Low-temperature photoluminescence of CuSe2 nano-objects in selenium thin films

Thin films of CuSe2 nanoparticles embedded in selenium matrix were prepared by vacuum evaporation method on a glass substrate at room temperature. The optical properties of the films were investigated by photoluminescence spectroscopy (T = 20–300 K) and UV-VIS spectroscopy (T = 300 K). Surface morphology was investigated by scanning electron microscopy. The band gap for direct transition in CuSe2 was found to be in the range of 2.72–2.75 eV and that for indirect transition is in the range of 1.71–1.75 eV determined by UV-VIS spectroscopy. On the other hand, selenium exhibits direct band gap in the range of 2.33–2.36 eV. All estimated band gaps slightly decrease with the increase of the film thickness. Photoluminescence spectra of the thin films clearly show emission bands at about 1.63 and 2.32 eV at room temperature, with no shift observed with decreasing temperature. A model was proposed for explaining such anomaly.

Possible application of Cu-Se strongly depends on its optical properties.Despite the numerous publications of the optical properties of Cu-Se thin films, the estimated value of band gap of Cu-Se is not well defined.Cu-Se has both direct and indirect transitions, so the presence of both band gaps, direct and indirect, can be observed.Literature data are quite controversial: direct allowed transitions are reported to have corresponding band gap in the range of 2 to 3 eV, and indirect band gap between 1.1 and 1.5 eV [15][16][17][18].The indirect band gap being near the optimum value for solar cell applications makes this material capable of potentially offer a high efficiency of conversion.However, Cu-Se nanoparticles have been reported to possess a direct band gap of 4 eV and indirect one of 1.87 eV [19].The reasons of such variation in band gaps could lie in the sharp cut off of the wavelength with the spectral transmittance instead of the slow increase, the presence of large number of dislocations, wide range of stoichiometric deviation and quantum confinement effect.
In our previous report [20], we prepared Cu-Se thin films of three different thicknesses by vacuum evaporation technique using Mo boat onto glass substrate at room temperature, and investigated their structural properties.XRD and far-infrared spectroscopy revealed the presence of CuSe 2 nanocrystals in predominant Se films.
The objective of this research was to determine optical properties of Cu-Se thin films, using UV-VIS and low-temperature photoluminescence spectroscopy, and to investigate how the film thickness could influence the band gap value and photoluminescence properties.SEM analysis was also performed in order to get information about the morphology of the obtained Cu-Se thin films.

II. Experimental
Thin films were obtained by evaporating commercially high purity CuSe powder (99.99%) supplied from Aldrich Company.The powder was deposited onto highly pre-cleaned glass substrates with use of Mo boat.The procedure was done in a high-vacuum environment with typical background pressures of 3 mPa.The deposition rate, 10 nm/s, was monitored by quartz crystal thickness monitor -FTM4, Edwards and the final films thicknesses was found to be 56 nm, 79 nm and 172 nm for the films labelled as F-55, F-80 and F-170, respectively [20].
SEM imaging was done using scanning electron microscope equipped with a high brightness Schottky field emission gun (FEGSEM, TESCAN) operating at 4 kV.The samples were coated with gold/palladium to make them conductive.
The UV-VIS diffusion reflectance and transmittance spectra were recorded in the wavelength range of 300-1000 nm on a Shimadzu UV-2600 spectrophotometer equipped with an integrated sphere.The diffuse reflectance and transmittance spectra were measured relative to a reference sample of BaSO 4 .
Photoluminescence measurements on various temperatures (T = 20-300 K) were obtained by Jobin-Yvon U1000 spectrometer, equipped with RCA-C31034A photomultiplier with housing cooled by Peltier element, amplifiers and counters.The 488 nm laser line of argon laser was used as excitation source.

SEM analyses
Scanning electron microscopy (SEM) images were obtained for the Cu-Se thin films deposited on glass substrate in order to study the surface morphology and agglomeration of the samples.Top view and tilted micrographs of the thin films are presented in Fig. 1.From the top view micrographs it can be observed that the surface of the samples is relatively uneven and rather rough, with presence of cracks and voids.Formation of the Cu-Se thin films most probably proceed unevenly, in the form of islands which later grew into agglomerates.Agglomerated clusters of few hundreds nanometers in diameter are distributed nonuniformly along the surface and form the structure consisting of CuSe 2 nanocrystals in predominant Se matrices.In order to determine the film thicknesses, the samples were tilted at 30°.The thicknesses estimated by SEM are: ∼56 nm, ∼78 nm and ∼171 nm for the films F-55, F-80 and F-170, respectively.The thickness values estimated by SEM analysis match the ones obtained during the preparation of thin films.

UV-VIS spectroscopy
In Fig. 2 diffuse reflectance R and transmittance T spectra of the thin films samples in the wavelength range 300-1000 nm (4.13-1.24eV) on room temperature (T = 300 K) are presented.As it can be seen, the transmittance increases with decrease in the film thickness, which is not the case for reflectance.This is typical for films with high electrical conductivity and implies a reflection coefficient nearing 1 for films with metallic conductivity.
In this study we used the Tauc plot for the determination of the optical band gap from diffuse reflectance measurements.The determination of band gap in semiconductors is significant for obtaining the basic solid state physics.The relation expression proposed by Tauc, Davis and Mott [21][22][23] is the following: where h is the Planck's constant, A is the transition probability constant depending on the effective mass of the charge carriers in the material, E g is the band gap, h • ν is the photon energy and α is the absorption coefficient which is defined as the relative rate of decrease in light intensity along its propagation path, i.e. a property of a material that defines the amount of light absorbed by it.The value of n denotes the nature of the transition.In case of direct transitions n equals 1 /2 and 3 /2 for allowed and forbidden transitions, respectively.As for indirect transitions, n equals 2 and 3 for allowed and forbidden transitions, respectively.Since CuSe exhibits both direct and indirect allowed transitions, n = 1 /2 and n = 2.
Then, the acquired diffuse reflectance spectra are converted to Kubelka-Munk function [24]: So using this function, a plot of (α The energy band gap is determined by extrapolating the linear portion of (α The intercept of these plots on the energy axis gives the energy band gap.Such plots are given in Fig. 3. Direct transitions (left part of Fig. 3) reveal band gap for both selenium and CuSe 2 , while indirect transitions (right part of Fig. 3) reveal band gap for CuSe 2 only.
The experimentally determined values of energy gaps for CuSe 2 show slight decrease with film thickness and their values range from 2.75 to 2.72 eV for direct transitions, and from 1.75 to 1.71 eV in case of indirect transitions.The estimated band gaps for selenium follow the same trend with film thickness and range between 2.33 and 2.36 eV.The estimated band gap positions for each sample are given in Table 1.The difference in the film thicknesses causes the small difference in band gaps in the second decimal place and they follow the wellestablished trend, the smaller the thickness, the wider the band gap is.Also, their values are quite wider than the ones that can be found in literature [7,16,[25][26][27][28][29][30][31][32][33].Broad range of energy band gap values for Cu-Se can be found in literature.For direct transitions those values are usually between 2 and 3 eV.Bari et al. [25] obtained the value of 2.51 eV for the sample with thickness of 150 nm, and with the increase of film thickness they reported the decrease of band gap width.Grozdanov [26], Garcia et al. [16] and Sakr et al. [27] obtained the value of 2.33, 2.13-2.38 and 2.74 eV, respectively.The latter is very similar to the results obtained in this paper.Rajesh et al. [28] got a diversity of band gaps ranging from For the pure selenium, the direct band gap was reported to be about 2 eV for the bulk [31,32] and 2.20-2.06 for the thin films of thickness 130-290 nm [33].Our films are thinner than the ones mentioned in the literature and it is not surprising that we obtained wider band gap values, about 2.3 eV.The film F-170 has the thickness between 130 and 290 nm, but the higher value in band gap is the result of specific border conditions between selenium matrix and CuSe 2 nanoparticles.We presume that CuSe 2 nanoparticles directly influence the band gap of predominant selenium, and vice versa.
The Urbach energy is also analysed.Urbach rule states that the optical absorption coefficient α just below the band edge in insulators and semiconductors varies exponentially with the incident photon energy [34]: where α 0 is a constant and characteristic parameter of the material, h • ν is incident photon energy, and the term E U which is the width of the exponential tail is called Urbach energy.The Urbach energy represents the width of defect states in the band gap. Figure 4 shows log α as a function of incident photon energy for the film F-170, as a representative one.By extrapolating the linear part of the plot and with use of Equation 3, the Urbach energy can be determined as the inverse of the slope and α 0 from the intercept of extrapolated plot.From the slope and inception of extrapolated plot it was determined that E U is 0.32 eV and α 0 is 3.02 cm -1 .

Photoluminescence spectroscopy
Photoluminescence (PL) spectra can be used for investigating the possible outcomes of photo-induced electrons and holes in a semiconductor, since photoluminescence emission results from the recombination of free charge carriers.There are two types of photoluminescence phenomenon according to its attributes and formation mechanism: the band-to-band photoluminescence and the excitonic photoluminescence [35][36][37].The band-to-band PL spectrum regards the separation situation of photo-generated charge carriers.The excitonic PL spectrum, however, cannot directly reflect the separation situation of photo-induced carriers.If discrete energy levels are present in the band gap, these may dominate the optical spectrum.PL measurements then yield information about the energetic positions of the electronic states in the gap.Such localized states can originate from various types of imperfections like vacancies, interstitial atoms, atoms at surfaces and grain boundaries.However, it is often difficult to determine the exact position and origin of these states.
Photoluminescence spectra of thin films on various temperatures are presented in Fig. 5.The spectra are rather complex, thus for their analysis the deconvolution method had to be employed.Two typical resolved spectra are presented in Fig. 6 (on 20 K and room temperature), and the deconvolution of the others is done in the same manner.Each spectrum is characterized with 5 bands.Band-to-band photoluminescence dominates the room temperature spectra.The band in red area at ∼1.6 eV is clearly seen.According to the UV-VIS results (see previous chapter), we can attribute this mode to band-to-band transition for indirect transition in CuSe 2 .In green area, a broad band is observed at ∼2.3 eV which originates from direct transitions in selenium (also see prev.chapter).Due to Stokes shift, the obtained positions are a bit lower than the ones obtained by UV-VIS spectroscopy.Between these two bands there are three defect modes at 2.1 eV, 1.9 eV and 1.8 eV, which are of small intensity on room temperature.Lowering the temperature, the band at ∼1.8 eV increases its intensity, and becomes the dominant one at 20 K.According to literature data [32] this band is attributed to selenium defect mode -negative U-centre.This mode is expected to appear at 0.5 eV from the band edges.Another two defect modes, at 2.1 and 1.9 eV, are attributed to the defect modes of CuSe 2 , according to the work of Urmila et al. [38].In their work they obtained bands at 2.1, 1.9 and 1.5 eV in PL spectrum of Cu 7 Se 4 thin film.They concluded that there is nonradiative transition from conduction band to defect levels with energies 2.1, 1.9 and 1.5 eV and from these lev- els radiative transitions occurred to valence band.The values 2.1 and 1.9 eV match the band positions we obtained in this work, thus we can assign these bands to defect modes in CuSe 2 .Due to predominant phase of selenium, these bands are hidden by the bands originated from selenium and cannot be observed without the deconvolution method.Chong [39] in his work also obtained a PL band at ∼2.16 eV without assigning it to any transition.In all spectra, the uprise of peak intensity with lowering temperature is observed, Fig. 7.The intensities have their maximum values at the lowest temperature (20 K) and show decrease with rising temperature.The most radical decrease is observed for the band at ∼1.8 eV, which intensity drops to ∼40% at 200 K, and only to ∼6% at room temperature.Temperature dependence of PL emission band positions is shown in Fig. 8. Let us discuss the band position of indirect transitions of CuSe 2 .At the room temperature, there are differences on the second decimal place for films of different thicknesses, the same as observed when analysing UV-VIS spectra.The same trend is observed on 20 K.If the shift of this band with temperature is analysed for each film, it can be noticed the temperature invariance, i.e. small non-monotonous differences in positions on the second decimal place.This is in contradiction with the expected red shift, characteristic for the semiconductors.The temperature invariance is observed for other bands as well.There is a question that needs to be answered: why the PL measurements show no shift with increasing temperature, instead of the conventional red shift characteristic for the semiconductors?A model proposed by Shen et al. [40] explains those discrepancies.It involves surface electron accumulation as a result of severe band bending in nanorods.However, the same trend was observed in thin film samples (including ours), whose curvature-less surface does not support a spatial charge separation such as in 1D nanostructures.Wei et al. [41] gave more exact explanation of this phenomenon in their work.They began the analysis by making difference between E PL and E g : where E Fn and E F p are the electron and hole quasi-Fermi levels measured from the bottom of conduction band and the top of valence band, respectively.So the temperature dependence of the band gap shift is the competition between the lattice dilation dE g /dT on the one hand, and the sum (dE Fn /dT − dE F p /dT ) on the other hand.The former results in the conventional red shift of the band gap with increasing temperature and the latter gets the blue shift.The resulting shift depends on the magnitude of these two contributions.Usually when the electron density n is high, the thermal response in the material is governed by electronic rather than photonic interactions, the sum (dE Fn /dT − dE F p /dT ) becomes dominant thus the blue shift of E PL is observed.However, if these two contributions are of the same magnitude, it will result in no shift with changing temperature, as in case of our samples.PL emission bands positions as a function of film thicknesses are presented in Fig. 9.It is common knowledge that as the confining dimension decreases, typically in nanoscale, the energy spectrum turns to discrete so the band gap of a semiconductor becomes size dependent.For one-dimensional confinement (film thickness), the quantization energies increase when the size along the confinement direction decreases [42,43].In an amorphous or structurally disordered film, the imperfection in the film causes the bands of localized states to get broaden and a band gap reduction may occur due to the Urbach edge [44].If we observe the shift of the band at ∼1.6 eV, which corresponds to indirect transitions in CuSe 2 , we can see the blue shift with decreasing size, as being expected.On the other hand, the band at ∼2.3 eV which corresponds to direct transitions in Se, we can see no size dependence with the band position.The reason of this behaviour lies in specific composition of our films, i.e.CuSe 2 nano-objects embedded in selenium matrix.The particles of CuSe 2 are small enough to react on the size reduction, but the selenium matrix the amorphous effect becomes dominant.
As it can be seen from the above, the temperature changes do not affect the band gap.The size changes, i.e. the reduction of film thickness affects the band gap only on a second decimal place.Thus, we can state that the low cost technique of vacuum evaporation gives us the opportunity to produce quality, stable thin films suitable for further applications in heterojunction solar cells and photo detectors.

IV. Conclusions
Cu-Se thin films of three different thicknesses, obtained by vacuum evaporation technique on glass substrate, underwent through photoluminescence investigation along with UV-VIS measurements and SEM analysis.Reflectance measurements revealed values for both direct and indirect band gap: ∼2.7 and 1.7 eV, respectively for CuSe 2 and ∼2.3 eV for Se.The existence of indirect band gap in CuSe 2 at this value, little wider than in literature, is confirmed by photoluminescence measurements.A band at ∼1.8 eV, registered by PL measurements at low temperatures, is attributed to defect level of selenium -negative U-center.In this paper we proved that simple and low-cost technique as vacuum evaporation is capable of producing high-quality thin films.

Figure 3 .
Figure 3. Dependence of (α • h • ν) 2 on photon energy (h • ν) -left side, and dependence of (α • h • ν) 1/2 on photon energy (h • ν)right side 1.95 (the thickest film) to 3.70 eV (the thinnest film).However, for indirect transitions they received less attention.Garcia et al. [16] obtained values in the range 1.22-1.34eV, whereas the value obtained by Bhuse et al. [7] is about 1.4 eV.Our values (∼1.7 eV) are bigger than the reported in literature.According to our opinion, the larger indirect band gap values are due to quantum confinement effect [29,30] whereby the electrons are localized in individual crystallites, and due to specific border conditions between CuSe 2 nanoparticles and selenium matrix.For the pure selenium, the direct band gap was reported to be about 2 eV for the bulk[31,32] and 2.20-2.06 for the thin films of thickness 130-290 nm[33].Our films are thinner than the ones mentioned in the literature and it is not surprising that we obtained wider band gap values, about 2.3 eV.The film F-170 has the thickness between 130 and 290 nm, but the higher value in band gap is the result of specific border conditions between selenium matrix and CuSe 2 nanoparticles.We presume that CuSe 2 nanoparticles directly influence the

Figure 4 .
Figure 4. Logarithmic dependence of the absorption coefficient on the photon energy for F-170 film

Figure 6 .Figure 7 .
Figure 6.Typical resolved photoluminescence spectra of thin films F-170: a) on 20 K and b) room temperature

Figure 8 .
Figure 8. Temperature dependence of photoluminescence emission bands positions

Figure 9 .
Figure 9. PL emission bands positions dependence of film thicknesses

Table 1 .
Estimated band gap energies of thin films determined with UV-VIS spectroscopy