Silicon based solar cells have been studied since the 1940s. The first silicon solar cell was accidentally discovered by Russell Ohl at Bell Labs in 1941 [65]. It was noticed while performing resistivity measurements on a piece of high-purity silicon that the current through the material was being affected by the presence of light emanating from a nearby lamp. It was found that upon illumination, a current flowed through the silicon even when no power source was connected. Ohl had inadvertently created a p-n junction; while cooling down his pieces of molten silicon, impurities fortuitously separated due to their differing atomic weights, with the lighter p-type doping elements (boron, aluminium) remaining at the top and the heavier n-type dopants (phosphorus) settling at the bottom. While an extremely important discovery, the cells produced had an efficiency of less than 1%.
It took until 1954 before what is known as the first ’practical solar cell’ was produced by Chapin et al. [66]. This cell was also made at Bell Labs, and achieved an efficiency of 6%. It was produced by diffusing p-type dopants into an n-type silicon wafer. Due to their high cost they were not widely used, but cells of this nature found use in the niche market of satellite power generation; Vanguard 1, the first solar powered satellite was launched in 1958 and made use of a 100 cm2 silicon solar panel which generated 0.1 W.
Shockley-Queisser Limit
While a silicon p-n homojunction is often what is imagined when reading about a solar cell, silicon is also used as an active component in other solar cell architectures. One of the primary reasons that silicon is widely used is due to its nearly-ideal band gap for a cell with a single absorbing layer, as calculated by Shockley and Queisser [67]. Another important reason is due to the abundance, and low cost of high purity silicon4.
When calculating the Shockley-Queisser the only electron-hole recombination mechanism considered is radiative; this ultimately demonstrates that a single cell utilising an absorbing material with a band gap of 1.1 eV, i.e., silicon, can reach a
maximum efficiency of ∼30%. It was noted in this paper that recombination caused by the Auger effect could result in lower real-device efficiencies. 23 years later, Green et al. showed that under an illumination intensity equal to that of one-sun, Auger recombination was indeed the dominant loss mechanism in silicon solar cells, and had a detrimental effect on the open circuit voltage [68]. The theoretical efficiency limit has been slightly revised several times [69, 70] since the original estimation by Shockley and Queisser, and currently stands at 29.43% for a 110 µm thick cell made of undoped silicon [71].
Since the first practical cell of Chapin, silicon p-n homojunctions have improved dramatically, and have now demonstrated efficiencies in excess of 24% [72]. While their efficiency is heading towards that of the Shockley-Queisser limit, efficiency improvements have stagnated in the past few decades.
Modern Heterojunction Architectures
In light of this stagnation, alternate architectures which still include silicon have begun to be explored recently. One of these is the silicon heterojunction solar cell. These cells are so called due to the heterojunction that is formed between the crystalline silicon wafer substrate and another material; typically amorphous silicon (a-Si) grown on the wafer surface for high performance cells. While they have a complex structure, the cells produced using this architecture are extremely efficient, and have recently demonstrated an efficiency of 26.3%,5 the highest of any silicon-based solar cell (excluding those which use concentrated solar radiation) [73]. A schematic of such a cell is shown in Figure 2.9. Many of these cells have a nano- pyramidal texture on their surface to trap light within the cell, reducing reflection losses. One of the main drawbacks of this cell is the chemicals required for growing both the undoped and doped a-Si. Plasma enhanced chemical vapour deposition is
4As an example, 99.999% purity silicon can be purchased for less than €200/kg.
5This was improved to 26.6% and is included in the famous NREL efficiency chart
https: //www.nrel.gov/pv/assets/images/efficiency-chart.png
Rear Electrode Rear TCO a-Si (n) a-Si (i) c-Si (n) a-Si (i) a-Si (p) Front TCO Front Electrodes
Figure 2.9: A typical silicon heterostructure solar cell formed from an n-type
wafer. The doping of the amorphous and crystalline silicon layers is indicated in brackets. The thickness of the crystalline silicon layer is vastly scaled down for clarity. Efficiencies of up to 25% have been reached using this design.
typically used to grow this material in industrial environments, and this requires the use of gaseous silane (SH4). This chemical is toxic and pyrophoric, and can ignite upon exposure to air, even when diluted to 1% concentration in nitrogen. The chemicals required to dope the a-Si p-type (phosphorus doping with phosphine (PH4)) and n-type (boron doping with diborane (B2H6)) are also toxic and show
similar reactivity.
Si-TCO Solar Cell
Another cell which falls in the same family is the heterojunction between a transparent conducting oxide and silicon. This has been under investigation for at least the past 50 years. One of the first studies was SnO2 on n-Si [74]. The SnO2, which was grown in a rudimentary fashion by oxidising SnCl2 on a heated piece of silicon, was presumed to be n-type, and to be degenerately doped. This was found to produce solar cells with a poor efficiency of one hundredth of that of typical silicon solar cells at the time. Solar cells of similar efficiencies were produced with CdO [75], ITO [76], SrTiO3 [77], and ZnO [78]. Despite the poor performance of these cells, the results remained promising due to the facile nature of their fabrication. While these papers demonstrated new devices, they did not undertake detailed studies to understand the electrical properties of the cells. As they are n-type TCOs deposited on n-silicon, they were most likely Schottky junction solar cells; the barrier being between semiconducting silicon and “metallic” TCOs. Some evidence for this is provided by the low resistivity and high carrier concentration values provided for the prepared ITO of Matsunami et al., which suggests that their ITO was degenerately doped [79].
A cell with notably improved characteristics was observed when polycrystalline In2O3 [79] was used as the TCO layer. The cell had an efficiency comparable to that of a commercial silicon solar cell when the n-type In2O3 was deposited on a p-type
silicon substrate. The diode and solar cell characteristics were found to be much worse when n-type silicon was used as a substrate. ITO, when grown on n-Si using a variation of spray pyrolysis, was also found to deliver similar performance [80].
Investigations began into the achieving the same efficiencies with indium-free TCOs for the same reason that alternate high-performance n-type TCOs are currently being developed: the high cost and relative rarity of raw indium. One example which combines facile synthesis (spray pyrolysis) with abundant materials is the ZnO/Si solar cell, which achieved an efficiency of 8.5% [81]. Both n and p-type silicon were used as substrates; the n-type was found to produce much better solar cells. These cells however experienced a rapid degradation in their efficiency over a period of five days which was attributed to the formation of an SiO2 layer at the interface between Si and ZnO, which was formed due to atmospheric oxygen diffusing through the ZnO layer along the grain boundaries. Illumination was also found to degrade the cells, a highly undesirable property in solar cell materials.
Various other Si/material heterojunction solar cell systems have been explored recently: a p-type Oxychalcogenide (CuS)x:(ZnS)1-x/n-Si cell [82], p-Si and n-Si nanowires with an n-type TiO2 shell deposited on them using atomic layer deposi- tion [83], p-type carbon nanotubes on n-Si [84], vertically aligned nanorods made of (Mg doped, with gold nanoparticles as a catalyst) p-type GaN grown on n-Si [85],
and monolayer MoS2 deposited on top of p-Si [86].
One material system which has yet to be explored is the heterojunction between modern high performance p-type TCOs and n-type silicon. Their transparency allows a large proportion of the incident light to reach the silicon layer, while their p-type nature could lead to a p-n junction with silicon which would result in efficient carrier extraction. This device will be explored in this thesis.