Analysis of strained Si/SiGe
structures by Raman spectroscopy
Piotr Dobrosz
Strained
silicon layers deposited on SiGe virtual substrate structures
have attracted interest because of their importance for high-speed
microelectronic devices. In order to achieve such speed increases, devices are
made from silicon with a tensile strain introduced by epitaxial deposition onto
a relaxed SiGe alloy with a larger lattice parameter.
Sample architectures with more complex layer structures, such as strained
silicon on strained SiGe, have been developed to achieve
further enhancement in electron and hole mobility. Characterisation
of the strain in such structures is critical to optimise processing and
performance.
Micro-Raman spectroscopy is a well
known characterisation technique and it can be used to assess the residual
strain in strained silicon/silicon germanium devices [1-4]. The peak shift
associated with the 520cm-1 silicon Raman peak
can be used to directly measure the macrostrain in the cap layer provided that
the strained silicon peak can be deconvoluted from the more intense Si in SiGe
peak which occurs at slightly lower wavenumbers. However, though peak position
gives a measure of the macrostrains in the layer it is not useful for the
assessment of microstrains associated with point defects which may also
influence device performance; such microstrains influence the intensity of the
Raman peaks and can, in principle, be monitored by this method.
In this study, a detailed analysis of
peak shape has been undertaken. Changes in peak shape and intensity were
investigated as a function of processing conditions for strained silicon on
SiGe with different sample architectures and compositions. Changes in peak
position are correlated with macrostrains and macrostrain relaxation around
extended defects such as dislocations [4]. Changes in peak width can be
correlated with processes which lead to changes in composition and defect
density (microstrain [3]). Such changes are not
necessarily correlated with changes in macrostrain but indicate that
microstrain could also be an important factor influencing device performance.
In many
cases the signal collected from the strained silicon layer is very weak and
overlapped by the Si in SiGe peak. In such situations the direct peak position
reading is very difficult and incorrect values may be produced unless an
appropriate analysis procedure is used. In order to find the Si cap peak
position in this study, two methods were used [4]. The first method is to
obtain a difference spectrum before and after selectively etching the Si cap
and the second method involves peak-fitting and deconvolution by software. Both
of these methods show very good agreement, however the peak fitting technique
is a faster and non-destructive technique and is suitable for analyzing devices
at various stages of processing.
A range of
strained Si/SiGe architectures have been assessed as
part of this project [5-8]. The dual channel strained Si/SiGe architecture has received
a lot of attention due to performance benefits being predicted for both n- and
p-channels; enhanced mobilities are expected for
holes in the compressively strained SiGe and
electrons in the tensile strained Si cap layer. However, the dual channel
structures investigated here show compromised mobility enhancement compared
with the single channels due to the effects of Ge
diffusion and interface roughening during processing which can be correlated
with Raman response. Due to problems with self-heating in devices made from
strained Si/SiGe thick virtual substrates (related to
the poor thermal conductivity of SiGe compared to
silicon), thin virtual substrate layers have been
developed to provide similar quality and performance to the thick virtual
substrate. Raman spectroscopy indicates that these SiGe
layers are not fully relaxed and that relaxation of the strain in the silicon
cap will occur during process anneals if the thickness of the cap exceeds its
critical thickness.
[1] Measurement of the residual macro and microstrain in strained Si/SiGe using Raman spectroscopy, P. Dobrosz, S.J. Bull, S.H. Olsen and A.G. O'Neill, Mat. Res. Soc. Symp. Proc., 809 (2004) 109-114.
[2] Impact of Ge diffusion and wafer cross hatching on strained Si MOSFET electrical parameters, L. Driscoll, S. Olsen, S. Chattopadhyay, A. O'Neill, K. Kwa, P. Dobrosz, and S. Bull, Mat. Res. Soc. Symp. Proc., 809 (2004) 225-230.
[3] Study of strain relaxation in Si/SiGe metaloxide-semiconductor field-effect transistors, S.H. Olsen, A.G. O'Neill, P. Dobrosz, S.J. Bull, L.S. Driscoll, S. Chattopadhyay and K.S.K. Kwa, J. Appl. Phys., 97 (2005) 114504.
[4] The use of Raman spectroscopy to identify strain and strain relaxation in strained Si/SiGe structures, P. Dobrosz, S.J. Bull, S.H. Olsen, A.G. O’Neill, Surf. Coat. Technol., 200 (2005) 1755-1760.
[5] On the relationship between electrical performance and Raman spectroscopic results for strained Si/SiGe devices, S.J. Bull, P. Dobrosz, S.H. Olsen and A.G. O’Neill, Proc. ICSI-4, (2005).
[6] Impact of
strained-Si thickness and Ge out-diffusion on gate
oxide quality for strained –Si surface channel n-MOSFETs,
Goutam Kumar Dalapati,
[7] Mobility-limiting
mechanisms in single and dual channel strained Si/SiGe
MOSFETs, S.H. Olsen, P. Dobrosz,
E. Escobedo-Cousin, S.J. Bull and A.G. O’Neill,Mat. Sci.
[8] Control
of self-heating in thin virtual substrate strained Si MOSFETs,
Sarah H. Olsen, Enrique Escobedo-Cousin, John B. Varzgar,
Johan Seger,