Electronic and Defect Structures of CuSCN - American Chemical Society

Apr 28, 2010 - of Cu 3d levels hybridized with S 3p states, the conduction band minimum (at the K ..... conduction bands; solid symbols are valence ba...
0 downloads 0 Views 2MB Size
J. Phys. Chem. C 2010, 114, 9111–9117

9111

Electronic and Defect Structures of CuSCN John E. Jaffe,* Tiffany C. Kaspar,* Timothy C. Droubay, Tamas Varga, Mark E. Bowden, and Gregory J. Exarhos Pacific Northwest National Laboratory, P.O. Box 999, Richland, Washington 99352 ReceiVed: February 22, 2010; ReVised Manuscript ReceiVed: March 30, 2010

Copper thiocyanate (CuSCN) is a candidate as a transparent solid p-type conductor for optoelectronic and photovoltaic applications, such as solar cells. We calculate the band structure, bonding characteristics, and basic native defect configurations of hexagonal β-CuSCN. β-CuSCN is predicted to be an indirect-gap semiconductor with an unusual orbital character: although the highest valence bands have the expected character of Cu 3d levels hybridized with S 3p states, the conduction band minimum (at the K point of the hexagonal Brillouin zone) has mostly cyanide antibonding character. This quasi-molecular character results in some unusual properties, including that the electron effective masses are comparable to or even larger than the hole effective masses. Calculated results match well with the valence band spectrum of thin film CuSCN, although optical absorption measurements do not conclusively confirm the predicted indirect nature of the lowest transitions. The dominant p-type character of this material is explained in terms of copper vacancies; CN unit vacancies, which are also expected to be acceptors, are proposed as a mechanism to increase p-type conduction. 1. Introduction Transparent conductors are wide-band-gap semiconductors that, at appropriate carrier doping levels, combine high electrical conductivity with high transparency to visible light.1,2 These conductors have various applications in optoelectronic devices, such as solar cells and light-emitting diodes: they can serve as electrical contacts on the boundary of a device where light enters or is emitted and can also be an active element in devices, such as transparent transistors. One factor presently limiting the use of transparent conductors is that almost all are n-type materials; only a few p-type materials are known.3 Being able to freely employ both n- and p-type materials in transparent active devices is thus not yet practical. One reason for this situation is that most transparent conductors are oxides, and the dominant native defects in oxides are usually oxygen vacancies (or sometimes metal atom interstitials) that are donors. Thus, it is important to understand the origin of p-type conductivity in the few widegap materials where it occurs so as to be able to enhance the performance of these materials or design even better performers. One interesting candidate compound for a p-type transparent conductor is copper thiocyanate, CuSCN.4 CuSCN has recently found wide application as a solid hole-transporting electrolyte in dye-sensitized5,6 and other nanostructured7,8 solar cells. Its intrinsic acceptors are believed to be associated with copper deficiency in its composition,9,10 but little is known specifically about its native defects or about the nature of electronic states and hole transport in a CuSCN crystal. In the present work, we investigate these issues by first-principles density functional theory (DFT) calculations on CuSCN bulk and defected supercells, with an emphasis on unique properties related to its quasi-molecular structure, specifically, the (SCN)-1 thiocyanate ion, which makes the system a “pseudohalide.” We also provide optical absorption and X-ray photoelectron spectroscopy (XPS) measurements on a polycrystalline film of β-CuSCN. We * To whom correspondence should be addressed. E-mail: john.jaffe@ pnl.gov (J.E.J.), [email protected] (T.C.K.). Phone: 509-375-6382 (J.E.J.), 509-371-6503 (T.C.K.). Fax: 509-375-2644 (J.E.J.), 509-371-6242 (T.C.K.).

attempt to account for the special properties of this material, make comparisons to a few other p-type wide-gap conductors, and suggest how new materials with similar characteristics might be designed. 2. Background on β-CuSCN To begin, it may be helpful to explain why the thiocyanate molecular ion is monovalent, accepting only one electron from a cation (Cu, in this case), despite the fact that separated sulfide S2- and cyanide (CN)-1 anions would have accepted three electrons in an ionic compound. The reason is that there is a covalent single bond between sulfur and carbon in (S-CtN)-1 so that there is a shared electron pair that enables both S and C to attain closed shells with two fewer transferred electrons required than in the case of separated sulfide and cyanide. Using dots to represent electrons, the (CN)-1 ion may be represented by :C:::N:, where the leftmost electron pair is the one used to form the covalent bond with S, the middle three pairs are the CtN triple bond, and the rightmost pair is a lone pair on nitrogen, not shared covalently but forming a weak “dative bond” with copper. The sulfur atoms are also coordinated approximately tetrahedrally to three copper atoms in mixed ionic-covalent (but predominantly ionic) bonding. Thiocyanate is linear, somewhat like CO2 or CS2, but with very unequal bond lengths (S-C ) 1.701 Å; CtN ) 1.156 Å in molecular thiocyanogen11). Two structures are known for CuSCN, the hexagonal12 or rhombohedral13 β-phase and the orthorhombic14 R-phase. In both phases, the Cu atoms are coordinated approximately tetrahedrally by one N and three S atoms and the S by one C and three Cu atoms. However, in the β-phase, the SCN units and N · · · Cu bonds are strictly collinear, whereas there is a slight bend at the N atom in the R-phase. Moreover, the SCN units are all parallel to each other along the c axis in β-CuSCN but are canted in an alternating fashion in R-CuSCN. There are two stacking polytypes of β-CuSCN, depending on whether the Cu-S bonds at the two ends of an SCN · · · Cu unit are staggered (3R rhombohedral) or eclipsed (2H hexagonal polytype). Here,

10.1021/jp101586q  2010 American Chemical Society Published on Web 04/28/2010

9112

J. Phys. Chem. C, Vol. 114, No. 19, 2010

Figure 1. Hexagonal β-CuSCN structure, a 3 × 3 × 1 supercell.

we consider the hexagonal polytype only, with lattice parameters of a ) 3.850 Å and c ) 10.938 Å. This structure is a close analog of the wurtzite structure, with the same symmetry but with a rather extreme c/a ratio of 2.841 (as opposed to, for example, 1.603 in ZnO) and with three independent anion displacement parameters uS ) 0.433, uC ) 0.277, and uN ) 0.176 (uO ) 0.381 in ZnO), where the u parameter gives the displacement of an anion from the nearest cation along the c axis, in units of the lattice constant c. A sketch of the β-CuSCN structure is shown in Figure 1. CuSCN has been recognized as a p-type wide-gap semiconductor for some time. A relatively high conductivity together with good transparency has been observed, but even higher carrier concentrations would be useful for applications. The conductivity is clearly associated with stoichiometric copper deficiency, but details of the atomic-level defect structure are not yet clear, so it is also not clear how conductivity may be enhanced. Insulating material can be obtained by increasing the Cu content, but n-type conduction in CuSCN requires high levels of Cu excess,9 suggesting that native acceptor defects tend to compensate for any intrinsic or extrinsic donors that may be introduced. Some authors10 have reported a defect level in the band gap of β-CuSCN about 1 eV above the top of the valence band and proposed that this level may be the intrinsic acceptor, but this level would be too deep in the gap (too high in energy) to account for strong room-temperature conduction. We employ first-principles calculations as well as new measurements to improve the understanding of this material; to our knowledge, no electronic structure calculations on β-CuSCN have been reported thus far. 3. Methods 3.1. Computational Approach. The band structure and defect calculations on β-CuSCN were performed with the VASP15-20 density functional code. We represented the wave functions in a projector-augmented-wave basis21 with an energy cutoff of 280 eV and a 9 × 9 × 3 Monkhorst-Pack22 k-space grid; a similar k-space density was used for the smaller Brillouin zone of the supercells described below. The generalized gradient approximation (GGA) density functional of Perdew, Burke, and

Jaffe et al. Enzerhof23 was employed. Geometry optimizations for both the bulk structure and the supercells with defects (vacancies) were carried out until changes in total energy were less than 0.001 eV, corresponding to forces less than about 0.03 eV/Å. Charged defect states were treated by the formal assumption of a neutralizing positive background, with Coulomb and finite-cellsize effects being corrected within the general scheme proposed by Zunger et al.24 We used the potential-alignment correction given by their equation 7 and the image-charge correction in the form of their equation 24 with 1 + f ) 0.65. In applying the latter correction, we estimated the static dielectric constant for CuSCN to be ε ) 10, based on an assumed similarity to the Cu-VII zinc-blende and Cu-III-VI2 chalcopyrite semiconductors. Defect calculations were performed by removing atoms (creating vacancies) in 32- and 72-atom supercells (2 × 2 × 1 and 3 × 3 × 1 of the primitive cell). Spin-orbit and finitetemperature effects were not included. 3.2. CuSCN Thin Films. For the deposition solution, CuSCN powder (0.2 g) was dissolved in propyl sulfide (10 mL, ∼0.02 M CuSCN) and stirred for 48-72 h.6 The solution was allowed to settle for 24 h before use. The fused quartz substrate was mounted in a conventional spin-coater, and an infrared lamp mounted above the coater heated the substrate surface to approximately 70-80 °C. The substrate was rotated at 3000 rpm during dropwise deposition of CuSCN solution through a 0.45 µm filter. Rotation was stopped after deposition, and the sample was left at an elevated temperature for a few minutes. Film thickness was controlled by the number of drops deposited; a thick, hazy film was obtained with 20 drops, whereas a thin, optically transparent film was obtained with 2 drops. Film crystallinity was evaluated with grazing incidence X-ray diffraction (GIXRD, Philips MPD) with monochromatized Cu KR X-rays (λ ) 1.5406 Å) at fixed ω ) 5°. MicroXRD data were collected using a Rigaku D/Max Rapid II instrument with a 2D image plate detector. X-rays were generated with a MicroMax 007HF generator fitted with a rotating Cr anode (λ ) 2.2897 Å) and focused on the specimen through a 50 µm diameter collimator at a grazing incidence of 1°. Optical properties of the films were measured in transmission with a UV-visible spectrometer (Varian Cary 5) in the wavelength range of 200-800 nm. X-ray photoelectron spectroscopy (XPS, Scienta SES-200) was utilized to investigate the stoichiometry, charge state, and valence band structure of the CuSCN films. A low-energy electron flood gun was utilized to compensate for charging during measurement. Charge compensation was not complete, and thus, the reported binding energies (BEs) cannot be considered absolute. 4. Computational Results 4.1. Electronic Structure of CuSCN. The experimental crystal structure parameters of hexagonal β-CuSCN were accurately reproduced by the DFT calculation utilizing the generalized gradient approximation (GGA; see the Methods section for further details). We predict a lattice constant of a ) 3.781 Å versus 3.850 Å experimentally and with c/a ) 2.906, uS ) 0.430, uC ) 0.279, and uN ) 0.172, in good agreement with the experimental values12 cited above. The band structure of bulk hexagonal β-CuSCN is shown in Figure 2. Open (solid) dots represent conduction (valence) band states. The lowest valence bands arise from C and N 2s states and S 3s states, whereas the bands between -3 and -10 eV (relative to the valence band maximum) are associated with covalent bonds between S, C, and N, as discussed below. The highest valence bands are derived mostly from Cu 3d orbitals and show some

Electronic and Defect Structures of CuSCN

J. Phys. Chem. C, Vol. 114, No. 19, 2010 9113

Figure 3. Total and partial densities of states (DOS) calculated for β-CuSCN. The zero of energy has been placed at the top of the valence band, and the vertical zero is shifted for clarity.

Figure 2. GGA band structure of β-CuSCN. Open symbols are conduction bands; solid symbols are valence bands.

similarities to the upper valence bands of tetrahedrally coordinated I-VII compounds,25 for example, wurtzite structure β-AgI, with the valence band maximum occurring at the Γ point. However, whereas the I-VII compounds have a conduction band minimum at Γ associated with Cu 4s or Ag 5s states, β-CuSCN has a more complex conduction-band structure with a minimum at the K point, leading us to predict an indirect gap. Total and partial densities of states for β-CuSCN are shown in Figure 3. Partial DOS are shown only for orbital symmetry components that carry significant weights. The sub-band centered around -9 eV has predominantly S 3p and C 2p character and may be associated with the S-C covalent bond, whereas the next three sub-bands mainly contain cyanide triple bond states, with some weight near -3.5 eV from the Cu-S semi-ionic or Cu-N dative bonds. The highest valence bands have predominant Cu 3d character, as already mentioned, with a small amount of S 3p hybridization also present. The conduction band has almost no weight on the Cu atom but, instead, shows mainly C and N 2p character. We attribute these bands to unoccupied quasi-molecular antibonding orbitals of the cyanide portion of the thiocyanate ion. The full weight of the total density of states is not recovered because the antibonding states lie primarily outside the atomic radii used for the partial density of states projection (0.62 Å for carbon and 0.54 Å for nitrogen). These unoccupied states have no counterpart in conventional semiconductors, such as the I-VII, II-VI, or III-V compounds. They should be somewhat localized near the CN units and so explain the relatively flat conduction bands with especially low dispersion along the c axis, for example, Γ-A (see Figure 2). This situation may have implications for the transport properties of the material (electrons of similar or even lower mobility than holes; see the discussion below).

4.2. Defects in CuSCN. We now consider some of the simpler point defects that may occur in CuSCN. Because the p-type character of the material is clearly associated with Cu deficiency,9 we consider first the Cu vacancy. Removal of a Cu atom from the 32-site supercell resulted in only slight relaxation in the positions of the surrounding atoms. Properly referenced to the total energy of bulk Cu, this neutral vacancy had a formation energy of 0.71 eV in the copper-rich limit. This is fairly similar to the value26,27 in CuInSe2 and indicates an easily formed vacancy, but under Cu-rich conditions and roomtemperature growth, this formation energy would still imply a fairly low concentration of Cu vacancies. However, under Cupoor conditions, much higher concentrations of native acceptors should be expected. The limits of chemical stability of CuSCN are not well known; however, we computed the enthalpy of formation of β-CuSCN from Cu metal and thiocyanogen gas to be approximately -1.14 eV per formula unit, suggesting that the structure might still be stable at a Cu chemical potential of µCu ∼ -0.5 eV, resulting in a Cu vacancy formation energy of ∼0.2 eV and a high equilibrium concentration on the order of 1019 cm-3. For the larger supercell, we obtained a neutral VCu formation energy of 0.55 eV, with similar conclusions about the attainability of high vacancy concentrations. With proper corrections24 to the charged defect energies, we obtain the acceptor level of VCu to be at EVBM + 0.11 eV (31-atom cell) or EVBM + 0.09 eV (71-atom cell), showing that a high proportion of these acceptors will be ionized at room temperature. These results are consistent with ref 9, which reported shallow acceptor levels in p-type β-CuSCN at about 0.2 eV above the valence band edge. Little experimental evidence is available about other native defects that may exist in β-CuSCN. An exhaustive search of all theoretical possibilities is beyond the scope of the present work, but we can speculate as to what is possible. The hexagonal structure is reasonably close-packed, likely making interstitials high in energy, and the large physical asymmetry between the Cu atoms and SCN units makes high concentrations of antisites similarly unlikely. Hence, we focus on other possible vacancies. If a complete thiocyanate ion is missing, the resulting molecular vacancy should be a donor, which might tend to compensate the native acceptors. We modeled this situation in the 2 × 2 ×

9114

J. Phys. Chem. C, Vol. 114, No. 19, 2010

1 supercell (29 atoms) and calculated the formation energy for neutral VSCN under the most favorable (Cu-rich) conditions to be 3.17 eV. Under more Cu-poor conditions, this defect will be even higher in energy; thus, we would not expect the VSCN molecular vacancy to occur in high concentrations in its neutral state. We predict the (0|+1) donor level to lie 1.05 eV above the valence band maximum EVBM, so even under the expected strong p-type conditions (EF ) EVBM, favoring the formation of compensating donors) the formation energy of the charged donor is still 2.05 eV; this is much higher than the formation energy of the native acceptor VCu. Thus, we do not predict VSCN to occur in numbers sufficient to affect the doping behavior of CuSCN under equilibrium conditions, even in the presence of excess metallic Cu. We note that ref 9 reported the growth of n-type CuSCN through the use of applied electrochemical potentials or pH control by additional reagents to force excess Cu into the β-phase of the material. Possibly such samples were out of thermodynamic equilibrium but were kinetically stable. For n-type CuSCN, these authors9 reported a donor level 1.3-1.6 eV above the valence band edge, in rough agreement with what we have predicted for the SCN vacancy. Later experimental research on p-type β-CuSCN reported10 an acceptor level in the CuSCN gap at about 0.9 eV above the valence band edge, which is far too high in energy to account for hole conduction at room temperature. However, if a small concentration of deep donors coexists with a large concentration of shallow acceptors, then the donor levels will be ionized (unoccupied by electrons) due to overcompensation by the acceptors, and the empty donor states may appear to act like deep acceptors. Thus, we propose that these observed10 deep levels at EVBM + 0.9 eV may actually result from the VSCN molecular vacancy donor level we predicted at EVBM + 1.05 eV. Another possibility for a vacancy in CuSCN would be for only a part of the SCN ion to be missing. An S vacancy (or equivalently, (CN)SCN) would probably be electrically inactive because both the cyanide and the thiocyanate ions are monovalent. Removal of only C or only N would likely be very costly in energy due to the strength of the CtN triple bond. A more interesting possibility would be a compound vacancy VCN (or equivalently, SSCN). With the divalent sulfide ion replacing monovalent thiocyanate, such a defect could be another acceptor, in addition to the Cu vacancy. This might permit an enhancement of the p-type doping of CuSCN, by combining Cu-poor and moderately S-rich growth conditions. Modeling of such an acceptor in a 32-site (30-atom) supercell leads to an acceptor level in the gap at about 0.30 eV above the valence band maximum; for the larger supercell, the level is degenerate with the valence band. This level is thus predicted to be sufficiently shallow that it could contribute to p-type doping at room temperature, provided the CN vacancy occurs at a significant concentration. 5. Experimental Results 5.1. Structure of CuSCN Thin Films. The GIXRD patterns for two thicknesses of CuSCN films spin-cast on fused quartz are given in Figure 4. The thick film (20 drops of ∼0.02 M CuSCN solution) was hazy, whereas the thin film (2 drops, approximately 48 nm thick by X-ray reflectivity) appeared optically smooth and transparent by visual inspection. Likewise, the GIXRD pattern for the thick film indicates clear peaks arising from crystalline β-CuSCN, whereas the thin film does not show any crystalline peaks. The broad amorphous hump present in both spectra at ∼22° arises primarily from the fused quartz

Jaffe et al.

Figure 4. (a) GIXRD patterns of thick (20 drops of CuSCN solution) and thin (2 drops) CuSCN films spin-cast on fused quartz at ∼80 °C. Peaks arising from β-CuSCN are indicated. (b) MicroXRD pattern of thin CuSCN on fused quartz (top) and after subtracting the pattern of the bare fused quartz substrate and removing the remaining background (bottom). Peaks arising from β-CuSCN are indicated.

substrate. The thin film was also measured by microXRD, which has significantly improved sensitivity to crystalline order. As seen in Figure 4b, clear peaks belonging to β-CuSCN appear in the microXRD pattern, indicating that this film possesses weak crystalline order. 5.2. Composition and Valence Band Structure. An XPS survey spectrum of the thin CuSCN film is shown in Figure 5a. This spectrum was collected without any sample cleaning to remove the adventitious carbon overlayer. In addition to the expected signatures of Cu, S, C, and N, a clear O 1s peak is present, which is likely associated primarily with adventitious contamination. No peaks of other contaminants are present. Due to energy-dependent photoelectron attenuation, the presence of the carbon overlayer complicates the accurate determination of film stoichiometry. The film stoichiometry was estimated from peak areas (Cu 3p, S 2p, C 1s at 286 eV, N 1s) after subtracting a Shirley background from each peak. Sensitivity factors for each element were derived from known experimental peak areas for each element28 after applying the appropriate transmission function for the Scienta spectrometer. The contribution to the C 1s peak at 285 eV (see the high-resolution data below) was assumed to arise from the adventitious carbon overlayer; the film stoichiometry was determined after accounting for this carbon overlayer.29 The composition results, shown in Figure 5a, indicate that Cu, S, C, and N are present in roughly stoichiometric (1:1:1:1) proportions, with slightly more C and less S than expected.

Electronic and Defect Structures of CuSCN

J. Phys. Chem. C, Vol. 114, No. 19, 2010 9115

Figure 6. (top) XPS valence band spectrum of thin CuSCN at a 90° takeoff angle. The spectrum was shifted to place the Cu 3d peak at 3.0 eV. (bottom). The same valence band spectrum after subtraction of a Shirley background. The calculated DOS is also plotted on the same energy scale after broadening by convolving with a Gaussian of 1.0 eV fwhm and shifting for optimal overlap with the experimental spectrum. Dotted lines indicate calculated conduction band states that are not observable by XPS.

Figure 5. (a) XPS survey spectrum of thin CuSCN at a 90° takeoff angle. An adventitious carbon overlayer is present. Film stoichiometry was calculated as described in the text. (b) High-resolution spectra of N 1s, C 1s, S 2p, and Cu 2p regions. The C 1s peak was fit with two Voigt functions (thin lines) after subtraction of a Shirley background (dotted line).

High-resolution spectra are shown in Figure 5b. Charge compensation with the flood gun was optimized for these spectra, resulting in only small shifts in binding energy from the absolute values. The C 1s peak is dominated by adventitious carbon at approximately 285 eV. However, a clear shoulder can be observed at higher binding energy. Deconvolution results in two peaks that can be assigned to adventitious carbon (284.8 eV) and CtN (286.0 eV).30 Using only the area of the high binding energy peak, the C content in the film (1.15) is found to be close to stoichiometric (1.00), as indicated in Figure 5a, supporting the C 1s peak assignments. The N 1s peak has a single component at BE ≈ 400 eV, consistent with CtN bonding.30 Likewise, the S 2p peak consists of a single component centered at BE ≈ 165 eV; the spin-orbit splitting is poorly resolved. The high-resolution Cu 2p spectrum is also shown in Figure 5b. Because of the slight binding energy shift arising from sample charging, the binding energy position of this peak (BE ) 933.3 eV) cannot be used to distinguish between Cu(I) and Cu(II). However, the weak intensity of the shakeup satellites present at 7-13 eV higher binding energy argues for Cu(I), analogous to XPS spectra of Cu(I)2O versus Cu(II)O.31 Further confirmation of Cu(I) is obtained from the position of the Cu LMM feature in the survey spectrum (Figure 5a). For CuSCN, the peak separation between the Cu 2p3/2 and the primary LMM peak (KE ≈ 916-918 eV) is found to be 361.8 eV, which matches well with the value of 362.3 eV for Cu2O; both Cu metal and CuO have larger separations of 364.4 and 364.8 eV, respectively.31

The XPS valence band (VB) spectrum of the thin CuSCN film is shown in Figure 6. Valence band spectra at a low photoelectron takeoff angle (not shown) confirm that the contribution from the adventitious carbon overlayer is essentially featureless, and the VB spectrum obtained at normal emission in Figure 6 is characteristic of the CuSCN film, although the intensity tail extending into the band-gap region arises from the carbon overlayer. The most prominent feature in the VB spectrum is the Cu 3d level, which occurs as a well-resolved peak at approximately 3 eV;32,33 the VB spectrum was shifted to place the Cu 3d level at 3.0 eV. Jiang et al.32 attributed the peak at 12.4 eV, which they observed in CuO, to Cu-O bonding; the peak is not present in the VB spectrum of Cu metal. Also shown in Figure 6 is an overlay of the calculated VB DOS for β-CuSCN (from Figure 3), where the VB maximum was taken as BE ) 0 and the sign of the energy scale was reversed to coincide with the XPS data. To account for XPS instrumental resolution (0.46 eV) and intrinsic lifetime effects, the calculated DOS was convolved with a Gaussian of 1.0 eV full width at half-maximum (fwhm). To facilitate comparison with the calculated results, a Shirley background was subtracted from the CuSCN VB spectrum. The main features of the experimental spectrum are reproduced in the calculated DOS, including the Cu 3d peak at low binding energy and the smaller peak at higher binding energy; the overall width of the valence band spectrum is closely matched by the calculated DOS. As discussed above, the partial DOS plotted in Figure 3 reveals that the intensity of the calculated states between ∼3 and 10 eV arises from covalent bonding in CuSCN (Cu-S, S-C, and CtN bonds). The correlation between the predicted and experimental CuSCN VB confirms that the calculations successfully reproduce the essential electronic structure of CuSCN. 5.3. Optical Properties. Optical absorption data for both CuSCN films, as well as a bare fused quartz substrate, are shown in Figure 7a. High-quality optical data were obtained for the thin, smooth film. For the thick, hazy film, a 1 mm aperture was employed to minimize light scattering, and the transmission spectrum was collected from a thinner, smoother region near the edge of the film. Thickness interference fringes are apparent in the data above 400 nm; below ∼300 nm, light-scattering effects dominate, although the shoulder in the absorption edge near 290 nm in the thin film is reproduced in the thick film. To determine the optical band gap for both CuSCN films, each transmission spectrum was normalized by the spectrum for the

9116

J. Phys. Chem. C, Vol. 114, No. 19, 2010

Figure 7. (a) Optical absorption data for thick and thin CuSCN films on fused quartz, as well as a bare fused quartz substrate. A small (1 mm) aperture was utilized for the thick CuSCN film; the thin CuSCN film and the fused quartz spectra were collected without an aperture. (b) Normalized absorption data plotted as (Rhν)2 for both CuSCN films. Band-gap values determined by extrapolated linear fits are indicated.

bare fused quartz substrate collected with the same aperture. To calculate the absorption coefficient (R), 48 and 480 nm were estimated as the film thickness of the thinner and thicker films, respectively. Assuming a direct band gap, the absorption data were then plotted as (Rhν)2 versus photon energy (hν), and the linear region of the onset of absorption was extrapolated to zero, as shown in Figure 7b. The band gap for the thin film is thus found to be 3.90 eV, and a similar value (3.94 eV) is determined for the thick film. These values are comparable to those reported previously.34-36 For both films, the absorption tail extends 1.0-1.5 eV below the band edge. 6. Discussion The onset of strong optical absorption at about 3.9 eV in the experimental data on β-CuSCN appears to fit a direct transition, as noted both in section 5.3 and in earlier work elsewhere.34-36 The question remains, however, whether the lowest transition is direct or indirect. Our calculated results do, in fact, show a direct transition about 0.4 eV above the lowest transition, but we predict the latter to be indirect, as discussed in section 3.1. We do not believe that the experimental evidence for a lowest direct gap is conclusive because it shows a tail of absorption extending for at least 1 eV below the extrapolated direct edge. This tailing is much more extensive than typically seen in known direct gap materials, such as ZnO, where the tail of the absorption only extends on the order of 0.1 eV below the direct

Jaffe et al. edge. The extended tail in CuSCN could have any of several causes, including secondary crystalline phases or amorphous regions in the sample; however, it may also derive from an indirect transition below the direct one, as we predict from our calculations. Further measurements on well-crystallized samples will be needed to resolve this issue. The origin of the indirect lowest transition in the calculation is of some interest. It derives from the fact that the lowest conduction band has molecular antibonding π* character instead of metallic s-character, as in, for example, CuCl. For the latter, the wave function overlap between unit cells has bonding character at the Brillouin zone center (the Γ point) but increasingly antibonding character as the wave vector approaches the zone boundary; hence, band energies tend to increase as we move away from Γ, making a direct lowest transition at Γ likely. With a lowest conduction band derived from p-like states, the opposite is true: The intercell overlap has antibonding characteristics at Γ but becomes more bondinglike, at least in certain directions, as we move toward the edge of the Brillouin zone. Thus, the band energy will tend to drop as we move away from Γ, resulting in a conduction band minimum elsewhere in the zone. This means that the indirect gap in CuSCN is a consequence of the electronic structure of the thiocyanate molecular ion, which has no counterpart in conventional semiconductors, such as GaAs or ZnO. The unusual electronic structure of CuSCN has several implications for its useful properties. First, we consider the effective masses near the band edges, which are connected to carrier mobility via the semiclassical group velocity (we do not, however, consider scattering or trapping rates in this paper). The band structure leads to relatively large effective masses for electrons in a semiconductor: m* ∼ 2m for wavevectors in the ab plane, and m* ∼ m along the c axis. These high masses are a consequence of the quasi-molecular character of the lower conduction band, in contrast to the more metallic character of the conduction band minimum in a typical compound semiconductor. Thus, we might expect that electron mobilities in CuSCN will be lower than those in traditional semiconductors. On the other hand, we find that heavy (light) hole masses in β-CuSCN average around m* ) 2m (0.5m) in the ab plane, and both are ∼0.8m along the c axis. These values are fairly typical for hole masses in many semiconductors and suggest that holes have similar or even higher mobility than electrons in CuSCN, an unusual situation. The reasonably high valence band dispersion may be due to the strong hybridization between Cu 3d and S 3p electrons. Thus, the likelihood of high hole mobility, combined with the predicted low concentration of hole traps (the native deep donor VSCN) and the high concentration and low ionization energy of native acceptors (mainly VCu), makes β-CuSCN an excellent p-type conductor. Some features in the electronic structure of β-CuSCN may also help explain its high optical transparency. The material has a wide band gap and is expected to have very few deep native defect levels (color centers) in the gap, and we predict the lowest transitions to be indirect; the latter are expected to have very low optical absorption because phonon assistance is required to conserve crystal momentum. This circumstance may possibly account for our inability to observe an unambiguous indirect transition below the direct edge. Furthermore, even the lowest direct transitions should have only weak optical matrix elements because the initial and final states are centered on different atoms (Cu and C, N, respectively) with low orbital overlap between them.

Electronic and Defect Structures of CuSCN 7. Conclusions The electronic structure of the p-type semiconductor β-CuSCN has been elucidated by DFT calculations and corroborated by experimental measurements of thin CuSCN films. The overall structure of the calculated valence band DOS matches well with the valence band of CuSCN measured by XPS, confirming the accuracy of the electronic structure calculations. In contrast to similar semiconductors, such as ZnO, the calculations predict that the lowest band-gap transition in CuSCN is indirect as a consequence of the molecular character of the thiocyanate anion. Optical absorption data on thin films of CuSCN reveal a strong direct transition at 3.90 eV, with a long absorption tail possibly consistent with a weak indirect transition at lower energy. Further analysis of the DFT results reveals an unusually high effective mass for electrons in CuSCN, potentially contributing to low electron mobility, which favors intrinsic p-type conductivity of the material. This p-type conductivity is predicted to be enhanced through the formation of Cu vacancies (VCu) and possibly also through the formation of CN vacancies (SSCN). The wide band gap and specific orbital character of the bandedge states in CuSCN give it excellent optical transparency, so it may have many applications as a transparent conductor in optoelectronic devices, provided that complementary materials with suitable band offsets to CuSCN can be found. Acknowledgment. A portion of this research was performed using EMSL, a national scientific user facility sponsored by the U.S. Department of Energy’s Office of Biological and Environmental Research and located at the Pacific Northwest National Laboratory. This research was supported by the Laboratory Directed Research and Development Program at PNNL. References and Notes (1) Exarhos, G. J.; Zhou, X. D. Thin Solid Films 2007, 515, 7025. (2) Granqvist, C. G. Sol. Energy Mater. Sol. Cells 2007, 91, 1529. (3) Banerjee, A. N.; Chattopadhyay, K. K. Prog. Cryst. Growth Charact. 2005, 50, 52. (4) Tennakone, K.; Ariyasingha, W. M. Electrochim. Acta 1980, 25, 731. (5) Snaith, H. J.; Schmidt-Mende, L. AdV. Mater. 2007, 19, 3187.

J. Phys. Chem. C, Vol. 114, No. 19, 2010 9117 (6) O’Regan, B.; Schwartz, D. T.; Zakeeruddin, S. M.; Gratzel, M. AdV. Mater. 2000, 12, 1263. (7) Levy-Clement, C.; Tena-Zaera, R.; Ryan, M. A.; Katty, A.; Hodes, G. AdV. Mater. 2005, 17, 1512. (8) Dittrich, T.; Kieven, D.; Belaidi, A.; Rusu, M.; Tornow, J.; Schwarzburg, K.; Lux-Steiner, M. C. J. Appl. Phys. 2009, 105, 034509. (9) Tennakone, K.; Jayatissa, A. H.; Fernando, C. A. N.; Wickramanayake, S.; Punchihewa, S.; Weerasena, L. K.; Premasiri, W. D. R. Phys. Status Solidi A 1987, 103, 491. (10) Perera, V. P. S.; Senevirathna, M. K. I.; Pitigala, P. K. D. D. P.; Tennakone, K. Sol. Energy Mater. Sol. Cells 2005, 86, 443. (11) Frost, D. C.; Kirby, C.; Lau, W. M.; MacDonald, C. B.; McDowell, C.; Westwood, N. P. C. Chem. Phys. Lett. 1980, 69, 1. (12) Smith, D. L.; Saunders, V. I. Acta Crystallogr. 1982, B38, 907. (13) Smith, D. L.; Saunders, V. I. Acta Crystallogr. 1981, B37, 1807. (14) Kabesova, M.; Dunaj-Jurco, M.; Serator, M.; Gazo, J.; Garaj, J. Inorg. Chim. Acta 1976, 17, 161. (15) Kresse, G.; Furthmu¨ller, J. Comput. Mater. Sci. 1996, 6, 15. (16) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169. (17) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, 558. (18) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 48, 13115. (19) Kresse, G.; Hafner, J. Phys. ReV. B 1994, 49, 142518. (20) Kresse, G.; Hafner, J. J. Phys.: Condens. Matter 1994, 6, 8245. See also http://cms.mpi.univie.ac.at/vasp/. (21) Blo¨chl, P. E. Phys. ReV. B 1994, 50, 17953. (22) Monkhorst, H. J.; Pack, J. D. Phys. ReV. B 1976, 13, 5188. (23) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. Lett. 1996, 77, 3865. (24) Lany, S.; Zunger, A. Phys. ReV. B 2008, 78, 235104. (25) Victora, R. H. Phys. ReV. B 1997, 56, 4417. (26) Zhang, S. B.; Wei, S.-H.; Zunger, A. Phys. ReV. Lett. 1997, 78, 4059. (27) Zhang, S. B.; Wei, S.-H.; Zunger, A.; Katayama-Yoshida, H. Phys. ReV. B 1998, 57, 9642. (28) Wagner, C. D.; Davis, L. E.; Zeller, M. V.; Taylor, J. A.; Raymond, R. H.; Gale, L. H. Surf. Interface Anal. 1981, 3, 211. (29) Castle, J. E. J. Vac. Sci. Technol., A 2007, 25, 1. (30) Ronning, C.; Feldermann, H.; Merk, R.; Hofsass, H.; Reinke, P.; Thiele, J.-U. Phys. ReV. B 1998, 58, 2207. (31) Poulston, S.; Parlett, P. M.; Stone, P.; Bowker, M. Surf. Interface Anal. 1996, 24, 811. (32) Jiang, Z. T.; Thurgate, S. M.; Wilkie, P. Surf. Interface Anal. 2001, 31, 287. (33) Andrews, P. T.; Hisscott, L. A. J. Phys. F: Met. Phys. 1975, 5, 1568. (34) Ni, Y.; Jin, Z.; Fu, Y. J. Am. Ceram. Soc. 2007, 90, 2966. (35) Gao, X. D.; Li, X. M.; Yu, W. D.; Qiu, J. J.; Gan, X. Y. Thin Solid Films 2008, 517, 554. (36) Wu, W.; Jin, Z.; Hua, Z.; Fu, Y.; Qiu, J. Electrochim. Acta 2005, 50, 2343.

JP101586Q