Conclusions and Outlook

Abstract

The global conclusions of my work, some critical remarks and possible future developments are presented in this chapter.

5.1 General Conclusions and Remarks

A large number of different types of light-trapping structures for thin-film solar cells have been proposed over the last decades. In order to provide a unified, objective and quantitative assessment of the different structures, I have developed a figure-of-merit in Chap. 2, called the light-trapping efficiency (LTE),  which aims to assess the performance of the nanostructure itself. The LTE separates the light trapping effect from the material, the absorber thickness, the fabrication method and the light-trapping method. The LTE enabled me to compile a ranking list of state-of-the-art proposals in Tables 2.1, 2.2 and 2.3.

Compiling this list allowed me to draw four general design rules, which I will comment on in the following paragraphs.

• 1. Sub-wavelength structures offer the possibility of suppressing the surface reflection over the useable solar spectrum (see Sect. 2.2.2), and can replace, in fact outperform, antireflection coatings.

• 2. Since the absorber slab acts as a long-pass filter for the incident sunlight, structures on the back side of the absorber only need to address a reduced wavelength range.

• 3. Whether a structure presents a periodic or random arrangement of scatterers can be considered as less important than its local geometrical features (see Sect. 2.6).

• 4. Parasitical loss mechanisms are of critical importance and must be avoided when the active absorber material becomes weakly absorptive.

The first two design rules are addressed by research studies focusing on dual-grating structures. In fact, structuring a thin-film absorber layer from both sides allows us to take advantage of the full benefits that textures can offer. However, theoretical proposals are often limited to supercell-type calculations (where the period on one side is the multiple of that on the other), whereas experimental structures were restricted to those with the same period on both sides (where the structural features of the substrate go all the way through the absorbing material).

By using an optical adhesive to aid the transfer of a silicon thin-film to a new carrier, I developed a simple method for patterning before and after the transfer process without restrictions on the design parameters. In a proof-of-principle-demonstration, I etched two different 2D gratings into the top and bottom of a 400 nm thin-film of hydrogenated amorphous silicon. The absorption measurements could indeed highlight the promise of this approach, because the dual structure showed an appreciable improvement over single gratings patterned either on the top or bottom of the film (see Sect. 4.2).

The third design rule is derived from the fact that neither completely random nor completely periodic structures are able to maximize the absorption enhancement over a broadband spectrum (see Sect. 2.2). The most promising light trapping structures tend to lie between these two domains. This observation can be understood in the framework of diffractive optics:

While the lower diffraction orders generally being more intense in random and periodic designs, the higher diffraction orders are more likely to be totally (internally) reflected at the cladding of a silicon slab, thereby experiencing a longer path length in the absorber layer. As typical structural features form the building blocks of diffraction patterns, their local arrangement thus needs to guarantee a suppression of the lower orders so as to enhance diffraction into the higher orders – which in turn will increase the optical absorption due to their longer paths.

The appropriate level of (dis)order in the diffractive structures used in my thesis has been tuned by using its Fourier spectrum as a tool. For the fabrication, I used electron beam lithography and reactive ion etching (see Chap. 3). Since electron beam lithography is a pixel by pixel exposure, and thus is relatively slow, I also developed a tool for exploring the technical possibilities offered by nanoimprint lithography in Appendix A. The setup I used for characterizing the absorption of my samples is described in Sect. 3.4.

Having gained theoretical and experimental expertise in the design of efficient light trapping structures, I noticed the need for theoretical descriptions to be more comprehensive, as not all relevant optical losses are generally taken into account (fourth design rule). This observation is based on my literature review, because the light trapping efficiency of theoretical structures significantly exceeds that of experimental ones (see Fig. 2.17).

To study and to quantify the impact of parasitic effects on the silicon absorption enhancement, I decided to compare two different light-trapping mechanisms. In a joint project with CNR-IMM Catania, each research group fabricated their own design on an identical silicon slab. I used a quasi-random diffractive structure, which I had etched into the silicon surface, whereas the research group of Catania focused on self-assembled silver nanoparticles as strong scatterers for the incident light. While Mie theory predicts a low parasitic absorption for the metallic particles, this theory neglects that scattered light can come back to the particles and be scattered or absorbed again. I developed a model for the extraction of the parasitic influence by considering the cumulative parasitics that may arise from many round-trips of trapped light (see Sect. 4.3). My analysis turned out to be robust even for different functional dependencies of the parasitic absorption, such that I could generally state that diffractive light-trapping structures tend to be a better choice for photovoltaic applications, because of the cumulative parasitics arising from multipath interactions at the metallic nanoparticles.

In summary, this thesis outlines the assessment of light-trapping structures, the critical importance of parasitic loss mechanisms and demonstrates a simple layer-transfer technique to enable the study of more efficient light-trapping designs.

5.1.1 Limitations and Future Work Left

As the electron beam system I used at York was the first ever VOYAGER system produced by Raith, I often encountered exposure issues during my work. Although receiving outstanding support from the company, the regularly recurring software bugs and breakdowns of the pattern exposures slowed the overall lithography process down. Therefore, if I were given a second chance, I probably would have moved on to nanoimprint lithography earlier. For example, time did not allow me to fine-tune the imprint parameters for the replication of structures with simultaneous small and large structural features, such as the quasi-random supercell pattern used in my experiments. Future work thus will need to focus on

• 1. the influence of the applied pressure during the UV curing step,

• 2. the matching of the relief height of the stamp to the resist thickness,

• 3. the determination of the etching rate of AMONIL,

• 4. the fabrication of bi-layer hard-PDMS/PDMS stamps,

• 5. the replacement of the current LEDs with a 360 nm wavelength LED array.

In addition, if time had permitted, I would also have tried to drop cast the nanoresist AMONIL directly onto the stamp (instead of the silicon substrate), because it may make the spinning step redundant, could reduce the necessary amount of resist material and even further facilitate the process. This method is known as reverse nanoimprint lithography. Secondly, I would have tried to set up a step-and-repeat imprint process in order to prove the scalability of light-trapping patterns on our simple home-made setup described in Appendix A.

After the nanolithography process, I used RIE etching to transfer the diffractive nanostructures into the silicon material. However, the kinetic ion energies tend to severely damage the material. While surface defects may be irrelevant in terms of the texture’s optical performance, material defects degrade the electrical performance of a thin-film solar cell device. For example, when light trapping structures are made via RIE processes and used in solar cells, the carrier transport can be subject to surface recombination velocities as high as $10^5\,\text {cm/s}$ [1]. The need for curing these defects requires efficient surface passivation techniques – solar cells would otherwise not benefit from the absorption enhancement offered by diffractive light-trapping structures.

However, although efficient surface passivation techniques were recently introduced, which allow the curing of defective states, most methods require additional processing steps, reduce the throughput capability and tend to be difficult to integrate into production lines. Why should the combination of nanoimprint-lithography and anisotropic wet etching not be an alternative to reactive ion etching then? The high damage of crystal planes is avoided (less electrical defects), no vacuum equipment is needed (higher throughput and cost-effective), and the optical performance of graded textures has been shown to be advantageous for anti-reflection and light trapping (see Sect. 2.2.2).

For these three reasons, I initiated wet-etching tests to explore the feasibility of nanostructuring the silicon surface with tetramethylammonium hydroxide (TMAH). Unfortunately, I had to interrupt these tests, because the mask materials used (flowable oxide, FOx, i.e. spin-on glass; Al and Ni metal layers) turned out to be unsuitable for this purpose. Future work could therefore concentrate to develop this idea further, once one is able to deposit or to grow typical masking materials for silicon, such as silicon nitride or thermal silicon oxide.

However, the need for thin crystalline silicon layers for my experiments also motivated me to use TMAH for the etching of a standard silicon wafer down from $500\,\upmu$m thickness to a slab of tens of microns. Figure 5.1 shows my simple setup for this test along with the resulting outcome. The total etching time may take up to 6 h at temperatures around 200 $^\circ$C.

Fig. 5.1

TMAH wet etching allows to thin silicon standard wafers down to a few tens of micrometers. When the sample is left in TMAH at 200 $^{\circ }$C for 6 h, the thickness is reduced from 500 to 50 $\upmu$m – where the material starts to become flexible

It was not until recently when I identified the master thesis from Erik Janssen [2] as inspiration for optimizing this etching process, because he reviews five different procedures for thinning substrate wafers down to a few microns. In addition, Janssen also constructed an etching apparatus of Teflon that can be totally submersed in an etching solution of potassium hydroxide (KOH). If time would have permitted, I would have studied Janssen’s thesis and optimized the thinning process in light of his suggestions for obtaining a smoother surface roughness, a faster etching rate or to limit the etching attack only to one side of the wafer. The latter may even allow to transfer a high-quality layer of 200 nm crystaline silicon from the silicon substrate of a standard SOI wafer to a glass carrier for handling.

Future work also needs to focus on the absorption measurement setup. A better collimation of the laser beam is required on the sample, because the measured Fabry–Perot fringes of unstructured thin-films show a reduced dynamic range. In addition, since diffracted light from the sample can be directly reflected towards the entrance port, I would recommend to use a smaller entrance port of a few $\text {mm}^2$ area.

Last but not least, this PhD project has been a great experience for me. The developing of theoretical and experimental methods was a comprehensive learning curve in respect of the understanding of light management and of experimental skills in fabrication and characterization of nanostructures.

I would definitely study light trapping again: although the interplay of light and matter is only the first step in the photovoltaic conversion mechanisms, it must be well understood in order to pursue power efficiencies well above the Shockley–Queisser limit of 30 %. The gained knowledge in this research field therefore becomes important for tandem solar cells that require an intermediate scatterer between the top and bottom cell, i.e. a selective Lambertian reflection of short-wavelength light (top cell) alongside Lambertian transmission of long-wavelength light (bottom cell). Such an arrangement could allow to reduce the absorber thickness of both cells and hence facilitate the collection of photogenerated charge carriers.

5.1.2 Outlook

The next step would be the realization of a working thin-film silicon solar cell $({<}20\,\upmu \text {m})$ based on the layer-transfer technique and the diffractive supercell structure (see Chap. 4). Its realization could allow the study of how light-trapping influences the power conversion efficiency, i.e. the electrical performance. Here, I will briefly outline one possible fabrication process for such a device, using the steps shown in Fig. 5.2:

• 1. structuring and coating/passivating the front surface (steps 1–4),

• 2. passivating the back surface (step 5),

• 3. fabricating the p-type electrode (steps 6–9),

• 4. fabricating the n-type electrode (steps 10–13),

• 5. defining the metal contacts (step 14),

• 6. applying a spacer layer and a white back reflector (step 15).

Since electrons as the minority-charge carriers are generally able to diffuse longer distances in p-doped crystalline silicon than holes in n-doped Si, the absorber layer can be chosen to be doped lightly p-type already. A lightly doped material also reduces the probability of Auger-recombinations.

Fig. 5.2

Possible fabrication process for thin-film silicon solar cells (see text)

In order to avoid unnecessary optical losses, thin-film solar cell should use the entire front surface as the active area. Light-trapping structures then can exploit the entire device area at the front, yet both electrical contacts have to be arranged at the back side of the device (in an interdigitated fashion).

Before applying the layer transfer technique, the absorber material must first be structured (step 1) and be passivated, e.g. via the PECVD deposition of SiN$_x$ (step 2), since the foreign glass carrier will not allow the access to the front side any more (step 3). Sunlight, however, is able to pass through the new glass substrate and will be diffracted by the coated supercell structure into the silicon absorber.

After the transfer (step 4), the thin and heavily doped electrodes are applied to the back side. A thin layer of a-Si:H (ca. 10 nm) acts as the passivation layer for the back side and as a formation layer for the electrodes. Metal masks allow to selectively dry-etch the contact regions (step 6 and 10) prior the evaporation of the electrode materials (step 7 and 11) and the thin conductive oxide layer (step 8 and 12). Subsequent cleaning steps will address shunting issues (step 9 and 13) and complete the fabrication of the electrodes. The conductive oxide, here In$_2$O$_3$, and the sputtered SiO$_x$ (step 15) optically act as a cladding layer for the absorber layer and aim to enhance the total internal reflection of light.

The sputtering of the metal contacts (step 14) and the coating/spray deposition of a white back reflector (step 15) will finalize the fabrication process.

5.2 Can Solar Power Solve the World’s Energy Crisis?

Two important issues I did not address in this thesis so far. First, solar power will only become truly competitive, if the storage problem can be overcome. Once this additional hurdle is solved, I think solar energy can make a significant contribution towards the world’s energy production. Second, transport problems can be an issue, because low-latitude countries receive the most hours of bright sunshine during the year, whereas mid- to high-latitude countries have the highest per capita energy consumption.

However, I would like to finish my thesis with a few more general thoughts about our research efforts for a more efficient energy technology. In short, I am asking the question whether the world’s energy crisis can be solved by technologies based on solar power.

Energy drives economies, guarantees growth and sustains the existence of billions of people. While human population quadrupled in the last century, our energy consumption increased tenfold in the same period. Some studies even see the increase in population as a consequence of the higher energy consumption, because new technologies have allowed farmers to cultivate increasingly larger fields and to generate greater crop yields. In light of this statement, a 10 % growth of human population would be based on a 50 % greater thermal coal consumption. In fact, China, the country with the fastest growing population, is the largest energy consumer and the world’s largest coal producer as well as importer, according to the U.S. Energy Information Administration (EIA).

In the end, the fast transitions from the agricultural to the digital age have made our living costs more and more energy-intensive.

The continuing population growth might cease when either the energy supply or the space on our planet fails to keep up with our demand for natural capital. If the required space takes into account the amount of biologically productive land and sea area – not only necessary to supply the resources a human population consumes, but also to assimilate associated waste and pollution –, one can define the footprint of human population by its impact on Earth’s biocapacity. The global ecological footprint is a standardized figure-of-merit of our demand on the Earth’s ecosystems. For example, a footprint of 1 would mean that humanity uses ecological services as quickly as Earth can renew them. The majority of environmental studies, however, agree that we already need a biocapacity greater than our planet to meet our standards of living; we are living too energy intensive: fossil fuels, like coal and petroleum, are identified as the leading causes for our large footprint, but still remain rooted as the response to our rapidly increasing energy demands – even at much higher spot prices and progressively poorer quality (see Fig. 5.3). Ultimately, our strong dependence on fuels has resulted in many political, social and ethical conflicts. I believe these global conflicts will further increase if fossil fuels remain rooted in our society. Consequently, we may are not limited by energy, but by space in support of human sustainability.

The focus recently moved to “green” technologies, that are able to replace the usage of coal and petroleum. In particular, solar power experiences a remarkable revolution, demonstrating new capacities at record installation rates. The enormous potential of solar energy is one of the motivating forces, that drives the new ambitions. England, for example, receives $2.4\,\text {kWh/m}^2$ on an average day from sun. Therefore, the sunshine on a London cab can already meet the electricity needs per capita ($8\,\text {kWh/day}$), whereas the sunlight over all England even exceeds world’s electricity demands by six times.

Ironically and despite all the efforts over the last decades, solar power contributes to our net electricity by less than 0.2 % – coal still by over 40 %. The promotion of a green technology may be a challenging task, but its success will probably not solve our hedonism on energy: yes, solar power allows to reduce the ecological footprint and so can facilitate humanity’s sustainability, but no, I think, it probably will not satisfy the world’s screams for energy.

Fig. 5.3

The global ecological footprint is a standardized measure for the human impact on Earth’s biocapacity. While in the last decade population and the global footprint grew at almost the same rate, the energy requirements disproportionally increased at much higher costs. Here, the spot prices of the West Texas Intermediate (WTI), of the central Appalachian coal (CAPP) and of the Australian thermal coal (AP) were used in the analysis (see text)