IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 47, NO. 5, MAY 2000
`
`987
`
`Comprehensive Study of Rapid, Low-Cost Silicon
`Surface Passivation Technologies
`
`A. Rohatgi, Fellow, IEEE, P. Doshi, J. Moschner, T. Lauinger, A. G. Aberle, and D. S. Ruby
`
`Abstract—A comprehensive and systematic investigation
`of low-cost surface passivation technologies is presented for
`achieving high-performance silicon devices such as solar cells.
`Most commercial solar cells today lack adequate surface passiva-
`tion, while laboratory cells use conventional furnace oxides (CFO)
`for high-quality surface passivation involving an expensive and
`lengthy high-temperature step. This investigation tries to bridge
`the gap between commercial and laboratory cells by providing
`fast, low-cost methods for effective surface passivation. This paper
`demonstrates for the first time, the efficacy of TiO2, thin ( 10
`>
`nm) rapid thermal oxide (RTO), and PECVD SiN individually and
`in combination for (phosphorus diffused) emitter and (undiffused)
`back surface passivation. The effects of emitter sheet resistance,
`surface texture, and three different SiN depositions (two direct
`PECVD systems and one remote plasma system) were investigated.
`The effects of post-growth/deposition treatments such as forming
`gas anneal (FGA) and firing of screen-printed contacts were
`also examined. This study reveals that the optimum passivation
`scheme consisting of a thin RTO with a SiN cap followed by a very
`short 730 C anneal can 1) reduce the emitter saturation current
`15 for a 90
`/sq. emitter, 2) reduce
`density, 0 , by a factor of
`<
`Lr
`0 by a factor of 3 for a 40
`/sq. emitter, and 3) reduce back
`<
`Ss
`Lr
`below 20 cm/s on 1.3
`cm p-Si. Furthermore, this double-layer
`RTO+SiN passivation is relatively independent of the deposition
`conditions (direct or remote) of the SiN film and is more stable
`under heat treatment than SiN or RTO alone. Model calculations
`are also performed to show that the RTO+SiN surface passivation
`scheme may lead to 17%-efficient thin screen-printed cells even
`s.
`with a low bulk lifetime of 20
`nt
`Index Terms—Passivation, rapid thermal oxide, silicon, silicon
`nitride, suface, solar cells.
`
`I. INTRODUCTION
`
`M INIMIZING recombination of minority-carriers at the
`
`surfaces of silicon is crucial for the performance of many
`Si devices including solar cells, BJT’s, CCD’s, and power de-
`vices. The objective of this paper is to provide a comprehensive
`and systematic study of different surface passivation technolo-
`gies available for diffused and nondiffused silicon, planar (flat)
`and chemically textured surfaces. The information is immedi-
`
`ately applicable for junction devices such as solar cells, which
`+
`typically have a n p structure. For such devices, surface passi-
`vation is the key to higher performance especially because the
`trend is toward thinner substrates, which bring the surface closer
`to the collecting junction.
`The passivation schemes investigated include evaporated
`films of TiO , thin SiO films grown in a conventional furnace
`(CFO) and in a rapid thermal processor (RTO), plasma-de-
`posited (PECVD) SiN, and selected combinations of RTO,
`TiO , and SiN. RTO films are of particular interest because
`thin 8–10 nm films can be grown in an extremely short time.
`Films like TiO and SiN are investigated because they provide
`good antireflection properties for silicon, which are essential
`for photovoltaic devices. Since the properties of SiN depend
`strongly upon deposition conditions and the type of PECVD
`equipment used, SiN films from three different sources were
`compared.
`In this study, emphasis is placed on rapid, low-cost technolo-
`gies like RTO and PECVD SiN that can provide effective sur-
`face passivation in short time and with a much lower thermal
`budget than a CFO. Individually, their effectiveness for solar cell
`passivation has been demonstrated previously [1]–[3]. However,
`their combined effect and their ability to withstand subsequent
`thermal treatments necessary for complete solar cell fabrication
`has never been studied. Therefore, the impact of solar cell fabri-
`cation steps like forming gas anneal (FGA) and screen-printed
`contact firing on the surface passivation quality of individual
`and double-layer stacks of dielectrics has also been quantified.
`To address the issue of throughput of RTP and PECVD, sev-
`eral manufacturers and researchers are pursuing the goal of mul-
`tiwafer RTP machines [4] and high-throughput PECVD systems
`[5]. Additionally, ASE Americas, Inc. has already employed
`PECVD dielectrics in commercial cells [6]. Thus, some of these
`promising schemes are currently available for mass production
`and continued effort is expected to deliver higher throughput.
`
`II. EXPERIMENTAL
`
`Manuscript received March 8, 1999. The review of this paper was arranged
`by Editor P. N. Panayotatos.
`A. Rohatgi is with the University Center of Excellence for Photovoltaics
`Research and Education, School of Electrical and Computer Engineering,
`Georgia Institute of Technology, Atlanta, GA 30332-0250, USA (e-mail:
`ajeet.rohatgi@ece.gatech.edu).
`P. Doshi is with Hewlett-Packard, Altanta, GA 30319 USA.
`J. Moschner and T. Lauinger are with the Institut für Solarenergieforschung
`GmbH, D-31860 Emmerthal, Germany.
`A. G. Aberle is with Photovoltaics Special Reserch Centre, University of New
`South Wales, Sydney NSW 2052, Australia.
`D. S. Ruby is with Sandia National Laboratories, Albuquerque, NM
`87185-0752 USA.
`Publisher Item Identifier S 0018-9383(00)02741-6.
`
`To assess the surface passivation of p-type silicon, effective
`(Bex)
`minority carrier lifetime
`measurements were performed
`(00T)
`on 1.3 cm p-type
`FZ silicon wafers coated with various
`+
`passivating films. The investigation of n -emitter passivation
`was performed by
`measurements by the photoconductance
`207
`decay (PCD) technique on phosphorus diffused, high-resistivity
`-<
`(750
`cm), high bulk lifetime (
`ms) FZ Si wafers. Some of
`the wafers were subjected to a chemical random surface tex-
`turing before processing. Surface texturing is commonly used
`in solar cells for reducing reflection losses and optically con-
`fining (trapping) light. Samples for the emitter passivation ex-
`
`Authorized licensed use limited to: Sidley Austin LLP. Downloaded on June 22,2024 at 19:34:05 UTC from IEEE Xplore. Restrictions apply.
`
`0018–9383/00$10.00 © 2000 IEEE
`
`

`

`988
`
`IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 47, NO. 5, MAY 2000
`
`TABLE I
`PLASMA DEPOSITIONS USED IN THIS
`INVESTIGATION
`
`bb
`
`System | Excitation
`No.
`mode
`direct, HF
`SiN1
`(13.6 MHz)
`direct, HF
`(13.6 MHz)
`remote,
`2.45 GHz
`
`SiN2
`
`SiN3
`
`Deposition | Gases
`Temp. [°C]
`SiH, No, NHs
`300
`
`350
`
`400
`
`SiH, (5%) in He,
`N2, NH3
`SiH4, NH3
`
`Set
`were cal-
`samples) and the surface recombination velocity
`culated. The PCD measurement of
`is discussed in Kane and
`Joe
`Set
`Swanson [10] and
`was calculated from the measured
`value using the following two equations [11]:
`
`Teft
`
`+
`ean( BW) _ Set
`=
`B- D,,
`(T» > ov)
`In this study, an infinite bulk lifetime
`was assumed
`Set
`Q
`for high-quality 1.3
`-cm FZ wafers so the calculated
`ac-
`tually represents the worst-case (maximum) value.
`To insure high accuracy from the PCD measurements, each
`data point reported is an average of at least four raw measure-
`ments made over different regions of each sample. Each passi-
`vation was applied to half of a 4-in diameter wafer and the four
`measurements were made evenly across the half-wafer. In some
`cases, such as the unpassivated and RTO passivated samples,
`as many as twelve measurements were averaged because three
`samples were prepared for the three different SiN films used in
`this study. Excluding the infrequent data significantly above or
`below the average, these multiple PCD
`measurements on a
`Teft
`sample showed 5–25% variation. This variation represents the
`combined effects of errors in instrumentation and nonuniformity
`in surface passivation, sheet resistance, and texturing.
`mea-
`Joe
`QO
`surements, especially over the 40 /
`. diffusion were quite uni-
`form because this emitter is less sensitive to surface passivation
`variation.
`
`an
`
`2
`
`III. RESULTS AND DISCUSSION
`
`B
`
`B+ Dy,
`
`Teft
`
`Tb
`
`(1)
`
`(2)
`
`periment were diffused on both sides in an RTP system using
`spin-on dopant sources. Emitters with sheet resistances of 40
`QO
`and 90
`/
`, which correspond to emitters that can accommo-
`date screen-printed and evaporated contacts, respectively, were
`investigated.
`The RTP emitter depth profiles were characterized by
`OEI
`/
`. diffusion had
`spreading resistance analysis. The 90
`x1
`a surface concentration
`of about
`cm and
`(Cs)
`102°
`junction depth
`of 0.18 m [7]. The profile is modified
`(x5)
`slightly after an RTO cycle (discussed in the following para-
`6 x 10/9
`graph).
`reduced to about
`and
`increased to
`Cy
`about 0.25 m. It is important to note that this modest change
`in profile due to RTO did not appreciably alter
`for the
`JoeQO
`same passivation scheme [7]. Similarly, the 40
`/
`. diffusion
`73
`2.5 x 107°
`had a
`of
`cm and junction depth of 0.37 m
`Cs
`C, = 2.1 x 107°
`vj; = O04pe
`which modified to
`and
`m after a
`subsequent RTO cycle.
`After removal of the residual phosphosilicate glass, part of
`the diffused and nondiffused p-type samples were oxidized in
`the same RTP system used for the diffusions. This rapid thermal
`oxidation at 900 C for 150 s resulted in an oxide thickness of
`approximately 8–10 nm on diffused surfaces and about 6 nm on
`Q
`nondiffused 1 -cm p-Si. The oxidized low-resistivity samples
`were then annealed in forming gas at 400 C for 15 min. After
`this, deposition of SiN was performed in three different labo-
`ratories. The thickness of these films was approximately that
`~60
`of an antireflection (AR) coating (
`nm). The refractive in-
`dices of these films measured at 632.8 nm were between 2.15
`and 2.27, which is in the optimum range for single-layer AR
`coatings under glass, or the first film (directly in contact with
`silicon) of double-layer AR coatings in air [8].
`Although its passivation is known to be poor, TiO films are
`also compared for completeness because it is by far the dielec-
`tric most commonly employed dielectric by the PV industry as
`an AR coating. The deposition of TiO was performed by evap-
`2
`orating titanium in an oxygen atmosphere under a low pressure
`of 15 mPa. For the deposition of SiN, three different PECVD
`systems were used. Two of these systems have a parallel plate
`reactor and high frequency excitation, with deposition temper-
`atures of 300 C and 350 C, respectively. The third system is
`a remote PECVD system with microwave excitation and a de-
`position temperature of 400 C [9]. Table I summarizes the dif-
`ferences in key parameters of these systems. The plasma depo-
`A. Passivation of Heavily Doped Emitter Surfaces
`sition systems differ in a number of other aspects, such as the
`reactor geometry, and the plasma power and pressure. However,
`The passivation of solar cell front surfaces was investigated
`QO
`QO
`all three SiN films are used as a standard in the respective labo-
`on both 40
`/
`and 90
`/
`emitters. On a relatively opaque
`QO
`ratories.
`40
`/
`. emitter (which is generally needed to accommodate
`After film deposition, the effective minority carrier lifetime
`screen-printed contacts), the surface is largely decoupled from
`was measured on all samples. Subsequently, a forming
`the emitter bulk, because of the high surface doping concentra-
`(Tet)
`gas anneal (FGA) at 400 C was performed on all samples. As
`tion and depth of the doping profile. Thus, the introduction of
`a final step, the samples were subjected to a short (30 s) temper-
`RTO or SiN passivation resulted in a moderate decrease in
`of
`Joe
`ature cycle with a maximum temperature of 730 C, which is
`about a factor of two to three, as can be seen from Fig. 1. While
`typically used as a firing cycle for screen-printed contacts. This
`TiO showed hardly any passivation, SiN1 was clearly inferior
`step was performed in a beltline furnace with tungsten-halogen
`to RTO or SiN3, which, in combination, resulted in the best pas-
`lamp heating in a compressed air ambient.
`sivation. Note that the high-temperature treatment during RTO
`growth changed the doping profile and lead to a lower surface
`The minority carrier lifetime was measured after each step
`doping concentration, which allowed for better surface passi-
`using a commercially available inductively-coupled PCD tester.
`vation. The
`From these data, the emitter saturation current
`values for textured samples were about 1.5 to 2
`(for diffused
`Joe
`Joe
`Authorized licensed use limited to: Sidley Austin LLP. Downloaded on June 22,2024 at 19:34:05 UTC from IEEE Xplore. Restrictions apply.
`
`

`

`ROHATGI et al.: LOW-COST SILICON SURFACE PASSIVATION TECHNOLOGIES
`
`989
`
`Gi planar
`O textured
`
`464
`
`imme Mol:
`TiO_2 SIN2 SIN1
`
`SiIN3
`
`a“
`eer el | ae
`
`104
`
`86
`
`54] | 3584
`
`RTO RTO+ RTO+ RTO+ RTO*
`SIN3
`TiO_2 SIN2 SIN4
`
`T
`
`1,200
`
`1,424
`
`4,000
`
`800
`
`9
`Z S00
`2
`
`400
`
`200
`
`0
`
`IHA
`
`I
`
`1
`
`I
`
`516
`
`597
`531
`
`442
`
`260
`
`planar
`5 textured
`
`321
`
`174
`
`|
`1,200 7
`
`1.000
`
`=
`0
`g 600 +
`3
`3
`
`400 +
`
`200 +
`
`L
`
`4,241
`
`475
`
`407
`
`21
`
`157
`75
`
`74
`
`10
`
`no
`pass.
`
`Ti0_2
`
`SiN1
`
`SIN3.
`
`RTO RTO+ RTO+ RTO+
`SINS
`SIN4
`TIO_2
`
`Fig. 1. Emitter saturation current densities for different passivation schemes
`on 40
`/ RTP emitters.
`im
`
`times higher than those for planar surfaces, which resembles the
`1.73 times increase in surface area resulting from regular pyra-
`midal texturing.
`QO
`On the relatively transparent 90
`emitters, (which are
`/
`generally used for evaporated or photolithographically-defined
`contacts) the difference in the degree of passivation for various
`schemes was more apparent, as shown in Fig. 2. Again, TiO
`2
`does not provide any appreciable reduction in
`. For the planar
`Joe
`surface, RTO growth reduced
`by more than a factor of ten to
`Joe
`2
`below 100 fA/cm , as does the deposition of SiN . However, on
`3
`the textured surface, RTO is not as effective, resulting in a mod-
`+
`2
`erate
`value of 400 fA/cm . Here, SiN3 and the RTO SiN
`Joe
`double layers were clearly superior resulting in
`values in the
`Joe
`2
`range of 60–215 fA/cm .
`As-deposited double layers of RTO and SiN were better than
`the nitrides alone in all cases, resulting in low
`values of
`Joe
`2
`2
`50 fA/cm for planar and 100 fA/cm for textured emitter sur-
`faces (see Fig. 2). A subsequent forming gas anneal did not
`change the surface passivation appreciably. The same applies for
`QO
`°
`the contact firing cycle (730 C-30 s) on the 40
`/
`emitters.
`This indicates that double layer passivation with a SiN cap pre-
`serves the passivation quality of heavily-doped silicon during
`contact firing. For comparison, thin conventional furnace ox-
`ides (CFO’s) and double layers of CFO and SiN were grown on
`the same emitters. This passivation resulted in identical or only
`slightly lower
`values than the RTO-based schemes.
`Joe
`
`Q
`B. Passivation of Undiffused (1 cm p-Si) Surfaces
`As discussed earlier, this study also focused on evaluating the
`Q
`passivation of undiffused 1.3
`cm p-type silicon surfaces. The
`maximum
`resulting from the many passivation schemes is
`Sot
`Set
`shown in Fig. 3. (Note that
`values above
`cm/s could
`10+
`not be measured reliably by the method used in this study.) The
`impact of the subsequent FGA is also included in Fig. 3. Sim-
`ilar to the heavily-doped surface passivation, TiO again gave
`no measurable surface passivation. The lack of passivation was
`also evident for the as-grown RTO, however, unlike TiO it im-
`proved considerably after FGA. Similar behavior was observed
`for CFO passivation [3]. This indicates that very thin oxides
`rely upon hydrogen to improve surface passivation. Our pre-
`vious capacitance–voltage (C–V) measurements of thin RTO
`
`Fig. 2. Emitter saturation current densities for 90
`/ RTP emitters.
`im
`
`10,000 4
`
`1,000 +
`
`E
`
`a
`=
`2
`
`5
`:
`
`3 100 +
`
`10 +
`
`7
`
`OAs
`
`deposited/grown|
`FGA
`eater
`Fi
`@After FGA & Firing |
`
`a
`396
`
`143
`P|
`
`103
`
`100
`
`pe
`
`TiO_2 SIN1 SIN2 SIN3 RTO RTO+RTO+RTO+RTO+
`TiO_2 SIN1 SIN2 SIN3
`
`Fig. 3. Maximum surface recombination velocities and the effect of
`subsequent heat
`treatments for different passivation schemes on planar
`surfaces.
`
`films showed an order of magnitude reduction in the density of
`interface states at midgap
`following FGA [3].
`(Dix)
`Like the RTO, SiN1 initially gave very poor passivation but
`improved after the FGA. In contrast, as-deposited SiN3 gave
`very good passivation, but degraded slightly after the FGA. The
`superiority of as-deposited SiN3 compared with the other as-de-
`posited SiN films can be due to the many differences in de-
`position parameters, chamber geometry, etc. that can exist be-
`tween different PECVD systems. However, major differences
`are the completely damage-free deposition and the higher de-
`°
`position temperature of 400 C for SiN3. SiN1 and SiN2 were
`deposited at lower temperatures because of constraints of indi-
`vidual PECVD systems. Studies have shown that under proper
`conditions, SiN can result in quite good passivation when de-
`posited by either direct high-frequency (HF) or remote plasma
`systems; however, temperatures up to 400 C are required for
`the best passivation [12]. Thus, SiN1 improved after the FGA
`because it was deposited at only 300 C and the additional 400
`C heat treatment served to deliver the hydrogen more effi-
`ciently to the Si/SiN interface.
`Better results were obtained by double layer combinations of
`RTO with any of the three nitrides (see Fig. 3). Here, the pres-
`ence of SiN on top of the RTO gives the dual benefit of positive
`
`Authorized licensed use limited to: Sidley Austin LLP. Downloaded on June 22,2024 at 19:34:05 UTC from IEEE Xplore. Restrictions apply.
`
`

`

`IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 47, NO. 5, MAY 2000
`
`sod |
`
`[2207
`
`1,990
`11,282 4.174
`
`Bas-deposited
`after FGA
`oO
`
`3.973
`
`TABLE II
`SURFACE RECOMBINATION VELOCITIES USED FOR THE CALCULATIONS IN FIG.
`AND S OBTAINED FROM FIGS. 1 AND 3, RESPECTIVELY. S
`5. (NOTE: J
`WAS CALCULATED AS IN [7])
`
`990
`
`10,000 +
`
`1.000 -
`
`100 -
`
`10
`
`Settmax[cm/s]
`
`|
`
`TiO_2 RTO+
`Tio_2
`
`SiN4
`
`RTO
`
`SIN2 RTO+ RTO+
`SiN 1
`SIN 2
`
`SIN3 RTO+
`SIN3
`
`Fig. 4. Maximum surface recombination velocities for textured surfaces.
`
`charge [13] and possible hydrogenation during PECVD [14].
`+
`The RTO SiN3 initially gives the best passivation because of
`the better hydrogenation during the 400 C deposition. How-
`Set
`ever, after the 400 C FGA,
`values below 50 cm/s are ob-
`+
`served for all RTO SiN combinations. As a result, all three
`double-layer passivations are fairly similar in their quality after
`the anneal. This improvement is attributed to the annealing-in-
`duced release of hydrogen from the SiN that reaches the inter-
`Dit
`face and reduces
`.
`Fig. 4 shows that the same trend was observed for textured
`surfaces, with SiN3 giving considerably better passivation than
`+
`the other nitrides. After FGA, all RTO SiN double layers
`showed good passivation, resulting in a very low
`value of
`Sett
`+
`+
`39 cm/s for RTO SiN3. This shows that RTO SiN passivation
`can be maintained even on textured surfaces.
`+
`As a last step, the SiN and RTO SiN double layers was sub-
`jected to a screen-printed contact firing cycle with a maximum
`temperature of 730 C for about 30 s to test for thermal stability
`(see Fig. 3). In general, the nitrides alone failed to maintain good
`passivation. This may be the result of hydrogen escaping from
`+
`the interface and the SiN films. However, the RTO SiN double
`layers not only withstood the firing, they usually improved after
`+
`firing. This annealing-induced improvement in RTO SiN pas-
`sivation is attributed to the release in hydrogen from the SiN
`film coupled with the passivation of the RTO/Si interface under-
`neath. This is supported by FTIR measurements, which show a
`+
`significant reduction in the hydrogen content of the RTO SiN
`films after high-temperature anneals [15], [16]. Fig. 3 shows that
`Set
`exceptionally low
`values below 25 cm/s resulted regardless
`+
`of the type of nitride used. RTO SiN1, resulted in the lowest
`Set
`value of 12 cm/s on a planar surface. Note that this value
`Set
`gives the same value as the record low
`value of 4 cm/s re-
`sulting from SiN3 passivation [9] which was calculated using
`a bulk lifetime of 1.7 ms. Instead, all calculations in this study
`assumed an infinite bulk lifetime, which results in the higher
`value of 12 cm/s corresponding to the worst-case or maximum
`Set
`as reported in Fig. 3. Thus, the combination of RTO, SiN,
`and the standard high-temperature firing (already common in
`commercial cell manufacturing) results in exceptional surface
`passivation that is virtually independent of different SiN depo-
`sition conditions.
`
`Film
`
`TiO,
`SiN1
`
`SiN3
`RTO+SiN1
`
`RTO+SIN3
`
`Jee (fAfem) | S; (ens) | S, (cms)
`350,000 | Not used
`
`516
`
`419
`
`260
`
`234
`
`174
`
`200,000
`
`1,250
`
`55,000
`
`35,000
`
`5,000
`
`100
`
`12
`
`Not used
`
`C. Impact of Surface Passivation on Photovoltaic Device
`Performance
`Model calculations were performed to predict the impact of
`the various promising surface passivation schemes on the en-
`ergy conversion efficiency of photovoltaic devices. PC-1D for
`Windows [17] version 5.1 was used to solve the one-dimen-
`sional semiconductor transport equations and compute cell per-
`formance. The default material recombination parameters set by
`PC-1D were used. (These include the intrinsic carrier concen-
`nj = 1x 10!
`tration
`cm , the band gap narrowing model
`of Klaassen, et al. [18], mobility model used in Cuevas, et al.
`Cna = 2.2 x 107%
`[19], and Auger coefficient
`of Wang, et
`al. [20]). Other parameters held constant for all calculations in-
`clude a front internal reflectance of 92% (specular), back sur-
`Q
`face reflectance of 70% (diffuse), bulk resistivity of 1.3
`cm,
`QO
`Q
`measured 40 /
`emitter profile, series resistance of 0.6 cm ,
`Q
`shunt resistance of 10 k cm ,
`of 5 nA/cm ,
`of 1.7, and
`Jo
`grid shading (broadband reflectance) of 6%. The latter values
`are consistent with the parameters typically obtained for screen-
`printed cells and result in an FF of about 0.77–0.78. All calcu-
`lations assumed a planar (flat) surface and an optimized double
`layer antireflection coating.
`Fig. 5 shows the calculated cell efficiencies as a function
`of measured front and/or back surface passivation (
`and
`Jo
`) for two different values of cell thickness (
`100
`Ww
`Soff max
`i
`or 300 m) and bulk lifetime (
`20 s or 200 s). Table II
`le
`be
`shows the values of front and back surface recombination
`velocities (
`and
`, respectively) used for the simulations.
`Sy
`Sb
`Q
`was calculated using the measured
`and 40
`/sq.
`Sy
`Joe
`emitter profile as done in [7]. Some of the best commercial
`screen-printed cells are about 14–15% efficient today and
`do not usually have front or back surface passivation. Fig. 5
`shows that up to about 0.5% (absolute) gain in efficiency can
`be derived from improving just the front surface passivation.
`An even greater improvement can be gained by employing the
`high quality passivation schemes on the back surface as well.
`>17
`The calculations show that
`%-efficient screen-printed cells
`+
`are possible with RTO SiN front and back surface passivation
`even on materials with a bulk lifetime of only 20 s.
`le
`It is important to note that the calculations assumed neg-
`ligible contact recombination, which may not be valid espe-
`cially for gridded back contacts. One solution for limiting con-
`
`Authorized licensed use limited to: Sidley Austin LLP. Downloaded on June 22,2024 at 19:34:05 UTC from IEEE Xplore. Restrictions apply.
`
`

`

`ROHATGI et al.: LOW-COST SILICON SURFACE PASSIVATION TECHNOLOGIES
`
`991
`
`sa
`
`cont Zzwna?58
`“
`
`b=]
`
`se]2 =
`Zn
`© & Front& Back
`Passivation
`bj
`
`aU
`
`|
`
`ia
`
`{l
`
`nana
`
`85
`
`175
`
`17
`
`16.5
`
`15.5
`
`15
`
`14.5
`
`14
`
`Efficiency(%)
`
`z
`n
`
`Ti02
`
`
`
`=Ly8 u8
`
`W=300, tb = 200
`
`W=100, tb = 200
`
`W=300, tb = 20
`
`Thickness (4m), Bulk Lifetime (Hs)
`
`Front Passivation Only
`(Sb= 10° cm/s)
`
`emitter, 6% grid
`Impact of front and/or back surface passivation on photovoltaic device performance. All calculations were performed with a 40
`/
`Fig. 5.
`oO
`shading factor, and fill factor of 0.77–0.78 to be consistent with screen-printed solar cells. See Table II for the calculated S and measured S corresponding to
`ff
`each passivation.
`
`17
`
`17
`
`17
`
`17.1
`
`1840
`
`179
`
`188
`tr
`
`Thickness (Hm)=
`— 100
`~* 300
`
`1.3.9Q-cm
`S,= 100 cmis
`
`200
`
`400
`
`800
`1,000
`1,200
`600
`Diffusion Length (Hm)
`
`1,400
`
`1,600
`
`1,800
`
`197
`
`18+
`
`17 +
`
`16 +
`
`15 %
`
`14
`
`13
`
`|
`
`0
`
`Efficiency(%)
`
`Fig. 6.
`
`Impact of thinning cells when S is low. In this case, thinner cells are superior to thick cells when the bulk lifetime is below about 50 s (L = 375 m).
`
`tact recombination is to employ a highly effective local back
`surface field (BSF). Several investigators including ourselves
`Spt
`[3], [21], [22] have demonstrated
`values as low as 200
`cm/s using an optimized screen-printed Al BSF. Thus, the com-
`+
`bination of high-quality RTO SiN silicon surface passivation
`and a gridded (local) BSF to reduce contact recombination may
`help in realizing the high efficiency cells predicted in Fig. 5.
`Additionally, this results in a bifacial structure, which offers the
`possibility of increased power output when rear illumination is
`made available.
`
`Fig. 5 also illustrates the increased importance of back
`surface passivation for thinner cells, which consume less
`T) = 20 pt
`silicon and therefore reduce cost. In fact, for the low
`s
`La => 237 fb
`(
`m) case, the calculations show that decreasing the
`cell thickness from
`300 to 100 m actually improves
`le
`efficiency because of the high-quality back surface passivation.
`This effect is highlighted in the model calculations shown
`in Fig. 6. Bringing the well-passivated back surface within a
`diffusion length of the front collecting junction increases the
`cell's open-circuit voltage
`because of the reduction in the
`(Voc)
`Authorized licensed use limited to: Sidley Austin LLP. Downloaded on June 22,2024 at 19:34:05 UTC from IEEE Xplore. Restrictions apply.
`
`

`

`992
`
`IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 47, NO. 5, MAY 2000
`
`pb
`
`volume of recombination. In this case, the short-circuit current
`density
`does not decrease considerably because the
`(Jsc)
`La
`photogenerated carriers absorbed deeper than
`in the thick
`device have low collection probability. In contrast, the 300 m
`fs
`Bb La=
`thick cells with a bulk lifetime of 200 s (
`750 m) have
`sufficiently long diffusion length to benefit from the good back
`passivation, and thinning this planar cell degrades the efficiency
`because of reduced long-wavelength photon absorption and
`substantial decrease in
`. (This is, of course, a function of
`Js
`the light trapping properties, i.e., thinning passivated cells
`can still help if good texturing and back surface reflector are
`<100
`employed.) Thus, if
`is maintained to the low levels (
`Sb
`W < La
`cm/s) presented in this study, it is better to have
`especially for materials with low lifetimes. These calculations
`are encouraging especially since cost limitations are forcing the
`recent trend to reduced cell thickness with PV grade (defective)
`silicon.
`
`IV. CONCLUSION
`
`This study provides a thorough investigation of silicon
`surface passivation by RTO, TiO , different PECVD silicon
`2
`nitrides, and double-layer oxide/nitride combinations of these
`films. The deposition or growth of these films can be performed
`in a matter of minutes, and all of the passivation schemes used
`in this study provide or allow for near-optimum antireflec-
`tion properties. Thus, they can enhance the performance of
`current industrial solar cells significantly. We have found that
`both RTO and silicon nitride films can individually reduce
`surface recombination substantially. However, double-layers of
`+
`RTO SiN can improve the surface passivation even further,
`2
`resulting in exceptionally low
`values below 50 fA/cm
`Joe
`QO
`QO
`2
`on 90
`/
`emitters, 200 fA/cm on 40
`/
`emitters, and
`Set
`maximum
`values approaching 10 cm/s on a planar 1.3
`Q
`cm Si surface. The combination of RTO and SiN also reduces
`the gap in passivation quality between the different nitrides
`allowing for a high degree of freedom in the SiN deposition
`conditions. Furthermore, this combination has been shown to
`enhance the stability of the surface passivation under thermal
`treatments such as screen-printed contact firing. Textured sur-
`faces revealed a similar trend as planar surfaces but showed an
`expected greater amount of surface recombination. Therefore,
`+
`effective RTO SiN passivation is even more essential for
`textured surfaces since surface recombination can frequently
`limit performance. Finally, model calculations showed that
`+
`the combination of RTO SiN double-layer passivation and
`standard screen-printed contact firing anneal can result in sig-
`nificant improvement of current industrial cells. Calculations
`show that this passivation on the front and back may lead to
`17%-efficient screen-printed cells on thinner substrates (100
`m) with low bulk lifetimes (20 s), resulting in considerable
`cost reduction of photovoltaic cells.
`
`REFERENCES
`[1] S. Sivoththaman et al., “Rapid thermal processing of conventionally
`and electromagnetically cast 100 cm multicrystalline silicon,” in Rec.
`25th IEEE Photovoltaic Specialists Conf.. Piscataway, NJ, 1996, pp.
`621–624.
`
`[2] B. Lenkeit, R. Auer, A. G. Aberle, and R. Hezel, “ Bifacial silicon
`solar cells with screen-printed rear contacts ,” in Proc. 14th Eur.
`Commission Photovoltaic Solar Energy Conf. . Bedford, MA, 1997,
`pp. 853–856.
`[3] P. Doshi et al., “Characterization and application of rapid thermal oxide
`surface passivation for the highest efficiency RTP silicon solar cells,”
`in Rec.26th IEEE Photovoltaic Specialists Conf.. Piscataway, NJ, 1997,
`pp. 87–90.
`[4] M. Lefrancois, “Product information regarding the “RTP solar cell fur-
`nace,” and “High-throughput RTP solar cell processor model VS2400,”
`Vortek Ind., Vancouver, B.C., Canada.
`[5] A. G. Aberle, J. Moschner, and R. Hezel, “PECVD deposition of silicon
`nitride: Fundamentals and prospects for industrial application,” in Proc.
`8th Workshop on Crystalline Silicon Solar Cell Materials and Processes.
`Copper Mountain, CO, 1998, pp. 77–84.
`[6] R. Gonsiorawski and G. Czernienko, “Method of fabricating solar cells
`with silicon nitride coating,” U.S. Patent 4 751 191, assigned to Mobil
`Solar Energy Corp., June 14, 1988.
`[7] P. Doshi and A. Rohatgi, “18% efficient silicon photovoltaic devices by
`rapid thermal diffusion and oxidation,” IEEE Trans. Electron. Devices,
`vol. 45, pp. 1710–1716, Aug. 1998.
`[8] P. Doshi, G. E. Jellison, and A. Rohatgi, “ Characterization and
`optimization of
`absorbing PECVD antireflection coatings
`for
`silicon photovoltaics,” Appl. Opt., vol. 36, pp. 7826–7837, Oct. 20,
`1997.
`[9] J. Schmidt, T. Lauinger, A. G. Aberle, and R. Hezel, “Record low surface
`recombination velocities on low-resistivity silicon solar cell substrates,”
`in Rec. 25th IEEE Photovoltaic Specialists Conf.. Piscataway, NJ, 1996,
`pp. 413–416.
`[10] D. E. Kane and R. M. Swanson, “Measurement of the emitter satura-
`tion current by a contactless photoconductivity decay method,” in Rec.
`18th IEEE Photovoltaic Specialists Conf.. Piscataway, NJ, 1985, pp.
`578–583.
`[11] D. K. Schroder, Semiconductor Material and Device Characteriza-
`tion. New York: Wiley, 1990, ch. 8.
`[12] T. Lauinger, A. G. Aberle, and R. Hezel, “ Comparision of direct
`and remote PECVD silicon nitride films
`for
`low-temperature
`surface passivation of p-type crystalline silicon,” in Proc. 14th
`Eur. Commission Photovoltaic Solar Energy Conf.. Bedford, MA,
`1997, pp. 853–856.
`[13] R. Hezel, “High charge densities in Si-nitride and their effect on the
`inversion layer mobility of silicon MIS/IL solar cells,” in Rec. 16th IEEE
`Photovoltaic Specialists Conf.. New York, NY, 1982, p. 1237.
`, “Silicon nitride for the improvement of silicon inversion layer
`solar cells,” Solid-State Electron., vol. 24, pp. 863–868, 1981.
`[15] A. Rohatgi et al., “Improved understanding and optimization of RTP and
`PECVD processes for high-efficiency silicon solar cells,” in Proc. 13th
`Eur. Commission Photovoltaic Solar Energy Conf., Nice, France, Oct.
`23–27, 1995, pp. 413–416.
`[16] L. Cai and A. Rohatgi, “Effect of post-PECVD photo-assisted anneal on
`multicrystalline silicon solar cells,” IEEE Trans. Electron. Devices, vol.
`44, pp. 97–103, Jan. 1997.
`[17] P. Basore and D. Clugston, “PC1D version 4 for windows: From anal-
`ysis to design,” in Rec. 25th IEEE Photovoltaic Specialists Conf.. Pis-
`cataway, NJ, 1996, pp. 377–381.
`[18] D. Klaassen, J. Slotboom, and H. C. de Graaf, “Unified apparent
`bandgap narrowing in n- and p-type silicon,” Solid-State Electron., vol.
`35, pp. 125–129, 1992.
`[19] A. Cuevas et al., “ Extraction of surface recombination velocity
`of
`passivated
`phosphorus-doped
`silicon
`emitters ,” in Rec.
`24th IEEE Photovoltaic Specialists Conf.. Piscataway, NJ, 1994, pp.
`1446–1449.
`[20] C. Wang and A. Neugroschel, “Minority-carrier transport parameters in
`n-type silicon,” IEEE Electro