IEEE ELECTRON DEVICE LETTERS, VOL. EDL-6, NO. 12, DECEMBER 1985
`
`655
`
`A Polysilicon Emitter Solar Cell
`
`N. GARRY TARR
`
`Abstract—A new solar cell structure is reported in which the emitter
`consists of a thin layer of in situ phosphorus-doped polysilicon deposited
`by a low-pressure chemical vapor deposition (LPCVD) tehniques. The
`highest process temperature required to fabricate this structure is only
`627°C. Although the use of a polysilicon emitter
`results in some
`degradation in blue response, both theoretical and experimental results
`are presented indicating that photocurrent densities in excess of 30
`mA:‘cm~? are attainable under AM1 illumination. The low back-
`injection current associated with the polysilicon emitter has allowed a
`very high open circuit voliage of 652 mV to be obtained at 28°C in a cell
`illuminated to give a short circuit current density of 30 mA:cm~?.
`
`I. INTRODUCTION
`
`T HAS been known for over a decade that polysilicon
`emitter contacts can reduce the back-injection current and
`thereby improve the performance of bipolar transistors [1]. It
`is, therefore, somewhat surprising that consideration has only
`recently been given to the use of polysilicon contacts in the
`production of high-efficiency solar cells. The first suggestion
`that polysilicon might be used to advantage as a contact grid
`material appears to have been made by Van Overstraeten [2].
`This idea was explored experimentally by Green and Blakers,
`who reported briefly on the enhancement
`in open-circuit
`voltage provided by the use of polysilicon contacts in an MINP
`structure [3]. Closely related research into the use of semi-
`insulating polysilicon (SIPOS) as a solar cell contact material
`has been carried out by Yablonovitch, Swanson, and Kwark
`[4]. Very recently, Lindholm, Neugroschel, Arienzo, and Hes
`demonstrated that both n~ and p* polysilicon layers could
`function as efficient back surface field regions [5]. This letter
`reports on a solar cell structure in which a thin, continuous
`layer of in situ doped polysilicon actually forms the entire
`emitter region, rather than simply serving as a contact.
`
`I. Device FABRICATION
`
`The solar cells were fabricated on 5-cm diam silicon wafers
`of (100) orientation obtained from three different vendors. A 2
`x 2-cm cell and a variety of smaller test diodes were placed
`on each wafer. To provide some degree of surface passivation,
`the wafers were first coated with 250 nm of SiO, deposited by
`low-temperature techniques. After device wells were opened
`in this oxide, the wafers were RCA cleaned, briefly etched in
`10 percent HF, and then spun dry and loaded into a low-
`pressure chemical vapor deposition (LPCVD)
`reactor as
`rapidly as possible to minimize growth of an interfacial native
`
`Manuscript received July 23, 1985; revised October 14, 1985. This work
`was supported by the Canadian Natural Sciences and Engineering Research
`Council (NSERC).
`The author is with the Department of Electronics, Carleton University,
`Ottawa, Ont., Canada K1S 5B6.
`
`oxide layer. The emitter deposition was carried out from a
`300/1 silane/phosphine mixture at a temperature of 627°C and
`a pressure of 0.3 torr. Polysilicon thicknesses of 50 and 80 nm
`were examined. It should be stressed that no high-temperature
`anneal was carried out after this deposition, so the highest
`temperature to which the wafers were exposed in the process
`sequence was 627°C. The resistivity of the polysilicon film
`was estimated to lie in the range 0.01-0.02 Q-cm.
`Contacts to the cells were made with aluminum. The back
`contact was annealed at 450°C in a hydrogen ambient. The 1-
`wm thick front contact grid was deposited by thermal
`evaporation and given no annealing treatment, for the anneal-
`_ing of Al contacts to polysilicon/silicon junctions has been
`found to lead to drastic degradation in device characteristics.
`The grid consisted of a central busbar collecting current from
`nominally 18-um wide fingers with centers spaced 317-ym
`apart. Some scatter in finger width from cell to cell resulting
`from variations in etching conditions was noted. The busbar
`itself was tapered linearly from a width of 500 to 100 um. The
`grid finger coverage was 6 percent, and the busbar coverage
`1.6 percent.
`
`Il. Test RESULTS
`
`The short-circuit photocurrent of the polysilicon emitter
`cells was first measured in natural sunlight in Ottawa with the
`sun at approximately 30° from the zenith. Under these test
`conditions a2 x 2 cm back-surface field-space cell manufac-
`tured by Applied Solar Energy Corporation (ASEC) gave a
`short circuit current within 2 percent of its AM1 rating. Later,
`an antireflection coating consisting of 700 A of thermally
`evaporated SiO was applied to three of the cells, and the
`photocurrent density was remeasured under ELH lampillumi-
`nation. The lamp intensity was adjusted to give the rated AM1
`photocurrent in the ASEC cell. Results of these measurements
`are presented in Table I. Measurements of open-circuit voltage
`and fill factor were made at a controlled temperature of 28°C
`with the incident light intensity to produce a photocurrent
`density of 30 mA+cm~? in each device; results are given in
`Table I.
`.
`Previous work has established that it
`is possible to form
`high-quality n* p junctions by the deposition of heavily in situ
`doped polysilicon on monocrystalline substrates [12]. Confir-
`mation of these findings is provided by Fig. 2, which presents
`the dark J-V characteristics for a small-area diode fabricated
`following the procedure outlined above. The diode law is
`obeyed with an ‘‘ideality factor’’ n of approximately 1.1 over
`nearly seven decades of current. At high-current densities,
`resistive effects which may be associated with the tunneling of
`carriers across the polysilicon-silicon junction are observed,
`
`0741-3106/85/1200-0655$01.00 © 1985 IEEE
`
`HANWHA1050
`
`HANWHA 1050
`
`

`

`656
`
`IEEE ELECTRON DEVICE LETTERS, VOL. EDL-6, NO. 12, DECEMBER 1985
`
`—
`
`» GRID FINGER
`I]
`Td
`n+ POLY EMITTER
`
`
`
`
`
`pSi SUBSTRATE
`
`Al BACK CONTACT
`
`
`Structure of the polysilicon emitter cell.
`
`Fig. 1.
`
`TABLE I
`2 cm
`xX
`TOTAL-AREA SHORT-CIRCUIT CURRENT DENSITY FOR 2
`POLYSILICON EMITTER CELLS. Jy. WAS MEASURED ON UNCOATED
`CELLS IN NATURAL SUNLIGHT; J,..2 WAS MEASURED UNDER ELH LAMP
`ILLUMINATION AFTER THE APPLICATION OF AN SiO ANTIREFLECTION
`COATING; THE LAMP--INTENSITY WAS SET TO SIMULATE AM1
`CONDITIONS
`
`
`Cell #:
`l
`
`Substrate:
`Semimetals 1,8 - 2.60cm
`
`Polysilicon Thickness:
`80nm
`
`Jee,1'
`22,5mAcm”
`
`Isc 42!
`30. SmAcm™~
`
`2
`3
`
`4
`
`"
`Monsanto
`
`"
`1.5 - 2.5 0cm
`
`"
`
`"
`
`50nm
`80nm
`
`50nm
`
`22.2
`22,2
`
`22.0
`
`-
`_
`
`30.2
`
`5
`Wacker Float Zone 0.1.¢cm
`80nm
`20.0
`28.5
`
`
`34
`
`a
`33
`a
`= 32
`
`34
`B30
`
`
`
`DIODE AREA = 9x10~4cm?
`a a ee
`DIODE VOLTAGE (V)
`
`art
`=
`z
`2
`
`-=
`
`Fig. 2.. Dark current-voltage characteristics for a small-area polysilicon-
`silicon diode.
`
`but these should not interfere with solar cells operating in
`unconcentrated sunlight.
`
`IV. Discussion
`
`9
`
`‘400
`POLYSILICON THICKNESS (nn)
`Fig. 3. Maximum available AM1.5 photocurrentas a function ofpolysilicon
`emitter thickness. Assumptions madeare a perfect antireflection coating, a
`quantum efficiency of unity, a substrate thickness of 300 um, complete
`collection of carriers photogenerated in the monocrystalline base, and
`complete loss of carriers generated in the emitter.
`
`200
`
`Atfirstinspection, it might be thoughtthat the incorporation
`of polysilicon in the emitter region of a solar cell would lead to
`a sharp degradation in short-wavelength response,
`since
`minority carriers photogenerated in the polysilicon are likely
`to recombine before they can reach the junction and contribute
`to the photocurrent. In fact, theoretical, calculations summa-
`rized in Fig. 3 reveal that for a polysilicon emitter thickness of
`50 nm,
`the maximum available AM1.5 photocurrent
`is
`reduced by less than 5 percent, while for an emitter thickness
`of 100 nm the photocurrent loss is only 8 percent. Fig. 3 was
`produced using a model in whichall carriers generated in the
`polysilicon emitter recombine before reaching the junction,
`while all carriers generated in the monocrystalline base are
`collected. A perfect antireflection coating and a quantum
`efficiency of unity are assumed. Data on the AM1.5 spectral
`
`composition and the absorption coefficientof silicon were both
`obtained from [6]. The substrate thickness was taken as 300
`um. Fig. 3 actually provides a worst-case estimate for the
`photocurrent loss expected when using a polysilicon emitter,
`since there is some evidence that minority carrier holes can
`diffuse across a thin polysilicon layer
`[7]. The J,, data
`presented in Table II support the contention that the photocur-
`rent loss associated with a polysilicon emitter is not severe, in
`that AMI current densities in excess of 30 mA:cm~? have
`been obtained for cells equipped with primitive single-layer
`SiO antireflection coatings formed on substrates of indifferent
`quality.
`The main motivation in using a polysilicon emitter in a solar
`cell
`is
`to exploit
`the reduced back-injection current
`this
`structure provides in order to achieve a high open-circuit
`
`

`

`TARR: POLYSILICON EMITTER SOLAR CELL
`
`voltage. In polysilicon emitter transistors fabricated with the
`same basic process used here, a base-emitter bias of approxi-
`mately 700 mV must be applied to give a base current density
`of 30 mA:cm ~? [8]. Assumingthat /g is dominated entirely by
`back injection into the polysilicon emitter, this suggests that
`the limiting open-circuit voltagé of a polysilicon emitter solar
`cell is approximately 700 mV. To achieve such a high V,, it
`would, of course, be necessary to completely eliminate
`recombination in the monocrystalline substrate and at the back
`contact. Previously reported high V,, cells have made use of
`high-quality float-zone refined silicon with resistivities in the
`range 0.1-0.3 Q:cm. One such wafer was included in the
`present fabrication run, and gave a V,, of 652 mV at 28°C
`with J,. = 30 mA+cm ~? (see TableII). This high open-circuit
`_ voltage compares favorably with the record values obtained by
`Blakers e¢ a/. using MIS tunnel junction contacts in an MINP
`structure [9]. The highest V,, reported by Blakers et al.
`[9]
`under AM1 illumination at 28°C is 655 mV, although V,,’s of
`up to 694 mV have been recorded by this group for similar
`cells under AMOillumination at 25°C. As discussed at greater
`length elsewhere [8],
`it seems quite probable that the low
`back-injection current obtained in the polysilicon emitter
`structure results from a quantum reflection mechanism similar
`to that operating in the MIS tunnel junction contacts of the
`MINPcells.
`The minimum resistivity obtainable in an as-deposited in
`situ phosphorus-doped LPCVD polysilicon LPCVD polysili-
`con is approximately 0.01 Q-cm. In consequence, the sheet
`resistance of a 50-nm thick polysilicon emitter is approxi-
`mately 2 kQ/C). Although this sheet resistance valueis high, it
`should not preclude the attainment of reasonablefill factors in
`polysilicon emitter cells with properly designed contactgrids.
`Fill factors of 0.78 have been obtained for inversion layercells
`[10], even though in these devices the electron density in the
`‘‘emitter’’ is unlikely to exceed 10! cm ~*, corresponding toa
`sheet resistance in the 5-10 kQ/O range. Indeed, the original
`‘‘violet cell’’ had an emitter sheet resistance of approximately
`1 kO/O, yet provided a fill factor of 0.8 [11]. It should be
`possible to reduce the resistivity of a polysilicon emitter
`approximately an order of magnitude. by annealing at a
`temperature of 900°C,
`in which case the emitter sheet
`resistance would have a negligible effect on the cell fill factor.
`However,it is not yet clear whether exposure to temperatures
`in this range will degrade the minority carrier reflecting
`properties of the polysilicon contact. In general, optimization
`of the efficiency of a polysilicon emitter cell requires that a
`compromise be made between the loss of blue response
`encountered with a thick polysilicon layer, and the high series
`resistance associated with a thin layer.
`Examination of Table II reveals that the fill factors of this
`first group of polysilicon emitter cells are rather poor. This
`can be attributed entirely to an inadequate busbar metal
`thickness, giving a parasitic series resistance of approximately
`1 Q. In comparison, the parasitic resistance associated with the
`grid fingers and the polysilicon emitter itself was estimated as
`being roughly 0.2 9. Fill factors as high as 0.7 were obtained
`for cells with areas on the order of a square millimeter, adding
`further support to this interpretation.
`
`TABLE Il
`OPEN-CIRCUIT VOLTAGE V,,,AND FILL FACTOR FF FOR CELLS OF TABLE
`I WHEN ILLUMINATED TO GIVE A PHOTOCURRENT DENSITY OF
`30 mA:cm~? AT A CONTROLLED TEMPERATURE OF28 C
`
`1
`
`2
`
`3
`
`4
`
`5
`
`584 mV
`
`582
`
`583
`
`583
`
`652
`
`0.622
`
`0.614
`
`0.611
`
`0.614
`
`9.605
`
`In the introduction several references which dealt with the
`prospect of forming a high-efficiency solar cell by using a
`polysilicon grid to contact a shallow diffused or ion-implanted
`junction were cited. Such a structure should in principle allow
`the high open-circuit voltages characteristic of polysilicon
`contacts to be combined with the enhanced blue responses
`provided by shallow, oxide-passivated monocrystalline emit-
`ters. However,
`fabrication of such a cell would require
`exposure of the substrates to temperatures in the range 850-
`900°C. Although such temperatures are low by the standards
`of integrated circuit processing, they mightstill lead to some
`degradation in the minority carrier lifetime of high-quality
`material. In contrast, the simple polysilicon emitter structure
`described here can be formed without raising the substrate
`temperature above 630°C, and requires nocritical diffusion or
`implant annealsteps.
`
`V. CONCLUSION
`
`It has been shownthat thin in situ doped polysilicon layers
`hold considerable promise in forming the emitter regions of
`silicon solar cells. Although the use of a polysilicon emitter
`inevitably leads to some degradation in blue response, if the
`polysilicon thickness is kept below 50 nm theoretical caicula-
`tions indicate that
`the AM1.5 photocurrent should not be
`reduced by more than 5 percent, Experimental results support
`this prediction, in that total-area photocurrent densities of over
`30 mA-cm~? have been recorded under simulated AM1
`illumination in polysilicon emitter cells with simple SiO
`antireflection coatings. As expected,
`the polysilicon emitter
`has proven effective in reducing the back injection component
`of the dark current, enabling an open circuit voltage of 652
`mVto be obtained under simulated AM1 conditions for a cell
`formed on a high-quality 0.1 Q-cm float-zone substrate. This
`last
`result supports recent suggestions that
`the use of a
`polysilicon contact grid should enhance the open-circuit
`voltages of more conventional cells with shallow, diffused, or
`implanted emitters.
`,
`
`ACKNOWLEDGMENT
`
`The author wishesto thank D. L. Pulfrey, L. P. Berndt, and
`G. Papadopoulos for their assistance in this project.
`REFERENCES
`
`[1]
`
`‘‘Comparison of experimental and
`P. Ashburn and B. Soerwirdjo,
`theoretical results on polysilicon emitter bipolar transistors,’” ZEEE
`Trans, Electron Devices, vol. ED-31, pp. 853-859, 1984.
`
`

`

`IEEE ELECTRON DEVICE LETTERS, VOL. EDL-6, NO. 12, DECEMBER 1985
`
`R. J. Van Overstraeten, ‘*Advancesin silicon solar cell processing,’ in
`Proc, 15th IEEE Photovoltaic Spec. Conf. New York:
`IEEE,
`1981, pp. 372-375.
`“
`M. A. Green and A. W. Blakers, “‘Advantages of metal-insulator-
`semiconductor structures for silicon solar cells,’’ Sofar Cells, vol. 8,
`pp. 3-16, 1983.
`‘‘An n-~
`E. Yablonovitch, R. M. Swanson, and Y. H. Kwark,
`SIPOS:pSIPOS Homojunction and a SIPOS-Si-SIPOS Double Heteros-
`tructure’’,
`in Proc, 17th IEEE Photovoltaic’Spec. Conf., New
`York: IEEE, 1984, pp. 1146-1148.
`F. A. Lindholm, A Neugroschel, M. Arienzo, and P. A. Iles, ‘‘Heavily
`doped polysilicon contactsolar cells.,’’ JEEE Electron Device Lett.,
`vol. EDL-6, pp. 363-365, 1985.
`M. A. Green, Solar Cells: Operating Principles, Technology and
`System Applications. Englewood Cliffs, NJ: Prentice Hall, 1982.
`P. M. Solomon, D. D. Tang and R. D. Isaac, *‘A polysilicon base
`
`[8]
`
`19]
`
`[10]
`
`{11]
`
`[12]
`
`[5]
`
`[6]
`
`[7]
`
`New York: IEEE,
`
`bipolar transistor,’’ in 1979 IEDM Tech, Digest,
`1979, pp. 510-514.
`‘‘A true polysilicon emitter
`M. B. Rowlandson and N. G. Tarr,
`transistor,’’ ZEEE Electron Device Lett., vol. EDL-6, pp. 288-290,
`1985.
`A. W. Blakers, M. A. Green, S. Jiqun, E. M. Keller, S. R. Wenham,
`R. B. Godfrey, T. Szpitalak, and M. R. Willison, ‘18% efficient
`terrestrial silicon solar cells,’ JEEE Electron Device. Lett., vol.
`EDL-S, pp. 12-23, 1984.
`‘‘655 mV open-circuit voltage,
`R. B. Godfrey and M. A: Green,
`17.6% efficient silicon MIS solar cells,’’ Appl. Phys. Lett., vol. 34,
`pp. 790-793, 1979.
`J. Lindmayer and J. F. Allison, ‘‘The violet cell: An improvedsilicon
`solar cell,’’ Comsat Tech. Rev., vol. 3, pp. 1-6, 1973.
`Z. Lieblich and A. Bar-Lev, “‘A polysilicon-silicon n-p junction,”’
`IEEE Trans. Electron Devices, vol. ED-24, p-p, 1025-1031, 1977.
`
`
`
`