WO1997019473A1

WO1997019473A1 - Conformal films for light-trapping in thin silicon solar cells

Info

Publication number: WO1997019473A1
Application number: PCT/AU1996/000753
Authority: WO
Inventors: David Christopher Thorp; Patrick Campbell; Stuart Wenham
Original assignee: Unisearch Limited
Priority date: 1995-11-23
Filing date: 1996-11-22
Publication date: 1997-05-29
Also published as: AUPN679295A0

Abstract

A light-trapping structure (10) comprises a base material (11) having at least a first face (12) textured with a texture (13) of period T and comprised of facets (14, 15, 16, 17...). Laid on or otherwise attached to the facets of the texture (13) is a film (18) which is continuous and sufficiently thin that it conforms in its profile to the texture (13). The film (18) will thus be shaped in conformance with the texture (13) to have a period T. Ideally the film (18) is of substantially constant thickness (t). In use according to a preferred embodiment of the invention the base material (11) is light reflective, at least at the facets (14, 15, 16, 17) comprising the texture (13) whereby light rays such as light ray (19) treat the facets as a mirror and are thereby reflected according to the laws of light reflection from a surface so as to be reflected off other facets of the texture (13). Where the refractive index of the film (18) is different from the refractive index of the layer on top of it, internal reflection can occur at the boundary between these layers. In the case of a thin film silicon solar cell the base material (11) can comprise a substrate of glass to which a thin film (18) of doped silicon is applied. Additional films (not shown) can be applied on top to create a multi-layer thin film silicon solar cell with the various layers doped appropriately and the whole covered in a glass encapsulant (20). In one form the texture (13) is three-dimensional as defined in the specification.

Description

Conformal Films for Light-Trapping in Thin Silicon Soiar Celb

Thin film polycrystalline silicon is one of several promising options for low-cost photovoltaic solar cells. Important to its success is the effective implementation of "light-trapping". In theory a thickness of only 5μm could produce a short-circuit current as high as 38mA/cm² (Fig. 8 ), with resulting efficiencies over 18% Despite this, there have been few practical, low cost proposals for ways of implementing effective light-trapping in such thin films, and even less experimentally demonstrated. In deteπriining the texturing scheme to be used for light-trapping, it is important to consider the whole solar cell fabrication process. In order for light-trapping to occur, it is essential to have a reflecting substrate, and this rules out crystalline substrates on which epitaxy may be performed. Chemical Vapour Deposition (CVD) is one technique suited to the large area deposition of silicon onto inexpensive module substrates. However, CVD of silicon onto non-crystalline substrates has so far not been demonstrated to produce anything other than very poor quality material, resulting in poor carrier collection efficiency and low voltages. These shortcomings can be minimised by reducing the thickness, as long as good absoφtion is maintained with effective light-trapping. A parallel multi-junction solar cell has been proposed which could better tolerate poor quality material, but even here there is a large benefit to be made from reducing the thickness. In addition to the efficiency consideration, thinner films will have shorter deposition times and require less silicon material, important for the low-cost, high-throughput manufacturing of photovoltaics. These considerations suggest that for CVD at least, there is a strong incentive to make solar cells as thin as possible (perhaps little more than lOμmV ', and a need for suitable and effective light-trapping schemes.

Conventional texturing methods used for wafer cells are not applicable to thin CVD grown polycrystalline silicon films. The grains in the film will most likely have random orientations and so silicon's crystallographic planes cannot be used to define the texture faces. Also, the cost of using photolithography to define a texture pattern over a large area module may well be prohibitive. Mechanical texturing of silicon, as proposed for thin wafer cells, would almost certainly be too expensive, if at all possible with the small texture sizes necessary for films thinner than lOμm. Further problems may be encountered due to degraded electronic properties as a result of damage from any silicon texturing process. If the silicon is textured, then it must be done before a thin top surface emitter is formed, otherwise one would be left with a discontinuous emitter. With CVD it is possible to deposit the emitter as a final in-situ doped film, and if high temperature diffusion steps are to be avoided (eg. if using a glass substrate), then this is essential. Disrupting the deposition process in order to texture the silicon could create many defects at the p-n junction. An alternative to these difficulties is to transform the problem into one of texturing the substrate. Texturing a substrate, for example rolling or embossing a metal sheet, could be easier than texturing silicon. Figure 1 shows lOμm wide grooves embossed in aliimi ium foil using a textured silicon wafer.

One cheap option for texturing the substrate (or glass superstrate) is simply to roughen it up, for example by sandblasting. However, it is not yet clear whether this is sufficient to produce effective light-trapping. In addition, anti-reflection properties must be considered. Rough surfaces are not as good at reducing reflection as precisely defined, sharp, geometric textures, and both of these options will only be of use when applied to a glass superstrate (ie. the front surface of the solar cell). There are other possible problems to be considered: Texturing a reflecting substrate generally lowers its reflectivity, since light rays undergo more than one bounce before leaving the surface. Also, for films as thin as 5μm, the texture size would have to be rather small. Some theoretical studies have concluded that very small textures (less than lμm) will degrade light-trapping, although one piece of experimental work with 0.5μm wide, 0.2μm deep, close-packed posts in lμm thick amoφhous silicon films claimed to achieve the maximum possible absoφtion enhancement. These feature sizes happen to be close to the optimum calculated for light-trapping with one dimensional gratings. It is not yet clear what effect texture sizes of l-2μm will have on light-trapping, but certainly experiments show an increase in reflectance of textured surfaces with dimensions less than 1 μm.

Besides these optical questions, growth of polycrystalline silicon on a microtextured substrate may result in the nucleation of a grain at every ridge and valley of the texture. In the case of a glass superstrate this would result in very small grains and poor quality material at the top surface of the cell, which is the most important part. Finally, if precise geometric textures are desired, they may be more easily formed by low cost techniques such as embossing if their size is increased. Outline nf Invention

Accordingly, in one broad form of the invention, there is provided in a thin film solar cell; a method of fabrication of said cell which includes the step of (1) texturing ^{^}s^i4 substrate so as to form a textured surface in at least one face thereof and (2) depositing a film of predefined film thickness on said at least one face which substantially conforms to said textured surface whereby both light trapping and anti-reflective properties are conferred on said solar cell.

Preferably said textured surface has a predefined cross-section in at least one plane which repeats according to a predefined period.

Preferably said surface has a predefined cross-section of predefined period in a first plane and at least a second predefined cross-section of a second predefined period in a second plane.

Preferably said first predefined cross-section and said second predefined cross-section are identical.

Preferably said film is locally smooth and said predefined period is greater than the effective thickness (as defined in the specification) of the film and less than 20 times the effective thickness, more preferably 10 times less than the effective thickness.

Preferably said substrate comprises any one of glass, metal sheet (Al, Cu, stainless steel), ceramic, foil, carbon, quartz.

Preferably said substrate comprises a composite substrate.

Preferably said composite substrate comprises a highly reflective top layer placed on top of a layer suitable for texturing.

Preferably said textured surface comprises a two-dimensional texture in the form of a channel or groove structure.

Preferably said textured surface comprises a three-dimensional texture.

Preferably said three-dimensional texture is comprised of regularly or randomly located faceted structures.

Preferably said textured surface comprises pyramids in the form of square base pyramids or triangular base pyramids.

Preferably said textured surface comprises pyramids which are upright or uligluly- tilted from vertical. Preferably said textured surface comprises inverted pyramids.

Preferably said textured surface has features wherein a feature in any one dimension is less than or equal to 20 times said predefined film thickness. Preferably said textured surface has features wherein a feature in any one dimension is less than or equal to 10 times said predefined film thickness.

Preferably the size of features forming said textured surface in any one dimension is no greater than 200 microns. Preferably the size of features forming said textured surface in any one dimension is no greater than 100 microns.

Preferably the size of features forming said textured surface in any one dimension is no greater than 50 microns.

Preferably the size of features forming said textured surface in any one dimension are no greater than 10 microns or 2 times said predefined film thickness.

Preferably said textured surface further includes microscopic random texturing. In a further broad form of the invention there is provided a solar cell made according to the above method.

Preferably the solar cell further includes reflective means whereby light rays can be reflected back into or through said film of predefined film thickness. Preferably said reflective means forms part of said substrate. Figures

Figure 1. Scanning Electron Micrograph of lOμm wide grooves embossed in aluminium foil Figure 2(a). Conformal silicon film on reflecting V-grooved substrate Figure 2(b). Light-trapping in conformal film with small texture period

Figure 3. Short circuit current for encapsulated, conformal V-grooves under normally incident illumination, 100mW/cm², global AM 1.5 spectrum. Effective thickness = lOμm (for all values of α).

Figure 4. Light intensity on a module tilted towards the north by the latitude in Sydney. Azimuth (labelled) and elevation are measured relative to the module surface. North is at azimuth = 0°. Elevation is 90° at the centre of the circle and 0° at the edge. Reproduced courtesy of S. Bowden ³⁰'

Figure 5. Short circuit current for encapsulated, conformal V-grooves under illumination with incident angles representative of a year in Sydney, lOOmW/cm , global AM 1.5 spectrum. Effective thickness - lOμm (for all values of α).

Figure 6. Fraction of rays (excluding reflected rays (with pathlength = 0)) with a pathlength greater than the pathlength given by the abscissa. Figure 7. Evidence of the importance of the re-entry mechanism - fraction of rays that enter the silicon film more than once, as a function of α. (Sydney yearly illumination) Figure 8. Limiting current collection for a planar cell with zero rear reflectance, and a randomising cell (top lambertian surface) with perfect rear reflector. Both cells riave zero top surface reflection. Illumination is normally incident, 100mW/cm², global AM 1.5 spectrum. Figure 9. Tilted conformal grooves

Figure 10. Short circuit current for encapsulated, tilted, conformal V-grooves, with Sydney yearly illtunination, 10OmW/cm², global AM 1.5 spectrum. Effective thickness = lOμm (for all values of β). (a) α = 30° (b) α = 75° Figure 11. Tapered grooves

Figure 12. Short circuit current for encapsulated, tapered silicon films on symmetric and tilted V-grooves. Data point labels are the angle of the top surface of the silicon relative to α or β. Sydney yearly illumination, lOOmW/cm², global AM 1.5 spectrum. Effective thickness = lOμm (for all values of taper). Figure 13. Change in short-circuit current with the rear reflectance for rays that are inside the Si / SiO₂ critical angle. All points use a DLAR. Refractive index of copper from ref. 14. Figure 14. Change in short-circuit current with the reflectance of the glass/silicon interface. Refractive index of Si₃N₄ single layer AR coating (SLAR) = 2.0, thickness = 75nm. Thickness of TiO₂ SLAR = 60nm. All points use SiO₂ / Ag reflector. Figure 15. inverted plus upright tetrahedrons, or equivalently, tilted grooves.

Figure 16. Short circuit current for regular inverted tetrahedra under normally incident illυjnination, lOOmW/cm², global AM 1.5 spectrum. Effective thickness = lOμm (for all values of α).

5A Figure 17. is a graph of short circuit current against facet angle for a textured surface compπsing conformal inverted tetrahedra of varying periods and facet angle.

Figure 18. is a graph of rays against path length relating to a comparison of the performance of inverted tetrahedra as a textured surface.

Figure 19. is a graph of a short circuit current against lambertian fraction.

Detailed Description of Preferred Embodiments

With reference to Fig. 2(a) and Fig 2(b) structures are shown adapted for incorporation in preferred embodiments of thin film silicon solar cells of the invention.

With reference to Fig. 2(a) the structure 10 compπses a base material 11 having at least a first face 12 textured with a texture 13 of period T and comprised of facets 14, 15, 16, 17 ....

Laid on or otherwise attached to the facets of the texture 13 is a film 18 which is continuous and sufficiently thin that it conforms in its profile to the texture 13.

The film 18 will thus be shaped in conformance with the texture 13 to have a period T. Ideally the film 18 is of substantially constant thickness t.

In use according to a preferred embodiment of the invention the base material 11 is light reflective, at least at the facets 14, 15. 16, 17 compπsing the texture 13 whereby light rays such as light ray 19 treat the facets as a mirror and are thereby 5B reflected according to the laws of light reflection from a surface so as to be reflected off other facets of the texture 13 Where the refractive index of the film 18 is different from the refractive index of the layer on top of it, internal reflection can occur at the boundary between these layers.

In the case of a thin film silicon solar cell the base material 11 can comprise a substrate of glass to which a thin film 18 of doped silicon is applied. Additional films (not shown) can be applied on top to create a multi-layer thin film silicon solar cell with the various layers doped appropriately and the whole covered in a glass encapsulant 20.

In this specification the term "substrate" is used to denote a base material such as base material 11 which is textured on at least one face and adapted to receive a film such as film 18 applied to that face so as to form a conformal film thereon In certain embodiments it is possible for the base material 11 to comprise a superstrate. In this specification the term "substrate" is to be taken to include situations where the substrate, in the final product of which it is a part, is acting as a superstrate.

Fig. 2(b) illustrates the situation where the peπod T is shortened for the same film thickness t. The detailed description of preferred embodiments which follows indicates this arrangement to be particularly preferred in certain circumstances including those relating to three dimensional texturing. 5C

In this specification two dimensional texturing describes the situation where a given texture 13 is of constant cross-section in one dimension as in the case of Fig. 1 which indicates textures formed as grooves of constant cross section throughout their length dimension

A three dimensional texture is one where there is a variation in cross section along the length dimension such as, for example as shown in Fig. 15 where the three dimensional textures comprise tetrahedrons

Conformal Films - First Preferred Embodiment of the Invention Some of the problems mentioned in the introduction can be avoided by texturing the substrate or base material 11 with feature sizes greater than the thickness of the deposited film 18. By doing this, the top surface of the silicon film conforms to the texture of the substrate, thus retaining the texture's anti-reflection properties. The regular V-grooves in Fig 2 are probably the simplest implementation of this idea

The structure of Fig. 2(a), with very large texture period, has been investigated for amoφhous silicon cells. In this case "light-trapping" occurs not by rays being trapped within the silicon film, (which cannot occur since the top and rear surfaces are parallel), but by re-entry into the opposing face after escape from the film. In other words, a structure with good anti- reflection properties will also provide more passes through the silicon for light rays that are not reflected, thus giving good light-trapping performance. Note that although the reflecting substrate is textured, it may be smooth on a microscopic level and so have a higher reflectivity, since reflected rays do not undergo a double bounce before leaving the rear surface. Note also that it could be a glass superstrate that is textured, with the silicon deposited onto the textured glass.

The corrugated structure of Fig. 2(b), with texture period of the same order of magnitude as the film thickness, was proposed by Uematsu et al. for crystalline silicon wafers. In this case light can now be trapped within the silicon. Their analysis was limited to unencapsulated cells with V-groove angles near to the 70.5° defined by silicon's crystallographic planes. They predicted an improved performance by making the rear V- groove angle slightly smaller than the front one, thus producing a taper in the silicon thickness. Whilst the corrugated structure might present manufacturing difficulties for silicon wafers, the deposition of a silicon film onto a textured substrate could be quite straightforward.

This specification uses ray tracing to investigate the performance of such structures for a large range of groove angles and texture periods, as well as tilted (asymmetric) grooves and tapered films. Altemative three dimensional structures are proposed which are expected to improve the performance above that obtained for these two dimensional textures.

Ray tracing

The ray tracing model is expected to be valid even for films as thin as 5μm, since within the wavelength range of interest (700 - 1 lOOnm), the wavelength within the silicon will be less than 0.3 μm, significantly less than the feature sizes involved. In addition, the coherence length of white light is very small, of the order of 1 μm in air, and therefore less than 0.3 μm in silicon, and so interference effects between the top and rear surfaces of the film are expected to be small.

A program, "Tracer", was written that traces light rays through the defined texture planes, reflecting or refracting at interfaces according to reflection probabilities that are calculated as a function of wavelength and angle of incidence for any number of (possibly absorbing) thin films, (using the matrix method The reflection probabilities used are the average of s and p polarisations as this approximation has been shown to have negligible effect By recording the number of rays within certain pathlength ranges, the ray tracing program is able to quickly calculate the short-circuit current for all silicon film thicknesses from one trace, as long as the ratio of texture size to film thickness remains constant. Although it is therefore not necessary to trace at many different wavelengths, all the results presented here were traced with 10,000 rays at each of eight wavelengths, in order to accurately model reflection properties. All short-circuit currents are calculated for lOOmW/cm² global AM 1.5 spectrum and are reproducible to within +/- O.lmA/cm². Several of the data points for symmetric grooves were accurately verified with the commercially available program, "Sunrays .

Effective thickness

In comparing the results for different textures of the same thickness, it is important that the definition of thickness does not bias one texture against another. This is not a simple matter for the textures considered here. There are two possible definitions of thickness: the film thickness, as labelled in Fig. 2(a), and the effective thickness, defined as the volume of silicon divided by the area of the module front surface, which for the symmetric grooves in Fig. 2, is simply the vertical height of the silicon film. In deciding which term to use one must consider the two issues of production cost and cell efficiency. For CVD the major cost factors will be the deposition time and the volume of source gases used. Comparing shaφ and less shaφ V-grooves of the same film thicknesses, the two textures will have similar deposition times (if the deposition rate is surface limited), whilst the sharper grooves will have a higher volume of silicon (and higher effective thickness) and will therefore consume more source gases. With regard to efficiency, the likely poor quality of non-epitaxial CVD polycrystalline silicon means that the recombination current and voltage in such solar cells will most probably be dominated by bulk or junction recombination. The more shaφly grooved cell will have a larger bulk volume and larger junction area, and so lower voltage. Comparing shaφ and less shaφ V-grooves of the same effective thicknesses, the bulk volumes will now be equal, whilst the more shaφly grooved cell will still have a larger junction area. However, the film thickness of the shaφly grooved cell will be smaller and so the carrier collecting efficiency will be greater (for a single junction cell); also the deposition time could be shorter. Considering these issues, the parameter used in this specification for comparing textures of equal thicknesses, is the effective thickness.

Results

The calculated values of current density for symmetric grooves as a function of α, the angle to the horizontal (see Fig. 2) for an effective thickness of lOμm and period widths of 20μm and lOOμm are shown in Fig. 3. The illumination is normally incident. Two points for a theoretically perfectly randomising cell are included for comparison (top or rear lambertian surface). All cells are encapsulated under glass of refractive index, n = 1.5, and have a double layer anti-reflection coating (DLAR) of 80nm Al₂O₃ (n = 1.7) and 50nm TiO₂ (n as 2.5, increasing at short wavelengths '), and a silver rear reflector separated by 120nm of SiO₂. ' The refractive index of silicon was taken from reference 2. The DLAR was optimised by ray tracing a lambertian cell, and keeps reflection at the glass/silicon interface at normal incidence below 5% over the wavelength range from 420-1050nm. The major reason for the difference between the two lambertian cells is the increased reflectance from the planar top surface of the cell with the rear lambertian surface. In reality, a top lambertian surface would also have lower reflection due to double bounces of incident rays at the textured glass/silicon interface (as well as due to reflected rays that totally internally reflect at the air/glass interface and return to the silicon ). On the other hand, a perfectly randomising surface is unlikely to be achieved in practice.

Note that for normally incident light, α only has to be greater than 21 ° (half the air/glass critical angle), in order for total intemal reflection at the air/glass interface to occur, thus providing low reflection and allowing the re-entrant mechanism to operate. The point of retum of a ray to the silicon after traversing the cover glass will depend on the thickness of 9

the glass. In the case of a small texture period the natural variations in the thickness of the glass will result Ln a random point of return. For the larger texture sizes of lOOμm modelled in this work, the glass thickness was set at 3mm, with a random variation of +/- lOμm. The difference in results for this case and for a fully random point of retum is very small.

As the texture period increases, light-trapping within the silicon is reduced and for most regions of the silicon film, light rays will simply enter, reflect from the rear and escape from the top surface, leaving only the re-entry mechanism and light-trapping in the glass to increase the pathlength.

Most of the results presented in this specification are for an effective thickness of lOμm. However, if the ratio of texture period to effective thickness is maintained, the trends shown in the graphs will be the same for other thicknesses, though to a greater degree for thinner films due to the increased importance of light-trapping.

Yearly illumination

Experience has shown that when performance depends critically on texture angles, illumination with light away from the normal causes a reduction in performance. This was tested by tracing the textures with rays at angles representative of typical Sydney yearly illumination. The module is tilted towards the north by the latitude, with the grooves running East - West. The number of rays traced at each angle was proportional to the energy received at that angle. The amount of energy falling on the module at 108 discrete solid angles was calculated from the global and direct radiation data, using equations that determine the position of the sun throughout the year and assuming the diffuse component to be perfectly diffuse. Figure 4 shows the intensity as a function of angle, where azimuth and elevation are measured relative to the module surface. The bright bands corresponding to the daily movement of the sun at +/- 23.5° (the declination of the earth) are clearly seen, although the diffuse radiation falling on the module accounts for a high 40% of the total.

Figure 5 shows the performance under yearly illumination for the same textures as traced at normal incidence for Fig. 3, as well as for textures of other period widths. The dip at α = 34° for the period width of lOμm, is difficult to explain, but in any case, it is questionable whether a silicon film of thickness little less than the period width could be conformally deposited (ie. maintaining the substrate texture on the top surface of the film). Note on the other hand, that the peak performance is obtained at α = 56°, where the film thickness is only 5.6μm. This angle of α happens to be close to that obtained from silicon's natural crystallographic planes. The poor performance for all periods at α = 45° is due to rays returning along their incident path, thus reducing scatter.

A large part of the drop in performance compared to normal incidence is due to the increased reflection from the cover glass at large angles (from 4.0 % to an average of 6.4 %). This illustrates the important general principle that even in very thin films, anti-reflection properties are of just as great concern as light-trapping properties. Setting the glass reflection to a fixed 4.0 %, independent of the angle of incidence, increases the current for all textures in Fig. 5 by 0.8 mA/cm², or about 2.3% relatively.

Tracing the textures of 20μm period with the grooves aligned North - South reduced the current by 0.3 to 0.6 mA/cm compared to the values in Fig. 5 where the grooves are aligned East - West. This shows the importance of considering the angular sensitivity of light- trapping structures to incident light, and that it is better to simulate these structures under illumination typical for a year, rather than simply using isotropic illumination. Unless otherwise stated, all further results are for the more realistic yearly illumination, with the grooves aligned East - West.

Re-entrv Mechanism. Pathlength Enhancement and Absoφtion limits

The average pathlength of rays for these textures is typically about 14 times the effective thickness. Considering how far this is below the theoretical maximum of 4n_Sj² (= 51 at 1 μm wavelength), the results predicted are quite reasonable. The reason for this can be seen in a plot of pathlengths (Fig. 6). A high proportion of rays stay in the cell for the first few passes, which are the most important ones as this is when the intensity of the ray is at its greatest. This is due to the re-entry mechanism - rays escaping from the film still have a chance of re-entering it on the opposing face, either directly, or via total intemal reflection at 11

the air/glass interface. Figure 7 plots the percentage of rays that enter the silicon more than once. A strong correlation is shown with the short-circuit current in Fig. 5, with the difference being that the short-circuit current reduces at high angles of ct because the thickness of the silicon film reduces (so that the effective thickness is constant with α). By contrast, in a lambertian cell it was found that only 52% of rays enter the silicon more than once.

normal yearly illumination incidence α, period 56°, lOμm 56°, lOμm 34°, 20μm 34°, lOOμm

J_sc (mA/cm²) for lOμm thickness 36.7 35.1 34.6 333 mean pathlength 17.7 14.8 14.4 14.1 standard deviation of pathlength 13.9 15.4 16.4 28.0

Table 1 : Jsc for an effective thickness of lOμm, mean pathlength and standard deviation of pathlength (excluding reflected rays), in units of effective thickness, for the structures whose pathlengths are plotted in Fig. 6.

Table 1 shows that when comparing these textures, the standard deviation of the pathlengths is an important parameter, and perhaps a better indicator of performance than the average pathlength. A small number of rays with very long pathlengths can give a high average pathlength, but these rays provide little extra current - either they are already nearly all absorbed after a few passes or they are so weakly absorbed that they will never contribute greatly to the current. The inversely exponential dependence of absoφtion on pathlength can be illustrated by comparing the current versus thickness curves of a planar cell (with perfect

AR coating and non-reflecting rear) and lambertian cell (perfect AR and reflector). (Fig. 8)

The straight lines show that although over 20 passes (for all rays) are required to match the lOμm thick lambertian cell, 9 passes will give 95% of the current and a mere 4.7 will give

90%. In other words, if a high proportion of rays can be kept in the cell for the first few passes, then it is not so hard to obtain reasonable light-trapping performance.

The average pathlength achieved with the conformal schemes is quite respectable when one considers that as yet, no advantage has been taken of scattering of the rays in the third dimension (along the grooves). The maximum average pathlength for a two dimensional 12

system under isotropic iUumination is only πn ^» 1 1.2 at 1 μm wavelength. The factor of π comes from an average ray-length across the cell of π/2, multiplied by 2 due to the presence of a rear reflector. This can be exceeded in the structures here since they are not truly two dimensional and therefore incident rays with a non-zero component in the direction along the groove will slightly increase the average ray-length, although this component will still be small within the silicon as it is restricted by the escape cone for silicon to air. (For rays to reach the silicon/glass critical angle would require near grazing incident angles at the silicon in the direction along the groove, but this angle is restricted by the glass/air critical angle after refraction into the glass.) The maximum average pathlength can also be exceeded due to non- isotropy of the incident light (note the increase for normal incidence in table 1) and non-zero reflectivity of the front interfaces. This latter point was verified by ray tracing a lambertian cell with a perfect rear reflector, under isotropic illumination. With the front reflectivity set to zero the average pathlength was calculated as 50.8 times the thickness at lμm wavelength. Using the correct values of reflectivity for air/silicon, the average pathlength increased to 77.9. Even tracing an encapsulated cell with zero glass silicon reflection, but with the correct reflectivity for the air/glass interface, gave an average pathlength of 56.0. However, in terms of short-circuit current, one always loses out due to increased reflection.

Combined conformal and random texturing

In ref. 35 it was pointed out that it is impossible for a surface textured planar solar cell to reach the maximum absorbance limit, where all rays have a pathlength exactly equal to 4n_Si ².

In order to achieve the maximum average pathlength it is necessary to have complete randomisation of rays, but in a surface textured cell, this automatically results in some rays having a pathlength greater than 4n_Sl in just one pass of the cell. However, this argument does not apply to corrugated conformal films; the corrugated nature of the cell restricts the very long rays, whilst in compensation, the re-entry mechanism increases the pathlength of rays that would otherwise be lost after escape from the film. To verify this argument, an unecapsulated conformal cell was ray traced under isotropic illumination, with α = 30°, period equal to twice the effective thickness, and with a lambertian front surface. The front reflectivity was set to zero and the rear to 100%. The average pathlength was calculated to be 50.5 times the effective thickness, the same as for an unecapsulated planar lambertian cell, but the standard deviation of the pathlengths was lower (50.5 compared to 53.9), and the short-circuit current for an effective thickness of lμm was 32.8 mA/cm², significantly higher than the 32.4 mA/cm² predicted for the planar cell. The fraction of rays entering the silicon more than once was 14%, compared to 0% for the planar cell (since the front surface is considered flat and the cell is unencapsulated), confirming the re-entrant mechanism as the reason for the improved performance. The percentage of rays with a pathlength greater than twice the effective thickness was 96.6% for the conformal cell, compared to 92.2% (= 1 - 1/ n_Si ²) for the planar cell. The values in this section were obtained by tracing 30,000 rays at 1 μm wavelength and the differences observed are larger than the statistical errors. In view of this higher theoretical upper limit, it may be that conformal films on a three dimensional textured substrate may also in practice achieve a higher level of light-trapping than surface textured planar cells.

Any practical attempt at a lambertian cell is likely to result in a fraction of rays escaping from the top surface that is greater than 1/ n_Si , due to incomplete randomisation of light, and in this case the re-entrant mechanism is of even greater value. To demonstrate this, planar and corrugated (α = 34°, period = lOμm and a more practical 5mm) cells were ray traced with a lambertian character of 0.5, so that the probability of a ray being randomised in direction at any reflection or refraction event at the top surface of the silicon was 0.5. All cells were encapsulated and had zero front reflection, a perfect reflector and isotropic illumination. Table 2 compares the reduction in short-circuit current for each cell when the lambertian character is changed from 1 to 0.5. The percentage of rays with a pathlength greater than twice the effective thickness is also given, demonstrating the value of the re-entrant mechanism when randomisation is incomplete. These results were also confirmed with "Sunrays".

lambertian planar corrugated corrugated character 1 Oμm period 5mm period

J_sc (mA/cm²) for 5μm thickness 1.0 37.9 38.1

0.5 34.8 37.3 36.1 % of rays with path > lOμm 1.0 92.... 96.6

0.5 61.3 93.0 83.4 14

Table 2. effective thickness of 5μm

Tilted f asymmetric', grooves

Previous work has shown that reflection can be reduced, and light-trapping improved, by tilting textures. Given the close relationship between anti-reflection mechanisms and the re_¬ entrant light-trapping mechanism, tilted grooves were studied to see if performance could be increased. In tilting the grooves, the thickness of the silicon film was made the same on both sides of the grooves, as might be expected from the deposition process and as illustrated in Fig. 9. The side edges of the defined texture must be at an angle to the normal to the glass surface so that the texture can be repeated periodically. The effective thickness is still the volume of silicon divided by the area of the module front surface; this is no longer the vertical height of the silicon film, since this is different for each side of the texture.

Figure 10(a) shows the varying performance of textures where the angle to the horizontal of one half of the groove, ct, is kept constant at 30°, while the angle of the other half, β, is varied. The curves have broader peaks than for the symmetric grooves. This is preferable when one considers manufacturing errors and different illumination patterns to that modelled here (eg. for modules set at non-optimum angles). Note the dip in performance again when the two sides of the groove meet at a right angle, (ie. α + β = 90°). This pattem is repeated for curves with other values of α (not shown). For example, with α = 55°, a dip occurs where β = 35°, just where one might otherwise hope to see a peak. Best performance is obtained when the grooves are pushed right over (β > 75°). Results are presented in Fig. 10(b) for α = 75°, with varying β. For α = 75°, α + β = 90° when β = 15°, which occurs well before the expected peak in performance. Setting α to 80° tended only to shift the peak to lower values of β.

Tapered films

The clearest factor limiting light-trapping performance is the period of the texture. The larger the period, the greater the regions of silicon film where light rays will simply enter, reflect from the rear and escape from the top surface, leaving only the re-entry mechanism and light- 15

trapping in the glass to increase the pathlength. If small texture periods present manufacturing difficulties, light-trapping can be encouraged within the silicon film for larger periods by putting a taper in the thickness of the film. This idea has been presented before in a limited fashion for wafered silicon cells. Such a taper often occurs naturally during the deposition of silicon on textured substrates by Plasma Enhanced CVD (PECVD), and can be controlled to some extent. The variable, taper, is defined as in Fig. 1 1.

When "taper" = 1.0, the top and rear surfaces of the silicon film are parallel. For any other given value of taper, the angle between the surfaces will be smallest for the larger periods. This angle is given alongside certain data points in Fig. 12. The two angles given for tilted grooves are relative to α and β respectively. Clearly some values could not realistically be achieved simply by depositing silicon, especially for the smaller periods. For example, at the limit of taper = 0.2 for the texture with α = β = 34° and period of 40μm, the top surface is actually flat. The ratio of the minimum thickness of the silicon film at the top of the groove, to that at the bottom of the groove ("film thickness"), is actually a little more than "taper". For example for taper = 0.6 for the texture with α = β = 34° and period of 40μm, this ratio is 0.67. A similar difference exists for other textures. The opposite is the case for taper greater than 1. Remember that the effective thickness is the volume of silicon divided by the area of the module front surface.

Clearly the less effective textures investigated so far can be improved with a taper, but it seems the best results previously predicted for texture periods less than 20μm, can not be exceeded. This is hardly suφrising considering the limits imposed by the restriction of scatter to two dimensions. The small values of taper required to reach this limit for larger angles of oc, or larger periods, imply possible problems with series resistance where the film is very thin.

Performance is seen to improve most consistently for "taper" less than 1. This is not suφrising as in this case the top surface is rotated so that a normally incident ray that is reflected from the rear, makes a larger angle with the normal to the top surface and so may be totally internally reflected. Unfortunately, the more natural value for "taper" after PECVD of silicon, is more than 1 (ie. thicker films on the peaks of the grooves). If however, a glass 16

superstrate is used, rather than a substrate, then the peaks of the grooves during deposition become troughs under illumination, resulting in a "taper" of less than 1.

Reflectors and Anti-reflection coatings

The preceding results have attempted to define the peak performance that could be expected for these textures, and have accordingly used the best possible anti-reflection (AR) coatings and reflectors. Cost considerations may prevent the use of such materials. Even using a silver reflector, practical imperfections may result in lower reflectivities than modelled.

Figure 13 shows that the performance of the lambertian cell is less dependent on reflector quality than the conformal textures because it scatters a larger number of rays outside the silicon / silicon dioxide critical angle at the rear. (Note that the abscissa refers to the reflectance for rays that are not totally internally reflected by the silicon dioxide layer.) The tapered film also steers rays outside this critical angle. For the aluminium reflector, the SiO₂ layer is probably not thick enough to prevent the aluminium frustrating total internal reflection. Also, the conformal textures have a smaller film thickness than the lambertian cell (for the same effective thickness), especially with large angles of α, and this smaller thickness increases the reflector requirements. The drop in performance from the use of an aluminium

(19) reflector could perhaps be compensated for somewhat by using MgF₂ (π = 1.38) in place of SiO₂ as the separating dielectric layer.

The V-groove textures are expected to have good AR properties, and indeed, Fig. 14 shows they are less dependent on AR coatings than the lambertian cell, although the difference is smaller for real AR coatings, perhaps because of a greater need for low reflection over a wide range of incident angles, due to the importance of the re-entry mechanism. (The reflectance given by the abscissa is independent of wavelength and angle of incidence.) The tapered texture drops off as rapidly as the lambertian cell because the top silicon surface is at an angle of only 23° to the glass top. (34° - 1 1°) A planar cell with a lambertian reflector and smooth top surface would show an even more drastic drop in performance.

3D textures 17

The two main factors limiting performance for the textures investigated are reflection from the glass surface (which is a problem faced by all approaches) and the lack of scatter in three dimensions. The glass reflection could be reduced by making grooves in the top surface of the cover glass. If these grooves are aligned North - South, then rainwater can run along them downhill, and may still prevent the build up of excessive dust on the module surface. This has been investigated by Sche decker et al. The increased surface area may also allow the modules to run cooler, thus increasing the operating voltage. Also, since the silicon grooves are best aligned East - West, the grooves on the top surface of the glass will introduce a third dimension to the scatter, although only within the glass, not the silicon. It will slightly increase the pathlength in the silicon by allowing rays to reach the silicon/glass critical angle in the direction along the groove, where as previously they were restricted to the silicon/air critical angle (because the incident angle at the silicon was limited to the glass/air critical angle). To obtain scatter in three dimensions within the silicon will require the deposition of silicon onto a substrate textured with a three dimensional pattern, such as inverted pyramids, or tetrahedrons (triangular based pyramids). A combination of inverted, tilted tetrahedrons adjacent to upright tetrahedrons is shown in Fig. 15.

This texture is in fact exactly equivalent to grooves tilted at an angle to the glass surface in the direction along the groove. There is a further advantage to this structure: The ridges and troughs of any texture embossed into a substrate may not be perfectly shaφ; if this is so then the resulting flat regions will increase surface reflection. However, in the texture of Fig. 15, the ridges and troughs are still at an angle to the module surface, which if greater than 30° will enable total intemal reflection at the air/glass interface to maintain low reflection. The only flat regions (ie. parallel to a flat glass top surface) are the small points at the peaks of the tetrahedrons. The use of a grooved top glass surface will also have this advantage of reducing the effects of not perfectly shaφ textures. Opposing this concept however is the possibility that rounded edges could improve light-trapping performance by effectively scattering rays in all directions. This is a possible explanation for the exceptional levels of light-trapping demonstrated in PERL cells,' which exceeds that predicted by ray tracing shaφ edged textures. 18 3D Conformal Textures - Second Preferred Embodiment of the Invention

To achieve the maximum average pathlength enhancement requires scattering in

3 dimensions. Raytracing studies of conformal films deposited on a wide variety of 3D textures revealed an optimum texture of inverted tetrahedra (triangular based pyramids). The predicted J_sc of such textures is presented in Fig. 17 for an effective thickness of lOμm. The cells are encapsulated and an AI₂O₃ / TiO₂

DLAR (80nm/50nm) and a Ag reflector displaced by 120nm of Si0₂ is assumed

Ray incident angles are typical of those experienced by an optimum tilt module over a year in Sydney, Australia.

Although AR properties and re-entry generally improve with increasing α, the decreasing film thickness reduces absorption tor a given number of passes across the film and also exposes rays more often to the possibility of escape from the front surface or absorption at the rear surface. Thus, optimum facet angles are quite moderate. For a = 36°, re-entry usually occurs via total-internal-reflection at the glass-air interface.

With a small enough period, performance exceeds that possible with a planar cell with perfectly randomising (lambertian) surfaces. Figure 18 which plots the pathlength of rays for tetrahedra of 36° facet angle with a perfect AR coating and reflector and isotropic illumination, shows how the re-entry mechanism prevents rays escaping after the first few passes. Under these limiting conditions, the J_sc for the 20μm period would be 40.25mA/cm², 0 4mA/cm² higher than for an ideal planar lambertian cell. Whilst Fig 18 shows that light-trapping with the more 19 practical 0.1mm period is inferior to lambertian, a high percentage of rays remain in the cell for the first few passes, which are the most important ones for current generation as this is when the light intensity is greatest. The better AR properties of the conformal cell enable it to develop the same J_sc as a lambertian cell under real conditions, as seen in Fig. 17. All 3 textures in Fig. 18 achieve the maximum possible average pathlength of 4n_Sl ² (= 51 at lμm wavelength), demonstrating the inadequacy of this parameter as a measure of light-trapping performance.

Table 1. J_sc of optimum textures for lOμm effective thickness, (encapsulated, DLAR, Si0₂/Ag reflector, yearly illumination).

α 40μm period lOOμm period grooves 34° 34.2 33.3 inverted pyramids 35° 36.3 35.2 inverted tetrahedra 36° 37.2 36.4

The performance of inverted tetrahedra is compared to conformal grooves and inverted pyramids (square based) in Table 1. The superiority of tetrahedra can be credited to: (1) all facets having different azimuthal orientations, which maximises the rate of randomising, (2) capacity for a triple bounce antireflection trajectory with low facet angles (which also enhances re-entry), and (3) the absence of opposing facets of equal and opposite angle, which by symmetry encourages escape with grooves and pyramids. Inverted textures are used rather than upright structures, since they provide a higher level of re-entry. Whilst the loss mechanism of point (3) can be reduced by tilting the grooves or pyramids, the results presented here for symmetric tetrahedra have not been exceeded. The differences observed in table 1 will increase for thinner films. 20

Substrate Embossing

The ray-tracmg results in Fig. 17 show that texture periods less than about 0.1mm are required for the best light-trapping in a conformal film of lOμm effective thickness. Since the ray-tracing analysis is purely geometrical, thinner films will require correspondingly smaller periods for the same level of light-trapping. This can be done by an embossing process. The embossing of lOμm wide grooves in Al foil, using a textured silicon wafer to form the imprint, has been demonstrated on a limited scale and is illustrated in Fig. 1 Ceramic substrates are another possibility, which can be imprinted while in powder form, before fusing.

Microtextured Surfaces - Third Preferred Embodiment of the Invention

Large period textures may be more economic for manufacturing in the near term. In this case, the reduced level of scatter can be compensated by adding a rough microtexture to the facets. Fig. 19 shows that most of the benefit from this method is achieved with quite a mild level of scattering. In Fig. 19 the "lambertian fraction" = the probability of ray being randomised at any reflection event at front or rear surfaces. The probability of a ray being randomised on transmission is set to half this value, since refraction into the escape cone reduces the scattering effectiveness If the ray is not randomised then it is treated specularly.

Experimental evidence and ray-tracing of rough surfaces suggest that microtexturmg is unlikely to produce scattering equivalent to a lambertian fraction much greater than 0.5. Thus a significant advantage can be expected form combining a rough microtexture with a conformal texture of large period. With a microtexture of reasonable scattering effectiveness the advantage of a 3D conformal texture is reduced because the scattering in grooves is no longer restricted to 2D.

The drawbacks of an additional microtexture are, (1) AR properties are degraded, (2) an increase in area related recombination (surface and junction), and (3) possibly degrading material quality through enhanced crystal nucleation 21

The small degree of scattering required to enhance conformal textures of large period, suggests mild alternatives to deliberate microtexturing may be of benefit. For example, roughening caused by crystallisation, or scattering at multiple, rough Si/Ge interfaces (due to the refractive index difference). The incorporation of Ge layers into thin silicon cells has been predicted to significantly increase absorption and efficiency when effective light-trapping is incorporated.

Industrial Applicability

The process of embodiments of the invention comprising texturing of substrates and the application of conformal films thereto can be applied with advantage to the manufacture of thin film silicon solar cells.

Claims

-22- CLAIMS

1. In a thin film solar cell of the type having a substrate and at least a layer of a first conductivity type and a layer of a second conductivity type; a method of fabrication of said cell which includes the step of (1) texturing said substrate so as to form a textured surface in at least one face thereof and (2) depositing a film of predefined film thickness on said at least one face which substantially conforms to said textured surface whereby both light trapping and anti-reflective properties are conferred on said solar cell.

2. The method of claim 1 wherein said textured surface has a predefined cross-section in at least one plane which repeats accordmg to a predefined period.

3. The method of claim 2 wherein said surface has a predefined cross-section of predefined period in a first plane and at least a second predefined cross-section of a second predefined period in a second plane

4. The method of claim 3 wherein said first predefined cross-section and said second predefined cross-section are identical.

5. The method of any one of claims 2-4 inclusive wherein said film is locally smooth and said predefined period is greater than the effective thickness (as defined in the specification) of the film and less than 20 times the effective thickness, more preferably 10 times less than the effective thickness.

6. The method of any previous claim wherein said substrate comprises any one of glass, metal sheet (Al, Cu, stainless steel), ceramic, foil, carbon, quartz. -23-

7. The method of claim 6 wherein said substrate comprises a substrate.

8. The method of claim 7 wherein said substrate comprises a highly reflective top layer placed on top of a layer suitable for texturing.

9. The method of any previous claim wherein said textured surface comprises a two-dimensional texture in the form of a channel or groove structure.

10. The method of any one of claims 1-8 wherein said textured surface compπses a three-dimensional texture

11. The method of claim 10 wherein said three-dimensional texture is comprised of regularly or randomly located faceted structures.

12. The method of claim 10 or 1 1 wherein said textured surface comprises pyramids in the form of square base pyramids or triangular base pyramids.

13. The method of any one of claims 10-12 wherein said textured surface comprises pyramids which are upright or tilted from vertical

14. The method of any one of claims 10-13 wherein said textured surface compπses inverted pyramids

15. The method of any previous claim wherein said textured surface has features wherein a feature in any one dimension is less than or equal to 20 times said predefined film thickness.

16. The method of any one of claims 1-14 wherein said textured surface has features wherein a feature in any one dimension is less than or equal to 10 times said predefined film thickness.

17. The method of any previous claim wherein the size of features forming said textured surface in any one dimension is no greater than 200 microns. -24-

18. The method of any previous claim wherein the size of features forming said textured surface in any one dimension is no greater than 100 microns.

19. The method of any previous claim wherein the size of features forming said textured surface in any one dimension is no greater than 50 microns.

20. The method of any previous claim wherein the size of features forming said textured surface in any one dimension are no greater than 10 microns or 2 times said predefined film thickness.

21. The method of any previous claim wherein said textured surface further includes microscopic random texturing.

22. The method of any previous claim further including reflective means whereby light rays can be reflected back into or through said film of predefined film thickness.

23. The method of claim 22 wherein said reflective means forms part of said substrate.

24. The method of any previous claim wherein said textured surface includes facets which are angled in a range between 25° and 65° to a horizontal reference plane.

25. The method of claim 24 wherein said facets are angled in a range between 30° and 60° to said horizontal reference plane.

26. The method of claim 24 wherein said facets are angled in a range between

30° and 50° to said horizontal reference plane.

27. The method of ciaim 24 wherein said facets are angled in a range between 30° and 45° to said horizontal reference plane. -25-

28. The method of claim 24 wherein said facets are angled at approximately

40° to said horizontal reference plane.

29. The method of any previous claim wherein additional scattering is effected by Ge or SiGe alloy layers in said cell.

30. The method of any previous claim wherein said solar cell includes a reflective layer as part of or on top of said textured surface.

31. A thin film solar cell made by the method of any one of claims 1-30.

32. A solar cell comprised of at least a substrate having deposited thereon at least a first layer of a first conductivity type and a second layer of a second conductivity type arranged so as to form at least one PN junction adapted to convert incident light into an EMF across said at least one junction; said substrate having a textured surface to which said first layer conforms.