3310
Langmuir 2000, 16, 3310-3316
Determination of the Orientation of 6-Mercaptopurine Adsorbed on a Silver Electrode by Surface-Enhanced Raman Spectroscopy and Normal Mode Calculations Alberto Vivoni,‡ Shi-Ping Chen,† David Ejeh,† and Charles M. Hosten*,† Department of Chemistry, Howard University, Washington, DC, 20059, and Department of Mathematics and Physical Sciences, Inter American University, San German Campus, San German, Puerto Rico 00683 Received October 5, 1999. In Final Form: December 22, 1999 6-Mercaptopurine, 6MP, is a sulfur analogue of adenine which is commonly used as an antitumor agent. The coordination chemistry of 6MP with metals is of interest because the drug contains a number of electronically versatile binding sites. In this study the interaction between 6MP and a silver electrode surface was investigated using surface-enhanced Raman spectroscopy (SERS). High-quality solution Raman spectra of 6MP along with surface-enhanced Raman scattering spectra of 6MP adsorbed on a roughened silver electrode surface are presented. Urey-Bradley force field and semiempirical calculations with the PM3 method were used to generate vibrational frequencies for 6MP based on a specific set of possible molecular interaction sites. An error analysis of the calculated frequencies and the experimentally determined SERS frequencies allowed for the identification of the experimental data which most closely matched the calculated SERS data. It was concluded that 6MP attaches head-on through the N1 atom when the molecule is adsorbed onto a silver electrode surface.
1. Introduction 6-Mercaptopurine, 6MP (Figure 1), a sulfur analogue of naturally occurring purine and adenine, was first synthesized1 with the idea that chemotherapy with abnormal nucleic acid bases might effectively inhibit tumor growth. The initial clinical trials of 6MP in the treatment of leukemia occurred in the early 1950s2 and the use of the drug has since been expanded to include the treatment of childhood and adult leukemias. It is also used as an immunosuppressant and antiinflamminant.3 6MP has been shown to behave as a competitive inhibitor of the enzymes hypoxanthine-guanine phosphoribosyltransferase and xanthine oxidase4 and it has been suggested that incorporation of thiopurine nucleotide into cellular nucleic acids is the mechanism for the drug’s cytotoxicity5 while the inhibition of de novo purine ribonucleoside synthesis and interconversion is the mechanism for growth inhibition.6 Metal complexes of 6MP such as Pt and palladium have also exhibited antitumor behavior, in some cases, at a level more enhanced than that of the free ligand.7 A number of metal complexes of 6MP have been synthesized and their crystal structure determined by X-ray crystallography.8,9 6MP and adenine are effective in the direct rapid electron transfer of cytochrome c at gold electrode surfaces * To whom correspondence should be addressed. Telephone: (202)806-6829. E-mail:
[email protected]. † Howard University. ‡ Inter American University. (1) Montgomery, J. A. Prog. Med. Chem. 1970, 7, 69. (2) Law, L. W. Proc. Soc. Exp. Biol. Med. 1959, 84, 409. (3) Burchenal, J. H.; Murphy, M. L.; Ellison, R. R.; Sykes, M. P.; Tan, T. C.; Leon, L. A. (4) Jewers, K. In Progress in Drug Research; Jucker, E., Ed.; Birkhauser Basel: Switzerland, 1981; Vol. 25. (5) Hitchings, G. H.; Elion, G. B. Proc. Am. Assoc. Cancer Res. 1959, 3, 27. (6) DeMiranda, P.; Beacham, L. M.; Creagh, T. H.; Elion, G. B. J. Pharmacol. Exp. Ther. 1973, 187, 588. (7) Chalmers, A. H.; Burdorf, T.; Murray, A. W. Biochem. Pharmacol. 1972, 21, 2662.
Figure 1. Structure of 6MP.
modified by dipping the gold electrode into a saturated solution of 6MP which adsorbed irreversibly onto the electrode surface.10 Other purine derivatives, guanine and hypoxanthine, did not serve as effective promoters of cytochrome c, suggesting that the lone pair of electrons at the N atom of position 1 is important for promoter action. The orientation of adsorbed 6MP is therefore critical in its action as a mediator in electron transfer. The coordination chemistry of 6MP is of interest because of the electronically and stereochemically versatile binding sites on the molecule. As a purine it offers possible binding sites to metals at N1, N3, N7, and N9 while the presence of the sulfur atom attached to C6 of the purine ring introduces a new range of chemical, structural, and spectroscopic features to the molecule, setting it apart from its parent purine. A range of optical methods including IR reflectance,11 ellipsometry,12 and SGH measurements13 have been employed in determining the orientation of adsorbed (8) Harris, T. E.; Baga, R. C. In Antineoplastic and Immunosuppressive Agents; Sartolli, A. C., Johnason, D. G., Eds.; Springer-Verlag: Berlin, 1974. (9) Dubler, E.; Gyr, E. Inorg. Chem. 1988, 27, 1466. (10) Atkinson, M. R.; Murrary, A. W. Biochem. J. 1965, 94, 64. (11) Porter, M. D. Anal. Chem. 1988, 60, 1143A. (12) Collins, R. W.; Kim, Y. Anal. Chem. 1990, 62, 887A. (13) Shannon, V. L.; Koos, D. A.; Kellar, S. A.; Huifang, P.; Richmond G. L. J. Phys. Chem. 1989, 93, 6434.
10.1021/la9913194 CCC: $19.00 © 2000 American Chemical Society Published on Web 02/17/2000
6-Mercaptopurine Adsorbed on a Silver Electrode
species. Surface-enhanced Raman scattering (SERS)14 spectroscopy provides a versatile and elucidating approach to studying the interaction and conformational behavior of biomolecules and allows for the in situ characterization of the chemical identity, structure, and orientation of surface species in the adsorbed state.15 The high enhancement factor of the Raman scattering intensity by the adsorbate, up to 6 orders of magnitude compared to the same modes for the species in solution, results in high-resolution vibrational spectra of molecules from very dilute solutions, as low as 10-8 M. Along with this high sensitivity, SERS also offers the molecular specificity inherent in vibrational spectroscopy. These factors along with water being a weak Raman scatterer result in SERS being the most appropriate technique to probe the adsorption of 6MP on an anodized silver electrode. The surface selection rules developed for SERS16,17 aid greatly in determining the orientation of planar ring systems with respect to the metal surface. If a planar ring is standing up on the metal surface with its molecular plane perpendicular to the surface, the intensity of inplane vibrational modes which transform to polarizability component zz should be enhanced because the electric field of the exciting light normal to the surface (z direction) is enhanced by electromagnetic resonances in small metal surface particles. A simple method for determining the orientation of lowsymmetry adsorbates on a metal surface using SERS spectroscopy has been applied to determine the orientation of a series of molecules containing four carbon atoms.18 The orientation of a number of aromatic halogenated thiols and disulfides has been determined on the basis of SERS spectral data.19,20 6MP has also been shown to form selfassembled monolayers on a mercury surface similar to those formed by alkanethiols.21 In this paper surface-enhanced Raman spectroscopy and normal coordinate calculations using a Urey-Bradley force field (UBFF) and semiempirical calculations are used to determine the mode of interaction between 6MP and a silver electrode surface. SERS spectra of 6MP adsorbed on an anodized silver electrode surface along with solution Raman spectra of 6MP are presented and frequency shifts in the SERS band frequencies relative to the solution Raman spectra are reported. Four possible binding geometries are identified and normal mode calculations for each geometry are performed. The calculated and experimentally determined frequency shifts are compared and from this comparison a possible mode of interaction for 6MP adsorbed on an anodized silver electrode is presented. 2. Experimental Section 2.1. Chemicals. 6-Mercaptopurine monohydrate (99% purity) was purchased from Aldrich Chemical Co. and used without further purification. For the acquisition of normal, non-surfaceenhanced Raman spectra, 10-1 M solutions of 6MP were prepared (14) Birke, R. L.; Lombardi, J. R. In Spectroelectrochemistry: Theory and Practice; Gale, R. J. Ed.; Plenum Press: New York, 1988; p 263. (15) Creighton, J. A. In Spectroscopy of Surfaces; Clark, R. J. H., Hester, R. E., Eds.; Wiley and Sons: New York, 1988; Vol. 16, p 37. (16) Suh, J. S.; Moskovits, M. J. Am. Chem. Soc. 1986, 108, 4711. (17) Moskovits, M.; Suh, J. S. J. Phys. Chem. 1984, 88, 5526. (18) Pemberton, J. E.; Bryant, M. A.; Sobocinski, R. L.; Joa, S. L. J. Phys. Chem. 1992, 96, 3776. (19) Szafranski, C. A.; Tanner, W.; Laibinis, P. E.; Garrell R. L. Langmuir 1998, 14, 3570. (20) Szafranski, C. A.; Tanner, W.; Laibinis, P. E.; Garrell R. L. Langmuir 1998, 14, 3580. (21) Sevilla, J. M.; Pineda, T.; Madueno, R.; Roman, A. J.; Blazquez, M. J. Electroanal. Chem. 1998, 442, 107.
Langmuir, Vol. 16, No. 7, 2000 3311 by dissolving the solid in doubly distilled deionized water. The resulting solution was placed in a rotator liquid cell (Spex Industries) and Raman spectra were obtained. 2.2. Raman Spectroscopy. An argon ion laser (Coherent, model I70C-4) tuned to the 488-nm line served as the excitation source. For the SERS experiments, laser power at the electrode surface was approximately 75 mW. The laser output was guided through a laser line interference filter (Oriel, 52630) by a surfacecoated mirror, focused by a 42 × 250 mm2 double-convex lens (DCL), and directed at a 90° incident angle to the silver working electrode by a mirror. The scattered radiation was collected at 90° by a Nikon f 1.2, 50-mm camera lens and collimated by a 50 × 200 mm2 DCL (Melles Griot) on the slit of a single spectrograph (SPEX, model 500M) which was configured with a 1200 grooves/ mm grating (SPEX) blazed at 500 nm. A holographic Notchplus filter (Kaiser Optical Systems) was placed between the collimating lens and the entrance slit of the spectrograph to attenuate the Rayleigh scattered radiation. The detection system was a chargecoupled device (CCD) detector (EG&G Princeton Applied Research, model 1530-AUV) containing a 256 × 1024 CCD chip in air-cooled housing. This arrangement gave a spectral window of approximately 1200 cm-l across the exiting plane of the spectrograph. A Gateway 2000 computer (model 4DX2-66) controlled both the CCD detector and the spectrograph. The solution Raman and SERS spectra were calibrated using the known frequencies of indene and fenchone. 2.3. Electrochemical System for Surface-Enhanced Raman Spectroscopy. For the SERS experiments, dilute aqueous solutions of 6MP containing 0.1 M KCl as the supporting electrolyte were used. The SERS spectroelectrochemical cell consisted of a standard working, reference, and auxiliary threeelectrode system. The working electrode was 1.5 mm in diameter by 100 mm in length of silver rod (99.999%, Aldrich), which was sealed in a Teflon tube with epoxy. The reference electrode was a miniature saturated calomel electrode (SCE) (Accumet, 13620-79), and a 3.00-cm long platinum wire served as the auxiliary electrode. A scanning potentiostat (EG&G Princeton Applied Research, model 362) was used to maintain potentiostatic control of the cell and to electrochemically roughen the silver electrode for maximum SERS intensity. A waveform generator (EG&G Princeton Applied Research, model 175) was connected to the potentiostat and employed to provide a pulse function to the working electrode. The silver electrode was hand-polished with 5-, 0.3-, and 0.05µm alumina slurries on a felt pad for 2 min. Residual polishing materials were removed from the electrode surface by sonication in a water bath for 30 s. The silver electrode was rinsed with doubly distilled deionized water and assembled into the cell. The in situ roughening procedure consisted of stepping the potential of the working electrode from 0.0 to +0.40 V versus SCE and maintaining the positive potential for 2 s. The electrode potential was then returned to 0.0 V and SERS spectra were acquired. To eliminate the contribution from solution species in the SERS spectra, the SERS cell was physically washed with an electrolyte blank at the completion of the ORC. The resulting SERS spectra were from surface species devoid of any solution contributions. 2.4. Normal Mode Calculations. Two calculation methods were used in performing the complete normal-mode analysis of the 6MP Raman spectra. These were a frequency fit of the inplane vibrations using a Urey-Bradley force field (UBFF) and a semiempirical molecular orbital calculation using the PM3 method. The UBFF calculation was done with the set of normal coordinate calculation programs of Schachtschneider adapted for a PC by M. Diem.22 The NOCO.FOR program of this set was further modified by adding a SIMPLEX subroutine to it, which constrains the force constant contributions to the normal modes through the potential energy distribution (PED).23 The PM3 calculations were done with the MOPAC program of the CHEM3D software from CambridgeSoft. Both types of calculations were done on a 166-MHz PC and a 225-MHz Macintosh. The 6MP frequency fit was done by transferring and modifying the force constants from adenine and thiophenol calculations. (22) Barlow, A.; Diem, M. J. Chem. Educ. 1991, 68, 35. (23) Vivoni, A. Ph.D. Dissertation, City University of New York, New York, 1995.
3312
Langmuir, Vol. 16, No. 7, 2000
The adenine experimental frequencies were obtained from Majoube.24 The force constants were obtained by optimizing the force constants transferred from aniline, imidazole, and pyrimidine with the SIMPLEX described above. Many of the force constants of vibrations with the same atom types were equaled to each other. This was done to reduce the number of parameters varied during the optimization, thereby reducing the error which accrues from a large number of possible solutions to the vibrational secular equation. The resulting force field had 29 force constants including the aromatic interaction for the fiveand six-membered ring components of the molecule. These force constants were used to fit 27 fundamental frequencies of adenine and 2 isotopomers, C8-D and N9,10-D3. The possible solutions of the vibrational secular equation was further reduced by constraining the force constant contribution on the PED in a manner that reflected the experimentally known assignments. The optimization of the force constants of the adenine force field was done by fitting the polycrystalline frequencies because more isotopic data were available in this state. The adenine geometrical parameters were obtained from Nowack et al.25 A total of 78 frequencies were fitted. Only 5 of these differed from the observed value by more than 2.0%. The average error for the 3 isotopomers was 0.89%. The thiophenol frequency fit was done by optimizing the force constants transferred from methanethiol and phenol calculations. The frequencies of the neat spectrum were obtained from Carron and Hurley.26 The ring structure of the molecule was approximated to a regular hexagon. Values from a PM3 geometry optimization were used for the bond lengths and angles. The minimum energy was obtained when the S-H bond was coplanar with the molecular plane. The frequency fit with the UBFF had an average error of 0.76%. Of the 23 frequencies fitted, only 1 differed by more than 2.0% from the observed value. The normal modes were also calculated with the PM3 method. The calculations yielded both the in-plane and out-of-plane vibrations. The average error was 3.68%. However, most of this error came on the 1584- and 1576-cm-1 frequencies. These were overestimated by 11.8% and 12.9% respectively. These errors are within the range found by Coolidge et al.27 The rest of the frequencies were reproduced with a 2.77% average error. The geometrical parameters and force constants of adenine were retained for the UBFF calculation of 6MP. The thiol group geometrical parameters and force constants were transferred from thiophenol. The energy minimum of the PM3 geometry optimization was obtained when the S-H was coplanar with the plane of the molecule, pointing toward the N1 position. After adjusting the C-S stretching and the S-H deformation force constants, the average error of the frequency fit was 1.35%. No simplex optimization was performed. The UBFF normal-mode analysis of the SERS spectra was done by modeling the surface by means of a silver atom. With the PM3 method, a zinc atom was used to model the surface because the method is parametrized for this element and not for silver. A positive charge was added to simulate a partial positive charge on the surface.28 From these calculations, the heats of formation, geometrical parameters, and π-bond orders were obtained. Variations in the π-bond orders were used to adjust some force constant parameters on the UBFF calculation.
3. Results and Discussion 3.1. Raman and SERS Spectra of 6MP. Figures 2 and 3 are the solution Raman and SERS spectra of 6MP adsorbed on a silver working electrode. Intense, wellresolved spectra were obtained which showed no significant variation in either band frequency or intensity as a function of time. SERS spectra were also recorded as a (24) Majoube, M. J. Raman Spectrosc. 1985, 16, 98. (25) Nowack, M. J.; Lapinski, L.; Kwiatowski, J. S.; Lesczynski, J. Spectrochim. Acta 1991, 47A, 87. (26) Carron, K.; Hurley, L. G. J. Phys. Chem. 1991, 95, 9979. (27) Coolidge, M. B.; Marlin, J. E.; Stewart, J. P. J. Comput. Chem. 1991, 12, 948. (28) Bagus, P. S.; Hermann, K.; Buaschilcher, C. W. J. Phys. Chem. 1984, 81, 1996.
Vivoni et al.
Figure 2. Solution Raman spectrum of 0.1 M 6MP with 488nm laser excitation.
Figure 3. SERS spectrum of 10-3 M 6MP with 488-nm laser excitation.
function of excitation line power and no photodegradation was observed with the laser power used in this study. A very intense 1285-cm-1 band dominates the SERS spectrum of 6MP while intense bands are observed at 1430 and 865 cm-1. Bands of moderate intensity are observed at 1572, 1532, 1375, 1233, 1202, 1187, 1140, and 991 cm-1. Weaker bands are observed at 1056, 951, 800, 694, 665, and 633 cm-1. Excitation with the 406.7-nm laser line gave spectra (not shown) whose band frequencies and relative intensities were comparable with those obtained with 488-nm excitation. Taniguchi et al.29 reported Raman spectra of 6MP as well as SERS spectra of 6MP adsorbed on a gold electrode surface, with 632.8-nm laser excitation. The solution Raman spectrum contained bands at 880, 1010, 1040, and 1310 cm-1, along with a number of other bands whose frequencies were not listed. The visible absorption spectrum of 6MP shows no absorption maxima in the 632- or 488-nm region. As a result neither Taniguchi’s spectra, recorded with 632.8-nm excitation, or our results, recorded with 488-nm excitation, are resonance-enhanced, and therefore close agreement should be expected between the two sets of data in terms of band frequencies and relative intensities. Band frequencies, between the two sets of data, show agreement to within (10 cm-1. Taniguchi29 observed bands at 880, 1010, 1140, and 1310 cm1 in the solution Raman (29) Taniguchi, I.; Higo, N.; Umekita, K.; Yasukouchi, K. J. Electroanal. Chem. 1986, 206, 341.
6-Mercaptopurine Adsorbed on a Silver Electrode Table 1. Observed Band Frequencies of Solution Spectrum of 6MP along with Band Assignments observed frequencies 1559 1525 1430 1359 1294 1238 1210 1192 1129 1001 950 870 693 617
assignments* ring I str C2-N3 str C2-H def C4-C5 str, N9-H def C8-N9 str, C8-H def C5-N7 str C4-N9 str, C2-H def C2-N3 str, C8-H def N7-C8 str, C8-H def C8-N9 str, N9-H def N1-C2 str, C6-N1 str C6-N1 str, C8-N9 str C6-S10 str, ring I def ring I and II def ring I and II def N3-C4 str, ring I def ring I def ring I def C6-S10 str C6-S10 def
spectrum and these bands can correlate with bands at 870, 1001, 1129, and 1294 cm-1 in our solution Raman spectrum. A moderately intense band at 1525 cm-1, appearing as a shoulder, and a pair of bands at 1210 and 1192 cm-1, which are present in our study, are absent from Taniguchi’s spectrum. A more detailed comparison between this work and Taniguchi’s was inhibited by the absence of frequency assignments in the spectra reported by Taniguchi. Table 1 lists the vibrational assignments of the 6MP bands based on calculations performed by the authors. 3.2. Interaction of Thiols with Metal Surfaces. The effect of adsorption of thiols on the S-H bond has been of interest to researchers. Cleavage of the S-H bond for 1-butanethiol has been confirmed by the disappearance of the 2575-cm-1 υS-H vibration upon adsorption of the molecule onto an electrode surface.30 The υC-S stretch vibrations were also shifted to lower frequencies, from 648 to 632 cm-1 and from 723 to 700 cm-1, indicating interaction between the sulfur atom and the silver surface. Ag-S interaction has been identified, using SERS spectroscopy, for the interaction between phenyl and benzyl mercaptide and a silver surface. In the spectrum of phenyl mercaptan, the substituent-sensitive 1092-cm-1 band is downshifted 22 cm-1 upon adsorption, while the C-S stretch in benzyl mercaptan shifts from 680 to 652 cm-1.31 The adsorption of benzene thiol on silver has also been investigated using SERS spectroscopy. Bands at 2567 and 917 cm-1 in the solution Raman spectrum were assigned to the S-H stretching and bending vibrations, respectively. These vibrations were absent from the SERS spectrum of benzene thiol, indicating dissociative chemisorbtion by rupturing of the S-H bond.32 A number of aromatic thiols have also been adsorbed onto roughened gold electrode surfaces, and SERS spectroscopy has indicated that all have been dissociatively adsorbed as monolayers of their corresponding thiolates.20,21 In the SERS spectrum of thiophenol, Carron and Hurley did not observe the band found at 914 cm-1 in the neat spectrum. This band was assigned to the S-H deformation, and its absence indicates a cleavage of the S-H bond. From the study by Carron and Hurley it was (30) Sobocinski, R. L.; Bryant, M. A.; Pemberton, J. E. J. Am. Chem. Soc. 1990, 112, 6177. (31) Sandoff, C. J.: Herschbach, D. R. J. Phys. Chem. 1982, 86, 3277. (32) Joo, T. H.; Kim, K.; Kim, S. J. Raman Spectrosc. 1987, 18, 1857.
Langmuir, Vol. 16, No. 7, 2000 3313
concluded that the S-H bond cleaves when a number of thiols bind to metal surfaces. Taniguchi29 observed bands at 870, 1000, and 1300 cm-1 in the SERS spectrum of 6MP, which were not observed in the Raman spectrum of solid 6MP. Although Taniguchi did not assign these new bands, they were found to be in good agreement with the bands of the solution spectrum of 6MP under basic conditions. Taniguch29 postulated that adsorption of 6MP resulted in cleavage of the S-H bond and that the molecule was adsorbed onto the electrode surface through the S atom and the nitrogen at position 7 (N7) in a vertical orientation. Further support for N7 interaction came from the presence of bands at 1270 and 1345 cm-1 in the SERS spectrum, which were assigned to the in-plane C-H deformation and the imidazole ring vibration, respectively. 3.3. Orientation. Because of the close similarity in chemical structure between 6MP and adenine, a survey of the SERS and Raman spectra of adenine along with the mode of interaction between adenine and metal surfaces would be instructive in deciphering the interaction between 6MP and an electrode surface. Ab initio molecular orbital calculations of the force constants of adenine have been reported.33,34 Most of the Raman lines that appear in the spectra of adenine and its residues are assignable to one of the molecule’s skeletal vibrations.35 A detailed assignment of the normal frequencies and normal modes of the in-plane vibrations of adenine was given by Tsuboi, Nishimura, and Hirrakewa.36 Two mechanisms have been employed to explain the surface-enhancement phenomena observed in surfaceenhanced Raman spectroscopy, the charge-transfer mechanism, and the electromagnetic mechanism.37 These mechanisms have given rise to surface selection rules for surface-enhanced Raman spectroscopy.38 The application of these surface selection rules has led to conflicting conclusions on orientation based on the same spectral data. Taniguchi29 proposed a vertical orientation of adenine based on the absence of out-of-plane adenine vibrations, which would be prominent if the molecule was orientated parallel to the metal surface. If adenine is orientated vertically on the basis of the selection rules, the C-H vibration is expected to be significantly enhanced and its’ absence in the SERS spectrum of adenine led Moskovits and Suh39 to propose a vertical orientation. Otto et al.40 concluded that when adenine is adsorbed unto a silver electrode, it is not adsorbed parallel to the surface but perpendicular to it; Watanabe et al.41 proposed that adenine was bound to the surface through N7 and orientated perpendicular to the surface when they observed peaks at 1330 cm-1 assigned to the N7-C5 vibration and at 736 cm-1 for the adenine ring breathing mode. Siiman et al.42 concluded, from the large intensity of the 739-cm-1 band, that adsorbed adenine had a (33) Koglin, E.; Sequaris, J. M. J. Mol. Struct. 1980, 60, 421. (34) Koglin, E.; Sequaris, J. M. J. Mol. Struct. 1982, 79, 185. (35) Lafleur, L.; Rice, J.; Thomas, G. J. Biopolymers 1972, 11, 2423. (36) Tsuboi, M.; Nishimura, Y.; Hirakawa, A.; Peticolas, W. L. Resonance Raman Spectroscopy and Normal Modes of Nucleic Acid Bases. In Biological Applications of Raman Spectroscopy; Spiro, T. G., Ed.; Wiley: New York, 1987; Vol. 2, Chapter 3. (37) Cotton, T. M.; Kim, J.-H.; Chumanov, G. D. J. Raman Spectrosc. 1991, 22, 729. (38) Chang, R. K., Furtak, T. E., Eds. Surface Enhanced Raman Scattering; Plenum Press: New York, 1982. (39) Suh, J. S.; Moskovits, M. J. Phys. Chem. 1984, 88, 5526. (40) Otto, C.; de Mul, F. F. M.; Huizinga, A.; Greve, J. J. Phys. Chem. 1988, 92, 1239. (41) Watanabe, T.; Kawanami, O.; Katoh, H.; Honda, K.; Nishimura, Y.; Tsuboi, M. Surf. Sci. 1985, 158, 341. (42) Siiman, O.; Rivellini, R.; Patel, R. Inorg. Chem. 1988, 27, 3940.
3314
Langmuir, Vol. 16, No. 7, 2000
Vivoni et al. Table 2. Observed and Calculated Band Frequencies of the In-Plane Modes of the Neat Spectrum of Thiophenol and Differences between the Neat and SERS Spectra (Frequencies in cm-1)
Figure 4. Structure of thiophenol.
predominantly flat orientation of the planar ring relative to the Ag surface. A number of experimental variables have also been shown to influence the frequency and intensity of SERS bands of adsorbates. Watanabe et al.41 observed correlations between the roughening procedure used to activate the Ag electrode and the SERS spectra, suggesting that the degree of surface roughness is an important variable in SERS studies of molecular orientation. A bulk concentration dependence of the adenine ring breathing mode41 has also been established. The ring-breathing mode undergoes a 11-cm-1 upshift as the adenine concentration is increased. This concentration-dependent shift has been interpreted as reflecting the molecular reorientation taking place on the silver surface with the molecular orientation changing from flat at low concentration to a perpendicular orientation at higher concentrations. These contradictions have led us to explore the use of UBFF and semiempirical calculations to determine molecular orientations based on shifts in band frequencies between solution Raman and SERS band frequencies. In an attempt to validate our methodology we first applied UBFF calculations to predict changes in SERS band frequencies of a system where the interaction between the adsorbate and the surface has been well-characterized. The expectation was that the shifts in SERS bands calculated from the UBFF and semiempirical method would closely mirror those obtained experimentally and therefore allow for a more accurate determination of the orientation of adsorbates from a comparison of SERS and solution Raman spectral band frequencies. The system which was selected was the adsorption of thiophenol on a silver surface. This system has been well-characterized by Carron et al.26 and these authors have presented solution Raman spectra and SERS spectra of the adsorbed thiphenol. These data were used in our UBFF calculations of thiophenol. In thiophenol, the S-H bend at 914 cm-1 in the neat spectrum does not appear in the SERS spectrum. Therefore, it was concluded that this bond cleaves when the molecule interacts with the surface.26 To analyze the SERS spectrum of thiophenol with the normal mode calculations, the S-H hydrogen was removed from the geometrical parameters that describe the structure and a silver atom was attached to the sulfur atom (Figure 4). The distance between the Ag and S atoms was determined by their covalent-metallic radii. It was estimated as 2.48 Å. The C-S-Ag angle was approximated to 103°. Table 2 shows the observed and calculated in-plane frequencies of the neat and SERS spectra of thiophenol. The average errors of the frequency fit for the spectra were 0.81% for both. This table also shows the shifts on the SERS spectrum. Carron and Hurley26 reported two more bands than those expected for thiophenol, with both bands possessing a1 symmetry. A likely corresponding band was the 1380cm-1 band in the neat spectrum. This band also appeared
obs. freq.
calc. freq.
obs. shift
calc. shift
1584 1576 1478 1440 1380
1575.5 1588.5 1488.7 1453.6
16 -1 -3 0 -5
-3.5 -0.4 -0.6 -3.2
1180 1156 1118 1092 1069 1000 914 698 616 414
1332.3 1250.2 1171.5 1171.0 1101.1 1076.3 1037.8 1001.1 915.5 706.1 611.7 412.1 318.7
0 4 -8 6 1 0 -3 4 6
-2.2 -7.4 -0.2 -0.7 -8.8 -1.0 -0.3 -1.0 +2.2 +0.6 +5.7 +7.2
in the SERS spectrum at 1375 cm-1. Neither PM3 nor the initial UBFF calculation yielded an a1 band within a (70cm-1 range from this frequency. The other extra band was one of the 1069-, 1024-, or 1000-cm-1 bands. The PM3 calculation yielded three modes in this range at 1078, 1015, and 997 cm-1. Of these, the 1015-cm-1 mode had b1 symmetry. Because of the precision with which the PM3 calculations reproduced these three frequencies, the 1024cm-1 band was assigned to the b1 mode. During the frequency fit of the UBFF calculation, a slot was left open on the observed frequency matrix for either the 1069- or 1024-cm-1 band. The optimization yielded a frequency of 1037.8 cm-1. Although closer to 1024 cm-1 than to 1069 cm-1, this value was still less than 3.0% off from 1069 cm-1. Therefore, the b1 assignment of the 1024-cm-1 mode was retained. Apart from the +16-cm-1 shift of the 1584-cm-1 band in the SERS spectrum, the largest shift was the -8-cm-1 shift on the 1118-cm-1 band. This mode is mostly C-C stretching, but the C-S stretching also contributes in a large measure. Removing the S-H hydrogen overestimated this downshift by 9 cm-1. However, the PM3 calculation showed that the π-bond order of the C-S bond increased by 0.013. Because the force constant is directly proportional to the bond order, increasing the C-S stretching force constant brought the calculated shift up to the experimental value. The 230-cm-1 band on the SERS spectrum of thiophenol was assigned to a Ag-S stretching by Carron and Hurley.26 The calculated frequency of this mode depended on the Ag-S stretching force constant and the angle at which the silver attached to the surface. This angle also affected the 414-cm-1 mode. Although the 414-cm-1 mode was mostly C-S stretching, the C-S stretching force constant was set by reproducing the shift on the 1118-cm-1 band as described previously. Therefore, the 230-cm-1 frequency and the shift on the 414-cm-1 band were reproduced by varying only the Ag-S stretching force constant and the Ag-S-C angle. The values of these parameters were 0.8 mdyn/Å and 103°. The 103° angle Ag-S-C was 8° larger than the one determined by Carron and Hurley. The energy minimum of the Zn-thiophenol adduct occurred with the Zn-S bond coplanar with the aromatic ring of the molecule. But a local minimum occurred with the Zn-S bond perpendicular to the ring. The difference between these minima was 10.11 kcal/mol. At the local minimum, the Zn-S-C angle was 110.5°. This angle
6-Mercaptopurine Adsorbed on a Silver Electrode Table 3. Observed and Calculated Frequencies of the Solution Spectrum of 6MP and the Difference between the Solution and SERS Spectra (All Frequencies in cm-1) obs. freq. calc. freq. obs. shift calc. shift 1559 1525 1430 1359 1294 1238 1210 1192 1129 1001 950 870 693 617
1623.3 1595.4 1489.4 1480.7 1474.3 1369.7 1341.3 1257.5 1266.5 1162.7 1220.8 1146.2 990.5 945.9 865.6 649.9 607.2 560.4 419.9 255.9
+13 +7 0 +16 -9 -5 -8 -5 +11 -10 +1 -5 +1 +16
+19.0 +13.2 +8.6 +5.7 +0.4 +5.5 +13.8 +1.4 +1.2 +6.8 -2.6 +8.6 -7.4 +13.8 -10.2 +7.5 +16.0 +6.4 -23.4 +1.1
assignments* ring I str C2-N3 str C2-H def C4-C5 str, N9-H def C8-N9 str, C8-H def C5-N7 str C4-N9 str, C2-H def C2-N3 str, C8-H def N7-C8 str, C8-H def C8-N9 str, N9-H def N1-C2 str, C6-N1 str C6-N1 str, C8-N9 str C6-S10 str, ring I def ring I and II def ring I and II def N3-C4 str, ring I def ring I def ring I def C6-S10 str C6-S10 def
causes the aromatic ring to lay near the surface and allows for a possible interaction between the delocalized electrons on the ring and another silver atom on the surface. The Ag-S-C angle estimated through the UBFF calculation was 7.5° less than the Zn-S-C angle from the PM3 calculation. In the manner described above, two of the four largest shifts on the SERS spectrum were reproduced fairly accurately. The other two largest shifts may be due to an interaction between the aromatic ring and the surface. These are the +16-cm-1 shift of the 1584-cm-1 band and the +6-cm-1 shift on the 1092-cm-1 band. As already mentioned, at an Ag-S-C angle of up to 110.5°, the plane of the thiophenol ring is close enough to the surface so that the delocalized electrons could interact with another atom bulging up on the surface. Both the 1584- and 1092cm-1 modes are mostly ring stretches. Although the results from the UBFF calculations are not conclusive evidence, they are consistent with the S-H cleavage proposition as expressed by Carron and Hurley and reproduce the most significant shifts in the SERS spectrum. These calculations establish the validity of the technique as a predictor of molecular orientation. After the accuracy of the normal mode method in predicting the molecular orientation of thiophenol was established, the procedure was applied to 6MP. Table 3 shows the experimentally observed and calculated SERS frequencies of 6MP, along with the frequency shifts between the solution and SERS data. A comparison of the band intensities of the solution and SERS spectra indicates that the ring II bands increase in intensity with respect to the ring I bands on the SERS spectrum. These changes are relatively small; however, more significant shifts are observed between the frequencies of the solution and SERS spectral bands. The calculated frequencies of the 1294- and 1238-cm-1 bands and those of the 1210- and 1191-cm-1 bands were inverted because of the normal-mode analysis of the solid and solution spectra of 6MP. The frequency fit had an average error of 2.16%. This error is acceptable considering the fact that no optimization of the force constants was done except for the adjustments due to changes on the π-bond order. Various forms of interaction between the 6MP anion and the electrode surface were studied to analyze the SERS spectrum. A face-on interaction, where the silver atom
Langmuir, Vol. 16, No. 7, 2000 3315
interacts with the delocalized electrons on the aromatic rings, was ruled out because preliminary calculations indicated that such an interaction does not result in the observed shift of the ring-bending frequency of 6MP to 633 cm-1. Suzuki and Orville-Thomas43 observed similar shifts when pyridine interacts through the nitrogen with a metal atom. For thiophenol, Carron and Hurley26 showed that the interaction with the surface occurs at the sulfur atom; however, the PM3 calculations showed that the faceon model for 6MP had a heat of formation which was 80.61 kcal/mol higher than that of the lowest head-on model. The various forms of interaction which were finally considered are shown in Figure 5. All the geometrical parameters which were used in the anion calculation were retained. A silver atom was attached at the N1, N3, N7, and S10 positions. The distances between the silver atom and the nitrogen and sulfur atoms were determined by the covalent-metallic radii as was done for thiophenol, and the Ag-N distance was determined to be 2.19 Å. The Ag-N-C angles were approximated to 120°. Suzuki and Orville-Thomas42 calculated the metal-nitrogen stretching force constant for various metals and observed values between 0.9 and 1.3 mdyn/Å. These values served as a guide for this study. In the SERS spectrum of 6MP, the 633-cm-1 band is upshifted 16 cm-1 compared to its solution frequency. The 633-cm-1 band is assigned to a ring I bending mode and indicates that the interaction with the surface is head-on. Table 4 shows the calculated frequency shifts, of the 633cm-1 band, which resulted when a silver atom was attached at various positions on the 6MP structure. The mode of attachment that best reproduced the shift is the one at the N1 position. The Ag-S stretching force constant was 1.3 mdyn/Å, consistent with the values obtained by Suzuki and Thomas-Orville. This mode of attachment also reproduced part of the shift of the 1129-cm-1 band. Besides the shifts of the 1129- and 617-cm-1 bands, other major shifts in the SERS spectrum of 6MP occurred to the 1559-, 1359-, and 1001-cm-1 bands. These modes have large contributions from the C2-N3 stretching, C4N9 stretching, and C6-S10 stretching, respectively. The PM3 calculation showed that when a Zn+ atom is attached to the N1 position, the π-bond order of the C2-N3 and C4-N9 bonds increases and that of the C6-S10 decreases. By adjustment of the stretching force constants of the bonds according to the change in the π-bond order, the calculations reflected the observed shifts of the 1559-, 1359-, and 1001-cm-1 bands. Table 3 shows that most of the frequency shifts on the SERS spectrum are reproduced by the method employed here. From these data it is concluded that, under the experimental conditions described in this report, 6MP attaches to the silver electrode through the N1 position. The largest discrepancies between the observed and calculated shifts were on the 1294-, 1238-, 1210-, and 950cm-1 bands. While the first three of these shifted downward on the spectrum, the calculations showed them shifting upward. These modes all had significant contributions from nonbonded interactions and their downshifts may be due to the loss of charge on the rings as shown by the PM3 calculations. The calculations overestimated the upward shift on the 950-cm-1 band by 12 cm-1. This mode also had a significant contribution from nonbonded interactions and may not shift as much as the UBFF calculation showed because of the loss of electron charge as well. (43) Suzuki, S.; Orville-Thomas, W. J. J. Mol Struct. 1977, 37, 321.
3316
Langmuir, Vol. 16, No. 7, 2000
Vivoni et al.
Figure 5. Possible forms of attachment between 6MP and a silver electrode. Table 4. Observed Shift in cm-1 of the 633-cm-1 Frequency on the SERS Spectrum and the Calculated Shifts of the Possible Forms of Attachment of a Silver Atom to the 6-Mercaptopurine Ion observed shift
16
mode of attachment
calculated shift
Ag-N1 Ag-N3 Ag-N7 Ag-S10
18.1 6.2 -0.4 -0.4
To further explore the conclusion of a head-on attachment of 6MP to the silver surface at the N1 position, the heats of formation of the various forms of attachment were also calculated with the PM3 method. These calculations showed that an attachment through the S10 and N7 is 5.80 kcal/mol more stable than the one through N1. However, an attachment through the N1 position allows the sulfur to bond with another silver atom on the surface. This interaction represents added stability for the system.
4. Conclusion A major application of SERS has been as an in situ probe to characterize the interaction of adsorbates at metal surfaces. This has been achieved by a qualitative interperation of shifts in band frequencies between solution and SERS spectra. The results presented have used UBFF and PM3 calculations to generate molecular vibrational frequencies based on a specific set of molecular orientations. These calculated frequencies were then compared with experimentally determined molecular frequencies and the molecular orientation of the molecule determined. This quantitative-based method serves as a more reliable predictor of molecular orientation. Acknowledgment. This work was supported by Grant 5 R03 CA68993 from the U.S. National Institutes of Health, National Cancer Institute. We acknowledge Dr. Laurent Mars for valuable technical assistance. LA9913194