Expand this Topic clickable element to expand a topic
Skip to content
Optica Publishing Group

Sub-picosecond all-optical gate utilizing aN intersubband transition

Open Access Open Access

Abstract

All-optical gate-switch operation utilizing GaN intersubband transition has been achieved by reducing edge dislocation density in the epitaxial layers. The diminution of dislocation was accomplished by MBE regrowth on an MOCVD-grown layer. Excess propagation loss for transverse magnetic polarization decreased due to the reduction of the dislocation. By the improvement of the propagation property, sub-picosecond all-optical gate with an extinction ratio of more than 10 dB was accomplished with an input pulse energy of 150 pJ. Moreover, the insertion loss with the switch on was improved.

©2005 Optical Society of America

1. Introduction

The expansion of high-speed optical communication systems is spurring demand for ultrafast all-optical switching devices and signal processing devices. A promising way of realizing such devices is to utilize intersubband transition (ISBT) in semiconductor quantum wells (QWs) since the absorption recovery time is extremely short due to the particularly fast carrier relaxation process [1]. So far, the ISBTs have been accomplished for optical communication wavelengths with InGaAs/AlAsSb, CdS/ZnSe/BeTe and GaN/Al(Ga)N QWs [24]. Above all, the ISBT in GaN QWs has many advantages, that is: (1) the absorption recovery time is considerably below 1 ps [58], (2) a wide range of wavelengths is available with a less complicated QW structure [4,7,9], (3) the homogeneous spectral line width is sufficiently broad due to the short dephasing time, (4) two-photon absorption does not interfere with the saturable absorption since the band gap is sufficiently wide, and (5) the materials are robust and less toxic.

Since the authors proposed applying the GaN ISBTs for all-optical switches [5], there have been many reports on this topic. Recently, ultrafast all-optical modulation was verified for the first time with a waveguide structure [10]. The extinction ratio of the modulator, however, was rather low (~2 dB) for high control pulse energy (120 pJ). Moreover, the insertion loss when the modulator was “on” was as large as 32.5 dB. The low extinction ratio and the large insertion loss were attributable partly to the excess polarization dependent loss (PDL), that is, PDL whose origin is not the ISBT, caused by dislocations in the epitaxial layers grown by molecular beam epitaxy (MBE) [11]. On the other hand, generally speaking, an epitaxial layer grown by metal organic chemical vapor deposition (MOCVD) has fewer dislocations. So far, however, an ISBT has not been achieved for optical communication wavelengths with MOCVD-grown QWs. In this study, all-optical gate switches are fabricated with MBE-regrown QWs on a MOCVD-grown GaN. These layer structures make it possible both to reduce dislocations and to achieve an ISBT at optical communication wavelengths. It is confirmed that the excess PDL decreased and characteristics of the absorption saturation were improved. Furthermore, an all-optical gate-switch operation with the extinction ratio of more than 10 dB is reported.

2. Fabrication

A schematic cross section of the optical switches is shown in Fig. 1. On a 0.8-µm thick MOCVD-grown GaN layer, GaN (0.2 µm), QW structure and GaN (1 µm) were successively grown by RF-MBE. Two types of samples were fabricated. Sample A consisted of 2 pairs of QWs and 2×1020 cm-3 doping with Si was performed on the wells. Sample B consisted of 3 pairs with 1×1020 cm-3 doping. According to the observation by transmission electron microscopy (TEM), the dislocation density of sample A was 1.7×109 cm-2.

 figure: Fig. 1.

Fig. 1. Schematic cross section of the optical switch.

Download Full Size | PDF

The intersubband absorption wavelength was nominally 1.7 µm for both samples. The ridge structures were fabricated by the electron cyclotron resonance-reactive ion beam etching (ECR-RIBE) technique. The ridge height was 1 µm. The mesa width was 1.5 µm and it tapered to 2 µm near the cleaved facets. The facets were anti-reflection (AR)-coated. The waveguide length was 400 µm. The devices were mounted in modules with 0.5-m-long polarization-maintaining dispersion-shifted fiber (PMDSF) pigtails. The length was shortened compared with those of the sample in ref.[10] (1 m) for the broadening of the pulse width of 1.7-µm wavelength to be suppressed.

3. Measurement setup

The excess PDLs were measured at a wavelength of 1.3 µm at which wavelength the intersubband transition shouldn’t occur. Absorption saturations and switching performances were characterized by a setup schematically shown in Fig. 2. The optical pulse source was an optical parametric oscillator (OPO) excited by a mode-locked Ti:sapphire laser with a repetition rate of 80 MHz. The signal pulses of a wavelength of 1.55 µm and idler pulses of a wavelength of 1.7 µm were used as the signal and control pulses for switching, respectively. Both pulses were polarization-controlled by wave plates. Pulse intensities were controlled by ND filters, wave plates and polarizers. The pulse widths of the OPO outputs were nominally 130 fs for both wavelengths. The width was unchanged for a wavelength of 1.55 µm through PMDSF. However, the pulse width is considered to be broadened to some extent at the device for a wavelength of 1.7 µm due to dispersion in the fiber. Switching operations were characterized by a pump-probe technique with a variable delay. Output signals (probe pulses) were detected by an InGaAs detector, whereas control (pump) pulses were cut by a band-pass filter in front of the detector.

 figure: Fig. 2.

Fig. 2. Schematic diagram of the measurement setup.

Download Full Size | PDF

4. Results and discussions

4.1 Excess PDL and absorption saturation

Comparison of the insertion loss for transverse electric (TE)-polarization with that for transverse magnetic (TM)-polarization at a wavelength of 1.3 µm indicated that the excess PDL was 5 dB for both samples A and B. The excess PDL for the sample reported in ref [10] was measured to be 10 dB. The TEM observations indicated that the edge dislocation density of sample A (of the order of 109 cm-2) was much less than that of the MBE-grown GaN layer (of the order of 1010 cm-2). The edge dislocations introduce acceptor centers along the dislocation lines that extend perpendicularly to the surface of the substrate, and the acceptor centers capture residual electrons [12, 13]. The authors consider that the perpendicularly aligned charges act as a wire-grid polarizer [11]. Thus, the improvement of 5 dB was attributable to reduction of dislocation density. The residual excess PDL should be eliminated by further improvement of the quality of the epitaxial layer.

For a wavelength of 1.55 µm, transmittances for TE-polarization, where the ISBT shouldn’t occur, were -6.6 dB and -5.5 dB for samples A and B, respectively. These losses were independent of the pulse energy. Thus, the major cause of the losses is considered to be coupling losses between the fibers and the waveguides that was attributable to roughness of surfaces of the cleaved facets. On the other hand, for TM-polarization, clear increases of the transmittance with the input pulse energy were observed for both samples, as shown in Fig. 3. These increases are due to the saturation of intersubband absorption. According to the results in Fig. 3, the pulse energy that was required for the increase by 5.5 dB was approximately 45 pJ for both samples. The energy is smaller by 4 dB than that of the previous sample in ref. [10]. The major cause of the decrease in the saturation energy is attributable to the reduction of the background propagation loss by 5 dB. The discrepancy by 1 dB was caused by the difference in the sample structure such as the ridge width. With sample A, an increase by as large as 10 dB was obtained when the pulse energy was 100 pJ. For sample B, the increase was 7 dB with the same pulse energy. The increase is smaller than that for sample A. However, the insertion loss at the pulse energy of 100 pJ (-12.5 dB) is much smaller than that of sample A (-19.5 dB) and the sample in ref. [10] (-29.5 dB). Major factors for insertion loss are coupling loss, excess propagation loss and residual saturable absorption component. Since the coupling loss and the excess loss were almost the same for samples A and B, the insertion loss differed due to the difference in the residual saturable absorption component that originated from the difference in the amount of the carrier doping into the wells. Because of the reduction of the dislocation density and the carrier doping in samples A and B, the insertion losses decreased compared with the sample in ref. [10].

 figure: Fig. 3.

Fig. 3. Dependence of transmittance for TM-polarization on pulse energy for samples A and B. The pulse width was 130 fs.

Download Full Size | PDF

4.2 Gate switch performance

Figure 4 shows the results of the pump-probe measurements for samples A and B. The energies of the control and the signal pulses were 150 pJ and 5 pJ, respectively. The increases in the transmitted signal intensity when the timing of control and signal pulse coincides are as large as 11.5 (sample A) and 7.4 (sample B) dB. The full widths at half maximum (FWHM) are as narrow as 230 (sample A) and 270 (sample B) fs. For both samples, the changes in the signal transmittance did not completely return to zero even in the case of a time delay of 1.5 ps. Since this phenomenon wasn’t observed with the sample in ref. [10] for which the number of wells and the doping level were different, this slow recovery component is expected to be eliminated by optimizing the QW structure.

For sample A, considering that the insertion loss for the signal pulse with the energy of 5 pJ was -29.5 dB as shown in Fig.3, the loss when the switch is “ON” should be -18 dB. For sample B, though the extinction ratio is smaller than that of sample A, the insertion loss with the switch on is as low as -11.6 dB. Thus, the results shown in Fig. 3 are very encouraging considering that they are for the preliminary fabrications of devices. If the quantum well structure is fully optimized and the coupling loss and the excess loss are further reduced, the authors expect sub-picosecond gate-switching operation with higher extinction ratio and lower insertion loss to be achieved.

 figure: Fig. 4.

Fig. 4. Results of pump-probe measurements for samples A and B.

Download Full Size | PDF

The input pulse energy required for the extinction ratio of 2.3 dB, which is the same extinction ratio for the sample in ref. [10], was 30 pJ for sample A. This is improvement by 6 dB, which is greater than the progress expected form the decrease by 4 dB in saturation energy. This is an advantage of shortening of the pigtail fibers from 1 m to 0.5 m, which suppressed the broadening of the control pulse width.

5. Summary

Sub-picosecond all-optical gate switches utilizing intersubband transition in GaN QWs were fabricated by MBE regrowth on an MOCVD-grown GaN. Compared with conventional MBE-grown layers, the dislocations in the MBE-regrown layer decreased since the dislocation density of the underlying MOCVD-grown layer was less. The reduction of the dislocation density led to a decrease in the excess PDL by approximately 5 dB. Absorption saturation as large as 10 dB was achieved for a sample with a pulse energy of 100 pJ. The other sample showed an insertion loss as low as 12.5 dB with a pulse energy of 100 pJ. Switching operations were performed with the wavelengths of 1.55 µm for signal pulses and 1.7 µm for control pulses. The FWHM of the gate operation was as short as 230 fs. With the control pulse energy of 150 pJ, a sample showed an extinction ratio of more than 10 dB. Low insertion loss of 11.6 dB was obtained when the switch was on for a sample with a different QW structure. If the quantum well structure is fully optimized and the coupling loss and the excess loss are further reduced, the authors expect switching operation with higher extinction ratio and lower insertion loss to be achieved. These results are encouraging with respect to fabrication of sub-picosecond all-optical switches and signal processing devices.

Acknowledgments

This work was performed under the supervision of The Femtosecond Technology Research Association (FESTA), which is supported by New Energy and Industrial Development Organization (NEDO).

References and links

1. S. Noda, T. Uemura, T. Yamashita, and A. Sasaki, “All-optical modulation using an n-doped quantum-well structure,” J. Appl. Phys. 68, 6529–6531 (1990). [CrossRef]  

2. T. Akiyama, N. Georgiev, T. Mozume, H. Yoshida, A. V. Gopal, and O. Wada, “Nonlinearity and recovery time of 1.55 µm intersubband transition in InGaAs/AlAs/AlAsSb coupled quantum wells,” Electron. Lett. 37, 129–130 (2001). [CrossRef]  

3. R. Akimoto, K. Akita, F. Sasaki, and T. Hasama, “Sub-picosecond electron relaxation of near-infrared intersubband transition in n-doped (CdS/ZnSe)/BeTe quantum wells,” Appl. Phys. Lett. 81, 2998–3000 (2002). [CrossRef]  

4. H. M. Ng, C. Gmachl, S. N. G. Chu, and A. Y. Cho, “Molecular beam epitaxy of GaN/AlxGa1-xN superlattices for 1.52–4.2 µm intersubband transitions,” J. Crystal Growth 220, 432–438 (2000). [CrossRef]  

5. N. Suzuki and N. Iizuka, “Feasibility study on ultrafast nonlinear optical properties of 1.55-µm intersubband transition in AlGaN/GaN quantum wells,” Jpn. J. Appl. Phys. 36, L1006–1008 (1997). [CrossRef]  

6. C. Gmachl, S. V. Frolov, H. M. Ng, S. N. G. Chu, and A. Y. Cho, “Sub-picosecond electron scattering time for λ~1.55 µm intersubband transition in GaN/AlGaN multiple quantum wells,” Electron. Lett. 37, 378–380 (2001) [CrossRef]  

7. N. Iizuka, K. Kaneko, and N. Suzuki, “Near-infrared intersubband transition in GaN/ALN quantum wells grown by molecular beam epitaxy,” Appl. Phys. Lett. 81, 1803–1805 (2002) [CrossRef]  

8. J. Hamazaki, S. Matsui, H. Kunugita, K. Ema, H. Kanazawa, T. Tachibana, A. Kikuchi, and K. Kishino, “Ultrafast intersubband relaxation and nonlinear susceptibility at 1.55 µm in GaN/AlN multiple-quantum wells,” Appl. Phys. Lett. 84, 1102–1104 (2004) [CrossRef]  

9. K. Kishino, A. Kikuchi, H. Kanazawa, and T. Tachibana, “Intersubband transition in (GaN)m/(AlN)n superlattices in the wavelength range from 1.08 to 1.61 µm,” Appl. Phys. Lett. 81, 1234–1236 (2002) [CrossRef]  

10. N. Iizuka, K. Kaneko, and N. Suzuki, “Sub-picosecond modulation by intersubband transition in ridge waveguide with GaN/AlN quantum wells,” Electron. Lett. 40, 962–963 (2004) [CrossRef]  

11. N. Iizuka, K. Kaneko, and N. Suzuki, “Effect of edge dislocation of polarization dependent loss of MBE-grown GaN ridge waveguide at optical communication wavelengths,” presented at 2004 MRS Fall Meeting, Boston, MA, 29 Nov. -3 Dec. 2004

12. W. T. Reed, “Theory of Dislocation in Germanium,” Phil. Magn. 45, 775–796 (1954)

13. H. M. Ng, D. Doppalapudi, D. Korakakis, R. Singh, and T. D. Moustakas, “MBE growth and doping III–V nitrides,” J. Crystal Growth , 189/190, 349–353 (1998) [CrossRef]  

Cited By

Optica participates in Crossref's Cited-By Linking service. Citing articles from Optica Publishing Group journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (4)

Fig. 1.
Fig. 1. Schematic cross section of the optical switch.
Fig. 2.
Fig. 2. Schematic diagram of the measurement setup.
Fig. 3.
Fig. 3. Dependence of transmittance for TM-polarization on pulse energy for samples A and B. The pulse width was 130 fs.
Fig. 4.
Fig. 4. Results of pump-probe measurements for samples A and B.
Select as filters


Select Topics Cancel
© Copyright 2024 | Optica Publishing Group. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.