Open Access
Add to BibSonomyAbstract
We report experimental evidence of nonvolatile all-optical memory operation using the two linear polarization states emitted from a GaAs oxide-confined VCSEL. The two polarization states coexist in a large range of pumping currents and substrate temperatures, and they can be controlled all-optically by exposing the device to polarization selective feedback, to crossed polarization reinjection orby injecting external light pulses. The active polarization state is recovered after powering off and on the VCSEL, while memory is lost if the substrate temperature is varied.
©2012 Optical Society of America
1. Introduction
Since the discovery of bistability in optical resonators, optical devices have been proposed as units for processing information [1]. The advent of semiconductor lasers with their revolutionary properties has further boosted the motivation for implementing all-optical functional devices based on bistability [2–4]. In Vertical-Cavity Surface-Emitting Lasers (VCSELs) bistability may be obtained by taking advantage of the polarization degree of freedom. Because of their symmetric structure,VCSELs do not exhibit anisotropies strong enough to fix the polarization direction. Residual dichroism and birefringence, which are inevitably present, lead to different net gains of the polarization modes, thus slightly favoring one mode over the other at VCSEL threshold. Nevertheless, polarization switching and instabilities occur when changing VCSEL control parameters [5, 6]. The most striking evidence is usually obtained in the Light output (L) vs. Pumping Current (I) curve (L/I curve),where, at a given current value, an abrupt switch occurs from one linear polarization (LP) state to the orthogonal one, often accompanied by a narrow hysteresis cycle. The most common situation observed is a switch from the shorter wavelength LP mode to the larger wavelength mode. This can be explained considering the redshift of the gain peak with respect to the cavity resonances when increasing the semiconductor temperature; as I is increased, Joule heating leads to a temperature increase in the active region, thus leading to a larger net gain for the red polarization mode [5]. A very different situation has been observed by Kawaguchi, reporting bistability between two orthogonal LP states in a large current interval [7]. This bistability is not disclosed by hysteresis in the L/I curve and it can be revealed only by injecting polarized light pulses into the VCSEL for inducing polarization switching. Under optimal conditions polarization control is obtained with commutation times of a few ps [8, 9].
In this paper we demonstrate the coexistence of two orthogonal LP states which extend throughout nearly the entire current range from threshold to the absolute maximum rating. We show that the polarization emission can be controlled all-optically by exposing the VCSEL to crossed polarization reinjection (XPR), to polarization selective optical feedback (PSF) or to injection of external laser beam pulses. We also show that the polarization state induced by perturbing the VCSEL is recovered after turning the device off and on. This surprising behavior discloses a modification of the VCSEL parameters which is not reset by powering the device off and it may have applications as non-volatile optical bit memory.
2. Experimental set-up
The VCSEL used is a 1 mW single-transverse mode oxide-confined GaAs device fabricated by ULM-photonics and lasing at 850 nm [10].The VCSEL is powered by a current driver having an accuracy of 1 μA while its substrate temperature is stabilized by using a thermo-controlled laser mount (up to 1 m°C). As shown in Fig. 1 , the VCSEL output is collimated and split by a polarization preserving beam splitter. One part of the output beam undergoes polarization splitting along its two orthogonal directions and is used for monitoring. The other part of the output is used for controlling the polarization state. Different schemes have been implemented to achieve polarization control. The first scheme is based on crossed polarization reinjection (XPR) where the VCSEL output beam passes through an optical system where its polarization is rotated byπ/2 before being reinjected into the VCSEL. This optical system consists of an optical isolator, where the input polarizer has been removed, and a mirror. While a defined LP component of the VCSEL beam is twisted byπ/4 and it reaches the mirror, the orthogonal LP component is filtered by the output polarizer of the isolator. Upon reflection on the mirror the selected polarization component passes again into the Faraday rotator and is further twisted byπ/4, thus returning to the VCSEL with a total π/2 twist.
Fig. 1 Experimental set-up with the three different schemes used for controlling the polarization. C: collimator, λ/2 waveplate, D: Detector, BS: Polarization preserving beamsplitter, PBS: Polarizing beam splitter, FR: Faraday rotator, P: Polarizer, F: Variable neutral density filter, M: Mirror, OI: Optical Isolator, ML: Master Laser.
The second scheme we have used is based on polarization selective feedback (PSF), where the VCSEL output beam passes through a polarizer and the selected LP component is fed back into the VCSEL by an external mirror. The λ/2 waveplate at the VCSEL output enables one to align one of the two emitted LP directions with the polarization direction selected by the XPR (or PSF) circuit. The XPR (or PSF) rate is defined as the ratio between the light power returned into the VCSEL over the light power emitted. This rate can be varied by using the neutral density filters in the XPR (or PSF) circuit and it has been quantified by opening reinjection (or feedback) loop in order to avoid interaction with the VCSEL. Finally, the third scheme we have used (Injection Set-up) is based on optical injection into the VCSEL of a coherent, linearly polarized external beam. Commonly for all three schemes, detector D3 monitors the addressing signal sent to the VCSEL for controlling the polarization.
3. Polarization bistability of the VCSEL output
In Fig. 2 we show the polarization resolved L/I curves of the VCSEL for different substrate temperatures (T). At T = 40°C, in the neighborhood of threshold, two linear and orthogonal polarizations are active but one (that we call LP-x) takes over rapidly and it is stable in the entire current range explored. When T is decreased it appears that the VCSEL starts to lase at threshold on a single polarization (that we call LP-y) orthogonal to LP-x. Besides, for some value of current depending on T, we observe a polarization switching to LP-x. For T = 25°C this switch occurs at I = 1.1 mA and, as T is decreased, it occurs at higher current values until, for T lower than 20°C, it disappears and the VCSEL emits on LP-y in the whole range of scanned I. T is then increased upward and this situation, where LP-y dominates for the entire range of current explored, is maintained till the highest value of T available in our set-up (40°C). We note that the situation of Fig. 2(a) cannot be reestablished by increasing T (in the limit of the values experimentally accessible). As we will show in the following, this situation can be instead restored by a proper optical perturbation of the VCSEL. Figure 2 shows that, for a large interval of currents and temperatures, LP-x and LP-y polarizations coexist and, therefore, it should be possible to optically perturb the VCSEL to control its polarization emission. Spectral measurements reveal that the LP-y polarization is 7.5 GHz blue detuned with respect to the LP-x polarization. We also notice that the two different situations observed in the polarization-resolved L/I curves for the same value of T (Figs. 2(c) and 2(e), for example) lead to the very same total (non polarization-resolved) L/I curve.
Fig. 2 Polarization resolved L/I curves versus T (red trace LP-x, blue trace LP-y). a) T = 40°C and then T is decreased to T = 30°C(b), T = 25°C (c) and to T = 15°C (d) then T is increased to T = 25°C (e)and T = 40°C (f).
4. Control of the polarization emission
Control of the polarization state emitted by VCSEL has been achieved implementing three different methods described in section 2. The first one is based on XPR, where only one polarization component (LP-x, say) is rotated into the orthogonal polarization direction (LP-y) and then is injected back into the VCSEL. If LP-x is the active component, XPR depletes the carriers available for it and eventually it brings LP-x below threshold, thus favoring the activation of LP-y. Since LP-y does not experience XPR, and the VCSEL is in the bistable regime, a polarization switch occurs. The LP component used for XPR must be the active one and it can be selected by adjusting the waveplate at the VCSEL output. In this way a switch from LP-x to LP-y and vice versa can be obtained.
In Fig. 3 we show the polarization switching obtained by using this method. In panel a) LP-y is initially active (the monitored polarization LP-x is off).We apply XPR for about 1.3s and, when XPR is removed, the VCSEL has switched to the LP-x state. If the VCSEL is exposed to XPR for shorter times, it returns to emitting in the LP-y state when XPR is removed. The XPR rate chosen is the minimal one able to induce a polarization switching but no significant reduction of the times required is observed when increasing this rate. In panel b) we show the situation where the VCSEL is emitting in the LP-x mode (the monitored LP-y mode is off). Even in this case the minimal time of exposure to XPR inducing a switch is extremely long compared with semiconductor timescales. We also remark that the exposure to XPR induces a polarization dynamics which are different if XPR is using the LP-y or the LP-x component. In the first case (panel a)), XPR induces an almost stationary elliptical polarization state with sporadic upward intensity spikes. In the second case (panels b)), XPR induces a slow increase of the LP-y intensity at the expense of the LP-x intensity (and thus of the XPR signal). This envelope contains fast dynamics whose amplitude decreases as the switching to a pure LP-y emission is completed. Polarization emission control using XPR has been obtained in the whole bistable domain shown in the polarization-resolved L/I curve. Required XPR exposure times vary with I and T and their orders of magnitude span from 0.2s to several seconds.
Fig. 3 LP-x (red trace), LP-y (blue trace) polarization intensities and XPR signal injected into the VCSEL (green trace), T = 30°C, I = 2.1mA, XPR rate = 2%.
Another method for controlling the polarization state is based on PSF. In this scheme an external cavity with a polarizing element admitting only the inactive LP mode is placed in front of the VCSEL. PSF selectively decreases the losses of this mode, thus effectively changing its dichroism and, above a critical rate, it leads to polarization switching. The control sequence of the polarization state by PSF is shown in Fig. 4 . The PSF rate required is as low as 0.044% (for inducing polarization switching from LP-x to LP-y, panel a)) and as low as 0.017% (for switching from LP-y to LP-x, panel b)).Polarization switch occurs immediately after the PSF is applied and, at the same time, a feedback signal (green trace) is detected. Nevertheless, in order to induce a permanent switch, i.e. surviving upon PSF removal, it is necessary to apply PSF for time windows of several hundreds of milliseconds.
Fig. 4 LP-x (red trace), LP-y (blue trace) polarization intensities and PSF signal injected into the VCSEL (green trace), for clarity green trace is shifted vertically of 0.1 units. T = 30°C, I = 2.9mA.
As for XPR, these times change depending on VCSEL current, but their order of magnitude is considerably larger than semiconductor timescales. Polarization emission control using PSF has been obtained in the whole bistable domain shown in the polarization-resolved L/I curve.
The last method we used to control the polarization emission is based on external optical injection. This method has been already successfully implemented by the Kawaguchi group [8, 9] and it consists of addressing the inactive LP mode by injecting a coherent electromagnetic field linearly polarized with the same orientation as the addressed LP mode and tuned to the same frequency. This method allows us to control the polarization emission in the whole bistability range of the VCSEL with very small injection intensities. On the other hand, as for the other methods implemented, we noticed that long exposure times to optical injection are required to induce a switching. The lowest injection levels required are obtained when the injected field is resonant with the addressed LP mode.
This implies that, because of the birefringence, the injected field frequency must be adjusted for inducing switching from LP-x to LP-y and vice versa. Figure 5 has been obtained in these resonant conditions and the injected powers are of 80nW for panel a), where we induce the switching from LP-x to LP-y, and 4nW for panel b) where we induce the switching from LP-y to LP-x. In the first case a minimum injection time of 25s is required, in the second case this time is less than 0.2s. As in Fig. 3, we note that the switching from LP-x to LP-y is accompanied by fast polarization dynamics, while the reverse switching occurs abruptly.
Fig. 5 LP-y polarization (blue trace) and injected external signal (green trace).In panel b) the green trace is magnified vertically of a factor of 30 with respect to panel a). I = 2.1 mA, T = 30°C.
5. Nonvolatile optical memory
While performing polarization control experiments we realized that after powering off the VCSEL, it always recovered the last polarization state upon being powered on again. We performed systematic measurements to analyze this phenomenon. In Fig. 6 the VCSEL is periodically switched on and off by modulating I with a rectangular signal and we monitor the LP-x polarization. The current modulation drives the laser periodically at I = 0 mA for 35s and at I = 2.36mA for 15s. As shown before, for T = 30°C and I = 2.36 mA the VCSEL exhibits polarization bistability. In panel a), initially the VCSEL is emitting on the LP-x polarization and surprisingly this state is recovered whenever it switches on. During the third cycle, when the VCSEL is on, we apply PSF to induce a polarization switching to LP-y. When PSF is removed, the VCSEL has switched to LP-y (LP-x is now off) and thereafter the VCSEL recovers this polarization state whenever is switched off and on. In panel b) the VCSEL is initially emitting LP-y mode (LP-x is off) whenever it switches on until, at the third cycle, we apply PSF in order to induce polarization switching to LP-x. Once XPR is removed, the VCSEL emits on LP-x and thereafter it recovers this polarization state after any current cycle.
Fig. 6 Laser driving current (green curve, right axis) and LP-x polarization intensity (red curve, left axis), T = 30°C, exposure to XPR is indicated by the blue pulse.
We have performed other tests where the VCSEL was switched off (also with the electrical contact short-circuited) during several hours and still, when driving the VCSEL in the bistable current range, it recovered the polarization state which was active before being powered off. This result was conditioned to leaving the substrate temperature T stabilized at the same value during the entire experiment. If T is varied during the time the laser is off, the polarization memory is lost, thus indicating that some parameter dependent on Tis at the origin of this memory effect. This can also explain why bistability can be disclosed in the polarization resolved L/I curves by scanning T. It is worth pointing out that this parameter is not affected by active region temperature variations induced by variations of I via Joule heating.
6. Discussion and conclusions
Different mechanisms can influence polarization stability through T. It is well known that polarization dichroism depends on mechanical stress on the laser chip [11]; it is possible that a change in T may induce some mechanical stress in the laser structure and that this stress has a hysteresis behavior versus T, thus explaining the result of Fig. 2 and the memory effect of the polarization state. This hypothesis is supported by the observation of an output beam steering in the VCSEL far-field of 0.004° horizontally and 0.0008° vertically when T is varied from 15°C to 45°C. This steering may be a consequence of the(T dependent) mechanical stress in the laser structure. It is important to point out that beam steering was not found when varying I and it was not observed in correspondence with a polarization switching. Whether of mechanical origin or not, the question of how this T-dependent mechanism might also be influenced by very weak optical perturbations is still unsolved. It is possible that optical perturbations activate the depressed polarization via the intracavity field and/or the semiconductor carrier with a fast timescale and that, if the addressed polarization is kept activated long enough, it may induce some change in the laser structure such as the one provoked by T. More work is on going to clarify this point.
In conclusion, we have reported experimental evidence of nonvolatile control of the polarization state in a wide-range bistable VCSEL. This control can be realized by using optical injection, by XPR or by PSF. Times required for polarization switching are of the order of seconds, but the situation may improve once the mechanism responsible for the polarization switching is identified.
Acknowledgment
We gratefully acknowledge funding from Agence Nationale de la Recherche through grant number ANR-12-JS04-0002-01.
References and links
1. H. Gibbs, Optical Bistability,Controlling Light With Light(Academic Press, 1985).
2. H. Kawaguchi, Bistabilities and Nonlinearities in Laser Diodes (Artech House, 1994).
3. H. Kawaguchi, “Bistable Laser Diodes and Their Applications: State of the Art,” IEEE J. Sel. Top. Quantum Electron. 3(5), 1254–1270 (1997). [CrossRef]
4. M. J. Adams, A. Hurtado, D. Labukhin, and I. D. Henning, “Nonlinear semiconductor lasers and amplifiers for all-optical information processing,” Chaos 20(3), 037102 (2010). [CrossRef] [PubMed]
5. K. D. Choquette, D. A. Richie, and R. E. Leibenguth, “Temperature dependence of gain-guided vertical-cavity surface emitting laser polarization,” Appl. Phys. Lett. 64(16), 2062 (1994) [CrossRef]
6. K. D. Choquette, R. P. Schneider, K. L. Lear, and R. E. Leibenguth, “Gain-dependent polarization properties of vertical-cavity lasers,” IEEE J. Sel. Top. Quantum Electron. 1(2), 661–666 (1995). [CrossRef]
7. H. Kawaguchi, I. S. Hidayat, Y. Takahashi, and Y. Yamayoshi, “Pitchfork bifurcation polarisation bistability in vertical-cavity surface-emitting lasers,” Electron. Lett. 31(2), 109 (1995). [CrossRef]
8. J. Sakaguchi, T. Katayama, and H. Kawaguchi, “All-optical memory operation of 980-nm polarization bistable VCSEL for 20-Gb/s PRBS RZ and 40-Gb/s NRZ data signals,” Opt. Express 18(12), 12362–12370 (2010). [CrossRef] [PubMed]
9. T. Katayama, A. Yanai, K. Yukawa, S. Hattori, K. Ikeda, S. Koh, and H. Kawaguchi, “All-optical flip-flop operation at 1-mA bias current in polarization bistable VCSELs with and oxide confinement structure,” IEEE Photon. Technol. Lett. 23, 1811–1813 (2011). [CrossRef]
10. M. Grabherr, R. Jäger, R. Michalzik, B. Weigl, G. Reiner, and K. J. Ebeling, “Efficient single-mode oxide confined GaAs VCSELs emitting in the 850 nm wavelength regime,” IEEE Photon. Technol. Lett. 9(10), 1304–1306 (1997). [CrossRef]
11. K. Panajotov, B. Nagler, G. Verschaffelt, A. Georgievski, H. Thienpont, J. Danckaert, and I. Veretennicoff, “Impact of in-plane anisotropic strain on the polarization behavior of vertical-cavity surface-emitting lasers,” Appl. Phys. Lett. 77(11), 1590–1592 (2000). [CrossRef]
