18.5% EFFICIENT FIRST-GENERATION MIS INVERSION-LAYER
`SILICON SOLAR CELLS
`
`Axel Metz, Rudiger Meyer, Burkhard Kuhimann, Manfred Grauvogl, and Rudolf Hezel
`institut fur Solarenergieforschung Hameln/Emmerthal (ISFH)
`D-31860 Emmerthal, Germany
`
`ABSTRACT
`
`CELL DESIGN
`
`In this paper, recent progress in the developmentof
`high-efficiency metal-insulator-semiconductor
`inversion-
`layer (MIS-IL) silicon solar cells at ISFH is presented. We
`fabricated MIS-IL solar cells showing independently con-
`firmed energy conversion efficiencies of up to 18.5%.
`This represents the highest value reported to date for
`MIS-IL silicon cells. The increase in cell efficiency has
`been possible by improvements along several
`lines:
`(i) reduced perimeter recombination losses, (ii) a reduced
`contact resistance of the MIS front grid, and (iii) reduced
`rear surface recombination losses. The cells are charac-
`terised in detail and design modifications for further im-
`provements towards 20% efficiency are presented.
`
`INTRODUCTION
`
`Metal-insulator-semiconductor inversion-layer (MIS-IL)
`silicon solar cells are a promising alternative to con-
`ventional diffused p-n junction Si solar cells. Their low-
`temperature fabrication process is simple and cost effec-
`tive. Furthermore, MIS-IL solar cells have already been
`produced in
`an industrial pilot production and are
`successfully operating in PV power plants [1,2]. Despite
`of being fabricated in a simple process consisting of only
`seven steps,
`the
`efficiency of
`these conventional
`(classified by us as‘first-generation’ cells reaches 15.7%
`on 2x2 cm? FZsilicon (ISFH) and 15.3% on 10x10 cm*
`CZ silicon (ASE, Alzenau) [3]. Detailed two-dimensional
`(2D) numerical modelling and experimental investigations
`have been performed in order to analyse and sub-
`sequently eliminate the device limiting factors of these
`first-generation inversion-layersilicon solar cells [4,5,6].
`in this work we focus on the improvements offirst-
`generation MIS-IL cells
`along the following
`lines:
`()
`reduction of
`the perimeter
`recombination losses,
`(ii) reduction of the losses associated with the contact
`resistance of the MIS front grid, and (iii) reduction of the
`rear surface recombination losses. These improvements
`allowed
`to
`increase
`the
`independently
`confirmed
`efficiency of MIS-IL silicon solarcells to 18.5%.
`
`Standard process
`
`An industrially feasible process for the fabrication of
`MIS-IL cells was developed in the 1980s by R. Hezel’s
`group at the University of Erlangen [7,8]. This MIS-IL cell
`processing sequenceis characterised as follows:
`
`4. Chemical cleaning (optional: texturing of the Si wafer
`with random pyramids before the cleaning)
`2, Rear surface: vacuum evaporation ofAl
`ad
`Preparation of a tunnel oxide at 500°C
`4. Front surface: vacuum evaporation of Al
`metal mask (finger spacing 280 pm)
`5. Al etch (removal of excess Al along the grid lines)
`6. Caesium dip (for increased positive charge densities
`along the Si surface)
`7. Deposition of a low-temperature PECVDsilicon nitride
`film onto the entire front surface at 250°C
`
`through a
`
`Fig. 1 depicts a schematic representation of a MIS-IL
`fabricated with the standard process described
`cell
`above. The optional surface texture consisting of random
`pyramids has been omitted in the drawing for the sake of
`clearness.
`
`Fixed positive
`charges
`
`Front finger
`
`SiN
`
`SiO, \inversion layer
`
`p-Si
`
`Rearelectrode
`
`first-
`standard
`a_
`Schematic diagram of
`Fig. 1.
`generation MIS-IL silicon solar cell. The surface texture
`with random pyramids is not shown.
`
`31
`
`0-7803-3767-0/97/$10.00 © 1997 TEEE
`
`26th PVSC;Sept. 30—Oct. 3, 1997; Anaheim, CA
`
`HANWHA1028
`
`HANWHA 1028
`
`

`

`For a more detailed analysis of the impact of the
`different features on the cell parameters their relative
`deviations with
`respect
`to
`the
`reference
`cells
`incorporating none of the features A, B, or C are shown
`in Fig. 4.
`
`Ss
`es
`xBSpS
`So
`xve,SoS
`+
`reiSeS£0
`oxgSSRS
`SC <>
`sesaseceresSeestatatetso4
`2s
`KSS
`
`MP KeKx?oes+:
`oC
`xes<5056
`Sox
`BRR
`
`5b
`
`18
`
`17
`
`16
`
`15
`
`
`
`Efficiency(%)
`
`CRSRO, 14
`
`io
`
`None
`
`A
`
`All
`
`Measured impact of different high-efficiency
`Fig. 3.
`features on the 1-sun energy conversion efficiency of
`non-textured MIS-IL solar cells on 0.5 Qem p-type FZ
`silicon (AM1.5G, 100 mWicm?, 25°C, aperture area
`~4cm’). The values shown are the average of four
`identically processed cells.
`
`
`
`
`
`Relativedeviation(%)
`
`
`
`SSSSSSSSd
`
`SSSSSSS
`
`
`Reee
`
`
`
`S
`
`Ped Li
`
`on
`
`a,
`
`None
`
`Measured relative deviations of the averaged
`Fig. 4.
`solar cell parameters with respect
`to reference cells
`incorporating none of the features A, B, or C.
`
`32
`
`Design modifications
`
`individual
`In order to investigate the influence of
`high-efficiency
`features
`on
`the
`above-mentioned
`standard MIS-IL cells, their fabrication process has been
`modified. These modifications
`include
`the following
`features:
`
`A) Definition of the active cell area by an oxide window
`B)
`Implementation of Cs underneath the MIS front grid
`C) Implementation of a rear surface with point contacts
`and oxide passivation
`
`Figure 2 shows a modified MIS-IL cell incorporating
`the features A, B, and C. For our studies we prepared
`five different sets of MIS-IL solar cells on 0.5-Cacm p-type
`FZ silicon.
`In order to consider the impact of processing
`related variations, each set consisted of four identically
`processed solar cells. The first set incorporates none of
`the high-efficiency features. Set two uses feature A, set
`three uses feature B, and set four uses feature C. Finally,
`set five exploits features A, B, and C. Since inversion-
`layer emitters are sensitive to variations of the front
`surface roughness, the texturing of the wafers has been
`omitted in these optimisation experiments.
`
`
`
`the modified MIS-IL
`Schematic diagram of
`Fig. 2.
`cells: A) oxide window, defining the active cell area, B)
`implementation of Cs underneath the MIS front grid and
`C)
`rear electrode with
`point
`contacts
`and oxide
`passivation.
`
`RESULTS AND DISCUSSION
`
`Impact of design modifications on cell parameters
`
`Fig. 3 shows the effect of the high-efficiency features
`A to C on the 1-sun energy conversion efficiency of non-
`textured MIS-IL silicon solar cells. The values shown are
`the average of four identically processed cells. As can be
`seen from Fig. 3,
`the incorporation of a single feature
`improved the cell efficiency between 0.2% absolute for
`feature A and 0.9% absolute for feature C.
`If the features
`A-G are implemented into a single solar cell,
`the
`efficiency improves by 2.0% absolute (13.5% relative,
`see right bar in Fig. 3).
`
`

`

`Table |. Measured impact of different high-efficiency features on the 1-sun performance of non-texturedfirst-
`generation MIS-iL solar cells. The values given are the average of four identically processed cells.
`
`
`Feature
`
`None
`
`A
`
`B
`
`Cc
`
`All
`
`Voc
`[mV]
`613 +8
`
`Uso
`[mA/cm’}
`31.5 +0.4
`
`FF
`[%]
`76.8 +0.7
`
`7
`1%]
`14.840.5
`
`Rsdark
`[Qcem’]
`0.77+40.10
`
`Reight
`[Qcm?]
`1.05+0.20
`
`Rspunt
`[kQcm’]
`10 +4
`
`617 +3
`
`371.2 +0.4
`
`77.7 +1.2
`
`15.0 +0.3
`
`0.68 +0.12
`
`4.00 +0.20
`
`698 +134
`
`619 +4
`
`625 +3
`
`638 +7
`
`31.4 +0.1
`
`78.6 +0.6
`
`15.3 +0.4
`
`0.45 +0.03
`
`0.63 +0.02
`
`33.3 40.2
`
`75.0 +1.5
`
`15.7 +0.3
`
`0.88 +£0.02
`
`1.25 +0.10
`
`11 +4
`
`10 +5
`
`33.4 40.2
`
`78.9 +0.2
`
`16.8 +0.2
`
`0.68 +0.05
`
`0.94 +0.08
`
`902 +297
`
`Featura C, the point-contacted rear electrode, has
`the strongest impact on the efficiency improvementof the
`MIS-IL solar cells. Due to the reduction of the rear sur-
`face recombination velocity from about 2x10* cm/s for the
`full-area ohmic back contact [9] to below 100 cm/sfor the
`point-contacted electrode [10],
`the cell efficiency im-
`proves by 0.9% (i.e.
`relatively by 6%). This efficiency
`increase can largely be attributed to an improvement of
`the short-circuit current density Jg- by almost 2 mA/cm?
`(i. e. relative by 6%), see Fig. 4 and Table |. The second
`important feature is the Cs-treatment underneath the MIS
`front grid (feature B). From an evaluation of the series
`resistance Rejgn¢ Of the cell under 1-sun illumination at
`the maximum power point it can be seen that the Cs-
`treatment significantly reduces Rgjg, from 1.05 to 0.63
`Qem*, see Table |. This leads to a significant improve-
`ment of the fill factor approaching 79%, an excellent
`value for MIS-IL solar cells. The definition of the active
`cell area by an oxide window, feature A, has the smallest
`impact on
`the
`cell
`efficiency. The
`corresponding
`reduction of the perimeter recombination increases the
`cell efficiency by 0.2%. This efficiency increase can
`largely be attributed to an improvementofthefill factor by
`1.2% relative, see Fig. 4.
`
`Results for textured cells
`
`In order to improve the efficiency beyond the best
`value (16.8%) of Table |, additional batches of cells incor-
`porating ali features discussed above were fabricated
`with (i) a textured front surface (‘random pyramids’), (ii) a
`double-layer AR coating (PECVDsilicon oxide onto the
`SIN film), and (ii) a front grid with a reduced finger width
`(8 um instead of 14 ym). Table Il shows the 1-sun
`parameters of
`the best
`first-generation MIS-IL cells
`fabricated prior to this work in comparison to the best
`cells fabricated in this work. Our best MIS-IL silicon solar
`cells reach an independently confirmed efficiency of
`18.5%. This increase in cell efficiency by 2.8% absolute
`compared to textured first-generation MIS-IL solar cells
`fabricated with the standard processis largely due to the
`increase in Jsc¢ of 3.7 mA/cm?. Fill factor and Vog could
`also be improved by 3% relative, see TableIl.
`
`33
`
`(mA/cm?)
`
`Currentdensity
`
`18.5% MIS-IL solar cell
`modified process (this work)
`
`15.7% MIS-IL solar cell
`standard process
`
`0.0
`
`01
`
`02 03 04 O.5
`
`Voltage (V)
`
`Measured 1-sun I-V curves of the 15.7% and
`Fig. 5.
`the 18.5% efficient first-generation MIS-IL solar cells
`(aperture area ~ 4 cm?) of Table Il.
`
`the best
`of
`Table Il. Measured 1-sun parameters
`textured first-generation MIS-IL cells fabricated at ISFH ,
`on 0.5-Ocm p-type FZ silicon, (AM1.5G, 100 mWicm?,
`25°C, aperture area ~ 4 cm’, all values independently
`confirmed by Fraunhofer-ISE).
`
`
`First-generation
`MIS-IL solarcell
`
`First generation
`MIS-IL solar cell
`
`Cell type
`
`standard process
`
`modified process
`(this work)
`14 um
`8 um
`14 um
`finger width
`37.7
`39.3
`35.6
`sc [mA/cm’]
`621
`615
`595
`Voc [MV]
`77.4
`76.7
`74.4
`FF [%]
`
`
`
`n[%] 18.5 15.7 18.1
`
`

`

`
`
`TOWARDS 20% MIS-IL SOLAR CELLS
`
`CONCLUSIONS
`
`In order to analyse the efficiency potential of MIS-IL
`cells, we performed two-dimensional numerical simula-
`tions with the semiconductor device simulation program
`DESSIS [6,11]. The cell efficiency was calculated as a
`function of the front and the rear recombination velocity
`parameters Sopon ANA So carThe corresponding results
`are shownin Fig. 6, whereby Spis defined as Sp = ovy,N,.
`In agreement with the experimental
`results of
`the
`previous
`section,
`a
`reduction of
`the
`rear
`surface
`recombination velocity leads to a significant improvement
`of the cell efficiency (transition from 1 to 2 in Fig. 6).
`
`NoNhOoNo
`
`Efficiency
`
`(%) _ co
`
`Calculated efficiency of a MIS-IL silicon solar
`Fig. 6.
`cell as a function of the front and rear surface recom-
`
`bination velocity parameters So7.,; ANd Soja, (0.5-Q¢m
`p-type FZ silicon, Auger-limited bulk lifetime, thickness
`350 um, Jee adjusted to 39 mA/cm’).
`
`is evident from Fig. 6 that an im-
`it
`Furthermore,
`provement of the front surface passivation of the non-
`metallised emitter areas, i.e. a reduction of the front sur-
`face recombination velocity, can further increase the cell
`efficiency (transition from 2 to 3). This can technologically
`be realised by increasing the SiN deposition temperature
`from 250°C to 400°C. So-called truncated-pyramid celis
`have already demonstrated a significant improvement of
`the open-circuit voltage due to the superior surface
`passivation properties of 400°C silicon nitride films [12].
`Further improvements can be obtained by applying me-
`chanicaily structured surfaces consisting e.g. of parallel
`groves [13]. Considering the results discussed in this
`work, this should enable us to reach energy conversion
`efficiencies above 20% by applying the inherently simple
`technology of MIS inversion-layersilicon solar cells.
`
`34
`
`improved metal-insulator-semiconduc-
`In this study;
`tor inversion-layer silicon solar cells are presented with
`independently confirmed ‘1-sun efficiencies of 18.5%.
`This value is the highest ever reported for this cell type.
`The improvement can clearly be attributed to three major
`effects: reduced perimeter recombination losses, a re-
`duced contact resistance of the MIS front grid, and re-
`duced rear surface recombination losses.
`It has been
`shown that the latter had the strongest impact on the
`improvement of the solar cell efficiency. For further im-
`provements towards 20% efficient MIS-IL silicon solar
`cells, novel
`cell
`structures
`incorporating passivation
`schemes that are able to reduce the front surface
`recombination are under investigation at ISFH.
`
`ACKNOWLEDGEMENT
`
`The authors thank all members of the photovoltaic
`department at ISFH for their contributions to this work.
`Financial
`support
`of
`this work
`by
`the
`state
`of
`Niedersachsen and the German Bundesministerium fir
`Bildung, Wissenschaft, Forschung
`und Technologie
`(BMBF)
`is gratefully acknowledged. The ISFH is a
`member of
`the FORSCHUNGSVERBUND SONNEN-
`ENERGIE, Germany.
`
`REFERENCES
`
`[i]
`
`[6]
`
`[7]
`
`[8]
`
`[9]
`
`R. Hezel, W. Hoffmann, and K. Jager, Proc. 10th
`European PVSEC, Lisbon, 1991, p. 511.
`M. Alonso, R. Pottrock, R. Voermans, J.J. Villa, and
`B. Yordi, Proc. 12th European PVSEC, Amsterdam,
`1994, p. 1163.
`ISE PV-Charts, Edition 9 (June 1997), Fraunhofer-
`ISE, Freiburg, Germany.
`B. Kuhlmann, A.G. Aberle, and R. Hezel, Proc. 13th
`European PVSEC,Nice, 1995, p.1209.
`M. Grauvogl, PhD Thesis, University of Hannover,
`1997.
`B. Kuhimann, M. Grauvogl, R.Meyer, R. Hezel, G.
`Heiser, and A.G. Aberle, Proc. 14fh European
`PVSEC, Barcelona, 1997, in press.
`R. Hezel and R. Schérner, JU. Appl. Phys. 52, 3076
`(1981).
`K. Jager and R. Hezel, Conf. Rec. 19th IEEE PVSC,
`New Orleans, 1987, p. 388.
`J. Schmidt, T. Lauinger, A.G. Aberle, and R. Hezel,
`Conf. Rec. 25th [EEE PVSC, Washington D.C.,
`1996, p. 413.
`;
`[10] A.G. Aberle, P.P. Altermatt, G. Heiser, S.J.
`Robinson, A. Wang, J. Zhao, U. Krumbein, and M.A.
`Green, J. Appl. Phys. 77, 3491 (1995).
`DESSIS 4.0, ISE AG, Zurich Switzerland, 1997.
`M. Grauvogl, A.G. Aberle, and R. Hezel, Appi. Phys.
`Left. 69, 1462 (1996).
`R. Hezel, Proc. 13th European PVSEC, Nice, 1995,
`p.115.
`
`[11]
`[12]
`
`[13]
`
`