Local-Loading Effects for Pure-Boron-Layer
Chemical-Vapor Deposition
V. Mohammadi, W.B. de Boer, T.L.M. Scholtes, L.K. Nanver Delft Institute of Microsystems and Nanoelectronics (Dimes),
Delft University of Technology, Feldmannweg 17, 2628 CT Delft, The Netherlands, Phone: +31 (0)15 27 86294, Fax: +31 (0)15 27 87369,
E-mail: v.mohammadi@tudelft.nl
The chemical-vapor deposition of pure boron (PureB) has been used very successfully as a means of fabricating extremely shallow, less than 10-nm deep, silicon p+n junction diodes. In this technology a nm-thin amorphous boron layer is deposited selectively on silicon through openings in an oxide isolation layer at temperature from 400ºC - 700ºC, in all cases creating an effective p+ layer at the interface. For the application as photodiodes, particularly impressive performance has been achieved for the detection of low penetration-depth beams for which purpose 2-nm-thick PureB-layers are reliably implemented as the front-entrance window [1]. Ideal low-leakage diode characteristics are achieved at low temperatures, which together with the fact that the deposition is conformal and highly selective to Si, also makes PureB technology an attractive candidate for creating junctions on silicon nanowires and advanced CMOS transistors including source/drain in p-type FinFETs [2]. In the latter applications, sub-3-nm thick layers are required to avoid excess series resistance through the high-resistivity boron layer. Moreover, for the photodiode application any thickness variations even in the range of angstrom can have a large impact on the responsivity to beams that only penetrate a few nm into the Si such as VUV light and less-than 1 keV electrons. Therefore, a very good control of the layer thickness is crucial.
The PureB deposition is susceptible to loading effects [3] the kinetics of which are investigated in this paper in order to achieve better control of the layer thickness on patterned wafers. For this purpose an analytical model is proposed that takes many of the important factors into account: the diffusion mechanism of the diborane species through the stationary boundary layer over the wafer, the gas phase processes and the related surface reactions by applying the actual parabolic gas velocity and temperature gradient profiles in the reactor to describe the deposition kinetics and the deposition chamber characteristics that determine the deposition rate over the non-rotating bare silicon wafer [4]. During the PureB deposition, there are two boron source components (vertical and lateral) that contribute incoming reactant molecules (Fig. 1a). The vertical component is determined by monitoring the deposition on a bare wafer where no lateral diffusion exists. A kinetic model for this term was proposed in [4] and can be predicted by the equation (1)
x u h D C x Cv 0 2 0 52 . 2 exp 692 . 0 ) ( (1)
The lateral component is a so-called local-loading effect that occurs because the boron is not deposited on oxide but will diffusion along the surface of the oxide. The importance of this component will depend on the patterning of the wafer and the boron diffusion lengths on the oxide and silicon, respectively. This is studied here by measuring the PureB-layer thickness of a die in the center of the wafer as a reference Si opening of width, WSi, that
is surrounded by a ring of oxide of width Woxand then a ring of isolating open silicon as shown schematically in the inset of Fig. 2. The overall oxide coverage ratio is also important as described in [3]. Here we define the local-oxide ratio (LOR) as the ratio of Wox to WSi, and the local-oxide ratio is varied by changing the width of the oxide, Wox.
Fig. 2 shows the PureB deposition rate versus local-oxide ratio. As described in the figure text, three regions can be distinguished. An empirical model has been developed to describe this behavior and it can be seen that all three regions are well described by the empirical formula. This formula can be used to model the lateral diffusion component of the boron atoms and to develop a comprehensive model to predict the PureB deposition rate on any 2-D uniform or non-uniformly patterned wafer. Examples will be given in the final paper of the impact of the local-loading effect on the performance of specific devices such as, for example, low-energy electron-beam detectors made up of different photodiode areas.
Figure 1. Schematic illustration of the (a) vertical and lateral boron source components that contribute incoming reactant molecules during the PureB deposition, (b) boron concentration distribution over a patterned wafer indicating the differences occuring due to the differences in the width of the Si windows and the oxide areas.
Figure 2. The PureB deposition rate versus local-oxide ratio. The lower inset shows a cross section of the central Si window where the PureB-layer thickness is measured and the surrounding rings of oxide and silicon. In the graph the red circles and blue stars represent the experimental results for the regions with LOR < 0.1 and 0.1 < LOR < 1 respectively. As the Wox increases, the depositon rate increases because more boron travels from the oxide to the Si-region. The dashed blue line is calculated by the empirical formula and is seen to fit very well with the experimental results. In the inset the LOR 4.5 is very large and the increase in PureB-layer thickness saturates for such high values, i.e., the diffusion across the oxide is limited by the diffusion length and not the presence of the Si ring.
It can be concluded that the local-loading effect plays an important role on the final PureB-layer thickness distribution. It has been possible to model the deposition parameters and the effect of the local-oxide ratio so that the thickness can be well predicted and controlled to the degree that 2-nm-thick PureB-layers can be deposited with only a few angstrom variation of the thickness.
[1] A. Šakic, et. al., IEDM 2010, pp.712-713. [2] L.K. Nanver, et. al., RTP 2010, pp. 136-139. [3] V. Mohammadi,et.al.,#862, 221st
ECS meeting,2012. [4] V. Mohammadi,et.al.,#904, 221st ECS meeting,2012.
I II III
(b) (a)