All-optical manipulation of signals carried by lightwaves is attractive because controlling the light directly can be more efficient, allows a multitude of signal formats, and can also prove most cost effective. We implemented a novel scheme for ultrafast optical switching using very small control energy that relies on the use of a saturated fiber-optic parametric amplifier. Approximately 19 aJ (150 photons) of control pulse energy was needed for 50% extinction of the signal which is three to four orders of magnitude smaller than in other all-optical switching demonstrations. This allows the consideration of novel practical approaches to implement all-optical switching devices and all-optical subsystems for telecommunications and other applications.
©2008 Optical Society of America
The manipulation of information signals carried by lightwaves, such as the switching or routing of data or the conversion to another wavelength by using all-optical methods is attracting significant interest in the research community. This is because controlling the light directly - as opposed to first detecting it for subsequent electronic manipulation and conversion back into a lightwave (OEO conversion) - can potentially be more efficient, can handle a multitude of different signals (modulation formats, bit-rates) and can also prove most cost effective, at least if an all-optical approach could replace an approach needing many parallel OEO based switches. In addition, all-optical switching can operate at higher speed than OEO switching. In such schemes, a lightwave at some particular wavelength, polarization, or orientation relative to the signal wave is used to modify the way in which the signal wave is propagated. The nonlinear properties of glasses  or optical fibers  have been utilized to demonstrate such functionalities. However, the all-optical building blocks are not yet practical and a fundamental reason for this is the high switching power/energy needed, typically several Watts corresponding to pJ pulse energies in high speed applications requiring picosecond resolution.
One approach to implement all-optical switching devices with lower control power is to develop fibers with much smaller core area thus resulting in much stronger nonlinearity, since this is proportional to the intensity of the light. A five hundred-fold increase in nonlinearity compared to standard fiber (corresponding to potentially a five hundred-fold reduction of the needed switching power) has been demonstrated in so-called holey optical fibers . However, such fibers still suffer from several impairments such as excessive attenuation, large birefringence, and poor coupling efficiency. All-optical switching has also been demonstrated by using a semiconductor optical amplifier as the nonlinear medium. In , an all-optical AND-gate was demonstrated with a control pulse energy of about 20 fJ (excluding input coupling losses) needed to operate the switch having about 3 ps temporal resolution. At a much more fundamental level, free-space all-optical switching has also been demonstrated in Rubidium vapor . Here, the optical control power used to switch the signal light was 600 times smaller than the signal light and a pulse energy of about 10 fJ (40,000 photons) was needed to actuate the switching. While this work is still very fundamental, a drawback appears to be the observed slow response time of approximately 4 μs.
Here we investigate all-optical switching in a fiber optical parametric amplifier (FOPA) [3,4] operating in saturation  and show that it is possible to switch an optical signal using very small control pulse energies at a picosecond time scale. With our silica-fiber-based FOPA we switch a signal using a control power being 40,000 smaller than the signal power and show switching with 10 dB of extinction using a control pulse energy of 70 aJ or about only 550 photons and 3 dB extinction with 150 photons. This is over four orders of magnitude smaller than what is normally observed in optical fiber-based switching and represents a substantial advancement as it allows the consideration of novel practical approaches to implement all-optical switching devices and all-optical subsystems for telecommunications and other applications.
2. Fiber optic parametric amplifier operated in saturation
In general, the optical power needed to cause significant saturation of the gain of an optical amplifier is reduced if the unsaturated gain is large. Therefore, the use of a high gain amplifier, such as the FOPA, can be advantageous if low power all-optical switching is desired. The ability to control a strong signal with a weak one (rather than the opposite, which is normally the case for optical switches) is an important feature for cascadability, i.e. the output of a switch is capable of controlling several other optical switches, for example in subsystems performing advanced computational tasks. In addition, the FOPA has an ultrafast response time making it compatible with very fast phenomena or events. For example, it has been utilized to demonstrate all-optical monitoring of optical waveforms at bit rates as high as 640 Gbit/s with sub-ps temporal resolution .
When operated in saturated mode, the FOPA pump depletion can be almost complete. In  we demonstrated a record high 31 dB depletion when inserting an input signal of 3 mW, which means that more than 99.9% of the power was transferred from the pump to the signal and idlers. Here, we take advantage of this phenomenon and demonstrate ultrafast switching with a peak power of less than 0.01 mW. Similar to a transistor, we are able to control (e.g. switch on or off) a strong signal (in our experiment about 1 W) with a very weak control signal. It should be understood, however, that there is no inherent need to use a large signal power. What matters is that the unsaturated gain of the FOPA is large and this, in turn, is governed by the product γPpL, where γ (W-1 km-1) is the nonlinear parameter of the fiber, Pp (W) is optical pump power used to excite the FOPA, in this case also serving as the signal power to be switched by the weak control signal, and L (km) is the effective length of the fiber. We used γPpL ≈ 6.4, which in the simplest theory  would give a maximum unsaturated peak gain of 50 dB. When operating at a wavelength slightly off the gain peak, where the highest pump depletion was observed, the gain was decreased to 45 dB. Theoretically, the ratio between the signal power Pp and the required control signal, Psat, is in the same order as the gain, given by Psat ≈ 0.4Pp/G  for a single pump FOPA, where G is the FOPA gain at the signal wavelength. Similar performance and gain should be expected with much lower FOPA pump power by instead using an optical fiber having a much larger nonlinearity. With a hundred-fold increase of the fiber nonlinearity, the needed signal power in our example would be reduced to about 10 mW, and at the same time the required control signal power would also be reduced by a factor of hundred. Therefore, an effective way to decrease both the required signal power and control power is to increase γL (preferably γ). However, fibers with very high nonlinearity tend to have large loss which may reduce the benefit somewhat as shorter fibers would likely have to used in practice. When optimized for maximum amplification we have achieved FOPA gain values of up to 70 dB  this would also decrease the required control power Psat, but would generate a significant amount of amplified spontaneous emission noise.
The switching operation was analyzed experimentally by co-propagating a strong CW signal and a weak control signal either consisting of CW light, or short pulses in order to also demonstrate the temporal resolution of the gate. Figure 1 schematically shows the setup used in the experiments. The CW pump laser signal (λp = 1546 nm) was phase-modulated with four RF signals (100 MHz, 300 MHz, 900 MHz, and 2.7 GHz), in order to avoid stimulated Brillouin scattering, and then amplified to 1.6 W and filtered before launched into the highly nonlinear fiber (HNLF), where the parametric amplification occurred. The control signal, centered at λs = 1553 nm, co-propagating with the pump signal, was either provided by a tunable CW laser, or consisted of 2.7 ps wide control pulses provided by a 10 GHz mode-locked fiber ring laser. The HNLF had a nonlinear parameter γ = 8 W-1km-1, a dispersion slope 0.02 ps/(nm2km), a zero dispersion wavelength λ0 = 1539 nm, and a loss of 1.48 dB/km. The length of the fiber was 500 m. The principle of the optical switch is that the control signal is amplified as the pump signal is depleted. This means that the power at λp will be switched off very fast by a weak signal pulse at λs. After filtering out the signal at λp the switched signal can be analyzed.
The switch was first operated in steady state, i.e. CW laser light served as control signal to the switch. The output signal power was measured as a function of input control power on the optical spectrum analyzer (OSA). Figure 2 shows the result. Already at very low control power, the signal power starts to decline, and a maximum depletion of 18 dB was observed at -13.4 dBm input control power. For higher control signals, the power starts to transfer back to the signal.
In order to demonstrate the temporal resolution, and the required control pulse energy, the dynamic performance of the gate was investigated by launching 2.7 ps wide pulses as control signal. The output signal was then analyzed with an optical sampling oscilloscope (PicoSolve 101B) with 1 ps resolution and capable of measuring the inverted low duty-cycle pulses. The extinction ratio (ER) of the output signal was measured as a function of input peak power. The results are shown in Fig. 3. An ER of 3 dB (i.e. when 50% of the signal is switched) is reached at -21.5 dBm (7 μW), which corresponds to a pulse energy of 19 atto-Joule (≈ 150 photons), while 10 dB ER is obtained at -15 dBm (32 μW, 71 aJ, 550 photons). A maximum ER of 13 dB was reached at -10.8 dBm. The 2.7 ps input pulses result in a 4.8 ps switching window width at 3 dB ER, which is broadened by dispersion. One can see from the insets in Fig. 3 that the switching window width increases for higher ER (to 7.9 ps at 13 dB ER), which is due to the nonlinear transfer function of the output signal power, shown in Fig. 2. The results also agree well with numerical simulations using a split-step algorithm to solve the nonlinear Schrödinger equation .
The results in pulsed mode agree well with the CW measurements, i.e. the required peak power to reach a certain ER is similar as the CW power, even though we only obtain a maximum ER of 13 dB in pulsed mode to compare with 18 dB in CW mode. This is primarily due to fiber dispersion and optical filtering of the output signal. With a lower pump power and gain (30 dB), we were able to reach 17 dB ER in pulsed mode and 30 dB in CW mode. With 30 dB gain (instead of 45 dB) the required control power was increased by approximately 13-14 dB, as expected from .
In conclusion, we have demonstrated a novel optical switch based on fiber optical parametric amplification. With our configuration, taking advantage of the pump depletion of the amplifier, we were able to - in analogy with the function of a transistor - switch a strong signal with a very weak control signal being three to four orders of magnitude smaller than in other optical switches. Since the switch is also very fast, the required switch energy was extremely low (71 aJ or 550 photons to reach 10 dB ER) in our experiment. The implementation is practical and robust with low insertion loss since it is all-fiber based. While not implemented here, the FOPA is the only known optical amplifier that, in principle, can reach a 0-dB noise figure (using phase-sensitive mode of operation). In addition, the optical gain profile can, in principle, be translated to any other wavelength range as this is governed, not by fundamental material properties, but by the dispersion property of the fiber which is possible to engineer. By increasing the nonlinearity of the fiber (while maintaining the gain by simultaneously reducing the pump power), the required control energy would be further significantly reduced. The prospects of using this principle for single photon switching, applicable, for example, in quantum communication systems, photon counting, or quantum computing, are subject to further investigations.
This work was supported by the Swedish Research Council (VR), by the Swedish Foundation for Strategic Research (SSF) and by the Air Force of Scientific Research, Air Force Material Command, USAF, under grant number FA8655-06-1-3084. The U.S. Government is authorized to reproduce and distribute reprints for Governmental purpose notwithstanding any copyright notation thereon. Key equipment used in the experiments was funded by the Knut and Alice Wallenberg Foundation. Sumitomo Electric Industries, Japan provided the highly nonlinear fibers used in the experiments and PicoSolve Inc. provided the 500 GHz bandwidth optical oscilloscope needed to capture the optical waveforms. The authors also thank Dr. Mathias Westlund and Prof. Magnus Karlsson for fruitful discussions. T. Nishitani expresses his special thanks for The Global COE Internship Program of Osaka University, Japan.
References and links
1. W. R. Boyd, Nonlinear Optics (Academic Press, New York, 1992).
2. G. P. AgrawalNonlinear Fiber Optics (Academic Press, New York, 2001).
3. R. H. Stolen and J. E. Bjorkholm, “Parametric amplification and frequency conversion in optical fibers,” IEEE J. Quantum Electron. 18, 1062–1072 (1982). [CrossRef]
4. J. Hansryd, P. A. Andrekson, M. Westlund, J. Li, and P. O. Hedekvist, “Fiber-based optical parametric amplifiers and their applications,” IEEE J. Sel. Top. Quantum Electron 8, 506–520 (2002). [CrossRef]
6. T. M. Monro, K. M. Kiang, J. H. Lee, K. Frampton, Z. Yusoff, R. Moore, J. Tucknott, D. W. Hewak, H. N. Rutt, and D. J. Richardson, “High nonlinearity extruded single-mode holey optical fibers,” in Optical Fiber Communication Conference, Anaheim, (Optical Society of America2002), paper FA1-1-FA1-3.
7. S. A. Hamilton, B. S. Robinson, T. E. Murphy, S. J. Savage, and E. P. Ippen, “100 Gb/s optical time-division multiplexed networks,” J. Lightwave Technol. 20, 2086–2100(2002). [CrossRef]
9. P. A. Andrekson and M. Westlund, “Nonlinear optical fiber based high resolution all-optical waveform sampling,” Laser Photonics Rev. 1, 231–248 (2007). [CrossRef]
10. P. Kylemark, H. Sunnerud, M. Karlsson, and P. A. Andrekson, “Semi-analytic saturation theory of fiber optical parametric amplifiers,” J. Lightwave Technol. 24, 3471–3479 (2006). [CrossRef]
11. T. Torounidis, P. A. Andrekson, and B. E. Olsson, “Fiber-optical parametric amplifier with 70-dB gain,” IEEE Photon. Technol. Lett. 18, 1194–1196 (2006). [CrossRef]