Content uploaded by Daniel Blumenthal
Author content
All content in this area was uploaded by Daniel Blumenthal on Feb 02, 2015
Content may be subject to copyright.
NOVEL APPLICATION OF QUANTUM WELL INTERMIXING IMPLANT
BUFFER LAYER TO ENABLE HIGH-DENSITY PHOTONIC INTEGRATED
CIRCUITS IN InP
Steven C. Nicholes*, Milan L. Mašanović†, Jonathon Barton†, Erik J. Norberg†,
Erica Lively†, Biljana Jevremović†, Larry A. Coldren†, and Daniel J. Blumenthal†
*Department of Materials
†Department of Electrical and Computer Engineering
University of California, Santa Barbara 93106
Abstract
We demonstrate a novel technique for free-carrier absorption reduction using an InP buffer layer with quantum
well intermixing. Application of this technique enabled fabrication of monolithic tunable optical routers with
more than 200 functions.
I. Introduction
Photonic integrated circuits (PICs) in InP provide an
attractive alternative to discrete component systems, offering
a reduction in overall system footprint, improved reliability,
and reduced packaging costs. However, as on-chip
component demands increase in both total number and
functional requirements, novel approaches to epitaxial design
and fabrication that minimize process complexity to achieve
high device yields are desirable [1]. A number of device
fabrication platforms have been reported that provide a path
for active and passive component integration [2-4]. Each
approach has advantages and disadvantages depending on the
application, but should provide an optimal combination of
high gain and low loss regions with a minimum number of
regrowths for improved yield. For instance, with the offset
quantum well platform, active and passive regions are
defined by a simple wet etch, which minimizes fabrication
complexity [2]. This approach only yields two band-edges
and therefore limits the diversity of components that can be
integrated on-chip. Alternatively, quantum-well intermixing
(QWI) can provide multiple band-edges on-chip, but this
requires additional post-growth processing [3]. Epitaxial
regrowths have also been used to increase integration
flexibility, such as butt-joint regrowths [4] for additional
band-edges on chip and unintentionally doped (UID) InP
regrowths for very low-loss propagation regions. However,
these regrowth techniques increase cost and complexity and
can result in a significant reduction in device yield.
Optical routers are one promising area in which the
footprint and power benefits of photonic integration could
make a significant impact [5]. In this paper, we discuss a new
application for a bulk InP implant buffer layer used in QWI to
realize a 640 Gbps, 8x8 monolithic tunable optical router
(MOTOR). This method allows for reduced passive loss
without a UID regrowth, thus improving device yield. The
device contains more than 200 building blocks, representing
(a)
(b)
Fig. 1. a) Schematic overview of MOTOR architecture; (b)
Photograph of fabricated device
the forefront of integration in terms of the total number of
components and functional complexity.
II. Device Integration Approach
The MOTOR chip consists of an array of eight, 40 Gbps
RZ, widely-tunable wavelength converters (WC) and an
arrayed-waveguide grating router (AWGR) (Fig. 1). The WCs
operate using cross phase modulation effects in a carrier-
based, differential Mach-Zehnder interferometer (MZI) [6].
This approach to wavelength conversion uses an integrated
differential-delay line to overcome carrier recovery time limits
in semiconductor optical amplifiers (SOA) in the MZI at 40
Fig. 2. Epitaxial base structure for quantum well intermixing
Gbps. The converted data are routed to a specific output port
of the AWGR, based upon the new wavelength set by the WC
[7]. The chip integrates active and passive components, such
as SOAs, sampled grating (SG)-DBR lasers, variable optical
attenuators (VOA), phase modulators, light couplers,
waveguides and delay lines simultaneously, thus requiring a
robust integration platform to properly place different band
edges and reduce waveguide losses. To overcome this key
challenge of integrating a low loss AWGR with the optimized
active tunable wavelength converters, we use a QWI epitaxial
design, three different waveguide architectures, and a novel,
non-regrowth approach to reduce free-carrier absorption loses
with the QWI implant buffer layer.
A. Epitaxial Platform and QWI Details
To define the active/passive regions of the chip, we utilize
an impurity-free quantum well intermixing (QWI) process [3].
The initial base growth consists of ten compressively-strained
(+0.9%) 6.5-nm InGaAsP quantum wells (QW) and eleven
tensile-strained (-0.2%) 8.0-nm InGaAsP barriers centered in a
quaternary waveguide (Fig. 2). This structure maximizes the
optical overlap with the QWs, yielding an optical confinement
factor (Г) of ~13%. Above the waveguide, a 450 nm
unintentionally doped (UID) InP buffer layer is grown to
collect the subsequent phosphorous implant for QWI.
Selective intermixing of the wells and barriers is achieved
through a rapid thermal anneal (RTA) process at 675°C to shift
the as-grown peak PL wavelength from 1545 nm to 1420 nm
as in [8]. A single, blanket, p-type regrowth is used later to
define the waveguide cladding.
B. Waveguide Designs and Implementation
Because of the diversity of on-chip functions, three
waveguide architectures are used in MOTOR: surface ridge
waveguides, deeply etched ridge waveguides, and buried rib
waveguides. All active components and most passive
waveguide regions in the WCs employ a surface ridge design
(Fig. 3a,d) for reasons explained below. This design is
accomplished using a combination of dry and wet chemical
etching. A 400-nm PECVD SiO2 hard mask is defined above
the cladding by photolithography and CHF3 inductively-
coupled plasma (ICP) dry etching. The InP ridge is etched to
a depth of roughly 1.8 µm in an ICP system using a Cl2:H2:
Ar chemistry to achieve straight and smooth sidewalls [9].
(a) (b) (c)
(d) (e) (f)
Fig 3. Ridge architectures used in MOTOR: (a), (b), (c) show schematic cross-sections of epitaxial layers with the optical mode
profile superimposed for surface ridge, deep ridge, and buried rib, respectively. (d), (e), (f) show corresponding SEM cross sections
for surface ridge, deep ridge, and buried rib, respectively
Next, the surface ridge regions are wet etched another ~0.6 µm
in a 3:1 H3PO4:HCl mixture. The quaternary waveguide
below the cladding regrowth acts as a stop-etch layer. By
etching the InP down to only the top of the waveguide, we
avoid etching the QWs and eliminate surface recombination
losses. Although the surface ridge waveguide is simple to
fabricate, there is significant modal overlap with the Zn-
doped cladding material above the waveguide, resulting in
free-carrier absorption losses.
The device also utilizes a deeply etched waveguide for the
differential-delay line to accommodate a compact structure
with tight bend radii (Fig. 3b,e). This region is fabricated by
two separate dry etches. The first 1.8-µm etch is
accomplished simultaneously with the dry-etch step of the
surface ridge waveguide. The delay line region is masked
with photoresist during the surface ridge wet etch, due to the
crystallographic nature of the wet etch chemistry. Following
the surface ridge wet etch, a lift-off process with photoresist
and 350 nm PECVD SiO2 is used to open the delay line
region. A 2-3 µm deep dry etch through the waveguide (using
identical etch conditions) is then performed.
In the AWGR, a 70-nm rib waveguide is etched into the
upper portion of the waveguide prior to the cladding regrowth,
which subsequently buries it (Fig. 3c,f). The rib allows for
large bend radii in the AWGR with low scattering losses since
the waveguide is only partially etched.
C. Novel Use for Quantum Well Intermixing Buffer Layer
In our previously reported work on QWI [2,8], the
undoped implant buffer layer grown above the waveguide is
used only for QWI purposes. The thickness of this layer is
designed to ensure that the 100-keV phosphorous implant
generates vacancies above the waveguide, avoiding damage to
the QW region. After intermixing is complete, the layer is
typically removed by wet etching. Here, however, we
deliberately leave the buffer layer in certain regions of the
chip, creating a UID setback layer between the optical mode
and the Zn dopant atoms in the p-type cladding in order to
reduce optical scattering losses. At 1.5 µm, the absorption loss
as a function of Zn concentration has been shown to be high,
according to the following [10]:
α = 20(p/1018 cm-3) cm-1. (1)
where α is the loss in cm-1 and p is the Zn concentration. By
leaving the implant buffer layer in passive sections, the Zn
doping in the cladding is pushed further away from the optical
mode and a significant reduction in loss can be achieved
without comprising the quality of active components.
Furthermore, a 200-nm sacrificial InP layer is grown above the
implant buffer layer during the initial epitaxial growth to
protect the buffer layer from implant damage and ensure a
high-quality regrowth interface (Fig. 2).
Both the deeply etched waveguide and the buried rib
waveguide designs are well suited to using the implant buffer
layer in this manner. As shown in Fig. 3b, because the deep
etch process etches completely through the waveguide
Fig. 4. Effect of the implant buffer layer thickness on the
scattering loss due to Zn doping in a buried rib and deeply
etched waveguide (model assumes no Zn diffusion)
region, the optical mode can only interact with Zn-doped
material directly above the waveguide. Therefore, the
presence of an undoped InP setback layer is expected to
provide a major improvement in optical loss. In the buried
rib region (Fig. 3c), however, the mode is not tightly confined
and expands laterally outside of the rib region. Leaving the
buffer layer directly above the rib will only reduce optical
loss over a portion of the mode volume. Because the lateral
portion of the mode remains in the Zn-doped cladding, the
expected reduction in optical loss is not as great as that in the
deeply etched waveguide.
To verify this reduction in loss, we used Eqn. 1 with 3D
beam propagation methods to simulate the expected loss due
to Zn in the cladding as a function of implant buffer layer
thickness. Our cladding regrowth utilized a graded doping
profile to ensure the formation of a p-i-n junction across the
QWs while minimizing the concentration of Zn near the
optical mode. Using these doping concentrations and
neglecting Zn diffusion during the regrowth, we can
approximate the reduction in scattering loss due to the
presence of the buffer layer. Fig. 4 shows the relationship
between loss and buffer layer thickness. As anticipated, loss
in the deeply etched region drops off more quickly than in the
buried rib structure, as there is no lateral mode interaction with
Zn dopant atoms. This result also shows that there is no
measureable advantage to increase the implant buffer layer
beyond its current thickness of 450 nm. The magnitude of
these loss values will be larger in the actual sample due to
diffusion effects, but the general trends should be unchanged.
This technique cannot be readily applied to the surface
ridge waveguide, however, because the surface ridge is
finished by wet-etching. In order to facilitate removal of the
implant buffer layer before the cladding regrowth, a
quaternary stop etch layer is included directly below the buffer
(Fig. 2). If the implant buffer layer was left in the surface
ridge region, the wet etch would stop on this quaternary layer
instead of the waveguide, resulting in a detrimental reduction
in optical confinement in these regions.
III. Device Performance
The AWGR region was first characterized using amplified
spontaneous emission (ASE) generated by forward biasing
SOAs in the WC and measuring the output in an optical
spectrum analyzer. Fig. 5 shows the ASE spectrum from input
channel #3 at every output channel of the AWGR. The free
spectral range was measured to be 11.1 nm. Next,
wavelength-based switching was examined by tuning the SG-
DBR to an allowed wavelength for each output port (Fig. 5).
Output powers of more than -5 dBm were measured in the
OSA. These powers are reasonable given the long propagation
length of the AWGR region and fiber coupling losses.
Measurements of the MZI transfer function in the WC
show more than 25 dB of extinction in the MZI for multiple
input ports (Fig. 6). Input WCs #1 and #5 had ridge defects in
the MZI and thus data are not shown for these ports. This level
Fig. 5. Wavelength switching using the SG-DBR of input
channel #3 and measuring at each output port (superimposed
on ASE spectra from all output ports)
Fig. 6. MZI transfer functions for six of the eight input
wavelength converters measured from output port #3
of extinction indicates that there is sufficient phase swing in
the MZI to perform wavelength conversion of the input data.
IV. Conclusion
We demonstrate a new technique to reduce free-carrier
absorption in passive regions of large-scale PICs by exploiting
an InP implant buffer layer used for quantum well intermixing.
The buffer layer, which is typically removed everywhere, was
left in deeply etched and buried rib waveguide regions to
provide an undoped setback layer between the optical mode
and the Zn-doped cladding. In addition to a reduction in
optical loss, this method eliminates the need for an additional
UID regrowth in the AWGR region, thus improving device
yield. Using this method, we fabricated the first 8x8
monolithic tunable optical router capable of 40 Gbps operation
per channel with more than 200 integrated functions.
Acknowledgements
This work was supported by DARPA MTO and the Army
under the DOD-N LASOR Project (W911NF-04-9-0001).
Device fabrication was done in the UCSB nanofabrication
facility, part of the NSF funded NNIN network.
References
[1] R. Nagarajan , et al., “Large-scale photonic integrated circuits,” IEEE J.
Sel. Topics Quantum Electron., vol. 11, pp. 50, Jan./Feb. 2005.
[2] M.N. Sysak, J.W. Raring, J.S. Barton, M. Dummer, D.J. Blumenthal, and
L.A. Coldren, "A single regrowth integration platform for photonic circuits
incorporating tunable SGDBR lasers and quantum-well EAMs," Photonics
Technology Letters, vol.18, no.15, pp.1630-1632, Aug. 2006.
[3] E.J Skogen, J.S. Barton, S.P. Denbaars, and L.A. Coldren, "A quantum-
well-intermixing process for wavelength-agile photonic integrated circuits,"
Journal of Selected Topics in Quantum Electronics, vol.8, no.4, pp. 863-869,
Jul/Aug 2002.
[4] J. J. M. Binsma, M. van Geemert, F. Heinrichsdorff, T. van Dongen, R. G.
Broeke, and M. K. Smit, “MOVPE waveguide regrowth in InGaAsP/InP with
extremely low butt joint loss,” in Proc. IEEE/LEOS Symp. (Benelux Chapter)
Brussels, Belgium, Dec. 2001.
[5] D. Wolfson, V. Lal, M. Masanovic, H.N. Poulsen, C. Coldren, G. Epps, D.
Civello, P. Donner, and D.J. Blumenthal, "All-optical asynchronous variable-
length optically labelled 40 Gbps packet switch," Optical Communication,
2005. ECOC 2005. 31st European Conference on , vol.6, no., pp. 49-50 vol.6,
25-29 Sept. 2005.
[6] V. Lal, M.L. Masanovic, J.A. Summers, G. Fish, and D.J. Blumenthal,
"Monolithic Wavelength Converters for High-Speed Packet-Switched Optical
Networks," Selected Topics in Quantum Electronics, IEEE Journal of , vol.13,
no.1, pp.49-57, Jan.-feb. 2007.
[7] M. Smit and C. van Dam, “PHASAR-Based WDM-Devices: Princples,
Design and Applications,” Journal of Selected Topics in Quantum Electronics,
vol 2, no. 2, 236-250, 1996.
[8] J. Raring et al., “Advanced integration schemes for high-functionality
high-performance photonic integrated circuits,” presented at the Proc. SPIE,
San Jose, CA, Jan. 23–25, 2006, Paper 61260H.
[9] F. Karouta, B. Docter, E. Kok, E.-J. Geluk, J. Van der Tol, and M. Smit,
“High Aspect Ratio Etching and Application in InP-Based Photonic Integrated
circuits,” Meet. Abstr. - Electrochem. Soc. 802, 2050 (2008).
[10] H. C. Casey Jr. and P. L. Carter, “Variation of intervalence band
absorption with hole concentration in p-type InP,” Appl. Phys. Lett., vol. 44,
no. 1, pp. 82-83, 1984.
