Elsevier

Thin Solid Films

Volume 518, Issue 9, 26 February 2010, Pages 2317-2322
Thin Solid Films

Hydrogen in amorphous Si and Ge during solid phase epitaxy

https://doi.org/10.1016/j.tsf.2009.09.145Get rights and content

Abstract

Studies into the effect of hydrogen on the kinetics of solid phase epitaxy (SPE) in amorphous Si (a-Si) and Ge (a-Ge) are presented. During SPE, H diffuses into surface amorphous layers from the surface and segregates at the crystalline–amorphous interface. Some of the H crosses the interface and diffuses into the crystalline material where it either leaves the sample or is trapped by defects. H segregation at concentrations up to 2.3 × 1020 H/cm3 is observed in buried pha-Si layers with the SPE rate decreasing by up to 20%. H also results in a reduction of dopant-enhanced SPE rates and is used to explain the asymmetry effects between the SPE velocity profile and the dopant concentration profile observed with shallow dopant implants. Conversely, H diffusion is enhanced by dopants in a-Si. These studies suggest that H diffusion and SPE may be mediated by the same defect. The extent of H in-diffusion into a-Ge surface layers during SPE is about one order of magnitude less that that observed for a-Si layers. This is thought to be due to the lack of a stable surface oxide on a-Ge. However, a considerably greater retarding effect on the SPE rate in a-Ge of up to 70% is observed. A single unifying model is applied to both dopant-enhanced SPE and H diffusion processes.

Introduction

In the fabrication of a broad range of Si and Ge complementary metal-oxide-semiconductor (CMOS) devices an amorphous layer may form through the implantation of heavy ions with keV energies. Alternatively, self-amorphization implants may be used before implantation of lighter ions in order to avoid implantation channeling. Regrowth of the crystal layer via solid phase epitaxy (SPE) has been identified as a pathway for achieving high dopant activation with a low thermal budget [1]. Stringent demands are placed on device fabrication modeling where devices must be made efficiently in order to meet the requirements of future technology nodes. Accordingly, an extensive SPE literature exists (for comprehensive reviews see Refs. [2], [3], [4], [5], [6], [7], [8]).

The velocity of the c–a interface during SPE has a strong dependence on many parameters, all of which need to be known for process modeling to be accurate. These parameters include the substrate crystallographic orientation [9], pressure [10], and the presence of dopants [6]. SPE is also thermally activated, the c–a interface velocity being described by an Arrhenius-type equation of the form,v=voexp(Ea/kT)where vo and Ea are the pre-exponential factor and activation energy, respectively. For SPE in Si, vo = 4.64 × 108 cm/s and Ea = 2.7 eV has been determined up to the melting point [3]. The corresponding Ge values are vo = 2.6 × 109 cm/s and Ea = 2.15 eV determined in a temperature range of 300–540 °C [7]. The SPE rate is often unknowingly retarded by the presence of non-doping impurities such as hydrogen [4]. Hydrogen is an interesting case as it is often present during SPE unless special steps are taken. The behaviour of H during SPE then will also need to be understood and incorporated into fabrication models.

During thermal treatments on surface a-Si layers, H diffuses from the native oxide into the layer. Once the H meets the c–a interface it strongly segregates in the amorphous phase and retards the SPE rate by up to ~ 50% [4]. This in-diffusion occurs whenever there is water vapor in the ambient or a surface oxide present. For thin a-Si layers (< 400 nm), such as those produced during shallow junction processing, a nearly constant concentration of ~ 2 ×109 H/cm is expected at the c–a interface throughout the SPE process [4]. Amorphous layers formed by cluster implantation of decaborane (B10H14) [11] also contain H and therefore may affect the SPE rate.

For thick a-Si layers, H can infiltrate several microns into the a-Si layer before meeting the c–a interface. The SPE rate is found to decrease linearly with H concentration up to [H]  3 × 1019 cm 3. For greater concentrations up to 7 × 1019 cm 3 the SPE rate depends only weakly on the H concentration. This threshold value has been correlated with the density of dangling bonds (DB) in a-Si formed by ion implantation and has been cited as evidence for the possible involvement of DBs in the SPE process [4], [12].

Hydrogen diffusion is also thermally activated and for low H concentrations ([H]) has been described by an equation similar to Eq. (1) with v and vo replaced by a pre-exponential factor, Do = 2.2 × 104 cm2/s and an activation energy Ea = 2.7 eV [13]. Since a common activation energy is shared by SPE and H diffusion, one is led to suspect that the rate limiting step in these two seemingly unrelated processes may be the same [13]. Hence, studies into H diffusion and its role during SPE can also provide important insight into the SPE mechanism.

In this paper, experiments concerning the role of H during SPE are presented. Firstly, H implantation into buried a-Si layers is used to study the SPE retardation at H concentrations greater than 7 × 1019 cm 3. H is also shown to have a temperature dependent segregation coefficient. The interaction of H with dopants is also observed through SPE and H diffusion studies presented here. The presence of dopants can enhance the diffusion of both H in-diffusing from the surface and implanted H. The SPE rates in a-Ge are also shown to suffer from H contamination. These observations are drawn together into a single model, the generalized Fermi level shifting model, which links structural changes at the interface or H diffusion to the local electronic structure [6], [10]. The impact of these results on device modeling for fabrication processes is considered.

Section snippets

Theoretical background

The generalized Fermi level shifting (GFLS) model is used in this work to describe the effect of H on SPE and dopant-enhanced SPE and H diffusion. The GFLS model has been successful in describing the dopant dependence of SPE in both a-Si and a-Ge [6], [7], [10]. The effect of pressure has also been recently incorporated into the model [14].

According to the model, the number of events, R, is determined by the concentration of neutral defects, X0 and its positively or negatively charged

Sample preparation

The kinetics of SPE were measured in amorphous layers formed by self-ion implantation into <100> wafers. A National Electrostatics Corp. 1.7  MV tandem accelerator was used for all implants. The samples were tilted 7° off the incident beam axis to avoid channeling and also rotated about the surface normal by a similar amount to prevent any remaining possibility of planar channeling [16]. Substrates were affixed with Ag paste to the implanter stage, which was held at 77 K. Good thermal contact was

Hydrogen retardation of SPE

Fig. 1a) and b) are schematics of the expected H concentration ([H]) at two different times during an anneal of a surface a-Si layer at 660 °C. In Fig. 1a) the c–a interface has moved 0.25 μm from its original position. During this time H has infiltrated into the amorphous layer from the surface oxide. In Fig. 1b), the H has reached the c–a interface and causes the SPE rate to decrease while being strongly segregated on the amorphous side of the interface. During segregation a fraction of the H

Conclusions

The effect of H on the SPE rate is complex. The H infiltrates from the surface into the amorphous surface layer during thermal processing. The H diffusion coefficient is dependent not only on the temperature but also the dopant concentration in the amorphous layer. Once H meets the c–a interface it strongly segregates on the amorphous side of the interface. The segregation coefficient is found to have a temperature dependence. The in-diffusion of H and its dopant-enhancement can be used to

Acknowledgment

The Department of Electronic Materials Engineering at the Australian National University is acknowledged for their support by providing access to SIMS and ion implanting facilities. This work was supported by a grant from the Australian Research Council.

References (26)

  • G.L. Olson et al.

    Mat. Sci. Rep.

    (1988)
  • G. Olson et al.
  • A. Armigliato et al.

    Mater. Sci. Eng.

    (1989)
  • B.C. Johnson et al.

    Mater. Sci. Eng. B

    (2009)
  • International Technology Roadmap for Semiconductors (2003...
  • J.W. Mayer et al.

    Can. J. Phys.

    (1968)
  • O. Hellman, Mater. Sci. Eng. R16 (1996)...
  • B.C. Johnson et al.

    Phys. Rev. B.

    (2007)
  • B.C. Johnson et al.

    Phys. Rev. B

    (2008)
  • N.G. Rudawski et al.

    J. Appl. Phys.

    (2009)
  • L. Csepregi et al.

    J. Appl. Phys.

    (1978)
  • G.-Q. Lu et al.

    J. Appl. Phys.

    (1991)
  • J. Matsuo et al.
  • Cited by (0)

    View full text