25 atoms thick. This oxide layer acts as the glue that fuses the
two materials together. When a forward bias is applied to the
indium phosphide, it emits photons into the silicon-waveguide
resonator, where they accumulate and subsequently pass into
the next silicon-photonic device.
We have monolithically integrated four hybrid silicon lasers
onto a single silicon substrate with four traveling-wave optical
modulators, a multiplexer and an inverted taper that couples
light into a fiber. The four lasers were designed so that their
wavelengths matched the ITU-T G.694.2 grid wavelengths of
1,291, 1,311, 1,331 and 1,351 nm.
The InP bonding can be done at the wafer level in the
back end of the silicon processing flow. Therefore, hundreds
or thousands of hybrid silicon lasers could be created from a
single bond, enabling a high-volume path to light emission on
a silicon photonics chip.
Packaging and assembly elements
for the transmitter module.
Courtesy of Intel Corp.
for transmission. An optical connector that includes a plastic
lens and alignment pin couples light into fiber.
After passing the output grating and exiting the laser resonator,
the light enters a traveling-wave silicon modulator. The modulator shutters the light on and off to encode ones and zeros on
the beam at rates up to 12. 5 Gbps. There are two reasons for
using these modulators rather than directly modulating the
laser. First, at high data rates, modulating the laser directly
would consume much more power. Second, direct modulation
of a laser introduces chirp, which diminishes performance at
high speeds. Since the long-term plan is to produce devices
running at 40 Gbps per channel and beyond, this is essential
for future scaling.
The silicon modulator relies on the free-carrier plasma-dispersion effect, in which silicon’s refractive index is changed
when the density of free carriers (electrons and holes) is varied.
This modulation can be very fast, because free carriers can be
swept out of the junction in approximately seven picoseconds.
The speed is thus limited by the parasitic effects such as RC
time constant limits. Accordingly, we adopted a traveling-wave
scheme allowing the electrical and optical signal to copropa-gate across the device. To operate the traveling-wave modulator, our driver IC feeds a small RF signal into the transmission
line, and the transmission line is terminated with an external
resistor. We have shown in previous experiments that such
modulators can have a 3-dB bandwidth of about 30 GHz and
data transmission capability up to 40 Gbps.
Light beams from the four modulators are combined into
a single output beam with an optical multiplexer. Both the
transmitter’s multiplexer and the receiver’s demultiplexer use
an integrated optical grating. In the transmitter, this consists
of a free-space region inside the chip where light beams reflect
off the curved surface and the grating, and they are combined
onto a single waveguide. The light then leaves the chip through
a taper, a passive device that gradually changes the size of the
optical beam to couple more efficiently into the optical fiber
The optical signals are carried by a standard optical fiber and coupled into the Rx chip through a second optical connector. Signals
travel into the demultiplexer to be separated once again into
individual channels. The alignment of the wavelength channels
between the hybrid silicon laser, the multiplexer and the demultiplexer is within one nanometer with only passive heat sinking.
An array of four photodetectors converts the photons back into
electrons. We use germanium-on-silicon detectors to receive
optical data because germanium is highly compatible with
silicon IC manufacturing, and it detects light efficiently at the
wavelengths emitted by the lasers. The challenge with these
devices is minimizing the dark current due to crystal-lattice
differences between germanium and silicon. For the 50 Gbps
link, the measured 3-dB bandwidth of the four photodetectors
was in the range of 9. 4 to 10. 6 GHz, with a measured responsiveness of approximately 0.9 A/W. As with the modulators, we
have previously demonstrated that individual PIN photodetectors based on this technique can operate up to 40 Gbps.
Packaging and assembly
We packaged the Tx and Rx chips onto two different printed
circuit boards. The system is designed to operate at 10 Gbps
at room temperature with a passive heat sink. The Tx is “
flip-chip” mounted to the substrate along with a silicon driver IC.
Flip-chip bonding is the standard method of packaging modern microprocessors; it was developed as a high-speed alternative to wire-bonding. In this type of bonding, the silicon chip
is flipped upside-down and placed on an array of solder bumps
that form the connection between circuitry and the package.
The integrated receiver is flip-chip attached to a second
substrate, which is copackaged with a commercially available
four-channel receiver IC. Fiber coupling between the two chips
is aligned passively with precise metal alignment pins seated into
v-grooves in the silicon substrate. These pins allow the chips to
28 | OPN Optics & Photonics News