ewt solar cell 1993 origin

8/2/2019 EWT Solar Cell 1993 Origin

1/6

EMITTER WRAP-THROUGH SOLAR CELLJames M. Gee, W. Kent Schubert, and Paul A. BasoreSandia NationalLaboratories

Albuquerque,NM 87185ABSTRACT

We present a new cell concept (Emitter Wrap-Through orEWT) for a back-contact cell. The cell has laser-drilled vias towrap the emitter on the front surfwe to contacts on the backsurface and uses a potentially low-cost process sequence.Modeling calculations show that efficiencies of 18 and 2 1% arepossible with large-area solar-grade multi- and mono-crystalline siliconEWT cells, respectively.INTRODUCTION

Commercial one-sun silicon solar cells have the photocanier-collection junction (emitter) and a current-collection grid on thefront surface. A front-surface emitter is necessary for goodinternal collection efficiency; solar-grade silicon materialstypically have diffision lengths less than the device width. Thefront-surface grid is necessary for low series resistance. Front-gridded cells have the well-known grid optimization tradeoff -increasing the grid coverage reduces losses due to seriesresistance but increases losses due to grid obscuration. Otherproblems associated with the front-grid cell design include thefollowing: increased complexity for cell manufacturing due tothe requirement for fine-line grid definition, poorly optimizedemitter due to requirements of low-cost metallizations (i.e.,high surface concentration), and reduced packing density ofcells in the module and increased complexity of moduleassembly due to front-to-back cell stringing [l]. Commercialone-sun silicon solar cells use screen-printed silver lines thathave relatively poor aspect ratios (grid height divided bywidth), conductivity, and contact resistance. Wenhamestimates that losses due to grid obscuration, grid resistance,and poor emitter characteristics reduce the cell performance ofcommercial screen-printed solar cells by nearly 30% relative tohigh-efficiency (i.e.,> 20%) laboratory cells [I].Placing both the n-type and the p-type current-collection gridson the back surface of the solar cell (i.e., "back-contact") hasseveral advantages compared to the front-grid cell. The mostobvious advantage of a back-contact cell is higher performancefrom no grid obscuration and low grid resistance; the onlyconstraint on the grid design is the technological limits of themetallization technology. Another important advantage ofback-contact cells is simplified module assembly and betterpacking factor of cells in the module; front-to-back stringing offront-contact cells requires additional space between cells forthe tabs, is difficult to automate, and has a non-uniform surfacefor encapsulation. Back-contact cells also present a more

aesthetic uniform appearance, which is important for someconsumer applications (e.g., photovoltaic sunroofs forautomobiles). Finally, backantact ed cells allow for thepossibility of a high-voltage cell through monolithic seriesconnection of cells on a single substrate.There are two approaches for placing both contacts on the backsurface of a solar cell. In the first approach, photocanier-collection junctions and grids for both polarities are located onthe back surface ("back-junction" cell). Swanson e t aliademonstrated an efficiency of 21% with a one-sun, 35-cm2,back-junction silicon solar cell [2]. The photogeneratedcarriers must diffise to the back surface for collection in back-junction cells; hence, these cells require materials withdiffusion lengths much longer than the device width for goodphotocarrier-collection efficiency. The back-junction cells aretherefore not usell with many solar-grade silicon materialsthat have diffision lengths shorter than the device width.The second approach for placing both contacts on the backsurface of the solar cell keeps the collection junction on thefront surface, which is more desirable for one-sun solar cells.This approach requires holes ("vias") through the substrate forthe current-collection grid on the back surface to contact thecollection junction on the front surface. Hall and Soltysdescribed such a cell in 1980 that used alignedphotolithography and wet-chemical etching to form the vias [3].K. Fujui, K. Shirasawa, and H. Watanabe recently described aback-contact cell with a front-surface collection junction thatthey named the Through-Hole Back-Contact Solar Cell [4].They did not provide a description of the process sequence.They reported results of 2-dimensional modeling that indicatedefficiencies a proaching 28% were possible, and reported a Jof 40 mA/cm!ind a Voc of 543 mV for a 4-cm2 test device.In this paper, we describe a new approach for a back-contactcell with a front collection junction. We refer to the new cellconcept as the Emitter Wrap-Through cell since the emitter iswapped through vias to the back surface for contacting. Animportant feature of the EWT cell is the process sequence; theEWT-cell process is based on the buried-contact cell process,which has been shown by several authors to be potentially costcompetitive with conventional screen-printed cells and is inpilot production at several locations [ 5 ] .This work was performed by Sandia National Laboratories forthe United States Department of Energy under contract DE-ACO4-76DPOO789.

2650-7803-1220-1/93 $3.0 0 0 1993 IEEE


2/6

CELL AND PROCESSDESCRIPTIONFigure 1 presents a schematic of the EWT cell. For thepurposes of this discussion, we assume that the substrate isptype silicon. The front surfacehas only a phosphorus diffusion(emitter) and an anti-refection coating. The front surface mayalso be textured for low reflectance and light trapping. Theback surface has two sets of interdigitated grids. The two setsof grids are the ptype and the n-type contacts. The n-type gridis a laser-scribed groove that has been heavily diffised withphosphorus and filled with metal. The front-surface emitter iscontacted from the back surface through laser-drilled vias [6 ] .The vias are aligned with the n-type grooves on the backsurface and are also heavily phosphorus d i W d and metallizedat the same time as the n-type grooves. The heavy groovediffision reduces contact resistance and contact recombination.As will be discussed later, there are several options for theinterdigitatedptype grid and substrate contact.Figure 2 presents a relatively simple variation of the EWT cellwith a double-junction emitter, i.e., there is a photocarrier-collection unction on both surfaces. A double-junction emittercell collects photogeneratedcarriers from both surfaces, whichimproves the collection efftciency for materials with difhsionlengths shorter than the device width [7]. The conversionefficiency of a double-junction cell should also be quite goodfor rear-surface illumination, which is imprtant for moduleapplications that use bifacial illumination of the cell [SI.The process sequence for an EWT cell is similar to a buried-contact cell. First, phosphorus is d i W d o around 100 W Oon the front surface for an EWT cell or on both surfaces for adouble-junction EWT cell. Next, a dielectric layer is eithergrown (Si02) or deposited (Si3N4) on both surfaces. The n-type grooves and vias are next scribed and drilled with a l w ,ideally, the grooves and vias are formed in a single step toensure good alignment. The grooves and vias are subsequentlyetched, heavily diffised with phosphorus, and filled with metalby a selective process. For the selective metallization, buried-contact cells typicallyuse electroless nickel sintered in forming

Top Surface

gas for low contact resistance followed by electroless copper orelectroplated silver for good conductance. The dielectric layerserves several hct ions in the buried-contact cell sequence: anetch stopduring the groove etching, a diffusion mask during thegroove diffusion, a platingmask during the metallization step,and the antireflection coating and surface passivation in thefinished cell.Several options exist for f&g the interdigitated grid on theback o contact the ptype substrate. One option is to perform apattemed-aluminum alloy, possibly using an aligned screen-printed aluminum paste fired through the dielectric layer. Asecond option is to use a laser to scribe a second set of groovesand diffuse the grooves with boron. Note that the ptypealuminum alloy or boron diffusion could be performed eitherbefore or after the phosphorus groove diffusion. The p-typegrid is plated with metal h r ow resistance at the same time asthe n-typegrooves and vias.The high perfomance and potential cost effectiveness of theburied-contact cell arises h m he following features: separateemitter and contact difiisions for high emitter efftciency, selfalignment of the metallization with the groove diffusion, highaspect ratio for the gridline due to the groove, and to theseveral functions served by the dielectric layer [SI. The onlydifference in equipment between the EWT and buried-contactcells is the laser system; the laser system must be capable ofdrilling vias aligned to the scribed grooves with a minimumthroughput of one 100-cm2 sample per minute. For a typicalvia spacing of 1 mm and 10OOO vias per sample, the desiredthroughput corresponds to a drilling and sample (or laser beam)translation time less than 6 ms per via. We have identifiedcommercial laser systems that are capable of drilling holes insilicon in a single pulse with a repetition rate greater than 1kHz. The grooves can be scribed at the same time as the viadrilling with a second laser collinear with the drilling laser orthe grooves can be mechanically scribed in a separate step. Inany case,commercial laser systems are available that meet ourminimum throughput requirements.

Holes

Bottom Surface (-GroovesFigure 1. Schematic of Emitter WrapThrough Cell.

266


3/6

Front Surface

Emitter Contac tDiffusion, n + +Bulk, p-typect Diffusion, p + +

Back Emitter, n+....................................................................... .......................................................................................Bulk Contact ViaHoleFigure 2. Schematic (side view) of Double-Junction Emitter Wrap-Through Cell

CELL DESIGNWe first develop a model for the series resistance of an EWTcell; the series resistance model is required for cell design andfor device models to estimate EWT cell perfoxmance.Important resistance terms include emitter resistance (includingspreading resistance as the current converges into the vias), viaresistance (transport of current through the vias to the backsurface), grid resistance, and base resistance. Important designparameters for series resistance include emitter sheetresistance, via diameter, groove and via spacing, and groovecross-sectional area. The important optimization parameter isthe via and groove spacing; reducing the spacing reduces theseries resistance but increases the time required for laserscribing and drilling.

We developed a model for the series resistance by dividing thecell geometry into regions with onedimensional current flows.For calculation of the emitter spreading resistance, the modeldivides the emitter into circular "unit cells" with an areaequivalent to the square collection area of each via [3].Similarly, the model separates current flow in the base intoregions of linear and cylindrical flow [9]. The grid resistancedepends upon the geometry of the busbars. Figure 3 presentstwo possible configurations for the busbars with interdigitatedback contacts. Adding more busbars reduces the gridresistance at the expenseof increased complexity in the moduleassembly. Table 1 presents expressions for the seriesresistance components of an EWT cell.

...I........I..........................................................a

...........-.... I.......................................................

................... I................... 1.1.- .............I......I_..

-U_.-......-

-..........-.. .....l

---.-.-..1 . 1 1 - -.-....-...._.--................

Figure 3. Two possible busbar arrangements for an interdigitated back-contact cell. (A) has a single busbar for eachpolarityon each end of the cell. Module assembly using this cell would be easy. (B) has an additional set of busbars in thecenter of the cell, which reduces the grid resistance by a factor of fourcompared to (A) at the expense of more complexmodule assembly.

267


4/6

PgSL2Rg=- 3A&?

we calculated the series resistance of a 100-2 EWT cell as afunction of via diameter, vi a and p v e pacing, and emittersheet resistance. We used a cross-sectional area for the grid of10OOO pm2,a resistivity of 2 flan fix the metal, a busbar ateach end of the cell, and a base resistivity of 1 f2cm. Figure 4presents an example of these calculations. The two largestcomponents of series resistance aredue to the grid and theemitter. The emitter resistance is relatively high due to currentcrowdingas he current converges on the via hole, so that thevia diameter is an mportant parameter in the cell design. Notethat the grid resistance would be negligible if we eitherassumed a larger cross-sectional area for the grid or usedadditional busbars in the design. A range of values for spacing,sheet resistance, and via diameter yield a series resistance ofless thanour target of 0.3 2cm2.

Table 1. Expressions for emitter, grid, and base resistancecomponents&,Rg, and Rb)in ncm2 of an EWT cell. s isthe groove and via spacing, L is the device length, pe is theemitter sheet resistance, pg is the grid resistivity,pb is the baseresistivity,Ag is the cross-sectional area of the gridline, W isthe device wdth, rc is the radius of the via divided by theradius of the unit cell, and r b is the radius of the p-typecontact. The terms in W e t s for R, and Rb represent theeffect of current crowdmg in the emitter and 2dimensionalcurrent flow in the base, respectively.

We used a numerical onedimensional semiconductor devicemodeling code (PC-1D) with a series resistance of 0.25 ncm2to estimate the cell ~erfomulnce f a loo-cm2 EWTcell [IO].We used material parameters (e.g., base doping, lifetime,width, and surface type) appropriate for highquality solar-grade Cz and multicrystalline silicon. Table 2 presents resultsof our calculations. The calculations show that efficiencies inacess of 21 and 18% are possible with solar-grade Cz andmulticrystalline silicon EWT cells, respectively. The double-junctioncellprovides a performance enhancement of nearly 1%absolute for both materials compared to the front-junction cell.Note that these Calculations overestimate the open-circuitvoltage slightly because the onedimensional device simulationignored the contribution to recombination current of thediffused grooves on he rear surface.

150Sheet Rholohmslsq)

Spacing (mmlFigure 4. Calculated series resistance of EWT cell with a via diameter of 100pm as a function of via

and groove spacing and of emitter sheet resistance.

268


5/6

I 20.5 I 21.6 1 17.3 I 18.0zh 30 15

Width (mm) 300 250I Ph(Rcm) 0.8 1.5L . , I ITable 2. Results of EWT cell simulations using a seriesresistance of 0.25 Clan2 and material parameters appropriatefor solar-grade mono- and multicrystalline silicon. We used atextured surface with a fixed reflectance of 2%for simulationsof the Cz cell and a planar surface with the modeled reflectanceof a single-layer antireflection coating for simulations of themulticrystalline cell.

PROCESS DEVELOPMENTMost of the processes required for fabrication of an EWT cell(e.g., multiple uses of a dielectric, groove formation, groovediffision, metalization, etc.) have already been demonstratedby other groups for fabrication of buried-contact silicon solarcells. The new processes required for fabrication of an EWTcell include the following: laser drilling of the vias aligned to alaser-scribed groove, diffusion of phosphorus through the viaswith low electrical resistance, and incorporation of borondiffusion or patterned aluminum alloy into the processsequence. Note that the EWT cell requires at least onealignment in the process sequence for the interdigitated n-typeand p-type grids.We have a pulsed (Q-switch) Nd:YAG (1.064pm) laser-scribe.system, which we have upgraded with a four-axis (x, y, z, and0) sample stage and a computer-based control system [I 11. Thecontrol system uses files from a CADD program for easyadjustment of patterns. Our laser system drills a via in lessthan 100 ms using a pulse energy of 0.5 d,pulse width of25 ns, and a pulse rate of 1 kHZ. (Our laser has a lowthroughput for hole drilling because it does not have theappropriate pulse width and energy for drilling holes in a singlepulse.) The grooves are scribed in one pass and the vias aredrilled in a subsequent pass in immediate succession, so thatthe groove and vias are always aligned.We also modified our system to accomodate alignment ofdifferent patterns. We found that mechanical alignment wasnot sufficient for alignment of the n-type and the p-type

interdigitated grooves. The problem with mechanicalalignment was rotational accuracy; the two patterns must berotationally accurate to less than 0.6" for spacing of 1 mm andcell length of 10 cm. However, we found that we couldmanually align patterns with the aid of a HeNe alignment laserbeam that is collinear with the laser-scribe beam.We successllly demonstrated diffusion of phosphorus usingPH3 in the vias. We conf ii ed , by measuring the electricalresistance of the diffused vias, that the walls of the vias haveapproximately the same sheet resistance as the diffision on ahorizontal surface. This result means that the resistance of thevias, even without plating of metal into the vias, is negligiblefor an EWT cell. We have also found that metal (electrolessnickel and electroplated silver) readily plates into and mostlyfills the vias, which further reduces the via resistance.A boron diffusion or an aluminum alloy is required for lowcontact resistance of the sintered nickel to the p-type basebecause solar-grade silicon typically has a resistivity of onlyabout 1 ncm. A boron diffusion or aluminum-alloyed junctionis also required for the double-junction EW T cell to counterdope the phosphorus emitter diffusion (Figure 2). We havedemonstrated that evaporated aluminum can be alloyed throughsilicon nitride in a tube h a c e at 1000C in our laboratory.However, we are concentrating on developing an EWT processsequence with a boron diffusion and aligned interdigitatedgroove because we do not presently have a means in ourlaboratory to pattern evaporated aluminum over a grooved backsurface.Integration of boron diffusion into the process requiresdemonstration of several process-related items: the borondiffusion must be compatible with the silicon nitride, the borondiffusion in the grooves must be protected from the phosphorusgroove diffusion, and the boron diffusion must be compatiblewith the metalkt ion. (We use silicon nitride rather thansilicon dioxide for the dielectric because silicon nitride is abetter anti-reflection coating (higher refractive index), does notrequire a long fiunace oxidation, is a better diffusion barrier,and is more easily integrated into the process due tocompatibility with chemicals used for deglazing the diffusionglasses (HF) and for etching the grooves (KOH).) We havedemonstrated all the boron-diffusion items individually. Borondiffusion using diborane gas (B2%) does not interact stronglywith silicon nitride, although there appears to be an interactionof the boron and phosphorus diffusions together with the siliconnitride. We conf ii ed that a 200-nm CV D oxide protects theboron diffusion from the phosphorus groove diffusion. We alsodemonstrated plating to boron-diffused grooves, although wehave not yet measured the contact resistance. Table 3 presentsour process sequence using a boron diffusion. We expect tohave the full process integrated for fabrication of an EWT cellby the end of 1993.

269


6/6

Table 3. Process sequence for EWT cell.Phosphorusemitterd i h i o n (100 R/sq).Si3N4 deposition by LPCVD (100 nm).

Grooveptype grid.KOH etch grooves.

B2Hg boron diffusion (approximately 50 Wsq).CVD oxide (200 nm) on back surfice.

Groove and drill n-type grid.KOH etch grooves and vias.

PH3 phosphorus dimion to (approximately 10Nsq).Deglaze grooves and tune nitride to ARC.

Electroless nickel.sinter in forminggas, 500C.

Electroplate silver.Forming gas anneal, 4OOOC.

CONCLUSIONSWe described a new cell concept (EWT) that has theperformance advantages of a back-contact cell, that can usematerials with short d i h i o n lengths, and that uses apotentially cost-effective process sequence. We havedemonstrated all the individual processes required forfabrication of an EWT cell. From ou r modeling, we believethat the EWT cell is capable of achieving efficiencies of 18 and21% on large-area (100 an2) multi- and monocIystalline solar-grade silicon substrates, respectively.

ACKNOWLEDGMENTS

REFERENCESPVSC refers to Proceedings of the ZEEE PhotovoltaicSpecialists Conference. EPVSEC refers to Proceedings of theEuropean Communities PhotovoltaicSolar Energv Conference.1. S . R. Wenham and J. M. Gee, "Advanced Process ing ofSilicon Solar Cells," Tutorial at 2 3 dZEEEPVSC, 1993.2. R. M. Swanson et alia, "High-Efficiency Silicon SolarCells,"llth EPVSEC, 1992,pg. 35.3. R. N.Hall and T. J. Soltys, "Polka-Dot Solar Cell,"llrhPVSC, 1980, pg. 550.4. K. Fujui, K. Shirasawa, and H. Watanabe, "Through-HoleBack-Contact Solar Cell," Tech. Digest Fourth "Sunshine"Workshop on CrystallineSolarCells, 1992, pg. 25.5. M. A. Green et alia, "Present Status of Buried-Contact SolarCells," 2 F d PVSC, 991, pg. 46.6. Jacob Murad (Mobil SolarEnergy Corp.) first suggested tous the useof laser drilling for back-contact solar cells. See alsoS.R.Wenham and M.A. Green,U.S. Patent 4,626,613.7. M. A. Green, Hieh Efficiencv Silicon Solar Cells. TransTech Publications, VT, 1987, and T. Warabisako et alia, 11thEPVSEC.1992, pg. 172.8. A. Cuevas, A. Luque, and J. M. uiz, "A N+PP+ Double-SidedSolar Cell for Optimal Static Concentration,"l4thPVSC,1980, pg. 76.9. I. Tobias and A. Luque, "Series Resistance Calculations inSilicon Solar Cells at Very High Concentration," 21st PVSC,1990, pg. 289.10. P. A. Basore, PC-ID Installation Manual and User'sGuide, Version 3, Sandia Report SAND91-1516, Albuquerque,NM, 991.11.D.L. King, B. R. Hansen, and W. M. Lehrer,"Developmentof a Multi-Purpose, Pulsed-Laser System for Solar CellProcessing Applications," 2lstPVSC, 1990, pg. 278.

We would like to thankE. Buck, B. Silva, and J. Tingley forassistancein processing and M. Narayanan and J. Wohlgemuthfor many useh1 discussions regarding buried-contact celltechnology.

270

ewt solar cell 1993 origin

Documents