Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`Defect passivation of industrial multicrystalline
`solar cells based on PECVD silicon nitride
`
`F. Duerinckx*, J. Szlufcik
`
`IMEC, Kapeldreef 75, 3001 Leuven, Belgium
`
`Abstract
`
`Low surface recombination velocity and significant improvements in bulk quality are key
`issues for efficiency improvements of solar cells based on a large variety of multicrystalline
`silicon materials. It has been proven that PECVD silicon nitride layers provide excellent
`surface and bulk passivation and their deposition processes can be executed with a high
`throughput as required by the PV industry. The paper discusses the various deposition
`techniques of PECVD silicon nitride layers and also gives results on material and device
`properties characterisation. Furthermore the paper focuses on the benefits achieved from the
`passivation properties of PECVD SiNx layers on the multi-Si solar cells performance. This
`paper takes a closer look at the interaction between bulk passivation of multi-Si by PECVD
`SiNx and the alloying process when forming an Al-BSF layer. Experiments on state-of-the-art
`multicrystalline silicon solar cells have shown an enhanced passivation effect if the creation of
`the alloy and the sintering of a silicon nitride layer (to free hydrogen from its bonds) happen
`simultaneously. The enhanced passivation is very beneficial for multicrystalline silicon,
`especially if the defect density is high, but it poses processing problems when considering thin
`(o200 mm) cells. r 2002 Elsevier Science B.V. All rights reserved.
`
`Keywords: Solar cells; Multicrystalline silicon; Defects; Hydrogen passivation
`
`1. Introduction
`
`With the current stable growth of the PV shipments of around 30% the total world
`PV production will reach a GWp level already by the year 2005. Crystalline silicon
`solar cells constitute more than 85% of the world PV market with a tendency to
`increase the market share. To follow the production growth, cell manufacturers are
`
`*Corresponding author. Tel.: 32-16-281923; fax: +32-16-281501.
`E-mail address: filip.duerinckx@imec.be (F. Duerinckx).
`
`0927-0248/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved.
`PII: S 0 9 2 7 - 0 2 4 8 ( 0 1 ) 0 0 1 7 0 - 2
`
`

`

`232
`
`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`forced to shift from good quality mono-Si wafers to different types of lower quality
`polycrystalline substrates: cast multi-Si wafers, ribbons and thin polycrystalline-Si
`films. All these materials are characterised by large inhomogeneities caused by grain
`boundaries and intra-grain defects like dislocation, twins and others. These inherent
`defected regions act as recombination centres of minority carriers.
`Due to silicon supply problems, often a low quality feedstock is used, introducing
`various impurities in an uncontrolled way. Metallic impurities, oxygen and carbon
`behaving themselves as recombination centres,
`interact additionally with the
`crystallographic defects enhancing their recombination properties [1]. It is obvious
`that such imperfect multicrystalline silicon substrates are characterised by a very low
`lifetime of minority carriers. A wrongly designed cell processing sequence can
`increase impurity segregation at the defected regions and consequently increase their
`recombination activity. Therefore the thermal treatments must often be tuned to the
`properties of specific multi-Si materials. There are, however, several processing
`techniques which, when combined in a well thought-out solar cell processing
`sequence, can drastically improve the starting lifetime. Impurities like oxygen and
`metallic fast diffusers (Fe, Cu, Ni, Cr) can be efficiently extracted by external
`gettering occurring during the emitter diffusion from POCl3 source [2] and/or during
`the aluminium silicon alloying process commonly used for BSF formation.
`Defects,
`impurities and segregated impurities on extended defects can be
`passivated by hydrogen [3]. There are several techniques developed for hydrogen
`passivation of silicon bulk and surfaces. The best known are: annealing in forming
`gas ambient, H ion implantation, direct and remote plasma hydrogenation and
`deposition of H-containing silicon nitride layers (SiNx:H).
`
`2. Deposition techniques of the SiNx: H layers
`
`The hydrogen containing silicon nitride layers are becoming widely introduced in
`industrial crystalline silicon solar cell processes thanks to the unique possibility of
`combining in one processing step an antireflection coating deposition along with
`surface and bulk passivation [4,5]. Chemical, mechanical, optical and electrical
`properties of silicon nitride as well as the effectiveness of surface and bulk
`passivation strongly depend on the selected deposition technique. For solar cell
`application the most suitable are the deposition processes from the gas phase by
`means of chemical vapour deposition (CVD) using silane, ammonia and/or nitrogen
`as the reactant gases. The resulting silicon nitride layers are usually non-
`stoichiometric and contain up to 40 at% of hydrogen and are usually denoted in
`the literature as SiNx:H. The three basic CVD processes are: atmospheric pressure
`(APCVD),
`low pressure (LPCVD) and plasma enhanced (PECVD). APCVD
`involves reaction of silane and ammonia at atmospheric pressure in the temperature
`range of 700–9001C. LPCVD involves the reaction of dichlorosilane and ammonia at
`reduced pressure (B0.1 Torr) and temperatures around 7501C. PECVD uses plasma-
`enhanced reaction of silane and ammonia (or optionally nitrogen) at reduced
`pressure (B1 Torr) and temperatures below 5001C. The PECVD method is of
`
`

`

`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`233
`
`particular interests for solar cell application. The main advantages of the PECVD
`method over APCVD and LPCVD are:
`low processing temperature, higher
`deposition rate, the possibility to tune the refractive index over a wide range and
`much larger concentration of hydrogen in the deposited layers. The refractive index
`of PECVD SiNx layers can be varied in a range of 1.8 and 2.3 by means of the silane/
`ammonia ratio. A trade-off has to be found between minimisation of reflection and
`reduction of energy loss by absorption of short wavelength photons in the SiNx:H
`layer. The main reason for the strong interests in PECVD SiNx:H stem however not
`from the antireflection properties but mainly from the fact
`that
`there is
`overwhelmingly strong evidence for the very good surface and bulk passivation
`properties of the PECVD SiNx:H layers. The appropriate implementation of
`PECVD SiNx:H into a low-cost multicrystalline silicon solar process leads to
`comparable surface passivation and bulk passivation to respective dry oxide and
`prolonged hydrogenation passivation processes. Important pioneering work regard-
`ing SiNx-passivation and its application to silicon solar cells has been conducted
`during the past decades by J. Hanoka and co-workers [6].
`There are two basic different methods of PECVD SiNx:H deposition: direct and
`remote PECVD. In direct PECVD reactors all processing gasses (silane, ammonia,
`and nitrogen) are directly injected between the electrodes and excited by an
`electromagnetic field. The silicon substrates are placed within the plasma. Depending
`on the generator frequency, one can distinguish a ‘‘low frequency’’ (10–500 kHz), a
`‘‘high frequency’’ (13.56 MHz) or a ‘‘very high frequency’’ (30–100 MHz) direct
`plasma excitation. The excitation frequency influences the effectiveness of the surface
`and bulk passivation. Heavy ion bombardment present
`in a low frequency
`direct plasma produces a surface damage and therefore lower quality silicon
`surface passivation is achieved than in case of high or very high frequency direct
`plasma deposited SiNx:H layers [7]. On the other hand it is believed that some
`surface damage enhances in-diffusion of hydrogen into silicon and helps in bulk
`passivation [3].
`In the remote plasma CVD technique, the plasma excitation takes place outside
`the deposition chamber usually by means of microwaves, hollow cathode or arc jet.
`In most cases only ammonia or a nitrogen/hydrogen mixture are excited and directed
`onto a silicon substrate. Silane is injected downstream directly into the deposition
`chamber and is mainly dissociated by the atomic hydrogen from the plasma source.
`The main advantages of the RPECVD SiNx:H are the much higher deposition rate
`and the absence of silicon surface damage. Important for the solar cell working
`conditions is that RPCVD and high frequency direct plasma are reported to provide
`UV stable and excellent surface passivation [7]. The broad overview on silicon nitride
`surface passivation of crystalline silicon solar cells can be found in [7]. There is
`however still a question about the extent of bulk passivation from RPECVD
`deposited SiNx layers in comparison to Direct PECVD layers.
`A great interest from the PV industry in implementation of the PECVD SiNx:H
`layers in an industrial multi-Si solar cell process triggered the developments of
`dedicated high throughput deposition systems. There are several commercial systems
`being currently in the advanced development phase or already introduced on the
`
`

`

`234
`
`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`market. These systems are based either on direct plasma parallel plate reactor [8],
`microwave remote plasma [7,9] or very high throughput expanding thermal plasma
`[10].
`In the next chapters the properties of direct plasma PECVD silicon nitride layers
`combined with a low cost screen printing process is described in detail and analysed.
`
`3. Screen-printed multicrystalline cells with SiNx-passivation
`
`3.1. Process specifications
`
`The previous paragraphs have introduced the different methods available for
`deposition of SiNx-layers including direct PECVD, which is very well suited for
`application on silicon solar cells. At IMEC’s solar cell pilot line extensive process
`optimisations have been conducted to improve the efficiency of multicrystalline
`silicon solar cells using the inherent passivation properties of silicon nitride.
`Experiments have shown that it is very advantageous to apply a heat treatment after
`silicon nitride deposition to free hydrogen, which is present in the layer at high
`concentrations, from its bonds to silicon and nitrogen. Consequently hydrogen can
`diffuse out of the layer into the cell and bind itself to dangling bonds in the defected
`multicrystalline material (passivation). A comparison of two process sequences, with
`and without a thermal treatment of the silicon nitride layer, is shown in Fig. 1.
`The process with thermal treatment of the silicon nitride layer is referred to as a
`‘‘firing-through’’ process since the front cell contacts (Ag) are fired though the nitride
`layer to make contact with the underlying emitter during this high temperature step.
`The duration and the temperature profile for this firing step are chosen to achieve a
`good contact (a high Fill Factor (FF)) between the emitter and the front contact. On
`state-of-the-art multicrystalline material, improvements of 1% absolute have been
`
`Fig. 1. Process sequence with and without thermal treatment.
`
`

`

`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`235
`
`obtained with the firing-through process compared to the reference process without
`thermal treatment of silicon nitride. Internal Quantum Efficiency measurements [11]
`of these cells have shown a higher response for photons of all wavelengths indicating
`that the improvement due to the hydrogen passivation reaches far into the bulk of
`the cell. For the reference process (no thermal treatment), the bulk of the cells is not
`affected and therefore it can be stated that no bulk passivation occurs if the nitride is
`not thermally treated.
`The effect of the bulk passivation from the firing-through process strongly
`depends on the material quality. In general it can be stated that the more defected the
`material, the higher the improvement will be from the firing-through-SiNx process.
`This is confirmed by the results of Table 1 that compare firing-through-TiOx (no
`bulk passivation) and firing-through-SiNx processes on material of lower starting
`quality.
`This process, which is straightforward and can easily be implemented in existing
`production lines, can nevertheless effectively passivate the bulk of multicrystalline
`cells resulting in an important efficiency improvement. Although the prime purpose
`of the firing through process is to incorporate a bulk passivation in the process
`sequence, the newly developed process also brings other advantages specifically
`related to the use of silicon nitride. Some of these are discussed below.
`
`3.2. Advantages of firing-through SiNx process
`
`3.2.1. Emitter passivation
`The silicon nitride effectively passivates the emitter of silicon solar cells. This has
`been experimentally verified by making a comparison of cells with and without a dry
`oxide beneath the silicon nitride layer. The results of both groups are similar for
`industrial-type emitters on condition that the silicon nitride is thermally treated. If
`the nitride coating is not
`thermally treated, damage resulting from the ion
`bombardment during plasma deposition degrades the surface quality resulting in a
`J0e that is twice as high (1.3  1012 A/cm2 compared to 6.2  1013 A/cm2). With
`thermal treatment at temperatures above 7001C, this damage is annealed.
`
`3.2.2. Diffusion barrier
`Silicon nitride acts as a diffusion barrier during contact firing. This inherent
`property of nitrides is well known in the microelectronics industry. Likewise the
`silicon nitride used in our process acts as a diffusion barrier against unwanted
`impurities. As a consequence, metal spikes deep into the silicon can be avoided
`
`Table 1
`Comparison of firing-through-SiNx and TiOx processes on defected material
`
`Process
`
`Jsc (mA/cm2)
`
`Voc (mV)
`
`FF (%)
`
`Eff. (%)
`
`Firing through TiOx
`Firing through SiNx
`
`28.0
`30.6
`
`585
`600
`
`76.5
`76.9
`
`12.5
`14.1
`
`

`

`236
`
`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`reducing the risk of shunting the junction and the effective metal–silicon interface
`stays closer to the highly doped surface region. The consequence of this is a lower
`series resistance and a lower ideality factor around the maximum power point. It also
`means that shallow emitters with a sheet resistance around 60 O sq. can be contacted
`while maintaining a good fill factor.
`
`3.2.3. Al-BSF formation
`On the back surface of most modern screen-printed solar cells, aluminium is used
`to create a BSF. A thick aluminium layer (20 mm) is alloyed with silicon during the
`sintering process of the contacts. The resulting alloyed region at the back surface of
`the solar cell is around 5 mm deep with an acceptor concentration that is dependent
`on the firing temperature used. If high enough temperatures (>8001C) are used, the
`aluminium concentration in the BSF can rise to 5  1018 cm3 while the base doping
`level for material with a typical resistivity of 1 Ocm is only 1.5  1016 cm3. This
`high/low junction effectively repels minority carriers from the back surface. Since the
`front contacts have to be fired through the silicon nitride layer to make contact with
`the emitter, an elevated firing temperature is needed which is in the optimal range for
`a BSF-creation. This created the opportunity to merge these two separate processing
`steps in one single step.
`This in situ BSF-formation using the firing through process is a complementary
`advantage to the effect of hydrogen bulk passivation. While materials of lower
`quality (low base diffusion length) will be more dependent on bulk passivation,
`materials of higher quality (high base diffusion length) that are less dependent on
`bulk passivation, will benefit more from the created back-surface field. Other effects
`of Al-alloying and especially its synergy with SiNx-passivation will be further
`discussed in Section 4.
`
`3.2.4. Anti-reflection coating
`The silicon nitride coating used in the firing-through process has a double aim: it’s
`not only a source of hydrogen (passivation) but also acts as an excellent anti-
`reflection coating or ARC. This is an important advantage in comparison with other
`processes that incorporate bulk passivation (e.g. based on plasma hydrogenation). In
`that case, a separate step is still needed to deposit an ARC.
`The purpose of an ARC is to reduce the reflection of incoming light photons by
`favouring destructive interference of photons impinging and reflecting from the Air/
`SiNx/Si interface. This effect is mainly influenced by the thickness of the layer and its
`refractive index. Ideally the refractive index is around 2.0 for the Air/SiNx/Si
`structure and around 2.3 for the encapsulated condition (Air/glass-EVA/SiNx/Si).
`Since all cells are finally put into a module, the value of 2.3 is the one that’s most
`important. However this observation makes abstraction from another feature: light
`absorption. When the extinction coefficient of the deposited layer rises more light
`will be absorbed in the layer itself. These photons will therefore not contribute to the
`cell current. Characterisations based on spectro-ellipsometry have indicated that this
`effect becomes especially important for short-wavelength light (Fig. 2). This effect is
`directly influenced by the NH3/SiH4 ratio during deposition. If more silicon is
`
`

`

`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`237
`
`NH3/SiH4=7.3
`
`8.8
`
`12.9
`
`18.6
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0
`
`Extinction coefficient
`
`300
`
`400
`
`500
`
`600
`
`700
`
`800
`
`Wavelength (nm)
`
`Fig. 2. Extinction coefficient of silicon nitride as a function of NH3/SiH4 flow ratio.
`
`incorporated in the layer (lower NH3/SiH4 ratio), the extinction coefficient (and the
`absorption losses) as well as the refractive index will rise. At higher refractive indices,
`the reflection losses will be lower but this will be overcompensated by the absorption
`loss (higher extinction coefficient) in the layer itself. Experiments have shown that,
`for this reason, even for the encapsulated condition, it is favourable to limit the
`refractive index to values around 2.0.
`
`3.2.5. Passivation
`As was discussed earlier, a thermal treatment after the deposition of silicon nitride
`can release hydrogen from the Si–H and N–H bonds present in the nitride layer into
`the cell, passivating the silicon surface and the bulk. In our process sequence, the
`firing of the contacts performs this role. During the short time that the cells are at
`high temperature in an IR-furnace, the following actions take place simultaneously:
`First, the front contact is fired through the dielectric layer and makes contact with
`the underlying emitter. Secondly, hydrogen is released from its bonds in the nitride
`layer and diffuses into the cell (bulk passivation). Thirdly, the surface damage, which
`is a consequence of the nitride deposition in a direct PECVD-system, is annealed so
`that a good surface passivation is obtained. Laser beam induced current (LBIC)
`measurements of a typical multicrystalline cell made with the firing through silicon
`nitride process confirm the uniform passivation of the surface and the upper part of
`the bulk, independently of the grain orientation. For short wavelengths (Fig. 3(a))
`the grain boundaries are invisible, passivated by the in-situ hydrogenation. The
`LBIC-response is displayed relative to the maximum value with a minimum value
`around 90%. This indicates the extent of the passivation for the upper part of the cell
`since most defects have been made inactive by the passivation effect. Fig. 3(b) shows
`the response for long wavelengths (1060 nm). These photons are absorbed uniformly
`throughout the cell. Therefore, the grain boundaries are more visible since the lower
`part of the bulk is less passivated.
`
`

`

`238
`
`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`Fig. 3. LBIC measurements of cell made with a firing-through process (a) wavelength (540 nm), (b) long
`wavelength (1060 nm).
`
`Table 2
`Effective bulk diffusion length extracted by the spectral response analysis from 800 to 1000 nm
`
`SiNx present?
`
`Firing through SiNx?
`
`Effective diffusion length (mm)
`
`Yes
`Yes
`
`No
`Yes
`
`76.2
`119.1
`
`To quantify the effect of the passivation, a comparison experiment was made on
`neighbouring multicrystalline silicon solar cells between the firing-through process
`and the process without thermal treatment of the nitride layer. The cells were
`analysed using spectral response measurements and the effective diffusion length was
`extracted using Basore’s method [12]. The results are shown in Table 2. It can be
`concluded that the effective diffusion length has increased by 56% due to the short
`thermal treatment of the silicon nitride (without any process complication).
`
`3.3. Additional passivation features of SiNx
`
`Besides the features described above, several other issues determine the usefulness
`of PECVD-silicon nitride for a particular solar cell application. It is for instance well
`known that a high positive charge is incorporated in the first 20 nm of silicon nitride
`[13]. This could be confirmed by Capacitance–Voltage measurements [14] for the
`
`

`

`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`239
`
`Table 3
`Positive fixed charge and interface-trap-density of as-deposited and sintered silicon nitride layers
`
`Silicon nitride condition
`
`As-deposited
`Thermally treated
`
`Qf (cm2)
`
`3  1012
`1  1012
`
`Dit (cm2/eV)
`
`2  1011
`1  1011
`
`direct PECVD coatings deposited at IMEC’s pilot line. From the CV-measurements
`also the interface trap density could be determined. Both parameters are given in
`Table 3 for as deposited and thermally treated silicon nitride layers.
`It is interesting to notice that the interface trap density is reduced considerably by
`applying a thermal treatment. Earlier measurements confirmed the bulk passivation
`while these measurements indicate surface passivation. Therefore, sintering the
`silicon nitride layer can, at least partly, passivate damage that is a result from ion
`bombardment during the direct deposition process. Moreover, the positive charge is
`high enough to induce an inversion layer on moderately doped (around 1 Ocm)
`p-type Si. Therefore silicon nitride layers should also be very suited for rear surface
`passivation of silicon solar cells. This could indeed be confirmed by measuring the
`effective surface recombination velocity (Seff ) of Float-zone p-silicon wafers
`passivated with PECVD-SiNx. Values as low as 30 cm/s were obtained. This is an
`important observation for next generation solar cells.
`
`3.4. Results of an optimised SiNx-process on large area multicrystalline silicon solar
`cells
`
`At IMEC, much effort has been put in upgrading the firing-through process and
`bringing it as close as possible to an industrial process ready to be implemented in
`production lines. While the main features of the firing-through process are similar to
`what was discussed in the previous paragraphs, each process step has been optimised
`with regard to all other process steps. Some of the most important features of the
`final process are:
`
`* Isotropic texturisation [15]: a new chemical isotropic texturisation technique has
`been developed that replaces the standard caustic saw damage removal. This one
`new step removes the saw damage and at the same time randomly textures the
`wafers to decrease the reflection of light from the front surface.
`* POCl3-diffusion: a shallow emitter is used with a sheet resistance around 60 c sq.
`The phosphorus profile has been optimised for a maximal collection of high-
`energy photons.
`* Parasitic junction removal: A plasma etching step has been optimised to remove
`the parasitic junction at the edge of the cells.
`* SiNx-deposition: As described above, the silicon nitride layer has been tailored to
`reduce the reflection and absorption and to increase the uniformity of the
`deposition.
`
`

`

`240
`
`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`Table 4
`Results of large area cells (12.5  12.5 cm2) fabricated with the firing-through-SiNx process
`
`Material
`
`Emitter
`
`ARC
`
`Jsc (mA/cm2)
`
`Voc (mV)
`
`FF (%)
`
`Eff. (%)
`
`Polix
`Eurosil
`Baysix
`Eurosil
`Polix
`
`Homogeneous
`Homogeneous
`Selective
`Selective
`Selective
`
`SiNx
`SiNx
`SiNx+MgF2
`SiNx+MgF2
`SiNx+MgF2
`
`33.5
`33.5
`35.1
`34.9
`35.3
`
`621
`614
`621
`623
`635
`
`78.1
`77.9
`77.0
`77.0
`76.0
`
`16.3
`16.0
`16.8
`16.7
`17.0
`
`* Screen-printed metallisation: This step is the most critical one. The composition
`of the front side paste (Ag), the printing conditions (width, height and number of
`the fingers, width of the busbars, the printing speed, pressure and snap-off
`distance) and the firing conditions (temperature profile and speed) all have to be
`optimised in order to achieve a low series resistance in the lines as well as a low
`contact resistance with the emitter. This optimisation has been done taking into
`account the properties of the SiNx layer (the front contact has to be fired through
`this layer) and the underlying emitter.
`
`With this process, efficiencies between 15.3% and 15.7% have been achieved on
`large wafer batches (50–100 wafers, size: 12.5  12.5 cm2, resistivity: 0.5–1 Ocm) from
`several different manufacturers. The series resistance of the busbar limits the
`efficiency of the cells. By tabbing the busbars, as is done for encapsulation of the cells
`into a module, the Fill Factor of the cells rises strongly and efficiencies above 16%
`are obtained (Table 4). These are among the highest ever measured for large cells
`fabricated with a straightforward process scheme.
`Experiments to increase the cell efficiency further have focussed on implementing a
`selective emitter structure. In this more complicated process, different emitters are
`foreseen for the contact area (deep emitter with a high P-concentration to achieve a
`good contact) and for the active area (shallow emitter for a high collection of carriers
`created by short wavelength photons). Results of these cells with a double anti-
`reflection coating are also shown in Table 4.
`
`4. Synergy between Aluminium and SiNx
`
`4.1. Screen-printed Al-BSF as back surface passivation
`
`Screen printed aluminium has found a widespread use for silicon solar cells, not
`only in laboratory but also in production lines. It is the easiest way to create an
`effective BSF without unnecessarily complicating the process sequence. The Al-layer
`is normally printed over the whole back surface except for two strips of Ag/Al for
`soldering purposes. The printed layers are subsequently dried and fired. In our
`‘‘firing-through’’ process this happens together with the front contact firing (co-
`firing). During this firing at high temperature, the aluminium melts and forms an
`
`

`

`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`241
`
`alloy with silicon according to their binary phase diagram [16]. This typically results
`in a 5 mm deep, highly doped p+ layer with an acceptor concentration around
`5  1018 cm3. The effective surface recombination velocity as experienced by the
`base minority carriers is considerably reduced due to the strong electric field present
`at this high/low (p/p+) junction. Values below 1000 cm/s are normally obtained.
`Because of the simplicity and the effectiveness of the process, Al-BSF became a
`standard process years ago for the back surface passivation of screen-printed silicon
`solar cells. When PECVD-SiNx was introduced in solar cell processes, the screen-
`printed aluminium back contact was already an established technology. It is more
`recently that researchers became aware of
`the dependence for a good bulk
`passivation on the presence of both features simultaneously, silicon nitride as well
`as aluminium [17,18]. This is discussed in the next paragraph.
`
`4.2. The Al/SiNx synergetic effect
`
`To investigate the possible interaction between silicon nitride passivation and
`aluminium, the following experiment was carried out: multicrystalline Baysix wafers
`(size 10  10 cm2) were processed into solar cells according to different process
`sequences (Table 5). All cells were identically saw damage etched followed by a
`POCl3-emitter diffusion. The rest of the process differed for the four groups that
`were made:
`
`* Group 1: Standard process involving the co-firing of the PECVD-silicon nitride
`layer, the front contacts (through the silicon nitride) and the Al-layer. In this
`process, the creation of the BSF and the release of hydrogen from the nitride layer
`happen simultaneously.
`* Group 2: No BSF is created. The nitride is fired together with the front (Ag) and
`back contacts (Ag/Al grid: only a very small amount of Al is present in this paste
`in order to have a good contact with p-Si).
`* Group 3: The Al is fired first resulting in a BSF. The eutectic layer is removed by
`hot HCl, followed by deposition of SiNx, the printing of front (Ag) and back (Ag/
`Al grid) contacts and firing. The resulting cell structure is exactly the same as for
`
`Table 5
`I–V results of different sequences involving SiNx passivation and Al-BSF
`
`Group
`
`Process sequence
`
`Jsc
`(mA/cm2)
`
`Voc
`(mV)
`
`1
`2
`3
`
`4
`
`SiNx (front)FAg (front)Ffull Al (back)Ffiring
`SiNx (front)FAg (front)FAg/Al grid (back)Ffiring
`Al (back)FfiringFAl removalFSiNx (front)FAg (front)FAg/Al
`grid (back)Ffiring
`Al (back)FfiringFAl removalFAg (front)FAg/Al grid
`(back)FfiringFSiNx (front)
`
`31.1
`29.8
`29.9
`
`27.8
`
`614
`604
`605
`
`589
`
`

`

`242
`
`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`the first group. The only difference is that the firing of the nitride and the creation
`of the Al-BSF do not happen simultaneously.
`* Group 4: Similar to group 3, only the nitride is deposited on top of the front
`contacts (no annealing of the nitride layer).
`
`Group 1 of the experiment is processed according to the standard screen printing
`process. In order to check the importance of Al as back contact, the second group
`was made using an Ag/Al grid instead of a full Al coverage. At first sight, a
`straightforward explanation for the rather large difference in cell results between the
`two groups (Table 5) would be the influence of the BSF, which is present for group 1
`but not for group 2. However, this explanation was doubtful since the diffusion
`length is smaller than the wafer thickness of 330 mm. Therefore the BSF should not
`have a very large effect. To test this assumption the third group was processed: In
`this case, the BSF was made first followed by removal of the Al-eutectic layer,
`deposition of the nitride layer, screen printed metallisation of the contacts (Ag/Al
`grid for the back) and firing. The same results were obtained as for group 2 (no BSF)
`which indicates indeed that the BSF has no influence on the final cell results. It seems
`that the creation of the BSF together with the annealing of the nitride has a strong
`beneficial effect. This is seen in the IQE-curves of Fig. 4 as an improvement in the
`long wavelength region and indicates an extended hydrogenation effect (since the
`BSF does not play a role) deep into the bulk of the cell.
`Group 4 shows however that, if the nitride layer is not fired at all, the results are
`still a lot lower. This indicates that the passivation effect from firing the nitride layer
`is also present when no Si–Al alloy is simultaneously created on the back surface
`(group 2). The effect can however be enhanced by the simultaneous creation of an
`alloy as can be seen from group 1. It has been proposed [18] that this enhanced effect
`is due to the creation of vacancies during the alloying process, which can facilitate
`the in-diffusion of hydrogen into the bulk.
`If the enhanced passivation effect is due to the creation of vacancies during the
`alloying then the effect should scale with the thickness of the p+ layer that is
`
`Group 1: standard SiNx/Al-process
`
`Group 2: SiNx + Ag/Al
`
`Group 3: Al-BSF (eutectic removed)
`+ sintering of Ag/Al and SiNx
`
`Group 4: Al-BSF (eutectic removed)
`+ Ag/Al (SiNx not annealed)
`
`100.0%
`
`80.0%
`
`60.0%
`
`40.0%
`
`20.0%
`
`0.0%
`
`IQE
`
`300
`
`500
`
`900
`700
`Wavelength (nm)
`
`1100
`
`Fig. 4. IQE measurements of different sequences involving SiNx passivation and Al-BSF.
`
`

`

`F. Duerinckx, J. Szlufcik / Solar Energy Materials & Solar Cells 72 (2002) 231–246
`
`243
`
`Table 6
`I–V results of cells with back contact paste (full coverage) with different Al-concentrations
`
`Back side paste
`
`Jsc (mA/cm2)
`
`Voc (mV)
`
`FF (%)
`
`Eff. (%)
`
`L (mm)
`
`Seff (cm/s)
`
`Al
`75% Al, 25% AgAl
`50% Al, 50% AgAl
`25% Al, 75% AgAl
`AgAl
`
`32.1
`31.6
`31.4
`30.5
`30.3
`
`618
`614
`612
`606
`603
`
`76.8
`76.4
`76.5
`76.1
`74.2
`
`15.2
`14.8
`14.7
`14.1
`13.6
`
`261
`224
`172
`115
`(107)
`
`1000
`1315
`2297
`34,000
`F
`
`100% Al
`
`75% Al - 25% AgAl
`
`50% Al - 50% AgAl
`
`