Advanced silicon processing for active planar photonic devices
ABSTRACT Using high quality, anisotropically etched Si waveguides bonded to InGaAsP, the authors demonstrate a hybrid laser, whose optical profile overlaps both Si and III-V regions. Continuous wave laser operation was obtained up to 45 ° C , with single facet power as high as 12.7 mW at 15 ° C . Planar Si optical resonators with Q=4.8×106 are also demonstrated. By using a S F 6/ C 4 F 8 reactive ion etch, followed by H 2 S O 4/ H F surface treatment and oxygen plasma oxide, the optical losses due to the waveguide and the bonding interface are minimized. Changes of optical confinement in the silicon are observed due to waveguide width variation.
- SourceAvailable from: Gene Tsvid[Show abstract] [Hide abstract]
ABSTRACT: The integration of thin film edge emitting lasers onto silicon enables the realization of planar photonic structures for interconnection and for miniaturized optical systems that can be integrated in their entirety at the chip scale. These thin film emitters are compound semiconductor lasers that are optimized for operation without the growth substrate. Removal of the laser growth substrate, coupled with bonding to the silicon host substrate, enable the integration of high quality edge emitting lasers with silicon. This paper explores the challenges, approaches, fabrication processes, and progress in the integration of thin film edge emitting lasers integrated onto silicon.Proc SPIE 02/2010;
Advanced silicon processing for active planar photonic devices
Michael Shearn,a?Kenneth Diest, Xiankai Sun, Avi Zadok, Harry Atwater,
Amnon Yariv, and Axel Scherer
Thomas J. Watson Laboratory, California Institute of Technology, 1200 E. California Blvd., Pasadena,
?Received 8 July 2009; accepted 5 October 2009; published 7 December 2009?
Using high quality, anisotropically etched Si waveguides bonded to InGaAsP, the authors
demonstrate a hybrid laser, whose optical profile overlaps both Si and III-V regions. Continuous
wave laser operation was obtained up to 45 °C, with single facet power as high as 12.7 mW at
15 °C. Planar Si optical resonators with Q=4.8?106are also demonstrated. By using a SF6/C4F8
reactive ion etch, followed by H2SO4/HF surface treatment and oxygen plasma oxide, the optical
losses due to the waveguide and the bonding interface are minimized. Changes of optical
confinement in the silicon are observed due to waveguide width variation. © 2009 American
Integrating silicon electronics with other materials sys-
tems is an important way to extend the capabilities of the
microelectronics platform. Hybrid systems enable applica-
tions that were previously impossible or ineffective in sili-
con, such as lasers1and biosensing.2Hybrid optical systems
are of particular interest, as they use silicon’s capability for
low loss passive waveguides3but use other materials to
avoid functionality limitations from silicon’s indirect band-
gap. However, direct integration through epitaxy of these
systems is problematic due to differences in the lattice con-
stant and thermal expansion coefficients.4Recently, wafer-
bonded III-V materials on Si waveguides have produced
lasers1,5and modulators.6The bonded structure is designed
to support a joint optical mode, whose profile overlaps both
materials. This architecture is a promising step toward mono-
lithic integration of electronics with optics.
In this article, we demonstrate operation of a hybrid
Si/InGaAsP Fabry–Pérot laser and describe the processing
steps and optimization in its production. Si waveguide qual-
ity is maximized using a low damage, high fidelity reactive
ion etch, and subsequent surface treatment, characterized by
Si ring resonator measurements. Low temperature, plasma-
assisted wafer bonding7is then used to integrate the two
material systems. Finally, optical and electrical measure-
ments are made to characterize the hybrid laser.
II. FABRICATION METHODS
A. Pattern definition and plasma etching
Waveguide patterns were defined on silicon-on-insulator
?SOI? wafers with a Si layer thickness from 220 to 900 nm
on top of a 2 ?m SiO2buried oxide layer using electron
beam lithography. Zeon ZEP520A was spun as 5000 rpm
and baked at 180 °C for 20 min before exposure with a
100 keV electron beam. After development, patterns were
placed in an oven at 150 °C for 5 min to cause the resist to
reflow, reducing pattern roughness.
The waveguide patterns were transferred using an aniso-
tropic dry etch in an inductively coupled plasma reactive ion
etcher ?Oxford Instruments PlasmaLab System 100 ICP-RIE
380?, with a typical etch profile seen in Fig. 1. We utilized a
mixed mode plasma etch using SF6?an etch gas? and C4F8?a
polymer source gas?, with process conditions as described in
other publications.8In contrast to more traditional chopped
processes such as the Bosch etch,9simultaneous etching and
sidewall passivation offers improved sidewall smoothness
and reasonable etch rates at room temperature.10Also impor-
tant is the low forward rf power utilized by this etch, which
reduces roughness from mask erosion. The etch depth varied
in order to leave varying heights of Si remaining.
B. Wafer bonding, mesa definition, and metallization
After etching, the SOI wafer was cleaned by solvents and
a 3:1 mixture of H2SO4:H2O2?10 min at 170 °C?, followed
immediately by a HF dip to remove chemical oxide from the
SOI wafer and native oxide from the III-V wafer, respec-
tively. The reason for this cleaning step is twofold. First, it
removes any remaining organics that would interfere with
bonding. Second, it chemically prepares the surface for
bonding by hydrogen terminating the surface, maximizing
the effect of plasma activation.7Similar processing steps
have also shown improvements in optical properties of
waveguides through removal of absorbing surface states.11
After cleaning, the surfaces of the wafers were then acti-
vated through exposure to a low bias oxygen plasma and
bonded together under a pressure of 0.1 MPa at 350 °C for
1 h. Following the bonding, the InP substrate was removed
by HCl etching and mesa and contact structures were defined
lithographically and etched.Amesa structure, centered above
the Si waveguide, was formed in the InGaAsP layers using
photolithography and a three-phase wet etch. The etching
solutions were ?i? 1:1:10 mixture of H2SO4:H2O2:H2O
a?Electronic mail: firstname.lastname@example.org
31803180J. Vac. Sci. Technol. B 27„6…, Nov/Dec 20091071-1023/2009/27„6…/3180/3/$25.00©2009 American Vacuum Society
?p-InGaAs layer, 60 s?, ?ii? 2:1 mixture of HCl:H2O ?p-InP
layer, 30 s?, and ?iii? 1:1:10 mixture of H2SO4:H2O2:H2O
Cr/AuZn/Au for the p-side metallization ?on top of mesa?
and Cr/AuGe/Au for the n-side metallization ?to side of
mesa? were deposited by thermal evaporation. The current
flow was laterally confined by means of proton implantation
at an areal dose of 5?1014cm−2and an energy of 170 keV,
resulting in high resistance except in a 5 ?m channel. Fi-
nally, the Si substrate was thinned mechanically to increase
thermal conductivity to the testing chuck, resulting in a
structure, as shown in Fig. 2.
III. DEVICE MEASUREMENTS
A. Microring resonators
To determine the compatibility of the wafer bonding
chemical treatments with low loss photonic structures, we
fabricated bare silicon microrings using the same processing
steps used for wafer-bonded samples. The devices were fab-
ricated from a SOI wafer with a 220 nm thick Si device layer
on a 2 ?m SiO2buried oxide layer, and measured using
evanescent coupling from a tapered fiber.12On these devices,
a thin 30 nm thermal oxide was grown; this differs from
wafer-bonded devices which we discovered only required an
ultrathin oxide formed from O2plasma treatment.
The quality factor Q of the resonator was determined ex-
perimentally from the normalized fiber transmission spec-
trum under weak coupling to the ring. Due to poor phase
matching between the fiber and microring modes, coupling is
very low even when the fiber is nearly in contact with the
ring. However, in this regime the measured Q is nearly the
intrinsic Q. Figure 3 show results from a ring with Q=4.1
?106and Q=4.8?106for the short and long wavelength
modes, respectively. This corresponds with a loss of approxi-
mately 0.14 dB/cm. The two modes in the spectrum are a
result of coupling due to scattering between the otherwise
degenerate clockwise and counterclockwise circulating
modes, and are only present because of the otherwise low
loss of the waveguide structure.13,14
This initial result of a high quality planar structure, as
opposed to freestanding geometries,15led us to pursue wafer-
bonded structures that incorporate gain with the waveguide.
B. Hybrid lasers
We tested Fabry–Pérot hybrid lasers with lengths from
300 to 1500 ?m. The Si waveguides were defined on a SOI
wafer with 900 nm thick Si device layer on a 2 ?m SiO2
buried oxide layer. Results of a typical hybrid laser are
shown in Fig. 4, which shows a device with a lasing thresh-
old voltage Vthof 1.3 V and a threshold current density Jthof
1.25 kA/cm2. The maximum power output from a single fa-
FIG. 1. Profile of a fully etched disk structure on a SOI wafer with resist.
The slope of the resist is caused by reflow.
FIG. 2. Cross section of a typical bonded device. N-side contacts are to
either side of the device ?not pictured?.
FIG. 3. Measurements and schematic of a high-Q microring resonator. ?a?
Taper transmission versus wavelength of a doublet mode for a microring
with R=40 ?m, w=2 ?m tSi?200 nm, tox?30 nm. ?b? Top view of mi-
croring. ?c? Cross-sectional view of microring.
FIG. 4. L-I-V curve of a 960-?m-long laser operating in CW mode at 15 °C.
?Inset? Laser spectrum.
3181Shearn et al.: Advanced silicon processing for active planar photonic devices3181
JVST B - Microelectronics and Nanometer Structures
cet was 12.7 mW at 15 °C. The inset of Fig. 4 shows the
laser spectrum with a central wavelength of 1490 nm.
On varying the Si waveguide width w, Vthwas found to
have a local minimum at w?1.5 ?m, with dependence
shown in Fig. 5. This behavior can be understood qualita-
tively by considering the two limiting cases of waveguide
width. As width decreases, less of the mode resides in the
silicon, and thus experiences less feedback from the Si fac-
ets, which are of higher reflectivity than those in the III-V.
On the other hand, as width increases, the mode is less con-
fined to the quantum well region and thus requires a higher
pump level to reach threshold. Further work on investigating
this effect and optimizing our gain structure to utilize the
relative strengths of these two regimes is underway.
We have shown how care in silicon processing steps for
hybrid systems can lead to good optical performance. Low
loss Si waveguides can be fabricated using resist reflow, low-
bias dry etching, and surface chemical treatments. Subse-
quent integration of these optical structures at low tempera-
ture is possible with plasma-assisted wafer bonding. By
using the CMOS compatible processes described, we can
integrate optics with electronics seamlessly without signifi-
cantly altering the constraints on upstream processing steps
on the silicon. Further improvement is possible with super-
mode engineering16or new materials systems, and will make
this architecture even more attractive.
This work was supported by Defense Advanced Research
Projects Agency ?DARPA? Contract No. N66001-07-1-2058
and HR0011-04-1-0054, the U.S. Air Force Office of Scien-
tific Research ?AFOSR? Grant No. FA9550-06-1-0480, and
the Center for Science and Engineering of Materials, a Na-
tional Science Foundation ?NSF? Materials Research Science
and Engineering Center at Caltech. The authors thank the
Kavli Nanoscience Institute, Caltech, for supporting fabrica-
tion. M.S. thanks the NSF Graduate Research Fellowship
program. A.Z. acknowledges postdoctoral fellowships from
the Center for the Physics of Information, Caltech, and the
Rothschild fellowship from Yad-Hanadiv Foundation, Israel.
1A. W. Fang, H. Park, O. Cohen, R. Jones, M. J. Paniccia, and J. E.
Bowers, Opt. Express 14, 9203 ?2006?.
2R. Pantoja, J. M. Nagarah, D. M. Starace N. A. Melosh, R. Blunck, F.
Bezanilla, and J. R. Heath, Biosens. Bioelectron. 20, 509 ?2004?.
3D. A. B. Miller, IEEE J. Sel. Top. Quantum Electron. 6, 1312 ?2000?.
4H. Kroemer, T.-Y. Liu, and P. M. Petroff, J. Cryst. Growth 95, 96 ?1989?.
5X. Sun, A. Zadok, M. J. Shearn K A. Diest, A. Ghaffari, H. A. Atwater,
A. Scherer, and A. Yariv, Opt. Lett. 34, 1345 ?2009?.
6H.-W. Chen, Y.-h. Kuo, and J. E. , Opt. Express 16, 20571 ?2008?.
7D. Pasquariello and K. Hjort, IEEE J. Sel. Top. Quantum Electron. 8, 118
8M. D. Henry, S. Walavalkar, A. Homyk, and A. Scherer, Nanotechnology
20, 255305 ?2009?.
9S. Rauf, W. J. Dauksher, S. B. Clemens, and K. H. Smith, J. Vac. Sci.
Technol. A 20, 1177 ?2002?.
10H. Jansen, M. de Boer, R. Legtenberg, and M. Elwenspoek, J. Micro-
mech. Microeng. 5, 115 ?1995?.
11M. Borselli, T. J. Johnson, and O. Painter, Appl. Phys. Lett. 88, 131114
12M. Borselli, K. Srinivasan, P. E. Barclay, and O. Painter, Appl. Phys. Lett.
85, 3693 ?2004?.
13M. Borselli, T. Johnson, and O. Painter, Opt. Express 13, 1515 ?2005?.
14M. L. Gorodetsky, A. D. Pryamikov, and V. S. Ilchenko, J. Opt. Soc. Am.
B 17, 1051 ?2000?.
15D. K. Armani, T. J. Kippenberg, S. M. Spillane, and K. J. Vahala, Nature
421, 925 ?2003?; M. Cai, O. Painter, and K. J. Vahala, Phys. Rev. Lett.
85, 74 ?2000?.
16X. Sun and A. Yariv, J. Opt. Soc. Am. B 25, 923 ?2008?.
FIG. 5. Threshold voltage of hybrid lasers with various Si waveguide widths.
Measurements were taken in pulsed mode at 15 °C, with dcdenoting the
duty cycle of the applied voltage.
3182Shearn et al.: Advanced silicon processing for active planar photonic devices3182
J. Vac. Sci. Technol. B, Vol. 27, No. 6, Nov/Dec 2009