Introduction
In 2013, Chirilă et al. [
1] of the Swiss Federal Laboratories for Materials Science and Technology (Empa) achieved an energy conversion efficiency of 20.4% on a polyimide (PI)-substrate-based Cu(In,Ga)Se
2 (CIGS) solar cell by alkali fluoride (NaF and KF) post-deposition treatment (PDT), which is excellent progress for CIGS-based solar cells. This world-record-setting device, fabricated by the low substrate temperature process, was not only the most efficient CIGS-based solar cell, but also equaled the champion cell efficiency of polycrystalline silicon-based solar cells. As researchers continued to study the PDT process, several institutes rapidly boosted the best device efficiency. In the past three years, the efficiency increments that have been achieved are even larger than those made during the prior 15 years, thus establishing a milestone in progress. In June 2016, Jackson et al. [
2] of the Center for Solar Energy and Hydrogen Research Baden-Württemberg (ZSW) set the current world record for efficiency for CIGS-based solar cells, of up to 22.6%, by using heavy alkali element fluoride RbF-PDT treatment.
In 1993, Hedström et al. [
3] found that the crystalline structure and device performance of CIGS absorber developed on soda lime glass (SLG) substrate is significantly better than that developed on borosilicate glass. The reason for this phenomenon is the incorporation of sodium (Na), which increases the open-circuit voltage (
VOC) and the fill factor (FF), and therefore increases device efficiency [
4]. In 1994, Holz et al. [
5] revealed that no matter whether the Na was diffused from SLG or introduced by other methods, the material conductivity increased only when the atom density of Na approached 10
15 cm
−3. In 1997, Granata et al. [
6] suggested that the device performance deteriorated as the Na concentration approached 1 at%. The optimal concentration was in the range of 0.05 at% to 0.5 at%. In the same year, Contreras et al. [
7] from the National Renewable Energy Laboratory unprecedentedly investigated doping with other alkali elements, and revealed that potassium (K) and cesium (Cs) also improved the device
VOC, but that Na improved it the most. In 2005, Rudmann et al. [
8] from Empa improved the efficiency of CIGS devices on PI substrates up to 14.1% by doping Na on absorbers with PDT treatment. In 2013, as mentioned earlier, Chirilă et al. [
1] from Empa demonstrated that the KF-PDT process could improve the conversion efficiency of flexible CIGS devices on a PI substrate to 20.4%.
It is commonly agreed that alkali element doping passivates the defects at the p-type CIGS absorber surface or at grain boundaries. It does not change the acceptor concentration, but decreases the compensating donor concentration [
9]. As the free carrier density is determined by the difference in acceptor and donor concentration, the p-type carrier concentration increases; as a result, the Fermi level (
EF) is lowered. Thus, an enlarged
EF differential will produce higher
VOC and FF. In this article, we summarize some outcomes achieved by alkali incorporation, such as surface chemical composition and electronic structure variations, and analyze the reasons behind the boost to CIGS-based solar cell efficiency by the induction of alkali PDT.
Influences of incorporation strategies on Cu(In,Ga)Se2 absorbers
There are three commonly adopted alkali incorporation methods: pre-deposition, co-evaporation, and post-deposition. Pre-deposition means that the alkali element diffuses from an alkali-containing substrate. Co-evaporation means that the alkali element is evaporated during the absorber deposition. Post-deposition means that the absorber is deposited first and then annealed with the alkali element (also called the PDT method).
Pre-deposition incorporation
It is commonly recognized that diffusing Na from the SLG substrate during CIGS evaporation, without intentional incorporation, is a good Na doping method. However, for large-scale SLG, the Na content uniformity is insufficient and deteriorates the module efficiency. The solution is to deposit a thin layer of aluminium oxide (Al
2O
3) or silicon nitride (Si
3N
4) on SLG to prevent elements in SLG from diffusing into deposited CIGS absorbers. Next, Na is distributed uniformly via a method such as sputtering a Na-containing molybdenum (Mo) (MoNa) back contact, which acts as a Na source. Another method is to deposit a NaF layer (NaF precursor), with a certain amount on the Mo back contact to ensure a sufficient supply of Na during the absorber deposition. NaF is stable, and the residual fluorine ion (F
−) reacts with selenium (Se) to form vaporous SeF
6 that can desorb from the absorber [
10]. The Na doping is controlled by the NaF layer thickness. Insufficient NaF amount results in low
VOC and FF, and an excess amount leads to peeling-off problems during absorber deposition. Moreover, an excess amount of NaF also deteriorates the solar cell performance [
11]. Due to the good controllability of the NaF precursor, this method has been commonly adopted.
Salomé et al. [
12] compared the performance of CIGS-based solar cells prepared with different Na supply methods, including using SLG, NaF precursor, and a MoNa layer. According to Table 1 [
12] (where cell parameters are mean values), Sample 3, which is a Na-free device, shows the lowest
VOC and FF. Thus, the average conversion efficiency is only 8.8%. Of the rest of the samples with pre-deposition incorporation methods, the NaF precursor sample (Sample 4) shows the best performance, with a significantly higher
VOC and FF than the other devices. The second-best performer is the SLG sample (Sample 1), and the worst device among the Na-doped samples is the sample with the MoNa structure (Sample 5). This sample is doped with Na diffused from the Mo back contact layer. The lowest device parameters indicate the lowest doping content. In addition, Sample 2, which was prepared with a MoNa layer on SLG, shows inferior performance compared with Sample 1. We deduce that the MoNa layer blocks the diffusion of Na from SLG to the absorber.
Co-evaporation incorporation
In 2010, Güttler et al. [
13] prepared CIGS absorbers on PI substrate using the three-stage co-evaporation process, and incorporated Na by co-evaporation during the first, second, and third stages, respectively. They compared the film structures with Na-free CIGS film (Fig. 1) [
13].
The scanning electron microscope cross-sections in Fig. 1 show that incorporating Na during the first and second stages deteriorates the CIGS crystallinity. Fine grains were found at the film surface and back interface. However, incorporation at the third stage only slightly affects the crystallinity at the film surface, and the corresponding solar cell shows the best performance. The Na-free sample shows larger grain sizes with no stratification. By analyzing the secondary ion mass spectrum (SIMS) profiles (Fig. 2) and comparing the results with Fig. 1, we deduced that gallium (Ga) accumulates at the Cu(In,Ga)Se2 (CIGS)/Mo interface if Na is incorporated at the first or second stage. These fine grains were composed of high-Ga-content CIGS or CuGaSe2 (CGS). For the samples in which Na co-evaporated at the third stage, fine grains were located at the film surface, and the Ga gradient was similar to that of Na-free samples.
It is now commonly accepted that Na decreases the interdiffusion of indium (In) and Ga, and results in the stratification of CIGS films. In order to understand how Na influences the interdiffusion of In and Ga, Brummer et al. [
14] studied the phase transformation during the selenization of Cu/In/Se stack precursors with and without Na by high-temperature X-ray diffraction (HTXRD). It was found that without Na, the Cu
2–xSe phase could be detected until the substrate temperature rose to 275 °C. However, with Na, the detectable temperature decreased to 225 °C. This phenomenon proves that Na facilitates the formation of Cu
2–xSe at lower temperatures. Therefore, we can deduce that whatever NaF is deposited in the first or second stage facilitates the formation of Cu
2–xSe. Consequently, as In
2Se
3 is preferred to react with Cu
2–xSe, the formation of CuInSe
2 (CIS) is preferable to the formation of CuGaSe
2. The residual Ga
2Se
3 is left at the Mo/CIGS interface and forms small grains. This is the fundamental cause of the film stratification, which is not due to the decreasing interdiffusion of In and Ga. In the three-stage co-evaporation process, after the Cu/Se co-evaporation at the second stage, CIGS and Cu
2–xSe both exist in the films. In the third stage, NaF facilitates the formation of CIS, and Ga will still be left at the edge of the films. That is why fine grains are located on the film surface at the third stage. As the layer deposited at the third stage is only a small fraction of the whole film, the fine grain layer is quite thin. The whole film structure is similar to that of the Na-free sample.
We have developed CIGS absorbers on PI/Mo substrate, and incorporated Na during the first, second, and third stages, respectively. The film structures were quite similar to those in Fig. 1. All the film carrier densities increased by 1–2 orders of magnitude compared with the Na-free samples. Thus, the film sheet resistivity decreased by 1–2 orders of magnitude (Fig. 3). The sample with co-evaporated NaF at the third stage showed the best performance; this is because Na has little influence on crystallinity, but rather passivates defects at the film surface or grain boundaries. Excellent crystallinity and passivated surface and grain boundaries are positive factors for higher free carrier concentration.
Post-deposition incorporation
PDT, which involves the incorporation of alkali elements (Na, K, Rb, Cs) into the CIGS layer under the Se atmosphere after absorber growth, has a wide tolerance for doping level. The PDT process passivates defects without affecting crystallinity; therefore, it improves the electrical properties of CIGS-based solar cells. The substrate temperature for alkali PDT is very important. We have studied CIGS films with NaF-PDT at different substrate temperatures, as shown in Fig. 4. It was found that the sheet resistivity decreased and the carrier concentration improved as the substrate temperature increased. The carrier concentration was saturated at two regions: 250–350 °C and 400–450 °C. From 350 °C to 400 °C, the carrier concentration increased and the resistivity decreased correspondingly. Therefore, a substrate temperature of around 400 °C is commonly adopted for NaF-PDT. In the literature, a substrate temperature between 350 °C and 400 °C is chosen for alkali PDT.
Chirilă et al. [
1] fabricated CIGS absorber layers on PI films by co-evaporation; the corresponding solar cell achieved 20.4% conversion efficiency. X-ray photoelectron spectroscopy (XPS) measurement revealed that the surface composition of the CIGS absorber changed with (NaF+ KF)-PDT: The Ga concentration decreased obviously and copper (Cu) was depleted. The depletion depth of Cu and Ga was less than 30 nm at the CIGS surface. The fact that the majority of Na was substituted by K indicated that K was preferred to Na to exist in the film.
Effects of (NaF+ KF)-PDT on CIGS device characteristics
Researchers from ZSW improved their solar cell performance significantly by using KF-PDT (Table 2) [
2]. The diode quality factor decreased by 12%, the reverse saturation current density (
J0) decreased by about one order of magnitude, the solar cell efficiency improved by 6.9%, the
VOC approached 750 mV, and the FF reached over 79%.
As shown in Table 3 [
15], compared with the champion cell made in 2013, the new champion cell made in 2014 is improved in every parameter: The
JSC is improved by 4.9%, and the corresponding external quantum efficiency (EQE) is obviously increased between the wavelengths of 400–520 nm (Fig. 5(a)) [
15]. These variations are mainly attributed to the improvement in the CIGS absorber quality by (NaF+ KF)-PDT. The CdS buffer layer is thinner but denser, so the short wavelength absorption is better and the
JSC has increased. In addition, the cell made in 2014 clearly presents a steeper Ga gradient and a smaller bandgap minimum than the cell made in 2013, according to the SIMS profiles (Fig. 5(b)). Thus, the EQE of the champion solar cell fabricated in 2014 increased significantly in the range over 1000 nm compared with its predecessor, which is another reason for the increase of
JSC. Although such a steep Ga gradient may bring in more lattice defects, however, the
VOC or the FF does not decrease compared with the previous champion cell, and the
JSC is even able to increase because the minimum bandgap is decreased. It can be deduced that the PDT process passivated the structure defects caused by the steeper Ga gradient. From Table 2 [
2], we can see that the shunt resistance,
Rsh, as well as the
J0 of the solar cells were both significantly optimized. This indicates that the alkali PDT significantly reduced defects at the heterojunction.
Effects of heavy alkali elements (Rb and Cs) on output characteristics of CIGS-based solar cells
In July 2016, ZSW announced that a CIGS-PDT using the heavy alkali elements rubidium (Rb) and Cs, instead of K, had increased the solar cell efficiency to 22.6%, thereby creating a new world record [
2]. The efficiency of solar cells with heavy alkali PDT increased by 6.78% on average, an improvement that was mainly attributed to the increased
VOC and FF, as shown in Fig. 6 [
2] (without antireflection coating). The average efficiency of solar cells with (Rb or Cs)-PDT is 2.42% higher than that of KF-PDT solar cells. The primary cause is the reduced diode quality factors (Fig. 7) [
2]; that is, the diode characteristic of the cells is improved. This indicates that the heavier alkali elements are more effective than K in promoting cell efficiency.
Our team has also started research on the incorporation of heavy alkali elements. Here, we present part of our recent results. We developed four CIGS absorbers in a single run in order to maintain the uniformity of the experimental conditions. Three of the absorbers were post-deposition and were treated with different CsF contents. The last absorber, without any PDT treatment, served as a reference. The doping level was determined by the evaporation temperatures of the CsF. In this study, these temperatures were 320 °C, 290 °C, and 260 °C, respectively. The other experimental conditions were kept the same. As shown in Table 4, the samples with CsF-PDT show a clearly higher VOC and FF; meanwhile, the JSC shows little change because the thickness of CdS is not optimized. As shown in Fig. 8, the doping level of CsF plays a significant role in improving the VOC and FF, and thus in improving the conversion efficiency.
To analyze the cross-section composition, Jackson et al. [
2] measured the CsF-PDT samples by SIMS (Fig. 9). It was observed that the heavy alkali element substituted for the doped lighter alkali (Na, K), and that Cs was preferable over Na and K to exist in the CIGS films. At present, the mechanisms of alkali ion exchange and of where the substituted metal atoms (such as depleted Cu and Ga) move are still unclear.
Influence of the alkali PDT on CIGS surface and device performance
JSC improvement with thinner CdS buffer layer
Chirilă et al. [
1] analyzed the chemical composition of the CIGS surface with (NaF+ KF)-PDT by XPS. It was revealed that Na was substituted by K, while Cu and Ga were depleted within less than 30 nm depth at the CIGS surface. Thus, the acceptor defect (
) density increased in the original ordered vacancy compound (OVC) region of the CIGS surface. During the deposition of CdS film by chemical bath deposition (CBD), the cadmium (Cd) ions diffused to a certain extent into the surface of the absorber material to occupy the
accepters and form Cd
Cu donors. As the donor density was higher than that of the sample with no PDT, the surface inversion was more intense, and thus strengthened the buried homojunction. Therefore, it was not necessary to use a thick buffer layer to form a high-quality CIGS/CdS heterojunction. In addition, for samples with KF-PDT, the CdS layer presented good coverage and uniformity at an early stage of CBD (Fig. 10) [
15], so the CdS film could be thinner, which benefited the short wavelength absorption and thus increased the
JSC.
Increased hole concentration at the near-surface of the CIGS
In 2014, Pianezzi et al. [
9] prepared CIGS absorber layers on PI substrates using the three-stage co-evaporation process. Capacitance-voltage (C-V) measurement was applied to characterize the hole concentration of the CIGS absorbers with NaF-PDT, KF-PDT, (NaF+ KF)-PDT, and no PDT (Fig. 11). The results show that for the PDT samples, the hole carrier concentration at the CIGS near-surface increased significantly compared with the no-PDT sample. The carrier concentration of the NaF-PDT sample is half an order of magnitude higher than that of the (NaF+ KF)-PDT sample, which has a carrier concentration that is slighter higher than that of the KF-PDT sample.
The CIGS absorber surfaces of high-efficiency cells are composed of OVC. A large amount of shallow acceptor defects ( ) and donor defects ( ) form neutral defect pairs (2 + ), and NaF-PDT is able to decompose ( + ) defect pairs. Na replaces the donor defects ( ) to form neutral defects (NaCu) in In-rich film, and thus improves the hole carrier concentration in the CIGS layers. Na can also occupy the In vacancy (VIn) to form the acceptor defect ( ), which is shallower than the acceptor; thus, it further improves the absorber hole concentration. In addition, at the film surface will absorb O2 to transform Se vacancy (VSe) into OSe (Se substituted by O), and will reduce the compensating donor concentration. In general, Na increases the hole concentration in the films for many reasons.
Wider surface bandgap and lower valance band
Handick et al. [
16] provided a depth-dependent band structure measurement at the surface of CIGS absorbers deposited on PI substrates by using ultraviolet photoelectron spectroscopy (UPS), synchrotron-based hard X-ray photoelectron spectroscopy (HAXPES), and inverse photoelectron spectroscopy (IPES). The UPS spectra (He I excited, 21.22 eV) are the most surface sensitive, with an inelastic mean free path (IMFP) of approximately 0.6 nm. By using excitation energies of 2 keV and 8 keV, the values of the IMFP for the HAXPES measurements were increased to 4 nm and 12 nm, respectively. According to the above measurements, three different values of the absorber surface valence band maximum (VBM) were derived. IPES can be used to determine the conduction band minimum (CBM). In Fig. 12 [
16], the near-surface bandgap for the NaF-PDT sample was 1.61 eV (+0.14/−0.51 eV). For the (NaF+ KF)-PDT CIGS sample, the near-surface bandgap was 2.52 eV (+0.14/−0.51 eV), with the VBM showing a pronounced downward shift from the
FF. Fig. 13 shows the bandgap structure diagram that was calculated according to these experimental data.
Handick et al. [
16] proposed that the enlargement of the bandgap at the surface was due to the composition change, which varied the electrical structures. The characterization demonstrated that the surface region was mainly composed of K, In, and Se, and that the bandgap of the In
2Se
3 compound was pronouncedly thickness dependent—that is, the bandgap exceeded 2.5 eV when the thickness decreased to 2.6 μm. The bandgap of KInSe
2 was 2.7 eV. It is not possible to distinguish which compound it is, based on the current bandgap data. The two compounds may co-exist, but the ratio is not known yet.
Derived from the data of Fig. 12 [
16], the bandgap diagrams of the CIGS/CdS heterojunction with NaF-PDT and (NaF+ KF)-PDT are shown in Fig. 13. The p/n inversion of the CIGS/CdS heterojunction has entered the near-surface region of the CIGS. Holes become the minority carrier at the interface, and the hole concentration determines the magnitude of the interface recombination. As shown in Fig. 13, the VBM of the (NaF+ KF)-PDT sample is noticeably shifted downward, more than that of the NaF-PDT CIGS, and the hole concentration is decreased to a great extent. The interface recombination is decreased further, and the
VOC and conversion efficiency are obviously improved.
Conclusions
In the semiconductor industry, the use of the alkali element Na is a major issue that affects device stability and leads to device failures. Its contamination of silicon wafers must be strictly controlled. In the 1990s, researchers discovered that the alkali element Na could improve the efficiency of CIGS-based solar cells without negative effects on device stability, and re-recognized the mechanisms of how Na acts on multiple compound semiconductor devices. CIGS is a quaternary polycrystalline compound with a conductive mechanism that is totally different from those of elemental semiconductors such as Si and Ge. The conductive type of acceptor or donor depends on the defect structures arising in non-stoichiometry. Alkali elements can change the defect structures via an ion exchange mechanism that is induced by the ion element electronegativity, resulting in the transition of the conductive type. In general, the carrier concentration of the absorber layer can increase, no matter how the alkali element is doped. However, different doping methods will affect the polycrystalline structures of the absorber layer and influence the efficiency.
Alkali PDT can passivate lattice defects without changing the structure of CIGS, and can reduce the diode quality factor and the reverse saturation current so as to improve device performance. Moreover, it has a wide tolerance for the doping amount of the alkali element, especially for the heavy alkali elements, thus contributing to surface structure improvement without changing the CIGS lattice structure. The lower electronegativity of the heavy alkali elements changes the stoichiometry at the CIGS surface, increases the concentration of the acceptor defect ( ), causes more Cd ions to fill in the acceptor defects ( ), and increases the concentration of the donor defect ( ). The surface electrical inversion is stronger, and strengthening the buried homojunction at the CIGS surface. In addition, heavy alkali PDT improves the CdS coverage, allowing the CdS to be thinner in order to increase the short wavelength absorption. This is an important factor in increasing device efficiency. Heavy alkali PDT also broadens the CIGS surface bandgap and lowers the valence band maximum of the CIGS surface, while the lower hole concentration further suppresses carrier recombination at the interface. This is another important reason for the improvements in VOC and conversion efficiency.
The use of heavy alkali PDT has resulted in breakthrough progress for CIGS-based solar cells, and significantly decreases the efficiency gap between thin-film solar cells and single crystalline silicon-based solar cells. The European Union’s Sharc25 project (super high-efficiency Cu(In,Ga)Se2 thin-film solar cells approaching 25%) demonstrates that the CIGS surface and interface characteristics have a major impact on improving cell efficiency up to 25%; therefore, more innovations and developments in this field are expected in the near future.
2017 THE AUTHORS. Published by Elsevier LTD on behalf of the Chinese Academy of Engineering and Higher Education Press Limited Company. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
PDF
(2517KB)
