Article pubs.acs.org/cm
Photoluminescence and Photocurrents of GaS1−xSex Nanobelts
Chan Su Jung,† Kidong Park,† Fazel Shojaei,‡ Jin Young Oh,† Hyung Soon Im,† Jung Ah Lee,† Dong Myung Jang,† Jeunghee Park,*,† NoSung Myoung,§ Chang-Lyoul Lee,§ Jong Woon Lee,⊥ Jae Kyu Song,⊥ and Hong Seok Kang*,∥ †
Department of Chemistry, Korea University, Jochiwon 339-700, Korea Department of Chemistry and Bioactive Material Sciences and Research Institute of Physics and Chemistry, Jeonbuk National University, Jeonju 560-756, Korea § Advanced Photonics Research Institute, Gwangju Institute of Science and Technology, Gwangju 500-712, Korea ⊥ Department of Chemistry, Kyung Hee University, Seoul 130-701 Korea ∥ Department of Nano and Advanced Materials, College of Engineering, Jeonju University, Jeonju 560-759, Korea ‡
S Supporting Information *
ABSTRACT: Two-dimensional layered structures have recently drawn worldwide attention because of their intriguing optical and electrical properties. GaS and GaSe are attractive layered materials owing to their wide band gap. Herein, we synthesized GaS1−xSex belt-type multilayers (nanobelts) with uniform morphology ([21̅1̅0] hexagonal-phase long axis) by a chemical vapor transport method, and investigate their composition-dependent optical and optoelectronic properties. The GaS1−xSex exhibited strong visible-range photoluminescence at 490− 620 nm (2.0−2.5 eV), with a unique composition dependence: longer decay time for the S-rich compositions (x ≤ 0.5). Photocurrent measurements were performed on individual nanobelts by fabricating photodetector devices; higher photocurrents were found for x ≤ 0.5. First-principles calculations predicted that oxygen chemisorption can cause the direct and indirect band gaps of GaS to converge, similar to the band structures of GaSe, and thus enhance the optical properties. On the basis of the band alignment (predicted by calculation) for the Schottky barriers in the metal−semiconductor− metal photodetector, we proposed the origin of the higher photocurrent for GaS than for GaSe.
■
2.1 eV, respectively.11−18 Ab initio calculations have predicted that the band gaps of GaS and GaSe significantly increase with decreasing number of layers.19−24 Our group predicted that (indirect) band gaps of GaS and GaSe monolayers are ca. 3.325 and 3.001 eV, respectively; these predicted values were supported by experiments.24 Extensive works have recently reported the synthesis and photodetector application of fewlayered nanosheets (including monolayers) by mechanical exfoliation or epitaxial growth.24−37 Despite the scientific importance of previous results, comprehensive studies that characterize the exotic electronic structures of GaS and GaSe nanosheets remain lacking. In the present work, we synthesized composition-tuned GaS1−xSex [=(GaS)1−x(GaSe)x; 0 ≤ x ≤ 1] multilayers using the chemical vapor transport (CVT) method. These multilayers consisted of a hexagonal or a hexagonal mixed with a rhombohedral phase, with a uniform belt morphology. Hereafter we referred to as nanobelts or NBs. Alloy phases
INTRODUCTION Graphene-like transition metal dichalcogenide (TMD) layered materials, such as MoS2, MoSe2, WS2, and WSe2, have recently attracted worldwide attention for their intriguing optical and electrical properties that are absent in their bulk counterparts.1−10 One striking feature is the increase of the band gap with decreasing the layer numbers to the limit, following an indirect-to-direct band gap transition. The monolayers have a very high mobility (200 cm2 V−1 s−1 for MoS2) and strong photoluminescence (PL), characteristic that are attributed to their unique band gap structure.2,4 They are promising candidates for a wide range of applications related to optoelectronic devices such as photodetectors, light-emitting diodes, optically driven spintronics, etc. Therefore, exploring new layered materials that could exhibit excellent optical and electrical properties with different band gap energies is quite a challenge. GaS and GaSe are a class of two-dimensional (2D) layered materials. GaS (and GaSe) have layered S−Ga−Ga−S repeating units built by six-membered Ga3S3 rings; adjacent layers are held together by van der Waals interactions. GaS and GaSe bulk crystals have indirect band gaps of 2.5−2.6 and 2.0− © 2016 American Chemical Society
Received: May 24, 2016 Revised: July 29, 2016 Published: July 31, 2016 5811
DOI: 10.1021/acs.chemmater.6b02101 Chem. Mater. 2016, 28, 5811−5820
Article
Chemistry of Materials
are displayed. The GaS peaks corresponded well to β-GaS (JCPDS Card No. 30-0576; P63/mmc, a = 3.587 Å and c = 15.492 Å). GaSe exhibits peaks corresponding to both β-GaSe (JCPDS Card No. 03-065-3508; P63/mmc, a = 3.755 Å and c = 15.94 Å) and γ-GaSe (R3m, a = 3.755 Å and c = 23.92 Å). The peaks of γ-GaSe are marked by a subscript R, e.g., (006)R, (101)R, (110)R. The β and γ phases have the same a parameter and the same d spacing of (002) or (003)R planes (7.97 Å). As x increases, the peaks shift continuously from those of GaS to those of GaSe, accompanying the phase conversion from the β-GaS to a mixture of β- and γ-GaSe phases. The peaks broaden significantly as x approaches 0.5. We resolved the (100), (101), (101)R, and (012)R peaks using the Voigt functions. The intensity of the γ phase peaks is negligible for GaS, but it becomes equal to that of the β-phase peaks (marked in red) in the 0.7 ≤ x ≤ 1 compositions. We estimated the ratio of the β and γ phases based on the area of the (100) and (101)R peaks; the average ratios were 3:1 and 1:1 for 0.1 ≤ x ≤ 0.6 and 0.7 ≤ x ≤ 1, respectively. For GaSe, three polytypes have been found, i.e., two hexagonal (β and ε) and one rhombohedra1 (γ) phases.11,14 The structure of the bulk GaSe is dependent on the synthesis methods; ε−γ mixed phase by melting method and β phase by vapor transport or sublimation methods. We found that the γ phase becomes dominant for GaS and GaSe if the growth temperature was lowered (Supporting Information, Figure S3). Because the incorporation of heteroatoms (S or Se) in lattices induces inevitably the broader distribution of lattice parameters, the peak width of alloy composition becomes larger than that of GaS and GaSe, and maximized at x = 0.5. The composition (x) was determined using Vegard’s law [i.e., d = (1−x) dGaS + x dGaSe] based on the peak position of the end members. Analysis of the β phase (100) and (107)/ (110) and the γ phase (101)R peaks provided the same x value for each sample, which is consistent with the value given by SEM energy-dispersive X-ray spectroscopy (EDX) data (Supporting Information, Figure S2). The linear shift of the resolved (100) peak is illustrated by the green line. In contrast, the (004)/(006)R peak position deviates from the linear guide line. This indicates the nonlinear dependence of the c parameter on x, with a larger value than the linear one (see the plot in Figure S4, Supporting Information). This nonlinear behavior of the lattice constant c agrees with previous research on the alloy film.15,16 We will discuss this result using the firstprinciples DFT calculations later. Figure 2a shows SEM images of the high-density GaS NBs grown on the substrates. The widths ranged from 500 nm to 2 μm. The same morphology was observed for all other compositions. The high-resolution transmission electron microscopy (HRTEM) images of selected GaS and GaSe NBs reveal their single-crystalline nature (Figure 2b,c, respectively). The insets show selected-area electron diffraction patterns (SAED) at the [0001] zone axis, confirming the [21̅1̅0] direction along the long axis. The d spacings between the neighboring (100) planes are 3.1 and 3.2 Å, respectively, for GaS and GaSe, which are consistent with the literature values.14−16 The lattice-resolved and corresponding fast Fourier-transform (FFT) (zone axis = [011̅0]) images and intensity line profiles for the lateral side of the GaSe NB both show the d spacing between the neighboring (002) planes of β phase (or (003) planes of γ phase) to be 8.0 Å. Further TEM analyses revealed that two β and γ phases coexist within the single nanobelt or present individually, as shown in Figure S5
offer a chance of achieving continuous band gap tuning and of modifying localized defect energy levels, which can be critical for achieving high-efficiency optoelectronic devices. In a previous work, we observed that there were no restrictions on the concentrations of the components in GaS1−xSex alloys.24 As an extension of that work, our present goal is to investigate the composition-dependent electronic structures of multilayers, which could be distinctive from those of monolayers, and further to find a composition that possesses usable optoelectronic properties. We observed that GaS1−xSex alloys have strong photoluminescence (PL) in the visible range of 490− 620 nm, despite their intrinsic indirect band gap. Photocurrent measurements were performed on individual NB by fabricating photodetector devices, and the results showed a composition dependence in which the photocurrent was higher for the Srich composition. We performed first-principles density functional theory (DFT) calculations to explain the unique PL spectrum and photocurrents of the GaS1−xSex NBs.
■
RESULTS AND DISCUSSION High-density GaS1−xSex NBs were grown on a large area (∼1 cm2) of the substrates. A total of 11 samples with x = 0.1 step were prepared through CVT. The growth of the NB on Aucoated Si substrates follows a typical vapor−liquid−solid growth mechanism, which makes use of Au nanoparticles as catalysts for growth. The nanosheets having no uniform morphology were produced if H2 gas flow was not used.24 Full-range X-ray diffraction (XRD) patterns were acquired as shown in Figure S1 (Supporting Information). Figure 1 shows (004), (100) (101), (107), and (110) peaks of the hexagonal (H) β phase. The reference peaks of H-phase β-GaS (top), βGaSe (bottom), and rhombohedral (R) phase γ-GaSe (bottom)
Figure 1. XRD patterns of as-grown GaS1−xSex NBs: hexagonal (β) phase (004), (100), (101), (107), and (110) peak regions. The reference peaks of β-GaS (top), γ-GaS (top), β-GaSe (bottom), and γGaSe (bottom) are also displayed. 5812
DOI: 10.1021/acs.chemmater.6b02101 Chem. Mater. 2016, 28, 5811−5820
Article
Chemistry of Materials
Figure 2. (a) SEM images of high-density GaS NBs homogeneously grown on a substrate. HRTEM images and SAED patterns of selected (b) GaS and (c) GaSe NBs at the zone axis of [0001] and [011̅0], showing their single-crystalline natures and the [21̅1̅0] growth directions of the NBs. The lattice-resolved and FFT images, as well as the intensity line profile of the lateral side, show that the (002)H (or (003)R) planes of GaSe are separated by a distance of 8.0 Å. (d) EDX mapping and line-scan profiles (with STEM images) of GaS, GaS0.5Se0.5, and GaSe NBs.
(Supporting Information). The EDX spectra [with scanning transmission electron microscopy (STEM) images] show the homogeneous compositional distributions of Ga, S, and Se, providing ratios of Ga:S = 1:1, Ga:S:Se = 1:0.5:0.5, and Ga:Se = 1:1 throughout the NBs (Figure 2d). X-ray photoelectron spectroscopy (XPS) was employed to examine the electronic states of GaS1−xSex NBs as a function of the composition. The survey spectrum is displayed in the Supporting Information, Figure S6. The fine-scan Ga 3p1/2 peak for the GaS and GaSe NBs are blue-shifted by ∼1 eV from that of the neutral Ga (104 eV), as shown in Figure 3a. We resolved the peak using Voigt functions; Ga−S (or Ga−Se) bonding structure at 105 eV and Ga−O bonding structure at 108.7 eV. The Ga 3p3/2 peak (blue-shifted by ∼1 eV from neural one at 107 eV) was not resolved due to the overlap with the Ga−O bonding peak. The fine-scan S 2s and Se 3s peaks were resolved using Voigt functions, and the results show that the ratio of S and Se peaks matches the composition x of GaS1−xSex (Figure 3b). The S 2s and Se 3s peaks are red-shifted by 2.0 and 2.7 eV, respectively, from those of the neutral S (228 eV) and Se (232 eV), corresponding to the Ga−S and Ga−Se bonding structures. GaS and GaS0.7Se0.3 show a peak at higher energies, corresponding to the S−O bonding structures. In contrast, GaS0.5Se0.5, GaS0.3Se0.7, and GaSe show negligible peaks at
Figure 3. Fine-scan XPS spectrum of (a) Ga 3p peak, and (b) S 2s and Se 3s peaks of GaS, GaS0.7Se0.3, GaS0.5Se0.5, GaS0.3Se0.7, and GaSe NBs. The data points were resolved into bands (colored lines) using a Voigt function. The black line represents the sum of the resolved bands.
5813
DOI: 10.1021/acs.chemmater.6b02101 Chem. Mater. 2016, 28, 5811−5820
Article
Chemistry of Materials higher energies. The smaller red shift (0.7 eV) of the S 2s peak relative to that of the Se 3s peak may be related to the preferential binding of S with adsorbed oxygen. We therefore conclude that oxygen binding is more significant in the S-rich compositions because S can bind more strongly with oxygen than Se. The PL spectra were measured by delivering continuouswave excitation from a 325 nm He−Cd laser to the individual GaS1−xSex NBs at room temperature (Figure 4a). The emission
We measured the PL decay curve of the as-grown GaS1−xSex NBs (at 8 K), and the results are shown in Figure 4b. The decay curves were fitted with triple exponential decay functions. The average decay time ⟨τ⟩ was calculated using the equation ⟨τ⟩ = Σi f iτi, where f i is the amplitude fraction and τi is the decay time, yielding ⟨τ⟩, as shown in Table S1 (Supporting Information). The average decay times of GaS and GaSe were 1.53 ± 0.15 and 0.24 ± 0.02 ns, respectively. In the range of 0 ≤ x ≤ 0.5, ⟨τ⟩ = 1.4 ± 0.1 ns for x = 0.3 and 1.5 ± 0.1 ns for x = 0.5, showing values similar to those of GaS. However, ⟨τ⟩ = 0.38 ± 0.04 ns for x = 0.7. The incorporation of the heteroatom Se reduces the decay time after x = 0.5; the shortest decay time thus occurs in GaSe. We then examined the photocurrents of the GaS1−xSex NBs by fabricating electrodes. The SEM image (Figure 5a) shows a
Figure 4. (a) PL spectra (with photographs) of the individual GaS1−xSex NBs at various values of x. (b) PL decay profiles of the GaS1−xSex NBs monitored at the emission peak for x = 0, 0.3, 0.5, 0.7, and 1. The excitation photon energy is 3.49 eV (355 nm), obtained from frequency-doubled Ti−sapphire laser pulses. The data points are fitted with triple exponential function (shown as the lines).
Figure 5. (a) SEM images of NB aligned between Ti/Au bottom electrodes with a 5 μm gap; the top Pt electrodes were deposited using focused-ion beam (FIB). I−V characteristics of the GaS1−xSex NBs, with x = 0, 0.3, 0.5, 0.7, and 1, (b) under dark conditions and (c) 325 nm He−Ne laser irradiation at 1 mW. (d) I−t curves at a bias voltage of 2 V under on/off cycled irradiation. (e) Photocurrents of GaS1−xSex as a function of x.
intensity measured from the individual GaS1−xSex NB is nearly the same for all compositions. Photographs on the top correspond to the optical microscopy images. They show the color change of the strong PL emissions; blue (GaS), green (x = 0.3), yellow (x = 0.5), yellowish orange (x = 0.7), and red (GaSe). A contact printing method applied to the transfer of NBs to a Si substrate having 300 nm-thick thermally oxidized layers; the as-grown NWs on the Si substrate were rubbed onto a receiver substrate under an applied load. Therefore, most of NBs becomes broken and shorten after the transfer. The images were taken to show well the color of the PL emission by selecting the NBs with a wider width (1−2 μm). The PL peaks shift continuously as x changes, with no drastic changes in position or shape. The PL peak position, corresponding to the band gap (Eg), shows linear dependence on the composition between 490 nm (GaS) and 620 nm (GaSe). The average PL spectrum (at 8 K) measured from the as-grown NBs on substrates showed the same peak position as that of the individual NB (see Figure S7, Supporting Information). As x increases, Eg of GaS1−xSex decreases linearly (with no bowing effect) from 2.5 to 2.0 eV, which is consistent with that of film.11,13
NB aligned between the Au bottom electrodes, with a gap of 5 μm patterned on the SiOx substrate. We selected 1.5−2 μm width NB (having a thickness is 15−20 nm) that lies between the electrodes, keeping 5 μm path-length. The power density of the focused He−Cd laser (325 nm) beam was 1.3 kW/cm2; laser beam diameter = 10 μm, laser power = 1 mW. Figure 5b shows the current−voltage (I−V) curves of the device using GaS1−xSex NBs with x = 0, 0.3, 0.5, 0.7, and 1 under dark conditions. The I−V curves are nonlinear over the measured range of −2 to 2 V, indicating the Schottky behavior of the GaS1−xSex/Au electrode contact. GaS exhibited the highest current (0.4 nA) at a bias voltage of 2 V. Figure 5c shows the I−V curves under light irradiation at 325 nm (3.8 eV). The I−V curves are almost linear within the measured range of −2 to 2 V. I−t curves at a bias voltage of 2 V were collected in real time for a series of on/off illumination cycles (Figure 5d). The current in the device instantly increased (t < 0.1 s) when the light was turned on and decreased when the light was turned off. The photocurrent (ΔI), i.e., the current increase under 5814
DOI: 10.1021/acs.chemmater.6b02101 Chem. Mater. 2016, 28, 5811−5820
Article
Chemistry of Materials Table 1. Band Gap and Band Position (versus Vacuum Level) of GaS and GaSe Trilayers, Calculated Using HSE-06 Functionalsa Eg(indirect)b
Eg(direct)c
ΔEgd
E Ce
EFf
ECg
β-GaS
2.654
3.133
0.479
β-GaSe
2.173
2.181
0.008
γ-GaSe
2.199
2.204
0.005
−6.47 (−5.87)h −5.88 (−5.37) −5.81
−5.14 (−4.91) −4.79 (−4.67) −4.71
−3.81 (−3.95) −3.70 (−3.96) −3.61
All energies in eV. b((K-Γ)→Γ) transition. c(Γ→Γ) transition. dΔEg = Eg(direct) − Eg(indirect); energy difference between the direct and indirect band gaps. eConduction band minimum (CBM) level. fFermi level. gValence band maximum (VBM) level. hCalculated using PBE-D2 functionals. a
Table 2. Band Gaps of Pristine and O-Chemisorbed GaS and GaSe with Various Number of Layers NL, Calculated Using PBED2 Functionalsa Pristine b
c
O-chemisorbed ΔEgd
NL
indirect
1
2.580
2.881
0.301
2
2.129
2.364
0.235
3 1
1.924 (1.627)g 2.202
2.097 (1.824)g 2.309
0.173 (0.197)g 0.107
2
1.610
1.663
0.052
β-
3
γ-
3
1.383 (0.998)g 1.420
1.412 (0.998)g 1.445
0.029 (0)g 0.025
GaS
GaSe
direct
cell size
indirect
× × × × × × × × × × × × × ×
2.400 2.456 2.024 2.066 1.851 1.895 2.165 2.204 1.655 1.640 1.414 1.410 1.417 1.421
4 6 4 6 4 6 4 6 4 6 4 6 4 6
4 6 4 6 4 6 4 6 4 6 4 6 4 6
e
direct
ΔEgd
Eadf
2.450 2.460 2.061 2.097 1.858 1.901 2.255 2.206 1.703 1.693 1.439 1.436 1.441 1.446
0.050 0.004 0.037 0.031 0.007 0.006 0.090 0.002 0.048 0.053 0.025 0.026 0.024 0.025
−0.17 −0.23 −0.21 −0.32 −0.19 −0.30 −0.08 −0.16 −0.16 −0.24 −0.16 0.30 −0.11 −0.17
a All energies in eV. b((K-Γ)→Γ) transition. c(Γ→Γ) transition. dΔEg = Eg(direct) − Eg(indirect); energy difference between the direct and indirect band gaps. e(K→Γ) or ((K-Γ)→Γ) transition. fO chemisorption energy: total energy change from the chemisorption of an O atom onto a given cell. g Value of the bulk.
illumination, was ∼7 nA for x = 0, 0.3, and 0.5; ∼ 1 nA for x = 0.7; ∼ 0.5 nA x = 1. Responsivity, defined as ΔI/light intensity, at 2 V was calculated as ∼7 μA/W for x = 0−0.5, whereas that was ∼1 and ∼0.5 μA/W for x = 0.7 and 1, respectively. Figure 5e shows the photocurrents of the GaS1−xSex photodetector devices as a function of composition (x). The S-rich NBs exhibit higher photocurrents than the Se-rich NBs; the photocurrents at 0 ≤ x ≤ 0.5 are almost 10 times higher than those of 0.7 ≤ x ≤ 1. The characteristics of GaS and GaSe photodetectors in previous works of other research groups are summarized in Table S2 (Supporting Information).25−37 Recent studies on the photocurrents of Mo(S 1−x Se x ) 2 monolayers showed significant decreases in photocurrent for Se-rich alloys as compared to S-rich ones, consistent with our results.38 To the best of our knowledge, no studies have reported the composition-dependent photocurrent of GaS1−xSex. The first question that arises is whether S and Se mixed homogeneously within the layers of the alloys. Because the lattice constant c was not linearly dependent on x, we attempted to determine whether this deviation was related to inhomogeneous mixing. We calculated the electronic structures and band gaps of a GaS0.5Se0.5 bilayer using the HSE06 and PBE-D2 functionals (see Supporting Information, Table S3). Two bilayer structures were tested: (1) mixed S and Se layers of (S,Se)−Ga−Ga−(S,Se) with two configurations and (2) separated S and Se heterolayers of S−Ga−Ga−S and Se−Ga− Ga−Se. Configuration 1 was more stable than the heterolayer
configuration 2 by 0.019 eV/atom. The structural parameters and the geometries are shown in Table S4 (Supporting Information). The band gap of configuration 1 is closest to the linear value. Because the experimental band gap changes linearly with x, S and Se are assumed to be distributed homogeneously in the alloy structures. The lattice constant a of configuration 1 is nearly equal to the linear value. We found that the lattice constant c of the bulk crystals (as calculated separately) exceeds the linear value. The Se−Se or S−Se bond distance, which is greater than the S−S bond distance because of the larger radius of Se, is primary reason for the increase in lattice parameters along the c-axis. Therefore, we conclude that S and Se mixed homogeneously within the layers. The band gap (Eg) of GaS and GaSe NBs is 2.5 and 2.0 eV, respectively; these values are same as the indirect band gap of the bulk materials.11−13,17,18 Furthermore, we reported the firstprinciples calculations on GaS and GaSe using the HSE-06 functionals.24 The Eg values of GaS and GaSe trilayers calculated using HSE-06 functionals are listed in Table 1. For GaSe, β and γ phases were considered, based on our experimental data. The indirect band gaps of GaS and βGaSe (and γ-GaSe) trilayers are 2.654 and 2.173 (2.199) eV, respectively, close to the values measured in the NBs (2.5 and 2.0 eV, respectively). β- and γ-GaSe have similar band gaps, explaining the linear change in the band gaps of the GaS1−xSex NBs. Surprisingly, the GaS1−xSex NBs exhibited a strong PL spectrum for all compositions, despite their intrinsic indirect 5815
DOI: 10.1021/acs.chemmater.6b02101 Chem. Mater. 2016, 28, 5811−5820
Article
Chemistry of Materials
for the γ phase (−0.17 eV). β-phase GaSe has the same Ead value as GaS. The different adsorption behavior of the β and γ phases can be understood by considering that the β phase is more compact than the γ phase, with a shorter interlayer distance. Chemisorption of O increases the shortest interlayer Se−Se distance from 3.78 to3.84 Å for the β phase, while decreasing it negligibly, from 3.83 to 3.82 Å, in the γ phase. This indicates that the β phase compensates for electrostatic repulsion between layers, which are presumably positively charged after adsorbing O, by increasing the interlayer distance. Therefore, we conclude that O chemisorption is more favorable in the β-phase-pure GaS, considering that GaSe consists of a 1:1 mixture of β and γ phases. The more favorable O binding of GaS is supported by the XPS data. O−GaS, in practical terms, has a direct band gap for all layers; the indirect gap [K→Γ or (K-Γ)→Γ transition] converges to the direct band gap at the Γ point, with an energy difference of less than 0.01 eV. On the other hand, O chemisorption has only a marginal effect on the band structures of GaSe; the direct/indirect band gaps and their energy differences are equal to those of pristine GaSe. The band gap difference of O−GaS becomes even smaller than that of O− GaSe. The O-chemisorption can reduce the (indirect) band gap for GaS, but not for GaSe. However, the reduction effect becomes smaller as the number of layers increases; i.e., 0.124 eV for 6 × 6 monolayers, 0.063 eV for 6 × 6 bilayers, and 0.029 eV for 6 × 6 trilayers. It implies that the band gap of the oxygenated multilayers (the present NB) is similar to that of the pristine one. This result is consistent with the PL data, which do not show any appreciable deviation from the band gap calculated by HSE06 functionals. On the basis of these calculations, we suggest that the O chemisorption of GaS and S-rich composition alloy enhances the PL emission and the photocurrents, because of the converged direct band gap transition. One important result is the composition dependence of the PL decay time; the incorporation of Se heteroatoms with x > 0.5 into GaS significantly reduces the decay time. The composition-dependent photocurrents correlate well with the decay times of PL emission. Once excitons are created by light absorption, electron and holes are separated, resulting in a current flow toward the Au electrodes. The increased exciton lifetimes or longer recombination time of electron−hole pairs could create higher photocurrents. The question is why the decay times decrease for compositions of x > 0.5. One possible model is that the larger-portion of γ phases may be detrimental to the separation and transfer of photogenerated electrons and holes, which decreases the photocurrents in these Se-rich alloys. We measured the PL decay curves for the commercially available GaS and GaSe powders (bulk), and γ-phase-rich NBs (Supporting Information, Figures S9 and S10). For both cases, the GaS exhibits always the longer PL decay time than the GaSe, and the average decay time is not much dependent on the phase. Therefore, more studies need to explain the composition dependence of the PL decay times. To explain the photocurrents decrease for GaSe, we propose a plausible model based on the characteristics of “metal− semiconductor−metal (MSM) photodetectors”, where photocurrent is mainly generated from the photoexcited carriers at the Schottky contact with electric potential. As shown in Figure 5b, the I−V curves under dark conditions exhibited a Schottky behavior.45 Table 1 summarizes the position of the valence band maxima (VBM) and the conduction band minima (CBM)
band gaps. The calculation showed that direct band gap of the GaS trilayers lies above the indirect band gap by a difference of 0.479 eV (HSE06). In contrast, the direct band gap of the GaSe trilayers lies near the indirect band gap (0.008 and 0.005 eV above for β and γ phase). If the energy difference between the two k-points of the valence band (VB), corresponding to the direct and indirect transitions, is within the thermal energy range, the direct transition would be significantly enhanced. GaSe is thus expected to show enhanced optical properties (i.e., higher PL intensity) as compared to GaS, which has a larger direct/indirect band gap difference. However, these band structures do not properly explain the similar PL intensities of GaS and GaSe NBs. We considered a model based on the aforementioned XPS data showing the presence of the O binding structures and the favorable O binding of S compared to Se. Studies on TMDs have reported that the adsorption of O2 and/or H2O molecules enhances the PL emission of n-type MoS2 and MoSe2 monolayers.39,40 Recently, Oh et al. showed that oxygen physisorption of MoS 2 can enhance the PL, whereas chemisorption reduces it.41 Using DFT calculations, Liu et al. predicted that the oxygen (O)-chemisorption barrier would be lower for GaS and GaSe than for MoS2 or MoSe2.42 In our previous work, we suggested that the enhanced cathodoluminescence emission of GaS monolayers is due to the O chemisorption.24 Here, the effects of O2 adsorption on the electronic structures of multilayers were therefore investigated by calculating the O-chemisorption energy (Ead) of one-half of an O2 molecule (i.e., one O atom) onto 4 × 4 and 6 × 6 cells of GaS and GaSe trilayers (with mono- and bilayers for comparison) by PBE-D2 functionals, as shown in Table 2. The band structure is shown in the Supporting Information, Figure S8. O2 was assumed to initially physically adsorb and migrate as O atoms to different S or Se sites. The most stable configuration was an O atom bridging the Ga and S (or Se) atoms.24 The PBE-D2 functional calculations predicts the band gaps of pristine trilayers and bulk to be smaller than the experimental values; 1.924 and 1.627 eV for GaS trilayers and bulk, respectively; 1.383 and 0.998 eV for β-GaSe trilayers and bulk, respectively. The direct band gap of GaS lies above the indirect band gap by a difference of ∼0.2 eV. The direct band gap of GaSe lies near the indirect band gap. Therefore, the band gap features are consistent with those that predicted by HSE06 calculation. Importantly, the O chemisorption is appreciably exothermic (i.e., Ead < 0), and the corresponding energy release generally increases with increasing layer numbers (NL). O chemisorption is more energetically favorable on the 6 × 6 cell than on the 4 × 4 cells, indicating that O coverage will not exceed that of the latter cells. For instance, Ead = −0.19 and −0.30 eV for 4 × 4 and 6 × 6 cells of GaS trilayers, respectively. For the O-chemisorbed GaS (OGaS) structures, the SO bond length (1.60 Å) is longer than both the SO double bond (1.43 Å) and the partial double bond (1.49 Å).43 The OGa bond length (1.92 Å) is comparable to that of a partial single bond with a bond order of 0.35 (1.93 Å).44 In the O GaSe structures, the SeO and OGa bond lengths are 1.77 and 1.91 Å, respectively, showing bond orders similar to those of OGaS. These bond lengths are nearly equal for all layers, implying that the difference in Ead mostly originates from interlayer interactions. The Ead value of the GaSe trilayer depends on the crystal phase: it is appreciably larger for the β phase (−0.30 eV) than 5816
DOI: 10.1021/acs.chemmater.6b02101 Chem. Mater. 2016, 28, 5811−5820
Article
Chemistry of Materials
height ΦBn(m→s) for electrons from the metal to the semiconductor, as well as the built-in potential Φnbi(s→m) for electrons from the semiconductor to metal, can be calculated from the relations ΦBn = χ(s) − EF(m) and Φnbi= ΦBn − [EC(s) − E F (s)]. 46 Here, χ(s) denotes the electron affinity corresponding to EC. The barrier heights (HSE06), ΦBn, are 1.45 (1.31) and 1.56 (1.30) eV for GaS and GaSe, respectively. The numbers in the parentheses denote values calculated from the PBE-D2 calculations, respectively. On the other hand, the built-in potentials, Φnbi, are 0.12 (0.35) and 0.47 (0.59) eV, respectively. In the figures, we denote Φnbi(s→m) as Φbi. Figure 6c depicts the energy diagram showing the photocurrent from the GaS (or GaSe) to the Au electrodes at a bias voltage; the right and the left electrodes are positively and negatively biased, respectively. We suggest that the dominant mechanism for the photocurrent is a combination of thermionic emission and thermally assisted tunneling of photoelectrons, produced in the MSM device that consisted of a lightly doped semiconductor. The higher photocurrent of GaS would be ascribed to the lower Φnbi barrier (0.12 eV) toward the positively biased Au electrode (right side electrode), compared to that of GaSe (0.47 eV). GaS and GaSe are known to be ntype and p-type semiconductors, respectively.25 In that case, the Φnbi value for GaS decreases, whereas that for GaSe increases. The current from n-type GaS to the positive electrode will be larger than that for the intrinsic one because the fraction of conduction electrons with thermal energies larger than Φnbi is proportional to exp(−Φnbi/kT) at a bias voltage. The opposite argument also holds for GaSe. Therefore, the difference in photocurrent between the two materials will be more pronounced by doping. For bulk GaSe, the hole concentration (Na) was reported to be 3 × 1016 m−3, and the effective density of state for holes (NV) was estimated to be 1025 m−3.47 Then, EF − EV = kT ln(NV/Na) = 0.51 eV at 300 K and EF = −4.86 eV, indicating that EF(m) < EF(s) still holds, and thus our energy band diagram based on the n-type Schottky contact remains valid. The photocurrent of our photodetectors shows remarkably an almost linear I−V curves. The previous studies on GaSe photodetectors also reported the linear I−V curves for photocurrents.26−28,30 This observation can be explained in terms of the metal-induced gap states (MIGSs).48 Under light irradiation, photogenerated holes drifted toward the negatively biased Au electrode (left side electrode) are trapped in the MIGSs. There will be appreciable number of those states in the 2D nanostructured materials (nanobelts) because of their large surface-to-volume ratio. Owing to the increased positive charge density in the depletion region of negatively biased contact, its Schottky barrier ΦBn lowers and the space charge region shrinks. Consequently, both of the thermionic emission and the thermally assisted tunneling of electrons from the left electrode contributes in increasing the photocurrent for low applied voltages.
(versus vacuum level) of the trilayers calculated using HSE06 functionals. For comparison, we calculated the position using PBE-D2 functionals (see the value in parentheses). Figure 6
Figure 6. Energy diagram for the Schottky contact between the metal (Au) electrode and semiconductor (GaS or GaSe), calculated using HSE06 functionals, (a) before and after the contact is made at (b) zero bias and (c) applied bias under light irradiation. The numbers in parentheses denote values calculated from PBE-D2 calculations. Solid and dotted line arrows in panel c represent the thermionic and tunneling electron flows, respectively.
shows the band energy diagram for the Schottky contact between the Au electrode and the GaS (or GaSe) trilayers before and after contact (with electric field) is made using the calculated band position. We assume that the work function of the Au electrode is the same as that of the Au(111) surface, 5.26 eV.46 Because the work function of the low-indexed Au surface is almost the same, our assessment will not be altered. For isolated GaS (or GaSe) trilayers, the position of the VBM and CBM bands, EC and EV, are shown in Figure 6a. For simplicity, only the β phase is considered for GaSe trilayers. Assuming that GaS (or GaSe) is moderately doped, its Fermi level can be approximated by the intrinsic Fermi levels (EF), corresponding to the midgap of the VBM and CBM, EF= 1/2 (EC + EV). Later, we will show that the shift of the Fermi level due to unintentional doping does not alter our conclusion. As explained above, O chemisorption of the trilayers does not alter their band gap. Our separate calculation showed that O chemisorption lowers the CBM of the GaS by only 0.02 eV, which mean that but it does not alter the energy levels of the edge bands (VBM and CBM), unless the O concentration is much higher than one atom per 6 × 6 cell, which is actually unfavorable. The lowering of the CBM level will be even smaller for the higher numbered layers, implying that the edge band energy levels of the multilayers (the present NBs) are similar to those of the oxygenated ones. Figure 6b the energy level diagrams for the present MSM photodetector at zero bias. Both GaS and GaSe form a Schottky contact of the n-type semiconductors with the Au electrode because EF(m) < EF(s), where m and s stand for the metal and the semiconductor of Fermi level EF, respectively. The barrier
■
CONCLUSIONS GaS1−xSex NBs were synthesized with band gap tuning (2.0− 2.5 eV) throughout the compositional control by CVT. The NBs grew along a constant [21̅1̅0] direction. Although the GaS NBs contained only the β phase, the portion of the γ phase increased with increasing x; the two phases were equally prevalent in GaSe. A strong visible-range PL emission was observed for all composition, with longer decay times for S-rich compositions (x ≤ 0.5). We fabricated photodetector devices 5817
DOI: 10.1021/acs.chemmater.6b02101 Chem. Mater. 2016, 28, 5811−5820
Article
Chemistry of Materials
photomultiplier (PMA-182-P-M, PicoQuant), and recorded using a TCSPC module (PicoHarp300, PicoQuant). The PL decays were analyzed by the multiexponential model given by exp(−t/τi). Fabrication of Photodetectors and Measurement of I−V Curves. Photolithography was used to deposit the Ti (20 nm)/Au (80 nm) electrode structure onto a Si substrate with a 1 μm-thick thermally grown SiO2 layer by sputtering using a patterned mask. The Ti film underneath the Au film was just used for the better adhesion of Au film on the substrate. The gap distance between the electrodes was 5 μm. The NBs were dispersed in isopropyl alcohol and deposited on the patterned electrodes. In the focused ion beam (FIB) chamber, we switched carefully between the secondary electron imaging mode and the ion-beam deposition mode in the dual-beam FIB system during fabrication. By doing so, the Ga ion beam-induced Pt deposition was limited to the contact regions to avoid possible contamination over the entire NB. The length between the Pt electrode pads was approximately 10 μm. The thickness of the Pt deposition pads was 80 nm. We selected 1.5−2 μm width NB that aligned between two Au electrodes with an angle of