112113-2
Lieten et al.
Appl. Phys. Lett. 94, 112113 ͑2009͒
We have investigated the crystallization of amorphous
Ge layers deposited by PECVD with mass spectrometry and
reflection high energy electron diffraction in ultrahigh
vacuum. The release of hydrogen could be observed during
the beginning of the crystallization process around 400 °C.
This hydrogen originates from the breaking of Ge–H bonds
inside the amorphous Ge layer during crystallization. The
layers that were thermally annealed in vacuum at 600 °C
however showed a rough surface, which is in agreement with
other reports.11,14 Our layers showed a RMS roughness larger
than 2.5 nm, as measured with XRR. Much smoother layers
were obtained for annealing in N2 atmosphere at 600 °C. A
RMS surface roughness of 0.7 nm was measured for 100 nm
crystallized Ge on Si͑111͒ by XRR. Therefore annealing in
N2 seems effective in keeping the surface atoms immobile
during annealing and obtaining smooth layers. In this way
the surface roughness after crystallization is comparable to
the roughness of the as deposited layers. A possible explana-
tion could be the formation of a few monolayers of Ge ni-
tride during annealing, which protects the Ge layer from
roughening. Such a layer is formed when exposing Ge to
nitrogen plasma.18 However no Ge nitride layer was ob-
served by x-ray photoemission spectroscopy measurements
for annealing in N2 atmosphere at 600 °C. Therefore it
seems that the Ge atoms are immobilized by collisions with
N2 molecules. The crystalline Ge layers that were annealed
in N2 atmosphere and in vacuum showed in both cases com-
plete relaxation. This indicates that the surface roughening is
not related to the relaxation of the Ge layer which exceeds
the critical layer thickness, but to the Ge surface mobility.
During annealing, when the Ge layer is not yet fully crystal-
lized, Ge atoms at the surface can become mobile and clus-
ter. This surface mobility should be limited during crystalli-
zation to keep the surface smooth.
Structural characterization of various samples has been
carried out using high resolution XRD. The influence of the
Ge thickness was investigated by preparing layers of 100, 50,
10, and 5 nm. For all these thicknesses the crystalline Ge
layer was fully relaxed, as observed by XRD. This indicates
that the lattice mismatch between Si and Ge is compensated
by the introduction of dislocations in the first 5 nm. There
does not seem to be an upper or lower limit in layer thick-
ness for crystallization. The crystallization must start at the
Si–Ge interface and then propagate to the Ge surface. Crys-
tallization in the bulk, before the crystalline front has
reached the surface must be avoided. This would lead to
polycrystalline layers. Therefore, there is a limit on the speed
of the temperature ramp and possibly on the total thickness
of the Ge layer. The high crystal quality of our Ge layers
indicates PECVD is well suited for depositing highly disor-
dered layers. The limited roughness after crystallization in
N2 atmosphere indicates a layer-by-layer growth mode for
SPE of Ge on Si͑111͒.
FIG. 1. ͑Color online͒ XRD /2 scan of 100 nm crystallized amorphous
Ge on Si͑111͒. An intense and narrow Ge ͑111͒ diffraction peak is observed,
indicating high crystal quality. The presence of fringes indicates a smooth
Ge surface and interface between Ge and Si.
In a first series of experiments Ge layers were deposited
on Si͑111͒ substrates and were thermally annealed at 600 °C
in ultraclean N2 atmosphere at atmospheric pressure. The
resulted Ge layers are crystalline with the ͕111͖ crystal
planes parallel to the ͕111͖ Si surface, as observed from
/2 XRD scans, see Fig. 1. The Ge crystal quality is high.
A XRD rocking curve scan of the Ge ͑111͒ diffraction peak
showed a full width at half maximum ͑FWHM͒ of only 146
arc sec for 100 nm Ge. Fringes are visible in the /2 XRD
scan of Fig. 1 indicating a smooth Ge surface and interface
between Ge and Si. However when using a similar cleaning,
deposition and annealing procedure for Si͑001͒ substrates, a
much broader and less intense XRD ͑004͒ peak was ob-
served, see Fig. 2. From this experiment it is clear the crystal
orientation of the Si surface has an important influence on
the crystallization process. A difference in seed nucleation or
seed growth related to the crystal orientation could be the
reason for the difference in crystal quality. A difference in
grain growth of amorphous Ge can be expected on Si͑111͒
and Si͑001͒ substrates as crystallization of amorphous Si is
reported to be faster on Si͑001͒ surfaces than on Si͑111͒.17
For this reason one could expect that crystallization of Ge on
Si͑111͒ proceeds layer-by-layer, whereas on Si͑001͒ crystal-
line seeds will grow in an island growth mode. Layer-by-
layer growth will lead to superior quality in respect with
island growth, explaining the difference in crystal quality.
Structural investigation by XRD shows the presence of
Ge twins. This was observed by XRD phi scans of skew-
symmetric reflections, a typical result is shown in Fig. 3. The
͑202͒ reflection was measured for different phi angles, cor-
responding with a rotation of the sample in the plane of
growth. Two types of Ge diffraction peaks are observed: at
the same phi angle as the Si diffraction peaks and in between
the Si diffraction peaks. The latter Ge diffraction peaks origi-
nate from Ge that is 180° rotated in plane. Avoiding the
FIG. 2. ͑Color online͒ XRD /2 scan of 100 nm crystallized amorphous
Ge on Si͑001͒. A Ge ͑004͒ diffraction peak is observed, which is less intense
and broader than the Ge ͑111͒ peak on Si͑111͒. This indicates that crystal-
lization on Si͑001͒ leads to less good crystal quality than on Si͑111͒.
formation of these twins would improve the crystal quality
138.251.14.35 On: Sat, 20 Dec 2014 11:22:50