History of Guided Wave All Optical Logic
by David D. Olmsted (Copyright - 2000, 2006. Free to use for personal and
educational purposes)
Last Revised November 6, 2006
All-optical wave guide logic has been slow
to develop due to the problem of phase interference. The earliest research involved
using one optical pulse to switch a second optical pulse from one output branch
into another in what is in effect an AND operation. These switching devices and
later two input logic devices avoided destructive phase interference by separating
the two optical inputs either by frequency or by polarization and then used an indirect
nonlinear method to control their interaction. One such indirect method is
the Kerr effect, a nonlinear property of certain optically active materials which light is able to change the refractive index of the optical medium in proportion to the intensity
of the signal’s own optical field or in proportion to an externally applied electrostatic
field. This in turn changes the velocity of the signal thus resulting a phase shift
proportional to the distance.
One common problem of these early Kerr effect devices was that
the use of two different types of input signals (separated either by polarization
or frequency) means that the operations are generally not cascadable unless additional
and more complex components are added to discard one of the polarizations or frequencies.
In practice this usually meant discarding the high powered pulse responsible for
producing the Kerr effect resulting in very low power efficiency for those devices. The other is that
the product of optical power and wave guide length was often excessively large.
(For a tutorial review of all-optical wave guide switching see Stegeman and Wright
(1990).
One of the first two input Kerr effect logic devices was the EXCLUSIVE OR
operation based upon the three input Mach-Zehender interferometer of Lattes et al
(1983). The middle input arm is fed a continuous stream of optical pulses which
are split and then recombined at the output where they destructively interfere.
If one of the other two arms now has a high power pulse (pump pulse) synchronized with the
middle pulse stream and if the pump pulse power level is just right and if the pump
pulse is cross polarized then one arm will experience a 180 degree phase shift due
to the Kerr effect such that the pulse streams now constructively interfere producing
an output pulse. A polarization filter is placed at the output to block the pump
pulse.
One problem unique to large sized Mach-Zehnder Kerr effect devices of that
time is that they had to be phase tuned using an externally applied electric field
to compensate for thermal effects and unequal line lengths of the arms. One thermal
effect arises from the unequal heating of the arms which changes both the refractive
index and the line length. Even if the temperature change is equal on both arms
a mismatched arm length of even a few hundred wavelengths will be enough to produce
a significant phase shift. A better solution was the optical loop mirror.
The optical
loop mirror or Sagnac interferometer was first proposed in a generic form by Doran
and Wood (1988). It consists of a looped optical wave guide with some a directional
coupler at its neck. The coupler splits the input signal in half such that the two
half signals flow around the loop in opposite directions. The coupler retards the
phase of the crossover signal by a quarter of a wavelength so that upon recombining
at the neck the signal constructively interferes (adds) at the input side but destructively
interferes at the output side. Consequently, the signal energy appears to be reflected
back out the input side. Yet if the material of the loop is optically active any
field whether from an additional high intensity light pulse or an externally applied
electrostatic field will retard the velocity of the signal thus changing the phase
of one of the half pulses at the neck and thus prevent a return signal from being
produced at the input. The principle advantage of the optical loop mirror is that
the one input has a greater resistance to phase misalignment since both half signals
traverse the same path. Yet the question remained as to the best way to use another
input to change the phase of one of the half signals which produced a variety of
papers and patents culminating in a patent by Gabriel, Houh, and Whitaker in 1992
(5,144,375) which also has other references.
The logical possibilities of the Kerr
effect optical loop mirror were explored by Jinno and Matsumoto (1991) and shown
in figure 1. It had three inputs: a low power clock pulse (24 milliwatts) input
which ended up as the output and the two high powered (0.5 to 0.9 watts) oppositely
polarized logical inputs. Because of their separate polarizations each logical input
entered the loop on its own side and traveled around the loop in opposite directions.
The clock pulse is polarized at a 45 degree angle so the coupler at the neck of
the loop splits it in half so that it travels in both directions around the loop.
If only one of the logical inputs is present (the EXCLUSIVE OR condition) then it
affects the clock pulses unequally causing the recombined clock pulse to exit opposite
its input port. Yet if both logical pulses are present (the AND condition) then
the recombined clock pulse will exit out its input port. Because high powered logic
pulses are use the loop was only 200 meters in length but poor contrast in the EXCLUSIVE
OR case caused the authors recommend a longer fiber loop (>500 meters).
Figure 1
Jino and Matsumoto’s Sagnac Interferometer with Two Control Beams (Jinno and Matsumoto - 1991)
P.B.S. - polarization beam splitter, F.R. - 45o faraday rotator, D.M. dichroic mirrors (pass one polarization but not the other), lambda/2 - half wave plates.
 |
The problem
with optical loop Kerr effect devices was that these early devices used long loops
of glass fiber. During the early 90’s more compact devices were developed
based upon the non-linearity inherent in GaAs semiconductor optical amplifiers (SOA's) caused
when the carriers are depleted by a gain saturating light pulse. These devices came
to be called TOAD's (figure 2) for Terahertz Optical Asymmetric Demultiplexers after
Sokoloff, et al. (1993). Yet their use as fast all-optic AND operation logic devices
remains doubtful for the SOA saturating pulse and the SOA’s injection current must
be precisely controlled to achieve the proper phase shift. This means that the devices
are not readily cascadable for the saturating pulse would have to be one of the
AND operation inputs. Another limitation resulting from carrier depletion due to
saturation is that a certain amount of time is required for the carriers to replenish
themselves, usually 300 to 500 picoseconds limiting the error free repetition rate
to a few gigahertz.
Figure 2
A Terahertz Optical Asymmetric Demultiplexor (TOAD) (Sokoloff, Prucnal, Glesk, and Kane - 1993)
Top shows the window for the optical response of the non-linear element followed by the high power control pulse then the signal pulses.
 |
Yet as wave guide line sizes have shrunk and manufacturing tolerances
have improved phase misalignment has ceased to be much of a concern in present day
Mach-Zehnder devices such as those used for wavelength conversion (Spiekman, et
al - 1998)
In order to avoid the carrier depletion problem of optical semiconductors
Patel, Hall, and Rauschenbach (1996) developed the UNI, the Ultrafast Nonlinear
Interferometer, shown in figure 3. It is a threshold logic device not affected by
carrier depletion since it is based upon the less sensitive but faster transient
phenomena affecting power gain such as carrier heating, and two-photon absorption.
The 1996 UNI worked on a pulse stream of 40 gigabits/second. Later developments
raised the pulse stream rate to 100 gigbits/second (Hall and Rauschenbach - 1998).
It works by splitting one signal, called the signal pulse, into horizontal and vertical
polarizations by passing it through 7.5 meters of birefringent fiber in order to
separate the two polarizations by 12.5 picoseconds. The other signal, called the
control signal, is timed so that it intercepts the trailing polarization of the
first signal so as to affect its power level. The control pulse is orthogonally
polarized relative to the trailing signal pulse so that their phases will not interfere
and so that the control pulse can be filtered out before exiting the device. The
two polarizations next are recombined after passing through another 7.5 meter long
birefringent fiber. In this way the longer term carrier depletion affects both polarization
pulses equally.
Figure 3
The Ultrafast Nonlinear Interferometer (UNI) for 100 Gigabit/second Bitwise Logic (Hall and Rauschenbach - 1998)
DTF - dispersion tailered fiber, AMP - amplifier, BRF - birefringent fibers, PMT photomultiplier tube, E/O electro-optic, EDFA - erbium doped fiber amplifier
 |
To accomplish the AND operation the semiconductor optical amplifier
(SOA) is electrically biased just low enough so that the signal pulse is not able
to pass through in the absence of the control pulse in effect turning it into a
threshold logic device. To accomplish the NOT AND operation the SOA is electrically
biased upward to a high degree so that in the presence of the control pulses the
amplifier goes into saturation preventing the signal pulses from passing through.
Maximizing the transient nonlinear effects requires that the two signal and control
pulses have slightly different wavelengths. The pass through pulse was were 3 picoseconds
long having a center frequency of 1.553 microns. The control pulse was 8 picoseconds
long and had a center frequency of 1.560 microns. While an “INCLUSIVE OR” operation
was claimed by the authors it is not phase invariant. This operation is based upon
simply combining two inputs to become the new control pulse. If the phases of the
pulses happen to be out of phase the operation breaks down. Also the authors do
not reveal the power level of the signals which are very likely quite high and unequal.
In summery, no present method has the low power, small size, and cascadable combinations
to be a serious candidate for all-optical monolithic circuit logic.
Isolators /
circulators are important elements in photonic circuit design since most lasers
and optical amplifiers require isolation from reflections which tend to degrade
the output signal. For the most part present day optical isolators are multi-component
devices requiring ¼ wave or ½ wave plates to modify or block signal polarizations.
The simplest versions require a single predefined polarization as the input but
more complex devices are polarization insensitive able to handle arbitrary polarization
inputs. A good example of this type of isolator for use in fiber optics is made
by Wavesplitter Technologies, Inc. Its device is 3.1 cm long, 0.56 cm wide, has
an insertion loss of 0.4 dB and an isolation of 30 dB. Its size prohibits it use
in all-optical integrated circuits.
Smaller in size is the 0.5 cm long polarization
insensitive device demonstrated by Sugimoto, et al (1999). It is a hybrid device
made out of an optical wave guide on garnet with a magnet on top. It uses two polyimide
half wave plates placed in slots. It has a 3.0 dB insertion loss and a 14 to 23
dB isolation level. So while smaller it is not suitable for an all-optical integrated
circuit on some gallium arsinide (GaAs) substrate.
The only possibility of an integratable
optical isolator is that recently proposed (but not yet demonstrated) by Zaets and
Ando (1999). It is a single polarization device using a magneto-optic film made
out of a diluted magnetic semiconductor grown on a GaAs optical amplifier. Theoretically
a signal propagating in the forward direction has a gain while a signal propagating
in the opposite direction has a loss giving as much as a 180 dB/cm isolation ratio.
Yet this is a relative ratio and not an absolute isolation level which would be
preferable.
While fast all-optical logic arrays will have many applications all-optical
computing requires some sort of all-optical memory. Presently, this function is
performed by delay lines in which signals enter a long fiber loop and are optionally
accessed only when they arrive back at an output port. Obviously this is not integratable
nor is it applicable to computing in which the information needs to be recalled
on demand. The only proposal for an all-optical flip-flop seems to be by M. Fatehi
and G. Clinton in patent 5,537,243 (1996) which flip-flops between two different
frequencies. Yet using two different frequencies prevents different logical and
memory operations from being cascaded to form useful information processing devices.
References
Danielson, S.L., Hansen, P.B., Stubjaer, K.E., Schilling, M.,
Wunstel, K., Idler, P., Doussiere, P., and Pommerau, F. (1998) All Optical Wavelength
Conversion Schemes for Increased Input Power dynamic Range. IEEE Photonics Technology
Letters 10:6-62
Hall, K.L. and Rauschenbach, K.A. (1998) 100-Gbit/s Bitwise Logic.
Optics Letters 23 (16):1271-1273
Hess, R. et al (1998) All-Optical Demultiplexing
of 80 to 10 Gb/s Signals with Monolithic Integrated High Performance Mach-Zehnder
Interferometer. IEEE Photonics Technology Letters 10:165-167
Hofstee, et all (1998)
Designing for a Gigahertz. IEEE Micro: May-June Jahns, Jurgen (1997) Optical Digital
Computing. In chapter 15 of Handbook of Photonics. editor Mool Gupta. CRC Press
Jinno, M., and Matsumoto, T. (1991) Ultrafast All-optical Logic Operations in a
Non-linear Sagnac Interferometer with Two Control Beams (1991) Optics Letters 16:220-222
Lattes, A., Haus, H.A., Leonberger, F.J. and Ippen, E.P. (1983) IEEE Journal of
Quantum Electronics. QE-19:1718
Li, Y. (1998) Multigigabits per second board level
clock distribution scheemes using laminated end-tpered fiber bundles. IEEE Photon.
Tech. Lett. 10:884-886
McAulay, Alastair D. (1999) Optical Guided Wave Arithmetic.
Optical Engineering 38 (3): 468-476
Patel, N.S., Hall, K.L and Rauschenbach (1996)
40 Gbit/s Cascadable All-Optical Logic with an Ultrafast Nonlinear Interferometer.
Optics Letters 21 (18): 1466-1468
Sokoloff, J.P., Prucnal, P.R., Glesk, I., and
Kane, M. A (1993) Terehertz Optical Asymetric Demultiplexer (TOAD), IEEE Photonics
Technol. Lett. 5 970: 787-790
Spiekman, L.H., Koren, U, Chien, M.D, Miller, B.I.,
Wiesenfield, J.M, and Perino, J.S. (1997) All-Optical Mach-Zehnder Wavelength converter
with monolithically Integrated DFB Probe source. IEEE Photonics Technology Letters
9:1349-1351
Spiekman, L.H., Wiesenfield, J.M., Koren, U., Miller, B.I., and Chien,
M.D. (1998) All-Optical Mach Zehnder Wavelength Converter with Monolithically Integrated
Preamplifiers. IEEE Photonics Technology Letters 10:1115-1117
Stegeman, G.I. and
Wright, E.M. (1990) All-Optical Waveguide Switching. Optical and Quantum Electronics
22:95-122
Sugimoto, N. et al (1999) Waveguide Polarization-Independent Optical
Circulator. IEEE Photonics Technology Letters 11:355-357
Zaets, W. And Ando, K (1999)
Optical Waveguide Isolator Based on Nonreciprocal Loss/Gain of Amplifier Covered
by Ferromagnetic Layer. IEEE Photonics Technology Letters 11:1012-1014