Passivation and depassivation of silicon dangling bonds at the Si_SiO2 interface by atomic hydrogen

E. Cartier, J. H. Stathis, and D. A. Buchanan

Citation: Applied Physics Letters 63, 1510 (1993); doi: 10.1063/1.110758 View online: http://dx.doi.org/10.1063/1.110758

View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/63/11?ver=pdfcov Published by the AIP Publishing

Articles you may be interested in

Spurious dangling bond formation during atomically precise hydrogen depassivation lithography onSi(100): The role of liberated hydrogen

J. Vac. Sci. Technol. B 32, 021805 (2014); 10.1116/1.4864302

Hotelectron induced passivation of silicon dangling bonds at the Si(111)/SiO2 interface Appl. Phys. Lett. 69, 103 (1996); 10.1063/1.118088 Hydrogen bonding arrangements at Si–SiO2 interfaces J. Vac. Sci. Technol. B 13, 1613 (1995); 10.1116/1.587866

Dissociation kinetics of hydrogenpassivated (100) Si/SiO2 interface defects J. Appl. Phys. 77, 6205 (1995); 10.1063/1.359148

Passivation of paramagnetic SiSiO2 interface states with molecular hydrogen Appl. Phys. Lett. 53, 508 (1988); 10.1063/1.100620

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

162.105.246.102 On: Sun, 07 Sep 2014 03:41:56

Passivation and depassivation interface by atomic hydrogen of silicon dangling bonds at the Si/Si02 IBM Research Division, Thomas J. Watson Research Center, Yorktown Heights, Nelir York 10598 E. Cartier, J. H. Stathis, and D. A. Buchanan (Received 27 May 1993; accepted for publication 6 July 1993) Atomic hydrogen is found to simultaneously passivate and depassivate silicon dangling bonds at the Si( 11 l)/SiOz interface at room temperature via the reactions P6+Ho-P&I and P&I + Ho -+ Pb-!- Hz. The passivation reaction occurs more efficiently keeping the steady-state Pb density at a low value of only 3-6X 10” cmM2 during atomic hydrogen exposure. This low Pb density can only account for a small fraction of the total number of interface states produced by atomic hydrogen. For decades, hydrogen has been known to play a cru- cial role in the fabrication of high quality Si/SiO, inter- faces in metal-oxide-semiconductor (MOS) electronic de- vices. One important Si/SiO, interface charge trap, called the Pb center, was identified microscopically more than two decades ago using electron paramagnetic resonance (EPR)?* This defect is characterized by an unpaired elec- tron localized in a sp3-like silicon dangling bond at the interface.3 It has been known for a long time that this Pb center can be passivated by annealing in molecular hydro- gen via the reaction &,+HpW+H. (1) Here, P&I represents the hydrogen-passivated dangling bond. Recently, a detailed kinetic study of the interaction of molecular hydrogen with the Pb center at the Si( 11 l)/ Si02 interface was reported.4 Using EPR, Brewer and My- ers showed that reaction (1) has an activation energy of 1.66 eV. Furthermore, they argued from theoretical con- siderations that Pb passivation by atomic hydrogen, He, according to the reaction Ph+@*P&, as well as Pb depassivation according to the reaction P$+H”-&+H2 (3) should occur spontaneously, with essentially no energy barrier, if a hydrogen atom encounters a Pb center or a P,J3 center, respectively. If the energy barriers for each of these reactions were exactly zero, one would expect that the steady-state concentration of Pb centers during atomic hy- drogen exposure-where reactions (2) and (3 ) occur with equal probability-would reach half the value of the intrin- sic dangling-bond density at the Si( 1 11)/Si02 interface. Based on a detailed study of the intrinsic Pb density by Stesmans,’ one would thus expect to be able to produce about 5 x lOI cm-‘-P, centers by atomic hydrogen expo- sure. DoThanh and Balk6 were the first to show that atomic hydrogen exposure at room temperature indeed produces interface states, but no microscopic identification was attempted. In this letter we show experimentally that reactions (2) and (3) do indeed occur simultaneously at room tem- perature. However, we find that atomic hydrogen prefer- entially passivates dangling bonds leading to a maximum 1510 steady-state Pb density of only 3-6 X 10” cme2, indicating that both reactions may be thermally activated. Further- more, we find that this steady-state Pb density is about an order of magnitude smaller than the total number of elec- trically active interface states produced by atomic hydro- gen. A remote microwave hydrogen-plasma system similar to the one described by Johnson’ was used for atomic hy- drogen exposure. The system operates at a pressure of 250 mTorr using 99.9995% pure hydrogen at a microwave power of 100 mW. In our downstream configuration the sample chamber is separated from the discharge region by a baffle system to eliminate all light and charged particles, and by a metallic variable-length recombination tube which allows accurate variations of the hydrogen concen- tration in the sample chamber. The absolute atomic hydro- gen concentration is monitored bolometrically, via the tem- perature increase of an Ag-coated thermocouple caused by hydrogen surface recombination’ H+H-+H2+4.5 eV. (4) (2) We are able to maintain Ho concentrations of up to 2% about 50 cm downstream from the discharge region over periods of hours. For the experiments discussed below, the concentration was kept below 0.1% ‘to avoid direct sam- pling heating via reaction (4). Exposure times ranged from 1 to 200 min. In all experiments, gate-free oxides were exposed and a mercury probe was used to measure the interface state density after exposure using the high-low- frequency capacitance (HLFC) method. In a few cases, aluminum gates were evaporated after exposure to verify the reliability of the mercury probe method, yielding good agreement for the two procedures. EPR measurements were performed at room tempera- ture at -9.5 GHz, using a Bruker ER-200 spectrometer with a rectangular cavity. The absolute density of para- magnetic defects was obtained by double numerical inte- gration of the experimental first-derivative spectra and comparison with a NBS-calibrated ruby standard, taking into account differences in tilling factor and spin. The mag- netic field modulation amplitude was 2.2 G peak-to-peak, resulting in slight overmodulation of the signal in order to obtain a better signal-to-noise ratio. @ 1993 American Institute of Physics 1510 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

162.105.246.102 On: Sun, 07 Sep 2014 03:41:56

Appl. Phys. Lett. 63 (1 l), 13 September 1993 0003-6951/93/63(11)/1510/3/$6.00 10’3 &- ‘E 5% oap 10’2 m a ti ‘5 k 10” 0s ? 45 2.004 2.002 2.000 1.998 / ,I , I 10'0 ’ 100 11 0-l 10’ 102 Exposure time (minutes) I _I I I 3414 3418 3422 -3426 FIG. 1. Comparison of the measured interface state density at midgap with the measured P,, density at the Si/SiO, interface as a function of exposure time to atomic hydrogen. The initial interface state density in the as-grown sample (49.5nm-thick oxide) was @,=.5X 10” cm -‘eV-‘. No P,, resonance could be observed in the EPR samples (9.75-nm-thick oxide) before atomic hydrogen exposure. Only a small fraction of the atomic hydrogen-induced interface states are Pn centers. Magnetic Field (G) FIG. 2. EPR spectra of as-grown, vacuum annealed, and atomic hydro- gen exposed 9.75-nm-thick SiOz filmson Si( 111). The magnetic field is parallel to the [ll l] axis. After atomic hydrogen exposure, the density in the as-grown oxide increases to 3.0*0.5X 10” cmW2 while the density in the vacuum annealed sample decreases to 4.5 f 0.5 X 10” cm -‘. Passiva- tion and depassivation of silicon dangling bonds by atomic hydrogen [reactions (2) and (3)] occur simultaneously at room temperature. Si( 111) wafers with a resistivity of 20-60 R cm (n type) for electrical measurements and double polished, nominally undoped Si ( 111) wafers for EPR measurements were oxidized in dry oxygen at 900 “C. The oxide thickness was 49.5 nm for the HLFC samples and 9.75 nm for the EPR samples. The EPR samples were scribed into 3 mmX2 cm pieces. The principal results of our atomic hydrogen exposure experiments at room temperature are summarized in Figs. 1 and 2. In Fig. 1, the buildup of the interface state density at midgap, AD,,, is compared to the buildup of the Pb concentration during repetitive hydrogen exposures. The data are plotted as a function of total exposure time. As can be seen, ADit increases linearly with hydrogen dose at least up to a density of 1013 cm-* eV-’ with no indication of saturation. The measured interface state distribution across the silicon band gap shows that a broad peak, roughly 0.2 eV above midgap, continuously grows with Ho dose. (The general evolution of the interface state speo trum is found to be very similar to its evolution observed during high-field current stress in similar structures.‘) A figure of merit for the total interface state density can be obtained from the product Dit X Eg, where Eg= 1.1 eV is the width of the silicon band gap. Since the maximum possible’ number of Pb centers is of the order of lOi cme2, one might expect that the HO-induced interface states are essentially all P,, centers. Surprisingly, we find that the Pb density, shown by the full circles in Fig. 1, increases at a roughly ten times smaller rate than the density of electri- cally active sites. This difference is not due to the difference in substrate resistivity or oxide thickness. Measurements of the interface state density on the same samples used for EPR at the end of the exposure sequence of Fig. 1 yields a value of at least lOI cmm2 eV-’ at midgap, consistent with the measurements on the thicker, lower resistivity HLFC samples. Furthermore, in contrast to the ADit values, the Pb density clearly saturates at a very low value of only 3 X IO” cmw2. This is consistent with recent spin- dependent recombination measurements on electrically stressed transistors,” which showed that the Pb center can- not account for all of the defects produced by high-field stress. This result directly demonstrates that the depassi- vation reaction (3) occurs at room temperature but also suggests that the passivation reaction (2) might efficiently compete with the depassivation reaction. To prove that the low saturation value of 3X 10” cm-’ for the Pb density is actually the result of the simul- taneous passivation and depassivation of silicon dangling bonds, we compared the steady-state Pb density after long Ho exposures of as-grown and of vacuum annealed sam- ples. As can be seen from Fig. 2, a high Pb density (2.4 X 1012 cmw2> is observed after vacuum annealing under conditions similar to those used by Stesmans,s while no Pb resonance could be observed in the as-grown oxide. After a 120 min Ho exposure, the Pb density in the vacuum an- nealed sample decreased to a value of 4.5&0.5X 10” cmw2, a density which is very close to the saturation value measured after Hc exposure of the as-grown sample (see Fig. 2). Clearly, both reactions ( 1) and (2) occur simul- taneously at room temperature. This result strongly sug- gests that either or both reactions may be thermally acti- vated, the reaction barrier for passivation being smaller than that for depassivation. We have some evidence that the rate limiting step during the buildup in Fig. 1 is actu- This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:1511 Appl. Phys. Lett., Vol. 63, No. 11, 13 September 1993 Cartier, Stathis, and Buchanan 1511 162.105.246.102 On: Sun, 07 Sep 2014 03:41:56

ally the reaction at the interface itself and that we are not source or diffusion limited. For example, we find only a weak oxide thickness dependence for the interface state buildup rate and the Pb density in steady state is indepen- dent of the He density over at least two orders of magni- tude. A detailed study of the reaction kinetics of atomic hydrogen with the Pb center might thus become possible. Our results provide new insight into long standing is- sues concerning the degradation of MOS device structures during high-field stress or in a radiation environment. Many experiments on device degradation strongly suggest that atomic hydrogen-which is incorporated in as- fabricated device structures in abundance--can be released by ionizing radiation or by hot electrons. ‘*-I3 We have now demonstrated that interface state formation will be inevi- table under these conditions and that atomic hydrogen can account for much of the observed degradation of MOS devices. Cartier and DiMaria14 have recently incorporated such a hot-electron-induced Ho-release process into a Monte Carlo electron transport formalism and demon- strated that hot-electron-induced hydrogen release can ac- count for the interface state formation during current stress over a wide electric field and oxide thickness range. In conclusion, we have demonstrated that atomic hy- drogen at the Si/SiO, interface both passivates and depas- sivates silicon dangling bonds according to the reactions (2) and (3 ) . However, dangling-bond passivation is found to dominate at room temperature, such that the Pb center can only account for a small fraction of the total number of electrically active interface defects produced by atomic hy- drogen. Our experiments strongly support MOS device degradation models which involve hot-electron and/or radiation-induced hydrogen release during high-field cur- rent stress or irradiation. Finally, our experiments strongly suggest that the Pb center is not the dominant charge trap at the Si/SiOZ interface generated by atomic hydrogen. This appears to be in contrast with previous studies15 which showed that the number of interface states produced by radiation and the number of Pb centers differ by a factor of 2 or less. The authors would like to thank D. J. DiMaria and F. R. McFeely for many useful discussions, and A. H. Par- sons for oxide growth. ‘Y. Nishi, Jpn. J. Appl. Phys. 10, 52 (1971). *P. J. Caplan, E. H. Poindexter, B. E. Deal, and R. R. Razouk, J. Appl. Phys. 50, 5847 (1979). “K. L. Brower, Appl. Phys. Lett. 43, 1111 (1983). 4K. L. Brower and S. M. Meyers, Appl. Phys. Lett. 57, 162 (1990). ‘A. Stesmans and G. VanGorp, Appl. Phys. Lett. 57, 2663 ( 1990). ‘L. DoThanh and P. Balk, J. Electrochem. Sot. 135, 1797 ( 1988). ‘N. M. Johnson, in Hydrogen in Semiconductors, edited by J. I. Pankove and N. M. Johnson (Academic, San Diego, 1991), p. 113. *K. Tdnkala and T. DebRoy, J. Appl. Phys. 72, 712 (1992). 9D. A. Buchanan and D. J. DiMaria, J. Appl. Phys. 67, 7439 (1990). “J. H. Stathis and D. J. DiMaria, Appl. Phys. Lett. 61, 2887 (1992). “R. Gale, F. J. Feigl, C. W. Magee, and D. R. Young, J. Appl. Phys. 54, 6938 (1983). ‘*M. A. Briere and D. Braunig, IEEE Trans. Nucl. Sci. 37, 1658 (1990). “D. Buchanan, A. Marwick, D. DiMaria, and L. Dori, in Proceedings of the Second Symposium on the Physics and Chemistry of SiO, and Si-Si02 Interface, edited by B. Deal and R. Helms (The Electrochemical Soci- ety, Pennington, NJ, 1992) (in press). 14E. Cattier and D. J. DiMaria, in Insulating Films on Semiconductors edited by P. Balk (North-Holland, Amsterdam, 1993) ‘(in press). lsP. M. Lenahan and P. V. Dressendorfer, J. Appl. Phys. 54, 1457 (1983). This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

162.105.246.102 On: Sun, 07 Sep 2014 03:41:56

1512 Appl. Phys. Lett., Vol. 63, No. 11, 13 September 1993 Cartier, Stathis, and Buchanan 1512