11248
J. Phys. Chem. B 2000, 104, 11248-11252
Wavelength- and Time-Dependent Two-Photon Photoemission Spectroscopy of Dye-Coated Silicon Surface A. Samokhvalov and R. Naaman* Department of Chemical Physics, Weizmann Institute, RehoVot 76100, Israel ReceiVed: July 6, 2000
Two dye molecules, rhodamine B (RB) and sulforhodamine (SR), were adsorbed on doped silicon. The effect of photoexcitation of the adsorbed species on the electronic properties of the surface was investigated applying wavelength- and time-dependent two-photon photoemission spectroscopy (WD-TPPE). Despite the very similar electronic properties of the two molecules, their adsorption affects very differently the electronic properties of the substrate. It was found that while the adsorbed RB interacts weakly with the substrate and hence preserves its spectral properties, SR interacts strongly with the substrate, causes a decrease in the work function, and looses its “molecular” spectral features. It is shown that photoexcitation of a small fraction (0.01%) of the weakly bound RB species causes a significant change in the electronic properties of the substrate.
Introduction Photoinduced charge transfer between dye molecules and semiconductor substrates is playing a vital role in many processes, among others electrophotography and solar energy cells.1 In addition, adsorbed organic molecules on semiconductor surfaces are known to affect the semiconductor electronic properties. Those properties control many applications such as the operation of electronic devices, photostimulated adsorption/ desorption and photodecomposition of adsorbed molecules, catalysis, etc. It is therefore important that these properties can be modified in a predictable way by adsorption of molecules on the semiconductor surface.2-6 It is common to assume that it is possible to rationalize the effect of adsorbed molecules on electronic properties of the substrate by comparing the relative energy of the molecule’s highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) with the substrate bands edges.7 In the present paper two dye molecules, rhodamine B (RB) and sulforhodamine (SR)8 were adsorbed on the silicon surface (see Scheme 1). These molecules have similar electronic properties,9 namely, their ionization potential and their HOMOLUMO energies are very similar. However, the chemical interaction of these molecules with the surface is expected to be different due to the anionic sulfo group in SR that may interact strongly with the silicon oxide substrate. Hence, the goal of the present study is to reveal the changes in electronic properties of adsorbate-substrate systems, when two similar adsorbates interact in a different chemical way with the substrate. The effect of photoexcitation of the adsorbed species on the electronic properties of the surface was investigated applying wavelength- and time-dependent two-photon photoemission spectroscopy (WD-TPPE).4,10 In this method one uses two laser pulses. The first one, the “pump” laser, excites electrons either in the adsorbed molecules or near the surface of the solid and after some delay the second one, the “probe” laser, ejects photoelectrons from the surface. Tuning the pump laser wavelength allows one to excite electrons into a certain welldefined set of electronic levels. In semiconductors, the photoexcitation proceeds usually from the valence band (VB) to the
SCHEME 1: Structural Formulas of the Dyes Used: Rhodamine B (RB) and Sulforhodamine (SR)
conduction band (CB), from surface states to the CB, or from the VB to surface states. The dependence of photoelectron signal upon the delay between the pump and probe pulses allows determination of the lifetime of the corresponding excited state. It is important to realize that the photoelectrons are ejected from the ground state. TPPE is probing different properties than twophoton photoemission spectroscopy, in which the electrons are ejected from the excited state.11-13 Therefore TPPE is sensitive to much longer time scale dynamics. In the past, most of the studies on electronic properties and surface charge dynamics were performed using contact potential difference (CPD),14 surface photovoltage spectroscopy (SPV),15 and photoluminescence (PL) decay techniques. The WD-TPPE technique applied here provides insight both on the changes in the work function due to the adsorbate and on the electrostatic potential barrier induced by the adsorbed molecules. Early studies using time-dependent TPPE spectroscopy were performed on clean surfaces of semiconductors and metals. These experiments allowed one to study the energy and lifetime of intrinsic surface states. Most of the attention was paid to studying metals,12,16 and only few works were performed on semiconductors.10,17 Even less attention has been paid to the application of the time-dependent TPPE spectroscopy to study the energy structure and photorelaxation dynamics of surface states of semiconductors modified by the adsorption of organic molecules. Recently, WD-TPPE spectroscopy was used to investigate the surface states of CdTe before and after specially designed organic molecules were adsorbed on it.4 The energy
10.1021/jp002420g CCC: $19.00 © 2000 American Chemical Society Published on Web 11/01/2000
WD-TPPE of Dye-Coated Silicon Surface
J. Phys. Chem. B, Vol. 104, No. 47, 2000 11249
of surface states, relative to the valence and conduction bands, was found to correlate with the relaxation time of the excited states. The differences in the relaxation kinetics between the bare CdTe and CdTe covered with organized organic thin films (OOTFs) demonstrated that the properties of surface states of semiconductors can be manipulated by adsorption of suitable organic molecules. Experimental Section The present studies were performed on n-Si (100) crystals, phosphorus-doped 1015/cm3, that were etched with a 5% solution of HF to remove natural oxides, followed by rinsing with deionized water, followed by oxidation in an ozone chamber to grow 5-10 Å of protective oxide layer.18 SPV experiments were performed also on a p-doped silicon (doping by 1015/cm3 boron). Two dyes, rhodamine-like molecules, rhodamine B (RB) and sulforhodamine (SR), were used to coat the silicon surface. Scheme 1 shows the structural formulas of the dyes used. Rhodamine B (RB) was used as a perchlorate, and sulforhodamine (SR) as the sodium salt. The wafers of n- or p-Si were coated with the organic molecules by immersing them in a 5 mM solution of the molecules in the 1/1 mixture of dimethylformamide and triethylamine. After about 12 h in solution, the wafers were removed, rinsed with dichloromethane, and dried with nitrogen gas. To prepare the “bare” silicon samples, i.e., samples that are not covered with the organic molecules, the Si wafers were treated as described above, but in this case the solution did not contain the dye molecules. FTIR adsorption spectra indicated that under these conditions about one monolayer of the dye molecules is adsorbed. The samples were attached to a holder and inserted into an ultrahigh vacuum chamber, pumped to below 10-8 Torr. The photoelectron signal was measured either with a time-of-flight or a retarding field19 electron energy analyzer. A Nd:Yag pumped dye laser (Spectra-Physics) was used as the “pump” laser at 637 and 532 nm and a 193 nm laser beam was used as the “probe” (Lambda Physik Compex 102 excimer laser). The energy of the “probe” was about 50 nJ/pulse with a pulse length of about 20 ns. The energy was kept low to avoid nonlinear and charging effects. The “pump” laser energy was in the range from few nanojoules to 10 µJ/pulse with a beam area of about 1 cm2 and a pulse length of 7 ns. Results Figure 1A,B shows the kinetic energy distribution of the photoelectrons ejected from bare silicon surface or surface coated with SR, respectively. As it is well established, the spectra reflect the emitted photoelectron energy distribution and the energy-dependent transmission of the adsorbed layer.19 The photoelectron energy distribution was measured with the probe laser alone (solid line) or together with the pump laser when it was tuned to 532 or 637 nm. The high-energy cutoff of the energy distribution is at 1.5 and 2.2 eV for the bare silicon and for the silicon coated with SR. Since the photon energy is 6.4 eV, a work function of 4.9 ( 0.2 eV is obtained for the bare silicon. The adsorption of the dye causes a drop of about 0.7 eV in the work function to a value of 4.2 ( 0.2 eV. When photoelectrons are ejected while photoexciting with the pump laser at 532 or 637 nm (dashed and dotted lines), the highenergy edge of the electron energy distribution is shifted to low energy by 0.3 eV, both for the bare surface and for the dyecoated surface. The results for the RB-coated surface are similar to the bare silicon. The magnitude of the shift in the high-energy
Figure 1. The photoelectron energy distribution obtained for the “bare” silicon (A) and for silicon coated with sulforhodamine (B). The spectra were obtained by applying the “probe” laser alone, operating at 193 nm (solid line), or when it was applied together with the “pump” lasers operating at λoff ) 637 nm (dotted line) or at λres ) 532 nm (dashed line).
cutoff depends very weakly on the photoexcitation wavelengths the shorter the wavelength, the larger is the shift to low energy. Figure 2 presents the dependence of the integrated (over-all energies) photoelectron signal on the time delay between the pump and probe lasers for the “bare” silicon and the silicon covered with each dye. The pump laser was tuned to resonance absorption of the dyes, λres ) 532 nm. The data are displayed in three time scales, tens of microseconds (top), microseconds (middle), and hundreds of nanoseconds (bottom). The surface covered with SR shows the shortest relaxation time, and when coated with RB, the longest relaxation time is obtained. All curves could be fitted by a multiexponential decay function, S(t) ) ∑i ai exp(t/τi) where S(t) is the time-dependent signal, τi is the decay time constant, and ai is the coefficient indicating the relative contribution of each component. When the same samples were pumped at a wavelength which does not correspond to the absorption of the dye (off-resonance) λoff ) 637 nm, all the samples show about the same time dependence. The “off-resonance” decay time constants for “bare” and covered silicon are similar to those obtained with the “resonance” excitation of bare silicon. All of the time constants are listed in Table 1, both for resonance and off-resonance excitation. Figure 3 presents the integrated PE signal as a function of the delay between the pump and probe lasers, for different intensities of the “pump” laser at λoff ) 637 nm (A) and λres ) 532 nm (B) for Si coated with RB. When the low laser intensity (tens of nanojoules per pulse) was applied at the off-resonance wavelength λoff ) 637 nm, three time constants were observeds the “fast” one τ1 ) 0.07 ( 0.01 µs, the “middle” one τ2 ) 3.7 ( 0.6 µs, and the “slow” one with τ3 ∼100 µs. The same results were obtained for the bare Si. The fact that the time response
11250 J. Phys. Chem. B, Vol. 104, No. 47, 2000
Samokhvalov and Naaman
Figure 3. The dependence of the integrated photoelectrons signal on the delay between the “pump” and the “probe” lasers for different pump laser intensities for Si surface coated with RB. The sample was photoexcited at λoff ) 637 nm (A) and with the λres ) 532 nm (B).
Figure 2. The dependence of the total photoelectron signal on the delay between the “pump” and the “probe” lasers. The “pump” laser is at the “resonance” wavelength λres ) 532 nm. Squares denote data obtained with the monolayer of the dye SR adsorbed on silicon, circles denote data obtained with the “bare” silicon, triangles denote data obtained with the monolayer of the dye RB adsorbed on silicon. Different plots show kinetic data obtained with different time resolution. Thin lines denote best fits to the experimental data.
TABLE 1: Relaxation Time Constants for the Monolayers of Different Dyes and Bare Silicon Measured in the TPPE Experiment upon Photoexcitation with Pump Wavelength λres ) 532 nm and λoff ) 637 nm substrate
λpump (nm)
τ1
time constants, µs τ2
τ3
“Pump” Intensity about 10 µJ/Pulse (high-intensity limit) SR 532 0.05 ( 0.01 5 ( 1 >100 bare Si 532 0.07 ( 0.03 20 ( 5 100 ( 50 RB 532 0.09 ( 0.02 135 ( 25 >100 all substrates 637 0.08 ( 0.01 9 ( 0.5 ∼100 “Pump” Intensity about 10 nJ/Pulse (low-intensity limit) RB 532 none none ∼1000 bare Si and SR 532, 637 0.07 ( 0.01 3.7 ( 0.6 ∼100
must be presented by a multiexponential function indicated that there are many states involved with different time responses which can be presented by these “averaged” values. In the case of the resonance wavelength λres ) 532 nm, we observed very slow relaxation with the time constant τ of hundreds to thousands of microseconds with the full relaxation time estimated as thousands of microseconds. When the laser intensity is increased up to tens of microjoules per pulse, for the off-resonance excitation of RB on silicon (Figure 3A), again three time constants can be identified τ1 ) 0.08 ( 0.01 µs, τ2 ) 9 ( 0.5 µs, and τ3 ∼ 100 µs. About the same results are obtained for the bare Si pumped either at 637
Figure 4. The logarithm of I0/Ip as function of the pump laser intensity. I0 is the integrated photoelectron signal without photoexcitation, and Ip is the integrated signal immediately after excitation with the pump laser. The results are presented for pump laser with λoff ) 637 nm (A) and for λres ) 532 nm (B), for the bare silicon (circles), silicon coated with SR (squares), and silicon coated with RB (triangles). The data were fitted by the function y ) axb (solid line).
or at 532 nm. In the case of Si coated with RB pumped at the λres ) 532 nm (Figure 3B), the time constants are τ1 ) 0.09 ( 0.02 µs, τ2 ) 135 ( 25 µs, and τ3 > 100 µs. Table 1 summarizes the measured time constants for high-intensity (a) and lowintensity (b) illumination. Figure 4 presents the dependence of the logarithm of I0/Ip as function of the pump laser intensity. I0 is the integrated photoelectron signal without photoexcitation and Ip is the integrated signal immediately after excitation with the pump laser. In the case of a simple scattering process, in which electrons are scattered from photoexcited adsorbed molecules, we expect this function to depend linearly on the number of excited molecules, namely to depend linearly on the laser intensity. The results are presented for pump laser with λoff ) 637 nm (A) and for λres ) 532 nm (B) for the bare silicon (circles), silicon coated with SR (squares), and silicon coated with RB (triangles). In the case of resonance excitation of silicon covered with RB the dependence is linear for the low pump laser intensities, contrary to all other cases where the dependence can be described by a sub-linear dependence. The data in Figure 4 were fitted to the function ln(I0/Ip) ) axb. The parameters a and b are shown in the figure. The parameter b is smaller than
WD-TPPE of Dye-Coated Silicon Surface
Figure 5. The surface photovoltage signal measured for p-type (A) and n-type (B) silicon surfaces. The signal was obtained by measuring the work function of the bare silicon (circles), or silicon coated with RB (triangles), or SR (squares) relative to gold surface, while illuminating the samples with white light at various intensities.
unity for all cases except for the resonance excitation of the surface coated with RB where it is unity for all the low intensities till saturation is reached. Figure 5 presents the SPV results obtained by measuring the changes in the work function of the bare silicon (circles), or silicon coated with RB (triangles), or SR (squares) relative to gold surface, while illuminating the samples with white light at various intensities. Both p-type (Figure 5A) and n-type (Figure 5B) silicon surfaces were investigated. The results indicate that in the case of bare substrates or surfaces coated with RB, the work function changes in the direction expected for the relevant doping, namely, it increases upon illumination for p-doped and decreases for n-doped silicon. However, for the p-doped substrate coated with SR a very sharp increase in the work function is observed. Interestingly, for the n-doped silicon coated with SR, at low light intensity a slight lowering in the work function is observed, as expected, but with higher light intensity the work function increases. Discussion Despite the very similar structure and spectrum of the two dyes studied, their effect on the TPPE is very different. While SR shortens the lifetime, RB lengthens it. In addition, their effect on the intensity dependence of the TPPE is different (Figure 4). In the case of SR the intensity dependence is similar to that of the bare silicon, while RB shows a linear dependence. The key for understanding these observation is seen in Figure 5. From this figure it is clear that while SR interacts strongly with the substrate, causing a dramatic reduction in the work function, RB has much weaker interaction. Figure 1 also indicates that the adsorption of SR causes a sharp decrease in the work function. This decrease occurs both for n- and p-type substrates. Hence, we conclude that SR is a very good donor, which, upon adsorption, donates electrons to the substrate and
J. Phys. Chem. B, Vol. 104, No. 47, 2000 11251 forms a positively charged adlayer. This property stems from the chemical structure of SR (see Scheme 1) in which the SO3group is expected to interact strongly with the silicon oxide substrate and donate electrons to the surface. RB on the other hand does not have a distinguished donating group and therefore interacts much more weakly with the surface. Commonly upon illumination of a semiconductor, its work function will increase for the p-doped bulk and decrease for a n-doped material. This is indeed observed for the bare surfaces and for the surfaces coated with RB, indicating an expected band flattening (Figure 5). However, in the case of substrate coated with SR a sharp increase in work function is observed, upon illumination, for both n- and p-doped silicon. The positively charged layer formed due to the strong donor character of the SR can again explain this phenomenon. The photoexcitation, in this case, transfers electrons from the valence band to the empty molecular states on the surface. This effect causes the neutralization of the positively charged surface and an increase in the work function. In Figure 5B one can realize that for high light intensity, a similar effect appears for the RBcoated surface, indicating that RB is also a donor but much weaker than the SR. It is interesting to note that when illuminated with strong enough light, the work function obtained for the surfaces coated with both dyes reaches nearly the same value. This indicates that photoexcitation increases the coupling of the dyes with the substrate also for the more weakly coupled RB. This increase in coupling results from excitation of the dyes, namely transfer of an electron from their HOMO to the LUMO. Now, electrons from the bulk can refill the HOMO causing a net charging of the surface by electrons. As will be discussed below, this process results in reduction of the photoemission from the surface. We chose to present the time dependence as a tripleexponential function. The time constants obtained do not reflect the decay time of a specific state, but rather they have to be taken as “averaged values” characterizing the time response of all systems. All of the time-dependent results can be rationalized based on the fact that SR interacts strongly with the surface and therefore its states are indistinguishable from the surface states of the solid, while the states corresponding to the weakly interacting RB keep their molecular identity. As expected, when the photoexcitation occurs at a wavelength that does not correspond to the absorption of the dyes, all of the samples show the same time dependence (see Table 1). However, when exciting at 532 nm, which corresponds to the resonance in the absorption of the dyes, the substrate coated with RB is distinguishable from the two others. Much larger time constants are obtained for the RB-coated substrate as a result of the weak coupling of the molecular-related surface states to the bulk states. The same behavior is seen in the intensity-dependent studies (Figure 4). While the SR-coated substrate and the bare substrate show similar intensity dependence when the intensity of the “pump” laser at 532 nm is varied, the surface covered with RB shows the intensity dependence expected for resonance excitation of molecular states. When the surface coated with RB is excited at the resonance wavelength (532 nm), the pump-laser intensity dependence of the depletion shows a linear behavior (Figure 4B), till saturation is achieved at an intensity of about 2 nJ/cm2. At the saturation intensity, only about 5 × 108 molecules/cm2 are excited. Since a single monolayer includes about 1013 molecules/cm2, it means that exciting about 10-4 of the molecules on the surface is sufficient to saturate the effect. The question is why photoex-
11252 J. Phys. Chem. B, Vol. 104, No. 47, 2000 citation of the molecules is so efficient in quenching the photoemission. It is important to realize that the effect of excitation cannot be explained as a simple interaction of electrons with excited molecules. If this is done, a ridiculous scattering cross section of about 10-10 cm2 is obtained for the effect. The explanation lies rather in the change in the surface potential upon excitation. Namely, as explained above, due to the electronic excitation of the molecule, its HOMO is partially depleted and electrons are transferred from the substrate to the HOMO of the adsorbed RB molecules. As a result, the net negative charge on the surface increases, which causes an increase in the work function, hence reduction in the photoemission yield. Our results provide direct evidence that excitation of very few molecules on the surface is enough to affect in a substantial way the photoemission process. This phenomenon was observed in the case of metal atom adsorption on semiconductor surfaces,20 but here it is demonstrated in a quantified way for photoexcitation of adsorbed molecules. Conclusions In the present study it has been demonstrated that the effect of adsorbed molecules on the electronic properties of semiconductor surface can vary significantly, even if the molecules are very similar. The donor-acceptor interaction between the adsorbates and the surface varies significantly for the two dyes studied. Since the two molecules have similar aromatic moieties, the results indicate that their interaction with the substrate is not through their aromatic groups. Hence, it is expected that at least the SR is not lying parallel to the surface, but rather “stands” with the sulfuric group bound to the substrate. As expected, when the interaction is strong, the molecular states lose their “molecular identity”, while for weak interaction, one can recognize the molecular resonances. In the case of “weak interaction” between the adsorbed molecule and the substrate, photoexcitation of the adsorbates can cause a significant increase in the interaction, manifesting itself as a large change in the surface properties, e.g., the work function. We proved that it is enough to excite a small portion of the adsorbed molecules to have this effect expressed in full. Despite the known fast electron transfer from the excited adsorbed molecule to the substrate, the full relaxation time of
Samokhvalov and Naaman the photoexcited surface states proceeds on a long time scale, up to milliseconds. The time scale is apparently a result of the time it takes for electrons to tunnel between the photoexcited adsorbed molecule and the bulk, through the oxide layer. Acknowledgment. We thank Prof. D. Cahen for critical reading of the manuscript and for his comments. References and Notes (1) Rehm, J. M.; McLendon, G. L.; Nagasawa, Y.; Yoshihara, K.; Moser, J.; Gra¨tzel, M. J. Phys. Chem. 1996, 100, 9577. (2) (a) Skromme, B. J.; Sandroff, C. J.; Yablonovitch, E.; Gmitter, T. Appl. Phys. Lett. 1987, 51, 2022. (b) Sandroff, C. J.; Nottenburg, R. N.; Bischoff, J. C.; Bhat, R. Appl. Phys. Lett. 1987, 51, 33. (c) Yablonovitch, E.; Sandroff, C. J.; Bhat, R.; Gmitter, T. Appl. Phys. Lett. 1987, 51, 439. (3) Avouris, P.; Walkup, R. E. Annu. ReV. Phys. Chem. 1989, 40, 173. (4) Kadyshevitch, A.; Naaman, R.; Cohen, R.; Cahen, D.; Libman, J.; Shanzer, A. J. Phys. Chem. B 1997, 101, 4085. (5) Bruening, M.; Moons, E.; Yaron-Marcovich; Cahen, D.; Libman, J.; Shanzer, A. J. Am. Chem. Soc. 1994, 116, 2972. (6) Cohen, R.; Bastide, S.; Cahen, D.; Libman, J.; Shanzer, A.; Rosenwaks, Y. AdV. Mater. 1997, 9, 746. (7) See for example: Chasse´, T.; Wu, C.-I; Hill, I. G.; Kahn, A. J. Appl. Phys. 1999, 85, 6589. (8) Also referred to as rhodamine 610 and kiton red, respectively. Exciton laser dye catalogue. (9) The oxidation potential for RB is +1.2 V vs NHE as given by Eichberger, R.; Willig, F. Chem. Phys. 1990, 141, 159. The oxidation potential of SR is +1.4 eV vs NHE as given by Parkinson, B. A. Langmuir 1988, 4, 967. (10) Bokor, J.; Haight, R.; Storz, R. H.; Stark, J.; Freeman, R. R.; Bucksbaum, P. H. Phys. ReV. B 1985, 32, 3669. (11) Fauster, T.; Steinmann, W. In Photonics Probes of Surfaces; Halevi, P., Ed.; Elsevier: Amsterdam, 1995; pp 347-411. (12) Harris, C. B.; Ge, N.-H.; Lingle, R. L., Jr.; McNeill, J. D.; Wong, C. M. Annu. ReV. Phys. Chem. 1997, 48, 711. (13) Petek, H.; Ogawa, S. Prog. Surf. Sci. 1997, 56, 239. (14) Surplice, N. A.; D’Arcy, R. J. J. Phys. E.: Sc. Instrum. 1970, 3, 477. (15) Lagowski, J. Sur. Sci. 1994, 299/300, 92. (16) Knoesel, E.; Hertel, T.; Wolf, M.; Ertl, G. Chem. Phys. Lett. 1995, 240, 409. Hotzel, A.; Ishioka, K.; Knoesel, E.; Wolf, M.; Ertl, G. Chem. Phys. Lett. 1998, 285, 271. (17) Bokor, J.; Halas, N. J. IEEE J. Quantum Electron. 1989, 23, 2550. (18) Niwano, M.; Suemitsu, M.; Ishibashi, Y.; Takeda, Y.; Miyamoto, N.; Honma, K. J. Vac. Sci. Technol. A 1992, 10, 3171. (19) Detailed description of the experimental setup is given in Naaman, R.; Haran, A.; Nitzan, A.; Evans, D.; Galperin, M. J. Phys. Chem. B 1998, 102, 3658. (20) Cao, R.; Miyano, K.; Kendelewicz, T.; Lindau, I.; Spicer, W. E. J. Vac. Sci. Technol. 1987, B5, 998.
