Time evolution of the Mo ageing
The typical CIGS film was 2 µm thick with x close to 0.40 ± 0.02 and y close to 0.92 ± 0.02. There are small variations of the composition between each run, but they are within the X-ray fluorescence measurement error. In Figure 1, the median solar cell parameters are presented. For reference, in Table 1, the relative differences in the parameters between the solar cells from reference and the aged Mo layers are shown.
Table 1. Difference of the electrical parameters between the aged Mo devices and their references.
|ΔAge (days)||ΔVoc (mV)||ΔJsc (mA/cm2)||ΔFF (% absolute)||Δη (% absolute)|
|Average without 28 days outlier||13||–1.5||1.3||0.5|
For the fresh Mo, the efficiencies of the solar cells vary between 14.2% and 16.2%. This variation is slightly higher than we normally observe and was caused by the reference cell used in the 28-day test, which was worse than expected, mostly because of a low FF. This was an unidentified and isolated problem. It may be connected with a more resistive ZnO : Al layer or more resistive grid, but because it was an outlier, it was not further studied. For comparison purposes, this value will not be considered. For the aged Mo, the efficiencies vary between 15.5% and 16.9%. Although the values mentioned before are medians, the variation of the most efficient cells of each sample follow the same trend, between 15.5% and 16.9% for the reference cell and between 15.7% and 17.8% for the cells with aged Mo. Graphs showing J–V and EQE curves of typical cells are presented in Figures 2 and 3, respectively. In both cases, the reference cell has an efficiency of 16.7%, and the aged Mo one presents an efficiency of 17.2%. The EQE plot shows no significant difference in the cells.
Although the efficiency difference is small, it is still significant. The cells with the aged Mo have better median efficiencies than their reference fresh counterparts. Most of the films follow the similar tendencies in the other parameters, that is, an increase of the Voc and FF and a small decrease of the Jsc for the cells with the aged Mo. By considering these variations and by having in mind that the position of the samples was randomised, it is possible to state that the aged Mo provides a small but beneficial effect to the cell performance. It is also interesting to observe that there is no apparent correlation between the length of the ageing period and the evolution of the parameters. As it is shown in Table 1, the ageing of the Mo results, in average, in an increase of the Voc by 12.5 mV, a decrease of the Jsc by 1.1 mA/cm2, an increase of the FF by 2.4% and an increase of the efficiency by 0.8%, absolute percentages. These differences are also visible in the J–V plot shown in Figure 2. To have a better validity of these results, this experiment was reproduced using two other set of samples, and the observed results were similar, that is, a small improvement of the efficiency but with no correlation with the length of the Mo ageing after the initial 18 h. The best cell prepared in this set had an efficiency of 17.8% without AR coating, and it was achieved with a 35-day-aged Mo. Removing the 28 days outlier changes the variation of the FF from 2.4% to 1.3%, but it only lowers the difference of the average efficiency from 0.8% to 0.5%, where for the Jsc and the Voc, there is almost no difference, so basically the conclusions are valid with or without this point. Both variations are presented in Table 1, and for consistency purposes, the values without the outlier are going to be considered.
The differences in the Voc and Jsc seen in Table 1 may be explained by a higher carrier concentration in the CIGS layer for the cells with the aged Mo. A higher carrier concentration in the CIGS layer may lead to a higher built-in field between the n-type and p-type sides of the junction and thereby a higher Voc. On the other hand, higher doping in the CIGS part of the junction reduces the depletion layer width; therefore, the current collection may drop. It is well known that carrier concentration in CIGS increases with increasing concentration of Na [24-26]. A correlation between oxygen and Na in Mo layers has also been observed . It is therefore reasonable to believe that the ageing of the Mo layer prior to CIGS deposition would change the diffusivity of the Mo layer for Na from the SLG substrate and thereby change the Na concentration in the CIGS layer. Regarding the increase of the FF with the ageing of the Mo, this can only partly be explained with the increase in Voc. This difference can be explored using the following expression, which correlates FF with Voc :
where voc is the normalised voltage defined as Voc/(AkT/q), A is the ideality factor of the cell, k the Boltzmann's constant, T the temperature and q the electron charge. By using the two different average Voc values of Table 1, one reaches a difference in FF of 0.2–0.3% by using A values between 1 and 2. This is much less than 1.3% averaged difference seen in Table 1; therefore, the increase in the Voc does not explain why the FF increases so much.
Secondary ion mass spectrometry results
In Figure 4, SIMS analysis and a comparison of Na concentration of the CIGS layers shown in Figures 2 and 3 are presented. The left parts of the plots correspond to the CIGS/CdS interface, and the right part of the plots is the Mo/CIGS interface. For both films, the Se and Cu signals are flat all the way through the film, whereas In and Ga show concentration profiles with opposite slopes. There is more Ga at the back part of the film, that is, close to the Mo, and more In towards the CdS interface as intended . The interesting point to note in these profiles is the similarity of the Na concentration. Both samples show the same Na profile and the same Na concentration, from around 1.5 × 1019 atoms/cm3 to 2 × 1019 atoms/cm3. Both samples also show a Na bump near the CdS interface. Closer to the Mo interface, the Na signal is much higher for two reasons. One is an interface SIMS problem, which may be explained by an increased oxygen concentration at the CIGS/Mo interface, and the other is that the Na content in the Mo itself is higher than in the CIGS and that the sputtering induced roughness leads to a smearing at the interface. From the SIMS data, it is hard to find any significant difference between the samples that can explain the differences in cell behaviour. The minor differences in the shape of the profiles lie mostly within the measurement error of the method, but possibly, there is an indication of a higher Na concentration near the surface of the CIGS in the aged Mo sample.
Figure 4. Secondary ion mass spectrometry profile measurements of the (a) reference and (b) aged Mo cell with Mo aged for 35 days. The Na values are calibrated concentrations in atoms per cubic centimetre where the rest of elements are shown in atomic percentage. Note that the scale of the depth axis starts at 0.5 µm, where the Cu(In,Ga)Se2 starts, because the top of the sample consisting of the ZnO and CdS layers has been shifted outside of the graph. The right most part of the graph corresponds to the Mo back contact.
Download figure to PowerPoint
C–V measurements were performed to obtain an estimation of the carrier concentration. The doping profiles for the same cells as the ones used for the SIMS analysis are shown in Figure 5. Both cells are good representatives of the average behaviour of the six cells measured for each sample. Typical non-uniform profiles are observed, as a result of a different response from interface and bulk states . The net acceptor concentration for both cells is between 1 × 1016 and 2 × 1016 cm−3. In comparison with the reference cells, slightly higher concentration is observed for intermediate distances from the junction for the aged Mo cells while the whole profile is shifted by around 20 nm towards the junction. There is also a shift regarding the net acceptor concentration: the curves for the two samples are shifted around 0.5 × 1016 cm−3. If one considers the following equation, Vbi = (q/KBT)ln(/(NA*ND)) , where Vbi is the built-in voltage, q the electron charge, KB the Boltzman's constant, T the temperature, the concentration of the intrinsic carriers, NA the concentration of acceptors and ND the concentration of donors, and assume that between the samples, only the concentration of acceptors changes, then a ratio between the Vbi of the two samples is possible to estimate. For the first approximation, one can use the estimated net acceptor concentration as the concentration of acceptors and estimate the Vbi, which is closely related to the Voc. For the estimated values between 2 × 1016 and 2.5 × 1016 cm−3, = 0.994. If this difference of 0.6% is applied to the average reference Voc, one would have an absolute difference around 8 mV. Considering that for these calculations many approximations are carried out, one can say that the net acceptor concentration alone is not explaining the increase of the Voc for the aged devices, but it may explain part of it.
According to our hypothesis of increased carrier concentration as an explanation of increased Voc and decreased Jsc, one would expect CIGS deposited on aged Mo to have a higher carrier concentration. Keeping in mind the measurement errors associated with C–V measurements, we can tentatively say that the shape of the curves is in accordance with the SIMS profiles with a small increase of Na concentration near the surface of the CIGS for both samples. The aged sample also shows a slightly higher net carrier concentration than the reference sample. This effect would change the Voc in accordance with what we saw before, an increase for the aged samples, but it does not explain the effect totally.
Solar cells were prepared on top of the Mo layers that were oxidised in an oxygen furnace, and the average parameters for 12 cells are presented in Table 2. The J–V plots for a representative from these cells are shown in Figure 6. The oxidised Mo showed a higher Voc, a lower Jsc, a similar FF and an overall 0.2% improvement in the efficiency. Because in this case the FFs are similar for both devices, it is not clear that the benefits given by both studies are coming from the same mechanism. It has been suggested  that water vapour also influences the performance of CIGS devices, and although no water is expected to be absorbed by the Mo, one would need more detailed analysis, such as X-ray photoelectron spectroscopy , to investigate the oxidation states and composition of the surface of the Mo films. Such detailed study was performed by Yoon et al. , but the correlation of this effect with electrical performance was not made.
Table 2. Electrical parameters of a reference cell and of a cell where Mo was oxidised.
| ||Voc (mV)||Jsc (mA/cm2)||FF (%)||η (%)|