Interaction between process technology and material quality during the processing of multicrystallin

Interaction between process technology and material quality during the processing of multicrystalline silicon solar cellsDietmar Borchert ÆMarkus RinioReceived:20July 2007/Accepted:27February 2008/Published online:20March 2008ÓSpringer Science+Business Media,LLC 2008Abstract Multicrystalline silicon is the most used material for the production of silicon solar cells.The quality of the as grown material depends on the quality of the feedstock and the crystallization process.Bulk impurities,crystal defects like dislocations and of course the grain boundaries deter-mine the material quality and thus the solar cell conversion efficiency.Therefore minority carrier lifetime measure-ments are often done to characterize the material quality.But the measured values are from limited use because it is known that the solar cell process itself can dramatically change the minority carrier lifetime and the solar cell efficiency.In order to obtain more detailed information of the behaviour of different defect types additionally high-resolution LBIC (light beam induced current)-measurements have been done.Since LBIC needs a pn-junction for photocurrent generation the LBIC technique has been combined with the a-Si/c-Si heterojunction cell process,which makes it possible to manufacture solar cells even from as cut wafers without changing the material quality.With this combination of measurement and preparation techniques it was possible to analyze the influence of the diffusion process and the firing process on the behaviour of the three different defect types:grain boundaries,dislocation networks and bulk impurities.1IntroductionMulticrystalline silicon is the most used material for the silicon solar cell production.The material quality is limited by crystal defects like dislocations and grain boundaries and bulk impurities.However over the years the mean values of the cell efficiencies for production lines manufacturing multicrys-talline silicon solar cells have increased from 13%to values above 15%.This is due to a better understanding of the interaction between the solar cell production process and the material quality.It has been found for example that a well-chosen diffusion process enhances the material quality due to a strong gettering effects for impurities.A similar improvement of the material quality has been observed after the introduction of silicon nitride as antire-flection material.A suitable SiN film enhances the material quality additionally.Very often lifetime measurements using the microwave photoconductivity (l -PCD)technique are used to monitor the change of the material quality during the solar cell process.Modern measurement setups deliver lifetime top-ograms of the whole wafer area so that the change of the carrier lifetime for different grains and also around the grain boundaries can be seen.However the spatial resolu-tion of such systems is too low to have a closer look on other defect types like dislocation networks.On the other hand high-resolution LBIC measurements can distinguish between different defect types and their influence on the material quality.The disadvantage of the LBIC method is that a pn-junction and contacts have to be formed on the wafer to be investigated before a measure-ment can be done.Since conventional processing changes the material quality one had no access to the as cut material using the LBIC method.This problem can be overcome by using a low temperature a-Si/c-Si heterojunction solar cell process.As a consequence this paper shows that high-resolution LBIC measurements in combination with the a-Si/c-Si heterojunction processing gives one the possibility to monitor the influence of the different defect types on theD.Borchert (&)ÁM.RinioFraunhofer ISE,Laboratory and Servicecenter Gelsenkirchen,Auf der Reihe 2,45884Gelsenkirchen,Germany e-mail:dietmar.borchert@ise.fraunhofer.de J Mater Sci:Mater Electron (2009)20:S487–S492DOI 10.1007/s10854-008-9687-0material quality from the starting material throughout the whole process chain.2Experimental2.1The standard processThe process sequence for the applied cell manufacturing process is shown in Fig.1.This process is very similar to that which is commonly used in industrial production lines.The first step is the removal of the saw damage using wet chemical etching.This can be done using either an alkaline solution in order to remove just the saw damage or an acidic solution which gives the additionally possibility to get a surface tex-turization.After the saw damage etching the pn-junction formation is done using a high temperature phosphorous dif-fusion process.The typical process temperature for the diffusion process is in the range from850°C to950°C.The optimum diffusion temperature depends on the material which is used for the production process.During the diffusion process a so called phosphorous glass is formed on the surface of the wafer.This is a silicon oxide film with a high content of phosphorous inside.This phosphorous glass is removed by wet chemical etching with hydrofluoric acid.The next step is the deposition of the antireflection coating on the front side of the solar cell. Today production lines are using silicon nitride(SiN)as antireflection coating.The deposition is done by plasma enhanced chemical vapour deposition(PECVD).After the PECVD deposition of the silicon nitride follows the screen-printing of the front and the back contact.Since the SiN is an insulating material the contacts on the front side of the cell have to befired through the SiN layer.This is done by a short heating of the screen-printed cell to a temperature in the range between850°C and900°C.The last process step is the edge isolation which is done by laser cutting.2.2Lifetime measurementsLifetimes have been measured using the micro-wave photo-conductivity(l-PCD)technique[1].The measurements were performed using an Amecon Janus2000mapping system.This system uses a microwave frequency of22GHz.The microwave probes an area of approximately1cm2.For the carrier generation a GaAs laser diode is used.The laser pulse duration is around100ns.Additionally white bias light can be used,the intensity of the light can be chosen from0%to100% of1sun(100mW/cm2).The measurement software deter-mines the lifetime byfitting the microwave curves using an exponential decay and a baseline asfit parameters.In these measurements special care has to be taken on the laser pulse and bias light power since the measured lifetime strongly depends on the injection conditions.Lifetime measurements only give information about the bulk lifetime when no surface recombination is present. Therefore the samples have to be passivated which has been done by depositing a suitable PECVD SiNfilm on both sides of the sample.2.3LBIC measurementsThe LBIC measurements were done using the measure-ment setup shown in Fig.2[2,3].The system is operating at a single wavelength of about830nm and with1.7l W light power.In order to measure an LBIC topogram the sample is moved under the laser beam using a xyz stage. The system has a spatial resolution of12.5l m.The short circuit current I sc from the solar cell is measured together with the reflected light power in vertical and side direction using the lock-in amplifiers1,2,and3.Addition-ally the incident light power P L is measured with lock-in amplifier4.From I SC,P L,and the reflection coefficient R the local internal quantum efficiency(IQE)is calculated.In the LBIC-map the IQE is plotted as a function of the illumination position.2.4The a-Si/c-Si heterojunction cell processSince the LBIC method measures the generated photocurrent there has to be a pn-junction inside the device.If one wants to analyze the as cut material with high-resolution LBIC a pn-junction has to be formed without changing the material quality.Since the conventional high temperature process technology changes the material quality the a-Si/c-Si het-erojunction process has been used.This process is a complete low-temperature process,the applied process temperatures are below250°C for all process steps.Therefore the quality of the material which is processed,is not altered.A schematic drawing of the manufactured cell structure is shown in Fig.3,the used process is described in[4]and has been successfully applied even to large areas[5].3Results and discussion3.1Material modification during the diffusion process Neighbouring wafers of block casted multicrystalline sili-con have been processed in the following way.After damage etching set1has been deposited with a SiNfilm on both sides which is well known for its excellent surface passivation properties.The second set has been diffused and after the diffusion the pn-junction has been removed by wet chemical etching.Then the wafers got the same SiN passivationfilm on both sides as set one.The change of the material quality during the pn-junc-tion diffusion has been investigated by comparing the results of the l-PCD lifetime measurements made on both sets(Fig.4).The as grown wafers show at one wafer side a region with very low lifetimes.The wafers were taken from a column which had been located at the side of a multi-crystalline block.Impurities from the crucible wall have diffused inside the material and thus the material quality in this region is very low.After an optimized diffusion pro-cess this region disappears completely and even in this region high carrier lifetimes are found.In many regions of the wafer area the lifetime is increased by the diffusion process but there are also regions were the diffusion process could not increase the lifetime. Overall we see an improved of the material quality and the improvement is mainly due to a gettering process which reduces the concentration of impurities in the bulk of the wafer.Ballif et al.[6]have shown that for material coming from the bottom of a crucible the improvement of the material quality is mainly due to Fe gettering.Moreover the Fe content in the silicon material changes with the position of the wafer in the ingot[7].During the diffusion process the iron concentration is strongly reduced and the phosphorous layer acts as a sink for this impurity.Gerrlings et al.[8]show how the Fe concentration can be measured very accurate by Fe–B pair dissociation.Nevertheless by this kind of experiment we get no information about the regions in which the material quality has not changed.3.2Hydrogen passivation of defectsA further improvement can be achieved by hydrogen pas-sivation of defect.This hydrogen passivation takes place when a suitable SiN layer is used as antireflection layer. During the contactfiring process hydrogen effuses from the SiNfilm and passivates defects in the multicrystalline wafer.Figure5shows the results of two measurements of the external quantum efficiency on complete solar cells with different types of SiN layers.(The quantum efficiency gives the number of electron generated per incoming photon for different wavelengths.)There are remarkable higher values for EQE found in the red spectral region.This indicates a clear material quality improvement and since the area under the curve is a measure for the cell current,one can see that the cell with the passivating SiN has a higher short circuit current den-sity and thus a higher conversion efficiency.While the EQE analysis is an integral method which gives no information about small regions of the wafer the LBIC method delivers a high spatial resolution.Therefore the influence of hydrogen passivation on grain boundaries has been investigated on a small wafer area(Fig.6).For this experiment four neighbouring wafers have been(n) a-Si:H emitter ITO layerdeposited with four different SiN antireflection coatings.The difference between these four films is that different gas mixtures have been used during the film deposition.The relative amount of hydrogen in the plasma is highest for sample 215and lowest for sample 218.From the LBIC measurements one can clearly see that the recombination activity of all grain boundaries is reduced when the amount of hydrogen in the plasma is increased.While for sample 218all grain boundaries show high recombination rates represented by low IQE values for sample 215the upper two-grain boundaries show only a very low reduction for the IQE.This effect can also be seen in Fig.7were an LBIC linescan has been made across the most active grain boundary in the samples.For a low hydrogen concentration in the plasma during the deposition we see a strong reduction of the IQE values around the grain boundary.The improvement at the grain boundary be the hydrogen passivation is remarkable.The IQE is increased from 75%to 85%.after diffusionas-grown Fig.4Influence of a diffusion process on the minority carrier lifetime of a multicrystalline silicon wafer.Left:Measurement for the as-grown wafer after surface passivation.Right:Measurement for the diffused wafer after wetchemical pn-junction removal and surface passivationSiH 4, NH 3H 2216215217LBIC (832 nm):218LinescanLinescanLinescanLinescan3.3Detailed analysis for various defect typesFor a detailed analysis for various defect types three neighbouring wafers have been processed in the following way (see also Fig.8).After saw damage etching of all wafers a heterojunction solar cell has been manufactured form wafer one.Wafer two has been diffused at 900°C and then a magnesiumfluoride MgF 2antireflection coating has been applied in order to decrease the surface reflection.The MgF 2film has been evaporated at room temperature so that no additional material modification must be taken into account by this process step.Wafer number three was completely processed as a solar cell using the described standard process.After the processing LBIC scans have been made in the same small area of the neighbouring wafers.The results are shown in Fig.9.We see the material quality change for three different regions on the wafer.In the upper left corner we can see the change of the IQE values for the bulk material since in this region no other defects can be found.Also we can see grain boundaries and the circle marks an area where a dislocation cluster is present.It is remarkable that in the as grown material the dis-location cluster and the grain boundaries show no enhanced recombination activity compared to the bulk material.However the overall IQE values are lower than for the other two samples.After the phosphorous diffusion at 900°C we see the following:The material improves but the grain boundaries and the dislocation cluster become active as recombination areas.There is a dramatic decrease in the corresponding IQE values.This means that the gettering effect during the phosphorous diffusion only leads to an improvement of the bulk material but on the other hand this positive effect is partly compensated by the activation of grain boundaries and dislocation clusters.The situation is improved a lot,when hydrogen passivation due to the SiN firing takes place.The dislocation cluster activity vanishes completely and here we find the same high IQE values than in the defect free bulk regions around.Also the recombination activity of the grain boundaries is strongly reduced but does not vanish completely.Moreover even the bulk material quality increases again.Therefore it becomes clear why the introduction of suitable SiN films as antireflection layers in the cell process has lead to a strong increase of the solar cell conversion efficiencies in the cell production lines.The investigation method described here can also be used for the characterization of the defect behaviour in other silicon materials as shown for EFG silicon in [9].4ConclusionsUsing lifetime measurements to monitor the material quality change for multicrystalline silicon in a solar cell process gives only limited information due to the low spatial resolution of the available measurement ing high-resolution LBIC measurements one can see the behaviour of different defect types in the various process ing additionally the a-Si/c-Si heterojunction pro-cess one can extend the LBIC measurements on the as grown material as the reference and starting point.Fur-thermore the low temperature heterojunction process gives one the possibility to process solar cells for the LBICD i s l o c a t i o n -c l u s t e rLBIC LBIC LBIC96 %81 %66 %quasi as-grown P-diffusion c o n t a c t f i n g e rP-diffusion & SiN:H firingFig.9LBIC scans on a small area of three different neighbouring wafers after different process steps.The circle marks the location of a dislocation clustermeasurements at any stage of the process chain,when structure are removed and the bare wafer is available. References1.M.Kunst,G.Beck,J.Appl.Phy.63,1093(1998)2.M.Rinio,H.J.Mo¨ller,M.Werner,Solid State Phenom.63–64,115(1998),(Diffusion and Defect Data/B)3.M.Rinio,S.Peters,M.Werner,werenz,H.J.Mo¨ller,SolidState Phenom.82–84,701(2002),(Diffusion and Defect Data/B) 4.D.Borchert,G.Grabosch,W.R.Fahrner,Sol Energ Mater Sol Cell49(Part III),53(1997)5.D.Borchert,A.Kenanoglu,A.Gronbach,M.Rinio,G.Willeke,T.Brammer,H.Stiebig,H.Windgassen,W.A.Nositschka,O.Voigt,in Proceedings of the19th European PV Solar Energy Conference, Paris,France,p.1174(2004)6.C.Ballif,S.Peters, D.Borchert, C.Ha¨ssler,J.Isenberg,R.Schindler,W.Warta,G.Willeke,in Proceedings of the17th European PV Solar Energy Conference,Munich,Germany, p.1100(2001)uer,M.Ghosh,werenz,S.Dauwe,in Proceedings ofthe21st European PV Solar Energy Conference,Dresden, Germany,p.1362(2006)8.L.J.Geerlings,G.Colette,D.Macdonald,in Proceedings of the21st European PV Solar Energy Conference,Dresden,Germany, p.692(2006)9.S.Scholz,F.Dreckschmidt,T.Kaden,H.J.Mo¨ller,A.Seidl,inProceedings of the21st European PV Solar Energy Conference, Dresden,Germany,p.354(2006)。