`
`6E:6S:LLvz0zAintZO
`
` 02 July 2024 11:59:39
`
`
`
`A 720 mV open circuit voltage SiO, :c-Si:SiO, doubie heterosiructure solar
`cei!
`
`£. Yablonovitch and T. Gmitter®
`Bell Communications Research, Murray Hill, New Jersey 07974
`
`R. M. Swanson and Y. H. Kwark
`Stanford University, Stanford, California 94305
`
`(Received 29 July 1985; accepted for publication 10 September 1985)
`
`For maximal performancesolar cells should resemble semiconductorlasers,i.e., they should be
`constructed in the form of a double heterostructure. We have found rather good performancein
`SIPOS-crystalline silicon-SIPOS double heterostructure solar cells, where SIPOS=SiO,, . The
`processing ofthese solar cells gives insights into the truly outstanding performanceof the n7-
`SIPOS:p-Si heterojunction which has a forwardsaturation current coefficient Jy = 107 '* A/cm’,
`or equivalently an “emitter Gumme! number” G, = 3.3 X 10"° s/cm*. This suggests that
`crystalline silicon solar cells can be much moreefficient than had been suspected.
`
`It has been recognized for some timethat the structure
`of an ideal solar cell should resembie that of a semiconductor
`laser. The solar cell should be built in the form of a double
`heterostructure. In this configuration a narrow band-gapac-
`tive layer is sandwiched between two wide band-gaplayers
`of opposite doping. An example of a double heterostructure
`biased at a forward voltage V is shownin Fig. 1. The wide
`band-gap materials may be called ‘“‘minority-carrier mir-
`rors” although this term is more frequently applied to the
`high-low doping homojunctionsat the rear ofsolar cells.
`It is usually assumed that the wide band-gap heterocon-
`tact layers must be single crystal and lattice matched to the
`active layer to assure high performance. While this has been
`very successfulfor the ITI-V class of semiconductors,it is not
`actually a necessary condition. The wide band-gap layers
`need only be of sufficient electronic quality to support the
`quasi-Fermilevel separation in the high quality narrow-gap
`active layer. Due to its larger band gap, the heterocontact
`material may be disordered and of poor quality andstill be
`able to support the voltage generated in the active layer. That
`is the key point. The other main pointis that the interface
`states at the heterocontacts must be passivated.
`Since crystalline silicon has no useful lattice-matched
`heterocontacts, we turn to a disordered semiconductor
`heterojunction material, SIPOS(originally named semi-in-
`sulating polycrystalline silicon). In its annealed form it is a
`poorly understood mixture of microcrystalline silicon and
`silicon dioxide (SiO, ). The interface states between SIPOS
`and crystallinesilicon (c-Si) seem to be passivated as well as
`the high quality interface between thermally grown SiO, and
`c-Si.
`
`The first indication of a superior quality heterojunction
`using SIPOS camein the pioneering’ work at Sony. Matsu-
`shita et al.' showed that SIPOSheterojunction emitters in-
`corporated as part of a bipolar transistor structure showed
`greatly enhanced current gains. A more direct measure of
`heterojunction quality is the forward saturation current Jy,
`where the forward current is given by J = J) exp(qV /kT),
`where J, should be as small as possible. An equivalentfigure
`
`"Part of this work was performed while the first two authors were with
`Exxon Research & Eng’g. Co., Annandale, NJ 08801.
`
`of merit? is the “emitter Gummel number” G, = gn?/Jy
`which should be as large as possible. (Here q is the electronic
`charge and n, is the intrinsic carrier density.)
`As a result of the fabrication sequence to be discussed
`below we now routinely produce J, = 107 '* A/cm?for n*-
`SIPOS
`heterocontacts
`on
`p-Si
`(or
`equivalently
`G, = 3.310" s/cm*). This is about two orders of magni-
`tude superior to a conventional homojunction. Unfortunate-
`ly we have been unable thusfar to achieve a correspondingly
`good figure of merit for p-type SIPOS. Our best value of
`forward
`saturation
`current
`on
`p-type
`SIPOS
`is
`Jo = 8X 107 4 A/cm’, whichis nevertheless superiorto that
`of a homojunction. Dueto the lower quality ofp-type SIPOS
`the solar cell we will describe falls short of the idealof Fig. 1,
`in that both faces of the silicon wafer will be covered with
`n*-SIPOS.
`To achieve the highest possible open circuit voltage,
`electron-hole recombination in a solar cel! must be mini-
`mized. There are two major recombination paths; on the
`surface of the active region and in the bulk of the active
`region. The purpose of a heterojunction contact is to mini-
`mize electron-hole recombination on the surface whereelec-
`trical contact to one of the carriers is made. Equally impor-
`tant is the bulk recombination. This is minimized by using
`the highest quality silicon in as thin a crystal as permitted by
`optical absorption considerations. Recently, a quantitative
`analysis* of light trapping in semiconductor sheets showed
`that the effective internal path length for light rays in tex-
`tured sheets was 4? ~ 50 times greater than the sheet thick-
`
`EenTe poev
`~~.be erent itearnBg
`
`Active Reqion
`
`Heterocontact
`
`n- Type
`Heterocontact
`
`FIG.1. Ideal solar cell would be in the form of a double heterostructure, by
`analogy with semiconductorlasers. V = E,,, — Fr, represents the quasi-
`Fermi level separation.
`HANWHA1036
`
`1211
`
`Appi. Phys. Lett. 47 (11), 1 December 1985
`
`0003-6951 /85/231211-03$01.00
`
`© 1985 AmericanInstitute of Physics
`
`1211
`
`HANWHA 1036
`
`

`

`
`
`6E:6S:LL#zOz-AiNZO:
`
` 02 July 2024 11:59:39
`
`
`
`
`
`
`
`
`
`TABLEI. Properties of silicon wafer material.
`
`Orientation
`Resistivity
`Doping concentration
`Bulk minority-carrier
`lifetime
`Minority-carrier mobility
`Diffusion length
`Thickness
`J, of n*-SIPOS
`heterocontact
`Emitter Gummel number
`
`(100)
`0.5 2 cm p-type
`2.5 10'®/cem?
`
`450 us
`700 cm?/Vs
`0.9 mm
`50 zm
`
`10—'* A/cm?at 27°C
`3.3 10"° s/em*
`
`ness (where n is the index of refraction). Accordingly, the
`optimum‘ thicknessof silicon solar cells is =50 zm, much
`thinner than would be the case withoutlight trapping.
`In this work weused float zone p-type silicon wafers of
`0.5 Q cm resistivity which were chemically thinned to 50
`jem. The material properties are summarized in Table I, and
`the sequenceof processing steps for the SIPOS deposition in
`Table If. Step (2) in Table II is critical in achieving a bulk
`minority-carrierlifetime of 450 ys. The as-received material
`is only abouthalf as good.
`Thesilicon wafers, which were coated on both faces
`with 1000 A of n*-SIPOS, were fabricated into devices as
`shownin Fig. 2. The n*-SIPOS was contacted with a 0.5
`cm X 1 cm evaporatedsilver electrode. The p-Si was contact-
`ed by Ga-In metal at one edge of the wafer. No attempt was
`made to optimize the current collection or the fill factor in
`this design. The series resistance was dominated by the bulk
`resistance of the wafer itself between the edge contact and
`the active region. This limited thefill factor to about 0.5.
`The open circuit voltage was measured undera simula-
`tor at 1.3 suns. The light intensity was calibrated by two
`separate standardsolarcells obtained from SERI. An inten-
`sity slightly above one sun was chosen in orderto partially
`compensate for the absence of an antireflection coating on
`thesilicon wafer and to simulate one sun absorbed internal-
`ly. The measured open circuit voltage* was 720 mV at 25 °C,
`with relatively small variations from run to run.
`This outstanding voltage performance is made possible
`both by the excellent bulk quality of the crystalline silicon
`andthe excellent surface passivation ofthe n * -SIPOSheter-
`ocontact. The bulk leakage currentis 1.5 x 107 '* A/cm’?for
`the bulk parameters in Table I while the surface component
`of 10—'* A/cm? must be multiplied by 2 for the two faces of
`
`TABLEII. Processing sequence.
`
`1. Thinsilicon wafers to 50 sm in aqueous KOH.
`2.Getter in an oxidation furnace containing substantial chlorine.
`3.Strip gettering oxide in HF acid.
`4, Immediately load into cold APCVDfurnace while purging with N,.
`5.Ramp up to 650 °C while maintaining N, purge.
`6.Oxidize for 5 min with 100 ppm O,.
`7,Deposit SIPOS from 1000 ppm SiH,, 50 ppm PH,, and 10 ppm O,.
`8.Rampfurnace to 900°C,
`9.Anneal for 15 min.
`10.Ramp furnace down to 450 °C.
`11.Anneal in forming gas for 2 h.
`12.Pull wafers from cold furnace.
`
` E> Ga -In CONTACT
`
`LOAD
`
`FIG.2. Test configuration for the SIPOS: c-Si : SIPOS double heterostruc-
`ture solar cells. In this design thefill factor was limited by the bulkresis-
`tance of the p-Si waferitself.
`
`the wafer. This gives an overall J= 3.5 107'* A/cm’,
`whichis responsible for the observed opencircuit voltage.
`The forwardsaturation current J, of the SIPOS hetero-
`contact was checked in two waysin addition to the implicit
`check associated with the measured open circuit voltage.
`One method was to fabricate bipolar transistors using SIPOS
`emitters® and to measure the transistor characteristics. An-
`other method was to measure the photoconductivity decay
`lifetime’ on wafers with known®excellent bulk lifetime and
`known bulk doping density. Especially for thin wafers, sur-
`face recombination will dominate and the resulting data can
`be used to determine J,. The connection is expressed by
`J, = qn?S /p” , where S is the observed surface recombinatin
`velocity and p® is the bulk majority-carrier density. Al! these
`diagnostic methods agree and give Jy = 10~'* A/cm? for
`wafers subjected to the processing sequence in Table II.
`Ofcritical importance in achieving this performance
`level is the oxidation step (6). In omitting that one step the
`“emitter Gummel number’’ figure of merit degrades by al-
`most an order of magnitude. We found that purging with
`high-purity cylinder gas (<0.5 ppm O, and H,O) helps to
`reproducibly control
`this oxidation step. In some pub-
`lished'?:!° results on SIPOS heterojunction emitters and po-
`lysilicon emitters this step has been omitted, but we believe
`that oxidation frequently occurs inadvertently and is neces-
`sary toward achieving the most favorable figure of merit.
`Dewarnitrogen gas is notoriously unreliable in regard to
`oxygen concentration and was avoided. Likewise it was
`helpful to load the silicon into a cold furnaceto prevent inad-
`vertent oxidation.
`The annealing step (9) follows SEPOS deposition. This
`step is necessary to precipitate microcrystalline silicon
`grains within the material and to activate the phosphorus
`dopant. In addition during this step the phosphorusdiffuses
`about 20-50 A into the silicon wafer as measured by secon-
`dary ion mass spectrometry. This produces a very shallow
`n™* accumulation layer which is transparent to minority car-
`riers; i.e., the holes from the bulk “see”’ the Si-SiO,interface.
`Recent studies'' of electron-hole recombination at thermal-
`ly grown Si-SiO,interfaces have shownthat under accumu-
`lation the recombinationcurrentis exp(gV /kT} x 107 '* A/
`cm’, whichis consistent with the performancelevel achieved
`with the SIPOSemitter structure in Fig.3.
`Oxide thickness between 25 and 40 A was foundto be
`equally favorable with respectto the “emitter Gummel num-
`ber.” Therefore, differential
`tunneling between electrons
`and holes does not appearto be playing a majorrole because
`
`1212
`
`Appl. Phys. Lett., Vol. 47, No. 11, 1 December 1985
`
`Yabionovitch et a/.
`
`1212
`
`

`

`
`
`6E:6S:LLvz0zAintZO
`
` 02 July 2024 11:59:39
`
`microcrystalline structure with a significant surface density
`of individual grains protruding into the thin oxide layer. In
`polysilicon the grains are muchlarger, reducing the density
`of potential contacting points and causing reproducibility
`and repeatability problems.
`We see then that the functioning of the n*-SIPOS
`heterojunction shown in Fig. 3 is complex. The individual
`contacts provide more than enough conductivity for major-
`ity-carrier electrons, but the small fractional contact area
`acts to block minority-carrier holes from recombining in the
`SIPOS.Differential tunneling probability between electrons
`and holes is probably not playing a majorrole, since such a
`process would be very sensitive to oxide thickness contrary
`to what we observe. The functional mechanism ofthe n*-
`SIPOS: p-Si heterojunction emitter will be discussed at
`greater length in a forthcoming”? publication.
`Theresults in this letter suggest that a nonconcentrator
`crystalline silicon solar cell can be designed to achieve an
`efficiency of 24% underair mass| illumination. This design
`would rely on superior heterojunction emitters to improve
`the voitage and onlight trapping* to improve the current ina
`thin device. The recent achievement? of 25% efficiency ina
`c-Si concentratorcell (without the benefit of light trapping)
`should be regarded as further encouragement along these
`lines.
`
`'T. Matsushita, N. Oh-uchi, H. Hayashi, and H. Yamoto, Appl. Phys.Lett.
`35, 549 (1979).
`7H. C. DeGraaf, J. W. Slotboom, and A. Schmitz, Solid State Electron. 20,
`515 (1977).
`3E. Yablonovitch and G. Cody, IEEE Trans. Electron Devices ED-29, 300
`(1982); E. Yablonovitch J. Opt. Soc. Am. 72, 899 (1982).
`*T. Tiedje, E. Yablonovitch, G. D. Cody, and B. G. Brooks, IEEE Trans.
`Electron. Devices ED-31, 711 (1984).
`°E. Yablonovitch, R. M. Swanson, and Y. H. Kwark, The Conference Re-
`cord of the Seventeenth IEEE Photovoltaic Specialists Conference—1984
`(IEEE, New York, 1984), p. 1146.
`*y. H. Kwark, “SIPOS Heterojunction Contacts to Silicon,” Technical
`Report, Solid State Electronics Laboratory, Stanford University, 1985.
`7G. L. Miller, D. A. H. Robinson, and S.D. Ferris, Proc. Electrochem. Soc.
`78-3, 1 (1978)
`’Chemical methods(fluorination) were to be used to minimize surface re-
`combination and permit a measurementof the bulk minority-carrierlife-
`time. B. R. Weinberger, H. W. Deckman, E. Yablonovitch, T. Gmitter,
`W.Kobasz, and S. Garoff, J. Vac. Sci. Technol. A 3, 887 (1985).
`°M. B. Rowlandson and N. G. Tarr, IEEE Electron Device Lett. EDL-6,
`288 (1985).
`‘OB, Soerowirdjo and P. Ashburn, Solid State Electron. 26, 495 (1983).
`"E. Yablonovitch, R. M. Swanson, W. E. Eades, and B. R. Weinberger,
`Appl. Phys. Lett. (to be published).
`"E. Yablonovitch and T. Gmitter, Electron Device Lett. Nov. (1985).
`13RSinton, Y. H. Kwark, S. Swirhun, and R. M. Swanson, [EEE Electron
`Device Lett. EDL-6, 405 (1985).
`
`MICRO -CRYSTALLINE
`GRAINS OF SILICON
`EMBEDDED IN SiQp
`1
`|
`
`SiO,
`
`METALLIZATION
`
`1
`SHO OD
`
`|CRYSTALLINE
`(PIL Con
`WAFER
`
`
`
`
`1
`
`ORIGINAL | La
`THICKNESS
`;
`OF SiOz
`
`FIG.3. Mode!for the metallurgical configuration of the SIPOS-Sihetero-
`Junction. Electron conduction is thought to proceed via the microcrystal-
`line grains which protrude into the thin oxide layer.
`
`over part of this range the oxide is too thick for electron
`tunneling. We believe the majority-carrier electrical contact
`is by individual microcrystalline grains which occasionally
`protrude into the oxide as shown in Fig. 3, i.e., electrical
`contactis via thin spots and pinholes in the SiO,. The chemi-
`cal mixturein step (7) of Table IJ results in a SIPOS formula-
`tion with only 5% oxygen incorporation as determinedbyx-
`ray fluorescence. At such low levels of oxygen concentration
`one might ask what distinguishes a SIPOS heterojunction
`from a polysilicon emitter. Indeed theresistivity of the n*-
`SIPOS (p = 107? 2 cm)is similar to that of doped polysili-
`con. We believe the importance of the 5% oxygen concentra-
`tion is metallurgical rather than electrical. The small
`concentration of oxygen inhibits grain growth and causes a
`
`1213
`
`Appl. Phys. Lett., Vol. 47, No. 11, 1 December 1985
`
`Yablonovitch et a/
`
`1213
`
`