Device results

The distribution of efficiency values from 201 separate ‘cells’ (i.e. 5 × 5 mm² contacts) is shown in figure 6.2. Three distinct populations of devices were identified and these were attributed to being under-treated, optimised and over-treated by the CdCl₂ process. In preliminary experiments the shorter CdCl₂anneal times used (∼ 5 mins) typically led to the low device performance associated with under-treatment. However, for devices annealed in the optimum time window (9 - 11 min) it was often found that some parts of a given 5 × 5 cm² plane were visibly degraded. The cell performance parameters, η, F F , J_SC and V_OC are given in table 6.1 and selected J-V and EQE curves, that represent the typical

Table 6.1: Average performance characteristics of typical under-treated, over-treated and optimised device populations. The best device performance achieved is also included.

Figure 6.2: Distribution of efficiency of 201 5 × 5 mm² contacts taken from 12 separately processed sample plates. An average efficiency of 8.6±2.2 % was achieved for the group of contacts that received an optimum CdCl2 treatment.

behaviour of each population, are shown in figures 6.3 and 6.4 respectively. The behaviour associated with each population is described below.

Figure 6.3: J − V curves for under-treated (η = 3.1 %), optimised (η = 9.34 %) and over-treated (η = 2.5 %) devices.

Figure 6.4: The corresponding external quantum efficiency (EQE) of devices from figure 6.3. The curves are normalised in order to allow comparison of their shape.

For the under-treated device, the behaviour is characteristic of a buried junction.

Optimised devices:

The main group of cells in the distribution shown in figure 6.2 is for high performing devices that have an average efficiency of 8.6 ± 2.2%. EQE results for a typical cell from this group (see figure 6.4) indicate that the junction is at an optimised position at the CdTe/CdS interface, with a high response over the range 550-800 nm. Poor response in the sub 500 nm range is attributed to the absorption in the CdS, which was 200 nm thick.

The best cell in the group had the following parameter values: η = 12.4%, F F = 61%, JSC = 24.9 mA.cm⁻² and VOC = 0.82 V. Its high performance was associated with a relatively low series resistance, R_s = 7.6 Ω.cm², and a high shunt resistance, R_sh = 2011 Ω.cm², leading to good rectification behaviour and a high fill factor.

Over-treated devices:

As mentioned above, over-treatment was clearly identified since there was visible physical degradation of the layer. In particular, the films showed signs of bubbling and exfoliation, and had poor adhesion. Viewing from the glass side revealed discolouration from the more usual black to an orange-grey. The efficiencies (3.3 ± 1.3%) were correspondingly poor.

Although the shapes of their EQE were comparable to those for optimised cells, their

magnitudes were reduced, leading to a reduced J_SC. Another factor contributing to the low efficiency of the cells is the series resistance which was typically twice that of the optimised cell population.

Under-treated devices:

Short CdCl₂ processing times (∼ 5 mins) gave rise to the J-V and EQE curves shown in figures 6.3 and 6.4. The EQE curve shape is typical for buried junction behaviour, and this is consistent with the depth of chloride diffusion being insufficient to form a junction at the correct position in the CdTe.

The population of under-treated cells shown in figure 6.2 was for cells on the 5 × 5 cm² device plates which were treated for 9 - 11 mins, i.e. the same conditions as for the main series of samples. They were distinguished from optimised contacts on the basis of their low efficiency and/or low F F (≤ 35%), since they were visibly indistinguishable from other parts of the plate. The average efficiency of this population was 4.4 ± 1.6%, the low performance being due to a relatively low F F together with low J_SC and V_OC values.

Cross sectional SEM

SEM was performed on an FIB-milled cross section of a single cell (contact) that had a performance typical of those within the optimised population (η = 10%). Following the preparation of the cross section, according to the methodology described in section 4.4.4, imaging was performed using the secondary electron detector. Figure 6.5a shows the geometry of the cross-section of the thin films of the solar cell, which are clearly visible at “X”, even at low magnification. Figure 6.5b shows a ×50, 000 magnification image of the same polished cross-section. The complete layer structure shows high contrast and it was possible calculate estimates for the layer thicknesses: CdTe(1.7 ± 0.1 µm)/CdS(100 ± 4 nm)/ZnO(100 ± 10 nm)/ITO(180 ± 10 nm). Furthermore the grain structure of the CdTe is clearly visible and composed mainly of large grains, > 1 µm, but with smaller grains, < 500 nm, located at grain boundary intersections. This indicates that the film has undergone a high degree of re-crystallisation during the post-deposition treatment process, it being known from preliminary investigations that the maximum grain size within as-grown CdTe films does not exceed ∼ 200 nm.

Figure 6.5: SEM imaging of a cross section of a completed device (η = 10 %). (a) The sample was prepared via cutting and polishing with a focussed ion beam of Ga⁺ ions and imaged in secondary electron mode. (b) The full cross-section shows that all layers (i.e. Au/CdTe/CdS/ZnO/ITO) within the layer are distinct and can easily be distinguished from each other. (c) The interface between the Au and CdTe is slightly porous and a change in contrast of the CdTe at the interface indicates some compositional change (e.g. an excess of Te). (d) The CdS layer has a distinct grain structure and voids are clearly present within the layer.

In some of the larger CdTe grains, twin bands are seen (see for example figure 6.5d).

Such defects are common in bulk CdTe [8–10]. While coherent twin boundaries are not strongly electrically active, random grain boundaries orientated such that the photo-current must cross them, might be expected to have a deleterious effect on PV performance [11].

Figure 6.5c shows a high magnification (×150, 000) image of the interface between the back surface of the CdTe and the Au contact. Small voids are clearly observed along this interface. This is unusual, as previously [12] it has been assumed that such voids are a consequence of a nitro-phosphoric (NP) acid etch, commonly used prior to contact CSS deposited CdTe films but not used in the case of these sputtered devices. Furthermore, a very thin dark layer is observable at the interface indicating the presence of some other phase. In the case of NP etched surfaces prepared for contacting, enrichment of Te has been demonstrated. For this material however, the interface phase has not been identified. Fur-ther characterisation, via techniques such energy dispersive X-ray analysis (EDX), Auger electron spectroscopy (AES) or X-ray photoemission spectroscopy (XPS), is required to confirm this hypothesis. An alternative hypothesis is that a Cd-O phase may form at the back surface when annealing in air [13].

Another high magnification image, this time of the CdTe/CdS/ZnO interfaces is shown in figure 6.5d. The grain structure of the CdS consists of grains typically ∼ 200 − 400 nm in width and spanning the thickness of the film. Again, significant re-crystallisation has occurred in this layer following the post-deposition treatment, the grain sizes in as-grown sputtered films being generally too small to determine using SEM. Significant voids are present throughout the CdS layer and are thought to arise from the re-crystallisation process which is likely to involve a densification of the film. Note that while in some cases these voids span the entire width of the CdS layer, at no point along the CdTe/CdS interface does the CdTe layer break through and come into contact with the underlying ZnO layer. Therefore it can be assumed that such voids are passive and do not short the devices. This is consistent with high efficiency CSS devices in which voids within the CdS layer are even more numerous than those that are fully sputtered [12].