This is a repository copy of Ultra-low-threshold continuous-wave and pulsed lasing in
tensile-strained GeSn alloys.
White Rose Research Online URL for this paper:
http://eprints.whiterose.ac.uk/157738/
Version: Accepted Version
Article:
Elbaz, A, Buca, D, von den Driesch, N et al. (15 more authors) (2020) Ultra-low-threshold
continuous-wave and pulsed lasing in tensile-strained GeSn alloys. Nature Photonics, 14
(6). pp. 375-382. ISSN 1749-4885
https://doi.org/10.1038/s41566-020-0601-5
© The Author(s), under exclusive licence to Springer Nature Limited 2020. This is an
author produced version of a journal article published in Nature Photonics. Uploaded in
accordance with the publisher's self-archiving policy.
eprints@whiterose.ac.uk
https://eprints.whiterose.ac.uk/
Reuse
Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless
indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by
national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of
the full text version. This is indicated by the licence information on the White Rose Research Online record
for the item.
Takedown
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by
emailing eprints@whiterose.ac.uk including the URL of the record and the reason for the withdrawal request.
Ultra-low threshold continuous-wave and pulsed
lasing in tensile strained GeSn alloys
Anas Elbaz
1,2
, Dan Buca
3,*
, Nils von den Driesch
3,4
, Konstantinos Pantzas
1
, Gilles
Patriarche
1
, Nicolas Zerounian
1
, Etienne Herth
1
, Xavier Checoury
1
, S
´
ebastien Sauvage
1
,
Isabelle Sagnes
1
, Antonino Foti
5
, Razvigor Ossikovski
5
, Jean-Michel Hartmann
6
,
Fr
´
ed
´
eric Boeuf
2
, Zoran Ikonic
7
, Philippe Boucaud
8
, Detlev Gr ¨utzmacher
3,4
, and Moustafa
El Kurdi
1,*
1
Center for Nanoscience and Nanotechnology, C2N UMR 9001, CNRS, Universit
´
e Paris Sud, Universit
´
e Paris
Saclay, 91120 Palaiseau, France
2
STMicroelectronics, Rue Jean Monnet 38054 Crolles, France
3
Peter Gr
¨
unberg Institute (PGI 9) and JARA-Fundamentals of Future Information Technologies, Forschungszentrum
Juelich, 52428 Juelich, Germany
4
JARA-Institute Green IT, RWTH Aachen, 52062 Aachen, Germany
5
LPICM, CNRS, Ecole Polytechnique, Universit
´
e Paris-Saclay, 91128 Palaiseau, France
6
CEA, LETI and Univ. Grenoble Alpes, 38054 Grenoble, France
7
Pollard Institute, School of Electronic and Electrical Engineering, University of Leeds, Leeds LS2 9JT, UK
8
Universit
´
e C
ˆ
ote d’Azur, CNRS, CRHEA, 06560 Valbonne, France
*
corresponding authors: moustafa.el-kurdi@u-psud.fr, d.m.buca@fz-juelich.de
ABSTRACT
Strained GeSn alloys are promising for realizing light emitters based entirely on group IV elements. Here, we report GeSn
microdisk lasers encapsulated with a SiNx stressor layer to produce tensile strain. A 300nm GeSn layer with
5.4 at. %
Sn,
which is an indirect band-gap semiconductor as-grown, is transformed via tensile strain engineering into a direct band-gap
semiconductor that supports lasing. In this approach the low Sn concentration enables improved defect engineering and the
tensile strain delivers a low density of states at the valence band edge, which is the light hole band. We observe ultralow-
threshold continuous wave (cw) and pulsed lasing at temperatures of up to 70K and 100K respectively. Lasers operating at
a wavelength of
2.5 µm
have thresholds of
0.8 kW cm
−2
for
ns
-pulsed optical excitation, and
1.1 kW cm
−2
under cw optical
excitation. The results offer a path towards monolithically integrated group IV laser sources on a Si photonics platform.
Introduction
Si-Ge-Sn alloys are a promising, enabling material system for the monolithic integration of both passive and active optoelectronic
devices and circuits on Silicon
1
. Silicon (Si) photonics presently relies on integration of III-V materials for emitters
2
. Although
such an approach has recently demonstrated some impressive progress, it still faces challenges like the wafer throughput,
scalability and compatibility with the current Si complementary metal oxide semiconductor (CMOS) technology. The most
successful route for laser action within group IV materials nowadays is based on germanium-tin (GeSn) semiconductors. The
first demonstration of an optically pumped laser
3
and subsequent developments to improve the performance in respect to
threshold and operation temperature
4–7
, have shown the potential of these group IV materials for achieving Si-based light
sources, the final ingredient for completing an all-inclusive nano-photonic CMOS platform. Furthermore, (Si)GeSn materials
can help to extend the present Si photonics platform with a much broader application area than only near infrared data
communication. In the short-wave to mid infrared region of
2
–
4 µm
, in which GeSn laser emission has been obtained, potential
applications including gas sensing for environmental monitoring and industrial process control
8
, lab-on-a-chip applications
9, 10
or infrared imaging for night vision and hyperspectral imaging
11
can be envisaged.
A Sn content increase in GeSn alloys modifies the energy of the electronic bands. The band-gap at
Γ
-point (
E
Γ
) reduces
faster than the band-gap towards the
L
-valley (
E
L
), leading to a crossover from an indirect to a direct band-gap semiconductor
at an Sn concentration of
8 at. %
3
. The lattice mismatch between Sn-containing alloys and the Ge buffer layer, the typical
virtual substrate for their epitaxial growth, generates compressive strain in the grown layer, which counteracts the effect of Sn
incorporation, decreasing the directness
∆E
L−Γ
= E
L
− E
Γ
. On the contrary, applying tensile strain will increase the directness.
Finding a proper balance between a moderate Sn content to minimize crystal defects and to maintain thermal stability of
the GeSn alloy on one hand and making use of tensile strain on the other hand are the keys to bring lasing threshold and
operation temperature close to application’s requirements. The mainstream research to increase
∆E
L−Γ
focuses on high Sn
content alloys
5, 12
, obtained by epitaxy of thick strain-relaxed GeSn layers. A large directnesss is obtained, leading to higher
temperature operation, although at the expense of steadily increasing laser threshold
13
. We have recently theoretically proposed
an alternative approach, which is based on two key ingredients: employing moderate Sn content GeSn alloys, and inducing
tensile strain in them
14
. This study indicated that, if a given directness is reached via tensile strain rather than by increasing Sn
content, the material can provide a higher net gain. The underlying physics originates in the valence band splitting and lifting
up of the light hole,
LH
, band above the heavy hole,
HH
, band. Its lower density of states (DOS) reduces the carrier density
required for transparency, hence reduces the lasing threshold, as will be shown below.
GeSn alloys with a moderate Sn content offer a couple of advantages from the materials perspective. The epitaxial growth
temperature,
375
◦
C
compared to below
300
◦
C
for high Sn content alloys, yields a better crystalline quality and lower defect
density. Lattice mismatch, and therefore, the density of misfit dislocations at the GeSn/Ge interface scales with the Sn content
15
.
Both types of defects strongly influence the carrier recombination dynamics
16, 17
and contribute to the high pumping levels
necessary to reach lasing. Accordingly, laser thresholds of
100
–
300 kW cm
−2
were reported at
20 K
for GeSn lasers with
12–14 at. % Sn
4, 18
, while ∼MW cm
−2
values are required for very high Sn content alloys (>20 at. %) above 230 K
7, 12
.
The described material advantages and the underlying physics should then be combined with the technology able to induce
tensile strain in GeSn alloys. Strain engineering is a mature Si technology employed to modify the electronic band structure
of semiconductors
19
. In pure Ge,
1.7 %
biaxial
20, 21
or
4.5 %
uniaxial
22–24
strain is required to energetically align the
L
- and
Γ
-valleys (
∆E
L−Γ
= 0
), i.e. to reach crossover from an indirect to a direct semiconductor. Even higher strain is required to
achieve the mandatory directness,
∆E
L−Γ
>150 meV
, for room temperature operation of a strained Ge laser. Although such
high levels of tensile strain are technologically possible
25, 26
, they are challenging in a laser device geometry. Depending on
Sn content, which can be chosen in the range of
5
–
8 at. %
, significantly lower values of tensile strain are needed in GeSn
alloys to achieve a sufficient directness
14
. However, in order to obtain laser with low pumping threshold, the impact of the
defective GeSn/Ge interface region has to be removed. It is responsible for a considerable part of non-radiative recombination.
The growth of hetero- and quantum well structures appears to be a suitable technology to separate the gain material from the
defective interface
17
. Here, we focus on the impact of tensile strain and, for simplicity, we are using bulk GeSn layers. To
remove the defective interface, the layer transfer method is applied, as used to fabricate GeSn on insulator (GeSnOI) structures.
Due to the transfer, the defective part is at the top surface of the GeSn layer and can be easily removed by etching.
All GeSn lasers reported in the literature so far operate only under pulsed excitation, although continuous wave (cw) lasing is
the key milestone required for accessing the full potential of GeSn for technologically useful optical devices.
In this work, we demonstrate both cw and pulsed lasing using microdisk cavities fabricated from initially indirect band-gap
Ge
0.946
Sn
0.054
alloys. Tensile strain of
1.4 %
is applied ex-situ to the GeSn layer, using all-around SiN
x
stressors. The method
relies on the formation of GeSnOI by wafer bonding and layer transfer, followed by under-etching so that the final cavity is
supported by a metallic post, here Al, that acts as a heat sink. The combination of these factors, i.e. strain engineering, bulk
defect density reduction, and improved heat removal, allows the demonstration of lasing in tensile strained GeSn with record
low thresholds of
0.8 kW cm
−2
and
1.1 kW cm
−2
at
25 K
in pulsed and cw operation regimes, respectively. The threshold is two
orders of magnitude lower than previously reported for GeSn pulsed lasers, while no reports of cw lasing in GeSn are available.
Results
Material growth, characterization and cavity patterning. The GeSn layers were grown on Ge virtual substrates (Ge-VS) on
200 mm
Si(100) wafers
4, 18
TRON TRICENT
®
reactor
27
. Digermane (Ge
2
H
6
) and tin tetrachloride (SnCl
4
) were used as
precursors for elementary Ge and Sn, respectively. GeSn layers with thicknesses of
300 nm
and Sn content of
5.4 at. %
were
grown at
375
◦
C
. The layers were partially strain relaxed, with a residual compressive strain of
−0.32 %
as measured by X-Ray
diffraction (Figure 1a). Details on material characterization can be found in the supplementary information (SI).
The GeSn layers were processed into microdisk cavities with all-around SiN
x
stressors. The technology flow, developed for
pure Ge
28
and adapted for GeSn thermal budget limitations
29
, is presented in the SI. Yet, some key aspects are shown in
Figure 1b. The stressor layer is a
350 nm
thick SiN
x
layer with an intrinsic stress of
−1.9 GPa
. An additional Al metal layer was
added to the layer stack, in order to reduce heating during optical pumping
30, 31
. This is particularly important, since alloying of
Ge with Sn strongly decreases the thermal conductivity of the alloy
32
. After GeSn bonding and removal of the donor wafer,
the top
40 nm
of the GeSn layer, containing the defective GeSn/Ge interface, was also removed. It was shown that this dense
misfit network strongly reduces the photoluminescence of the layers at the onset of strain relaxation
33
. Up to this step the
compressive strain in the GeSn layer is preserved (Figure 1c,d). Tensile strain, coming from the SiN
x
stressor underneath, is
induced only by structuring the GeSn/ SiN
x
layer stack (Figure 1b). The under-etching process, i.e. selective and local removal
of Al, was optimized to maximize the tensile strain in the GeSn layer and have a wide Al post for heat sinking. Subsequently,
2/12
the suspended disks were conformally covered by a second
400 nm
thick SiN
x
stressor layer, leading to fully encapsulated
GeSn disks, standing on Al posts (Figure1b).This layer transfer technology and processing transforms the initial Ge
0.946
Sn
0.054
layer with residual compressive strain having an indirect band-gap into a microdisk exhibiting pronounced biaxial tensile strain
and consequently a direct band-gap.
a
280 290 300
Ge reference
as-grown GeSn
bonded GeSn
all-around
SiN stressor
Raman signal intensity (a.u.)
Raman signal (cm
-1
)
310
k
E
HH
LH
L-valley
Γ-valley
720 meV
660 meV
as-grown
20 K, -0.32%
-30 meV
k
E
HH
LH
L-valley
Γ-valley
466 meV
535 meV
final laser device
80 K, 1.40 %
170 meV
e
f
tensile
strain
GeSn layer
bonding &
substrate removal
c
b
micro disk
structuring
&
SiN deposition
GeSn
GeVS
as-grown
a
out-of-plane
(Å)
Ge-VS
GeSn
5.60 5.65 5.70 5.75
5.60
5.65
5.70
5.75
a
in-plane
(Å)
pseudomorphic line
strain-
relaxed line
Photoluminescence (a.u.)
0.4 0.5 0.6 0.7 0.8
Energy (eV)
bonded
x10
d
Γ-HH
Sim
80 K
8 mW
LN InSb
Γ-LH
as-grown
x10
20 K
15 mW
InGaAs
L-HH
all-around
SiN
2 µm
Al
SiN
GeSn
2 µm10 nm
Figure 1. Structural and optical characterization.(a) X-Ray diffraction reciprocal space map for quantitative assessment
of the layer strain (−0.32 %). (b) Sketch of the processing steps. TEM image of the as-grown layer with strain releasing
dislocations at the GeSn/Ge-VS interface. SEM images of the under-etched GeSn/SiN
x
stack and the final laser structure with
SiN
x
all-around.
(c)
Raman spectra for three cases: as-grown layer, bonded (i.e. transferred) layer and all-around disk of
9 µm
diameter. Bulk Ge is used as reference. (d) PL spectra taken at 20 K for the as-grown and bonded GeSn layers (the bonded
layer has the misfit dislocation network removed), and PL of patterned disk with all-around stressor at 80 K. The dotted line
shows the simulated PL spectra of the final structure. The calculated band structure for the
(e)
indirect band-gap as-grown layer
and (f) direct band-gap tensile strained GeSn layer.
Sample analysis
Raman spectroscopy was performed to follow the strain evolution during fabrication of GeSn microdisks. The positions of
the Ge-Ge vibration modes in GeSn alloys in the as-grown sample (blue line), after bonding onto the host Si substrate (green
line), and the final processed microdisk structure are shown in Figure 1c. The Raman modes of the unpatterned transferred
layer and the as-grown layer, at
297.3 cm
−1
and
296.8 cm
−1
, respectively, are very similar, within the experimental resolution of
0.5 cm
−1
. Therefore, it can reasonably be assumed that the layer transfer process itself does not change the strain in the GeSn
layer. However, after the processing of the final microdisk, a Raman shift of
−10.5 cm
−1
is detected. Using the equations from
Ref.
34
, including alloy disorder and strain effects, this Raman shift corresponds to a built-in biaxial tensile strain of
1.5 %
for
the all-around embedded GeSn disk. Note that Raman spectroscopy probes only the in-plane strain within a small depth below
the disk surface, while photoluminescence probes the whole disk volume, giving an average value of the strain, and is directly
related to the band structure.
28
.
Photoluminescence (PL) experiments were conducted to assess the strain-induced band structure changes, as well as
the quality improvement of the transferred layer. The PL signal from the as-grown GeSn layer is very weak, due to
(i)
the
3/12
indirect band-gap, with the conduction band energy splitting
∆E
L−Γ
=−60 meV
, and
(ii)
the presence of defects at the GeSn/Ge
interface. After the layer is transferred, these defects are removed and an increase in PL intensity is observed (Figure 1d). The
PL signal is found in the same energy range, indicating again that the transferred layer maintains its compressive strain and,
therefore, its indirect band-gap character with
L
and
HH
as conduction and valence band extrema, respectively. The optical
transition at
0.61 eV
is attributed to the recombination of electrons in the
L
-valley of the conduction band and holes near the
Γ
point of the
HH
valence band, thus across the fundamental indirect band-gap. Indirect carrier recombination dominates over
the direct transitions, because almost
100 %
of electrons at
20 K
are in the
L
-valley. The shoulder of the PL signal around
0.66 eV
is assigned to the direct transition, i.e. electrons and heavy holes around the
Γ
-point in k-space. Details on optical
transition identification can be found in SI. PL spectra in Figure 1d for the as-grown and the transferred GeSn layers are taken
under identical conditions with a sensitive InGaAs detector. This is emphasized because this, initially weakly-emitting, layer
will become the active laser medium after inducing the tensile strain.
The major limitation for the GeSn emission efficiency, the indirect band-gap, is overcome by tensile strain turning it into
a direct-gap semiconductor. Since the extended InGaAs detector has a cut-off wavelength of
2.4 µm
, another set-up with a
nitrogen cooled InSb detector, with a cut-off of
4.8 µm
, is used for the fully processed microdisk device. Even though the
InSb detector has a lower sensitivity, a strong increase of the integrated PL emission by 2 orders of magnitude compared to
the as-grown layer is measured. The PL signal is strongly red-shifted, showing the peak emission around
0.50 eV
which is
attributed to the tensile strain of the final structure with the all-around SiN
x
stressor. The band structure was modeled using
k·p
method, taking
5.4 at. %
Sn and a band-gap of
465 meV
of the strained film. The best fit to the experimental data in Fig. 1d is
obtained for biaxial tensile strain of
1.4 %
. This value is slightly smaller than the
1.5 %
obtained from Raman spectroscopy.
The discrepancy is attributed to the uncertainty of the parameters used in the two methods. The band structures of the as-grown
and tensile strained GeSn are shown in Figure1e and -f, respectively. In the final device, the tensile strain lifts the degeneracy of
the
LH
and
HH
band, with
LH
becoming the fundamental valence band. The valence band splitting is
E
LH
− E
HH
=
170 meV
.
More importantly, in the conduction band the tensile strain shifts the
Γ
-valley below the
L
-valley. The tensile strained GeSn
thus becomes a direct band-gap material with a directness of
∆E
L−Γ
= 70meV
. Consequently, the pronounced enhancement of
the PL intensity emission is due to the fundamental direct optical transition. The electrons recombine with
LH
at the
Γ
point in
the center of the Brillouin zone. This transition is labelled as Γ-LH optical transition.
CW device results
In order to obtain stimulated emission, microdisk devices with a diameter of
9 µm
were optically pumped using a
µ
-PL setup
with
1550 nm
wavelength cw pump laser focused on the sample surface into a
12 µm
diameter spot (see Methods). PL emission
spectra collected at various incident pump powers at
25 K
are shown in Figure 2a. At low excitation levels, the microdisks
produce a broad spontaneous emission background, attributed to
Γ → LH
direct transitions. By increasing the cw pump power
from
0.2 mW
to
0.8 mW
, whispering gallery modes (WGM) develop and grow in intensity on top of the spontaneous emission.
Higher excitation induces an exponential intensity increase of the main optical mode at
485 meV
. At
2.3 mW
pump power,
the lasing emission is four orders of magnitude stronger than the background, as shown in the high resolution spectrum in
Figure
2b. The two symmetric side lobes of the modes are only a measurement artifact, which stems from a finite range of
sampling points due to the apodization of the interferogram.
35
The observation of a clear threshold in the light-in light-out (L-L) characteristic visible in Figure 2 c,d, the
S
-shape L-L
characteristic (inset Figure
2d) and the collapse of the linewidth (Figure 2 c) unambiguously prove the onset of lasing. The
emission energy at
25 K
of the lasing mode at
485 meV
corresponds to
2.55 µm
wavelength. WGM simulations (see Methods)
indicate that this mode is the transverse magnetic TM
20,1
mode. The intensity of this mode in the threshold region is displayed
in Figure 2c as a function of incident pump power by the linear L-L characteristics. The observed narrowing is consistent with
the Schawlow-Townes equations, which predict a decrease by a factor of two at the transition from incoherent to coherent
emission. Note that the measured linewidth of 58 µeV is the smallest reported to date for any group-IV semiconductor lasers.
The laser threshold, clearly separating the spontaneous and the amplified emission regimes, is extracted as
1.3 mW
(Figure 2d).
This value corresponds to a pump power density of
1.1 kW cm
−2
. The dependence of the integrated signal on cw pumping
power at different temperatures is also shown in Fig.
2d. The typical laser S-shape emission is shown in logarithmic scales
in the inset. Below
45 K
an unambiguous threshold can be observed, while no signature of lasing is detected above
45 K
. A
roll-over of the integrated emission occurs for pump powers above
5 mW
. In III-V compounds this effect is typically associated
with thermal effects, leading to a sharp laser emission quenching with further increase of temperature. However, in the case of
GeSn microdisk laser under investigation, the directness
∆E
L−Γ
is in the range of only
70meV
. In fact, band filling effects of
the
Γ
-valley will reduce the electrons energy required for thermal (scattering) escape from
Γ
- into the
L
-valley even further.
Therefore, a small increase in temperature or excitation power may lead to an exponential increase of this thermal escape,
producing the observed roll-over. As will be discussed later, an increase in the electron population of the
L
-valley leads to a
decrease of the total gain. Microdisks with diameter of
12 µm
were also fabricated. The undereteching is kept constant, leading
4/12
