Field effect of fixed negative charges on oxidized silicon induced by AlF 3 layers with fluorine deficiency

Transcription

1 Applied Surface Science 234 (2004) Field effect of fixed negative charges on oxidized silicon induced by AlF 3 layers with fluorine deficiency D. König a,*, D.R.T. Zahn a, G. Ebest b a Institut für Physik, Technische Universität Chemnitz, Chemnitz, Germany b Institut für Elektrotechnik, Technische Universität Chemnitz, Chemnitz, Germany Available online 2 July 2004 Abstract We recently discovered that in an AlF 3 /SiO 2 /Si structure extrinsic electrons are trapped at fluorine (F) vacancies in AlF 3 at the interface with SiO 2, generating a high sheet density of fixed negative charges. p- and n-type Si substrates were oxidized using rapid thermal oxidation (RTO) or furnace oxidation (th); some samples were passivated in hydrogen (H 2 ). AlF 3 was deposited onto oxidized Si wafers by a modified PVD process, leading to a F deficiency (AlF x ). Samples were characterized by mercury probe (Hg) CVand microwave photo conduction decay (mw-pcd), determining charge and trap densities and effective carrier lifetime t eff, respectively. An effective charge density of up to ¼ cm 2 is reached due to electrons tunneling from Si into AlF 3, occupying F vacancies. Lifetime scans of p-type float zone (FZ) Si samples with 1.5 nm RTO and 20 nm AlF 3 show an increase in effective minority carrier lifetime by a factor of 8.4 compared to samples with 1.5 nm RTO only. The fixed negative charge density increases with exposure time to sunlight or at simulated ageing by a 24 h anneal at 200 8C in air. # 2004 Elsevier B.V. All rights reserved. Keywords: Surface passivation; Fixed charge; Aluminium floride 1. Introduction * Corresponding author. Tel.: þ ; fax: þ address: dirk.koenig@physik.tu-chemnitz.de (D. König). The use of a positive fixed charge in an insulator semiconductor (IS) structure as an external drift field source emerged in the 1970s, allowing for an effective passivation of Si surfaces and for the manufacture of a variety of solar cells employing the field effect. The AlF 3 /SiO 2 /Si structure contains a high density of fixed negative charges and is therefore a negatively charged IS (I S) structure, presenting the antipolar counterpart. Concepts of an I S structure were proposed in [1]. In 2001, the first experimental evidence was given [2]. By preparing p-inversion layer solar cells [3] we reached a proof-of-concept level. In addition to its field effect passivation the AlF 3 layer works as an anti reflective coating (ARC). Its transparency and refractive index in the range of the air mass 1.5 solar spectrum are very close to the values of magnesium fluoride (MgF 2 ) which is a standard material for ARC on optical devices. Fig. 1 shows the composition of the I S structure. The effective charge consists of the fixed negative charge Q(AlF 3 /SiO 2 ) being located at the interface AlF 3 /SiO 2 and of the fixed positive charge Q SiO being located at the interface SiO 2 /Si, weighted by /$ see front matter # 2004 Elsevier B.V. All rights reserved. doi: /j.apsusc

2 D. König et al. / Applied Surface Science 234 (2004) Fig. 1. Layer arrangement forming the I S structure with fixed charge and trap state densities. respective static dielectric constants. The interface trap state density at interface SiO 2 /Si is denoted by D it. 2. Experimental Samples on p- and n-type Si were prepared [4,5]. Three inch C z (100) p-si and four inch Cz (0 0 1) p- and n-si substrates were used for mw-pcd measurements and all other measurements, respectively. After an RCA cleaning process most wafers received an HF dip prior to RTO which was carried out in a Heatpulse 80 system. The RTO process parameters were as follows: ambient O 2 /N 2 ¼ 2/98, plateau temperature T ¼ 900 8C, time t ¼ 30 s for d SiO ¼ 1.5 nm and O 2 /N 2 ¼ 50/50, T ¼ C, t ¼ 15 s for d SiO ¼ nm. The remaining wafers received a conventional furnace oxidation (HCl/O 2 ¼ 3/97, T ¼ C, t ¼ 160 s), resulting in d SiO ¼ 9 nm. Some samples were annealed in H 2 (T ¼ 550 8C, t ¼ 5 min) for passivation of the interface SiO 2 /Si. The AlF 3 layer was deposited by a modified PVD in a Balzers BAK 640 system with a base pressure of p ¼ mbar and a substrate temperature of T ¼ 250 8C. Samples with an AlF 3 layer thickness of d AlF ¼ 10, 20, 60 and 355 nm are presented in this work. The impact of solar irradiation, of a strong electric field, and of simulated ageing was investigated. Solar irradiation was emulated by a halogen lamp field with a radiation temperature of T ¼ 3000 K and a photon flux equivalent to 1 sun over a cumulative time of t ¼ 16 h. Corona charging was carried out with a surface charge density of Q s ¼ cm 2 (equivalent field strength: F ¼ Vcm 1 ) for t ¼ 2 h. The deposition of Q s onto the AlF 3 surface was done by a bias voltage of 10 kv with the electrode being 10 cm above the AlF 3 surface. Since the charging was done in ambient air the energy of the electrons corresponds to the thermal energy (E kt). After corona charging, the sample was rinsed in deionized water in order to remove Q s so that the sample could be characterized by Hg-CV. Simulated ageing was carried out by annealing in ambient air at T ¼ 200 8C for t ¼ 24 h. Generally, samples were stored in ambient air (waterbox) at room temperature. 3. Measurements and discussion Samples were characterized by 49-point Hg-CV wafermaps with an SSM 495 setup and mw-pcd waferscans with a Semilab WT-85 lifetime scanner, yielding values for and for t eff, respectively. For measuring, Hg-CV measurements were carried out in the high frequeny regime (f ¼ 1 MHz). The average dot size of the mercury droplet was 2.2 mm 2. D it was measured by single Hg-CV probes with the capacitance difference method. Therefore, additional measurements in the low frequency regime (f ¼ 10 khz) were carried out. Fig. 2 shows the time evolution of for different values of d AlF with a constant oxide thickness of d SiO ¼ 1.5 nm. Fig. 3 shows the time evolution of for different values of d SiO with an AlF 3 thickness of d AlF ¼ 10 nm. For thick AlF 3 layers the standard deviation is rather large as Q(AlF 3 /SiO 2 ) is split up into a large fraction at the interface AlF 3 /SiO 2 and a small fraction at the (cm -2 ) nm / 10 nm 1.5 nm / 20 nm 1.5 nm / 60 nm 1.5 nm / 355 nm time (weeks) Fig. 2. as function of d AlF and t. The error bars show the standard deviation. The legend shows d SiO /d AlF. Lines are guides to the eye.

3 224 D. König et al. / Applied Surface Science 234 (2004) nm / 10 nm 3.6 nm / 10 nm 9.0 nm / 10 nm (cm -2 ) time (weeks) Fig. 3. as function of d SiO and t. The error bars show the standard deviation The legend shows d SiO /d AlF. Lines are guides to the eye. D it (cm -2 ev -1 ) /1: 9 nm / -- 42/7: 9 nm / --, H /6: 9 nm / 20 nm 42/8: 9 nm / 20 nm, H E - E V (ev) Fig. 4. D it as function of energy within the band gap of Si. The legend shows: d SiO /d AlF,H 2 passivation. AlF 3 surface (see macroscopic capacitor model described as follows). At both regions there is a significant F deficiency, leading to a high density of F vacancies [6]. The negative fixed charge Q(AlF 3 / SiO 2 ) is presented by F vacancies being occupied with extrinsic electrons originating from Si [5,6]. Average values of ¼ cm 2 are feasible; peak values are in the range of ¼ cm 2. The interface trap state density D it of the interface SiO 2 /Si is shown in Fig. 4. For samples with AlF 3 layer the decrease of D it by H 2 passivation is smaller when compared to samples with thermal SiO 2 only. Samples with AlF 3 have a midgap trap state density of D it ¼ cm 2 ev 1 ; it decreases to D it ¼ cm 2 ev 1 if a H 2 passivation is carried out. The measurement of t eff by mw-pcd wafer scans of samples based on FZ p-si revealed that t eff is increased by the deposition of AlF 3. Fig. 5 shows a typical waferscan of t eff of the sample passivated in H 2 with d SiO ¼ 1.5 nm and d AlF ¼ 20 nm. At the wafer edge t eff drops sharply as there the AlF 3 layer ends abruptly due to screening by the sample holder, underlining the passivating impact of the deposited AlF 3 layer. For samples with RTO, d SiO ¼ 1.5 nm and H 2 passivation the average effective minority carrier lifetime t eff increased from t eff ¼ 11 mstot eff ¼ 94 ms (effective surface recombination velocity S eff ¼ 173 cm s 1 ) by depositing an AlF 3 layer with d AlF ¼ 20 nm. For samples with d SiO ¼ 9 nm, d AlF ¼ 60 nm t eff ¼ 127 ms (effective surface recombination velocity S eff ¼ 120 cm s 1 ) was measured. Table 1 shows average values for t eff with standard deviation and calculated values for S eff of different samples. Table 1 Average effective minority carrier lifetime t eff, standard deviation s(t eff ) and calculated effective surface recombination velocity S eff from waferscans of samples with FZ p-si substrate Batch/sample d SiO, d AlF (nm) t eff (ms) s(t eff ) (%) S eff (cm s 1 ) Remarks 42/1 9.0 th, /2 9.0 th, /3 9.0 th, H 2 passivated 38/6 9.0 th, / RTO, / RTO, H 2 passivated 42/ RTO, / RTO, H 2 passivated 38/ RTO, RTO stands for rapid thermal oxide; th stands for furnace oxide.

4 D. König et al. / Applied Surface Science 234 (2004) Fig. 5. Waferscan of t eff of sample 42/5 (see Table 1)with d SiO ¼ 1.5 nm, passivated in H 2, d AlF ¼ 20 nm. At wafer edge t eff drops sharply as there the AlF 3 layer ends due to screening by sample holder. Limits at colour scale present lowest/highest 10% of t eff in ms. Exposing a sample with d SiO ¼ 1.5 nm, d AlF ¼ 10 nm to solar irradiation increased from ¼ cm 2 to ¼ cm 2 within t ¼ 8 h. This behaviour is in contrast to degradation of conventional IS structures under solar irradiation. Corona charging did not have any influence on fixed charge densities and interface states, i.e. and D it are constant. Simulated ageing of a sample with d SiO ¼ 9 nm and d AlF ¼ 120 nm left D it constant while increasing by a factor of 3.8 to ¼ cm 2. As a result the I S structure is stable over time (see Figs. 2 and 3) and very robust against environmental influences. The negative fixed charge Q(AlF 3 /SiO 2 ) is formed by F vacancies at the interface AlF 3 /SiO 2 and further at the AlF 3 surface being occupied with extrinsic electrons originating from Si [2,4 6]. The deposited AlF 3 layers possess a very low density of extrinsic defects like alkali ions or hydroxide ions [9] so that a localization of extrinsic electrons can only occur in F vacancies. This statement is also supported by density x(alf x ) 3 Q(d) 0 _ AlF 3 SiO 2 Si _ ++ + AlF 3 SiO 2 x(alf x ) 3 Si Q(d) Fig. 6. Macroscopic capacitor model explaining the impact of fixed charges on Si.

5 226 D. König et al. / Applied Surface Science 234 (2004) d SiO -5.0x10 12 Q(AlF 3 /SiO 2 ) -4.0x x10 12 (cm -2 ) (1.5nm) (9.0nm) Fit [ (1.5nm)] Fit [ (9.0nm)] d AlF (nm) Fit[ Q(AlF 3 /SiO 2 )] d SiO =1.5 nm -2.0x d AlF (nm) Q(AlF 3 /SiO 2 ) (cm -2 ) Fig. 7. (Left) Measured and empirical fittings as function of d AlF for samples with d SiO ¼ 1.5 nm RTO and d SiO ¼ 9 nm thermal SiO 2. Error bars indicate the standard deviation. (Right) Calculated Q(AlF 3 /SiO 2 ) and empirical fitting as function of d AlF for samples with d SiO ¼ 1.5 nm RTO. functional computations [5,10]. There is a high ohmic conduction path between the two AlF x regions allowing for a charge transfer by hopping conduction [7,8]. With decreasing d AlF3 both AlF x regions merge [4 6]; for d AlF 20 nm Q(AlF 3 /SiO 2 ) is assumed to be located at the interface AlF 3 /SiO 2. Assuming Q SiO being located at the interface SiO 2 /Si we can calculate Q(AlF 3 /SiO 2 ) by: jqðalf 3 =SiO 2 Þ e SiO ealf3 jq eff jq SiO (1) e eff e SiO2 The e denote static dielectric constants of respective material whereby e eff stands for the entire insulation layer stack in analogy to. As a result we get Q(AlF 3 /SiO 2 ) ¼ 2to cm 2 depending on considered sample. Fig. 6 shows the macroscopic capacitor model explaining the impact of fixed charges on Si. In accordance with the macroscopic capacitor model, a hyperbolic fit can be carried out on the measured effective charge densities for samples with an RTO layer thickness of d SiO ¼ 1.5 nm and for samples with a thermal oxide layer thickness of d SiO ¼ 9.0 nm. With Eq. (1) Q(AlF 3 /SiO 2 ) can be calculated from experimental values of. For samples with an RTO layer thickness of d SiO ¼ 1.5 nm a hyperbolic fit yields: jqðalf 3 =SiO 2 ; d SiO2 ¼ 1:5nmÞ ¼ cm 2 10 nm þ d AlF3 þ 8:5nm ð 3: cm 2 Þ (2) Fig. 7 shows the average values of over the thickness of the AlF 3 layer d AlF with respective fitting functions and calculated values of Q(AlF 3 /SiO 2 ) with the fitting function given in Eq. (2). Thereby values of and Q(AlF 3 /SiO 2 ) as a function of d AlF and d SiO can be predicted. 4. Conclusion The I S structure yields average effective fixed charge densities of ¼ cm 2, and peak values are in the range of ¼ cm 2. The I S structure is very robust against environmental influences. Corona charging (surface charge density Q s ¼ cm 2, t ¼ 2 h) did not show any effect upon. Exposing the I S structure to solar irradiation increased by 40% to ¼ cm 2. A simulated ageing (annealing in ambient air at T ¼ 200 8C for t ¼ 24 h) resulted in an increase

6 D. König et al. / Applied Surface Science 234 (2004) of while the trap state density D it at the interface SiO 2 /Si remained constant. The high density of fixed negative charges of the I S structure induces an efficient field effect passivation on Si, resulting in a p-inversion layer and very low surface recombination velocities. Additionally, the AlF 3 layer can be used as ARC. In conclusion, the I S structure is valuable for increasing solar cell efficiency [3,5]. Acknowledgements One of the authors (D.K.) wishes to thank the the Stifterverband zur Förderung der Deutschen Wissenschaften for financial support and the Deutsche Forschungsgemeinschaft (DFG) for a Ph.D. grant. References [1] D. König, G. Ebest, Solid State Electr. 44 (2000) 111. [2] D. König, M. Rennau, M. Henker, G. Ebest, Thin Solid Films 385 (2001) 126. [3] D. König, D.R.T. Zahn, R. Reich, K. Gottfried, G. Ebest, Poster 1P-D3-21, in: The Third WCPEC, May 11 18, 2003, Osaka, Japan. [4] D. König, G. Ebest, Solar Energy Mater. Solar Cells 75 (2003) 335. [5] D. König, Ph.D. Thesis, Institut für Physik, Technische Universität Chemnitz, Germany, [6] D. König, R. Scholz, S. Gemming, I. Thurzo, T.U. Kampen, D.R.T. Zahn, G. Ebest, Physica E 14 (2002) 259. [7] Y. Danto, J. Salardenne, Vacuum 27 (1977) 293. [8] A.M. Phale, Thin Solid Films 46 (1977) 315. [9] J. Heber, private communication. [10] D. König, R. Scholz, S. Gemming, D.R.T. Zahn, G. Ebest, submitted to Phys. Rev. B, June 2004.