All-Optical Flip-Flops
Optical flip-flops constitute time-domain switches that can be turned on and off using an external control. Such devices attracted considerable attention during the 1980s because they mimic the functionality of electrical flip-flops and provide the most versatile solution for optical switching, optical memory, and optical logic elements.
All flip-flops require an optical bistable device that can be switched between its two output states through a control signal. Semiconductor lasers and SOAs are often used for making flip-flops because of their compact size, low power consumption, and potential for monolithic integration with other photonic devices. The external control can be electrical or optical for such devices. When an optical control is used, the device is referred to as an all-optical flip-flop. The following figure shows the basic idea behind such a device. The device output can be switched to the "on" state by sending an optical set signal in the form of a short pulse. At a later time, a reset pulse turns the flip-flop off. Unlike the switching scheme discussed in the nonlinear techniques and devices tutorial, the output remains on for the duration between the set and reset pulses. In this sense, a flip-flop retains memory of the set pulse and can be used as an optical memory element.
1. Semiconductor Lasers and SOAs
An InGaAsP semiconductor laser was used in a 1987 experiment as a Fabry-Perot amplifier by biasing it at slightly below the threshold (97% level). Two other 1.53-μm lasers with a frequency difference of only 1 GHz were used for holding and control beams. The flip-flop could be switched on and off, but the switching time in this experiment was relatively long (>1 μs). In a 2000 experiment, a DFB laser was biased below the threshold, and the resulting SOA was employed as an optically bistable device. The holding beam at 1,547 nm was tuned toward the longer-wavelength size of the Bragg resonance. Set and reset pulses were 15-ns wide and were obtained from two InGaAsP lasers operating at 1,567 and 1,306 nm, respectively. The set pulse had a peak power of only 22 μW (0.33-pJ energy), while the peak power of reset pulses was close to 2.5 mW (36-pJ energy). The figure below shows (a) a sequence of two set and two reset pulses together with (b) the output of the flip-flop. Such a device is capable of switching on a time scale comparable to the carrier lifetime (~1 ns).
The physical mechanism behind such a flip-flop is related to the shift in the stopband of the grating as the refractive index changes in response to variations in the carrier density. The set pulse saturates the SOA gain, reduces the carrier density, and thus increases the effective refractive index 
and shifts the Bragg wavelength to longer wavelengths as the two are related by
where Λ is the grating period. In contrast, the reset pulse is absorbed by the SOA. The resulting increase in the carrier density decreases the refractive index and shifts the Bragg wavelength toward shorter wavelengths. The wavelength of set pulses must be within the gain bandwidth of the SOA so that it can saturate the amplifier. The exact wavelength of the reset pulse is not important as long as it is shorter than the holding-beam wavelength by an amount large enough that it falls outside the gain bandwidth and is thus absorbed by the amplifier. Thus, both control signals have a broad wavelength range of operation. The polarization of the reset pulse does not play any role. The dependence on the polarization of the set pulse can be reduced by designing the SOA appropriately. As control signals operate independently from the holding beam, they can propagate in a direction opposite to that of the holding beam; their role is only to change the carrier density. This transparency to the direction should be useful for system design.
Optical flip-flops have been constructed in recent years using several other designs. In a 1995 experiment, flip-flop operation at 1.2 GHz was realized using a vertical-cavity surface-emitting laser (VCSEL) by injecting optical set and reset pulses with orthogonal polarizations. The physical mechanism behind this flip-flop is related to polarization bistability. More specifically, the state of polarization of the output is switched from TE to TM by using set and reset pulses. In another experiment, an optical flip-flop was realized by switching between two modes of a semiconductor laser. FWM in photorefractive crystals can also be used for making an optical flip-flop when the feedback is provided by placing the crystal in a ring cavity. However, the speed of such a device is limited by the response time of the photorefractive crystal. A VCSEL laser biased below the threshold was used as a bistable amplifier in 2009 to realize a flip-flop operating in the reflection mode.
In another scheme, the injection of CW light into a DFB laser created optical bistability through the spatial hole burning effect. Switching between the low- and high-power states was realized by injecting low-energy (~0.2 pJ) set and reset pulses into the DFB laser from opposite directions. Such a flip-flop could be switched on within a duration <75 ps at repetition rates of up to 2 GHz.
Passive semiconductor waveguides can also be used for constructing an all-optical flip-flop. Such devices cannot employ gain saturation as the nonlinear mechanism. Instead, they often operate below the bandgap of the semiconductor material and employ the optical Kerr effect to introduce intensity-dependent changes in the refractive index. A Bragg grating is also fabricated along the waveguide length to make the device bistable. Because of the electronic nature of the Kerr nonlinearity, such optical flip-flops can respond at time scales of picoseconds or shorter. This is the main advantage of passive waveguides compared with SOAs whose response time is limited by the carrier lifetime.
2. Coupled Semiconductor Laser and SOAs
Several kinds of flip-flops have been made by coupling two lasers or SOAs. The use of two mutually synchronized semiconductor lasers was proposed in 1997. By 2001, two coupled semiconductor lasers were used for making flip-flops in which the output wavelength was switched between two values by selectively turning one of the lasers off. The following figure shows the experimental scheme. Two lasers, each built using an SOA and two fiber Bragg gratings acting as mirrors, operate at different wavelengths, say, λ1 and λ2. One of the lasers is selectively turned off using the technique of gain quenching by injecting light at a wavelength different than that at which the laser operates in isolation. As a result, the output wavelength can be switched between λ1 and λ2 using optical controls.
An optical flip-flop in which two coupled lasers were integrated on the same chip has also been fabricated. In this device shown schematically in the following figure, a VCSEL is integrated with an edge-emitting laser. The two lasers share the same active region and are mutually coupled through gain saturation since they compete for the gain in this shared region. The edge-emitting laser contains a short unbiased section that casts as a saturable absorber and makes it bistable. This laser is biased such that its output is relatively weak (off state). A set pulse injected into the absorber switches the laser to the "on" state because it reduces cavity losses by saturating the absorber. The device can be turned off by injecting a reset pulse through the VCSEL if the pulse is intense enough to saturate the gain in the active region shared by the two lasers. As the intracavity intensity is reduced in response to lower gain, eventually it becomes too low to saturate the absorber; this results in increased cavity losses, and the device returns to the original "off" state. This cycle can be repeated to switch the flip-flop on and off using set and reset pulses. In another scheme, two active waveguides were coupled using a multimode-interference (MMI) coupler and two saturable absorbers were incorporated within a semiconductor laser cavity such that it exhibited instability with respect to its two transverse modes when controlled by th set and reset pulses.
In an interesting scheme, optical feedback between an SOA and a DFB laser is employed to realize a flip-flop. The bidirectional coupling between the two is exploited to realize bistable operating by injecting low-energy (~5 pJ) set and reset pulses from the opposite directions. The flip-flop exhibited an on-off ratio of more than 15 and could be operated at a repetition rate of 0.5 GHz using 150-ps-wide pulses. In another scheme shown schematically in the following figure, two Mach-Zehnder (MZ) interferometers, with an SOA in one of their arms, were coupled to realize flip-flop operation. Coupling between the two interferometers provides the instability by allowing switching between the two CW holding beams launched at different wavelengths. The whole device, named optical static random-access-memory (RAM) cell, could be integrated monolithically using hybrid technology in which InP chips are flip-bonded onto a silicon platform. Such a device exhibited read an write functionality at a bit rate of 5 Gb/s.
In a 2010 experiment, a flip-flop was realized through heterogeneous integration of an InP microdisk laser (diameter 7.5 μm) that was coupled to a silicon waveguide, fabricated with the silicon-on-insulator technology. At low bias currents, the laser operated in both the clockwise and counter-clockwise directions, but it operated in one direction only at high currents because the two effective lasers were coupled through their common gain medium. The bistable operation between the clockwise and counter-clockwise directions was exploited to make the flip-flop. More specifically, when the set and reset pulses were injected into the laser from the opposite directions, the laser switched its direction of operation. Such a device exhibited switching in 60 ps with only 1.8 fJ of injected pulse energy.