1.1 Tandem and Multijunction Solar Cells

Researchers and industries working in the field of thin-film silicon solar cells have made extensive use of the tandem and multijunction concept. Various designs have been studied and commercialized; the main designs used are the following:

On the research front, many other designs for triple-junction cells are being studied. Results have been obtained for the following designs:

At this stage, it is not clear which of these designs will be the most successful; it may well be a completely different design using a novel microcrystalline alloy for one of the subcells. Such alloys are presently being developed (e.g., see [24]).

2.1 Structure of Amorphous Silicon

Crystalline solids, such as monocrystalline silicon wafers, have a fully regular and periodic structure; they possess what is called both short- and long-range order. In such a crystalline silicon network (or crystalline silicon matrix as it is also called), each silicon atom is bonded to four neighbouring silicon atoms. The bond angle—i.e., the angle between two adjacent bonds—is fixed at a value of 109° 28′ and remains the same throughout the whole crystalline network. Figure 1 schematically shows this situation. The bond length, or distance between two neighbouring silicon atoms within such a network, is also fixed and remains constant throughout the whole network at a value of approximately 0.235 nm.

In amorphous silicon thin films, both the bond angles and the bond lengths vary in a random fashion: there is a whole distribution of values. For instance, the bond angles have a random distribution centred around 109° 28′ and a standard deviation of 6° to 9°. If the amorphous silicon layer has just a low “amount of disorder,” then the distributions for bond angles and bond lengths will be very narrow. In this case, we will obtain “device quality” amorphous layers with satisfactory electronic properties. If the amorphous layer has a “high amount of disorder,” we will obtain broad distributions and unsatisfactory electronic properties. The disorder will directly affect the band tail states: the band tails will be more pronounced for strongly disordered layers. Amorphous silicon layers possess some amount of short-range order, the nearest atomic neighbours being in almost the same positions as would be the case for crystalline silicon. However there is no long-range order at all. Furthermore, due to the disorder prevailing in the network, about one in every 10⁴ silicon atoms is unable to have four regular bonds with neighbouring silicon atoms; it has a broken or “dangling” bond as it is called (Figure 2(a)). These dangling bonds give rise to “midgap states” that act as recombination centres. In hydrogenated amorphous silicon, most (but not all) dangling bonds are “passivated” by a hydrogen atom (Figure 2(b)); in this case, they no longer contribute to the midgap states and do not at all act as recombination centres. The density of remaining, unpassivated dangling bonds in device-quality hydrogenated amorphous silicon (a-Si:H) is somewhere between 10¹⁴ and 10¹⁷ cm⁻³. (The value 10¹⁴ dangling bonds per cm³ refers to the bulk of the very best a-Si:H layers in the as-deposited or annealed state—i.e., before light-induced degradation or after its removal by annealing; the value 10¹⁷ dangling bonds per cm³ refers to a-Si:H layers after light-induced degradation.) Under the influence of light shining on the amorphous silicon layer, a degradation effect takes place that is characterized by an increase of unpassivated dangling bonds; this is the Staebler–Wronski effect. After about 1000 hours of light exposure (with light intensity equivalent to full sunlight), the dangling bond density tends to saturate at a higher value. If there are many silicon atoms with two hydrogen atoms passivating a broken bond (so-called SiH₂ configuration, Figure 2(c)); it will be relatively easy to break the Si-H bonds and the SWE will be more pronounced. This is seen in porous layers, having a relatively high density of microvoids; a typical microvoid is schematically represented in Figure 3.

In general, one can consider that in amorphous silicon layers the microstructure is an important structural property, which often may not have much effect on the initial layer properties but may strongly influence the light-induced degradation effect (SWE). This is a particularly disturbing situation, because one may have amorphous silicon layers and cells with reasonably good initial properties, which show their deficiencies only after several hundred hours of light-induced degradation. During the 1980s and 1990s, a huge amount of work was undertaken to clarify, understand, and reduce the SWE (for a summary, see [25]). In spite of this tremendous effort, there is to date still no complete understanding of the SWE. Experimentally, one has made the following observations:

In amorphous silicon solar cells, the p–i–n-configuration is used as already stated. Here the light enters into the cell generally through the p layer. For these amorphous silicon p layers, it is customary to employ amorphous alloys of silicon and carbon (a-Si,C:H). Such alloyed p layers have a higher band gap than do unalloyed a-Si:H layers [29]. They are used as so-called window layers: they absorb less light than unalloyed a-Si:H layers. Now all doped amorphous layers have a poor electronic quality and a very high density of recombination centres so that the light absorbed in the p layer is lost and does not contribute to the collected photocurrent. Therefore, it is of advantage for the solar cell if the p layer absorbs less light. If one increases the carbon content in the a-Si,C:H layer too much (more than approximately 40%), the gap does increase further (more than 2.1 eV), and the unwanted absorption further decreases, but the electrical conductivity of the layer decreases to values below 10⁻⁶ S/cm, and the layers are no longer suitable for solar cells [30].

For the photoactive i layers of tandem and multijunction solar cells, the use of amorphous alloys of silicon and germanium (a-Si,Ge:H) has been extensively studied. Such alloys have lower band gaps than unalloyed a-Si:H layers and allow the tuning of the spectrum absorbed in each subcell of the multijunction device to complementary parts of the solar spectrum [31–34]. However, if one increases the germanium content in the a-Si,Ge:H layer over a certain threshold (about 40% Ge content) one obtains (up to now) layers that have high-defect densities, especially in the “degraded” or “stabilized” state—i.e., after light-induced degradation. This is one reason why multijunction cells containing a-Si,Ge:H alloys have so far not led to substantial improvements in solar module efficiencies. Another disadvantage of using a-Si,Ge:H alloys is that germane is much more expensive than silane; this leads to higher overall costs for source gases, even though the utilisation ratio of germane is higher than that for silane. Furthermore, the global availability of germanium as a raw material is at present a reason for concern.

2.2 Gap States in Amorphous Silicon: Mobility Gap and Optical Gap

Classical crystalline semiconductors, such as wafer-based crystalline silicon, have a well-defined energy gap between the valence band and the conduction band. Within this energy gap, practically no electronic states can be seen (except for those due to impurities and crystal defects). In amorphous semiconductors, such as hydrogenated amorphous silicon (a-Si:H) and its alloys, there is a continuous band of states throughout and no actual band gap (Figure 4), though valence and conduction bands can still be identified. These bands have delocalized electronic states; this means that electrons in the conduction band and holes in the valence band can move about, albeit with much lower values of mobility than in the corresponding bands of (mono)-crystalline semiconductors. The highest energy level in the valence band is now given by the mobility edge E_V; for energies E>E_V, the states are localized, which means that the charge carriers are “trapped” and cannot move about freely.

For energies above E_V, we first have the valence band tail, where the density of states N(E) decreases exponentially. Thereby, the energy constant denotes the energy needed for the exponential function to fall off by a factor of e (=2.718…). In the best device-quality a-Si:H layers, is about 45 to 50 meV. is a measure for the disorder in the amorphous network; the higher the value of , the higher the disorder. can be evaluated by measuring the optical absorption coefficient as a function of photon energy (see Figure 10). Due to the capture and recombination kinetics, states in the valence band tail do not play the role of recombination centres; rather, they act as traps for the free holes during solar cell operation. One can assume that for every free hole in the valence band (i.e., for every free hole contributing to the photocurrent in the solar cell), there are 100 to 1000 “trapped” holes in the valence band tail. These trapped holes constitute a positive charge that can deform and reduce the internal electric field within the i layer of p–i–n-type solar cells.

Now, if the light enters into a p–i–n-type solar cell from the p side, the majority of the photogenerated holes have less far to travel than if the light enters from the n side: in the first case, they are mostly generated near the p–i interface; in the second case, they are mostly generated near the n–i interface. Thus, the field deformation through trapped charge is less pronounced in a-Si:H p–i–n-type solar cells, if the light enters from the p side: this is what is generally done.

Note that the band tails are not modified by light-induced degradation (SWE).

Above the valence band tail (for higher energies—i.e., to the right in Figure 4) we have the midgap states, given by the dangling bonds. These states act as recombination centres and are therefore detrimental to the functioning of solar cells as they directly limit the collection of the photogenerated carriers. The density of midgap states is increased by a factor of 10² to 10³ by light-induced degradation (SWE).

At higher energies we can see the conduction band tail, where the density of states N(E) also follows an exponential law, but with an energy constant that is about half the value of . As the conduction band tail is much less pronounced than the valence band tail, it does not play a great role in solar cells—the electrons trapped within the conduction band tail do not noticeably deform the electric field within the i layer of p–i–n-type solar cells. (The conduction band tail does play an important role in n-channel thin-film transistors, where the electrons are the dominant charge carriers, and the density of holes is very low and of no importance at all.)

A mobility edge E_C separates now the localized states in the conduction band tail from the states in the conduction band with its delocalized electronic states. Instead of a “true” band gap as in (mono)-crystalline semiconductors, we now have, in amorphous silicon, a mobility gap (E_C−E_V), where there are the localized gap states. It is not straightforward to determine the mobility gap. It is easier to determine the so-called optical gap, a quantity that is extrapolated from measurements of the optical absorption coefficient. The optical gap is found to be about 100 meV smaller than the mobility gap [35–36].

There are different methods used for determining the optical gap. The most common method is the one proposed by Tauc et al. ([37], see also [38]). The method consists of measuring the absorption coefficient α as a function of photon energy E=hν=hc/eλ≈1.240 [eV])/(λ[μm]), for photon energies above the band gap energy. Here ν and λ are the frequency and the wavelength of light; h is Planck’s constant, c the velocity of light, e the charge of an electron (unit charge). One then plots (α(E)E)^1/2 as a function of E; this plot gives us more or less a straight line (Figure 5). The intersection of this straight line with the abscissa ((α(E)E)^1/2=0) gives us the value of the Tauc ga , generally used as estimate for the optical gap.

The values measured for the optical gap in a-Si:H layers are significantly higher than are the band-gap values for crystalline silicon (c-Si); they are in the range 1.6 eV to 1.85 eV, compared to 1.1 eV for c-Si. Furthermore, the band-gap values of a-Si:H layers vary according to the deposition conditions: layers deposited at higher temperatures have lower band-gap values; layers deposited with high values of hydrogen dilution have higher band-gap values, as long as they remain amorphous and do not become microcrystalline (with a substantial crystalline volume fraction—more than 20%). In fact, if one uses high hydrogen dilution values but just avoids crossing the transition from amorphous to microcrystalline silicon, one can obtain so-called protocrystalline [39] or polymorphous [40] silicon layers that have band gaps around 1.9 to 2.0 eV, more short-range and medium-range order than do standard a-Si:H layers and a very small fraction of tiny crystallites. These layers constitute a promising topic for future research. Protocrystalline p layers are apparently used as window layers in certain amorphous silicon solar cells.

2.3 Conductivity and Doping of Amorphous Silicon

3.1 Structure of Microcrystalline Silicon

Hydrogenated microcrystalline silicon (μc-Si:H), as deposited by PECVD from a mixture of silane and hydrogen, is a mixed-phase material containing a crystalline phase (with tiny crystallites grouped into “conglomerates” or “clusters”), an amorphous phase, and voids (which are very often not real voids but just regions with a lower density [44–45]). By varying the hydrogen dilution ratio R=[H₂]/[SiH₄] in the plasma deposition (where [H₂] denotes the rate of hydrogen gas flow into the deposition system and [SiH₄] the rate of silane gas flow), one can obtain many different types of layers: (a) at low hydrogen dilution—i.e., at low values of R, amorphous layers; (b) by slightly increasing R, layers with mainly an amorphous phase and a low concentration of very tiny crystallites (such as protocrystalline silicon layers); (c) at still higher hydrogen dilution, layers with about 50% amorphous phase and 50% crystalline phase; and (d) at very high values of R, highly crystalline layers, which tend to have a large concentration of cracks or voids and thus constitute low-density, porous material.

When deposited on a glass substrate, the μc-Si:H layers usually start off with an amorphous incubation phase and the nucleation of crystallites only begins later on. This is shown schematically in Figure 7 [46]. Within μc-Si:H solar cells, the situation is more complex, as the μc-Si:H intrinsic layer is deposited on a p-doped or an n-doped microcrystalline layer, and the latter on a rough substrate. One strives, in fact, to avoid the formation of an amorphous incubation layer (which leads to a reduction in solar cell performance) by starting off the deposition with a relatively high value of R. The value of R at which microcrystalline growth starts depends very much on the deposition parameters, such as plasma excitation frequency, substrate temperature, and deposition pressure. The values of R indicated in Figure 7 are merely given as an example and are typical of deposition at relatively high pressures (2 to 3 Torr), with a plasma excitation frequency of 13.56 MHz.

It is generally assumed today that most of the defects (i.e., most of the recombination centres) in μc-Si:H are located at the boundaries of the conglomerates or clusters. It is also assumed that these defects are passivated by the amorphous phase. For this reason, one uses, as intrinsic layers within p–i–n- or n–i–p-type μc-Si:H solar cells, layers with about 50% crystalline volume fraction.

Let us take a closer look at the microstructure of a typical μc-Si:H layer. Figure 8 shows part of a μc-Si:H layer taken as a high-resolution transmission electron microscopic (TEM) image within a conglomerate of silicon crystallites: the latter have diameters between 10 and 20 nm and are embedded into an amorphous silicon matrix. The conglomerates themselves are separated by a varying amount of amorphous silicon, cracks and voids, and low-density material [47]. The microstructure is highly complex. In addition, the μc-Si:H layer is neither uniform nor anisotropic, because the conglomerates form cones that widen up toward the top of the layer until they touch each other as schematized in Figure 7.

The most convenient way to assess the “crystallinity” of μc-Si:H layers is to use Raman spectroscopy. With this technique, one investigates the local atom–atom bonding structure of a material by studying the interaction of monochromatic incoming light (photons of a given energy) with the bond vibrations in the material (phonons). The energy of the incoming photons is shifted by the energy of the phonon involved in the interaction. Due to scattering of light, one is able to collect and analyse the outgoing photons with the energy shifts. The amplitude of the scattered light is measured as a function of the shift in photon energy: this constitutes the Raman spectrum. Thereby, a unit called wavenumber is used, which is simply the reciprocal of wavelength and is expressed in cm⁻¹. The conversion between photon energy E, wavelength , and wavenumber ν is . Crystalline silicon has a narrow peak in its Raman spectrum at 520 cm⁻¹ and, due to defective regions, a tail around 500 to 510 cm⁻¹ wavenumbers, whereas amorphous silicon exhibits a broad Raman signal centred at 480 cm⁻¹[48]. By suitably analysing the Raman spectrum, one finds the total signal intensity I_c due to crystalline contributions and the total signal intensity I_a due to the amorphous contribution. The ratio I_c/(I_c+I_a) is a semiquantitative indication for the crystalline volume fraction and is called Raman crystallinity. It is rather difficult and cumbersome to evaluate the actual crystalline volume fraction (this can be done, e.g., by a high-resolution TEM but requires the attribution of the various regions in the layer to the crystalline and amorphous phases, which is not an easy task). Therefore, the value of the Raman crystallinity is generally used for μc-Si:H layer optimization in connection with solar cells. The best solar cells are obtained for values of Raman crystallinity around 50% to 60%.

3.2 Optical Absorption, Gap States, and Defects in Microcrystalline Silicon

Compared with intrinsic a-Si:H layers, intrinsic μc-Si:H layers show the following striking differences.

Figure 10 shows the main differences in optical properties, between typical μc-Si:H layers and typical a-Si:H layers, by displaying, in a logarithmic scale, the absorption coefficient α(hν) for the applicable spectral range—i.e., for photon energies E_photon = hν from 0.7 eV to 3.5 eV—corresponding to wavelengths from 350 nm to 1750 nm.

While interpreting Figure 10, note the following:

3.3 Conductivities, Doping, Impurities, and Ageing in Microcrystalline Silicon

4.1 Role of the Internal Electric Field

All solar cells function according to the following two principles:

Figure 13 shows a comparison for the profile of the internal electric field E(x) between p–n and p–i–n diodes. Note that for zero external (applied) voltage V, the integral ∫ E_int(x)dx is, for both types of diodes equal to the built-in voltage V_bi, a parameter, which is approximately equal to 1 Volt, for all forms of silicon. If V≠0, one can write: V=∫−E_ext(x)dx; this means that if the applied voltage is negative, it gives a reverse bias to the diode and the electric field is augmented (as shown in the figure); if the applied voltage is positive, it gives a forward bias to the diode and the electric field is reduced. As V approaches the open-circuit voltage V_oc of the solar cell, the internal electric field becomes strongly reduced. Note that in relation to the corresponding energy gaps, V_oc values of p–i–n- and n–i–p-cells are fundamentally lower than the theoretical limit values found for p–n cells.

Due to the action of the internal electric field E(x) in the i layer of p–i–n- and n–i–p-type solar cells, transport (and collection) of the photogenerated carriers is now governed (as long as V is not too high) by the drift length L_drift of both electrons and holes within the i layer. For thin-film silicon, one generally writes L_drift=μτE, where μ is the mobility of the carrier (electron or hole), τ the lifetime, and E the magnitude of the prevailing electric field. One can consider that, for a given value of E, L_drift will be approximately equal for both carriers (electrons and holes) [67,68]. One finds also that, at short-circuit conditions, the drift length has a value that is about 10 times higher than the minority-carrier diffusion length [69]: this is, of course, the reason, why one uses the p–i–n (or n–i–p) configuration for all a-Si:H and for most μc-Si:H. As previously stated, the minority-carrier diffusion lengths in these materials would be too small, often quite a bit smaller than the i layer thickness d_i.

Thus, for thin-film silicon solar cells on glass substrates, the structure shown in Figure 14 is obtained. Note that for optimal performance of a-Si:H solar cells, the light has to enter the solar cell through the p layer; it is only then that the deformation of the internal electric field through trapped charge in the valence band tail can be kept negligibly small (see Section 2.2). Therefore, if the solar cell is deposited on a substrate, which is not fully transparent (or, unlike glass, does not remain for several decades fully transparent but eventually becomes yellowish like most polymers), then the deposition sequence n–i–p is used. A textured reflector layer (e.g., textured silver) is deposited on the substrate followed by the n, i, and p layers; on top of the p layer, a transparent conductive oxide (TCO) and a grid for current collection are used. The n–i–p-configuration is regularly employed not only for solar cells deposited on stainless steel substrates but also on most polymer substrates.

Although it would be feasible to let the light enter through the n side in single-junction μc-Si:H solar cells (because the trapped charge in the valence band tail is much less prominent than in a-Si:H), μc-Si:H solar cells are, in practice, almost exclusively used in tandem and multijunction structures, together with a-Si:H solar cells, so that once again one prefers to let the light enter from the p side.

Because the doped layers in thin-film silicon do not have a sufficiently high conductivity as does wafer-based crystalline silicon, a TCO layer has to be used as the contact layer adjacent to the p layer—on the side where the light enters the solar cell. In practice, the TCO layer is often textured and thus contributes to light scattering (see Section 4.5). In the case of p–i–n-type solar cells (as shown in Figure 14), the thin-film silicon layers are deposited on top of this TCO layer: the TCO layer has therefore to withstand the action of a silane plus hydrogen plasma. In this case, F-doped SnO₂, Al-doped ZnO, or B-doped ZnO layers are used as TCO layers. (The SnO₂ layer tends to be reduced by the hydrogen-rich plasma used for the deposition of μc-Si:H and should in this case be covered by a thin protective layer—e.g., by a thin ZnO or a thin TiO₂ layer [47]). In the case of n–i–p-type solar cells, the TCO contact layer is deposited after the thin-film silicon layer and very often indium tin oxide (ITO) layers are used.

4.2 Recombination and Collection

In amorphous silicon solar cells, a large part of recombination is bulk recombination and takes place in the centre of the i layer due to the dangling bonds acting as recombination centres. According to [74], it is mainly the neutral dangling bonds that play an essential role in this part of recombination. One may speculate that the situation is essentially the same in microcrystalline silicon solar cells. However, it is important to realize that interface recombination in thin-film silicon solar cells can also play a significant role, as drawn schematically in Figure 16. In this case, charged (or ionized) dangling bonds will act as recombination centres. Their density can be substantially higher than the dangling bond density in the bulk of the i layer. In amorphous silicon solar cells, such interface problems arise mainly from cross-contamination—i.e., from dopant atoms having diffused during the fabrication process—from one of the doped layers into the i layer. For microcrystalline silicon solar cells, there is the additional problem of crystalline growth: one often has at the beginning of the growing microcrystalline layer a layer of inferior crystallographic properties that in extreme cases can even be amorphous.

For amorphous silicon solar cells, the dangling bond density will be dramatically increased by light-induced degradation (the Staebler–Wronski effect). By keeping the solar cell very thin, i.e., by choosing an i-layer thickness d_i in the range of 200 to 300 nm, and by adopting all the other measures described earlier (see Section 2.1) one is able today to fabricate amorphous silicon solar cells with a relative efficiency loss of just 10% to 20% due to light-induced degradation.

If the recombination in a thin-film silicon solar cell becomes excessive, the resulting deficiency in photocarrier collection can be mainly identified by: (a) decrease in the fill factor FF; (b) a deficiency in the spectral-response/external quantum efficiency (EQE) curve of the cell. In case (b) it is particularly instructive to compare two EQE-curves: a first curve with no bias voltage and a second curve with a reverse bias voltage of −1 to −2 Volts. If these curves essentially do not differ, the internal electric field at no bias is sufficiently high to collect practically all photogenerated carriers. However, a difference between the two curves (see Figure 18) indicates collection problems, i.e. the internal field must be increased by an external bias to enhance the carrier collection. If the difference is seen at short wavelengths, it means that the collection problem occurs at the interface through which light enters in to the cell (at the p–i interface, for p–i–n-cells illuminated from the p side); if it occurs at longer wavelengths, it means that the collection problem occurs in the bulk of the i layer. Figure 17 represents the difference between (a) blue light and (b) red light entering into a p–i–n-type solar cell. In the case of blue light, there are photogenerated holes and electrons only near the p–i-interface, so that recombination can only take place there. We are, thus, only probing the region of the i layer near the p–i-interface. In the case of red light, holes and electrons are generated throughout the i layer and we are probing the whole i layer.

Figure 18 shows typical external quantum efficiency curves of microcrystalline p–i–n-type cells. In (a) we are looking at cells with different p layers (in the case of the “bad” p layer, it would appear that the recombination in the adjoining regions of the i layer is also increased, possible through cross-contamination or another effect). In (b) we are looking at cells with and without contamination; the contamination results in a reduction or deformation of the electric field and a collection problem throughout the i layer as described earlier.

Spectral-response and EQE measurements are a powerful tool for the diagnosis of thin-film silicon solar cells. Their detailed interpretation needs, however, considerable experience, and goes well beyond the scope of the present chapter. Their main advantage is that they allow us to assign defects and shortcomings to various regions of the cell. Their “geometrical sensitivity” is excellent for the zone where the light enters into the solar cell, but it is very much reduced at the far end of the solar cell toward the back reflector (BR) in Figure 17. In order to probe the far end of the cell, it is necessary to employ a bifacial configuration, in which one can let the light enter “from the back”—i.e., through the n layer—and perform EQE measurements in this arrangement [75].

4.3 Shunts

In thin-film silicon solar cells and modules, shunts are a common problem. Shunts result in a reduction of the fill factor FF already at standard illumination levels of 100 mW/cm² light intensity. In order to distinguish them from collection problems (which also result in a reduction of the fill factor FF at standard light levels), it is necessary to measure the J-V curve of the cell or module at low light levels. If there is a substantial further drop in the FF as the light level is decreased to less than 1 mW/cm², we are facing a shunt problem, not a collection problem.

To conceptually separate the effects of collection problems and shunts, look at Figures 19 and 20. Figure 19 shows the common equivalent circuit universally used for all solar cells. This equivalent circuit is indeed very simple and therefore convenient for the design of entire electrical systems containing solar cells and modules. However, it does not provide us with physical insight into what is happening within the solar cell, especially within a thin-film silicon solar cell. In fact, the effects of shunts and of collection problems are merged together in the parallel resistance R_p. In Figure 20. a modified equivalent circuit, introduced by [76] is given: here the recombination or collection problems are symbolized by a “current sink” J_rec (opposite of a current source). The recombination current J_rec is proportional to the photogenerated current J_ph and will therefore result in performance reduction and in a FF reduction that is independent of light intensity. The resistance R_shunt represents now “true” Ohmic shunts. From the electrical diagram of Figure 20, one can see that R_shunt will play an increasing role as the light level, and thus, both the photogenerated current J_ph and the recombination current J_rec are decreased proportionally to the light level. On the other hand, the current through R_shunt (approximately V/R_shunt) is not reduced to the same extent, because the J-V curves at low light levels show only minimal voltage decreases compared to the current decreases.

Shunts in thin-film silicon solar cells can originate from various fabrication problems.

Shunts and collection problems can be quantified with the help of the variable intensity measurement (VIM) method. VIM analysis is based on measuring the J-V characteristics of the solar cell or module for a whole range of light intensities. An evaluation of these curves as indicated in [79,80] will enable us to determine to what extent the performance of the cell or module—and, in particular, its fill factor—are affected by shunts and to what extent they are affected by collection problems.

4.4 Series Resistance Problems

In thin-film silicon solar cells and modules, the electrical contacts are made to the front and back contact layers. Generally the contact on the “front side” (where the light enters into the cell or module) is given by a transparent conductive oxide layer. Here a compromise between optical and electrical requirements must be found: if the TCO layer is chosen to be too thin to keep the optical absorption low, it will have a high sheet resistance and will contribute in a pronounced manner to the series resistance R_series shown in Figure 20. If the TCO layer is too thick, it will absorb too much light, which is lost for photocarrier generation. The thickness of the back contact layer is less critical, since these layers are either highly conducting metal layers or, in case of TCO layers with additional reflector layers, can be chosen thicker for lower sheet resistance. For small individual laboratory test cells, sheet resistances usually do not cause significant electrical losses.

However, for modules that feature a monolithic series connection of cells (see Section 6.3, especially Figure 33), the earlier mentioned compromise between optical and electrical losses must be carefully taken into account.

For a cell with a photoactive width w, the relative power loss at the maximum power point is given by the expression [42]

where R_sheet denotes the sheet resistance of the TCO layer and j_mp and V_mp are current density and voltage at the maximum power point, respectively. The term (ΔP/P)_mp×100% then gives us the reduction in fill factor due to the sheet resistance of the TCO. One can draw the following conclusions from the preceding expression:

In addition to the sheet resistance of the TCO, poorly doped p and n layers will also contribute to increase the value of R_series. Furthermore, contact problems between the TCO (which is usually n type) and the adjacent p layer, as well as unsatisfactory tunnel or recombination junctions within tandems and multijunction cells (see Section 5.1) will also lead to an increase in R_series. If the cell or module is properly designed and fabricated, the loss in fill factor FF due to series resistance R_series, should, at standard illumination conditions, not exceed a few percentage points. If FF increases when reducing the light intensity, this is typically a sign of losses due to series resistance R_series. If these losses are excessive, they can be evaluated by VIM analysis [79,80].

4.5 Light Trapping

In thin-film silicon solar cells and modules, it is imperative to limit the thickness d_i of the i layer (which is usually the only layer contributing to useful absorption and photogeneration) in order to keep the internal electric field sufficiently high for effective carrier collection. A higher value of d_i, such that the solar cell could absorb a large part of the solar spectrum, would lead to a very high light-induced degradation in amorphous silicon solar cells; in microcrystalline silicon solar cells, it would mean long deposition times and high material and fabrication costs. This can be seen in Figure 23, where the absorption coefficient α and the penetration depth d_pen = 1/α of monochromatic light are plotted as a function of the wavelength λ (or of the photon energy E_photon = hν) of the light. Figure 23 (from [81]) compares five materials commonly used for solar cells. It can be seen that for both CuInSe₂ and CdTe the penetration depth d_pen is approximately 1 μm or lower for most of the light in the visible range of the solar spectrum—i.e., for light with wavelengths <700 nm. For the various forms of silicon, this is clearly not the case: d_pen is larger than 1 μm for  = 700 nm, and it remains still higher than the corresponding penetration depths for CuInSe₂ and CdTe down to wavelengths >500 nm.

In order to absorb a sufficient portion of the incoming light in a thin-film silicon solar cell, it is necessary to increase the average optical path length within the i layer by a factor m. With the help of a back reflector, one may double the optical path and reach m = 2. By using improved light-scattering techniques. one presently reaches substantially higher values of m—around 10 or even higher.

In p–i–n solar cells, this is currently done by using randomly textured TCO layers. Since the texture of the front TCO layer is widely replicated for the back reflector layer, scattering in reflection further increases the optical path length, particularly for the weaker absorbed wavelengths toward the red portion of the incident light.

In n–i–p solar cells, one deposits first the back reflector. By depositing a randomly textured back reflector, such as a rough silver layer [82–84], one obtains light scattering at this point, and, to some extent, also at the entry point of the light—as the layers deposited on top of the back reflector take on more or less the same form.

All present solar cells and modules use randomly textured structures for light scattering; however, from a research point of view, periodical structures for light scattering are increasingly being investigated [85–87].

The design of light-management and light-trapping schemes and their incorporation into thin-film silicon solar cells is by no means trivial. There are a whole series of questions yet to be solved:

It can therefore be stated that improved light management and light trapping techniques hold great promise for further enhancing the performance of thin-film silicon solar cells.

5.1 General Principles

In thin-film silicon solar cells, one so far almost exclusively uses two-terminal tandem solar cells. These devices stack two subcells, one on top of the other as indicated in Figure 25.

In the ideal case, each of the subcells absorbs the same amount of photons and therefore basically photogenerates the same electrical current. This is the case of “current matching,” and it can only be achieved for a given spectrum of the incoming light. If the spectrum of the incident light changes, or if one of the two subcells changes its properties in a more pronounced way than the other (e.g., because of light-induced degradation or because of a difference in temperature coefficients), we will have a mismatch in the two currents; in this case, the subcell with the lower current will limit the current of the tandem. The voltage of the tandem is equal to the sum of the voltages of the two subcells.

As for the fill factor of a tandem, it is difficult to predict. In the case of current mismatch, the fill factor is artificially increased—but because of the loss in current, this does not result in a net increase of the conversion efficiency. One should therefore be very careful when trying to draw conclusions from the value of the fill factor in a tandem (or, even more so, in a multijunction cell) [89].

Why does one designate the intermediate zone between top subcell and bottom subcell as recombination junction or tunnel junction? The electrons photogenerated in the i layer of the top subcell will travel toward the n layer and then enter into the intermediate zone; similarly, the holes generated in the i layer of the bottom subcell will travel toward the p layer and then enter into the intermediate zone. To ensure current continuity, all electrons travelling downward (in Figure 25) will have to recombine with the holes travelling upward. This recombination process is in general facilitated by “tunnelling” of the electrons and holes from the conduction and valence band, respectively, of the doped layers, into imperfection states (midgap defects) localized at the interface (see Figure 26) so that one also uses the term tunnel junction.

The tunnel or recombination junction is composed of an n layer and a p layer; at least one of these layers (or, alternatively, the interface between these two layers) should contain a high density of midgap defects to facilitate recombination. Both layers are generally strongly doped, and, if possible, microcrystalline and not amorphous to enable tunnelling. Poorly designed tunnel or recombination junctions will contribute to increase the series resistance of the cell.

What are the advantages of using the tandem configuration instead of the single-junction configuration? There are some evident advantages.

On the other hand, there is the delicate problem of current matching and a significant increase in the complexity of the fabrication process.

5.2 a-Si:H/a-Si:H Tandems

The simplest form of tandem cell is the one shown in Figure 27: an amorphous silicon top subcell is sitting on top of a bottom subcell, which consists also of amorphous silicon. If the band gaps of both subcells were exactly the same, then the i-layer thickness d₁ of the top subcell would have to be very much smaller than the i-layer thickness d₂ of the bottom subcell; this follows from the exponential absorption profile of the incident light and might lead to either an exceedingly low value of d₁ (and thus an increased tendency for the top subcell to have shunts) or a very large value of d₂ (with a corresponding undesired light-induced degradation effect in the bottom subcell). Thus, one attempts to increase the band gap of the top subcell—either by alloying with some carbon—i.e., using an a-Si,C:H layer [90]—or by changing the deposition conditions—i.e., by increasing the hydrogen dilution in the deposition plasma) and to decrease the band gap of the bottom subcell (either by alloying with some germanium—i.e., using an a-Si,Ge:H layer—or by changing the deposition conditions—i.e., by decreasing the hydrogen dilution and increasing slightly the deposition temperature). At present, the most successful a-Si:H/a-Si:H tandems do not use any alloying [20]; they can attain a total area stabilized efficiency of more than 7%.

5.3 Triple-Junction Amorphous Cells with Silicon–Germanium Alloys

Triple-junction cells with amorphous silicon–germanium alloys and modules based on this triple-junction design have been produced since the mid-1990s by the firm United Solar Ovonic. These devices are deposited on stainless steel substrates in an n–i-p sequence (see Figure 28). The light enters into the solar cell through the topmost p layer on which an indium tin oxide layer has been deposited as a TCO. In order to improve (reduce) the contact resistance and correspondingly reduce also R_series, a metal grid is deposited on top of the ITO layer. The grid has, however, the disadvantage of “shading” the solar cell, thus reducing the photoactive area by almost 10%. In the laboratory, an active-area initial efficiency of 14.6% and an active-area stabilized efficiency of 13% were achieved for small-area cells [21]. The stabilized total-area efficiency is thereby (because of the presence of the grid) only 12.1%: this is at present roughly the same value as the stabilized total-area efficiency of the best micromorph tandem cells (see Section 5.4).

The stainless steel substrates are flexible and thus lend themselves to roll-to-roll fabrication of lightweight, unbreakable modules especially suited for building-integrated photovoltaics. Module efficiencies are, however, considerably lower than the previously mentioned cell efficiencies—about the same value as for tandem a-Si:H/a-Si:H modules on glass. The relatively large discrepancy between the cell and module efficiencies are mainly caused by the method of interconnecting cells to a module (see also Sections 6.2 and 6.3). Unlike the monolithic interconnection that is applicable to nonconducting substrates like glass, relatively large-area cells on the conducting stainless steel substrate are individually connected, very much like the stringing of crystalline silicon cells. These large areas result in large total currents and hence require special precautions to reach low series resistances R_series. Therefore, the grid mentioned previously is absolutely necessary to balance losses by series resistance with active-area losses. In addition, the deposition rate has to be increased for economic reasons, whereby the layer quality will generally suffer, resulting in lower cell performance.

5.4 Microcrystalline–Amorphous or “Micromorph” Tandems

One of the most promising tandem cells, is the so-called micromorph tandem, containing, as shown in Figure 29, a microcrystalline silicon bottom subcell and an amorphous silicon top subcell. This structure was introduced by the Neuchâtel group in 1994 [91]. It has since then been the focus of intensive R&D efforts and stabilized small-area cell efficiencies of 12% were reported a few years back [92]. The micromorph tandem is of special interest, because the band gap combination of 1.1 eV/1.75 eV, as approximately given by μc-Si:H and a-Si:H, would correspond to the “ideal” band-gap combination for a tandem cell, provided (1) that collection problems can be neglected and (2) that all the photons with energies above the band-gap energy are usefully absorbed in the corresponding subcell [22]. These conditions are, in practice, far from being fulfilled. The amorphous top subcell suffers from the light-induced degradation effect (Staebler–Wronski effect) and therefore has to be kept as thin as possible, with an i-layer thickness d_i of 200 nm (or less). This means that the top subcell will not absorb enough light.

The remedy used here to enhance photogeneration in the amorphous top subcell is to provide an intermediate reflector (IR) between the top and the bottom subcells in such a way that the short-wavelength components of the light are reflected back into the amorphous top subcell but the long-wavelength components are (as far as possible) passed on, with little attenuation, into the microcrystalline bottom subcell. As intermediate reflector, one needs basically a material with a refractive index n lower than that of amorphous or microcrystalline silicon (n≈4). Physically, a ZnO layer, with a refractive index n = 2, is an excellent candidate; it was used in the first trials [93]. But ZnO is not a very convenient solution from the fabrication point of view, so later solutions, including doped silicon oxide layers, were introduced. These layers can be conveniently deposited by plasma deposition, just like the a-Si:H and μc-Si:H layers themselves (see [94–96]). The design of an efficient and spectrally selective intermediate reflector remains one of the primary research topics for micromorph tandems [97].

The microcrystalline bottom subcell has also to be kept as thin as possible. Here it is not the light-induced degradation, but the fabrication cost, that is the issue, The fabrication costs associated with the deposition of the microcrystalline silicon intrinsic layer are one of the main cost factors for micromorph module production. On one hand, the plasma-deposition equipment represents a very large investment. Now operation and depreciation costs for the plasma reactor are proportional to layer thickness. If the layer, which has to be deposited, is thick, the deposition process will correspondingly take a long time, thus diminishing the production throughput. Furthermore, the materials costs for gas inputs into the plasma reactor (silane and hydrogen) are also proportional to the thickness of the deposited layer. Therefore, one currently proposes to use microcrystalline silicon bottom cells with a total thickness of less than 1 μm [23]. Referring to Section 4.5, and particularly to Figure 23, one can immediately see that this will only be possible with a very effective light-trapping scheme. The multiplication factor m by which the optical path becomes longer than the intrinsic layer thickness d_i should be as high as 10, or even higher, in order to absorb a sufficiently large part of the incoming light.

Thus, one may state that most of the research problems to be solved in the coming years, in order to increase the efficiency of micromorph tandems, are related to improvements in light management (e.g., see [98]).

An interesting further development of the micromorph tandem concept is the extension to triple-junction cells. In these triple-junction cells, one generally retains an amorphous subcell on top and a microcrystalline subcell on the bottom. For the middle subcell, however, a number of different possibilities have been tried out.

One comes therefore to the conclusion that today there does not exist so far a clear preference for the structure of micromorph triple-junction cells. The different possibilities need further in-depth investigation, especially with respect to the light-management problem, which now truly becomes quite complex. On the other hand, if one goes for the triple-junction configuration and the increased fabrication costs associated with such a configuration, then one would hope to be able to avoid the use of intermediate reflectors.

6.1 Deposition of the Thin-Film Silicon Layers

The deposition of the thin-film silicon layers is the central step in the fabrication of solar modules based on amorphous and microcrystalline silicon. The method generally used—by all manufacturers up to now is plasma-enhanced chemical vapour deposition (PECVD) in a capacitively coupled reactor, as indicated schematically in Figure 30. For the deposition of the intrinsic layers, a mixture of silane and hydrogen is used; for the deposition of the relatively thin doped layers, diborane or trimethlyboron is added for the p layer, and phosphine is added for the n layer (see Section 2.3).

At present, the deposition of the intrinsic silicon layer is the production step that takes the largest share of total module production costs. It is also one of the most critical steps from the point of view of module performance. The high production costs for this fabrication step stem principally from two factors:

A capacitively coupled plasma-deposition system, as indicated in Figure 30, contains electrically neutral molecules, some of which are the growth radicals, actively contributing to layer growth. It also contains electrons and positively charged ions. Increasing the deposition rate is generally accomplished by increasing the electrical power fed into the deposition system. This not only increases the densities of growth radicals but also enhances the ion bombardment on the growing silicon layer. High-energy ion bombardment generally leads to the creation of additional defects on the growing layer and must therefore be avoided. The following are the two most common methods to avoid high-energy ion bombardment.

There are a number of other factors that have to be taken into account when designing a deposition system for thin-film silicon layers.

The design of economically attractive deposition systems for thin-film silicon modules is therefore a very important topic, which will have a decisive effect on the future success of this type of modules.

To conclude, one may remark that hot wire deposition (also called catalytic CVD) is increasingly being proposed and tested as an interesting alternative to plasma deposition (e.g., see [109–112]). This method is, however, at present not yet used in industrial module fabrication. Hot wire deposition completely avoids ion bombardment; instead of a plasma, one uses a filament heated to very high temperatures (around 1800°C) to obtain silane dissociation and generation of growth radicals. This method also has the advantage of a relatively high gas utilisation. Potentially it should be able to lead to the fabrication of solar modules at high deposition rates and with excellent performances. Excellent solar cells have up to now been fabricated by hot wire deposition at low deposition rates, and individual layers with excellent properties have also been fabricated at high deposition rates. But the combination of a high efficiency and a high deposition rate has so far only been demonstrated in isolated cases. Furthermore, it would seem that a certain amount of ion bombardment with low-energy ions may, in some cases, be beneficial for the fabrication of solar cells. Thus, the combination of plasma deposition and hot wire deposition [113] may be especially interesting to investigate.

6.2 Substrate Materials and Transparent Contacts

There are basically two possible configurations for thin-film silicon solar cells.

Most manufacturers opt at the moment for the superstrate configuration. Here the glass is coated with a transparent conductive oxide serving as a front contact. For commercial modules, one so far employs mostly fluorine-doped SnO₂ as TCO material; SnO₂ is usually deposited by atmospheric pressure chemical vapour deposition (APCVD) “online” within the float-zone process employed for glass production [114]. This is probably the most cost-effective way of producing the glass substrate together with the transparent front contact. As mentioned in Section 4.5, the transparent contact has to be textured in order to obtain light trapping. In the case of the present SnO₂ layers, which are manufactured online, the texturing is generally not optimized for light trapping in a particular cell configuration. One is therefore currently developing highly textured SnO₂ layers with off-line processes on a pilot-line basis (e.g., see [115]). If these experimental processes lead to the desired result, one will probably attempt later to incorporate them (if possible) as an online version within the float-zone process of the glass producer. (Note that SnO₂ is chemically reduced by a high-power hydrogen-rich plasma as used for depositing μc-Si:H. Therefore, if one deposits a μc-Si:H solar cell on a SnO₂ layer, one has first to cover the latter by a thin protective layer of TiO₂ or ZnO. This consideration is of importance for the deposition of single-junction μc-Si:H cells, as studied in research laboratories, but it is irrelevant in case of micromorph tandems, where the SnO₂ is no longer exposed to the extreme deposition conditions mentioned previously but is instead protected by the amorphous top subcell, which is deposited first.)

As an alternative to SnO₂, one is at present increasingly introducing ZnO as transparent contact material [116]. ZnO acts as an effective diffusion barrier against any contaminants diffusing from the underlying glass substrate. Another advantage is that it can be deposited in a relatively easy manner by the manufacturer of the thin-film solar cell and at a lower deposition temperature than SnO₂. There are two main deposition methods for ZnO: (a) sputtering and (b) low-pressure chemical vapour deposition (LPCVD). Sputtering leads to smooth layers that have to be chemically etched in order to obtain the desired texture for light trapping. Sputtered ZnO layers are generally doped with aluminium in order to obtain the desired high conductivity. The LPCVD process leads directly to textured ZnO layers, which have a high effectiveness for light trapping. LPCVD ZnO layers are generally doped with boron.

In the substrate configuration, the TCO front contact layer is deposited at the end of the deposition process. It has correspondingly to be deposited at a relatively low temperature (typically <200°C). Therefore, SnO₂ is not an option to be considered. One can use ZnO. One can also use indium tin oxide (ITO=0.9 In₂O₃/0.1 SnO₂). Note that ITO is never used in the superstrate configuration because the In₂O₃ would be strongly chemically reduced by the exposure to hydrogen during the initial phases of the PECVD, and thus, ITO would widely loose its transparency. Besides, indium would diffuse into the silicon layers, rendering the latter unusable.

Substrates and transparent contact layers, as well as reflecting layers, are an important part of any thin-film silicon cell. They determine to a large extent the performance (especially the current density) of the solar cell. The interaction of textured front and back layers is complex and different for the superstrate and substrate configurations. (See [117] for an analysis, which is restricted, however, to the case of single-junction microcrystalline solar cells.) These front and back layers are therefore increasingly the focus of current research efforts, as already described in Section 4.5. The focus on these parts of the solar cell is all the more warranted, as they also account for a significant part of the total fabrication cost—together with other “external features,” cell-interconnection schemes (see Section 6.3), module encapsulation, and current collection schemes (e.g., metallic grids that can be helpful in reducing the series resistance).

6.3 Laser Scribing and Cell Interconnection

Individual cells have to be electrically interconnected, most often by a series connection in order to form a complete module. Whenever the thin-film silicon solar cell has been deposited on an electrically isolating material, such as glass or polymer, one uses a monolithic interconnection scheme as shown in Figure 33. (The monolithic interconnection scheme cannot be used if the thin-film silicon solar cells have been deposited on an electrically conductive substrate, like stainless steel, unless an isolating interlayer is first deposited. Such an interlayer would, however, be prone to electrical shorts, and would lead to additional complications or costs.)

In the monolithic interconnection scheme one proceeds as follows if one is using the superstrate configuration with a glass superstrate.

For optimal cell performance, the photoinactive interconnection width Δw should be as low as possible; it is given by the precision and reproducibility of the laser-scribing system. In the best cases, it is around 0.2 mm [119]. The interconnection area, defined by Δw×l, is lost for photocarrier generation. Thus, there is a relative loss of incoming light, given by (Δw)/(w+Δw).

This loss term (areal loss) decreases with increasing stripe width w. On the other hand, because of the sheet resistance R_sheet of the TCO layer, there is an increase in series resistance R_series and a corresponding electrical loss of the solar cell leading to a relative power loss, which is proportional to w², as given in Section 4.5. The sum of the relative areal and electrical losses reaches a minimum for an optimum value of w. This value is of the order of 1 cm, with slightly lower values for single-junction cells and slightly higher values for tandems.

The monolithic series connection can also be applied to the substrate configuration, as long as one is using an isolating substrate such as a polymer sheet. In this case, the laser-scribing process is, for technological reasons, a little more delicate to implement.

If the laser-scribing system is not properly adjusted, partial short circuits will be created on the module; this will lead to low values of shunt resistance R_shunt and to a low fill factor FF. Such a situation is definitely to be avoided.

The monolithic series connection of cells to form large-area modules is one of the key features of thin-film solar cell technologies, ensuring a higher degree of automation and an enhanced reliability, as compared to the stringing procedures used for interconnecting cells in wafer-based crystalline silicon technologies.

6.4 Module Encapsulation

Module encapsulation is indeed a key factor determining the long-term reliability of photovoltaic solar modules. It is an especially critical factor for thin-film modules. For all thin-film modules, one of the main goals of module encapsulation is to protect the semiconductor layers, as well as the transparent contact layers, against the influx of humidity. To this end, special polymer foils and dedicated sealing techniques have been developed. Many different combinations are used in commercial modules; most solutions are proprietary. Thus, at present we can only emphasize the importance of this step. For general information on this problem, the reader is referred to [120–122].

6.5 Module Performance

There are several factors influencing the performance of photovoltaic modules under actual application conditions:

The spectrum of sunlight tends to change over the course of the day (with a shift toward the red in the morning and evening). It also changes with climatic and geographical conditions (with a shift toward the blue in the presence of snow and water surfaces). In general, single-junction amorphous silicon modules perform better than crystalline silicon modules under blue shift of the incoming light and worse than the latter under red shift. Note that a blue shift is regularly observed for indoor conditions, especially for illumination with energy-saving light sources (e.g., fluorescent lamps). As for tandem and multijunction cells, they are more sensitive to changes in the incident spectrum than single-junction cells [123]. This means they must be carefully designed for the average spectrum prevailing in actual field conditions at the site of deployment.

As the intensity of the incoming light is varied from standard conditions (intensity of approximately 100 mW/cm²) to lower values, the conversion efficiency of the cell will be reduced. The extent of this reduction depends very much on the shunt resistance R_shunt within the cell, and this parameter in turn depends on the fabrication process. A low value of R_shunt (and a poor solar cell performance at low light intensities) is an indication of a faulty manufacturing process. Amorphous silicon solar cells with well-mastered fabrication processes can exhibit a very high value of R_shunt and thus also excellent performance for very low light intensities, making them particularly suitable for indoor applications [124]. On the other hand, present thin-film silicon solar cells and modules are not suited for higher light intensities—i.e., for applications with sunlight concentration.

As for the angle of incidence of the incoming light, it evidently also has, for optical reasons, an influence on the efficiency of the solar module. Its influence depends on the module design, especially on the first layers, sheets, or foils through which light enters the solar module. Thus, it is almost impossible to make any general statements. Nevertheless, as the design and the fabrication of solar modules become more standardized and as their application progresses, this effect should be taken into consideration and experimentally quantified (see also [125]).

One of the most important factors that result in reduced solar module efficiency is higher module operating temperature. Figure 35 shows how various solar cell technologies show a decrease in efficiency when their operating temperature is higher than 25°C (the temperature for which the module efficiency is specified in the data sheets—it corresponds to the standard test conditions, STC). One clearly notes from Figure 35 that wafer-based crystalline silicon modules suffer a relative loss in efficiency of about 20% when their operating temperature is increased from 25°C to 75°C (curves a, b, and c in Figure 35), whereas thin-film silicon modules suffer a relative loss of only half that value (curves d and e in Figure 35). In very many applications, the solar modules are typically operating at higher temperatures—between 60°C and 80°C. This is true for applications in tropical countries, and also for BIPV in temperate countries during the summer months. Thus, thin-film silicon solar modules have a “hidden” advantage over wafer-based crystalline silicon modules that does not transpire from the datasheets: their efficiency drop at higher temperatures is much less pronounced (see also [123,126]).

Because of the various effects described previously, the so-called relative performance of photovoltaic installations containing thin-film silicon solar modules is, in general, about 10% higher than that of installations with wafer-based crystalline silicon modules (see [1], Table 6.4). Note that the “relative performance” is given in kWh/kW_p and denotes the total amount of electrical energy in kWh delivered by the module during the period of one year, divided by the nominal installed power in kW_p.

6.6 Field Experience

The field experience is a very important aspect: only with actual data from modules exposed over several years to outdoor light and weather conditions can we judge whether a given solar technology is mature enough for widespread practical deployment as an energy source.

Amorphous silicon modules have now been in operation for almost 30 years. Whereas modules produced during the 1980s did have unexpected problems and failures, the next generation of modules that came on the market in the 1990s have shown remarkably stable and reliable performance. Chapter 7 in [1] gives several examples of photovoltaic installations, in Europe, in the USA and in Brazil, with performance data for 10 years or more; in all cases, the performance is indeed quite stable after an initial phase of 1 to 2 years with distinct degradation due to the Staebler–Wronski effect (see Section 2.1). In the long run, all photovoltaic modules show some drop in efficiency [127]—i.e., a yearly relative decrease Δη/η in efficiency, of about 0.5%, due partly to the imperfections of the encapsulation. For amorphous silicon modules, Δη/η may be slightly higher, possibly about 1%. Micromorph modules (i.e., modules with microcrystalline–amorphous silicon tandem cells) have not been around for very long so that we do not have long enough field experience with these types of modules; however, from what has so far been observed, one may reasonably expect their long-term performance to be slightly better than that of amorphous silicon modules. A further reason to have full confidence in “micromorph” modules is that the fabrication processes and individual layers used here are very similar (and partly identical) to those used for amorphous silicon modules.