- Email: [email protected]
Boron Laser Doping through High Quality Al2O3 Passivation Layer for Localized B-BSF PERL Solar Cells
Available online at www.sciencedirect.com
ScienceDirect Energy Procedia 38 (2013) 270 – 277
SiliconPV: March 25-27, 2013, Hamelin, Germany
Boron laser doping through high quality Al2O3 passivation layer for localized B-BSF PERL solar cells Samuel Galla*, Sylvain Manuela, Jean-François Leratb a
CEA- LITEN, INES, 50 av. du Lac Léman, BP332, 73370, Le Bourget du Lac, France b EXCICO Group NV, Kempische Steenweg 305/2, B-3500 Hasselt Belgium
Abstract A new doping process is investigated by using Excimer Laser Annealing (ELA) on various Atomic Layer Deposition (ALD)-Al2O3/PECVD-SiNx(B) passivation stack. In a first part the passivation quality of dielectric stacks is investigated on lowly doped p-type Si substrate. Similar passivation level is highlighted with Boron containing SiN x as compared to un-doped SiNx layer when a thin interfacial Al2O3 layer is first deposited on silicon. In a second part laser doping of the silicon substrate is highlighted by sheet resistance (Rsh) decrease. Pulse energy and pulse number influence the diffusion of Boron and Aluminum atoms from the dielectric stack into the silicon. Electro Chemical Voltage (ECV) profiles confirmed p+ region formation. XPS analysis confirmed the presence of both doping atoms in the p+ region. It is suggested that Al is rather bonded to N and O than to Si atoms while B plays a major role in the doping mechanism of the Si lattice.
© 2013The TheAuthors. Authors. Published by Elsevier © 2013 Published by Elsevier Ltd. Ltd. Selection and/or peer-review under responsibility of thec scientific committee of the SiliconPV Selection and/or peer-review under responsibility of the scientifi committee of the SiliconPV 2013 conference2013 conference Keywords: solar cells, p-type silicon; boron, aluminum; silicon nitride; co-annealing; laser doping; laser annealing
1. Introduction Passivated Emitter, Rear Locally-Diffused (PERL) solar cell with B-BSF structure provides high performance when low rear surface recombination velocities are reached 1,2. Thermally-grown SiO2 layer suites well for reducing defect density Dit at the Si interface3 but its formation requires long and high thermal (~1050°C) dry processing and may potentially affects the quality of the Si bulk. Wet oxidation processes at lower temperatures (~800°C) have demonstrated good performances with SiNx capping 4,5. Al2O3 synthesized by ALD has demonstrated high surface passivation level on lowly doped surface 6 and
* Corresponding author. Tel.: Tel: +33 479 792 081 E-mail address: [email protected]..
1876-6102 © 2013 The Authors. Published by Elsevier Ltd.
Selection and/or peer-review under responsibility of the scientific committee of the SiliconPV 2013 conference doi:10.1016/j.egypro.2013.07.277
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
on boron doped p+ region 7-10. High fixed negative charge density at the Si/Al2O3 interface provides effective field-effect passivation 10,11 and excellent thermal stability during firing step when capping with low temperature thermal SiO2 6 or PECVD-SiNx12. Mostly used techniques for rear side contacting are dielectric passivation with Local Back Surface Field (LBSF/PERL), dielectric passivation with ohmic contacts (PERC), dielectric passivation with laserfired contacts (LFC), full area Boron-back surface field (BSF) and evaporated aluminium (ohmic contact)13. Formation of efficient local p+ BSF region through SiO2/Al2O3 or SiNx/Al2O3 by laser ablation of the dielectric stack followed by Al screen printing has been reported14. The full-sheet screen printed Al paste contacts the rear locally at those areas where the dielectric layer has been removed. This paper focuses on a laser process combining high quality dielectric passivation and laser doping for localized B-BSF formation. Efficient boron laser doping from PECVD-SiNx(B) with a 150 ns pulse duration Excimer laser source emitting at 308nm was reported in a previous work15. Nevertheless the poor passivation level of the SiNx(B) layer requires its complete removal after the laser diffusion step before new passivation step 15,16. The passivation quality was strongly enhanced by adding a dry thermal SiO2 (900°C) interfacial layer between substrate and dopant layer without preventing subsequent B laser diffusion through the oxide layer. Alternative stack with low temperature wet SiO2 interfacial layer (700°C) did not provide similar passivation level. With the goal of reducing the thermal budget of the dielectric stack deposition and the complexity of the rear side process of the PERL solar cell, a low temperature ALD-Al2O3 layer is combined with PECVD-SiNx(B) dopant layer to address simultaneously high passivation level and boron dopant availability. 2. Experimental p-type 1Al2O3/SiNx(B) stacks were used for the passivation study. 5 and 10nm thick Al2O3 layers were deposited by plasma ALD technique at 250°C. 75nm thick un-doped hydrogenated silicon nitride (SiNx) and Bcontaining silicon nitride (SiNx(B)) layers were deposited by PECVD. Detailed deposition parameters of SiNx(B) could be founded elsewhere16. Samples underwent a firing step in an IR belt furnace at 800°C to simulate a typical screen printing process. Implied open-circuit voltages of Al2O3/SiNx(B)/cSi(p)/Al2O3/SiNx(B) symmetric samples were measured using the quasi-steady state photoconductance (QssPC) technique. For the doping study n-type 1substrates were covered with ALD-Al2O3 (5 nm) / PECVD-SiNx(B) (80 nm) dielectric stack. Wafers were laser-irradiated at various energy densities with a XeCl 150 ns pulsed excimer laser emitting at 308nm. Doping process was evaluated through 4-probes sheet resistance measurements and Boron and Aluminum SIMS analysis. Furthermore active profiles were evaluated by Electro Chemical Voltage (ECV) measurement using a NH4F electrolyte and a ring surface of 0.101 cm². Chemical analysis was performed by X-ray Photospectroscopy (XPS) on sample surface and after Ar sputtering at 3 keV with an estimated rate of 1.5 nm per minute. Binding energies were corrected for specimen charging by referencing the C1s peak to 285 eV. 3. Results and discussion 3.1. Passivation study The implied Voc values of double side passivated substrates with (i) single layer of SiNx, (ii) single layer of SiNx(B), (iii) Al2O3/SiNx stacks and (iv) Al2O3/SiNx(B) measured before and after the thermal treatment are presented on Fig 1. The un-doped SiNx shows implied Voc of 638 mV before annealing up to
271
272
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
654 mV after annealing. Plasma and thermal ALD-Al2O3 films are known to provide an outstanding level of surface passivation on lowly doped p-type surface6 but plasma ALD films provides virtually no surface passivation in the as-deposited state17. In our case the Al2O3 layers are partially activated during the subsequent deposition of the SiNx film at 450°C. This leads to 642 mV and 659 mV implied Voc measured for the Al2O3(5 nm)/SiNx and the Al2O3(10 nm)/SiNx as-deposited stacks. Further annealing at 800°C leads to surface passivation improvement with implied Voc up to 669 mV and 679 mV for 5 and 10 nm thick Al2O3 layer, respectively. Quality enhancement could be explained by cumulative field effect passivation (scaled with the fixed charged density Qfix) and interface state passivation (Dit) of the SiNx capped-Al2O3 stack while a single layer of SiNx provides only Qfix-induced passivation. As already reported on previous studies the presence of B in the SiNx layer negatively affects the surface passivation performances when directly deposited on lowly doped Si substrate15-16. The implied Voc of the SiNx(B) layer is close to 500 mV before and after annealing. CV measurements have demonstrated a Qfix decrease with the B concentration in the SiNx(B)18. Nevertheless PC1D simulations suggested that the passivation of SiNx(B) layer is limited by a high Dit level rather than by Qfix18. The initial passivation level of SiNx(B) could be improved by Al2O3 layer deposition prior the dopant layer. The damaging effect of the SiNx(B) layer on the surface passivation is hidden by the Al2O3 layer. As a result implied Voc values increased from 550-600 mV range before annealing to 669 mV and 672 mV after annealing with respectively 5 and 10 nm of Al2O3. This demonstrated that the highly defective interface between the silicon substrate and the SiNx(B) layer could be strongly enhanced by adding an interfacial Al2O3 passivation film. Similar shadowing effect was reported in the case of SiO2(10 nm)/SiNx(B) stack19 but passivation mechanisms are different. Thermal oxide yielded a high level of chemical passivation due to a low interface defect density while plasma ALD-Al2O3 provides high quality passivation by field effect. 700
as deposited after firing
Implied Voc [mV]
650 600 550 500
x
x
0n m
)/S
iN
)/S
iN
(B
)
(1 3
2
Al
O
3
3
O 2
Al
2
Al
O
(5
(1
nm
0n m
)/S
x
iN
x
iN )/S nm (5 3
O 2
Al
(B
) (B x
N Si
x
N Si
)
450
Fig. 1. Implied Voc of various dielectric stacks deposited on polished c-Si(p) substrate (measured at 1Sun, 5 samples average)
3.2. Doping study As shown in Fig. 2, the sheet resistance after one pulse irradiation of SiNx(B)/Al2O3/c-Si(p) sample decreases continuously from 600 at 1.5 J/cm² below 100 at 5.0 J/cm². Dopants introduced in the melted silicon phase became electrically active during recrystallization and formed a p+ emitter on the ntype substrate. Boron and aluminum atoms are both donor impurities available in the dielectric stack and
273
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
thus are expected to produce hole charge carriers when substituting for Si atoms in the lattice. Sheet resistance decrease suggests diffusion of at least one type of dopants has diffused into the Si crystal. Moreover two subsequent pulses at same fluence strongly reduced the sheet resistance. At 3.1 J/cm² energy density the sheet resistance drops to 25 after two pulses while 360 was measured after single pulse irradiation. As shown in Fig. 3, ECV analysis profiles exhibited deeper depth and more surface concentrated profiles with increasing laser energy density and pulse number. This indicates enhanced diffusion and/or activation rate of the impurities regarding both parameters. 700
1E20
1 pulse 2 pulses
600
3.0 J/cm²
1E19
400
4.0 J/cm²
1 pulse 2 pulses 1 pulse 2 pulses 1 pulse 2 pulses
-3
N [at.cm ]
Rsheet [ /sq]
500
300 200
1E18
1E17
100 0 1,5
2.0 J/cm²
2,0
2,5
3,0
3,5
4,0
4,5
5,0
Fluence [J/cm²]
Fig. 2. Sheet resistance after irradiation of SiNx(B)/Al2O3 stack versus energy density for single and two pulses irradiation
1E16 0,0
0,2
0,4
0,6
0,8
1,0
Depth [μm]
Fig. 3. ECV profiles after irradiation of SiNx(B)/Al2O3 stack at 2.0, 3.0 and 4.0 J/cm² for single and two pulses irradiation
The following study focused on 2 pulses irradiation at 2.0, 3.0 and 4.0 J/cm² energy densities since the resulting sheet resistances and dopant profiles suited better for BSF application20. SIMS analysis shown on Fig. 4 was performed on the 3 samples to quantify impurity diffusion during the laser process. Constant boron and aluminium top surface concentrations (~3.1021 at/cm3) whatever the laser fluence suggest the existence of a residual dielectric layer containing Al and B atoms on top of the silicon surface. The interface position between the silicon surface and this residual dielectric layer estimated by the slope change of the Si profile (green line) is around 50 nm for the 3 samples. The profiles measured in the silicon (i.e. from 50 nm depth) highlighted simultaneous diffusion of B and Al atoms from the dielectric stack. Profiles revealed depth increase and reduced surface concentration when fluence increases but Al profiles are straight while B profiles are bell-shaped. Such B profiles were already reported in a previous study B diffusion from single SiNx(B) layer, leading the sheet resistance decrease15. By analogy a key role of the boron impurities in the doping process of the Si substrate could thus be expected. This hypothesis is straightened by the comparison of ECV and Al and B SIMS profiles. Indeed the first technique detects only electrically actives atoms while all (i.e. active and inactive) are detected by the second one. B and Al concentrations measured in the silicon (not in the residual layer) at 3.0 and 4.0 J/cm² are both excessively high at the near-surface (>2.1021 at/cm3) while ECV concentration is closed to 1020 at/cm3. That suggest only some of these dopants occupy the silicon lattice sites under electrically activated substitutional state, promoting silicon doping. As the ECV profiles are closer to the SIMS Boron profile at high fluence, it could be suspected a leading role of the boron in the doping of the p+ region in contrast to Al atoms occupying inactive interstitial states. They could form clusters particularly in the heavily doped region. Nevertheless this assumption could not be drawn at lower fluence (2.0 J/cm²) because ECV
274
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
concentration is strongly lower than those measured by SIMS. Only low activation rate of both dopant types could be concluded.
Fig.4. Boron and aluminum SIMS and ECV profiles at 2.0, 3.0 and 4.0 J/cm² (two pulses irradiation)
For further understanding the competition between both dopant types involved in the doping mechanisms, a XPS study has been performed on the 3 samples. Chemical bonding investigations were carried out at the surface (0 min etching) and after 20 and 40 min etching time. Related etching depths are 40 and 80 nm, respectively. It is assumed the residual dielectric is analysed on surface and at 40 nm etching time while 60 nm etching depth ensures silicon substrate analysis. As shown in Fig. 5a, the increasing Si-Si bonds signature at 99.5 eV after surface etching confirms the effective removal of the dielectric layer during the etching step. Contributions of Si-O seen at 104.0 eV and Si-N at 103.0 eV on 0 min and 20 min etched samples are due to SiO2 and SixNy compounds remaining on the Si surface after irradiation. Both Si-N and Si-O intensities disappeared after 40 min etching because all the residual SiO2 and SixNy compounds have been removed. The detection of chemical bonding states of B, whose concentration in Si is in the order of 1% or less is difficult because the photoionization cross section of B is small. Nevertheless the peak suspected in the B1s spectrum at 192.2 eV of un-etched samples (Fig. 5c) could be assigned to oxidized boron contribution resulting from SiNx(B)/Al2O3 stack irradiation. As seen on the Al2s spectrum plotted in Fig. 5b, the Al-O peak observed at 120.2 eV of un-etched sample reveals residual Al2O3 compounds. This peak could be correlated with the high Al concentration detected by SIMS at the surface after laser irradiation. The Al-O peak is still present after 40 min of etching time, i.e. after complete removal of the dielectric layer. An additional peak at 116.5 eV related to Al-N bonds suggests AlxNy compounds. The N1s spectra (Fig. 6b) of un-etched samples disclose Si-N signature decreasing at higher energy density. Based on O1s, Si2p and B1s spectra, it could be assumed Al and B containing Six(O,N)y residual dielectric layer remaining on the silicon surface. On the other hand a low intensity peak centred at 102.0 eV suspected on the Si2p spectra after 40 min etching could be related to Si-B bonds. The highest Si-B intensity is measured for the 4.0 J/cm² sample and could be correlated with the ECV profile where most of the B dopants are substituted for the Si atoms in the lattice. Al-O and Al-N contributions on Al2s spectra are clearly visible for the 3 fluences but the Al-O peak is more intense for the 4.0 J/cm² sample. The N1s spectra (Fig. 6b) after 40 min etching reveal high intensity Si-N peak at 2.0 J/cm² then decreasing at higher fluence. At 4.0 J/cm², the Si-N contribution is very low and a slight N-N signature appears. It could be concluded that Al atoms diffusing at high fluence associate mainly with N and O and maybe with Si. It could be assumed that the Boron atoms are introduced in the Si matrix under substitutional state and participate to the effective doping of
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
the p+ region while the Al atoms are only driven in the silicon following interstitial diffusion and forming AlxOy and AlxNy compounds. Further XPS analyses are on-going to conclude on this hypothesis.
Fig. 5. XPS peaks of Si2p (a), Al2s (b) and B1s (c) on samples 2.0, 3.0 and 4.0 J/cm² (2 pulses conditions) after 0, 20 and 40 min of etching time
Fig. 6. XPS peaks of C1s, N1s and O1s on samples 2.0, 3.0 and 4.0 J/cm² (2 pulses conditions) after 0, 20 and 40 min of etching time
275
276
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
4. Conclusion We demonstrated similar passivation quality with SiNx(B) layer as compared to un-doped SiNx when a thin Al2O3 interfacial layer is first deposited on silicon. In a second part, we demonstrated simultaneous Al and B diffusion in the Si substrate by irradiation of Al2O3/SiNx(B) stacks with an Excimer laser source. Both dopants profiles are influence by laser energy density. The formation of a p+ doped region highlighted by ECV analysis confirmed the effective doping of the Si lattice by donor impurities. The activation rate and/or the concentration of active dopant increased with laser fluence. The XPS study reveals Al atoms being rather bonded to N and O forming Al-O and Al-N compounds. It could be concluded un-active state of most of the Al atoms. As the active profile fits the boron profile at high fluence, B atoms seems to substitute locally for Si in the substrate. Nevertheless the B-Si bonds contributions suspected on B and Si XPS spectrum are too weak to strengthen this hypothesis. The quality Al2O3/SiNx(B) stacks could thus be suitable for both back side passivation of PERL solar cells and dopant source for local B-BSF formation. Boron profiles ensuring efficient Back Surface Field effect require sheet resistance between 40 and 60 p-type substrate20. Surface concentration 20 3 below 10 at/cm and 1 μm range depth profile suites well for enhanced diffusion length of the charge carriers in the doped region. Our laser process could be easily tuned to provide adapted boron profiles. The effect of Al containing compounds has to be investigated in term of electrical quality of the doping region. Acknowledgement The authors thank Nicolas Auriac for ALD deposition, Rachid Hida (CEA-LETI) for PECVD processing, Lucasz Borowik (CEA-LETI) for XPS analysis and Thomas Wolf (WEP Control) for ECV measurements. This work was supported by the General Directorate for Competitiveness, Industry and Services (DGCIS ex DGE) under convention N° 10.2.90.6027 and by the French Environment and Energy Management Agency (ADEME) under convention No. 1005C0173. References 1. Phys. Status Solidi RRL 5, No. 4, 147 149 (2011) / DOI 10.1002/pssr.201105045 2. A. Das on Screen-Printed Cells with spin38th IEEE Photovoltaic Specialist Conference, Vol57, N°10, october 2012 3. Solar Energy Materials & Solar Cells, 90 (2006) 82-92. , Phys. 4. Status Solidi RRL 5, No. 1, 22 24 (2011) 5. O. Schultz et al, Prog. Photovolt. Res. Appl. 16, 317 (2008). 6. J. Benick et al., Prog. Photovolt. Res. Appl. (2010), DOI 10.1002/pip.1020. 7. a-SiNx:H stacks for surface passivation of crystalline silicon J. Appl. Phys. 2009; 106: 114907. 8. A. -type silicon solar Proc. 25th European Photovoltaic Solar Energy Conf., Valencia, Spain, p. 1453-9.
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
9. -chargeAppl. Phys. Lett. 2007; 91: 112107. 10. n-type Si solar cells on Al2O3-passivated Applied Physics Letters, vol. 92, pp. 253504/1-3, 2008 11. G. Agostinelli et al p-type silicon wafers passivated with a dielectric with fixed negative charges 90 (2006) 3438-3443 12. Aluminum Alloying in Local Contact Areas on Dielectrically Passivated Rear Surfaces of Silicon Solar 37th IEEE Photovoltaic Specialist Conference, June 19-24, 2011, Seattle 13. 20th European Photovoltaic Solar Energy Conference, 6 10 June 2005, Barcelona, Spain 14. F. S. Grasso F.S. et al, Proc. 25th European Photovoltaic Solar Energy Conf, Valencia, Spain, 2010; 371 15. Laser Doping Strategies Using SiN:P and SiN:B Dielectric Layers for Profile Engineering in High Efficiency Solar Cell Energy Procedia, Volume 27, 2012, Pages 449-454 -rich, P and B 16. S. Gall, R. Hida, B. PavietProc. of 26th European Photovoltaic Solar Energy Conference, Hamburg, Germany (2011) 1648-1652. 17. J. Schmidt, B. Veith, and F. Werner, D. Zielke, R. Brendel, Silicon Surface Passivation by Ultrathin Al2O3 films and Al2O3/SiNx stacks ,35th IEEE 07/2010; DOI:10.1109/PVSC.2010.5614132 In 18. B. PavietInvestigation of charges carrier density in phosphorus and boron doped SiNx:H layers for crystalline silicon solar cells , Materials Science and Engineering: B, Volume 178, Issue 9, 15 May 2013, Pages 580-585 19. High Quality Passivation Scheme Combined with Laser Doping from SiNx(P) and SiNx(B) Layer for Silicon Solar Cell , Energy Procedia, Volume 27, 2012, Pages 467-473 20. Proc. of 26th European Photovoltaic Solar Energy Conference, Hamburg, Germany (2011) 1274 - 1279
277
ScienceDirect Energy Procedia 38 (2013) 270 – 277
SiliconPV: March 25-27, 2013, Hamelin, Germany
Boron laser doping through high quality Al2O3 passivation layer for localized B-BSF PERL solar cells Samuel Galla*, Sylvain Manuela, Jean-François Leratb a
CEA- LITEN, INES, 50 av. du Lac Léman, BP332, 73370, Le Bourget du Lac, France b EXCICO Group NV, Kempische Steenweg 305/2, B-3500 Hasselt Belgium
Abstract A new doping process is investigated by using Excimer Laser Annealing (ELA) on various Atomic Layer Deposition (ALD)-Al2O3/PECVD-SiNx(B) passivation stack. In a first part the passivation quality of dielectric stacks is investigated on lowly doped p-type Si substrate. Similar passivation level is highlighted with Boron containing SiN x as compared to un-doped SiNx layer when a thin interfacial Al2O3 layer is first deposited on silicon. In a second part laser doping of the silicon substrate is highlighted by sheet resistance (Rsh) decrease. Pulse energy and pulse number influence the diffusion of Boron and Aluminum atoms from the dielectric stack into the silicon. Electro Chemical Voltage (ECV) profiles confirmed p+ region formation. XPS analysis confirmed the presence of both doping atoms in the p+ region. It is suggested that Al is rather bonded to N and O than to Si atoms while B plays a major role in the doping mechanism of the Si lattice.
© 2013The TheAuthors. Authors. Published by Elsevier © 2013 Published by Elsevier Ltd. Ltd. Selection and/or peer-review under responsibility of thec scientific committee of the SiliconPV Selection and/or peer-review under responsibility of the scientifi committee of the SiliconPV 2013 conference2013 conference Keywords: solar cells, p-type silicon; boron, aluminum; silicon nitride; co-annealing; laser doping; laser annealing
1. Introduction Passivated Emitter, Rear Locally-Diffused (PERL) solar cell with B-BSF structure provides high performance when low rear surface recombination velocities are reached 1,2. Thermally-grown SiO2 layer suites well for reducing defect density Dit at the Si interface3 but its formation requires long and high thermal (~1050°C) dry processing and may potentially affects the quality of the Si bulk. Wet oxidation processes at lower temperatures (~800°C) have demonstrated good performances with SiNx capping 4,5. Al2O3 synthesized by ALD has demonstrated high surface passivation level on lowly doped surface 6 and
* Corresponding author. Tel.: Tel: +33 479 792 081 E-mail address: [email protected]..
1876-6102 © 2013 The Authors. Published by Elsevier Ltd.
Selection and/or peer-review under responsibility of the scientific committee of the SiliconPV 2013 conference doi:10.1016/j.egypro.2013.07.277
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
on boron doped p+ region 7-10. High fixed negative charge density at the Si/Al2O3 interface provides effective field-effect passivation 10,11 and excellent thermal stability during firing step when capping with low temperature thermal SiO2 6 or PECVD-SiNx12. Mostly used techniques for rear side contacting are dielectric passivation with Local Back Surface Field (LBSF/PERL), dielectric passivation with ohmic contacts (PERC), dielectric passivation with laserfired contacts (LFC), full area Boron-back surface field (BSF) and evaporated aluminium (ohmic contact)13. Formation of efficient local p+ BSF region through SiO2/Al2O3 or SiNx/Al2O3 by laser ablation of the dielectric stack followed by Al screen printing has been reported14. The full-sheet screen printed Al paste contacts the rear locally at those areas where the dielectric layer has been removed. This paper focuses on a laser process combining high quality dielectric passivation and laser doping for localized B-BSF formation. Efficient boron laser doping from PECVD-SiNx(B) with a 150 ns pulse duration Excimer laser source emitting at 308nm was reported in a previous work15. Nevertheless the poor passivation level of the SiNx(B) layer requires its complete removal after the laser diffusion step before new passivation step 15,16. The passivation quality was strongly enhanced by adding a dry thermal SiO2 (900°C) interfacial layer between substrate and dopant layer without preventing subsequent B laser diffusion through the oxide layer. Alternative stack with low temperature wet SiO2 interfacial layer (700°C) did not provide similar passivation level. With the goal of reducing the thermal budget of the dielectric stack deposition and the complexity of the rear side process of the PERL solar cell, a low temperature ALD-Al2O3 layer is combined with PECVD-SiNx(B) dopant layer to address simultaneously high passivation level and boron dopant availability. 2. Experimental p-type 1Al2O3/SiNx(B) stacks were used for the passivation study. 5 and 10nm thick Al2O3 layers were deposited by plasma ALD technique at 250°C. 75nm thick un-doped hydrogenated silicon nitride (SiNx) and Bcontaining silicon nitride (SiNx(B)) layers were deposited by PECVD. Detailed deposition parameters of SiNx(B) could be founded elsewhere16. Samples underwent a firing step in an IR belt furnace at 800°C to simulate a typical screen printing process. Implied open-circuit voltages of Al2O3/SiNx(B)/cSi(p)/Al2O3/SiNx(B) symmetric samples were measured using the quasi-steady state photoconductance (QssPC) technique. For the doping study n-type 1substrates were covered with ALD-Al2O3 (5 nm) / PECVD-SiNx(B) (80 nm) dielectric stack. Wafers were laser-irradiated at various energy densities with a XeCl 150 ns pulsed excimer laser emitting at 308nm. Doping process was evaluated through 4-probes sheet resistance measurements and Boron and Aluminum SIMS analysis. Furthermore active profiles were evaluated by Electro Chemical Voltage (ECV) measurement using a NH4F electrolyte and a ring surface of 0.101 cm². Chemical analysis was performed by X-ray Photospectroscopy (XPS) on sample surface and after Ar sputtering at 3 keV with an estimated rate of 1.5 nm per minute. Binding energies were corrected for specimen charging by referencing the C1s peak to 285 eV. 3. Results and discussion 3.1. Passivation study The implied Voc values of double side passivated substrates with (i) single layer of SiNx, (ii) single layer of SiNx(B), (iii) Al2O3/SiNx stacks and (iv) Al2O3/SiNx(B) measured before and after the thermal treatment are presented on Fig 1. The un-doped SiNx shows implied Voc of 638 mV before annealing up to
271
272
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
654 mV after annealing. Plasma and thermal ALD-Al2O3 films are known to provide an outstanding level of surface passivation on lowly doped p-type surface6 but plasma ALD films provides virtually no surface passivation in the as-deposited state17. In our case the Al2O3 layers are partially activated during the subsequent deposition of the SiNx film at 450°C. This leads to 642 mV and 659 mV implied Voc measured for the Al2O3(5 nm)/SiNx and the Al2O3(10 nm)/SiNx as-deposited stacks. Further annealing at 800°C leads to surface passivation improvement with implied Voc up to 669 mV and 679 mV for 5 and 10 nm thick Al2O3 layer, respectively. Quality enhancement could be explained by cumulative field effect passivation (scaled with the fixed charged density Qfix) and interface state passivation (Dit) of the SiNx capped-Al2O3 stack while a single layer of SiNx provides only Qfix-induced passivation. As already reported on previous studies the presence of B in the SiNx layer negatively affects the surface passivation performances when directly deposited on lowly doped Si substrate15-16. The implied Voc of the SiNx(B) layer is close to 500 mV before and after annealing. CV measurements have demonstrated a Qfix decrease with the B concentration in the SiNx(B)18. Nevertheless PC1D simulations suggested that the passivation of SiNx(B) layer is limited by a high Dit level rather than by Qfix18. The initial passivation level of SiNx(B) could be improved by Al2O3 layer deposition prior the dopant layer. The damaging effect of the SiNx(B) layer on the surface passivation is hidden by the Al2O3 layer. As a result implied Voc values increased from 550-600 mV range before annealing to 669 mV and 672 mV after annealing with respectively 5 and 10 nm of Al2O3. This demonstrated that the highly defective interface between the silicon substrate and the SiNx(B) layer could be strongly enhanced by adding an interfacial Al2O3 passivation film. Similar shadowing effect was reported in the case of SiO2(10 nm)/SiNx(B) stack19 but passivation mechanisms are different. Thermal oxide yielded a high level of chemical passivation due to a low interface defect density while plasma ALD-Al2O3 provides high quality passivation by field effect. 700
as deposited after firing
Implied Voc [mV]
650 600 550 500
x
x
0n m
)/S
iN
)/S
iN
(B
)
(1 3
2
Al
O
3
3
O 2
Al
2
Al
O
(5
(1
nm
0n m
)/S
x
iN
x
iN )/S nm (5 3
O 2
Al
(B
) (B x
N Si
x
N Si
)
450
Fig. 1. Implied Voc of various dielectric stacks deposited on polished c-Si(p) substrate (measured at 1Sun, 5 samples average)
3.2. Doping study As shown in Fig. 2, the sheet resistance after one pulse irradiation of SiNx(B)/Al2O3/c-Si(p) sample decreases continuously from 600 at 1.5 J/cm² below 100 at 5.0 J/cm². Dopants introduced in the melted silicon phase became electrically active during recrystallization and formed a p+ emitter on the ntype substrate. Boron and aluminum atoms are both donor impurities available in the dielectric stack and
273
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
thus are expected to produce hole charge carriers when substituting for Si atoms in the lattice. Sheet resistance decrease suggests diffusion of at least one type of dopants has diffused into the Si crystal. Moreover two subsequent pulses at same fluence strongly reduced the sheet resistance. At 3.1 J/cm² energy density the sheet resistance drops to 25 after two pulses while 360 was measured after single pulse irradiation. As shown in Fig. 3, ECV analysis profiles exhibited deeper depth and more surface concentrated profiles with increasing laser energy density and pulse number. This indicates enhanced diffusion and/or activation rate of the impurities regarding both parameters. 700
1E20
1 pulse 2 pulses
600
3.0 J/cm²
1E19
400
4.0 J/cm²
1 pulse 2 pulses 1 pulse 2 pulses 1 pulse 2 pulses
-3
N [at.cm ]
Rsheet [ /sq]
500
300 200
1E18
1E17
100 0 1,5
2.0 J/cm²
2,0
2,5
3,0
3,5
4,0
4,5
5,0
Fluence [J/cm²]
Fig. 2. Sheet resistance after irradiation of SiNx(B)/Al2O3 stack versus energy density for single and two pulses irradiation
1E16 0,0
0,2
0,4
0,6
0,8
1,0
Depth [μm]
Fig. 3. ECV profiles after irradiation of SiNx(B)/Al2O3 stack at 2.0, 3.0 and 4.0 J/cm² for single and two pulses irradiation
The following study focused on 2 pulses irradiation at 2.0, 3.0 and 4.0 J/cm² energy densities since the resulting sheet resistances and dopant profiles suited better for BSF application20. SIMS analysis shown on Fig. 4 was performed on the 3 samples to quantify impurity diffusion during the laser process. Constant boron and aluminium top surface concentrations (~3.1021 at/cm3) whatever the laser fluence suggest the existence of a residual dielectric layer containing Al and B atoms on top of the silicon surface. The interface position between the silicon surface and this residual dielectric layer estimated by the slope change of the Si profile (green line) is around 50 nm for the 3 samples. The profiles measured in the silicon (i.e. from 50 nm depth) highlighted simultaneous diffusion of B and Al atoms from the dielectric stack. Profiles revealed depth increase and reduced surface concentration when fluence increases but Al profiles are straight while B profiles are bell-shaped. Such B profiles were already reported in a previous study B diffusion from single SiNx(B) layer, leading the sheet resistance decrease15. By analogy a key role of the boron impurities in the doping process of the Si substrate could thus be expected. This hypothesis is straightened by the comparison of ECV and Al and B SIMS profiles. Indeed the first technique detects only electrically actives atoms while all (i.e. active and inactive) are detected by the second one. B and Al concentrations measured in the silicon (not in the residual layer) at 3.0 and 4.0 J/cm² are both excessively high at the near-surface (>2.1021 at/cm3) while ECV concentration is closed to 1020 at/cm3. That suggest only some of these dopants occupy the silicon lattice sites under electrically activated substitutional state, promoting silicon doping. As the ECV profiles are closer to the SIMS Boron profile at high fluence, it could be suspected a leading role of the boron in the doping of the p+ region in contrast to Al atoms occupying inactive interstitial states. They could form clusters particularly in the heavily doped region. Nevertheless this assumption could not be drawn at lower fluence (2.0 J/cm²) because ECV
274
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
concentration is strongly lower than those measured by SIMS. Only low activation rate of both dopant types could be concluded.
Fig.4. Boron and aluminum SIMS and ECV profiles at 2.0, 3.0 and 4.0 J/cm² (two pulses irradiation)
For further understanding the competition between both dopant types involved in the doping mechanisms, a XPS study has been performed on the 3 samples. Chemical bonding investigations were carried out at the surface (0 min etching) and after 20 and 40 min etching time. Related etching depths are 40 and 80 nm, respectively. It is assumed the residual dielectric is analysed on surface and at 40 nm etching time while 60 nm etching depth ensures silicon substrate analysis. As shown in Fig. 5a, the increasing Si-Si bonds signature at 99.5 eV after surface etching confirms the effective removal of the dielectric layer during the etching step. Contributions of Si-O seen at 104.0 eV and Si-N at 103.0 eV on 0 min and 20 min etched samples are due to SiO2 and SixNy compounds remaining on the Si surface after irradiation. Both Si-N and Si-O intensities disappeared after 40 min etching because all the residual SiO2 and SixNy compounds have been removed. The detection of chemical bonding states of B, whose concentration in Si is in the order of 1% or less is difficult because the photoionization cross section of B is small. Nevertheless the peak suspected in the B1s spectrum at 192.2 eV of un-etched samples (Fig. 5c) could be assigned to oxidized boron contribution resulting from SiNx(B)/Al2O3 stack irradiation. As seen on the Al2s spectrum plotted in Fig. 5b, the Al-O peak observed at 120.2 eV of un-etched sample reveals residual Al2O3 compounds. This peak could be correlated with the high Al concentration detected by SIMS at the surface after laser irradiation. The Al-O peak is still present after 40 min of etching time, i.e. after complete removal of the dielectric layer. An additional peak at 116.5 eV related to Al-N bonds suggests AlxNy compounds. The N1s spectra (Fig. 6b) of un-etched samples disclose Si-N signature decreasing at higher energy density. Based on O1s, Si2p and B1s spectra, it could be assumed Al and B containing Six(O,N)y residual dielectric layer remaining on the silicon surface. On the other hand a low intensity peak centred at 102.0 eV suspected on the Si2p spectra after 40 min etching could be related to Si-B bonds. The highest Si-B intensity is measured for the 4.0 J/cm² sample and could be correlated with the ECV profile where most of the B dopants are substituted for the Si atoms in the lattice. Al-O and Al-N contributions on Al2s spectra are clearly visible for the 3 fluences but the Al-O peak is more intense for the 4.0 J/cm² sample. The N1s spectra (Fig. 6b) after 40 min etching reveal high intensity Si-N peak at 2.0 J/cm² then decreasing at higher fluence. At 4.0 J/cm², the Si-N contribution is very low and a slight N-N signature appears. It could be concluded that Al atoms diffusing at high fluence associate mainly with N and O and maybe with Si. It could be assumed that the Boron atoms are introduced in the Si matrix under substitutional state and participate to the effective doping of
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
the p+ region while the Al atoms are only driven in the silicon following interstitial diffusion and forming AlxOy and AlxNy compounds. Further XPS analyses are on-going to conclude on this hypothesis.
Fig. 5. XPS peaks of Si2p (a), Al2s (b) and B1s (c) on samples 2.0, 3.0 and 4.0 J/cm² (2 pulses conditions) after 0, 20 and 40 min of etching time
Fig. 6. XPS peaks of C1s, N1s and O1s on samples 2.0, 3.0 and 4.0 J/cm² (2 pulses conditions) after 0, 20 and 40 min of etching time
275
276
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
4. Conclusion We demonstrated similar passivation quality with SiNx(B) layer as compared to un-doped SiNx when a thin Al2O3 interfacial layer is first deposited on silicon. In a second part, we demonstrated simultaneous Al and B diffusion in the Si substrate by irradiation of Al2O3/SiNx(B) stacks with an Excimer laser source. Both dopants profiles are influence by laser energy density. The formation of a p+ doped region highlighted by ECV analysis confirmed the effective doping of the Si lattice by donor impurities. The activation rate and/or the concentration of active dopant increased with laser fluence. The XPS study reveals Al atoms being rather bonded to N and O forming Al-O and Al-N compounds. It could be concluded un-active state of most of the Al atoms. As the active profile fits the boron profile at high fluence, B atoms seems to substitute locally for Si in the substrate. Nevertheless the B-Si bonds contributions suspected on B and Si XPS spectrum are too weak to strengthen this hypothesis. The quality Al2O3/SiNx(B) stacks could thus be suitable for both back side passivation of PERL solar cells and dopant source for local B-BSF formation. Boron profiles ensuring efficient Back Surface Field effect require sheet resistance between 40 and 60 p-type substrate20. Surface concentration 20 3 below 10 at/cm and 1 μm range depth profile suites well for enhanced diffusion length of the charge carriers in the doped region. Our laser process could be easily tuned to provide adapted boron profiles. The effect of Al containing compounds has to be investigated in term of electrical quality of the doping region. Acknowledgement The authors thank Nicolas Auriac for ALD deposition, Rachid Hida (CEA-LETI) for PECVD processing, Lucasz Borowik (CEA-LETI) for XPS analysis and Thomas Wolf (WEP Control) for ECV measurements. This work was supported by the General Directorate for Competitiveness, Industry and Services (DGCIS ex DGE) under convention N° 10.2.90.6027 and by the French Environment and Energy Management Agency (ADEME) under convention No. 1005C0173. References 1. Phys. Status Solidi RRL 5, No. 4, 147 149 (2011) / DOI 10.1002/pssr.201105045 2. A. Das on Screen-Printed Cells with spin38th IEEE Photovoltaic Specialist Conference, Vol57, N°10, october 2012 3. Solar Energy Materials & Solar Cells, 90 (2006) 82-92. , Phys. 4. Status Solidi RRL 5, No. 1, 22 24 (2011) 5. O. Schultz et al, Prog. Photovolt. Res. Appl. 16, 317 (2008). 6. J. Benick et al., Prog. Photovolt. Res. Appl. (2010), DOI 10.1002/pip.1020. 7. a-SiNx:H stacks for surface passivation of crystalline silicon J. Appl. Phys. 2009; 106: 114907. 8. A. -type silicon solar Proc. 25th European Photovoltaic Solar Energy Conf., Valencia, Spain, p. 1453-9.
Samuel Gall et al. / Energy Procedia 38 (2013) 270 – 277
9. -chargeAppl. Phys. Lett. 2007; 91: 112107. 10. n-type Si solar cells on Al2O3-passivated Applied Physics Letters, vol. 92, pp. 253504/1-3, 2008 11. G. Agostinelli et al p-type silicon wafers passivated with a dielectric with fixed negative charges 90 (2006) 3438-3443 12. Aluminum Alloying in Local Contact Areas on Dielectrically Passivated Rear Surfaces of Silicon Solar 37th IEEE Photovoltaic Specialist Conference, June 19-24, 2011, Seattle 13. 20th European Photovoltaic Solar Energy Conference, 6 10 June 2005, Barcelona, Spain 14. F. S. Grasso F.S. et al, Proc. 25th European Photovoltaic Solar Energy Conf, Valencia, Spain, 2010; 371 15. Laser Doping Strategies Using SiN:P and SiN:B Dielectric Layers for Profile Engineering in High Efficiency Solar Cell Energy Procedia, Volume 27, 2012, Pages 449-454 -rich, P and B 16. S. Gall, R. Hida, B. PavietProc. of 26th European Photovoltaic Solar Energy Conference, Hamburg, Germany (2011) 1648-1652. 17. J. Schmidt, B. Veith, and F. Werner, D. Zielke, R. Brendel, Silicon Surface Passivation by Ultrathin Al2O3 films and Al2O3/SiNx stacks ,35th IEEE 07/2010; DOI:10.1109/PVSC.2010.5614132 In 18. B. PavietInvestigation of charges carrier density in phosphorus and boron doped SiNx:H layers for crystalline silicon solar cells , Materials Science and Engineering: B, Volume 178, Issue 9, 15 May 2013, Pages 580-585 19. High Quality Passivation Scheme Combined with Laser Doping from SiNx(P) and SiNx(B) Layer for Silicon Solar Cell , Energy Procedia, Volume 27, 2012, Pages 467-473 20. Proc. of 26th European Photovoltaic Solar Energy Conference, Hamburg, Germany (2011) 1274 - 1279
277