-250. (29) Fontana, B. J.; Thomas, J. R. J. Phys. Chem. 1961, 65, 480-487. (30) Enriquez, E. P.; Schneider, H. M.; Granick, S. J. Polym. Sci. B 1995, 33, 2429-2437. (31) Berquier, J.-M.; Arribart, H. Langmuir 1998, 14, 3716-3719. (32) Johnson, H. E.; Garnick, S. Macromolecules 1990, 23, 3367-3374. (33) Frantz, P.; Granick, S. Macromolecules 1995, 28, 6915-6925.
Langmuir, Vol. 22, No. 9, 2006 4129
Figure 8. Separation of the ν(CdO) band for pixels in the pore center (A) and along the web ridge (B). The spectrum in (A) is the average of all spectra in the black cluster of Figure 6. The spectrum in (B) is the average of all spectra in the brown and red clusters of Figure 6. Solid lines: experimental spectra; dotted lines: calculated envelopes; dashed lines: calculated component bands.
We decomposed the main carbonyl band around 1720-1730 cm-1 in order to identify and to quantify the dominating species. All spectra of the black cluster of Figure 6A (pore center) and of the brown and red clusters (web ridge) were pooled. The best fit for the decomposition of the pooled set of the main carbonyl band was obtained with two Lorentzian bands centered at 1718 ( 4 and 1732 ( 4 cm-1, respectively. The curve fitting was performed by a nonlinear least-squares fitting routine using the Grams package (Galactic Industries Corporation, Salem, NH). Bandwidths were permitted to float. Bound carbonyl (1718 ( 4 cm-1) is the dominating species in pore regions (Figure 8A), whereas the web region is clearly dominated by free carbonyl (1732 ( 4 cm-1, Figure 8B). Berquier and co-workers reported two bands at 1716 (bound carbonyl) and 1693 cm-1 (carboxylate) for PMMA adsorbed on silicon with surface bound hydroxyl groups serving as electron donors. Carboxylate ions in PMMA might have been generated by cleavage of the ester bond through hydrolysis.34,35 A similar behavior occurs when PMMA adsorbs on chromium38 or aluminum.36 Other authors assigned a band at 1680 cm-1 to the -COO- group and a band at 1698 cm-1 to the -COOH group.37,38 (34) Papier, E.; Perrin, J.-M.; Nanse, G.; Fioux, P. Eur. Polym. J. 1994, 30, 985-991. (35) Kobstadinidis, K.; Thakkar, B.; Chakraborty, A.; Potts, L. W.; Tannenbaum, R.; Tirrell, M.; Evans, J. F. Langmuir 1992, 8, 1307-1317. (36) Mallik, R. R.; Pritchard, R. G.; Horley, C. C.; Comyn, J. Polymer 1985, 26, 551-556. (37) Allara, D. L. Polym. Sci. Technol. 1980, 12B, 751-756. (38) Tannenbaum, R.; Hakanson, C.; Zeno, A.; Tirerell, M. Langmuir 2002, 18, 5592-5599. (39) Nagai, H. J. Appl. Polym. Sci. 1963, 7, 1697-1714. (40) Grohen, Y.; Auger, M.; Prud’Hamme, E.; Schultz, J. J. Polym. Sci. B 1999, 37, 2985-2995.
4130 Langmuir, Vol. 22, No. 9, 2006
Steiner et al.
Figure 9. Normalized FTIR spectra of the web ridge (solid line) taken from position #1 in Figure 7B and of the pore center (dotted line) taken from position #3 in Figure 7B.
The spectrum for PMMA on gold in Figure 8A exhibits two distinct small bands at 1680 and 1698 cm-1. The CdO group has a vacant π orbital in addition to lone pairs. These vacant orbitals accept electron density from the filled orbitals of gold to form a type of π-back-bonding that supplements the σ bonding arising from lone pair donation.26 High electron density on gold is thus delocalized onto the low-lying orbitals of CO. As the extent of the back-donation from gold to CO increases, the Au-O bond becomes stronger and the >CdO bond weaker. The latter results in a band shift toward lower wavenumbers. Two different adsorption schemes are proposed for carbonyl groups on gold.31 They involve only one electron pair of the carbonyl group (1:1 site) or a couple of oxygen lone-pair electrons (2:1 site). The band at 1716 cm-1 may be assigned to the 1:1 site mode and the band at 1698 cm-1 to the 2:1 site mode.26 The remaining band at 1680 cm-1 may be assigned to the carboxylate group. An important question concerns the ratio of the band intensities of the two components of the main carbonyl bands in Figure 8. Is it a direct expression of the concentration ratio? It has to be taken into account that the spectra are measured in a reflection geometry. Figure 9 shows the centroid spectra of pore and web regions. For comparison, the spectra are normalized to the maximum of the ν(C-O-C) vibration at 1150 cm-1. This band was chosen because it originates from a side chain group of PMMA and involves the CdO group in its vibration. If ester cleavage occurs during the lithographical development process, the ν(C-O-C) would similarly be influenced as the ν(CdO) vibration. Other bands such as the C-H stretching modes are not directly related to changes in side chains. The normalized intensities of the ν(CdO) bands in Figure 8 differ remarkably between pore and web regions. The normalized ν(CdO) band is distinctly smaller in the pore region than that in the web region. This might lead to the assumption that carbonyl groups are removed during the development process while the other parts of the side chain remain in the polymer network. It was already demonstrated in Figure 5C that this is not the case. Moreover, the ν(C-O-C) vibrations at 1150 and 1190 cm-1 involve the carbonyl group; hence, the low intensity of the ν(CdO) band in pore regions cannot be explained by a loss in carbonyl groups during the development process The weakness of the carbonyl stretching mode in pore regions does not result from alterations in chemical composition, it is due to the reflection measurements with nearly perpendicular light incidence as illustrated in Figure 10. The carbonyl groups of bound PMMA show a predominantly perpendicular orientation toward the gold surface. The electrical vector of the incident wave and the electrical vector of the carbonyl
Figure 10. Scheme of the orientation of bound PMMA on a gold layer, the orientation of the transition dipoles of its ν(CdO) and ν(C-O-C) vibrations, and the E vector of light incident at 23°. The tip angle arise from the numerical aperture of the Cassegrainian objective of 0.4.
stretching vibration are aligned orthogonally toward each other, which results in a weaker excitation of the stretching mode of the bound carbonyl groups compared to the less uniformly oriented carbonyl groups in bulk PMMA. The C-O-C oscillators of the side chains are located roughly parallel to the gold surface. An ideally smooth gold surface would completely shield the electrical field of the incident light due to dipoles induced within the gold surface. The roughness at microscopic level of the real gold surface (cf. experimental) prevents this shielding, and the C-O-C modes lead to absorption.
Conclusions Arrays of micropores with diameters between 2 and 80 µm were prepared in thin PMMA films adsorbed on gold by optical lithography. Although SEM and AFM images show perfect topology of the microstructures SPR imaging reveals that pores are incompletely carved out. Laterally resolved chemical information of microstructured PMMA films on gold substrate was obtained by FTIR spectroscopic imaging. The carbonyl band was used to characterize the microstructure. Spectroscopic results exhibits small amounts of PMMA residues inside the pores. A downshift of 5 cm-1 compared to the position of the PMMA bulk carbonyl band indicates interactions of the PMMA residue with the gold substrate. Additional small bands are observed which indicate the formation of carboxylate during PMMA microstructuring. Three possible types of strong PMMA-gold interactions are derived from the spectra and discussed. This strong bound residues inside the pores could not be removed by plasma etching. FTIR spectroscopic imaging was the only method, which revealed any PMMA residue inside the pores and which provided clues for the optimization of the microstructuring process in thin organic polymer films on metal substrates. Acknowledgment. We are indebted to the Deutsche Forschungsgemeinschaft (DFG) for the financial support within the “Sonderforschungsbereich Reaktive Polymere” (SFB 287). The authors gratefully acknowledge Dr. H.-G. Braun from the Institute for Polymer Research Dresden and Dr. M. T. Pham from the Research Center Rossendorf (Germany) for assistance with the SEM and AFM measurements. LA053221X
