Infrared Spectroscopy and Surface Chemistry of β-Ga2O3 Nanoribbons

Nov 1, 2007 - Search; Citation; Subject. Search .... Effect of heat treatment time on local structure of lanthanum-aluminum–gallium-borate xerogels...
0 downloads 0 Views 303KB Size
Download PDF
12566

Langmuir 2007, 23, 12566-12576

Infrared Spectroscopy and Surface Chemistry of β-Ga2O3 Nanoribbons V. M. Bermudez* and S. M. Prokes Electronics Science and Technology DiVision, NaVal Research Laboratory, Washington, District of Columbia 20375-5347 ReceiVed July 18, 2007. In Final Form: September 5, 2007 The structure and surface chemistry of crystalline β-Ga2O3 nanoribbons (NRs), deposited in a thin layer on various metallic and dielectric substrates (mainly on Au), have been characterized using vibrational spectroscopy. The results have been analyzed with the aid of a previous ab initio theoretical model for the β-Ga2O3 surface structure. Raman spectra and normal-incidence infrared (IR) transmission data show little if any difference from corresponding results for bulk single crystals. For a layer formed on a metallic substrate, IR reflection-absorption spectroscopy (IRRAS) shows longitudinal-optic (LO) modes that are red-shifted by ∼37 cm-1 relative to those of a bulk crystal. Evidence is also seen for a bonding interaction at the Ga2O3/Au interface following heating in room air. Polarization-modulated IRRAS has been used to study the adsorption of pyridine under steady-state conditions in ambient pressures as high as ∼5 Torr. The characteristic ν19b and ν8a modes of adsorbed pyridine exhibit little or no shift from the corresponding gas-phase values. This indicates that the surface is only weakly acidic, consistent with the theoretical prediction that singly unsaturated octahedral Ga sites are the only reactive cation sites on the NR surface. However, evidence for adsorption at defect sites is seen in the form of more strongly shifted modes that saturate in intensity at low pyridine coverage. The effect of H atoms, formed by thermal cracking of H2, has also been studied. No Ga-H or O-H bonds are observed on the pristine NR surface. This suggests that the previously reported presence of such species on Ga2O3 powders heated in H2 is a result of a partial reduction of the oxide surface. The heat of adsorption of atomic H on the pristine β-Ga2O3(100) surface at 0 K is computed to be -1.79 eV per H at saturation (average of Ga-H and O-H sites), whereas a value of +0.45 eV per H is found for the dissociative adsorption of H2. This suggests that rapid recombinative desorption of H2 may limit the coverage of chemisorbed H on this surface.

1. Introduction The surface properties of Ga2O3, particularly in the most stable monoclinic (β) form, are important in several areas of technology. This material can be made to be n-type semiconducting, either by doping or through the introduction of oxygen vacancies, and sensors capable of operation at elevated temperature have been developed for a variety of chemical species on the basis of the conductivity changes that result from adsorption. Because of its wide band gap (∼4.5-4.9 eV), semiconducting β-Ga2O3 is of interest as a transparent conducting oxide to replace indiumdoped tin oxide (ITO) in electro-optic devices. β-Ga2O3 has superior optical transmission in the near-ultraviolet and is easier than ITO to fabricate reproducibly. The oxide surface properties are important in determining the stability and electronic characteristics of the interface with the active material, e.g., an organic light-emitting diode. Furthermore, much work on Ga2O3 surfaces (see below) has been motivated by its use as a support in industrial catalysts for light alkane dehydrogenation and aromatization. Many studies of the surface chemistry of β-Ga2O3 have been reported,1-7 primarily involving infrared (IR) spectroscopy of high-surface-area (HSA) powders. The adsorbates investigated include CH3OH;1 CO2;2,6 H2 and D2;2,3,6 CO;4-6 pyridine;5,6 and * Corresponding author. Phone: +1-202-767-6728. Fax: +1-202-7671165. E-mail: [email protected]. (1) Branda, M. M.; Collins, S. E.; Castellani, N. J.; Baltana´s, M. A.; Bonivardi, A. L. J. Phys. Chem. B 2006, 110, 11847. (2) Collins, S. E.; Baltana´s, M. A.; Bonivardi, A. L. J. Phys. Chem B 2006, 110, 5498. (3) Collins, S. E.; Baltana´s;, M. A.; Bonivardi, A. L. Langmuir 2005, 21, 962. (4) Rodrı´guez Delgado, M.; Otero Area´n, C. Mater. Lett. 2003, 57, 2292. (5) Rodrı´guez Delgado, M.; Otero Area´n, C. Z. Anorg. Allg. Chem. 2005, 631, 2115. (6) Bozon-Verduraz, F.; Potvin, C. J. Chim. Phys. 1976, 73, 43. (7) Pohle, R.; Fleischer, M.; Meixner, H. Sens. Actuators B 2000, 68, 151.

10.1021/la7021648

H2O, C2H4 and NH3.6 One novel study,7 focusing on sensor applications, used IR emission spectroscopy combined with electrical conductivity measurements to monitor the interaction of β-Ga2O3 HSA powder with H2, H2O, C2H4, (CH3)2CO, and CH3CH2OH at temperatures as high as 635 °C. The adsorption of several reagents on R- and/or γ-Ga2O3 has also been studied,2,3,5,8-11 again using IR spectroscopy of HSA powders, and C2H212 and C2H613 have been investigated as adsorbates on unspecified forms of Ga2O3. Other techniques have, in a few cases, also been used to study adsorption on various Ga2O3 surfaces. Temperature-programmed desorption has been applied14 to the reaction of CH4 with β-Ga2O3 at elevated temperatures, and electron spin resonance spectroscopy has been used15 to characterize the Lewis acid properties of R-, β-, γ-, and δ-Ga2O3 HSA powders via detection of radical species formed in the adsorption of electron-donor molecules. Some of these results will be described in more detail, as appropriate, in the following discussion. (8) Vimont, A.; Lavalley, J. C.; Sahibed-Dine, A.; Otero Area´n, C.; Rodrı´guez Delgado, M.; Daturi, M. J. Phys. Chem. B 2005, 109, 9656. (9) Lavalley, J. C.; Daturi, M.; Montouillout, V.; Clet, G.; Otero Area´n, C.; Rodriguez Delgado, M.; Sahibed-dine, A. Phys. Chem. Chem. Phys. 2003, 5, 1301. (10) Rodrı´guez Delgado, M.; Morterra, C.; Cerrato, G.; Magnacca, G.; Otero Area´n, C. Langmuir 2002, 18, 10255; Otero Area´n, C.; Lo´pez Bellan, A.; Pen˜arroya Mentruit, M.; Rodrı´guez Delgado, M.; Turnes Palomino, G. Microporous Mesoporous Mat. 2000, 40, 35. (11) Meriaudeau, P.; Primet, M. J. Mol. Catal. 1990, 61, 227. (12) Ivanov, A. V.; Koklin, A. E.; Uvarova, E. B.; Kustov, L. M. Phys. Chem. Chem. Phys. 2003, 5, 4718. (13) Kazansky, V. B.; Subbotina, I. R.; Pronin, A. A.; Schlo¨gl, R.; Jentoft, F. C. J. Phys. Chem. B 2006, 110, 7975. (14) Becker, F.; Krummel, Ch.; Freiling, A.; Fleischer, M.; Kohl, C. Fresenius J. Anal. Chem. 1997, 358, 187. (15) Parenago, O. O.; Pushkar’, Yu. N.; Turakulova, A. O.; Murav’eva, G. P.; Lunina, E. V. Kinet. Katal. 1998, 39, 288; Kinet. Catal. 1998, 39, 268 (English translation).

This article not subject to U.S. Copyright. Published 2007 by the American Chemical Society Published on Web 11/01/2007

β-Ga2O3 Nanoribbons

Figure 1. Schematic diagram of the ideally terminated (100)-B surface (ref 16). Atomic radii and interatomic distances are approximately to scale. (a) View along the surface normal with dashed lines indicating the surface unit cell. For simplicity, only the outermost layer of each atom type [O(I), Ga(I), etc.] is shown. (b) The five outermost surface layers, viewed along the [010] direction, with vectors indicating the displacements of Ga(II) and O(III) that occur during relaxation. For clarity, the displacement Vectors haVe been magnified by a factor of 4 relatiVe to the interatomic distances, and displacements of translationally equivalent atoms have been omitted. Other displacements are too small to be seen, even when magnified 4 times. The notation is consistent with that used in ref 16. In the present work, Ga(I) is referred to as Ga(Td) and Ga(II) as Ga(Oh).

In bulk Ga2O3, depending on the crystal form, gallium occupies either octahedral [Ga(Oh)] sites or both octahedral and tetrahedral [Ga(Td)] sites. Different inequivalent oxygen sites are also involved. The HSA powders can exhibit surface planes with varying degrees of coordinative unsaturation of both Ga and O sites and also with varying concentrations of defects such as O vacancies and OH sites. Thus, the distribution of reactive sites can depend on both the bulk crystallographic structure and the method of surface preparation and can differ from that expected for an ideally terminated bulk structure. By analogy with Al2O3 HSA powders, the Ga(Td) site is thought to be a stronger Lewis acid than the Ga(Oh) one. This is supported by IR data,8 and the relative concentration and coordinative unsaturation of the two types of sites are important factors in the surface chemistry. Surface acidity is traditionally studied using the adsorption of a molecular charge donor such as pyridine. The use of nanoribbons (NRs), which are true single crystals with well-defined surface planes, offers advantages in studies of β-Ga2O3 surface chemistry. Nanoribbons have a higher surfaceto-volume ratio (SVR) than a thin film, meaning that a larger fraction of the atoms is on or near the surfaces. Hence, NRs are potentially useful as catalyst supports, an area traditionally dominated by HSA powders. The unique structural properties of NRs open the possibility of studying surface reactions on a substrate with well-defined surfaces that is, nevertheless, high in SVR. The wide face of the NR is the (100) plane, which is also the preferred cleavage plane of the bulk single crystal (ref 16 and work cited). The structures of several low Miller index surfaces of β-Ga2O3 have recently been examined theoretically.16 The bulk lattice involves both Ga(Td) and Ga(Oh) cation sites [labeled “Ga(I)” and “Ga(II)”, respectively, in Figure 1]. There are three inequivalent anion sites, conventionally labeled “O(I)”, “O(II)”, and “O(III)”. The so-called “(100)-B” surface16 (Figure 1), which has the lowest surface energy (0.68 J m-2), (16) Bermudez, V. M. Chem. Phys. 2006, 323, 193.

Langmuir, Vol. 23, No. 25, 2007 12567

is almost ideally terminated and undergoes little relaxation in comparison to other, less stable surfaces. On this surface, Ga(Oh) is missing one of six O nearest neighbors, and O(III) is missing one of four Ga nearest neighbors. All other species, in particular the Ga(Td) sites, are fully coordinated. Hence, in the absence of O vacancies, the NR (100) surface is expected to be a relatively weak Lewis acid in comparison to typical Ga2O3 HSA powders that involve unsaturated Ga(Td) sites. The high SVR also makes NRs potentially useful for highsensitivity sensors, an application for which thin films are currently used. The use of NRs permits the monitoring of chemisorptioninduced changes in electrical conductivity and, conversely, the effects on catalytic activity of changes in the NR electron density. This is important in sensor applications and also as a potential method for achieving electronic control of surface chemical reactions, as discussed in ref 17 for the case of SnO2 nanowires. The purpose of the present work is to characterize the chemically active surface sites on β-Ga2O3 NRs using IR spectroscopy as the primary experimental technique. This is a necessary first step in developing these NRs for the applications noted above. An understanding of the IR spectroscopy of the NRs themselves, prior to any surface-chemical treatment, is a prerequisite in achieving the stated purpose of this study. Hence, the first part of this work addresses the vibrational modes of the NRs when deposited in the form of a thin layer. The second part uses exposure to atomic H and to pyridine to probe the nature of coordinatively unsaturated surfaces sites. 2. Experimental Details The β-Ga2O3 NRs were grown by placing Ga metal in an alumina boat and inserting it into a quartz tube mounted in a tube furnace. The quartz tube was pumped at one end, and several gas feed valves and flow controllers were connected to the other end, allowing a controlled flow of Ar and O2 gas to pass over the alumina boat. The quartz tube was evacuated to a base pressure of 25 mTorr and the furnace was then raised to 900 °C. Once at temperature, the NR growth was initiated by flowing a 6:1 Ar:O2 gas mixture. The tube pressure remained generally in the 5-10 Torr range during growth. The growth of Ga2O3 nanowires proceeds by the vapor-liquidsolid (VLS) growth process, requiring the use of an Au metal catalyst. However, in the case of the nanoribbons, we have found that the Ga metal itself acts as the catalyst.18 For depositing a layer of NRs, a suspension in CH3OH was first prepared by ultrasonic agitation. The concentration was unknown, but the suspension appeared cloudy to the unaided eye. It is estimated that the mixture consisted of about 1 g of NRs in 20 mL of CH3OH. For IR reflectance measurements, the NRs were deposited on a metallic substrate that was heated to ∼180 °C on a hotplate in air, for rapid evaporation of the CH3OH, while the suspension was either slowly dropped from a pipet or sprayed as a fine mist from an airbrush19 using dry N2 as the propellant. The latter method gave a more uniform coverage and was used when making samples for adsorption studies. The average thickness of the deposited layer was estimated by comparing the observed and calculated (see below) intensities of the β-Ga2O3 phonon spectra. Some data were also taken in normal-incidence transmission for samples prepared by deposition onto a KRS-5 window. For IR studies of HSA oxide powders, a widely used sample preparation method involves compacting the material into a self-supporting wafer by means of a high-pressure ram. This technique was not used here in order to avoid possible damage to the NRs. (17) Zhang, Y.; Kolmakov, A.; Lilach, Y.; Moskovits, M. J. Phys. Chem. B 2005, 109, 1923. (18) Prokes, S. M.; Carlos, W. E.; Glembocki, O. J. Proc. Soc. Photo-Opt. Instrum. Eng. 2005, 6008, 60080C-1. (19) Model HS; Paasche Airbrush Co., http://www.paascheairbrush.com. (20) Chidsey, C. E. D.; Loiacono, D. N.; Sleator, T.; Nakahara, S. Surf. Sci. 1988, 200, 45.

12568 Langmuir, Vol. 23, No. 25, 2007 For experiments in ultrahigh vacuum (UHV), the metallic substrate was an optically thick Au film deposited on a mica sheet (0.13 mm thick) with no adhesion layer at the interface. Metals (e.g., Cr) used to enhance the adhesion of Au films on substrates tend to diffuse to the Au surface upon moderate annealing. This is unsuitable for the present experiments, since such metals are chemically reactive with the species studied here. However, Au adheres well to bare mica (e.g., refs 20,21) at temperatures as high as ∼600 °C and exhibits a (111) ordering upon annealing. The sample was clamped against a Mo block with holes in which heating elements were buried. These consisted of tightly wound coils22 of W wire coated with Al2O3 for electrical insulation. A Ta plate was sandwiched between the sample and the Mo block, and temperature was estimated with a chromel-alumel thermocouple spot-welded to the center of the back of the Ta plate. The Ta clamps made contact with the edges of the Au-coated mica surface so that the sample could be connected to ground. There was no line-of-sight between the sample and the Mo, in order to avoid deposition of MoO3 during heating in O2 (see below). For experiments under ambient conditions, which did not involve high-temperature annealing, the metallic substrate was an Au film vapor-deposited onto a glass substrate with a thin Cr adhesion layer at the interface, henceforth termed “Au/SiO2”. Some experiments were also done for a similar Pt film with a thin Ti adhesion layer, termed “Pt/SiO2”. The UHV studies were done in a conventional stainless steel chamber23 equipped with a single-pass cylindrical mirror analyzer for Auger electron spectroscopy. The base pressure after overnight baking at ∼150 °C was about 1 × 10-10 Torr. Auger spectra were recorded in first-derivative mode with a 4 µA, 3 keV primary beam incident normal to the surface and a 2 eV peak-to-peak modulation amplitude. The chamber included differentially pumped KBr windows for IR spectroscopy and suitable gas-dosing facilities. A quadrupole mass spectrometer was also available for residual gas analysis (RGA). The section of the chamber used for IR studies included a separate pumping system and could be isolated from the rest of the chamber via a gate valve for experiments in background gas pressures of up to ∼1 atm. High-purity H2 (99.9995%), O2 (99.995%), and pyridine (99.9%) were used as received. The pyridine was stored in a glass bulb and outgassed by repeated freeze-pump-thaw cycles (with freezing in liquid N2). The purity of all reagents was checked using the IR transmission spectrum through the chamber (30 cm path length) when filled with the vapor. The H2 and O2 were also checked by RGA. Gases were admitted through standard UHV leak valves, and pressures were measured with cold-cathode ionization, thermocouple, or capacitance gauges, depending on the range. Hot-filament gauges were not used in the case of pyridine, in order to prevent excitation and/or decomposition of the molecule. For H-atom exposure, a W filament resistively heated to about 1700-1800 °C and located about 5 cm from the sample was used to dissociate a fraction of the H2. These will be termed “H2* exposures” and will be given in langmuirs (L), where 1 L ) 1 × 10-6 Torr s, based on uncorrected ionization gauge readings of the H2 pressure. The H2 cracking efficiency is about 2.5% under these conditions.24 The sample was biased at -11 V to repel low-energy electrons from the hot filament, which could lead to electronstimulated desorption of H. During a typical exposure lasting 103 s, the sample temperature rose to ∼70 °C due to radiative heating by the filament. For NRs lying flat on the metal surface, the Au substrate functions as a buried metal layer (BML) to enhance the IRRAS sensitivity to adsorbates on the NR surface.25 The IR optical properties of the BML structure are essentially those of the metal substrate; hence, the so-called “metal-surface selection rule” is in effect.26 As a result, (21) Altman, E. I.; Colton, R. J. Surf. Sci. 1992, 279, 49. (22) HeatWave Labs, Inc., http://www.cathode.com. (23) Bermudez, V. M. J. Vac. Sci. Technol. A 1998, 16, 2572. (24) Sutoh, A.; Okada, Y.; Ohta, S.; Kawabe, M. Jpn. J. Appl. Phys. 1995, 34, L1379. (25) Bermudez, V. M. J. Vac. Sci. Technol. A 1992, 10, 152. (26) Chabal, Y. J. Surf. Sci. Rep. 1988, 8, 211.

Bermudez and Prokes

Figure 2. SEM of β-Ga2O3 NRs deposited on an Au/SiO2 substrate. (a) A region with a relatively high density of NRs. (b) An individual NR. the only adsorbate normal modes that can be detected are those having a finite projection of the dynamic dipole moment (DDM) along the surface normal, and these can be observed only in p-polarization and at a near-grazing angle of incidence. The polarization imposed on the adsorbate spectrum by the BML then permits the use of polarization modulation (PM), first described by Golden et al.,27 for the reduction of the (isotropic) gas-phase signal when data are recorded in a steady-state pressure of an absorbing reagent. The experiment records ∆R/R ≡ (Rs - Rp)/(Rs + Rp), where Rs (Rp) is the s (p) polarized reflectance. Spectra were recorded for the bare substrate and again during (or after) gas exposure from which δ(∆R/R) ≡ (∆R/R)exp - (∆R/R)bare was obtained, after calibration and correction of ∆R/R as described elsewhere.23 The IRRAS measurements employed a Fourier-transform spectrometer. The PM-IRRAS studies of weak adsorbate modes were done using a “narrow-band” HgxCd1-xTe (MCT-A) detector. The lower limit of ∼650 cm-1 on the accessible range was determined by the transmission of the photoelastic modulator and by the sensitivity of the MCT-A detector. The IR beam was incident on the sample at an angle of 82° with respect to the surface normal with a (4° angular spread. Typically 4000 scans were averaged over a period of ∼35 min at a resolution of 8 cm-1, and 2-fold zero-filling and triangle apodization were applied to the interferograms before transformation. Studies of the much stronger Ga2O3 phonon modes were done in static p-polarization (i.e., without modulation) with the sample mounted in a conventional variable-angle external-reflection accessory located in the spectrometer sample compartment. Here Rp was measured (at an 82° angle of incidence) for the metallic substrate with and without the layer of NRs, from which (δR/R)p was obtained. For Rp ≈ Rs and δRs ) 0, which apply in the case of metals in the IR, δ(∆R/R) ≈ 1/2(δR/R)p. The static experiments used a “wideband” MCT-B detector, with a lower limit of ∼400 cm-1 on the accessible range, and only 500 scans were averaged.

3. Results and Discussion 3.1. Characterization of β-Ga2O3 Nanoribbons. Figure 2 shows scanning electron microscopy (SEM) images of a typical NR sample formed by deposition on Au/SiO2. The NRs mostly lie flat on the substrate, although some appear to be standing on edge or otherwise not in full contact with the substrate. A variety (27) Golden, W. G.; Dunn, D. S.; Overend, J. J. Catal. 1981, 71, 395.

β-Ga2O3 Nanoribbons

Langmuir, Vol. 23, No. 25, 2007 12569

Figure 4. Schematic diagram (not to scale) showing typical growth patterns for β-Ga2O3 NRs (cf. ref 32). Growth in the [001] direction produces NRs bounded by ((100) and ((010) faces. Growth in the (010) direction produces NRs bounded by ((100) and ((101h) faces. The (100) face, which has the lowest surface energy (ref 16), is always the wide face. For the present samples, typical dimensions are 20 µm, respectively, in the x, y, and z directions. Figure 3. Raman spectrum (unpolarized) of β-Ga2O3 NRs (λexc ) 488 nm). The peak at 658 cm-1 corresponds to a pair of modes observed (ref 30) at 651 and 657 cm-1 in bulk single-crystal Raman data.

of sizes and shapes is found, but most can be described as “ribbons” or “plates”. Further discussion of the physical structure of the sample is given below. 3.1.1. Raman Scattering and IR Transmission. The β-Ga2O3 space group28,29 is C2/m (C2h3). The IR-active modes transform as the Au or Bu representation of the C2h factor group and the Raman-active modes as the Ag or Bg representation. Figure 3 shows unpolarized Raman data (λexc ) 488 nm) for an NR sample deposited on an Au/SiO2 substrate. The phonon peaks exhibit only a small shift (3-5 cm-1 to lower energy) in comparison to data for either a bulk single crystal30 or a crystalline powder.31 There is no indication of extraneous features related to phases other than the monoclinic (β) form. For example, the R-Ga2O3 Raman spectrum31 exhibits peaks at 432, 573, and 688 cm-1, none of which is evident in Figure 3. In the IR, Au modes exhibit a DDM parallel to the 2-fold rotational (C2) axis of the unit cell (i.e., the [010] direction) while the DDM of a Bu mode lies in the mirror plane normal to the [010] direction. Figure 4 shows a schematic diagram32 of the most common β-Ga2O3 NR growth patterns (ref 16 and works cited). Figure 2 shows that the NRs lie mostly with the growth direction parallel to the substrate. Hence, in normal-incidence transmission spectra for NRs randomly distributed on a transparent substrate, both Au and Bu transverse-optic (TO) modes can be excited for either growth pattern. Individual Au and Bu single-crystal spectra were computed using the Fresnel relations33 for thin-film transmission at normal incidence and are shown in Figure 5a,b. The optical constants (28) Geller, S. J. Chem. Phys. 1960, 33, 676. (29) Åhman, J.; Svensson, G.; Albertsson, J. Acta Crystallogr. 1996, C52, 1336. (30) Dohy, D.; Lucazeau, G.; Revcolevschi, A. J. Solid State Chem. 1982, 45, 180. (31) Machon, D.; McMillan, P. F.; Xu, B.; Dong, J. Phys. ReV. B 2006, 73, 094125. (32) The present work uses the same crystallographic convention as in refs 28 and 29. The unit cell dimensions are such that a > c > b, and the angles are R ) γ ) 90°, β > 90°. Some other works use different conventions, which leads to a different labeling of the NR growth direction. (33) Azzam, R. M. A.; Bashara, N. M. Ellipsometry and Polarized Light; North-Holland: Amsterdam, 1977; Chapter 4. There is a typographical error in this edition that affects calculations of multilayer transmission. Equation 4.170 should read “T ) 1/S11”.

Figure 5. Normal-incidence transmission spectra for β-Ga2O3. The Au and Bu data were computed using the Lorentz oscillator data for an “annealed” β-Ga2O3 bulk single crystal (ref 34) and were broadened by convolution with a Gaussian function having a 20 cm-1 full-width at half-maximum. The NR data were obtained, using an MCT-B “wide-band” detector, for a thin layer deposited on a KRS-5 window. The NR spectrum was referenced to that of bare KRS-5. The spectra have been displaced vertically for clarity.

were obtained from the Lorentz oscillator parameters34 for an “annealed” bulk single crystal of β-Ga2O3 with ∞ ) 3.57 for the high-frequency dielectric constant.35 A 20 cm-1 Gaussian broadening was applied to the computed spectra so that the line widths approximate those in the experimental data. In the computation of the Au spectrum, the incident radiation is polarized parallel to the C2 axis, which lies parallel to the film surface. The NR modes (Figure 5c) are all within 2 cm-1 of the corresponding bulk modes, except for the lower energy Au (at 452 cm-1), which is ∼7 cm-1 higher for the NRs. Previous studies of the IR transmission36,37 and Raman36-41 spectra of β-Ga2O3 nanostructures disagree as to the direction (34) Vı´llora, E. G.; Morioka, Y.; Atou, T.; Sugawara, T.; Kikuchi, M.; Fukuda, T. Phys. Status Solidi A 2002, 193, 187. (35) Rebien, M.; Henrion, W.; Hong, M.; Mannaerts, J. P.; Fleischer, M. Appl. Phys. Lett. 2002, 81, 250. (36) Geng, B. Y.; Liu, X. W.; Wei, X. W.; Wang, S. W.; Zhang, L. D. Appl. Phys. Lett. 2005, 87, 113101. (37) Rao, R.; Rao, A. M.; Xu, B.; Dong, J.; Sharma, S.; Sunkara, M. K. J. Appl. Phys. 2005, 98, 094312. (38) Chun, H. J.; Choi, Y. S.; Bae, S. Y.; Seo, H. W.; Hong, S. J.; Park, J.; Yang, H. J. Phys. Chem. B 2003, 107, 9042.

12570 Langmuir, Vol. 23, No. 25, 2007

Figure 6. Simulated spectra showing the evolution of (δR/R)p for β-Ga2O3/Au with increasing film thickness. The results are computed, as in Figure 5, for films that are (a) 100, (b) 1000, or (c) 5000 Å thick on an Au substrate in p-polarization at an angle of incidence of 82°. Dashed (solid) lines show Au (Bu) data. The spectra have been displaced vertically for clarity, but the relative intensities are quantitative.

and magnitude of the shift between corresponding nanocrystalline and bulk-crystalline modes, and in some cases substantial shifts (30 cm-1 or more) are observed. The shift may depend on the nanostructure growth direction, and various explanations have been proposed, including mode confinement, symmetry breaking, and/or strain. For GaP nanowires, surface phonon excitation has also been shown42 to produce red-shifts in the Raman spectrum that depend on the dielectric constant of the surrounding medium. In any case, the shifts observed in the present TO and Raman data are much smaller than in some previous studies. 3.1.2. IR Reflectance. We now consider the reflectance spectrum of a layer of NRs deposited on an Au substrate. Figure 6 shows a calculation of (δR/R)p vs β-Ga2O3 layer thickness at an 82° angle of incidence (as in the present experiments) using the Fresnel relations33 for the polarized reflectance of a multilayer structure (air/Ga2O3/Au). The optical constants for β-Ga2O3 were obtained as described above, and those for Au were obtained from the Drude parameters in ref 43. A 20 cm-1 Gaussian broadening was again applied to the calculated spectra. The model assumes an idealized structure having atomically abrupt, smooth, and flat interfaces between homogeneous and isotropic materials, and (δR/R)p was obtained from Rp computed for bare and for film-covered Au. The computed spectra in Figure 6 show the IRRAS results to be expected, under the present experimental conditions, for a hypothetical single-crystal layer lying flat on the Au substrate with the [010] direction either parallel (Au) or perpendicular (Bu) to the surface normal. Recall that, in the present IRRAS experiment, the only observable modes in the thin-film limit are those with a finite DDM component along the surface normal. Referring to Figure 4, one sees that the [010] direction lies (39) Gao, Y. H.; Bando, Y.; Sato, T.; Zhang, Y. F.; Gao, X. Q. Appl. Phys. Lett. 2002, 81, 2267. (40) Dai, L.; Chen, X. L.; Zhang, X. N.; Jin, A. Z.; Zhou, T.; Hu, B. Q.; Zhang, Z. J. Appl. Phys. 2002, 92, 1062. (41) Park, G.-S.; Choi, W.-B.; Kim, J.-M.; Choi, Y. C.; Lee, Y. H.; Lim, C.-B. J. Cryst. Growth 2000, 220, 494. (42) Gupta, R.; Xiong, Q.; Mahan, G. D.; Eklund, P. C. Nano Lett. 2003, 3, 1745. (43) Ordal, M. A.; Bell, R. J.; Alexander, R. W., Jr.; Long, L. L.; Querry, M. R. Appl. Opt. 1985, 24, 4493.

Bermudez and Prokes

perpendicular to the surface normal for either growth mode and that, therefore, only Bu modes should be seen for a layer of 10-nm-thick NRs lying flat with the (100) face in contact with the Au. Schmitz et al.44 prepared a (100)-oriented crystalline film of β-Ga2O3 by thermal oxidation of a CoGa(100) single-crystal substrate and obtained the longitudinal-optic (LO) phonon spectrum using high-resolution electron energy loss spectroscopy (HREELS). Analysis based on dielectric theory yielded TO phonon frequencies of 280, 360, 620, and 690 cm-1. For specular scattering, HREELS is subject to the same metal-surface selection rule as is IRRAS (see, for example, ref 45). In accord with the discussion above, the first three modes are close in energy to TO modes of Bu symmetry seen34 in single-crystal bulk β-Ga2O3 at 275, 370, and 619 cm-1. However, the 690 cm-1 mode does not agree well with any of the Au or Bu TO modes (or with any Raman-active mode, cf. Figure 3), and the HREELS data do not show any feature that correlates with the strong Bu TO mode at 523 cm-1 (Figure 5). The film studied in ref 44 was only one unit-cell thick and, although well-ordered, may not represent a bulk crystal.46 In the thin-film limit, only LO modes are observed in IRRAS, as shown in Figure 6a. For completeness, Figure 6 also shows results for thicker layers. With increasing thickness the TO modes also become apparent. At still higher thickness (Figure 6c), the spectrum is affected by the anomalous dispersion in the real index of refraction,47 and the highest-energy LO mode undergoes a broadening and an apparent shift to higher energy. This progression can be seen easily in the Au results, which show that the thick-film spectrum (Figure 6c) is, qualitatively, a sum of the thin-film TO (Figure 5a) and LO (Figure 6a) spectra with the highest-energy LO mode shifted and broadened relative to the thin-film results. Figure 7 shows (δR/R)p spectra measured for a layer of NRs and calculated (as above) for a thin single-crystal film of β-Ga2O3, both on an Au substrate. The intensity of the 768 cm-1 peak indicates a β-Ga2O3 thickness of ∼50 Å (averaged over the area sampled by the IR beam). Contrary to expectation, a finite absorption intensity is also observed experimentally in (δR/R)s. To test whether the data might be affected by “polarizer leakage” spectra were obtained using two different devices, a wire-grid polarizer on a KRS-5 substrate (Figure 7c) and a Ge Brewsterangle polarizer (not shown). The two (δR/R)s spectra were virtually identical. A nonzero (δR/R)s spectrum can occur only if some of the NRs are physically separated from the metal substrate by a finite distance. The sum of the incident and reflected electromagnetic waves constitutes a standing wave, the amplitude of which at the surface is fixed by the boundary conditions at the air/substrate interface.26 For a metallic or BML substrate at a near-grazing angle of incidence, the p-polarized electric field amplitude (Ep) at the interface is at a maximum, and the s-polarized amplitude (Es) is essentially zero. This is the basis for the metal-surface selection rule. With increasing distance away from the interface and into the ambient, Ep decreases to a node, and Es increases to an anti-node. For bare Au in the mid-IR at an 82° angle of incidence, the first node/anti-node occurs at a distance of ∼2λ along the surface normal (obtained using expressions given in ref 25), where λ is the IR wavelength. An NR can occur in a configuration other than lying-flat and in direct contact with the Au substrate if it rests against or on (44) Schmitz, G.; Gassmann, P.; Franchy, R. J. Appl. Phys. 1998, 83, 2533. (45) Ibach, H. Surf. Sci. 1977, 66, 56. (46) Freysoldt, C.; Rinke, P.; Scheffler, M. Phys. ReV. Lett. 2007, 99, 086101. (47) Ottesen, D. K. J. Electrochem. Soc. 1985, 132, 2250.

β-Ga2O3 Nanoribbons

Figure 7. IRRAS data for β-Ga2O3. (a, b) Calculated results for a hypothetical thin single-crystal layer on a metallic substrate. (c, d) Experimental results for an NR layer on Au. The Au and Bu spectra are those shown in Figure 6a and have been divided by a factor of 3. The experimental data were recorded using an MCT-B “wideband” detector and referenced to the reflectance of a bare Au substrate. All spectra are p-polarized except for trace c, which shows s-polarized data in the form of (δR/R)s. The spectra have been displaced vertically for clarity, and the 37 cm-1 shift of the highest-energy Bu mode between crystal and NRs is indicated. The inset shows a schematic diagram (not to scale) of an NR lying flat, which experiences only a p-polarized standing-wave electric field, and an NR making an angle with the surface, which experiences a smaller p-polarized and a finite s-polarized field.

top of other NRs. Visual evidence of this can be seen in Figure 2. As indicated schematically in Figure 7 (inset), such an NR experiences a reduced Ep and an increased Es relative to a NR lying flat on the metal substrate. Radiation of either polarization, penetrating through these NRs and reflecting specularly from the Au, will then add a transmission-like NR spectrum to the reflectance spectrum due to NRs in direct contact with the metal. This transmission-like spectrum will contain both Au and Bu contributions, and one notes the qualitative similarity between (δR/R)s and the normal-incidence transmission data in Figure 5c. The (δR/R)s spectrum in Figure 7c is a direct measure of the transmission-like contribution since, according to the metalsurface selection rule, NRs lying flat and in direct contact with the Au substrate can make no contribution to (δR/R)s. Figures 7c,d also show that, except for the two highest-energy peaks, the (δR/R)p data receive a significant contribution from the transmission-like spectrum. In the case of the 663 cm-1 the peak, the intensity is about the same in both polarizations, indicating that all the intensity in (δR/R)p arises from transmission. 3.1.3. Chemical Interaction at the Ga2O3/Au Interface. The 446 cm-1 peak in Figure 7d does not correspond to any of the IR-active β-Ga2O3 Bu LO modes (Figure 7b). It is close in energy to an Au TO mode seen at about 452 cm-1 in transmission (Figure 5); hence, it could arise from the transmission-like contribution discussed above. However, the (δR/R)s spectrum in Figure 7c (which is a direct measure of this contribution) shows no prominent feature near this energy. In an effort to determine whether the data were influenced in any way by chemical interaction between the NRs and the metal substrate, IRRAS data were obtained for NRs deposited on a Pt/SiO2 substrate as described above. Pt was selected because, like Au, it does not readily form an oxide or a carbide under

Langmuir, Vol. 23, No. 25, 2007 12571

Figure 8. IRRAS data for NRs deposited on (a) Pt/SiO2 and (b) Au/SiO2 substrates. Trace b is the same as that shown in Figure 7d. The arrow marks the 446 cm-1 peak, which is seen for Au but not for Pt. Trace c shows the ν(CdO) peak due to CO formed in the thermal decomposition of the CH3OH used as the NR suspension medium (see the text). Note the different sensitivity scales in c vs a and b. The spectra have been displaced vertically for clarity.

normal conditions. Figure 8 shows that the NR spectra are essentially identical except for the absence of an obvious 446 cm-1 mode for Pt. A thicker NR deposit on Pt/SiO2 (not shown) also exhibited no peak near 446 cm-1. This suggests that this mode is associated with an Au-O-Ga bond formed by reaction at the NR/Au interface when deposition occurs (see above) at a substrate temperature of ∼180 °C. An HREELS study48 of the adsorption of ozone on Au(111) found an Au-O mode fairly close to this energy, at 395 cm-1. The appearance of a feature due to Au-O-Ga bonding implies that adhesion at the β-Ga2O3/ Au interface, prepared as described here, should be fairly strong. Figure 8c shows a peak at 2084 cm-1 due to the ν(CdO) stretching mode of adsorbed CO formed by decomposition of the CH3OH used to suspend the NRs. No such feature was seen for the Au/SiO2 substrate. Adsorbed CO formed by thermal decomposition of CH3OH on a Pt(110)-(2 × 1) surface was found49 to persist to > 200 °C and to yield an HREELS peak at 2070 cm-1. A “blank” experiment (not shown), in which pure CH3OH was dropped onto a bare Pt/SiO2 substrate under the same conditions as those used in NR deposition, produced the ν(CdO) peak but no obvious IRRAS features in the 400-1000 cm-1 range. Dropping liquid CH3OH onto the hot substrate is apparently somewhat effective in cleaning the surface, as shown by the appearance of chemisorbed CO. In spite of the apparent chemical interaction at the Au/Ga2O3 interface, the integrity of the NRs is not compromised. Figure 3 (obtained for the sample used in recording the spectrum shown in Figures 7d and 8b) gives no indication of NR decomposition. Furthermore, IRRAS data (discussed below) after annealing at 500 °C in UHV show no significant difference in comparison to spectra of the as-deposited samples. Hence, the interaction appears to be confined to the immediate interface and not to involve any significant interdiffusion, a conclusion that is supported by sensor studies50 showing that Au/Ga2O3 contacts are stable up to 800 °C. (48) Saliba, N.; Parker, D. H.; Koel, B. E. Surf. Sci. 1998, 410, 270. (49) Wang, J.; Masel, R. I. Surf. Sci. 1991, 243, 199. (50) Fleischer, M.; Ho¨llbauer L.; Meixner, H. Sens. Actuators B 1994, 18-19, 119.

12572 Langmuir, Vol. 23, No. 25, 2007

Bermudez and Prokes

3.1.4. Red-Shift of the LO Modes. The simplest comparison between bulk (Figure 7b) and NR (Figure 7d) LO modes is for the two highest-energy features, observed at 730 and 768 cm-1 for the NRs. These are relatively uncomplicated by overlap with the transmission-like TO features shown in (δR/R)s. These two NR modes are shifted to lower energy by about 37 cm-1 vs the corresponding bulk Bu features. Shifts of this magnitude occur only for the LO modes and not for the TO modes seen in normalincidence transmission (Figure 5). This indicates that the shift is related to an electrostatic effect rather than to other possible mechanisms noted above (e.g., mode confinement), which should also affect the TO modes. Mahan et al.51 have discussed the effect of the shape (as distinct from the size) of a nanostructure on the phonon modes. In, for example, a semiconductor nanowire both the TO and LO modes are shifted to some degree, relative to those in the bulk material. Furthermore, the present (δR/R)p structure is as sharp as that seen in transmission data; whereas, the actual samples comprise an assortment of different shapes (cf. Figure 2). If shape were having a significant effect on LO frequencies then one would expect the reflectance spectra to exhibit a pronounced broadening. The Lyddane-Sachs-Teller relation describes the effect of electrostatic polarization (i.e., long-range dipole-dipole interaction) on the LO frequencies. For the general case of a non-cubic crystal this gives52 N

∏ i)1

( ) () VLOi

0

)

VTOi

1/2

∞

where the product extends over all N normal modes of a given symmetry (Au or Bu in the present case). In bulk β-Ga2O3, the static and high-frequency dielectric constants are reported to be 0 ) 9.5744 and ∞ ) 3.57.35 In the calculations described above, the complex dielectric constant is obtained from the Lorentz oscillator sum53

˜ (ω) ) ∞ +

4πe2N

fk

µ

ωk2 - ω2 - iγkω

∑k

where fk, ωk, and γk are the experimental34 oscillator strength, resonance frequency, and damping parameters, respectively, and N is the number of oscillators per unit volume of reduced mass µ. 0 does not enter the calculation directly but is instead defined in terms of the experimental quantities as

 0 ) ∞ +

4πe2N

fk

µ

ωk2

∑k

Calculations were done, as above, to observe the effect on (δR/ R)p of a small increase in ∞ which could arise if the optical absorption edge were to occur at a lower energy in the NRs. Increasing ∞ from the bulk value of 3.57 to an arbitrarily chosen value of 4.20 caused a red-shift of 15 cm-1 in the highestfrequency LO modes with no significant effect on the TO modes (not shown). This is in qualitative agreement with experiment, but the computed shift is less than the observed value of ∼37 cm-1. A further red-shift could occur if the NRs were less dense than bulk material. This would decrease N and thus 0, as defined (51) Mahan, G. D.; Gupta, R.; Xiong, Q.; Adu, C. K.; Eklund, P. C. Phys. ReV. B 2003, 68, 073402. (52) Barker, A. S., Jr. Phys. ReV. 1964, 136, A1290. (53) Burns, G. Solid State Physics; Academic: Orlando, FL, 1985; Chapter 13.

above, leading to a further decrease in 0/∞. However, there is no independent evidence at this time for either a smaller optical gap or a lower material density for the NR samples used here vs bulk β-Ga2O3. 3.2. Surface Preparation. Prior to a series of adsorption experiments, the sample was recleaned by first outgassing in UHV at 400 °C until any evolution of gas effectively ceased and then heating at 500 °C in ∼8 × 10-7 Torr of O2 for 1 h followed by cooling to