ALD-Zn(O,S) buffer layer for CIGS grown by LTPED

CBO value depends from both Zn(O,S) and CIGS compositions, in terms of S and Ga content, respectively. The best results are assumed to correspond about at the same CBO formed by CBD-CdS towards CIGS, i.e. +0.3 eV, and must vary in function of the absorber GGI at the surface.

Zn(O,S) thickness series shows an unexpected FF loss for thinner buffer layers. Sputtering damage and resistive layer thinning are cleared of the trend. A non-uniform covering of the absorber surface could be the cause since clues of nucleation growth started by islands was found in ZnS study.

Reference samples were fabricated depositing CdS by CBD and i-ZnO and Al:ZnO by RF-Magnetron Sputtering, following the same procedure described in chapter 5. . Al contacts were then deposited on the top of samples with CIGS GGI=37.5% by means of thermal evaporation. No metal contact was applied on the top of CIGS-GGI=30% devices. JV and QE measurements were carried out as described in chapter 5. .

Note that neither potassium treatment of CIGS surface nor anti-reflection coating was applied;

moreover, the JV and QE measurements were conducted without any post-fabrication annealing or light soaking treatment of solar cells.

For the series on CIGS with GGI=37.5%, all the CdS references showed almost the same behaviour: the best solar cells of each batch had Jsc=30.5 mA/cm², Voc=645 mV, FF=73.6% and η=14.5%. The composition study tested pure ZnO and Zn(O1-x,Sx) with x from 0.25 to 0.40, at a constant cycles number of 175, i.e. about 40nm-thick films. CIGS with GGI=30% were buffered with Zn(O,S) having S=0.30 and S=0.38, and their CdS reference had Jsc= 31.6mA/cm², Voc=565 mV, FF=69.6% and η=12.4%. The trends of Jsc, FF, Voc and efficiency in function of Zn(O1-x,Sx) composition are shown in Figure 7.19, for the best solar cell of both LTPED-CIGS (blue squares for GGI=37.5% and green circles for GGI=30%).

Figure 7.19 – a), b), c) and d) shows Jsc, FF, Voc and efficiency of the best solar cells from samples buffered with Zn(O,S) onto CIGS having GGI=37.5% (blue squares) and GGI=30% (green circles), and relative CdS references (red squares and

purple spots, respectively) on y-axis.

The JV of solar cells fabricated using CIGS with GGI=37.5% are shown in Figure 7.20. Buffer with x=0.40 gave rise to a strong blocking behaviour (purple curve), while the FF is maximized at x=0.30.

The current is almost steady around 29 mA/cm², but the one corresponding to S=0.33, which was lower.

The Voc grew until about S=0.33 and later dropped. The trends were similar to the series onto Solibro CIGS, shown in section 7.4, but shifted towards lower S content and with a great gap from the CdS reference. The devices current densities are expected to be higher since Zn(O,S) is the buffer, but in this series they were found rather lower than the one of CdS samples. EQE comparison of CIGS-GGI=37.5%

buffered with different Zn(O,S) compositions is shown in Figure 7.21.

Figure 7.20 – JV curves of best solar cells by CIGS GGI=37.5% buffered with CdS (red line) and different Zn(O,S) compositions.

Figure 7.21 – External quantum efficiency measurements of solar cells fabricated with CIGS GGI=37.5% and buffered with CdS (red line) or different Zn(O,S) compositions (other lines). The solar cells were neighbour devices of the best ones, with

almost the same current. The CdS reference sample was measured with a wavelength step of 50nm, while for the Zn(O,S) samples the step was 2 or 4nm and curves smoothing was performed in the 900nm-950nm region to get rid of set-up

interferences.

The optical match between CIGS, Zn(O,S) and TCO resulted in high reflectivity, as highlighted by reflectance measurements. Internal QE was calculated to determine how much current the device is capable to provide in absence of reflection losses. Figure 7.22 compares EQE and reflectivity measurements along with calculated IQE of the samples with x=0.25 (a) and x=0.33 (b). The IQE gain in Jsc for all the Zn(O,S) samples with GGI=37.5% is about 9.5-10.0%, corresponding to 2.6-3.0mA/cm², in respect of EQE. Thus, with a perfect anti-reflection coating, the current density would be increased up to 29.6 mA/cm² for the sample with x=0.33, and up to 31.6-31.9mA/cm² for the other S contents.

-35 -30 -25 -20 -15 -10 -5 0 5 10

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

J (mA/cm2) V (V)

CdS ref S = 0.25 S = 0.30 S = 0.33 S = 0.40

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

300 400 500 600 700 800 900 1000 1100

EQE

wavelength (nm)

CdS S=0.25 S=0.3 S=0.33 S=0.40

The sample with x=0.33 exhibited a poor current even in IQE data, probably due to a lower quality of the interface. Anyways, IQE-Jsc values of the whole series are still lower than expected for Zn(O,S) buffer layer, for example the Solibro Research AB’s CIGS series presented in section 7.4 (33-34mA/cm²).

In the CdS references, the current-density gain calculated by means of IQE is around 6%, resulting in 32.3 mA/cm². With respect to the CdS, IQEs of Zn(O,S) devices are lower in the entire region above 500 nm (where the buffer absorption plays no role). Since the CIGS material was from the same batch, it may be imputed to different depletion width or interface conditions. Considering the Jsc values obtained by IQE, the efficiency of LTPED-CIGS GGI=37.5% devices would increase by the same percentage as the current density enhancement, reaching η=12.4% for the best sample buffered with Zn(O1-x,Sx) (x=0.30) and 15.4% for the references with CdS.

Figure 7.22 - EQE (blue line), IQE (green line) and reflectivity (orange line) measurements of the best solar cell with CIGS GGI=37.5%, buffered with Zn(O0.75,S0.25) (a) and with Zn(O0.67,S0.33) (b). The current loss due to reflection were determined to be 2.9 mA/cm² and 2.6 mA/cm², respectively. EQE smoothing was performed between 900nm and 950nm to get rid of

set-up interferences.

Other than the current lower than expected, the very limiting parameter of solar cells with CIGS-GGI=37.5% was the FF. It was more than 11 points lower than the CdS reference. This variation can be only partially ascribed to the absence of Al contact in the devices with alternative buffers. Low-quality interface may be the main cause. To evaluate the ALD process, a pure ZnO buffer was grown by ALD, with the same number of cycles (175), onto CIGS-GGI=37.5%. Its FF resulted to be 61%, slightly lower than the Zn(O,S) samples but typical for a ZnO buffer layer. Consequent deduction was the ALD process had no role in the FF loss, in terms of contamination, and the CIGS preservation and etching were not deteriorating the CIGS surface. The sulphur interaction with the absorber may affect the buffer growth.

The LTPED-CIGS with GGI=37.5% may nucleate Zn(O,S) differently than CIGS by co-evaporation

previously studied and so, probably, a good matching between the LTPED absorber and Zn(O,S) is not possible at the present ALD growth conditions.

FF and IQE data underlined once again the importance of the interface between absorber and buffer, which is strictly dependant from the materials used and the process applied.

Using the same absorber composition (GGI=37.5%), a thickness series was carried out at the buffer sulphur content x=0.30. Figure 7.23 shows FF in function of Zn(O,S) thickness. Interestingly, no important FF variation was observed, except for the 26nm-thick sample for which FF was 2-3 points higher than the other samples. Voc and Jsc of devices was scattered, their product had a maximum at buffer. Even in this series, the current density was affected by high reflection. Starting with 40nm-buffer and moving towards thicker or thinner Zn(O,S) did not improve the efficiency, so the solar cells with 40nm-thick Zn(O,S) was the best one of the series.

Figure 7.23 – FF vs thickness composition of the Zn(O,S) buffer layer for PED-CIGS with GGI=37.5%.

Alternative-buffered solar cells with CIGS-GGI=30% exhibited different behaviour than the ones with CIGS-GGI=37.5%. Few samples were fabricated for GGI=30% and only two buffer composition were investigated, so it is not possible to look for a trend. The buffer with sulphur content of x=0.38 gave rise to more efficient devices than buffer with x=0.30 (green spots in Figure 7.19), thanks to higher solar-cells parameters, which were even extremely close to or larger than the CdS reference (purple spot). Jsc

measured by EQE were still affected by reflectance issues, but IQE showed an enhancement of about 10.0-10.8%, achieving values of 34.2 mA/cm² in the best performing devices (CdS reference achieved Jsc=33.5 mA/cm² thanks an IQE gain of about 6% with respect to EQE). The FF was extremely close to CdS devices, just 1 point lower for x=0.38. Note that no metal-grid contact was deposited on the top of the TCO, whether the buffer material was CdS or Zn(O,S). The same composition turned out to have a Voc (599 mV) larger than the CdS reference (565 mV), even if CdS devices with higher Voc (590 mV) were found on the same sample but resulted in lower efficiencies due to poor current density.

In the end, the efficiency of the best solar cells on CIGS-GGI=30% and buffered with Zn(O0.62,S0.38) was η=12.8% (14.1% with IQE-Jsc), higher than the CdS reference, η=12.4% (13.1% with IQE-Jsc). The JV curves of these two samples are shown in Figure 7.24. Considering the IQE currents plausible, with

60 65 70 75

0 20 40 60

FF (%)

Thickness (nm)

Zn(O,S) CdS

a perfect AR coating the Zn(O0.62,S0.38)-buffered CIGS-GGI=30% would have an efficiency about 1 point larger than the CdS reference.

Figure 7.24 – JV curves of PED-CIGS with GGI=30% buffered with Zn(O0.62,S0.38) (blue curve) and CdS (red curve).

Current densities are confirmed by EQE measurements. Because of the reflection loss, Zn(O,S) Jsc can be push up by roughly 10% applying an anti-reflection coatings, confirmed by IQE.