We demonstrate highly efficient, low threshold InAs quantum dot lasers epitaxially grown on on-axis (001) GaP/Si substrates using molecular beam epitaxy. Electron channeling contrast imaging measurements show a threading dislocation density of 7.3 × 106 cm−2 from an optimized GaAs template grown on GaP/Si. The high-quality GaAs templates enable as-cleaved quantum dot lasers to achieve a room-temperature continuous-wave (CW) threshold current of 9.5 mA, a threshold current density as low as 132 A/cm2, a single-side output power of 175 mW, and a wall-plug-efficiency of 38.4% at room temperature. As-cleaved QD lasers show ground-state CW lasing up to 80 °C. The application of a 95% high-reflectivity coating on one laser facet results in a CW threshold current of 6.7 mA, which is a record-low value for any kind of Fabry-Perot laser grown on Si.

Over the past decade, silicon photonics has emerged as a promising candidate for high performance, chip-scale interconnects to meet the rapidly growing demand for data bandwidths.1,2 Although impressive progress has been achieved for various passive silicon photonics devices, realization of a monolithically integrated, efficient light source on a Si substrate has yet remained as a challenging task.3 Self-assembled InAs quantum dot (QD) lasers show promise as an efficient light source for silicon photonic integrated circuits because the lateral carrier confinement of individual QDs makes them less sensitive to the high density of threading dislocations generated from the inherent lattice mismatch between GaAs and Si.4 

Since the first demonstration of 1.3 μm InAs QD lasers on Si substrates in pulsed operation by Wang et al. in 2011,4 much effort has been made to lower the threshold current density and to increase the output power.4–6 Liu et al. presented impressive results on high performance InAs QD ridge-waveguide lasers on Ge/Si substrates with low threshold current (16 mA), high output power (176 mW), and high continuous-wave (CW) operating temperature (119 °C).6 Also, Chen et al. reported efficient broad-area QD lasers on Si with threshold current densities as low as ∼60 A/cm2 at room temperature.7 All of these QD lasers were grown on offcut (4°–6°) Si substrates to avoid the formation of anti-phase domains (APDs), which are known to be electrically active and can be detrimental for laser diode efficiency.8 Unfortunately, offcut Si substrates are not fully compatible with standard CMOS processing foundries, and this hinders the viability of InAs QD lasers as light emitters for silicon photonics.

Recently, we reported the first demonstration of 1.3 μm QD lasers epitaxially grown on on-axis (001) GaP/Si substrates.9 The GaP layer, grown by metal-organic chemical vapor deposition (MOCVD), terminates APDs within ∼40 nm of the interface between GaP and Si, removing the need for offcut Si.10 GaAs on V-groove Si (001) templates also enable QD laser growth on Si without offcut because the KOH etched Si trenches allow APD-free III-V layers by growing on {111} surfaces.11 Chen et al. also reported QD lasers on on-axis (001) Si by growing an APD-free initial GaAs layer using MOCVD with a sophisticated Si surface preparation step.12 However, all the QD lasers grown on on-axis Si substrates showed relatively diminished performance in terms of threshold currents and output powers compared to the previous high performance QD lasers on offcut Si.

In this work, we demonstrate highly efficient 1.3 μm QD lasers epitaxially grown on on-axis GaP/Si (001) substrates enabled by improved GaAs templates with a threading dislocation density (TDD) of 7.3 × 106 cm−2. Narrow ridge waveguide lasers with as-cleaved facets achieve CW threshold currents as low as 9.5 mA at 20 °C. The lowest CW threshold current density is 132 A/cm2, and the highest output power is 175 mW from both facets of a device. The InAs QD lasers also show a high CW wall-plug-efficiency (WPE) of 38.4%. The CW ground-state operation persists up to 80 °C, producing an output power of 7 mW. Applying a high reflectivity (HR) coating to one facet lowers the CW threshold currents even further to 6.7 mA, which is the lowest value to date for any Fabry-Perot lasers on Si.

Figure 1(a) shows a schematic of the epitaxial QD laser structure grown on an on-axis (001) GaP/Si substrate. The GaP/Si (001) on-axis wafer is commercially available in 300 mm size from NAsPIII/V GmbH. We first grew a 3 μm thick GaAs buffer layer with In0.1Ga0.9As/GaAs strained superlattice dislocation filter layers in a solid-source molecular beam epitaxy (MBE) chamber. Thermal cycle annealing was performed four times at temperatures between 320 °C and 700 °C under As2 overpressure to reduce threading dislocations via dislocation annihilations. Then, 10 periods of In0.1Ga0.9As/GaAs strained superlattice layers were grown at 500 °C at a growth of approximately 1 μm/hr.13,14 The root-mean-square roughness of the surface was reduced from 5.45 nm to 2.48 nm from the atomic force microscopy (AFM) measurement (not shown here). More details about the GaAs template growth on GaP/Si and TDD characterization will be reported in a separate study. After the GaAs buffer growth, we removed the sample from the MBE chamber and conducted electron channeling contrast imaging (ECCI) measurements to investigate the TDD in the GaAs template.15 

FIG. 1.

(a) A cross-sectional schematic of the InAs QD laser structure grown on a GaP/Si substrate. (b) Electron channeling contrast images showing threading dislocations before and (c) after optimization of the GaAs/GaP/Si template.

FIG. 1.

(a) A cross-sectional schematic of the InAs QD laser structure grown on a GaP/Si substrate. (b) Electron channeling contrast images showing threading dislocations before and (c) after optimization of the GaAs/GaP/Si template.

Close modal

Figures 1(b) and 1(c) show representative ECCI images from before and after the optimization of the GaAs template layer. Threading dislocations are shown in either bright or dark spots. The electron channeling condition used here is the cross point of the (220) and (4¯00) channeling conditions to increase the contrast. The ECCI measurement revealed that the TDD from the optimized GaAs buffer is 7.3 × 106 cm−2 by surveying a total area of 2017 μm2. Compared to the non-optimized GaAs buffer as shown in Fig. 1(b), the TDD is reduced by a factor of ∼20. After the ECCI measurement, we loaded the sample back to the MBE chamber to grow a GaAs/AlGaAs graded-index separate confinement heterostructure laser with four stacks of QD layers as the active region. The QD layer consists of 2.55 monolayers of InAs embedded in a 7 nm In0.15Ga0.85As well, both grown at a substrate temperature of 495 °C measured using an optical pyrometer. The lower half of the laser structure including the GaAs/AlGaAs separate-confinement-heterostructure was grown at 580 °C, while the top half was grown at 560 °C. In addition to the laser sample, we also grew a photoluminescence structure with a single layer of InAs QDs on a GaAs/GaP/Si substrate to assess their optical and structural properties.

Figure 2 displays photoluminescence spectra at room temperature with a peak wavelength at 1285 nm. The full-width at half-maximum is ∼28 meV, indicating an excellent QD size homogeneity. The AFM image in the inset of Fig. 2 confirms the highly uniform self-assembled QDs grown on the GaP/Si substrate. The QD density is ∼5.2 × 1010 cm−2, which is comparable to the density of the QD grown on a GaAs native substrate.16 The energy separation between the ground state and excited state is 84 meV. In addition to the low TDD in the GaAs template, the superior optical and structural qualities of the QDs enable high performance QD lasers grown on GaP/Si substrates.

FIG. 2.

Room-temperature photoluminescence spectra of InAs QDs grown on a GaP/Si substrate. The inset shows an atomic force microscopy image of uncapped InAs QDs. The scale bar is 200 nm.

FIG. 2.

Room-temperature photoluminescence spectra of InAs QDs grown on a GaP/Si substrate. The inset shows an atomic force microscopy image of uncapped InAs QDs. The scale bar is 200 nm.

Close modal

Ridge waveguide lasers with various ridge widths were fabricated using standard dry etching techniques. The ridge widths range from 2.5 to 8.0 μm, and the laser cavity length is defined by cleaving after thinning the silicon substrate down to 150–200 μm. All laser performance measurements described in this report were carried out in the CW mode. Figure 3(a) shows representative light-current-voltage (LIV) curves from a 1485 × 2.5 μm2 device with as-cleaved facets at 20 °C. The threshold current is 9.5 mA, which is significantly lower than the previous results from QD lasers grown on-axis and offcut (4°–6°) Si substrates. We believe that the significant threshold reduction is attributed to the increased internal quantum efficiency in the devices grown on the low TDD GaAs templates. The output power of the laser is more than 71 mW per facet for ground-state lasing. Excited state lasing begins at 400 mA. The lasing wavelength is measured using an optical spectrum analyzer, and Fig. 3(b) displays the onset of lasing at 1270 nm with numerous Fabry-Perot longitudinal modes as the injected current increases from 5 mA to 15 mA.

FIG. 3.

(a) Single-side CW LIV curves at 20 °C from a 1485 × 2.5 μm2 laser with as-cleaved facets and (b) high-resolution lasing spectra from a 1090 × 2.5 μm2 laser showing the onset of Fabry-Perot lasing modes at 1270 μm.

FIG. 3.

(a) Single-side CW LIV curves at 20 °C from a 1485 × 2.5 μm2 laser with as-cleaved facets and (b) high-resolution lasing spectra from a 1090 × 2.5 μm2 laser showing the onset of Fabry-Perot lasing modes at 1270 μm.

Close modal

Figure 4(a) shows a maximum single-side output power of 175 mW from an as-cleaved 2600 × 8 μm2 device. The threshold current of this device is 27.5 mA (current density of 132 A/cm2), and the WPE is 29.5% at an injection current of 175 mA. It should be mentioned that appropriate dielectric coatings (anti/high-reflectivity) and facet passivation can further increase the output power of the device by avoiding possible catastrophic optical damage on the facets as the power density approaches ∼1 MW/cm2.17,18 We also tested the same device at elevated temperatures as shown in Fig. 4(b). The ground-state lasing persists up to 80 °C, still producing an output power of ∼7 mW. This demonstrates that the QD lasers on GaP/Si can operate in a harsh environment such as those found in datacenter and high performance computing applications. The characteristic temperature of the device between 20 and 80 °C is 32 K. This relatively low characteristic temperature is due to the undoped active region in the QD laser, and we expect to improve the temperature performance by employing p-modulation doping in the active region.19,20

FIG. 4.

(a) Single-side CW LIV curves at 20 °C from a 2600 × 8 μm2 laser with as-cleaved facets. The inset shows temperature-dependent single-side CW LI curves from 20 °C to 80 °C.

FIG. 4.

(a) Single-side CW LIV curves at 20 °C from a 2600 × 8 μm2 laser with as-cleaved facets. The inset shows temperature-dependent single-side CW LI curves from 20 °C to 80 °C.

Close modal

Figure 5(a) reveals a clear trend of lower threshold currents with smaller laser widths due to the reduced current injection area. The scattering in the threshold current at a given width is due to imperfect facet cleaving and device processing. It should be noted that the threshold current decreases almost linearly with the laser width down to 2.5 μm, indicating negligible sidewall recombination effects due to the excellent lateral carrier confinement of our InAs QD material. We expect that lasers with smaller ridge widths may achieve even lower threshold currents. The lowest CW threshold current density among the as-cleaved devices (device lengths = 1343–1633 μm) is 147 A/cm2 (36.75 A/cm2 per layer) at 20 °C, which is considerably lower than the previously reported value of ∼425 A/cm2 (60.71 A/cm2 per layer) from HR-coated QD lasers grown on-axis (001) Si substrates. We attribute this to the high quality GaAs template with low TDD in addition to the low transparency current from using a reduced number of QD layers (four instead of five to seven) in the active region. WPEs and single-side output powers of 161 as-cleaved devices are presented in Figs. 5(b) and 5(c). The highest WPE value of 38.4% is measured at an injection current of 74 mA from a 1366 × 4 μm2 QD laser, producing an output power of 18.6 mW. Figure 5(c) shows ground-state output power versus device width at 20 °C. The relatively large scattering in WPE and optical power of our QD devices on GaP/Si is from a combination of the device fabrication and imperfectly cleaved laser facets.

FIG. 5.

(a) A plot of threshold current versus device width for QD ridge waveguide lasers. The dashed-line represents a linear fit. (b) Double-side wall-plug-efficiency versus device width. (c) Single-side ground-state lasing output power versus device width. All lasers were measured in the CW mode with cleaved facets at 20 °C.

FIG. 5.

(a) A plot of threshold current versus device width for QD ridge waveguide lasers. The dashed-line represents a linear fit. (b) Double-side wall-plug-efficiency versus device width. (c) Single-side ground-state lasing output power versus device width. All lasers were measured in the CW mode with cleaved facets at 20 °C.

Close modal

We have coated one facet of the 1106 μm laser bar with a 95% (Ta2O5/SiO2) HR coating, and Fig. 6(a) shows a CW threshold current of 6.7 mA from a device with a width of 3 μm. We believe that this value is the lowest threshold current density to date for any kind of Fabry-Perot lasers on Si substrates. Its corresponding threshold current density is 205 A/cm2, and the maximum single-side WPE is 23.55%. Reliability tests on these devices are underway to investigate the lifetime at various aging conditions. We expect that QD lasers grown using the high quality GaAs templates will exhibit improved reliability relative to previous results due to their reduced threshold current densities and defect-driven degradation.21 Also, further laser characterizations such as extracting gain parameters will be investigated.

FIG. 6.

Single-side CW LIV and wall-plug efficiency curves from a 1090 × 3 μm2 laser with one 95% HR coated facet. The inset shows a CW threshold current of 6.7 mA at 20 °C.

FIG. 6.

Single-side CW LIV and wall-plug efficiency curves from a 1090 × 3 μm2 laser with one 95% HR coated facet. The inset shows a CW threshold current of 6.7 mA at 20 °C.

Close modal

In this work, we have demonstrated the significantly improved performance of 1.3 μm InAs quantum dot lasers epitaxially grown on on-axis (001) GaP/Si substrates. The high quality quantum dot active region and reduced threading dislocation density lead to a threshold current of 9.5 mA, an output power of 175 mW, and a wall-plug-efficiency of 38.4% from as-cleaved devices in the continuous-wave mode. High-reflectivity coating to one facet of the device further reduces the threshold current to 6.7 mA with a single-side wall-plug-efficiency of 23.55%. We believe that this work is very promising for highly efficient light sources for photonic integrated circuits on Si.

This research was supported by Advanced Research Projects Agency-Energy (ARPA-E) DE-AR000067. We are also grateful to Kurt Olsson and John English for their assistance in MBE maintenance and Alan Liu, Kei May Lau, Chris Palmstrøm, and Kunal Mukherjee for fruitful discussions.

1.
A.
Rickman
,
Nat. Photonics
8
,
579
(
2014
).
2.
M.
Asghari
and
A. V.
Krishnamoorthy
,
Nat. Photonics
5
,
268
(
2011
).
3.
J. E.
Bowers
,
T.
Komljenovic
,
M.
Davenport
,
J.
Hulme
,
A. Y.
Liu
,
C. T.
Santis
,
A.
Spott
,
S.
Srinivasan
,
E. J.
Stanton
, and
C.
Zhang
, “
Recent Advances in Silicon Photonic Integrated Circuits
,”
Proc. SPIE
9774
,
977402
(
2016
).
4.
T.
Wang
,
H. Y.
Liu
,
A.
Lee
,
F.
Pozzi
, and
A.
Seeds
,
Opt. Express
19
,
11381
(
2011
).
5.
A. D.
Lee
,
Q.
Jiang
,
M. C.
Tang
,
Y. Y.
Zhang
,
A. J.
Seeds
, and
H. Y.
Liu
,
IEEE J. Sel. Top. Quantum Electron.
19
,
1901107
(
2013
).
6.
A. Y.
Liu
,
C.
Zhang
,
J.
Norman
,
A.
Snyder
,
D.
Lubyshev
,
J. M.
Fastenau
,
A. W. K.
Liu
,
A. C.
Gossard
, and
J. E.
Bowers
,
Appl. Phys. Lett.
104
,
041104
(
2014
).
7.
S. M.
Chen
,
W.
Li
,
J.
Wu
,
Q.
Jiang
,
M. C.
Tang
,
S.
Shutts
,
S. N.
Elliott
,
A.
Sobiesierski
,
A. J.
Seeds
,
I.
Ross
,
P. M.
Smowton
, and
H. Y.
Liu
,
Nat. Photonics
10
,
307
(
2016
).
8.
H.
Kroemer
,
K. J.
Polasko
, and
S. C.
Wright
,
Appl. Phys. Lett.
36
,
763
(
1980
).
9.
A. Y.
Liu
,
J.
Peters
,
X.
Huang
,
D.
Jung
,
J.
Norman
,
M. L.
Lee
,
A. C.
Gossard
, and
J. E.
Bowers
,
Opt. Lett.
42
,
338
(
2017
).
10.
I.
Nemeth
,
B.
Kunert
,
W.
Stolz
, and
K.
Volz
,
J. Cryst. Growth
310
,
1595
(
2008
).
11.
J.
Norman
,
M. J.
Kennedy
,
J.
Selvidge
,
Q.
Li
,
Y. T.
Wan
,
A. Y.
Liu
,
P. G.
Callahan
,
M. P.
Echlin
,
T. M.
Pollock
,
K. M.
Lau
,
A. C.
Gossard
, and
J. E.
Bowers
,
Opt. Express
25
,
3927
(
2017
).
12.
S. M.
Chen
,
M. Y.
Liao
,
M. C.
Tang
,
J.
Wu
,
M.
Martin
,
T.
Baron
,
A.
Seeds
, and
H. Y.
Liu
,
Opt. Express
25
,
4632
(
2017
).
13.
M.
Yamaguchi
,
M.
Tachikawa
,
Y.
Itoh
,
M.
Sugo
, and
S.
Kondo
,
J. Appl. Phys.
68
,
4518
(
1990
).
14.
M.
Yamaguchi
,
M.
Sugo
, and
Y.
Itoh
,
Appl. Phys. Lett.
54
,
2568
(
1989
).
15.
K. N.
Yaung
,
S.
Kirnstoetter
,
J.
Faucher
,
A.
Gerger
,
A.
Lochtefeld
,
A.
Barnett
, and
M. L.
Lee
,
J. Cryst. Growth.
453
,
65
(
2016
).
16.
K.
Nishi
,
T.
Kageyama
,
M.
Yamaguchi
,
Y.
Maeda
,
K.
Takemasa
,
T.
Yamamoto
,
M.
Sugawara
, and
Y.
Arakawa
,
J. Cryst. Growth.
378
,
459
(
2013
).
17.
M.
Fukuda
,
M.
Okayasu
,
J.
Temmyo
, and
J.
Nakano
,
IEEE J. Quantum Electron.
30
,
471
(
1994
).
18.
S.
Kamiyama
,
Y.
Mori
,
Y.
Takahashi
, and
K.
Ohnaka
,
Appl. Phys. Lett.
58
,
2595
(
1991
).
19.
O. B.
Shchekin
and
D. G.
Deppe
,
Appl. Phys. Lett.
80
,
3277
(
2002
).
20.
A. Y.
Liu
,
C.
Zhang
,
A.
Snyder
,
D.
Lubyshev
,
J. M.
Fastenau
,
A. W. K.
Liu
,
A. C.
Gossard
, and
J. E.
Bowers
,
J. Vac. Sci. Technol. B
32
,
02C108
(
2014
).
21.
A. Y.
Liu
,
R. W.
Herrick
,
O.
Ueda
,
P. M.
Petroff
,
A. C.
Gossard
, and
J. E.
Bowers
,
IEEE J. Sel. Top. Quantum Electron.
21
,
1900708
(
2015
).