APPLIED PHYSICS LETTERS
VOLUME 81, NUMBER 10
2 SEPTEMBER 2002
Dual behavior of H¿ at Si–SiO2 interfaces: Mobility versus trapping S. N. Rashkeeva) Department of Physics and Astronomy, Vanderbilt University, Nashville, Tennessee 37235
D. M. Fleetwood Department of Electrical Engineering and Computer Science and Department of Physics and Astronomy, Vanderbilt University, Nashville, Tennessee 37235
R. D. Schrimpf Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, Tennessee 37235
S. T. Pantelides Department of Physics and Astronomy, Vanderbilt University, Nashville, Tennessee 37235 and Solid State Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831
共Received 8 May 2002; accepted for publication 10 July 2002兲 We report first-principles calculations showing that protons in the vicinity of a Si–SiO2 interface can behave in two different ways. At an abrupt interface without suboxide bonds 共Si–Si bonds at the oxide side of the interface兲 H⫹ does not become trapped but migrates laterally until it reacts with a point defect 共e.g., depassivates a hydrogenated dangling bond兲. On the other hand, when large concentrations of suboxide bonds are present, H⫹ can become trapped in a deep energy minimum with a highly asymmetric energy barrier. Thus, large H⫹ densities first saturate suboxide bonds, and the balance can be cycled back and forth between a pair of interfaces by reversing the electric field. These results account for the experimentally observed dual behavior of protons at Si–SiO2 interfaces. © 2002 American Institute of Physics. 关DOI: 10.1063/1.1504879兴 As a consequence of device processing, hydrogen is present in abundant quantities in bound forms in the oxide, the polysilicon gate, and the metal interconnects of metal– oxide–semiconductor 共MOS兲 devices. When a MOS device is irradiated, hydrogen is released and transports to the Si–SiO2 interface where it can induce interface defects.1 Hydrogen is released and migrates through oxide as a proton because H0 is not a stable charge state in Si,2 SiO2 , 3,4 and at Si–SiO2 interface.5 Before and after irradiation, hydrogen can react with the existing Si–SiO2 interfacial dangling bonds, both passivating and depassivating them.6,7 More recently, another phenomenon has been observed: under specific conditions 共high-temperature inert anneal that introduces high densities of oxygen vacancies near the interface, plus very high exposure levels of hydrogen兲, H⫹ ions are created in the oxide that can transport rapidly across the SiO2 under bias but do not react with the interface. Instead, they can be cycled back and forth between opposing Si–SiO2 interfaces by simply reversing the electric field.8 Thus, under certain conditions protons in SiO2 interact with the interface and create defects. In other cases, the protons do not react with the interface, and by reversing the electric field practically all of them can be moved from the interfacial region back into the oxide. Vanheusden et al.8 and others9–11 have shown that one can observe up to 104 cycles or more in oxides of different thickness, i.e., the loss of protons at the interfaces is negligible. In this letter we present first-principles calculations in terms of which we characterize the behavior of H⫹ at Si–SiO2 interfaces with different local coordination. We find a兲
Author to whom correspondence should be addressed; electronic mail:
[email protected]
that a suboxide Si–Si bond at the interface 共a Si–Si bond on the SiO2 side of the interface兲 provides a deep energy minimum with a highly asymmetric barrier where the proton can be captured. This behavior is different from the case of an abrupt interface without suboxide bonds where the proton can easily move laterally at the interface because the activation barrier for such a motion is extremely low 共0.3–0.5 eV兲. These results can account for the observed dual behavior of protons: normal Si–SiO2 interfaces in MOS devices do not contain too many suboxide bonds where protons become trapped. In this case H⫹ migrates along the interface and reacts with passivated interfacial dangling bonds creating additional defects.5 In contrast, in samples with high densities of interfacial suboxide bonds and O vacancies, a flood of protons become captured, and an interfacial layer of them forms an ‘‘electrostatic fence,’’ enabling us to observe cycling. The present calculations were based on density functional theory, the local-density approximation for exchange correlation, ultrasoft pseudopotentials, supercells, and plane waves.12 The ultrasoft pseudopotentials13 and the VASP codes14 were used. These pseudopotentials have been thoroughly tested in earlier work on a variety of Si–O–H systems.15–17 The energy cutoff for the basis set was 24 Ry, and integrations over the Brillouin zone were done using the Monkhorst–Pack scheme with two k points in the relevant irreducible wedge.18 For studying charged defects we introduced a homogeneous negative 共positive兲 background when removing 共adding兲 electrons in the supercell. The calculations were performed for supercells containing a model interface, as recently reported by Buczko et al.,15 namely Si–SiO2 – Si ‘‘superstructures’’ containing 7– 8 planes of Si layers separating the SiO2 layers. A large
0003-6951/2002/81(10)/1839/3/$19.00 1839 © 2002 American Institute of Physics Downloaded 29 Oct 2002 to 129.59.117.126. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
1840
Appl. Phys. Lett., Vol. 81, No. 10, 2 September 2002
FIG. 1. A schematic of the H⫹ transport: 共a兲 in an irradiated MOS device and 共b兲 in a hydrogen-anneal experiment in a Si–SiO2 – Si structure with high O vacancy density.
number of abrupt interfaces with distinctly different local bonding arrangements were examined in Ref. 15. In addition, one can construct interfaces with Si–Si suboxide bonds and interfaces with SiO2 -like protrusions into the Si side.19 Before introducing hydrogen into the system, all the supercells were relaxed until the total energy was minimized 共the force on each atom is smaller than a tolerance, in this work, 0.1 eV/Å兲. The properties of a suboxide bond are very different from the properties of O vacancies in bulk crystalline SiO2 . Once the O vacancy is created in crystalline SiO2 , it reestablishes a Si–Si bond only in the neutral state. An O vacancy can easily capture a hole and become positively charged. In this case it undergoes substantial rebonding; one of the Si atoms pushes back through the plane of its three back-bonded neighbors and bonds weakly with one of the O atoms in that vicinity.20 The other Si simply has a dangling bond that holds one electron and gives rise to the wellknown E ⬘ electron-paramagnetic-resonance signal.21,22 Recently, Lu et al.23 found that in amorphous bulk SiO2 , vacancies at the vast majority of O sites (⬃90%) do not undergo this kind of reconstruction, but simply form Si–Si dimers in both the neutral and positively charged state. We carried out extensive checks near the Si–SiO2 interface and found that suboxide Si–Si bonds do not undergo any major reconstruction either. First, let us consider a behavior of a proton at an intrinsic Si–SiO2 interface as shown schematically in Fig. 1. A key question is whether a proton that arrives at the interfacial region can capture an electron from the Si side 共by tunneling兲 and become neutralized. That would certainly be in disagreement with the reversibility of proton transport in the hydrogen annealing experiment.8 Our recent calculations establish that H⫹ is the only stable charge state at intrinsic Si–SiO2 interfaces,5 consistent with the requirement that a proton must not be neutralized in the hydrogen anneal case in Fig. 1. To get this result we examined a wide range of different possible sites for a proton near the interface 共in the Si–Si bonds of the first Si layer on the Si side or attached to inequivalent O atoms on the first SiO2 layer of an abrupt interface, in suboxide bonds, and in various configurations in small SiO2 protrusions into the Si side兲. The results were also independent of the local interfacial bondings 共cristobalite like, quartz like, or tridymite like兲.15 This result is related to the fact that the empty localized energy level associated with H⫹ is always much higher than the Si conduction band edge. When an H⫹ arrives at the abrupt intrinsic Si–SiO2 interface, it does not react at first and cannot become trapped. The most energetically favorable case is for it to migrate
Rashkeev et al.
laterally jumping from one interfacial Si–Si bond 共on the Si side, in this position the proton has minimal potential energy兲 to another 关Fig. 2共a兲兴. The migration barriers for this motion range between 0.3 and 0.5 eV 共the barrier depends on the migration path兲. The barriers to escape either to the Si or the SiO2 side are larger 关Fig. 3共a兲; see also Ref. 19兴. However, the estimated values of these barriers 共0.8 eV for the Si side and 1.0 eV for the SiO2 side兲 still suggest that the proton can move into both bulk materials, especially at high temperatures. Thus the potential energy surface for a proton at an intrinsic abrupt Si–SiO2 interface represents a collection of ¯ 0 兴 direction 共the slightly warped grooves oriented in the 关 11 direction of the interfacial Si–Si–Si . . . chains at the abrupt Si–SiO2 关100兴 interface兲. A proton that comes to one of the grooves at the interface will travel along it before it finds a defect that can trap it 共e.g., a suboxide bond, see below兲, or reacts with another defect 共e.g., a Si–H bond; see Ref. 5兲. Since two H⫹ arriving at a similar location along the interface would repel each other electrostatically, any complexing between them is very unlikely. The situation dramatically changes when a suboxide bond is created at the interface 关Fig. 2共b兲兴. This defect represents a strained Si–Si bond that is located on the SiO2 side of the interface. The proton that is captured by this bond in the oxide side of the interface behaves very differently from the proton captured by a regular Si–Si bond on the Si side. Calculations reveal that H⫹ at a suboxide bond is in a deep local energy minimum, with an asymmetric barrier, as shown in Fig. 3共b兲. The barrier is 1.5 eV on the Si side. On the SiO2 side, the barrier is 1 eV for transfer to the nearest site, followed by another barrier of 0.8 eV to escape to far distances from the interface. In addition, the proton is trapped, and cannot travel along the interface as in the case of an abrupt interface. It needs to overcome a 1.5 eV energy barrier in order to come to the nearest Si–Si bond on the Si side. On the proton potential energy surface one gets a deep energy minimum with a very anisotropic barrier. This asymmetric barrier is a key result for understanding the behavior of protons arriving at the Si–SiO2 interface that can comprehensively account for the seemingly contradictory behavior of protons in the two different experiments 共radiation and annealing兲 illustrated in Fig. 1. In radiation experiments, a fraction of suboxide bonds capture H⫹ and become positively charged interface defects. Other H⫹ ions react with Si–H bonds, creating interface traps, as discussed in detail in Ref. 5. In hydrogen annealing experiments,9–11 a high temperature processing introduces excess of vacancies near the interface.24 This oxygen deficiency very near the interface results in a very high density of suboxide bonds, at much higher concentrations than is typical for MOS devices. In these experiments, the interface is flooded with H⫹ . The excess suboxide bonds will stop them because the barrier is higher on the Si side. Reversal of the external electric field can then drive the H⫹ back into the SiO2 . A layer of H⫹ at suboxide bonds right at the interface may in fact remain stuck and provide an ‘‘electrostatic fence’’ to further arrest the arriving protons when the external electric field is pushing them toward the interface. Thus, the H⫹ flux can be cycled back and forth. The O deficiency near the interface may in fact corre-
Downloaded 29 Oct 2002 to 129.59.117.126. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
Rashkeev et al.
Appl. Phys. Lett., Vol. 81, No. 10, 2 September 2002
1841
FIG. 3. A schematic of the proton potential energy across the Si–SiO2 interface 共shown by the dashed line兲: 共a兲 for an abrupt intrinsic interface and 共b兲 for an interface with a Si–Si suboxide bond.
FIG. 2. Models for the Si–SiO2 interface with different local arrangements. The position of the interface is shown by the dashed line; Si atoms are shown in gray, O in black: 共a兲 abrupt interface, the arrow shows the most probable migration path for a proton in the interfacial region; 共b兲 interface with suboxide bond 共the Si–Si bond at the SiO2 side of the interface is indicated by the arrow兲.
rier favoring release into the oxide, as opposed to motion along the interface and subsequent reaction to form interface traps. This mechanism of proton trapping can account for the dual behavior of protons in radiation effects and hydrogen annealing experiments. This work was supported in part by AFOSR Grant No. F-49620-99-1-0289, the U.S. Navy, the Defense Threat Reduction Agency under Contract No. DTRA01-00-C-0010, and by the William A. and Nancy F. McMinn Endowment at Vanderbilt University. Research at ORNL was sponsored by the Division of Materials Sciences, U.S. Department of Energy, under Contract No. DE-AC05-00OR22725 managed by UT-Battelle, LLC. D. B. Brown and N. S. Saks, J. Appl. Phys. 70, 3734 共1991兲. C. G. Van de Walle, P. J. H. Denteneer, Y. Bar-Yam, and S. T. Pantelides, Phys. Rev. B 39, 10 791 共1989兲. 3 P. E. Bunson, M. Di Ventra, S. T. Pantelides, R. D. Schrimpf, and K. F. Galloway, IEEE Trans. Nucl. Sci. 46, 1568 共1999兲. 4 A. Yokozawa and Y. Miyamoto, Phys. Rev. B 55, 13 783 共1997兲. 5 S. N. Rashkeev, D. M. Fleetwood, R. D. Schrimpf, and S. T. Pantelides, Phys. Rev. Lett. 87, 165506 共2001兲. 6 K. L. Brower, Phys. Rev. B 38, 9657 共1988兲. 7 D. M. Fleetwood, M. J. Johnson, T. L. Meisenheimer, P. S. Winokur, W. L. Warren, and S. C. Witczak, IEEE Trans. Nucl. Sci. 44, 1810 共1997兲. 8 K. Vanheusden, W. L. Warren, R. A. B. Devine, D. M. Fleetwood, J. R. Schwank, M. R. Shaneyfelt, P. S. Winokur, and Z. J. Lemnios, Nature 共London兲 386, 587 共1997兲. 9 R. E. Stahlbush, R. K. Lawrence, and H. L. Hughes, IEEE Trans. Nucl. Sci. 45, 2398 共1998兲. 10 R. A. B. Devine, J. Appl. Phys. 89, 2246 共2001兲. 11 P. J. Macfarlane and R. E. Stahlbush, Appl. Phys. Lett. 77, 3081 共2000兲. 12 M. C. Payne, M. P. Teter, D. C. Allan, T. A. Arias, and J. D. Joannopoulos, Rev. Mod. Phys. 64, 1045 共1992兲. 13 D. Vanderbilt, Phys. Rev. B 41, 7892 共1990兲. 14 G. Kresse and D. Joubert, Phys. Rev. B 59, 1758 共1999兲. 15 R. Buczko, S. J. Pennycook, and S. T. Pantelides, Phys. Rev. Lett. 84, 943 共2000兲. 16 M. Ramamoorthy and S. T. Pantelides, Appl. Phys. Lett. 75, 115 共1999兲. 17 S. N. Rashkeev, M. Di Ventra, and S. T. Pantelides, Appl. Phys. Lett. 78, 1571 共2001兲. 18 D. J. Chadi and M. L. Cohen, Phys. Rev. B 8, 5747 共1973兲. 19 S. T. Pantelides, S. N. Rashkeev, R. Buczko, D. M. Fleetwood, and R. D. Schrimpf, IEEE Trans. Nucl. Sci. 47, 2262 共2000兲. 20 J. K. Rudra and W. B. Fowler, Phys. Rev. B 35, 8223 共1987兲. 21 P. M. Lenahan and P. V. Dressendorfer, J. Appl. Phys. 55, 3495 共1984兲. 22 Y. Y. Kim and P. M. Lenahan, J. Appl. Phys. 64, 3551 共1988兲. 23 Z. Y. Lu, C. J. Nicklaw, D. M. Fleetwood, R. D. Schrimpf, and S. T. Pantelides 共unpublished兲. 24 W. L. Warren, M. R. Shaneyfelt, D. M. Fleetwood, J. R. Schwank, P. Winokur, and R. A. B. Devine, IEEE Trans. Nucl. Sci. 41, 1817 共1994兲. 25 P. E. Bunson, M. Di Ventra, S. T. Pantelides, D. M. Fleetwood, and R. D. Schrimpf, IEEE Trans. Nucl. Sci. 47, 2289 共2000兲. 26 R. E. Stahlbush, A. H. Edwards, D. L. Griscom, and B. J. Mrstik, J. Appl. Phys. 73, 658 共1993兲. 27 D. M. Fleetwood, Microelectron. Reliab. 42, 523 共2002兲. 1 2
spond to a density of true O vacancies within a few monolayers from the nominal interface. Calculations reported in another article find that an H⫹ can be trapped in a true O vacancy in bulk SiO2 and the binding energy is 0.8 eV,25 which is close to the binding energy of a proton trapped at an interface suboxide bond. It was shown that O vacancies in bulk SiO2 slow the transport of protons across the oxide and generate a delayed increase in the postirradiation 1/f noise.7 We conclude that true O vacancies near the interface should play the same role as interface suboxide bonds in capturing H⫹ and then releasing it. We note, however, that true O vacancies at layers adjacent to the nominal interface layer may be surrounded by Si–O–Si chains so that the barrier that holds the H⫹ in place does not have any intrinsic asymmetry. The interface suboxide bonds with their asymmetric barriers would then form the ultimate wall that blocks penetration of H⫹ into Si and allows the cycling to and from opposing Si–SiO2 interfaces. There are of course other reactions between hydrogen and Si–SiO2 interfaces. For example, if hydrogen is introduced in a previously irradiated oxide in a molecular form, the H2 molecules can crack on radiation-induced positively charged defects 共e.g., E ⬘ centers or broken bond hole traps兲, and additional mobile protons are thereby generated.26 In this case, one can observe a simultaneous interface trap buildup and a decrease of radiation-induced trapped positive charge within the oxide. There can then be a wide range of reactions with H2 in irradiated oxides in addition to those discussed here.27 In conclusion, we have shown that a suboxide Si–Si bond at an intrinsic Si–SiO2 interface creates a deep energy minimum 共trap兲 for protons, with a highly asymmetric bar-
Downloaded 29 Oct 2002 to 129.59.117.126. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
