Nphoton 2014 321-s1

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

SUPPLEMENTARY INFORMATION

Supplementary information DOI: 10.1038/NPHOTON.2014.321

Lasing inLasing
direct-bandgap GeSn
in direct bandgap GeSn alloy
alloy grown grown
on Si (001) on Si
S. Wirths1*†, R. Geiger2,3†, N. von den Driesch1, G. Mussler1, T. Stoica1, S. Mantl1, Z. Ikonic4,

M. Luysberg5, S. Chiussi6, J.M. Hartmann7, H. Sigg2, J.Faist3, D. Buca1 and D. Grützmacher1

1
Peter Grünberg Institute 9 (PGI 9) and JARA-Fundamentals of Future Information
Technologies, Forschungszentrum Juelich, 52425 Juelich, Germany
2
Laboratory for Micro- and Nanotechnology (LMN), Paul Scherrer Institut, CH-5232 Villigen,
Switzerland
3
Institute for Quantum Electronics, ETH Zürich, CH-8093 Zürich, Switzerland
4
Institute of Microwaves and Photonics, School of Electronic and Electrical Engineering,
University of Leeds, Leeds LS2 9JT, United Kingdom
5
Peter Grünberg Institute 5 (PGI 5) and Ernst Ruska-Centrum, Forschungszentrum Juelich,
52425 Juelich, Germany
6
Dpto. Física Aplicada, E.E.Industrial, Univ. de Vigo, Campus Universitario, 36310 Vigo,
Spain
7
Univ. Grenoble Alpes, F-38000 Grenoble, France
CEA, LETI, MINATEC Campus, F-38054 Grenoble, France

These authors have equal contribution

This file contains:

1. Material characterization

2. Band structure calculations

3. Excitation power dependence of the PL

4. Modelling of temperature-dependent PL emission intensity

5. PL comparison between sample D and E

6. Calculation of near- and far-field of the mode pattern

7. References

NATURE PHOTONICS | www.nature.com/naturephotonics 1 1

© 2015 Macmillan Publishers Limited. All rights reserved


1. Material characterization

Figure S1: (a) X-Ray diffraction – reciprocal space maps (XRD-RSM) around the (224)

asymmetric reflection for the four differently alloyed GeSn layers (samples A, B, C and

D) are used to determine in- and out-of-plane lattice constants and, thus, the layer’s

residual compressive strain. (b) RBS random spectra of GeSn samples A (8 % Sn), B

(10.3 % Sn), C (11.5 % Sn) and D (13 % Sn). (c) RBS random and aligned spectra of

GeSn sample E (13 % Sn, 560 nm).

© 2015 Macmillan Publishers Limited. All rights reserved


The in-plane and out-of-plane lattice constants of the analysed layers were obtained

from RSM around the asymmetric (224) reflection. Three peaks arising from the Si(001)

substrate, the Ge buffer and the GeSn layer are identified for all samples, as shown in Fig. S1a.

The diagonal line represents the cubic crystal structure, i.e. an unstrained lattice. The GeSn x-

ray diffractograms were recorded using a triple-axis high-resolution Bruker D8 diffractometer.

Reciprocal space maps were carried out around the (224) reflection in a steep-incidence

geometry. To determine the Sn concentration xSn and the degree of strain relaxation R, a least

square fitting routine was employed. This routine calculates the in-plane and out-of-plane

lattice constants from the parameters xSn and R, and finds the parameters with the best

agreement between the experimentally obtained and calculated lattice constants. The equations

to determine the lattice constants and the elastic constants of GeSn were taken from ref.1. The

samples’ stoichiometry was also addressed by Rutherford backscattering spectrometry (RBS)

with 1.4 MeV He+ ions at a scattering angle of 170°. The samples have been tilted by 7° off

the incident ion beam and rotated about the (001) direction to avoid ion channelling effects.

The random spectra of the investigated samples are shown in Fig. S1b. The Sn concentration

was determined by fitting the random spectra using the RUMP simulation code (see Table I).

The accuracy for elemental concentration is ± 0.5 %. In this configuration, these measurements

are not sensitive to strain effects. Hence, a combination of XRD and RBS results provides

independent values for the Sn concentration. In order to investigate the crystalline quality of

the epilayers, RBS channelling was carried out. Here, the He+ ion beam is well aligned with

the <001> crystallographic direction of the sample and the resulting spectrum is divided by the

random spectrum known as minimum yield value, min, being a measure for the crystalline

quality. In Fig. S1c, the aligned and random spectra of sample E are shown. A min value of 5-

6 % determined within the Sn signal is comparable to values measured for pseudomorphically

grown SiGe layers on Si(001) and, hence, indicates high single crystalline quality and complete

© 2015 Macmillan Publishers Limited. All rights reserved


Sn substitutionality. Compared to recently published data by Bhargava et al.2 for MBE-grown

GeSn layers, our values are about four times lower.

2. Band structure calculations

Figure S2: Bandgaps at T = 300 K calculated for the partially relaxed Ge1-xSnx layers with

compressive strains of -0.70 % and -0.71 % and Sn concentrations of 8.0 % and 12.6 %,

respectively. Indicated are the energies of the band minima at the  and L-valley: their

energetic positions are predicted to inter change, i.e. amid the two samples the transition

from indirect to direct group IV semiconductor occurs.

The schematic band diagrams shown in Fig. S2 indicate the bandgaps for samples A

and D, respectively. The predicted difference in energy, E, of the - and L-valleys in the

conduction band are also indicated. Notably, due to the compressive strain, the heavy hole

(HH) valence band is shifted towards higher energies as compared to the light hole (LH)

valence band. The band structure around the  point was calculated using the 8 band k.p method

including strain effects3,4. Conduction band valleys which are not at the Brillouin-zone centre

may split depending on the applied strain as described via appropriate deformation potentials.

© 2015 Macmillan Publishers Limited. All rights reserved


The parameters used in this calculation are given in Table S1. The temperature dependence of

the bandgaps is accounted for via Varshni’s empirical relation for Ge. The alloy parameters are

calculated from the corresponding values for Ge and Sn, using Vegard’s-type interpolation for

the deformation potentials with the quadratic correction (bowing) for bandgaps and lattice

constant. The exceptions are the Luttinger  parameters, which are positive for Ge and negative

for Sn, making the Vegard’s extrapolation inapplicable. The parameters have been calculated

in ref.5 for several alloy compositions using the empirical pseudopotential method. These data

sets were used here to find a quadratic interpolation formula, which was slightly corrected to

reproduce the widely accepted values of Luttinger parameters for pure Ge, given in Table S1,

and employed in the calculations. Note, the electron effective mass at the  point is not an input

parameter, but rather a result from the k.p calculation.

© 2015 Macmillan Publishers Limited. All rights reserved


Table S1. Parameter used in the 8-band k.p method
Ge Sn
Lattice and stiffness constants
alatt (nm) 0.565756 0.649126
C11 (GPa) 128.536 69.06
C12 (GPa) 48.266 29.06
Band structure parameters
(1,2,3 are modified Luttinger (for Ge1-xSnx)
parameters, for 8-band k.p)
1 14.31-37.04x+560.24x2 5
2 4.95-19.16x+280.14x2 5
3 6.32-18.69x+280.59x2 5
p (eV) 26.30-2.30x 4
s.o. (eV) 0.284+0.744x-0.200x2 5
Eg, (T=0K) (eV) 0.8927 -0.4088
Eg,L (T=0K) (eV) 0.7447 0.12028
mL,l 1.576 1.4786
mL,t 0.08076 0.0756
Varshni’s parameters
ag, 5.82X10-4 7 -
T0, 2967 -
ag,L 4.774X10-4 7 -
T0,L 2357 -
Deformation potentials
av (eV) 1.246 1.586
b (eV) -2.96 -2.76
ac (eV) -8.246 -6.06
aL (eV) -1.546 -2.146
Alloy bowing parameters
(for Ge1-xSnx)
bowing(alatt) -0.0411
bowing(Eg,) 2.55-0.002 T 9
bowing(Eg,L) 0.89-0.0007 T 9

Table S2. Effective masses and valence band offsets used in the JDOS model

Sample m,x/y/m0 m,z/m0 mL,x/y/m0 mL,z/m0 (Ehh-Elh) (meV)


A 0.029 0.034 0.080 1.563 65
B 0.025 0.029 0.080 1.560 50
C 0.024 0.027 0.080 1.559 41
D 0.021 0.028 0.080 1.558 67
E 0.022 0.027 0.080 1.558 55

© 2015 Macmillan Publishers Limited. All rights reserved


3. Excitation power dependence of the PL

Figure S3: (a) Power dependent photoluminescence (PL) measurements for sample D at 20 K.

(b) Integrated PL intensity as a function of the laser power for samples B, C and D following

a linear dependence.

To perform PL measurements below room temperature, the samples were mounted in

a helium cold-finger cryostat. Figure S3 presents the power dependence of the integrated PL

signal measured at 20 K. For our samples B, C and D, we observe a nearly linear power

dependence (Fig. S3b) as is expressed by the relation I~ P, where I is the PL intensity, P is

the excitation power, and  is close to 1.

4. Modelling of temperature-dependent PL emission intensity

The qualitative description of the PL materializes from the fact that by lowering the

temperature, the electrons will condensate into the lowest state of the conduction band. If those

states belong to the L-valley (E < 0), the number of carriers in the L-valleys increases at the

expense of those in the -valley. If E >0, i.e. under the conditions of a direct bandgap

© 2015 Macmillan Publishers Limited. All rights reserved


semiconductor, the carriers gather at the -point leading to a strong increase in the radiative

recombination rate on lowering the temperature. Moreover, the PL intensity scales with the

number of excited carriers which is related to the excitation power and the recombination time

. The latter may depend on temperature, and thus contribute an additional temperature-

dependence to the PL behaviour upon cooling.

The temperature-dependent PL emission is estimated from the joint density of states

(JDOS) given by:

JDOS(E)  CmredE p E  Ecb  Evb f e E


cb, e ,Tf h E
vb, h ,T.
3
2
(2)

Here, C is a constant, mred is the reduced effective mass of electrons and holes, Ep is the dipole

matrix 
element, Ecb and Evb denote the conduction- and valence band extrema, µe/h are the

electron- and hole quasi Fermi levels, and T is the temperature. Under the assumption of

isotropic, parabolic bands, the Fermi functions for electrons and holes, fe and fh, are evaluated

at the primed energies

cb  E cb 
E 
m r ed

m cb E  E cb  E vb   (3)

and vb  E vb 
E 
m r ed

m vb E  E cb  E vb  .  (4)

The quasi Fermi levels are calculated for a given carrier density and temperature. For

comparison with the experiments, the above expression for the JDOS is evaluated for both,

heavy hole and light hole bands, integrated over the entire energy spectrum and normalized

with respect to the result at 300 K. For the dipole matrix element, a ratio between heavy- and

light hole band of Ephh/Eplh = 3/1 is used. The values taken for the effective masses and the

valence band splitting are obtained via the 8 band k.p approach described above and can be

found in Table S2. The conduction band offset Ecb between  and the L-valleys is obtained

by a recursive approach to account for the injected carrier density nc, which depends on the

non-radiative recombination time and, thus, on the temperature. The temperature

© 2015 Macmillan Publishers Limited. All rights reserved


characteristics of the lifetimes may be similar for all samples enabling to accurately fit the

experiments with a universal (i.e. identical for all samples) solution nc (T), together with the

sample-dependent conduction band offset Ecb. The best fit is obtained for nc (300K) = n0 ≈

4×1017 cm-3. The corresponding value for the nonradiative lifetime under the given pumping

conditions (2 mW, 532 nm, 5 µm diameter) is 0.35 ns, which is consistent with surface

recombination velocities recently determined for Ge on Si10.

Figure S4: Dependence of the recombination time  on the temperature obtained from the best

fit to the experimental PL intensities of samples A to D showing a Shockley-Read-Hall like T-

dependence.

The as obtained lifetimes at the measured temperatures are shown in Fig. S4 (filled diamonds)

together with the Shockley-Read-Hall (SRH) model defined as:


1
 1 A 
     . (5)
 0  SRH 

0 describes the lifetime at low temperatures, whereas SRH is the decay due to the capture of
 given by11:
charge carriers by mid-gap states

© 2015 Macmillan Publishers Limited. All rights reserved


ET 
 S RH 1 cosh  . (6)
 kT 

Here, ET denotes the difference between the trap level energy ET and the intrinsic Fermi level

EF,i, and k is the Boltzmann constant. With A being used to normalize  to 350 ps at 300 K, we

obtain from the fit shown in Fig. S4 0 = 2.1 ns and ET = 19.0 meV. The model fails, however,

to accurately describe the observed decrease of the lifetime for temperatures between 300 K

and 150 K. Moreover, the nearly linear excitation power dependence of the PL at low

temperature, c.f. Fig. S3b, is nicely reproduced by the JDOS model (not shown).

5. PL Comparison between sample D and E

To verify the directness of the bandgap of the most recently grown sample E, we repeated the

temperature-dependent PL measurements together with sample D. In Fig. S5, the blue triangles

and the blue line depict the normalized integrated PL intensity as well as the JDOS model for

sample D taken from Fig. 2b, whereas the green triangles and red circles show the new

experimental data from samples D and E, respectively.

Samples D and E show an analogous behaviour over the whole temperature range, indicating

that also sample E has a fundamental direct bandgap. There is, however, a deviation from the

data taken from Fig. 2b for temperatures < 50 K, when comparing sample D for both

measurements. This effect can be well reproduced by the JDOS model (c.f. red line in Fig. S5)

assuming the same conduction band offset of 25 meV and the same T-dependence for  but a

slightly increased carrier density of n0 = 6×1017 cm-3 compared to n0 = 4×1017 cm-3 due to

faintly different excitation conditions in the previous experiment.

10

© 2015 Macmillan Publishers Limited. All rights reserved


Figure S5: Temperature-dependent normalized integrated PL intensities for samples D and

E. Sample E shows a similar behaviour as sample D corresponding to a direct band gap with

conduction band offset of E = 25 meV. The deviation between the two data sets for sample

D can be explained by a slightly changed optical excitation.

11

© 2015 Macmillan Publishers Limited. All rights reserved


6. Calculation of near- and far-field of mode pattern

Figure S6: Simulated far-field pattern along (║) and perpendicular () to the growth

direction.

The near-field pattern of the fundamental TE waveguide mode was obtained using a 2D mode

solver based on plane-wave expansion12. The calculations were done for a layer stack

corresponding to sample E with a waveguide width of 5 µm and an etch depth of 900 nm. Input

parameters for the simulations are listed in Table S3, where the refractive index for

Ge0.874Sn0.126 was obtained from ellipsometry experiments.

From the near-field Enf shown in the inset of Fig. 3a of the main text, the far-field pattern was

calculated along x and y, i.e. the directions parallel and perpendicular to the growth direction

using13:

 e  Enf x, ydxdy


2
ik sin  x
Pffalong (r )  cos 2   (7)

 e II Enf x, ydxdy .


2
ik sin  y
and Pffperp (r )  cos 2  II (8)

12

© 2015 Macmillan Publishers Limited. All rights reserved


From the far-field pattern shown in Fig. S6, we calculated the expected maximal intensity

detected by a 1 mm2 detector in 60 mm distance from the waveguide facet. Hence, the obtained

collection efficiency is found to range between 0.015 % and 0.030 % for the cases of neglecting

or taking into account the light emitted towards the sample surface.

Table S3. Material input parameters for the near-field mode calculation
Material Thickness (µm) Refractive index

Ge0.874Sn0.126 0.56 4.2

Ge 2.70 4.0

Si 1.50 3.4

13

© 2015 Macmillan Publishers Limited. All rights reserved


7. References

1. Gencarelli, F. et al. Crystalline Properties and Strain Relaxation Mechanism of CVD


Grown GeSn. ECS J. Solid State Sci. Technol. 2, P134–P137 (2013).

2. Bhargava, N., Coppinger, M., Prakash Gupta, J., Wielunski, L. & Kolodzey, J. Lattice
constant and substitutional composition of GeSn alloys grown by molecular beam
epitaxy. Appl. Phys. Lett. 103, 041908 (2013).

3. Bahder, T. Eight-band k·p model of strained zinc-blende crystals. Phys. Rev. B 41,
11992–12001 (1990).

4. Bahder, T. Erratum: Eight-band k·p model of strained zinc-blende crystals [Phys. Rev.
B 41, 11 992 (1990)]. Phys. Rev. B 46, 9913–9913 (1992).

5. Quadratic fit to data from Lu Low, K., Yang, Y., Han, G., Fan, W. & Yeo, Y.
Electronic band structure and effective mass parameters of Ge1−xSnx alloys. J. Appl.
Phys. 112, 103715 (2012).

6. Chang, G.-E., Chang, S. & Chuang, S. L. Strain-Balanced GezSn1-zSix/GeySn1-x-y


Multiple-Quantum-Well Lasers. IEEE J. Quantum Electron. 46, 1813–1820 (2010).

7. http://www.ioffe.ru/SVA/NSM/Semicond/Ge/bandstr.html#Basic

8. Moontragoon, P., Soref, R. A. & Ikonic, Z. The direct and indirect bandgaps of
unstrained SixGe1−x−ySny and their photonic device applications. J. Appl. Phys. 112,
073106 (2012).

9. Linear fit, with low and high temperature values from Ryu, M.-Y., Harris, T. R., Yeo,
Y. K., Beeler, R. T. & Kouvetakis, J. Temperature-dependent photoluminescence of
Ge/Si and Ge1-ySny/Si, indicating possible indirect-to-direct bandgap transition at
lower Sn content. Appl. Phys. Lett. 102, 171908 (2013).

10. Geiger, R. et al. Excess carrier lifetimes in Ge layers on Si. Appl. Phys. Lett. 104,
062106 (2014).

11. Schubert, E. F. Light-Emitting Diodes. 236 (Cambridge University Press, 2002).

12. Schwarz, U.T. & Witzigmann, B. Optical Properties of Edge-Emitting Lasers:


Measurement and Simulation. 405–22 (Wiley-VCH, 2007).

13. Rosencher, E. & Vinter, B. Optoelectronics. (Cambridge University Press, 2002).

14

© 2015 Macmillan Publishers Limited. All rights reserved

You might also like