Accelerat ing t he world's research.

Recent progress in lasers on silicon

John Bowers

Nature Photonics

Cite this paper Downloaded from Academia.edu 

Get the citation in MLA, APA, or Chicago styles

Related papers Download a PDF Pack of t he best relat ed papers 

III-V-on-Silicon Phot onic Devices for Opt ical Communicat ion and Sensing
Sarah Uvin, Mart ijn Tassaert , Kasper Van Gasse, Sören Dhoore, Jochem Verbist , Jing Zhang, …

Silicon Phot onics

bahram jalali

Micro-opt ical resonat ors for microlasers and int egrat ed opt oelect ronics: recent advances and fut ure…
Ana Vukovic
 FOCUS | PROGRESS ARTICLE
PUBLISHED ONLINE: 30 JULY 2010|DOI: 10.1038/NPHOTON.2010.167

Recent progress in lasers on silicon

Di Liang* and John E. Bowers

Silicon lasers have long been a goal for semiconductor scientists, and a number of important breakthroughs in the past decade
have focused attention on silicon as a photonic platform. Here we review the most recent progress in this field, including low-
threshold silicon Raman lasers with racetrack ring resonator cavities, the first germanium-on-silicon lasers operating at room
temperature, and hybrid silicon microring and microdisk lasers. The fundamentals of carrier transition physics in crystalline
silicon are discussed briefly. The basics of several important approaches for creating lasers on silicon are explained, and the
challenges and opportunities associated with these approaches are discussed.

T
he photonics market today is shared by several materials sys- InP Si
tems, including compound semiconductors (indium phosphide,
InP, and gallium arsenide, GaAs), elementary semiconductors Free-carrier
(silicon, Si, and germanium, Ge), silica and rare-earth-doped glasses absorption
Electrons
(glass ibre, for example) and polymers. Each system targets particular L
applications or components. Today, the use of Si photonics is dwarfed Phonon

Energy
Indirect X
by compound semiconductors and Si microelectronics, mostly due to Г
hv recombination
the problems associated with making Si a host material for eicient Direct Auger
light emission, and thus subsequently realizing a laser. Fity years ago recombination recombination
the birth of the laser started a scientiic and technological revolu- Wave vector
Holes
tion. Two years later, diode lasers were demonstrated in group iii–v
compound semiconductors, and this was around the same time that
Si-based transistor radios achieved mass popularity. Since then many
scientists and engineers have researched lasing on Si substrates1. Rapid
advances in Si photonics over the past two decades have been driven Figure 1 | Energy band diagrams and major carrier transition processes
not only by the need for more complex, higher functionality and in InP and silicon crystals. In a direct band structure (such as InP, left),
lower cost photonics integrated circuits, but also by pin count and electron–hole recombination almost always results in photon emission,
power limits for communications, as summarized in the International whereas in an indirect band structure (such as Si, right), free-carrier
Technology Roadmap for Semiconductors (ITRS)2. Electronics giants absorption, Auger recombination and indirect recombination exist
such as Intel, IBM, Hewlett Packard, STMicroelectronics, IMEC and simultaneously, resulting in little photon emission.
Alcatel-hales have teamed up with research institutes around the
world with support from government, industry and academia to which results in slow optical transition rates. A major non-radiative
drive progress in Si photonics. he current momentum and potential process is Auger recombination, in which an electron (or hole) is
for making a useful laser in or on Si are signiicant. excited to a higher energy level by absorbing the released energy
from an electron–hole recombination. he Auger recombination
Fundamentals rate increases with injected free-carrier density and is inversely pro-
At the time of the demonstration of the irst laser ity years ago, portional to the bandgap. Free-carrier absorption (FCA) represents
the fundamental hurdle to realizing stimulated emission in Si was another major non-radiative process wherein the free electrons in
understood: optical transitions must obey the laws of conservation the conduction band can jump to higher energy levels by absorb-
of energy and momentum, but these conditions are not satisied ing photons. In high-level carrier injection devices (lasers and
simultaneously in crystalline Si. In direct bandgap materials (GaAs ampliiers, for example) or heavily doped layers, free-carrier loss is
and InP, for example) radiative recombination occurs rapidly and orders of magnitudes higher than the material gain1. For both Auger
eiciently via a simple two-particle process, as shown by the simpli- recombination and FCA, the electrons pumped to higher energy
ied band diagram in Fig. 1 (let). Direct bandgap materials have a levels release their energy through phonons, rather than by emitting
structure in which the lowest energy points of both the conduction photons. hey also have much shorter lifetimes (τnonrad) than those of
and valence bands line up vertically in the wave vector axis; that is, radiative processes (τrad) in Si, resulting in an extremely poor inter-
they share the same crystal momentum. his is the principal reason nal quantum eiciency ηi of light emission, which is deined as3
why GaAs-, InP- and GaN-based materials have been the dominant
material systems for semiconductor diode lasers since their irst τnonrad
demonstration in 1962. ηi = τ + τrad
nonrad
Si, like Ge, is an indirect bandgap material, and is not naturally
capable of accomplishing eicient radiative recombination. Free
electrons tend to reside in the X valley of the conduction band, which and is generally of the order of 10–6. Consequently, semiconductor
is not aligned with free holes in the valence band (Fig. 1, right). laser research over the past ity years has primarily focused on com-
herefore if a recombination is to lead to emission of a photon, a pound semiconductor substrates, but now there is intense interest
third particle must be involved to carry away the excess momentum, in lasers on Si.

Department of Electrical and Computer Engineering, University of California, Santa Barbara, California 93106, USA. *e-mail: [email protected]

NATURE PHOTONICS | VOL 4 | AUGUST 2010 | www.nature.com/naturephotonics 511

© 2010 Macmillan Publishers Limited. All rights reserved

PROGRESS ARTICLE | FOCUS NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.167

a Bus waveguide Directional coupler metal–oxide–semiconductor (CMOS) technology is suiciently

Pump Iinc advanced to fabricate virtually all Si photonic components, which
are mostly still in the micrometre regime. Both factors allow for
Laser Si waveguides with propagation losses that are typically one order
output Ip(L) Ip(0) z of magnitude lower than compound semiconductor waveguides.
Furthermore, Si has a high thermal conductivity, which is a very
n-region V bias useful characteristic for an active device substrate. SiO2, the high-
quality native oxide of Si, serves as a protective layer and a naturally
good optical waveguide cladding, owing to its large refractive index
p-region Ring cavity
diference from Si (Δn ~ 2.1). his is one of the major advantages of
Si over Ge and other semiconductors for use in integrated circuits.
b SiO2 passivation Further loss-reduction in Si waveguides by oxidation19 and hosting
Al contact Al contact
rare-earth doping in SiO2 brings additional beneits to passive Si
lightwave circuits. Although low waveguide loss does not change
p-region Si rib waveguides n-region the ultralow band-to-band radiative emission eiciency in Si, it
improves the eiciency of Si lasers that rely on a nonlinear efect
Buried oxide such as Raman scattering.
Si substrate 1 μm
Silicon Raman lasers
c 60 d 10 he Raman efect refers to the inelastic scattering of a photon by an
25V optical phonon. When incident light is absorbed by an atom or mol-
0
50 10V ecule at a vibrational state, the system energy is raised to an inter-
–10
mediate higher state. In most cases, the energy quickly drops back
Relative spectral power (dB)

5V
0V –20 to the original vibrational state by releasing a photon with the same
Laser output (mW)

40
–30 frequency, which is known as Rayleigh scattering, and is analogous
to elastic scattering. Yet it is also possible to observe very weak
80 dB

30 –40
(approximately one in ten million photons) additional components
–50 with lower and higher frequencies than the incident light due to the
20 absorption or emission of optical phonons, namely the Stokes and
–60
anti-Stokes transitions, respectively.
–70
10 If a scattering medium is irradiated with pump and signal beams
–80 simultaneously, the pump beam excites the constituent molecules
0 –90 or atoms to a higher vibrational level, while the signal beam, which
0 100 200 300 1,544 1,545 1,546 1,547 has a frequency resonant at the Stokes transition, triggers the gen-
Pump power (mW) Wavelength (nm) eration of another Raman Stokes photon. hus, ampliication can
be achieved through stimulation of the Stokes transition. his tech-
Figure 2 | Low-threshold Si Raman racetrack ring laser. a, Schematic of a nique is known as stimulated Raman scattering, and has enabled
device with a p-i-n junction design. Ip(0) and Ip(L) are the pump power at the realization of Raman glass ibre ampliiers with gain band-
the starting point and after a round trip in the cavity, respectively. The light widths of over 100 nm. he Raman gain coeicient in Si is around
propagation direction is given by z. b, SEM cross-section of a directional ive orders of magnitude larger than that in amorphous glass ibres
coupler and p-i-n junction region. c, Laser output power against coupled input because of the well-organized single-crystal structure20. However,
pump power, showing a higher output power achieved at a higher reverse Si waveguide loss is also several orders of magnitude higher than in
bias on p-i-n junction for a 3 cm cavity. The error bars here are derived from glass ibre, making fabrication of a low-loss Si waveguide one of the
diferent measurement traces. d, High-resolution spectrum showing a low- keys to realizing net Raman gain in Si. Furthermore, the tight optical
threshold Si Raman racetrack ring laser with a side-mode suppression ratio of coninement in an SOI waveguide leads to an ultrasmall waveguide
over 70 dB. Figure reproduced from ref. 25, © 2007 NPG. efective area, which in turn lowers the pump power threshold for
stimulated Raman scattering. A pump with energy well below the
he recent and widespread availability of nanotechnology has Si bandgap is typically used to avoid elevating the electrons up to
allowed the traditional phonon-selection rule in indirect band- the conduction band and also to suppress FCA — both of which
gap materials to be relaxed by breaking the crystal-symmetry or prevent lasing in Si. Initial studies demonstrated up to 0.25 dB of
by phonon localization through the creation of nanostructures in stimulated Raman gain for a Stokes signal at 1,542.3 nm for SOI
crystalline Si. he motivation is to achieve quantum coninement waveguides, using a 1,427 nm pump laser with a CW power of
of excitons in a nanometre-scale crystalline structure4. A number of 1.6 W (ref. 21). Such high pump powers, however, induce another
groups have reported enhanced light-emitting eiciency and opti- optical loss mechanism — two-photon absorption (TPA). TPA is
cal gain in low-dimensional (that is, of the order of the de Broglie a nonlinear loss mechanism in which two photons combine their
wavelength) Si at low temperatures. hey include porous Si5–8, Si energies to boost an electron in the valence band to the conduction
nanocrystals9–12, Si-on-insulator (SOI) superlattices13 and photonic- band. Free carriers further induce FCA and dump more optical
crystal-like nanopatterns14, and Si nanopillars15,16. However, achiev- power inside the cavity. TPA increases with the number of photons
ing room-temperature continuous-wave (CW) lasing based on these in a waveguide, and therefore becomes a limiting factor when using
temperature-dominated processes remains a challenge3,17,18. high optical pump powers. he irst demonstration of a pulsed Si
Despite being fundamentally limited by an indirect bandgap Raman laser 22 overcame TPA by using a long delay together with
and low mobility, Si exhibits a number of important properties that a short optical pulse, thus allowing the carriers generated dur-
make it a good substrate, if not necessarily a good gain medium ing TPA to recombine prior to the next pass of the optical pulse.
for diode lasers. First, Si wafers are incredibly pure and have low Following demonstrations used a p-i-n (p-type/intrinsic/n-type
defect density. Second, state-of-the-art 32 nm complementary layers) structure in the waveguide to sweep free carriers away under

512 NATURE PHOTONICS | VOL 4 | AUGUST 2010 | www.nature.com/naturephotonics

© 2010 Macmillan Publishers Limited. All rights reserved

NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.167 FOCUS | PROGRESS ARTICLE
a reverse bias23, as this reduced the free-carrier lifetime to minimize

Integrated emission intensity

TPA-induced FCA. An alternative method involved reducing the 25 100
volume-to-surface ratio of the waveguide to increase the surface 80
recombination rate of the carriers. he irst successful demonstra-
60
tion of a CW Si Raman laser followed soon ater 24, with a lasing 20
threshold at an efective pump power of ~182 mW for a reverse 40

Emission intensity (a.u.)

bias of 25 V. 20
he threshold pump power was recently reduced using a high 15
Q-factor racetrack ring resonator cavity and an optimized p-i-n 0
0 10 20 30 40 50
diode structure25. he cavity resonance efect enhances the light ield Pump energy per pulse (μJ)
inside cavity. Figure 2a is a top-view schematic of the racetrack ring 10 50 μJ
cavity, with the p-i-n regions highlighted. A large bend radius of
400 μm helps to minimize waveguide bending losses, even though
high-index-contrast SOI waveguides can typically support low-loss
light propagation in a more compact bending structure. To utilize 5 6.0 μJ
Ge
the pump power eiciently and achieve a low lasing threshold, the
directional coupler in the 1.6-cm-long bus waveguide is designed 1.5 μJ
Si 500 nm
so that it is close to the critical coupling for the pump wavelength 0
1,560 1,580 1,600 1,620 1,640 1,660 1,680 1,700
(1,550 nm) but has low coupling for the Stoke signal wavelength
Wavelength (nm)
(1,686 nm). he narrow gap in the evanescent coupler was illed
with boron phosphorus Si glass to eliminate any air voids that form, Figure 3 | Optically pumped Ge-on-Si laser demonstrating CW
which helps to reduce losses. A thin layer of SiO2 bufer layer on top operation at room temperature. Edge-emission spectra of a Fabry–Pérot
of the Si surface is deposited before coating with boron phosphorus Ge waveguide under three diferent levels of optical pumping from a
Si glass to prevent phosphor and boron from difusing into the Si Q-switched laser at 1,064 nm with a pulse duration of 1.5 ns and at a
during the thermal relow step and inducing FCA later on25. repetition rate of 1 kHz is shown. The arrow indicates the peak optical gain
TPA-induced FCA nonlinear optical loss can also reduced by wavelength. Top inset: integral emission intensity from the waveguide facet
optimizing the p-i-n reverse-biased diode. Balancing the trade-of versus optical pump power, showing the lasing threshold. Bottom inset:
between a good metal/Si contact and induced free-carrier absorp- cross-sectional SEM image of the Ge waveguide. Figure reproduced with
tion loss ensures that the difusion of electrons and holes under permission from ref. 45, © 2010 OSA.
reverse bias produces a uniform ield across the optical mode, which
allows eicient carrier removal without signiicantly increasing the
linear optical loss. Figure 2b shows a scanning electron microscopy GaSb34 bufer layers have enabled the realization of GaAs-based
(SEM) cross-sectional image of the directional coupling region and CW diode lasers on Si substrates at room-temperature. However,
incorporated p-i-n diode structure. he average optical loss of this their reliability still remains a big issue for any future practical
particular racetrack ring was measured to be 0.20 ± 0.05 dB cm–1, application. An exciting approach is the epitaxial growth of com-
which indicates a negligible contribution of the p-i-n diode to the pound semiconductors lattice-matched to Si, such as GaNAsP35,36.
linear propagation loss when the p- and n-region separation is Another exciting approach is Ge-on-Si (or SiGe-on-Si) epitax-
greater than 6 μm (ref. 25). An extremely short free-carrier lifetime ial growth. Key photonic components from this material system,
of <0.4 ns was obtained for this device, resulting in a substantial including p-i-n37 and avalanche photodetectors38,39 and modula-
reduction in the lasing threshold. Under a reverse bias of 25 V, the tors40,41, have demonstrated performances comparable or even bet-
laser had a threshold of 20 mW and a maximum output power of ter than their iii–v counterparts in certain aspects. Pure Ge has
50 mW (Fig. 2c). hese represent ive- and tenfold improvements, a signiicant mismatch with Si in terms of its lattice constant and
respectively, over the irst CW Si Raman lasers24. As the bias volt- thermal expansion coeicient. Germanium has an indirect band
age is lowered, the laser output begins to saturate earlier, owing to structure, but the energy gap (0.8 eV) from the top of the valence
the relatively longer efective carrier lifetime. However, the lasing band to the momentum-aligned Γ valley is close to the actual band-
threshold changes only slightly because the TPA-induced FCA is gap (0.66 eV), which increases the chance of radiative recombina-
much weaker at lower pump powers around the threshold. Silicon tion between the Γ valley and the valence band. he larger thermal
Raman lasers beneit signiicantly from high spectral purity, which expansion coeicient of Ge naturally leaves thermal tensile strain
results from the absence of a linewidth enhancement (a common in Ge ater growth on a Si substrate, and a moderate tensile strain
efect in diode lasers). For example, linewidths of <100 kHz and of 0.2–0.25% is able to reduce the energy diference between the
side-mode suppression ratios of over 70 dB (see Fig. 2d) are well Γ and L valleys to 115 meV (refs 42,43). In addition, strain raises
beyond the best performance of present diode lasers25. the light-hole band, which increases optical gain for high injec-
tion42. Free electrons, incorporated through heavy n-doping,
Epitaxial lasers on silicon quickly ill up the L valley to a level equal to that of the Γ valley,
Compared with Si, GaAs and InP have lattice mismatches of which increases the probability that those free carriers will begin to
4.1% and 8.1%, respectively, and thermal expansion coeicient occupy the Γ valley for radiative recombination. hese techniques
mismatches of 120.4% and 76.9%, respectively. hese result in a have enabled room-temperature direct-bandgap electrolumines-
threading or misit dislocation density of 108–1010 cm–2 when either cence43,44 and CW room-temperature optically pumped operation
compound is grown on a Si substrate26. Numerous approaches, of Ge-on-Si lasers45.
including special surface treatment 27, strained superlattices28,29, he irst Ge-on-Si laser operating at room-temperature was fab-
low-temperature bufers30 and growth on patterned substrates31 ricated by selectively growing 1.6 μm × 0.5 μm Ge waveguides epi-
have been used to reduce the dislocation density to around 105–106 taxially on Si (ref. 45). A thermally induced tensile strain of 0.24%,
cm–2, but this is still around two orders of magnitude higher than together with a phosphorous doping level of 1 × 1019 cm–3, allowed
in InP- or GaAs-based epitaxial wafers for room-temperature CW enhanced light emission from direct gap of 0.76 eV. A cross-sectional
lasers. Recent advanced epitaxial techniques with SiGe32,33 and SEM picture of the Ge waveguide is shown in the inset of Fig. 3. he

NATURE PHOTONICS | VOL 4 | AUGUST 2010 | www.nature.com/naturephotonics 513

© 2010 Macmillan Publishers Limited. All rights reserved

PROGRESS ARTICLE | FOCUS NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.167

a pumped rare-earth-ion laser on Si is nearing realization46,47. his p-i-n

diode structure consists of a stack of very thin Si layers alternating
n-contact p-contact
with very thin erbium-doped SiO2 layers. Carriers injected laterally
SOI p-InGaAs into the Si ilms make their way into the oxide ilms to excite the
+ p-InP erbium. Related work on Si LEDs also shows promise11,12. By introduc-
h Active region
+
h n-InP ing dislocation loops into Si along with erbium implantation, room-
temperature electroluminescence and optical gain at a wavelength of
1.5 μm have been achieved. Similarly, in related experiments, strong
electroluminescence at 1.2–1.3 μm has resulted from replacing the
boron implanted in Si with thullium.

Hybrid silicon lasers

b w = 1 μm
c -10 he SOI substrates used in Si photonics are usually made by wafer
-30

-40 bonding an oxidized Si wafer onto another Si carrier wafer. By

-20 -50 wafer-bonding compound semiconductors to SOI substrates, this
III–V
-60 same approach can be used to combine the superior gain charac-
Si
Bonding -70 teristics of compound semiconductors with the superior passive
interface -30
waveguide characteristics of Si waveguides. his idea has been
Relative spectral power (dB)

-80
00 demonstrated for wafer sizes of up to 150 mm in diameter, and the

0
60

20
w = 1.5 μm

4
8

1,6
1,5

1,5

1,6
1,6

-40 photoluminescence characteristics of the bonded compound semi-

conductor ilms actually improve ater bonding, probably due to the
III–V
release of strain during cooling ater epitaxial ilm growth48,49. In
Si -50
w this way, it is possible to combine epitaxial ilms with low thread-
ing dislocation densities to the lattice-mismatched Si substrate. his
w = 2.5 μm -60 has advantages over bonding individual iii–v lasers to a SOI host
substrate50,51. Shown schematically in Fig. 4a, the Si passive light-
-70
wave circuits are patterned prior to the transfer, and the iii–v ilms
III–V
are processed ater transfer using standard lithography-based pat-
Si
terning techniques used to fabricate iii–v lasers52. Lateral current
-80 coninement is achieved using h+ proton implantation. he mode is
1,595 1,600 1,605
Wavelength (nm) conined predominantly in the Si waveguide, with a fraction (typi-
cally 3–8%) contained in the iii–v quantum wells, thus forming a
Figure 4 | Hybrid Si device platform and hybrid Si distributed hybrid mode.
feedback lasers. a, Schematic of a hybrid Si Fabry–Pérot laser, showing he coninement factor can be dramatically changed by chang-
recombination scheme of injected carriers in iii–v epitaxial layers on top of ing the waveguide width (Fig. 4b); when the Si waveguide width
a Si waveguide. b, Simulated hybrid mode profiles at diferent Si waveguide increases, the mode tends to sink more into the Si, which decreases
widths, w, showing that modal confinement in Si (iii–v) increases coninement in the iii–v layer. his allows lasers, ampliiers, and
(decreases) as the waveguide width increases. c, Spectra of a hybrid Si photodetectors with diferent quantum-well coninement factors to
distributed feedback laser, showing a side-mode suppression ratio of 50 dB be integrated onto the same chip.
and a single-mode bandwidth of over 100 nm (inset). Figure c reproduced Following the demonstration of hybrid Fabry–Pérot lasers at
with permission from ref. 55, © 2008 OSA. room-temperature51,52, non-Fabry–Pérot cavity devices53–57 were
soon fabricated for on-chip light sources, including for distributed
feedback, distributed Bragg relector and segmented grating distrib-
study involved optically pumping a mirror-polished, 4.8-mm-long uted Bragg relector lasers. A single mode (>100 nm bandwidth)
Fabry–Pérot cavity with a 1,064 nm Q-switched laser of 1.5 ns pulse spectrum with a side-mode suppression ratio of 50 dB (Fig. 4c)
duration. he actual peak pump power density absorbed by the Ge from a hybrid Si distributed feedback laser is comparable to those
was estimated to be 300 kW cm–2. A broad photoluminescence peak based on iii–v materials55.
at 1,600 nm was observed under pumping at 1.5 μJ per pulse, with For optical interconnects, small size, low power consumption
emission peaks emerging at 1,599, 1,606 and 1,612 nm. A shoulder and a short cavity design are all critical. Compact microring lasers
is visible at 1,594 nm when the pump increases to 6.0 μJ per pulse, with diameters of 15–50 μm have been fabricated through a self-
which corresponds to the lasing threshold in the inset of Fig. 3. aligned process, allowing CW operation up to a stage temperature
In contrast with bulk direct-bandgap compound semiconduc- of 65 oC (ref. 57). Figure 5a shows a schematic of an electrically
tors, an interesting and important characteristic in strained Ge is the pumped microring resonator laser, its cross-section SEM image,
enhanced luminescence eiciency at high injection levels and high and a simulated fundamental transverse electric mode, showing
temperatures44. he theoretical explanation of the larger overlap of the shit to the waveguide edge. CW thresholds as low as 4 mA
the Fermi–Dirac distribution in the Γ valley due to Joule heating with reasonable output powers (3.5 mW) have been observed in
has been conirmed experimentally by electrically driven Ge-on-Si devices with diameters of 50 μm (ref. 57). Further reducing the laser
LEDs44. Researchers are currently investigating an electrically diameter and increasing the facet relectivity (that is, reducing the
pumped version of the Ge laser consisting of a Si/Ge/Si heterostruc- mirror loss for straight devices and the outcoupling for ring/disk
ture diode that is p+-, n+- and n-doped, respectively. devices) results in a drastic reduction in the threshold. he dotted
line in Fig. 5b shows how the threshold has decreased from pre-
Rare-earth-doped Si/SiO2 light-emitting devices viously demonstrated large hybrid Si racetrack ring lasers (blue
Silicon-based structures that use rare-earth-ion ‘guests’ in a dielec- dot)53 to the recently fabricated microring lasers57 and eventually
tric ‘host’ have great potential for lasing at telecommunications to ultracompact devices with diameters of 4.5 μm and thresholds
wavelengths. Researchers have recently reported that an electrically of ~400 μA (red dot). Figure 5c shows a photo of two 1 cm2 chips,

514 NATURE PHOTONICS | VOL 4 | AUGUST 2010 | www.nature.com/naturephotonics

© 2010 Macmillan Publishers Limited. All rights reserved

NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.167 FOCUS | PROGRESS ARTICLE
each containing ~400 devices lying aside a portion of a 150-mm- a
diameter hybrid Si racetrack laser wafer 58. hanks to this reduction
in size, the fabrication of millions of microring lasers on a single Si
wafer is now feasible.
Sub-milliampere-threshold CW lasing has been experimentally
demonstrated on a hybrid iii–v-on-Si integrated platform56 simi-
lar to the hybrid microring laser discussed above. By lasing inside
a compact microdisk iii–v cavity and coupling to an external Si r-metal
waveguide, a good overlap between the optical mode and electrical
p-InP
gain results in threshold currents as low as 350 μA (ref. 59). An SOI p-metal
waveguide is positioned underneath the iii–v microdisk to capture a MQW n-InP
Si
small fraction of the evanescent light vertically (Fig. 6a). Four devices BOX
with slightly diferent cavity lengths integrated onto the same Si
waveguide (Fig. 6b) will have diferent resonance wavelengths, allow- b Cavity length (µm)
ing such a waveguide to be used as a wavelength-division multiplex- 10 100 1,000
ing source60. he spectrum in Fig. 6c results from combining four 100
R=0.3
devices with diameters of 7.632, 7.588, 7.544 and 7.5 μm. he devices
are individually tuned to give an even spread inside one free spectral

Threshold current (mA)

R=0.8
range (~24 nm)60. However, increasing thermal impedance causes 10 R=0.9
laser performance to decrease dramatically 60,61 with smaller diam-
eters, which is a major hurdle in the realization of compact devices. D = 15 µm
1 R=0.98
Challenges and opportunities
Lasers on Si are now a reality, and the recent progress in making Si
lase — regardless of the particular lasing mechanism — is exciting.
Although the gain in nanoscale Si at low temperatures diminishes 0.1
quickly before reaching room temperature, and despite the gen- 1 10 100 1,000
Diameter (µm)
erated photon energy being larger than or similar to the Si band-
gap, steady progress is being made. Erbium-doped Si nanocrystals c
push emission to telecommunication wavelengths, and the use of
a metal–oxide–semiconductor structure results in an electrolumi-
nescence eiciency that is comparable to commercial iii–v LEDs.
As long as a good trade-of solution between eiciency and life-
time can be found, lasers on Si will ind applications in relatively
low-speed, large-volume optical interconnects. Materials scientists
are pursuing a variety of heteroepitaxy techniques to ind a way to
reduce the dislocation density in compound semiconductors on Si
to a level that is low enough for good and reliable laser perform-
ance. he recent demonstration of optically pumped Ge-on-Si lasers
is exciting, and is focusing attention on how to increase the gain
to levels comparable to that of iii–v materials. Hybrid iii–v-on-Si
technology currently has the most advanced devices and the most
advanced photonic integrated circuits on Si. Wafer-scale iii–v epi- Figure 5 | Compact hybrid Si microring lasers. a, Schematic of a hybrid
taxial transfer up to 150 mm in diameter 58 and individual iii–v dyes microring laser with a Si bus waveguide. Expanded view shows the
attaches to larger SOI wafers62 show high-volume manufacturability. simulated fundamental transverse electric mode shifting towards the
Monolithic methods are typically preferred, although the highest waveguide edge. Inset: cross-sectional SEM image of a microring laser,
quality SOI wafers today are made by wafer bonding. he ultimate showing the diferent layers and metal contacts. b, Calculated threshold
reliability, performance, uniformity and cost of the hybrid approach current as a function of device cavity length for diferent facet reflectivities
are still unknown. (R). Inset: top-view SEM image of a 15-μm-diameter device. c, A portion of
New opportunities continue to appear. Companies are now con- a 150-mm-diameter iii–v-on-SOI wafer containing ring cavity lasers (left)
sidering whether optical interconnects could be a possible solution and 1 cm2 chips containing 400 microring lasers (right).
to the problems of high power consumption and low bandwidths
of electrical interconnects, while also achieving smaller intercon-
nect delays, lower cross-talk and better resistance to electromag- photodetectors with passive multiplexers and demultiplexers,
netic interference. For example, it is challenging to extend the could be a more immediate application of existing technology. For
reach of a 10 Gb s–1 copper interconnect beyond 30 cm, even with example, a hybrid Si integrated triplexer containing a 1,310 nm
sophisticated electronic processing. Placing a wavelength-division laser for upstream data transmission as well as 1,310/1,500 nm and
multiplexing optical communication system in or between micro- 1,490/1,550 nm wavelength selective splitters and photodetectors
processors allows theoretically terahertz bandwidths for on- and for downstream digital and video reception was recently demon-
of-chip interconnections. Low-cost optical interconnects will be strated by transferring two types of iii–v epilayers onto a single
manufactured in much higher volumes than they are today when Si chip63. Quantum-well intermixing is a promising option that
applied to diverse applications such as high-deinition display avoids bonding two diferent iii–v materials simultaneously and
ports, memory server interconnects and on/of-chip intercon- so has enabled integration of hybrid lasers and modulators64. Over
nects. he emerging market of ibre-to-PC devices using diplex- time, hybrid photodetectors will probably be replaced by mono-
ers and triplexers, which require the integration of lasers and lithic Ge detectors. he ultimate integration scheme in practical

NATURE PHOTONICS | VOL 4 | AUGUST 2010 | www.nature.com/naturephotonics 515

© 2010 Macmillan Publishers Limited. All rights reserved

PROGRESS ARTICLE | FOCUS NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.167

a Active layer
Furthermore, recently demonstrated SiGe Raman ampliiers and
Tunnel junction n-type metal
lasers bring extra lexibility in the pump and signal wavelengths67.
Lateral he higher carrier mobility in SiGe reduces the carrier lifetime and
contact
subsequently the FCA. Lasing realized in ring or disk resonators
p-type
exhibits extremely useful resonance and nonlinear efects such as
SiO2 metal bistability. As shown in Fig. 5a, stimulated emission can propagate
bidirectionally (that is, clockwise or anticlockwise). Multiple and
SOI WG single iii–v ring lasers or hybrid disk lasers have been designed to
realize lip-lop memory 68,69 and wavelength conversion70. When
integrated on a single Si chip with low-loss Si or silica waveguides,
b Si ring/disk lasers promise to build ultrafast switches, bufers and
Si
complex nonlinear networks.
WG Silicon photonics is a rapidly evolving research ield with tre-
mendous potential. he term ‘semiconductor lasers’ now includes a
L4 L3 L2 L1
broad range of approaches to lasers on silicon.

References
1. Soref, R. A. & Lorenzo, J. P. All-silicon active and passive guided-wave
c -30
components for λ = 1.3 and 1.6 mm. IEEE J. Quant. Electron. 22,
L3 L2
L4 873–879 (1986).
-35 L1 3.08 mA
2.8 mA 2.35 mA 2. http://www.itrs.net/
2.65 mA
3. Pavesi, L. in Device Applications of Silicon Nanocrystals and Nanostructures (ed.
-40 Koshida, N.) Ch. 4 (Springer, 2009).
Power (dBm per 0.1 nm)

4. Bisi, O., Ossicini, S. & Pavesi, L. Porous silicon: a quantum sponge structure for
-45 silicon based optoelectronics. Surf. Sci. Rep. 38, 1–126 (2000).
5. Gösele, U. & Lehmann, V. Light-emitting porous silicon. Mater. Chem. Phys.
-50 40, 253–259 (1995).
6. Canham, L. T. Silicon quantum wire array fabrication by electrochemical and
-55 chemical dissolution of wafers. Appl. Phys. Lett. 57, 1046–1048 (1990).
7. Cullis, A. G. & Canham, L. T. Visible light emission due to quantum size efects
-60 in highly porous crystalline silicon. Nature 353, 335–338 (1991).
8. Hirschman, K. D., Tsybeskov, L., Duttagupta, S. P. & Fauchet, P. M. Silicon-
-65
based visible light-emitting devices integrated into microelectronic circuits.
Nature 384, 338–341 (1996).
9. Wilson, W. L., Szajowski, P. F. & Brus, L. E. Quantum coninement in size-
-70
selected, surface-oxidized silicon nanocrystals. Science 262, 1242–1244 (1993).
1.57 1.575 1.58 1.585 1.59 1.595 1.6 1.605 1.61 10. Pavesi, L., Dal Negro, L., Mazzoleni, C., Franzo, G. & Priolo, F. Optical gain in
Wavelength (nm) silicon nanocrystals. Nature 408, 440–444 (2000).
11. Minoru, F., Masato, Y., Yoshihiko, K., Shinji, H. & Keiichi, Y. 1.54 mm
Figure 6 | Hybrid Si microdisk laser array. a, Schematic of a photoluminescence of Er 3+ doped into SiO2 ilms containing Si nanocrystals:
heterogeneously integrated iii–v microdisk laser with a vertically Evidence for energy transfer from Si nanocrystals to Er 3+. Appl. Phys. Lett.
coupled SOI waveguide (WG) underneath. b, Microscope image of the 71, 1198–1200 (1997).
12. Iacona, F. et al. Silicon-based light-emitting devices: Properties and applications
multiwavelength laser source consisting of four microdisk lasers (L1–L4)
of crystalline, amorphous and Er-doped nanoclusters. IEEE J. Sel. Top. Quant.
with slightly diferent disk diameters coupled to a common, underlying Electron. 12, 1596–1606 (2006).
SOI waveguide. c, Balanced output spectrum obtained by adjusting the 13. Lu, Z. H., Lockwood, D. J. & Baribeau, J. M. Quantum coninement and light
individual microdisk drive currents to get four evenly spaced wavelength emission in SiO2/Si superlattices. Nature 378, 258–260 (1995).
channels. Figure reproduced with permission from ref. 60, © 2008 IEEE. 14. Cloutier, S. G., Kossyrev, P. A. & Xu, J. Optical gain and stimulated emission in
periodic nanopatterned crystalline silicon. Nature Mater. 4, 887–891 (2005).
15. Nassiopoulos, A. G., Grigoropoulos, S. & Papadimitriou, D. Electroluminescent
optical interconnect systems is likely to be a combination of hybrid device based on silicon nanopillars. Appl. Phys. Lett. 69, 2267–2269 (1996).
and monolithic approaches, thus taking full advantage of iii–v-, 16. Malinin, A., Ovchinnikov, V., Novikov, S., Tuovinen, C. & Hovinen, A.
Ge- and Si-based materials. Integration with CMOS circuits can Fabrication of a silicon based electroluminescent device. Materials Science and
provide low cost, integrated control, signals processing and error Engineering B 74, 32–35 (2000).
17. Pavesi, L. Routes toward silicon-based lasers. Mater. Today 8, 18–25 (2005).
correction. If Si photonics is to claim a large market in intrachip 18. Pavesi, L. Silicon-based light sources for silicon integrated circuits.
optical communication links, however, power consumptions must Adv. Opt. Tech. 416926 (2008).
be reduced to 2 pJ bit–1 or lower 65. 19. Lee, K. K. et al. Efect of size and roughness on light transmission in a Si/SiO2
Silicon Raman lasers are potentially ideal light sources for a waveguide: Experiments and model. Appl. Phys. Lett. 77, 1617–1619 (2000).
variety of wavelength-sensitive regimes, owing to their unmatched 20. Jalali, B. Making silicon lase. Sci. Am. 58–65 (February 2007).
wavelength purity and the possibility of extending the lasing 21. Claps, R., Dimitropoulos, D., Han, Y. & Jalali, B. Observation of Raman
emission in silicon waveguides at 1.54 μm. Opt. Express 10, 1305–1313 (2002).
wavelength into the mid-infrared region66. Example applications 22. Boyraz, O. & Jalali, B. Demonstration of a silicon Raman laser. Opt. Express
include high-resolution and ultrasensitive detection of molecules 12, 5269–5273 (2004).
for trace gas analysis, pollution and toxic gas monitoring, bio- 23. Jones, R. et al. Net continuous wave optical gain in a low loss
medical sensing, coherent free-space optical communications silicon-on-insulator waveguide by stimulated Raman scattering. Opt. Express
and metrology 25. Achieving mode-hop-free CW tuning without 13, 519–525 (2005).
sacriicing linewidth and extinction ratio is relatively straight- 24. Rong, H. et al. A continuous-wave Raman silicon laser. Nature 433,
725–728 (2005).
forward by tuning the emission wavelength of the pump laser. 25. Rong, H. et al. Low-threshold continuous-wave Raman silicon laser.
he sensing market is currently dominated by bulky and power- Nature Photon. 1, 232–237 (2007).
hungry solid-state and gas lasers, but Si Raman lasers will be very 26. Kawanami, H. Heteroepitaxial technologies of III–V on Si. Sol. Energ. Mat. Sol. C.
competitive in size and cost if a pump source can be integrated. 66, 479–486 (2001).

516 NATURE PHOTONICS | VOL 4 | AUGUST 2010 | www.nature.com/naturephotonics

© 2010 Macmillan Publishers Limited. All rights reserved

NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.167 FOCUS | PROGRESS ARTICLE
27. Xie, Y. H., Wang, K. L. & Kao, Y. C. An investigation on surface conditions 50. Park, H., Fang, A. W., Kodama, S. & Bowers, J. E. Hybrid silicon evanescent
for Si molecular beam epitaxial (MBE) growth. J. Vac. Sci. Tech. A 3, laser fabricated with a silicon waveguide and III–V ofset quantum wells.
1035–1039 (1985). Opt. Express 13, 9460–9464 (2005).
28. Samonji, K. et al. Reduction of threading dislocation density in InP-on-Si 51. Roelkens, G., Van hourhout, D., Baets, R., Notzel, R. & Smit, M. Laser emission
heteroepitaxy with strained short-period superlattices. Appl. Phys. Lett. 69, and photodetection in an InP/InGaAsP layer integrated on and coupled to a
100–102 (1996). silicon-on-insulator waveguide circuit. Opt. Express 14, 8154–8159 (2006).
29. Masafumi, Y., Mitsuru, S. & Yoshio, I. Misit stress dependence of dislocation 52. Fang, A. W. et al. Electrically pumped hybrid AlGaInAs-silicon evanescent
density reduction in GaAs ilms on Si substrates grown by strained-layer laser. Opt. Express 14, 9203–9210 (2006).
superlattices. Appl. Phys. Lett. 54, 2568–2570 (1989). 53. Fang, A. W. et al. Integrated AlGaInAs-silicon evanescent race track laser and
30. Nozawa, K. & Horikoshi, Y. Low threading dislocation density GaAs on Si(100) photodetector. Opt. Express 15, 2315–2322 (2007).
with InGaAs/GaAs strained-layer superlattice grown by migration-enhanced 54. Fang, A. W. et al. A distributed Bragg relector silicon evanescent laser.
epitaxy. Jpn. J. Appl. Phys. 30, L668–L671 (1991). IEEE Photon. Tech. Lett. 20, 1667–1669 (2008).
31. Yamaichi, E., Ueda, T., Gao, Q., Yamagishi, C. & Akiyama, M. Method to obtain 55. Fang, A. W., Lively, E., Kuo, Y.-H., Liang, D. & Bowers, J. E. A distributed
low-dislocation-density regions by patterning with SiO2 on GaAs/Si followed feedback silicon evanescent laser. Opt. Express 16, 4413–4419 (2008).
by annealing. Jpn. J. Appl. Phys. 33, L1442–L1444 (1994). 56. Van Campenhout, J. et al. Electrically pumped InP-based microdisk lasers
32. Groenert, M. E. et al. Monolithic integration of room-temperature cw GaAs/ integrated with a nanophotonic silicon-on-insulator waveguide circuit.
AlGaAs lasers on Si substrates via relaxed graded GeSi bufer layers. Opt. Express 15, 6744–6749 (2007).
J. Appl. Phys. 93, 362–367 (2003). 57. Liang, D. et al. Hybrid silicon (λ=1.5 mm) microring lasers and integrated
33. Groenert, M. E., Pitera, A. J., Ram, R. J. & Fitzgerald, E. A. Improved room- photodetectors. Opt. Express 17, 20355–20364 (2009).
temperature continuous wave GaAs/AlGaAs and InGaAs/GaAs/AlGaAs lasers 58. Liang, D. et al. High-quality 150 mm InP-to-silicon epitaxial transfer for silicon
fabricated on Si substrates via relaxed graded GexSi1−x bufer layers. J. Vac. Sci. photonic integrated circuits. Electrochem. Solid. Lett. 12, H101–H104 (2009).
Tech. B 21, 1064–1069 (2003). 59. Spuesens, T. et al. Improved design of an InP-based microdisk laser
34. Cerutti, L., Rodriguez, J. B. & Tournie, E. GaSb-based laser, monolithically heterogeneously integrated with SOI. Proc. 6th IEEE Int. Conf. Group IV
grown on silicon substrate, emitting at 1.55 um at room temperature. Photon. 202–204 (2009).
IEEE Photon. Tech. Lett. 22, 553–555 (2010). 60. Van Campenhout, J. et al. A compact SOI-integrated multiwavelength laser
35. Kunert, B., Volz, K., Koch, J. & Stolz, W. Direct-band-gap Ga(NAsP)-material source based on cascaded InP microdisks. IEEE Photon. Tech. Lett.
system pseudomorphically grown on GaP substrate. Appl. Phys. Lett. 20, 1345–1347 (2008).
88, 182108 (2006). 61. Sysak, M. N. et al. Experimental and theoretical thermal analysis of a hybrid
36. Kunert, B., Zinnkann, S., Volz, K. & Stolz, W. Monolithic integration of silicon evanescent laser. Opt. Express 15, 15041–15046 (2007).
Ga(NAsP)/(BGa)P multi-quantum well structures on (001) silicon substrate by 62. Fedeli, J. M. et al. Development of silicon photonics devices using
MOVPE. J. Cryst. Growth 310, 4776–4779 (2008). microelectronic tools for the integration on top of a CMOS wafer.
37. Yin, T. et al. 31 GHz Ge n-i-p waveguide photodetectors on silicon-on- Adv. Opt. Tech. 412518 (2008).
insulator substrate. Opt. Express 15, 13965–13971 (2007). 63. Chang, H.-H. et al. Integrated triplexer on hybrid silicon platform. Optical
38. Kang, Y. et al. Monolithic germanium/silicon avalanche photodiodes with 340 Fiber Communication Conf. paper OhC4 (2010).
GHz gain–bandwidth product. Nature Photon. 3, 59–63 (2008). 64. Sysak, M. N., Anthes, J. O., Bowers, J. E., Raday, O. & Jones, R. Integration of
39. Morse, M. et al. Monolithic Ge/Si avalanche photodiodes. Proc. 6th IEEE Int. hybrid silicon lasers and electroabsorption modulators. Opt. Express
Conf. Group IV Photon. 25–27 (2009). 16, 12478–12486 (2008).
40. Kuo, Y.-H. et al. Strong quantum-conined Stark efect in germanium quantum- 65. Miller, D. A. B. Device requirements for optical interconnects to silicon chips.
well structures on silicon. Nature 437, 1334–1336 (2005). Proc. IEEE 97, 1166–1185 (2009).
41. Roth, J. E. et al. Optical modulator on silicon employing germanium quantum 66. Soref, R. Toward silicon-based longwave integrated optoelectronics (LIO).
wells. Opt. Express 15, 5851–5859 (2007). Proc. SPIE 6898, 689809 (2008).
42. Liu, J. et al. Tensile-strained, n-type Ge as a gain medium for monolithic laser 67. Claps, R. et al. Raman ampliication and lasing in SiGe waveguides.
integration on Si. Opt. Express 15, 11272–11277 (2007). Opt. Express 13, 2459–2466 (2005).
43. Sun, X., Liu, J., Kimerling, L. C. & Michel, J. Room-temperature direct bandgap 68. Hill, M. T. et al. A fast low-power optical memory based on coupled micro-ring
electroluminescence from Ge-on-Si light-emitting diodes. Opt. Lett. 34, lasers. Nature 432, 206–209 (2004).
1198–1200 (2009). 69. Liu, L. et al. An ultra-small, low-power, all-optical lip-lop memory on a
44. Cheng, S.-L. et al. Room temperature 1.6 μm electroluminescence from Ge silicon chip. Nature Photon. 4, 182–187 (2010).
light emitting diode on Si substrate. Opt. Express 17, 10019–10024 (2009). 70. Raz, O. et al. Compact, low power and low threshold electrically pumped
45. Liu, J., Sun, X., Camacho-Aguilera, R., Kimerling, L. C. & Michel, J. Ge-on-Si micro disc lasers for 20Gb/s non return to zero all optical wavelength
laser operating at room temperature. Opt. Lett. 35, 679–681 (2010). conversion. Optical Fiber Communication Conf. paper OMQ5 (2010).
46. Jambois, O. et al. Current transport and electroluminescence mechanisms in
thin SiO2 ilms containing Si nanocluster-sensitized erbium ions. J. Appl. Phys. Acknowledgements
106, 063526 (2009). he authors thank Scott Rodgers, Ron Esman, Mike Haney and Jag Shah of DARPA/MTO,
47. Jambois, O. et al. Towards population inversion of electrically pumped Er ions the Kavli Foundation, Intel and Hewlett Packard for funding this research. Many others
sensitized by Si nanoclusters. Opt. Express 18, 2230–2235 (2010). contributed to the hybrid silicon research described here, including Hanan Bar, Dan
48. Liang, D. & Bowers, J. E. Highly eicient vertical outgassing channels Blumenthal, Hsu-Hao Chang, Hui-Wen Chen, Alexander Fang, Hyundai Park, Martijn
for low-temperature InP-to-silicon direct wafer bonding on the Heck, Richard Jones, Brian Koch, Mario Paniccia, Omri Raday and Matthew Sysak.
silicon-on-insulator (SOI) substrate. J. Vac. Sci. Tech. B 26,
1560–1568 (2008). Additional information
49. Pasquariello, D. & Hjort, K. Plasma-assisted InP-to-Si low temperature wafer he authors declare competing inancial interests: details accompany the paper at
bonding. IEEE J. Sel. Top. Quant. Electron. 8, 118–131 (2002). www.nature.com/naturephotonics.

NATURE PHOTONICS | VOL 4 | AUGUST 2010 | www.nature.com/naturephotonics 517

View publication stats

© 2010 Macmillan Publishers Limited. All rights reserved