Reduced recombination through CZTS/CdS interface engineering in monograin layer solar cells

The power conversion efficiency of Cu2ZnSnS4 (CZTS) solar cells is still limited by deep defects, low minority carrier lifetime and high recombination rates at the CZTS/CdS interface. The objective of this study was to find an effective method to reduce interface recombination of CZTS monograin layer solar cells. A two-step heterojunction formation process was applied by controlling the intermixing of Cd and Cu in the CZTS/CdS interface, which resulted in improved device efficiency of up to 11.7%. Surface analysis by x-ray photoelectron spectroscopy confirmed Cd diffusion into the surface of CZTS after CdS air-annealing by forming an ultra-thin Cu2Zn x Cd1−x SnS4 layer. Moreover, external quantum efficiency measurements showed that the absorption edge shifts to longer wavelengths with the addition of Cd into the CZTS surface layer. This surface modification and replacement of a CdS:Cu buffer layer by fresh CdS greatly reduced the interface recombination and improved the junction quality, contributing to an enhancement of J SC ∼3 mA cm−2 (from 20.5 to 23.6 mA cm−2) and fill factor ∼14% (from 59.4% to 67.7%). The serial resistance of the CZTS monograin layer solar cells was significantly reduced from 2.4 Ω cm2 to 0.67 Ω cm2. To understand the electrical behavior of the highest-efficiency CZTS monograin layer solar cell in more detail, the temperature-dependent current–voltage characteristics were analyzed.


Introduction
Photovoltaic technology can provide a significant fraction of the world's energy demands if solar devices are composed of Earth-abundant and nontoxic materials. Among the inorganic absorber materials, kesterite materials (Cu 2 ZnSn(S,Se) 4 ) are promising candidates because they combine near optimum direct band gap, a high optical absorption coefficient of ∼ 10 4 cm −1 in the visible light region, predicted theoretical power conversion efficiency (PCE) of more than 30%, and constituent elements that are Earth-abundant, cheap, and nontoxic. The major issue for kesterite is a low open circuit voltage (V OC ) due to low minority carrier lifetime [1] and high recombination rates at the heterojunction interface as well as in the bulk. The highest recorded efficiency for a Cu 2 ZnSnS 4 (CZTS)-based solar cell has been reported to be 11% [2]. The highest V OC of 809 mV is demonstrated for ordered CZTS [3]. Regardless of these achievements, the V OC deficit is still large compared to Cu(In,Ga)Se 2 (CIGS)-based solar cells and, therefore, significant research is still needed in this field. Several groups have shown improvements in PCE with elemental intermixing at the absorber/buffer interface by post-annealing process [2,4]. Su et al [5] reported that by post-annealing a complete Cd-alloyed pure sulfide device, which included ITO as a window layer, showed efficiency of 12.6%. Our group's latest study [6] showed that soft air-annealing of CZTS/CdS at 175 • C-200 • C improved device performance mainly due to decreased series resistance (R S ), but annealing at higher temperature degraded the devices due to the formation of a Cu-rich CZTS absorber surface close to the CZTS/CdS interface. Analysis results showed that our CZTS monograin layer solar cells have several recombination routes.
Unfortunately, none of these routes disappeared with air annealing of the CZTS/CdS interface. Therefore, the post-annealing of CZTS/CdS is not a definite solution for reducing interface recombination and solving the V OC deficit. It must also be considered that with post-annealing of CZTS/CdS, diffusion of Cu from the absorber to the buffer layer is possible, which increases the photoconductivity of CdS and increases the R S of devices [7,8]. Another option to reduce recombination and boost V OC is a Cd ion soaking treatment, which is also used in CIGS technologies [9][10][11]. Results have shown that the Cd ion soaking process facilitates Cd diffusion into the CZTS surface [10,12], increases the minority carrier lifetime and reduces the non-radiative recombination [10], and decreases the R S of the CZTS devices [9].
An enhancement in the performance of CZTS monograin layer solar cells has been achieved in several ways. For example, the addition of a low concentration of Ag (1% from Cu) into Cu 2 (Cd,Zn)SnS 4 lattice led to increased V OC values of the devices and improved the efficiency from 6.6% to 8.7% due to the reduction in non-radiative recombination and/or interface recombination, but the V OC deficit remained unsolved [13]. The influence of order-disorder in CZTS monograin powders was investigated in the study [14]. The results showed that the low-temperature thermal treatments increased the degree of Cu-Zn ordering and devices showed increased V OC values up to 784 mV and CZTS monograin layer solar cell efficiency of 9.1%. Nevertheless, the V OC deficit still remained large compared to corresponding band gap values (E g ). A photoluminescence (PL) study showed that reduced Cu-Zn disordering of the crystals led to increased band gap energy by about 100 meV and a change in recombination type from band to tail to band to impurity type. The latter involves deep trap (∼200 meV) related recombination and reduces the positive effect of the band gap increase to the V OC in the least disordered material. Moreover, the performance of CZTS monograin layer solar cells was effectively improved up to 9.4% by using oxidative chemical treatment before the formation of the heterojunction [15]. The effect was attributed to the formation of a passivation layer of SnO 2 on the CZTS surface. Although all our previous developments have shown good progress, the main issues have remained.
In this study, we applied a two-step process in the heterojunction formation to reduce the interface recombination and improve the junction quality, and the current-voltage (I-V) measurements showed device efficiency of up to 11.7%. The serial resistance of the CZTS monograin layer solar cells was significantly reduced from 2.4 Ω cm 2 [6] to 0.67 Ω cm 2 . In this work, we analyze the influence of Cd diffusion into the surface of CZTS and replacement of Cu-doped CdS by a fresh CdS buffer layer on the electrical behavior of the highest-efficiency CZTS monograin layer solar cell by temperature-dependent I-V characteristics.

Experimental
In this work, Cu 2 ZnSnS 4 monograin powder was synthesized from binary compounds CuS, ZnS and SnS with purity of 99.999% in the liquid phase of potassium iodide (KI) flux material in a sealed quartz ampoule at 740 • C. The composition of the monograins was analyzed by energy dispersive x-ray spectroscopy. The bulk of the crystals had average composition ratios of Cu/(Zn +Sn) = 0.91 and Zn/Sn = 1.08. The standard post-growth processes in use in our laboratory were applied. Post-growth chemical etching was done with 1 vol% Br in methanol for 5 min followed by 10 wt% Potassium cyanide (KCN) aqueous solution for 5 min at room temperature. After the etching process, the CZTS powder was heat-treated in a sealed ampoule in a dual temperature zone furnace at 840 • C in S vapor (p S = 2050 Torr) for 60 min.
The highest-efficiency CZTS monograin layer solar cell (hereafter named as CZTS-modified) was prepared by using a two-step process (scheme of the combined two-step process in figure 1) for formation of heterojunction between the CZTS and CdS. In the first step, CdS was deposited by a chemical bath deposition process on powder crystals followed by annealing in air at 225 • C for 10 min [6]. With this step, Cd diffusion into the surface of CZTS, forming an ultra-thin Cu 2 Zn x Cd 1−x SnS 4 layer with a decreased band gap at the very top of the absorber, is expected. At the same time, Cu out-diffusion into the surface of the CZTS and also into the buffer layer is possible and this could change the properties of the buffer layer. Therefore, after the annealing step, CdS was removed by concentrated HCl. As a second step, a fresh CdS buffer layer without additional post-annealing was deposited on the CZTS crystals. A CZTS monograin layer solar cell with single-step CdS buffer layer deposition was considered as a reference device in this study (hereafter named CZTS-ref). Subsequently, CdS-covered crystals were implemented in the so-called monograin membranes. The membranes consist of a thin layer of transparent, low-shrinkage polymer into which microcrystals are semi-immersed in the middle of the curing process. The front side of the crystals remains intact from the polymer and, after complete polymerization, the membranes are covered with a window layer. More details about preparation of the monograin layer solar cells are described in [6].
Scanning electron microscopy (SEM) and electron beam induced current (EBIC) measurements were performed using a Merlin, Carl Zeiss Microscopy, Germany. EBIC mapping was performed by a specimen The room temperature current-voltage characteristics of monograin layer solar cells were measured with a Keithley 2400 source meter under standard test conditions (AM 1.5, 100 mW cm −2 ) using a Newport Oriel Class A 91195A solar simulator. The temperature-dependent measurements were performed in a closed-cycle He cryostat (Janis CCS-150) using a halogen lamp. The position of the halogen lamp was calibrated to match the current value measured standard test conditions. The external quantum efficiency (EQE) was measured using a computer-controlled SPM-2 prism monochromator. The generated photocurrent was detected at 0 V bias voltage at room temperature by using a 250 W halogen lamp and DSP lock-in amplifier.
Compositional changes at the CTS interface were analyzed by a Kratos Analytical AXIS Ultra DLD x-ray photoelectron spectrometer (XPS). In order to collect the signal of photoexcited electrons, CZTS powder mounted on a rotational sample holder by ultra-high vacuum proof carbon tape was excited by a monochromatic Al Kα x-ray source (1486.6 eV) at 15 kV and 150 W. Hybrid lens mode in conjunction with an aperture slot view (700 × 300 µm) and pass energies of 160 eV and 20 eV were set to collect the survey and high-resolution photoelectron spectra, respectively. After subtraction of the Shirley background, high-resolution experimental peaks were deconvoluted with Voigt functions and a nonlinear squares fitting algorithm in order to determine core-level peak positions.

XPS analysis
The surface composition of the CZTS powder after the first CdS deposition process followed by air-annealing at 225 • C was investigated by XPS. Figure 2(a) shows the XPS survey spectrum of three different samples: I-the surface of the CZTS powder after removing the annealed CdS buffer layer (blue spectrum), II-the surface of the Cu 2 (Zn 0.8 Cd 0.2 )SnS 4 powder as a proof of Cd inclusion in the lattice (purple spectrum), and III-fresh CdS on top of CZTS without annealing (orange spectrum). Cu 2 (Zn 0.8 Cd 0.2 )SnS 4 powder was also investigated in the previous study [13]. Figure 2(b) shows the high-resolution spectra of the Cd 3d core level. The peaks of Cd 3d 3/2 and Cd 3d 5/2 , which are located at 412.2 and 405.4 eV, can be assigned to Cd 2+ in CdS. The Cd 3d peaks at 411.7 and 404.9 eV correspond to Cd 2+ in Cu 2 (Zn x Cd 1−x )SnS 4 solid solution. The latter peak positions for Cd 2+ in solid solution are in good agreement with literature reports [16,17]. This gives us a solid proof that after CZTS/CdS air-annealing an ultra-thin Cu 2 Zn x Cd 1−x SnS 4 layer on CZTS crystals was formed. The signal of O 1s and C 1s in every survey spectrum is due to atmospheric contamination.

Device characterization
The room temperature I-V characteristics and EQE of the studied CZTS monograin layer solar cells are shown in figures 3(a) and (b), respectively. The black curves in both figures belong to the monograin layer solar cell based on CZTS covered by a non-annealed CdS buffer layer as a reference device in this study (CZTS-ref). The blue curves belong to the CZTS monograin layer solar cell prepared by using a two-step process for formation of a heterojunction between the CZTS and CdS (CZTS-modified). The combined process for heterojunction formation slightly improved V OC values from 723 mV to 735 mV, the main improvement was in the current density values (J SC ) from 20.5 to 23.6 mA cm −2 , and in fill factor (FF) from 59.4% to 67.7%. The serial resistance of the monograin layer solar cells based on the CZTS-modified absorber was significantly reduced from 2.4 Ω cm 2 to 0.67 Ω cm 2 . As a result, enhanced device performance  with PCE of 11.7% was obtained. PCE is calculated by using an image analysis tool for the active surface area (yellow area ∼75% of total contact area), which is estimated from the EBIC top view image ( figure 4(a)).
EBIC allows imaging of the collection efficiency of charge carriers in the device. In EBIC, the electron beam illuminates a nanometer-scale region in the semiconductor specimen and generates electron-hole pairs. A portion of them are collected at the device contacts, giving rise to a photocurrent. By scanning the electron beam, a two-dimensional map of collected photocurrent in the device is constructed [18]. The EQE spectrum was used to study the spectral response of the CZTS monograin layer solar cells. Figure 3(b) shows the normalized EQE spectra of monograin layer solar cells based on the CZTS-ref and CZTS-modified heterojunction. The absorption edge shifts to longer wavelengths with addition of Cd into the CZTS surface layer. The effective band gap energy (E g * ) of the absorber material was evaluated from the linear segment of the low-energy side of the construction (EQE * E) 2 vs. E curves. A slight decrease of E g * from 1.566 eV to 1.548 eV was found, which might be due to the formation of an ultra-thin Cu 2 Zn x Cd 1−x SnS 4 layer on the CZTS crystals [2]. To understand in more detail the electrical behavior of the highest-efficiency CZTS monograin layer solar cell, the temperature-dependent current-voltage characteristics were analyzed.

Temperature dependence of I-V characteristics
The dark and illuminated I-V curves of the CZTS-modified device measured at different temperatures are presented in figures 5(a) and (b). The solar cell parameters obtained from direct measurements are presented in figures 6(a) and (b). The maximum efficiency was achieved at T = 270 K mostly due to higher FF and V OC . The calculated efficiencies are not calibrated because we used a halogen light source. The distance between the lamp and the device was set to match the short circuit current density value from RT I-V measurements under the solar simulator. All solar cell parameters gained from the fittings of temperature dependence measurements correspond to the total area of the device.
The temperature dependence of V OC is often used to determine the dominant recombination processes in solar cells. It is known that at temperatures close to room temperature the V OC (T) curve shows a linear part and the extrapolation of this line to 0 K results in the activation energy Φ [19][20][21]. In the case of Φ < E g , the dominant recombination is related to the interface recombination. However, in our case, the V OC (T) curve shows only a limited number of points in the V OC (T) curve, which show a linear trend. This is an indication that the activation energy Φ ≈ 1.3 eV is probably not completely correct.
In order to get additional information about solar cell properties we performed a fitting of all I-V curves. Dark curves were fitted using a double diode model: and illuminated curves were fitted using a single diode model: where J 0 = J 00 exp is the saturation and J L the photogenerated current densities. The expression kT is the thermal energy where k is the Boltzmann constant and T the absolute temperature. The diode ideality factor n, series resistance R S and parallel shunt resistance R SH account for the non-ideal behavior of the solar cell. Here J 00 is an only weakly temperature-dependent pre-factor, Φ B is the activation energy of the saturation current. All the fittings were done using an algorithm proposed in [22] and ideality factors were not fixed during the fittings. An example of this fitting for T = 300 K and T = 120 K is given in figures 7(a)-(d).
The dark curves in figures 7(a) and (b) show typical non-ideal solar cell double diode behavior, where the first diode has so-called 'normal' behavior and the second diode has a very high value of ideality factor at different temperatures related to different recombination processes. At room temperature n 1 and n 2 have values 1.62 and 3.7, respectively. Such a high value of the ideality factor for the second diode represents a tunneling component, which has a similar exponential dependence to a diode. It is often called a 'weak' diode [23]. Therefore we do not present parameters for this diode and show only results for the first 'normal' diode.
The results of light curve fitting in figures 7(c) and (d) show that our cell has quite large current loss due to parallel shunt resistance R SH at lower temperature (T = 120 K), while at room temperature due to the relatively high value of R SH only diode current can be used in the fitting. At higher voltages, the role of series resistance R S is also seen. Both resistances have a certain temperature dependence; see figures 8(a) and (b). R S shows a simple exponential dependence for both light and dark: with activation energies E a = 59 meV and E a = 99 meV for light and dark curves, respectively. In most cases this activation energy is assigned to the back contact barrier height. The difference between activation energies for the dark and light cases is an indication of a small photoconductivity. More complicated behavior can be observed for R SH in dark and light; see figure 8(b). At higher temperatures, R light SH of the light curves increases almost linearly with temperature and therefore the current loss due to R light SH is smaller at higher temperatures. This ohmic behavior of R light SH can be related to ohmic shunts due to pinholes in the polymer layer between crystals or via low-resistance grain boundaries, because some monograins are formed from separate crystal blocks. At low temperatures, R light SH increases again, but this exponential increase has a different origin related to thermal activation of the semiconductor layer. It has an activation energy E a = 19.9 ± 0.5 meV. We assume that this rather small activation energy is related to some shallow trap level in the CdS buffer layer. However, the R dark SH values obtained from the dark curves are almost 1000 times higher than the values obtained for the light curves. While the decrease of R light SH with illumination is also observed in  thin film solar cells [24], in our case the decrease is larger than expected. This means that the shunting current density is strongly influenced by the illumination and is probably related to the photoconductivity of the buffer layer. The CdS layer is around the crystals when the grains are embedded into the polymer layer. In the back contact formation process, it will be removed from the back side by etching and mechanical polishing, but a low probability remains that the CdS layer can be in touch with the back contact.
The temperature dependence of the diode ideality factor n usually gives valuable information about possible recombination mechanisms in solar cells [25]. In the case of tunneling-assisted interface recombination, it was shown that the temperature dependence of the ideality factor is given by [26]: Figure 9. Diode ideality factor n as a function of temperature and fitted curves for the tunneling-enhanced interface recombination model (equation (5)).
where the characteristic tunneling energy E 00 is a function of the net acceptor concentration N A : where ℏ-Planck constant, q-the elementary charge, ε-the semiconductor's dielectric constant, m * -the effective tunneling mass, and where ω p , ω n and ε p , ε n are the space charge region widths and dielectric constants of the absorber and the buffer, respectively. The temperature dependence of n and the fitting result with equation (5) are given in figure 9.
According to the fitting results, the dominant recombination in our solar cell is indeed related to tunneling-enhanced interface recombination with tunneling energy E 00 = 68 meV. The same model is also true for dark curves, where the tunneling energy is E 00 = 29 meV; however, the parameter α has a value of 0.868. This means that part of the space charge region extends into the buffer layer; see equation (7). The tunneling energy for light curves is slightly higher than the tunneling energy measured in thin film CZTSSe cells (E 00 = 41.5 meV) [27] and is an indication that the carrier concentration in our CZTS is quite high; see equation (6). Therefore, the series resistance R S of our cell is also fairly low, i.e. around 0.7 Ω cm 2 at room temperature.
More information can be obtained by examining the temperature dependence of saturation current density J 0 . From equation (2) for J 0 , the following expression can be derived: By reorganizing equation (6), we obtain n ln (J 0 ) = n ln (J 00 ) − Φ B kT .
A plot of n ln (J 0 ) versus 1/ kT should yield a straight line with a slope corresponding to the activation energy of the saturation current Φ B ; see figure 10.
The obtained activation energy values of the saturation current are lower than the estimated CZTS room temperature band gap energy E g = 1.548 eV. It is known that values Φ B < E g would indicate interface recombination as the dominant mechanism due to Fermi-level pinning or band gap narrowing at the interface. It is interesting that the temperature dependence of V OC provided lower activation energy Φ = 1.3 eV; see figure 6(a). It is known that as soon as tunneling becomes important, the ideality factor n becomes temperature-dependent inducing nonlinear terms into V OC vs T dependence and therefore, even for relatively high temperatures, the temperature dependence of V OC may not give an accurate activation energy.