 |
Description  |
|
|
BACKGROUND OF THE INVENTION
The invention relates to optoelectronic devices, and, more particularly, to
integrated optoelectronic transmitters and receivers and methods of
fabrication.
Optical communication systems have advantages over all electrical
communications systems with regard to bandwidth, noise immunity,
electromagnetic susceptibility, size, and weight are being extensively
developed. An optical source converts input electrical information signals
into amplitude modulated light for transmission over an optical
communications channel, and an optical detector reconverts the amplitude
modulated light into electrical signals for reception. Optical
communication systems typically employ semiconductor laser sources, glass
optical fiber communication channels, and various modulators and
detectors. Optical wavelengths on the order of 1 .mu.m permit use of
GaAs-Al.sub.x Ga.sub.1-x As diode lasers and silicon photodetectors. FIG.
1a heuristically illustrates a one-way communication system with
continuous wave laser diode source 102 modulated by Mach-Zehnder
interferometer 104 which incorporates phase modulator 106, optical fiber
communication channel 108, and photodetector 110. For amplitude-modulated
(analog) light signals, continuous wave lasers with following light
amplitude modulators, such as the Mach-Zehnder interferometer, are
preferred over just directly modulating the input power of the laser
because the laser and the modulator may be separately optimized. A full
duplex communication system would have a duplicate of the system on FIG.
1a for communication in the opposite direction. Single-optical-fiber
two-way communication could use separate optical wavelengths for the two
directions, but this would require wavelength filters. Alternatively, the
optical fiber could be split, as shown by splitter 112 in FIG. 1b, but
this results in a 50% loss in light intensity to detector 110.
Light modulators may be made from numerous materials in various structures
and have various modulation effects. In particular, reflection and phase
modulators vary the index of refraction of material in a light path, and
absorption modulators vary the absorptance of material in the path.
Materials such as ferroelectrics, organic polymers, and semiconductors
have been used in modulators. Index of refraction modulation can be had
from the Pockels effect, plasma effect, band-filling effect, quantum
confined Stark effect, magneto-optic effect, and acousto-optic effect;
absorption modulation can be had from the Franz-Keldysh effect, quantum
confined Stark effect, and Wannier-Stark localization. For example, Weiner
et al., Quadratic Electro-optic Effect due to the Quantum-confined Stark
Effect in Quantum Wells, 50 Appl.Phys.Lett. 842 (1987). compute change of
the index of refraction of a quantum well with a change of applied
electric field based on experimentally measured change of absorption. The
index of refraction and the absorption are related (Kramers-Kronig
relations) due to the causality of electric field induced dielectric
polarization. This allows computation of chirp in a quantum well-based
absorption modulator. FIGS. 2a-b show quantum well absorption for applied
electric fields of 0 and 65000 volts/cm together with the change in index
of refraction for perpendicular and parallel polarized light,
respectively, as a function of incident photon energy. Note that large
absorption changes enable absorption modulator operation and large index
or refraction changes enable electro-optic devices such as directional
couplers and modulators. Further increases in the electric field across
the quantum well spreads out and shifts the absorption peak as illustrated
in FIGS. 2c-d, taken from Weiner et al., Strong Polarization-sensitive
Electroabsorption in GaAs/AlGaAs Quantum Well Waveguides, 47 Appl. Phys.
Lett. 1148 (1985).
Semiconductor lasers in the form of heterojunction diodes with quantum well
active regions and made of materials such as Al.sub.x Ga.sub.1-x As with
GaAs quantum wells provide a compact and rugged source of infrared light
which can be easily modulated by simply varying the diode current. In
particular, a stripe geometry diode laser may be small (e.g., 5l .mu.m
wide by 100 .mu.m long with a 30 nm thick active area imbedded in a 400 nm
thick optical core). The reflecting ends of the lasing cavity may be
distributed Bragg reflectors to avoid cleaved mirror ends. See for
example, Tiberio et al., Facetless Bragg Reflector Surface-emitting
AlGaAs/GaAs Lasers Fabricated by Electron-beam Lithography and Chemically
Assisted Ion-beam Etching, 9 J.Vac.Sci.Tech. B 2842 (1991).
SUMMARY OF THE INVENTION
The present invention provides two-way optical communication with a
receiving detector which may be made transparent for transmission.
Preferred embodiment integrated circuits include a waveguide with a series
combination of a transparentable detector, an amplitude modulator, and a
laser source for both reception and transmission.
The present invention provides technical advantages including a detector
compatible with same wavelength transmission without a 50% splitting loss.
BRIEF DESCRIPTION OF THE DRAWINGS
The drawings are schematic for clarity.
FIG. 1a-b show known optical communication systems.
FIGS. 2a-d illustrate quantum well absorption and index of refraction
variation.
FIG. 3 is a plan view of a first preferred embodiment optical transceiver.
FIGS. 4a-b are plan and cross sectional elevation views of the first
preferred embodiment.
FIG. 5 is a cross sectional elevation view of an integrated circuit
waveguide.
FIGS. 6a-b show optical confinement in the waveguide.
FIGS. 7-8 are band diagrams illustrating absorption and emission.
FIG. 9 illustrates a laser with distributed Bragg reflectors and waveguide.
FIGS. 10a-b show preferred embodiment phased array radar systems.
FIG. 11 is a plan view of a second preferred embodiment transceiver.
FIGS. 12a-f illustrate steps in a preferred embodiment method of
fabrication.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Overview
FIG. 3 shows in heuristic plan view first a preferred embodiment optical
transceiver integrated circuit, generally indicated by reference numeral
300, including optical waveguide 302 with incorporated multiple quantum
well (MQW) optical detector 310, interferometer 320 including phase
modulator 322 for amplitude modulation, diode laser 330, dc bias plus
output microstrip line 316 with inductor 318 for detector 310, and dc bias
plus input microstrip line 326 with inductor 328 for modulator 322.
Communication channel optical fiber 350 has a lensed end and butt couples
to optical waveguide 302 on integrated circuit 300. Electrical transmit
and receive lines connect to microstrips 326 and 316, respectively, and
bias (control) lines connect to the detector and modulator through the
inductors. Other bias connections, such as resistors, could be used in
place of the inductors.
Transceiver 300 operates as follows. To receive information in the form of
amplitude modulated (analog) light of a single frequency (single free
space wavelength) on optical fiber 350, the dc bias on MQW detector 310 is
set so detector 310 is absorbing and laser 330 is turned off. Then the
light from optical fiber 350 couples into waveguide 302 at the edge of the
integrated circuit and proceeds to detector 310 where it is absorbed and
creates an output electrical signal on microstrip 316. Detector 310
basically is a reverse-biased photodiode, so detector 310 outputs a
current with amplitude proportional to the amplitude of the input light.
The amplitude of the input light has frequencies on the order of MHz to
GHz, so inductor 318 prevents coupling of the output current to the dc
source, and the output current connects to the receive line.
To transmit information, which is in the form of electrical signals at the
transmit line connected to microstrip 326 and with frequencies on the
order of MHz to GHz, change the dc bias of detector 310 to make it
transparent, turn on laser 330 (which is a continuous wave diode laser),
set the dc bias of MQW phase modulator 322, and apply the transmit signal
to modulator 322. The light emitted by laser 330 splits into the two
branches of waveguide 302 making up interferometer 320, and the light
traveling the branch with modulator 322 sees a varying path length because
the transmit signal varies the index of refraction of the portion of
waveguide 302 within modulator 322. This varying path length implies that
the light in the branch with modulator 322 is phase shifted with respect
to the light in the other branch, and thus recombining the light from the
two branches produces interference and results in amplitude modulation.
The amplitude modulated light then passes through transparent detector 310
and out into optical fiber 350. The change of detector 310 from absorbing
to transparency by change of the dc bias permits the series connection of
detection and transmission on the same waveguide. The following sections
describe the components of transceiver 300 in more detail.
Detector
FIG. 4a is a plan view of detector 310 and FIG. 4b is a cross sectional
elevation view of along section b--b of FIG. 4a. Detector 310 is formed
from a portion of waveguide 302 by isolating the metal contact for
application of bias voltage and receipt of output signal current. In
particular, FIG. 5 shows a cross sectional elevation view of waveguide 302
which consists of 0.1 .mu.m thick metal top contact 502, 0.1 .mu.m thick
p+ GaAs contact layer 504, 1.5 .mu.m thick p+ Al.sub.0.3 Ga.sub.0.7 As
("AlGaAs") cladding 506, 0.48 .mu.m thick undoped multiple quantum well
("MQW") 508 core, 1.5 .mu.m thick n+AlGaAs 510 cladding, 1 .mu.m thick
n+GaAs 512 contact layer, and semi-insulating GaAs substrate 514.
Beryllium at 10.sup.9 atoms/cm.sup.3 and 10.sup.8 atoms/cm.sup.3 provides
the p+ type doping for the GaAs and AlGaAs, respectively, and silicon at
8.times.10.sup.7 atoms/cm.sup.3 and 2.times.10.sup. 18 atoms/cm.sup.3
provides the n+ type doping for the AlGaAs and GaAs, respectively. MQW 508
consists of 37 periods of a 3.5 nm thick GaAs quantum well with a 10 nm
thick AlGaAs tunneling barrier. Metal contact 502 is 3 .mu.m wide; and the
portion of p+ AlGaAs 506 directly under contact 502 forms a ridge to
define the location of the waveguide core. The "shelf" portion of p+
AlGaAs 506 extending beyond contact 502 is roughly 0.2 .mu.m thick.
The index of refraction of GaAs (roughly 3.6 for infrared used for
communication) is larger than that of AlGaAs (roughly 3.4), so light is
confined vertically by the MQW, although significant intensity exists
outside of the MQW as illustrated by the electric field density as a
function of vertical distance in FIG. 6a. Further, the index of refraction
under the AlGaAs ridge structure is greater than the effective index of
refraction under the shelf adjacent to the ridge, so the ridge generates
lateral confinement of the light and defines a core location. FIG. 6b
illustrates the intensity as a function of distance perpendicular from the
ridge. The broken line in FIG. 5 suggests a core location about 3 .mu.m
wide by 0.5 .mu.m thick. The n+ AlGaAs and n+GaAs connect to ground
through vias (not shown in the Figures) to backside metal ground plane 516
on substrate 514. Metal contact 502 also connects to ground for waveguide
302.
Detector 310 is a 300 .mu.m long section of waveguide 302 defined by two
gaps 450-452 in metal contact 402 on top of p+ AlGaAs ridge 406 and in the
ridge itself. Thus the detector portion of the p+ AlGaAs ridge can be dc
biased negative with respect to the n+ cladding to create a reversed
biased p-i-n diode for detection capability. Also, topside metal contact
430 to the n+ AlGaAs cladding 410 insures good ground contact. The layer
of p+ GaAs 404 on top of the ridge improves ohmic contact to the metal
402, and the n+ contact may be to the n+ GaAs 412 under the n+ AlGaAs
cladding 410 to also improve ohmic contact. As FIG. 4a illustrates, air
bridge 422 connects metal contact 402 on top of the ridge to microstrip
line 316. Microstrip line 316 is in the form of a "V" so that one branch
may output the detected signal and the other branch may include a
termination network. Air bridge 422, microstrip line 316, and ground
contact 430 are electroplated gold and may be several micrometers thick.
Microstrip line 316 is separated from backside ground plane 416 by about
500 .mu.m of semi-insulating GaAs substrate 414 and thus has a
characteristic impedance of 50 ohms.
Application of a negative dc bias voltage of-10 to -12 volts through
inductor 318 to microstrip 316 reverse biases the p-i-n diode formed by p+
AlGaAs 406, MQW 408, and n+ AlGaAs 410. The 10 volt drop appears across
the undoped MQW and thus creams an electric field of about
2.times.10.sup.5 volts/cm across MQW 408. FIG. 7 shows the band diagram
and heuristically illustrates a 1.44 eV photon (free space wavelength of
0.863 .mu.m) being absorbed to create an electron-hole pair with the
electron collected in the n+ AlGaAs and the hole collected in the p+
AlGaAs to contribute to the output signal current. The MQW absorption
coefficient for photons of energy 1.44 eV when an electric field of
2.times.10.sup.5 volts/cm is applied across the MQW is roughly 3000/cm.
Thus detector 310 essentially absorbs 100% of the incident light. The
recombination time of the electron-hole pair in the p-i-n diode is very
fast, so variations in amplitude of the incoming light up into the GHz
range will generate corresponding variations in the output current on
microstrip line 316.
Contrarily, when the electric field is removed, the absorption coefficient
for 1.44 eV photons drops to roughly 2/cm, and detector 310 is essentially
transparent. This is used for the transmit mode of operation. FIG. 2d
shows the absorption curves for various applied electric fields from
1.6.times.10.sup.4 volts/cm to 2.2.times.10.sup.5 volts/cm. For photons
with energy about 1.44 eV, the absorption changes drastically with applied
electric field.
The connection of microstrip line 316 to detector 310 could have different
geometries, such as a travelling wave connection. This would compensate
for the highly capacitive nature of detector 310 by introducing inductors
with the artificial travelling wave structure and better matching to the
50 ohm microstrip line 316. A backside ground plane on the semi-insulating
GaAs substrate connects to the top side metal contacts on n+ GaAs.
Phase modulator
Phase modulator 322 has the same structure as detector 310, but is
negatively dc biased at about -5 volts which implies an electric field of
about 1.times.10.sup.5 volts/cm across the MQW. This yields a quadratic
electro-optic coefficient of about 5.times.10.sup.-20 m.sup.2 /V.sup.2
which implies index of refraction changes on the order of 0.001 for
applied signal voltages of a few volts. Thus phase modulator 322 with a
length of 500 .mu.m can shift the phase of incident light by up to .pi..
Amplitude modulation
Interferometer 320 provides amplitude modulation for the continuous output
of diode laser 330 by phase modulator 322 shifting the phase of the light
traversing the upper waveguide branch relative to the light traversing the
lower branch. Thus the light from upper branch interferes with the light
from the lower branch upon recombination to yield an amplitude modulated
output from interferometer 320. Each waveguide branch of interferometer
320 has a length of about 2000 .mu.m with phase modulator 322 occupying
about 500 .mu.m of the center of the upper branch.
Other approaches to amplitude modulation could also be used. For example,
the electrical input signal on microstrip 326 could directly control the
current through diode laser 330 and thus the amplitude of the output
light, thereby eliminating the interferometer.
Laser
The gain portion of diode laser 330 has the same multiple quantum well
structure as detector 310 and phase modulator 322. In contrast to detector
310 and phase modulator 322, diode laser 330 is forward biased to create
electron-hole recombination in the quantum wells and thus generation of
photons. Stimulated emission leads to lasing when the electron and hole
densities are above a threshold. FIG. 8 shows the band diagram; note the
analogy to FIG. 7. Rather than having mirror facets, laser 330 uses
distributed Bragg reflectors; and the periodicity of the Bragg reflector
fingers determines the wavelength (and frequency) of the lasing. A free
space wavelength of 0.863 .mu.m translates to a wavelength of roughly 0.24
.mu.m in the core; thus the Bragg reflector fingers have a periodicity on
the order of 240 nm. The distributed Bragg reflectors are firmed in the
AlGaAs cladding layers as illustrated in cross sectional elevation view in
FIG. 9. The Bragg reflectors each has a length of about 500 .mu.m and the
gain portion diode has a length of about 500 .mu.m.
Varying the quantum well thickness will move the quantized levels and
thereby permit adjustment of the recombination energy and shift the curves
of FIGS. 2a-d and thus allow for operation at different wavelengths.
Alternative material systems may be used for the multiple quantum-well
common to the detector, phase modulator, and laser and also give different
wavelength operation. For example, PbTe-Pb.sub.1-x Sn.sub.x Te material
systems are well known. Further, there are five strained material systems
(both compressive and tensile strain) for various wavelength lasers: (1)
In.sub.x Ga.sub.1-x As/InAs.sub.y P.sub.1-y, (2) InP, and (3) Al.sub.x
Ga.sub.y In.sub.1-x-y As/InP for telecommunications (typically infrared
from 0.8 to 1.55 .mu.m free space wavelengths, especially at 0.8, 1.3, and
1.55 .mu.m which are popular communications and fiber wavelengths), (4)
AlGa(In)As(P)/GaAs for near infrared, and (5) AlGaInP/GaAs for visible
light. Also, a CdZnSe quantum well in a ZnSe core with ZnSSe cladding can
provide blue-green lasers.
Phased Array
FIG. 10a heuristically shows transceiver 300 used in a phased array radar.
Timing signals as well as RF signals may be sent from controller 1010 over
optical fiber 1020 to each of the transmit/receive modules 1030 to steer
transmitted radar beams, and the received signals sent back on the same
optical fiber network. The detectors in the transceivers 300 receive the
control signals which drive the power amplifiers powering the antennas.
The received radar returns picked up by the antennas are electrically
amplified and used to control the phase modulators in the transceivers
300. This transmits the analog received signals back along the optical
fiber to the detector in controller 1010. The physical separation of the
antennas creates a time separation of the return signals so they appear
serially on the optical fiber, multiplexed it time.
An alternative configuration appears in FIG. 10b and uses two optical
fibers to each transceiver 1060. For transmission from controller 1050 to
the transceivers 1060 the output of laser 1052 is amplitude modulated by
modulator 1054 and the detectors 1066 in the transceivers 1060 convert the
amplitude modulation into RF electrical output. Contrarily, for
transmission from the transceivers 1060 to the controller 1050, the output
of laser 1052 is held constant, modulators 1064 in the transceivers
amplitude modulate this light, and detector 1056 in controller 1050
converts the amplitude modulation into electrical signals. In effect
optical fiber and splitter 1072 is used for signal transmission from
controller 1050 to transceivers 1060 and also for supplying the light for
signal transmission from the transceivers to the controller, and optical
fiber and combiner 1074 is used for signal transmission from the
transceivers to the controller. In effect, the second fiber replaces the
on board laser of the transceiver. Thus one large laser supplies all of
the transmission light for all of the transceivers in the array, reducing
cost and dc power requirements.
Second Preferred Embodiment
FIG. 11 illustrates a second preferred embodiment transceiver, generally
denoted by reference numeral 1100, in plan view as including
interferometer 1120 with modulators 1122 in the upper branch and modulator
1123 in the lower branch and diode laser 1130. For detection, both
modulators 1122 and 1123 are dc biased into high absorbing condition, so
even though each modulator arm only detects 50% of the incoming light due
to the split by the interferometer, the two RF signals are coherently
added at the outputs of the modulators and maintain full signal strength.
For transmission, modulator 1123 in the lower branch is transparent and
inactive while modulator 1122 in the upper branch is driven by the output
signal. Thus the interferometer operates as the interferometer of
transceiver 300 for transmission.
Fabrication
FIGS. 12a-f show in cross sectional elevation view steps of a preferred
embodiment method of fabrication.
(1) Begin with a semi-insulating GaAs wafer with (100) orientation and 500
.mu.m thickness. Then grow (such as by metalorganic chemical vapor
deposition or molecular beam epitaxy) the epitaxial layers as listed in
following Table I; note that AlGaAs denotes Al.sub.x Ga.sub.1-x As for X a
fraction such 0.3, and that p type doping is typically with beryllium and
n type doping with silicon.
TABLE I
______________________________________
Layer Material Thickness
______________________________________
Contact p+ GaAs 0.1 .mu.m
Cladding p+ AlGaAs 1.5 .mu.m
Spacer AlGaAs 100 nm
Quantum well GaAs 3.5 nm
Barrier AlGaAs 10 nm
Quantum well GaAs 3.5 nm
Barrier AlGaAs 10 nm
. . .
. . .
. . .
Quantum well GaAs 3.5 nm
Barrier AlGaAs 10 nm
Quantum well GaAs 3.5 nm
Barrier AlGaAs 10 nm
Quantum well GaAs 3.5 nm
Spacer AlGaAs 100 nm
Cladding n+ AlGaAs 1.5 .mu.m
Cladding n+ GaAs 1 .mu.m
Substrate s.i. GaAs 500 .mu.m
______________________________________
The . . . indicate more quantum well-barrier pairs for a total of 37
quantum wells and a total thickness of the multiple quantum well (MQW)
structure of 0.5 .mu.m. The spacers keep the dopants out of the quantum
wells.
(2) Photolithographically define the locations of the 3-.mu.m wide metal
contacts (402, 502,902) on the ridge for the waveguide, detector, phase
modulator, and laser plus reflectors; this includes definition of the end
gaps. Then deposit the metal contacts by liftoff. The metal may be 300 nm
of sputtered Ti, Ni, and Pt. See FIG. 12a.
(3) Use the deposited metal contacts as a mask for etching the p+ GaAs and
p+ AlGaAs with a plasma of BCl.sub.3. The etch is timed to leave about 0.2
.mu.m of p+ AlGaAs over the MQW as shown in FIG. 12b. The length of the
end gaps can be increased to increase the resistance between the grounded
p+ AlGaAs of the waveguide and the negatively biased p+ AlGaAs of the
detector and modulator or positively biased p+ of the laser. Further, a
silicon implant can counterdope and decrease the conductivity of the p+
AlGaAs at the end gaps.
(4) Remove the metal contact over the portion of the ridge at either end of
the laser to form the distributed Bragg reflectors; this may be performed
with a noncritical photoresist mask plus wet metal etch. Then use
holographic lithography to define the fingers for the distributed Bragg
reflectors on the ends of the laser ridge where the metal contact was
removed. Note that the period of the reflectors determines the lasing
frequency. In holographic lithography the areas away from the reflectors
being defined are masked off and a thin layer of photoresist is applied to
the reflector areas; two laser sources are used to create the interference
pattern for the exposure of the photoresist. Then the p+ AlGaAs ridge is
etch with the holographically patterned photoresist as mask to form the
fingers of the reflectors. The depth of the reflector fingers and their
distance from the core could be varied: an upper portion of the ridge
could be etched away prior to the holographic lithography to move the
fingers closer to the core.
(5) Photolithographically define the extension of the MQW laterally from
the ridge (about 3-5 .mu.m) and etch down into the n+ AlGaAs with a timed
plasma. Continuing the etch down into the n+ GaAs permits the topside
ground contact to be to n+ GaAs rather than to n+ AlGaAs. In either case,
the depth of the etch is not critical because the thickness of the n+
AlGaAs and n+ GaAs are thick enough. See FIG. 12c.
(6) Photolithographically define the extension of the n+ AlGaAs/GaAs
laterally from the ridge (about 10 .mu.m on the microstrip side and enough
for the topside contact on the other side) and etch down into the
semi-insulating GaAs substrate with a timed plasma. See FIG. 12d which
illustrates the n+ extending to the right hand portion of the Figure to
accommodate the topside metal ground contact.
(7) Photolithographically define the locations of the microstrip lines and
the topside ground contacts. Then deposit metal contacts by liftoff. The
metal may be 300 nm of sputtered Ti, Ni, and Pt. See FIG. 12e. Connections
from the backside ground plane to the n+ AlGaAs/GaAs are made by a via
etched through the semi-insulating substrate prior to the deposition of
the ground plane, so the ground metal will line the vias.
(8) Photolithographically define the location for electroplating air
bridges and the microstrip lines and topside contacts. Lastly,
electroplate gold to form the air bridges and microstrip lines and topside
contact; see FIG. 12f.
Modifications and advantages
The preferred embodiments may be varied in many ways while retaining one or
more of the features of a detector switchable between transparent and
absorbing and an integration of detector and modulator (and laser) for
single wavelength two-way optical communication.
For example, the dimensions of the devices and waveguide could be varied,
the frequency and wavelength of the light could be varied. Systems of
arrays of transceivers other than radar could use the communication. The
materials could be changed such as an InP substrate in place of the GaAs
substrate, or other GaAs on silicon, or other such.
* * * * *
|
|
 |
|
 |
Description |
|
