Long wavelength vertical-cavity semiconductor optical ... - IEEE Xplore

2 downloads 0 Views 142KB Size Report
E. Staffan Björlin, B. Riou, Patrick Abraham, Senior Member, IEEE, Joachim Piprek, Senior ... fier gain, optical bandwidth, and saturation output power were si-.
274

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 37, NO. 2, FEBRUARY 2001

Long Wavelength Vertical-Cavity Semiconductor Optical Amplifiers E. Staffan Björlin, B. Riou, Patrick Abraham, Senior Member, IEEE, Joachim Piprek, Senior Member, IEEE, Y.-J. Chiu, Member, IEEE, K. Alexis Black, A. Keating, Member, IEEE, and John E. Bowers, Fellow, IEEE

Abstract—This paper overviews the properties and possible applications of long wavelength vertical-cavity semiconductor optical amplifiers (VCSOAs). A VCSOA operating in the 1.3- m wavelength region is presented. The device was fabricated using wafer bonding; it was optically pumped and operated in reflection mode. The reflectivity of the VCSOA top mirror was varied in the characterization of the device. Results are presented for 13 and 12 top mirror periods. By reducing the top mirror reflectivity, the amplifier gain, optical bandwidth, and saturation output power were simultaneously improved. For the case of 12 top mirror periods, we demonstrate 13-dB fiber-to-fiber gain, 0.6 nm (100 GHz) optical bandwidth, a saturation output power of 3 5 dBm and a noise figure of 8.3 dB. The switching properties of the VCSOA are also briefly investigated. By modulating the pump laser, we have obtained a 46-dB extinction ratio in the output power, with the maximum output power corresponding to 7-dB fiber-to-fiber gain. All results are for continuous wave operation at room temperature. Index Terms—Fabry–Perot resonators, laser amplifiers, optical filters, optical pumping, optical resonators, optical switches, semiconductor optical amplifiers, wafer bonding.

I. INTRODUCTION

L

ONG-wavelength optical amplifiers are essential components in optical communication systems. Erbium-doped fiber amplifiers (EDFAs) have been very successful for amplification in the 1.55- m wavelength range, and are widely deployed in long-haul links. On the other hand, fiber amplifiers in the important 1.3- m communication window have evolved much slower than EDFAs, and are not widely used. Although much work has been done in recent years using praseodymiumdoped fiber amplifiers (PDFAs) [1] and Raman amplification [2], these types of amplifiers show poor efficiency, thereby requiring high pump powers. Furthermore, PDFAs use fluoride fiber, which is brittle and difficult to handle. Due to the high cost, high power consumption, and large size of fiber amplifiers, they are not suitable for applications where low-cost compact amplifiers are required, such as in metro-area networks (MANs), and in the anticipated deployment of fiber to the home (FTTH).

Manuscript received June 16, 2000; revised October 17, 2000. This work was supported by the Defense Advanced Research Projects Agency (DARPA), via the Heterogeneous Optoelectronics Technology Center (HOTC). E. S. Björlin, B. Riou, J. Piprek, Y.-J. Chiu, and J. E. Bowers are with the Electrical and Computer Engineering Department, University of California at Santa Barbara, Santa Barbara, CA 93106 USA (e-mail: [email protected]). P. Abraham is with Agility Communications, Inc., Goleta, CA 93117 USA. K. A. Black is with Zaffire, Inc., San Jose, CA 95134 USA. A. Keating is with Calient Networks, Goleta, CA 93117 USA. Publisher Item Identifier S 0018-9197(01)00869-7.

Low-cost optical amplifiers in the 1.3- m wavelength range are, therefore, much anticipated. Semiconductor optical amplifiers (SOAs) are an alternative to fiber amplifiers and have been an area of intense research for many years. Conventional in-plane devices suffer from poor coupling efficiency to optical fiber, are typically sensitive to polarization, and are not yet price competitive. Vertical cavity semiconductor optical amplifiers (VCSOAs) show a number of advantages over their in-plane counterpart, such as better coupling efficiency to optical fiber, and thus lower noise figure, lower power consumption, insensitivity to polarization, and the potential for being integrated into 2-D array architectures. The vertical cavity structure also allows for on-wafer testing, which significantly reduces manufacturing cost, making VCSOAs attractive components for small networks (LANs, MANs, access networks, etc.) and FTTH applications. The high performance of fiber amplifiers is not required for such applications, but a large number of components will be needed, and the low cost of VCSOAs would provide a major advantage. The fundamental geometrical difference between VCSOAs and in-plane SOAs causes two major differences in their characteristics. First, most in-plane SOAs have sufficient single-pass gain to function as traveling wave amplifiers (TWAs). The shorter active region of the VCSOA makes the single-pass gain much smaller, on the order of a few percent. To compensate for this, the VCSOA uses feedback, which is provided by a resonance cavity created by the two mirrors. The resonance cavity results in an optical bandwidth that is limited to the linewidth of the Fabry–Perot mode. The optical bandwidth is, hence, significantly smaller than that of a TWA. To optimize the performance of the VCSOA, it is desirable to maximize the single-pass gain. To achieve this, a multiple quantum-well (MQW) active region is required, and a stacked MQW active region is desirable. Consequently, the mirror reflectivity has to be reduced in order to fully utilize the potential gain of the material without the onset of lasing. The second major factor differentiating a VCSOA from an in-plane SOA arises from the fact that the optical mode passes at a right angle through the active region, resulting in a circular symmetric cross section of the mode. This, in turn, results in high coupling efficiency to optical fiber and insensitivity to the polarization of the input signal light. Structures with a large number of quantum wells (QWs) often suffer from nonuniform carrier distribution if electrical pumping is employed. Optical pumping has the inherent advantage of generating the carriers directly in the wells, leading to a more even distribution of carriers among the wells, thereby

0018–9197/01$10.00 © 2001 IEEE

BJÖRLIN et al.: LONG WAVELENGTH VCSOAs

eliminating the issue of carrier transport. Furthermore, optical pumping allows for the use of undoped material, thereby minimizing optical losses. In . [3], [4], it has been demonstrated that optical pumping is indeed an efficient and commercially viable way to pump vertical cavity surface emitting lasers (VCSELs). Previous work has demonstrated continuous wave (CW) amplification in VCSOAs at both 1.5 m and 980 nm. An electrically pumped VCSOA at 1.5 m exhibited a maximum gain of 18 dB measured at 217 K for an input signal of 45 dBm [5]. At 980 nm, a maximum gain of 20 dB and an optical bandwidth of 0.1 nm were measured for an input signal of 40 dBm [6]. However, the important issues of fiber-coupled gain and saturation power have not been addressed. In this work, we report results and analyze the performance of an optically pumped VCSOA operating at 1.3 m. The device was operated continuous wave (CW) in the reflection mode at room temperature. The reflectivity of the VCSOA top mirror was varied in the characterization of the device. Results are presented for 13 and 12 top mirror periods. For the case of 12 top mirror periods, we have achieved fiber-to-fiber gain of 13 dB and a 0.6 nm (100 GHz) optical bandwidth (for 11.3-dB gain). A noise figure of 8.3 dB was calculated for 10-dB gain in the unsaturated regime. The saturation output power was measured to be 3.5 dBm. II. BACKGROUND The optimization of the performance of a VCSOA is complicated by a fundamental balance of device properties: the amplifier gain increases with increased mirror reflectivity, while optical bandwidth and saturation power benefit from decreased mirror reflectivity [7]. For a given reflectivity, both gain and saturation output power increase with increased pump power. The optical bandwidth, on the other hand, narrows with increased pump power. Consequently, it is impossible to maximize all parameters simultaneously. However, the mirror reflectivity and the pump level can be used to adjust the optical bandwidth to meet requirements for different applications, and all parameters can, within some limitations, be simultaneously improved, provided that the single-pass gain is changed. The single-pass gain is a function of the total thickness of the QWs and the material gain, which depends on the carrier density. For reflection VCSOAs, both amplifier gain and noise figure benefit from a bottom mirror reflectivity close to 1, and variations should only be made in the top mirror reflectivity. The gain of a VCSOA operated in reflection mode can be calculated by considering an incident optical signal with electric and propagation constant in the cavity field of amplitude of length . The output electric field is given by a sum of the fields reflected off, and transmitted through, the top mirror of the amplifier [8]

(1) In this model, the distributed Bragg reflector (DBR) mirrors are (top mirror) and replaced by hard mirrors with reflectivity (bottom mirror). The penetration depth into the DBRs is included in the cavity length and is the single-pass gain. The

275

Fig. 1. Calculated maximum amplifier gain and optical bandwidth, as functions of top mirror reflectivity, for a reflection type VCSOA. The different curves represent different values of single-pass gain g . The dots indicate lasing threshold.

first term in (1) represents the reflection off the top mirror, which experiences a phase shift but no gain. The second term is the sum of the fields transmitted through the mirror. The field exiting the cavity and the field reflected off the mirror are 180 out of phase, and, hence, vanish if they have the same amplitude. This occurs when (2) This determines the gain for full extinction and is discussed below in the section on switching. Since the power is proportional to the square of the field, the . An expression for the gain is given by and then squaring gain is thus obtained by dividing (1) by the result

(3) Here, is the single-pass phase detuning normalized to the cavity resonance. If is set equal to zero, (3) can be used to calculate the maximum gain. Fig. 1 (bottom) shows calculated maximum gain as a function of top mirror reflectivity for values of single-pass gain ranging from 1% to 4% in 1% steps. The maximum gain is, in this case, limited by lasing threshold (showed in the graph as dots). For lower values of top mirror reflectivity, the amplifier gain is ultimately limited by the highest achievable carrier density. In practice, the cavity resonance frequency and the material gain spectrum changes with temperature. Maximum achievable gain might, thus, be limited by the separation of the cavity mode and the gain peak due to heating of the active region. The high values of gain (up to 50 dB) suggested by the graph are difficult to achieve in practice, as operation further away from lasing threshold is usually necessary. The graph shows that high reflectivity yields higher gain for a given single-pass gain, but high

276

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 37, NO. 2, FEBRUARY 2001

gain can be achieved at lower reflectivities if the single-pass gain is increased. The following equation for the optical bandwidth, defined as the full width at half maximum (FWHM) of the gain spectrum, can be obtained from (3)

(4) Here, is the refractive index of the cavity, and is the velocity of light in the vacuum. Fig. 1 (top) shows the calculated optical bandwidth as a function of the top mirror reflectivity for the same values of single-pass gain as for the calculated maximum gain. The cavity length used in these calculations is 2.2 m, which is the cavity length of the device presented later in this paper. For a gain of 3 dB, the power of the initial reflection off of the top mirror equals half the power of the amplified signal (coupling losses neglected), and the bandwidth approaches infinity. The graph reveals that in order to achieve wide optical bandwidth, the reflectivity of the top mirror should be decreased but, for a given top mirror reflectivity, the bandwidth increases with decreased single-pass gain. For low-input signal powers the gain is constant, resulting in a linear relation between input and output signal power. As the power of the input signal is increased, the carrier consumption increases, and the amplification of stronger input signals is naturally limited by the available carrier density. In order to push the saturation of the VCSOA toward higher input powers, it is desirable to have a high single-pass gain, but a short photon cavity lifetime. This would yield a high gain after fewer roundtrips in the cavity, and can be obtained by increasing the pump power while reducing the mirror reflectivity. In summary, the amplifier gain for small input signals at a given pump power is greatest for high top mirror reflectivity. On the other hand, it is desirable to decrease the top mirror reflectivity to achieve a wider optical bandwidth and higher saturation power. By increasing the pump power, or otherwise increase the single-pass gain, this can be obtained without sacrificing total amplifier gain [9]. The dominant noise in an SOA is signal-spontaneous beat noise, which arises from the amplified spontaneous emission , the noise figure , defined as (ASE). For amplifier gain output signal-to-noise ratio (SNR) over input SNR, for an SOA is given by [8] (5) is the population inversion parameter, defined as . Due to the high inversion typical of a VCSOA, [7]. is the noise excess coefficient, which equals one for TWAs, but is a function of mirror reflectivities for FPAs. For reflection-mode VCSOAs, is a function of the bottom , [7]. This mirror reflectivity only, and for gives a noise figure of an ideal VCOA of 2 (or 3 dB), just as

where

Fig. 2. Schematic of wafer bonded VCSOA structure, showing direction of pump beam and 1.3-m signal.

for an ideal TWA. However, the noise figure is degraded by the coupling efficiency. The coupling efficiency of signal light into the amplifier is higher for a VCSOA than for in-plane SOAs due to the circular symmetry of the vertical-cavity active region. This clearly favors the vertical-cavity design. Collection of the output signal should naturally be maximized while the collection of ASE is kept to a minimum. A collecting lens with a high numerical aperture (NA) makes the coupling of signal into the fiber easier. A low NA, on the other hand, makes coupling more difficult but cuts out more of the ASE, due to the higher collimation of the signal beam. The noise figure in the unsaturated regime can be calculated , the fiber-to-fiber gain from the fiber coupled ASE density , and the photon energy , and is given by [10] (6) The first term on the right-hand side represents the signal-ASE beating, and the second term is the shot noise. III. DEVICE STRUCTURE AND EXPERIMENTAL SETUP A schematic of our VCSOA is shown in Fig. 2. The device is a planar structure consisting of an InP/InGaAsP active region that was wafer bonded to two GaAs/AlAs DBR mirrors, thereby utilizing the high reflectivity and favorable thermal properties of the GaAs/AlAs material system. Conditions for wafer bonding are reported elsewhere [11]. The active region consists of three sets of seven compressively strainedInAs P QWs, surrounded by strain compensating In Ga P barriers, and was grown by metal organic chemical vapor deposition (MOCVD). The undoped mirrors were grown . by molecular beam epitaxy (MBE). The cavity length is The fused interfaces are located at nulls in the standing wave pattern, while the three sets of QWs are located at the three central peaks. No patterning or lithography was performed on the sample, except to facilitate wafer bonding; the lateral dimensions of the active region are defined by the pump laser beam. In order to minimize the loss of pump power, the entire structure is undoped and the QWs are the only layers that allow band-to-band absorption of 980 nm light. After wafer

BJÖRLIN et al.: LONG WAVELENGTH VCSOAs

277

Fig. 3. Schematic of experimental setup.

bonding, the GaAs substrate on the top mirror was removed, and the substrate on the bottom mirror was lapped down to about 150 m. The bottom substrate was then polished in order to minimize the scattering of pump light, and anti-reflection coated for 1.3- m light in order to minimize interference in the cavity due to back reflection. The bottom mirror has 25 periods, of 0.999. The giving a calculated bottom mirror reflectivity number of top mirror periods was varied in order to allow a more thorough characterization of the device. Individual mirror periods were removed by selective wet etching. A VCSOA for 1.55 m could be fabricated using the same technique. The experimental setup that was used is shown in Fig. 3. An external cavity tunable laser diode was used as a signal source and the input signal power was controlled by means of a variable optical attenuator. The signal was coupled into and out of the amplifier through a fiber and lens. The beam waist of the input signal was measured to be 4 m and the depth of focus was 20 m. A circulator was used to separate the output signal from the input signal and an optical spectrum analyzer (OSA) was used to monitor the spectrum of the output signal. The total coupling loss, including loss in the circulator, was measured to be 4 dB. The VCSOA was optically pumped through the substrate and bottom mirror. A 980-nm laser diode was used as a pump source. In order to achieve the high energy density needed to generate sufficient carrier density, the beam was first collimated and then focused down onto the VCSOA. The beam waist of the focused pump beam was measured to be 8 m and the depth of focus was 100 m, which is sufficient to maintain the same spot size throughout the QWs. IV. AMPLIFIER RESULTS Measurements were made at several different spots on the sample, for samples with 13 and 12 top mirror periods, corresponding to calculated top mirror reflectivities ( ) of 0.98 and 0.973, respectively. The VCSOA gain was measured as a function of wavelength, input power and pump power. Due to the growth nonuniformity of the sample, the wavelength at the gain peak and the required pump power to reach lasing threshold varied slightly between measurements at different spots. The peak wavelength at constant pump power varied between 1315 and 1318 nm, in good agreement with the thickness variation across MOCVD grown wafers. The peak wavelength also varied with pump power due to heating of the active region. Fig. 4 shows fiber-to-fiber gain as a function of pump power for three cases: 13 top mirror periods and input signal powers of 25 and 12 dBm, and 12 top mirror periods and input

Fig. 4. Fiber-to-fiber gain as a function of pump power for different number of top mirror periods and input signal powers. For 13 top mirror periods lasing 95 mW, for 12 top mirror periods lasing threshold of the VCSOA was P threshold was P = 125 mW.

=

signal power of 25 dBm. For the case of 25-dBm input signal power, the operation is well within the unsaturated regime. For 13 top mirror periods, a gain of 11 dB was measured for a pump power of 80 mW. Higher gain was observed for higher pump power, but as lasing threshold was approached, the gain was unstable and the optical bandwidth became very narrow. The VCSOA lased at a pump power of 95 mW. When the input signal power was increased to 12 dBm, saturation was approached and the gain decreased. For 12 top mirror periods, the lasing threshold increased to 125 mW, allowing for higher pump powers than in the previous case. For a pump power of 112 mW, a fiber-to-fiber gain of 13 dB was observed for 25-dBm input signal power. For the case of 12 top mirror periods, the top mirror reflectivity showed some nonuniformity across the sample, which can be attributed to surface roughness caused by the removal of mirror periods. In order to investigate the best values of optical bandwidth and saturation power that can be achieved with the present structure, measurements of these properties were taken on areas of the sample where the reflectivity is low. In these areas, fiber-to-fiber gain on the order of 13 dB could still be achieved, although higher pump power was required, but the VCSOA could not be brought to lasing threshold. Fig. 5 shows fiber-to-fiber gain versus wavelength for 13 top mirror periods, 25-dBm input signal power, and two different and . The dots are measurepump power levels, ments and the lines are curve fits based on (3). As expected, the bandwidth increased with decreased gain. For 8.3-dB gain, a bandwidth of 33 GHz (0.2 nm) was measured. For 5-dB gain, the bandwidth increased to 50 GHz (0.3 nm). The values used 0.999, , and % and in the fit are 1.006%, respectively. These values of single-pass gain are low was not optimized. The slightly for 21 QWs, indicating that lower value of compared to the previously calculated value of 0.98 based upon epi-layer design is attributed to surface roughness. Removal of another top mirror period resulted in improved optical bandwidth. Fiber-to-fiber gain versus wavelength for 12 top mirror periods is shown in Fig. 6. The maximum gain is 11.3 dB and the optical bandwidth is 100 GHz (0.6 nm). The pump power was 120 mW. Lasing threshold was never reached at the

278

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 37, NO. 2, FEBRUARY 2001

Fig. 5. Fiber-to-fiber gain versus wavelength for 13 top mirror periods and 0:8P and P = 0:7P . For the higher level an optical two pump levels, P bandwidth (FWHM) of 0.2 nm (33 GHz) was measured. Decreased pump power resulted in a wider bandwidth: 0.3 nm (50 GHz). The input signal power was 25 dBm.

=

0

Fig. 6. Fiber-to-fiber gain versus wavelength for the case of 12 top mirror periods. The input signal power was 20 dBm. An optical bandwidth of 0.6 nm (100 GHz) was measured for a maximum gain of 11.3 dB.

0

spot on the sample were these measurements were taken. Values 0.999%, 0.91%, used in the curve fit for this case are 1.035%. This value of corresponds to 8 GaAs/AlAs and periods, provided that the surface is even. Fig. 7 shows fiber-to-fiber gain versus input signal power for the case of 13 mirror periods and two different pump power and . The dots are measurements and the levels, , lines are curve fits based on the relation is the unsaturated gain, is the input signal power, where is the saturation input power. The gain in the unsatuand rated regime is flat within 0.5 dB. The gain saturated, corresponding to a 3-dB drop in gain, at input powers of 10 and 9.5 dBm, respectively. The output saturation power is 9.5 and 6.0 dBm for . Fiber-to-fiber gain dBm for vs. input signal power for the case of 12 mirror periods is shown in Fig. 8. Again, this was measured where the reflectivity is low and the VCSOA did not lase. The unsaturated gain is 10.2 dB, and the linearity in the unsaturated regime is impeccable. The saturation output power was measured to be 3.5 dBm. These measurements confirm theoretical predictions. The decreased top mirror reflectivity allowed for higher pump power,

Fig. 7. Fiber-to-fiber gain versus input signal power for 13 top mirror periods and two different pump levels. The saturation power corresponds to a 3-dB gain drop. Both maximum gain and saturation output power increased for increased pump power.

Fig. 8. Fiber-to-fiber gain versus input signal power for the case of 12 top mirror periods. A saturation output power of 3.5 dBm was measured.

0

which yielded higher saturation output power without sacrificing gain. Furthermore, the gain and optical bandwidth were simultaneously improved, which is in agreement with expected trends (Fig. 1). A temperature controlled stage or an optimized cavity mode-gain peak offset may increase the gain at higher pump powers. This would allow for a further decreased top mirror reflectivity, which would yield wider optical bandwidth. Preliminary noise figure calculations were made using (6). Spectra of the input signal, output signal, and ASE from the VCSOA (12 mirror periods) are shown in Fig. 9. A fiber-to-fiber gain of 10 dB and an optical SNR of 27 dB were measured. The input power was 25 dBm, which is well below saturation and hence the ASE level can be assumed to be unaffected by the input signal. The noise figure, given by (6), was in this case 8.3 dB. This value is slightly higher than expected, which can be attributed to a few different things. First, as Fig. 9 reveals, the ASE level was not constant throughout the linewidth of the output signal from the amplifier. The peak value of the ASE , which might result in too high was used to determine the a value. The collecting lens was chosen to maximize collection of the signal. A lens with a lower NA would cut out more ASE. A smaller active region (smaller pump spot size) would further

BJÖRLIN et al.: LONG WAVELENGTH VCSOAs

Fig. 9. Spectra of input signal, output signal, and ASE, showing 10-dB fiber-to-fiber gain and 27-dB optical SNR.

decrease the ASE. Optimization of the collecting lens, as well as the pump spot size might, thus, reveal a better noise figure. Further measurements are also required to determine the noise figure in the saturated regime. The VCSOA presented in this work is a gain-guided planar structure with no lateral carrier confinement. The device was optically pumped by an external pump laser. This structure is naturally associated with substantial losses and low efficiency for coupling of pump power into the active region. The size and alignment of the pump beam and the input signal is of major importance for the efficiency of the device. A small pump beam waist is vital to achieve a sufficient pump energy density, but if the beam waist is too small, the carriers will diffuse laterally out of the active area before they contribute to the gain. This has been verified by an increase in threshold pump power density for small spot sizes in optically pumped VCSELs [12]. Poor overlap of pump beam and signal mode decreases the internal efficiency. This could be improved by coupling the input signal and pump beam into the device through the same lens. Measurements of pump light being reflected off the surface and transmitted through the structure revealed that over 50% is lost due to these effects. Further pump power is lost due to scattering, and carriers are lost due to lateral diffusion out of the active region. Lateral carrier confinement, and confinement of the optical mode, by etched mesas and oxide apertures would certainly improve the efficiency of the device. Confinement of a single optical mode would also prevent carrier loss to amplification of side modes, and hence increase the saturation power. The lateral dimension of the active region in the VCSOA presented here is defined by the size of the pump beam. A small active region minimizes the amount of fiber coupled spontaneous emission and, thereby, improves the noise figure. The gain spectrum of a VCSOA is typically narrow and the gain peak wavelength shifts with temperature. A coupled cavity design has been proposed to broaden the optical bandwidth [13]. The incorporation of an electrostatically controlled mirror into the design would allow for high-precision tuning of the cavity length. Gain peak shifts due to temperature changes could, thus, be compensated. This would also enable tuning between different channels in multi wavelength systems. Electrostatically

279

Fig. 10. Fiber-to-fiber gain versus pump power for the case of 12 top mirror periods and an input signal power of 15 dBm. The dip is caused by interference between the cavity mode and the initial reflection off the top mirror. An extinction ratio of 46 dB was measured.

0

controlled mirrors have enabled the demonstration of tunable Fabry–Perot filters [14] and tunable VCSELs [4].

V. OTHER POTENTIAL APPLICATIONS VCSOAs, like in-plane SOAs, can be used for various types of optical processing such as wavelength conversion [15] and switching [16], a big advantage compared to other components being the fact that the amplifier gain compensates for coupling losses. For the case of switching, a VCSOA operated in the transmission mode absorbs an incoming signal when not biased/pumped, yielding a very large extinction ratio in the output signal. Switching has been demonstrated in VCSOAs operated in reflection mode. An extinction ratio of 14 dB and a switching time of 10 ps were reported for a vertical-cavity amplifying switch operated in reflection mode at 1.5 m [17]. In order to obtain a better on–off variance from a reflection VCSOA, the gain level given by (2), at which the output signal vanishes, can be used as the off state, provided that the signal is inserted at right angles through the top mirror. Full extinction, as well as maximum gain, are only achieved at the cavity resonance frequency. However, as the pump power is increased, the cavity resonance frequency is red-shifted due to heating. Good extinction ratios can be obtained if the gain spectrum is wide enough to cover the red shift, or if the shift corresponds to the distance between two transverse modes. The mode separation can be controlled by an oxide aperture [18], or, for the case of an optically pumped planar device, by the diameter of the pump beam that defines the active region [19]. For high-speed operation, however, operation in the fundamental mode is desirable. A wider bandwidth, temperature control, or otherwise stabilized cavity resonance frequency might be necessary. Fig. 10 shows fiber-to-fiber gain as a function of pump power for the case of 12 mirror periods. The optical bandwidth does not cover the red shift in this case, but as the pump power is increased, a side mode is shifted over to the signal wavelength ( 1 nm). The input signal is 15 dBm and the maximum gain is 7 dB, at a pump power of 80 mW. The dip occurs at 5 mW of

280

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 37, NO. 2, FEBRUARY 2001

pump power, and reaches down to 39 dB, yielding an extinction ratio of 46 dB for a 12-dB modulation of the pump power. This is a very promising result for switching applications. The typically narrow bandwidth of VCSOAs suppresses the buildup of ASE when devices are cascaded into switch arrays, and the switching time of a VCSOA should be fast enough for packet switching [17]. In-plane Fabry–Perot SOAs have been used for wavelength conversion [20], and this could also be performed by a VCSOA. A modulated incoming signal will modulate the refractive index of the cavity, and, thereby, shift the resonance frequency of the cavity. The information on the modulated signal could thus be transferred over to a CW signal close to, or at, the cavity resonance frequency. The polarity of the information can be maintained or inverted in the conversion. The incoming modulating signal would have to be within the Fabry–Perot mode, or outside the stop-band of the top DBR in order for it to enter the cavity. Wavelength conversion could also be achieved using cross-gain modulation (XGM) between a cw signal and a modulated signal. Other potential applications for VCSOAs include optical repeaters for inter-board connections [21], and optical modulators [22]. Inherent to the design and structure of VCSOAs is the possibility of fabricating integrated 2-D arrays. In addition, wafer bonding allows for tuning of the cavity length across the wafer, affording the possibility of arrays of amplifiers working at different wavelengths. An optically pumped, wavelength-division multiplexed (WDM) array of long-wavelength VCSELs has already been demonstrated [23]. The optical bandwidth of a VCSOA can be modified in the device design. If a proper bandwidth is chosen, the amplifier will function as an amplifying filter. Careful design might allow a 2-D array of amplifying filters working at wavelengths corresponding to channels in a WDM system. This would be very attractive as pre-amplifiers before a 2-D array of detectors in a receiver module.

narrow optical bandwidth of VCSOAs prevents the build-up of ASE when devices are cascaded into switch matrices. We also propose a 2-D array of amplifying filters, to be used as pre-amplifiers in receiver modules in a WDM system. The filter bandwidth is adjusted by choosing appropriate mirror reflectivites and pump level. Wafer bonding allows for tuning of the cavity length across the wafer, yielding amplification of different wavelengths. Other possible applications include amplifying modulators, optical repeaters for inter-board connections, and wavelength conversion.

VI. CONCLUSION We have outlined the properties and potential applications for vertical cavity optical amplifiers, as well as presented our results for a 1.3- m VCSOA. We have shown that by reducing the top mirror reflectivity, the amplifier gain, optical bandwidth, and saturation output power could be simultaneously improved. Our device was fabricated using wafer bonding, operated in reflection mode, and optically pumped by an external 980-nm pump laser. Results were presented for 13 and 12 top mirror periods. Our results for 12 top mirror periods are, in summary, 13-dB fiber-to-fiber gain, 0.6-nm (100-GHz) bandwidth, saturation output power of 3.5 dBm, and a noise figure of 8.3 dB. All results are CW operation at room temperature. These results could be further improved if etched mesas and oxide apertures were incorporated into the design, which would also improve the efficiency of the device. We have also briefly investigated the switching properties of our device. By modulating the pump, we have obtained a 46-dB extinction ratio in the output power, with the maximum output power corresponding to a 7-dB fiber-to-fiber gain. The gain would compensate for coupling losses, and the typically

ACKNOWLEDGMENT The authors would like to thank C. Harder and S. Mohrdiek of Uniphase Laser Enterprise AG, Zurich, Switzerland, for providing the pump lasers.

REFERENCES [1] Y. Nishida, M. Yamada, T. Kanamori, K. Kobayashi, J. Temmyo, S. Sudo, and Y. Ohishi, “Development of an efficient praseodymium-doped fiber amplifier,” IEEE J. Quantum Electron., vol. 34, pp. 1332–1339, Aug. 1998. [2] A. J. Stentz, “Progress on Raman amplifiers,” in OFC ’97 Tech. Dig., 1997, Paper FA1, p. 343. [3] V. Jayaraman, J. C. Geske, M. H. MacDougal, F. H. Peters, T. D. Lowers, and T. T. Char, “Uniform threshold current, continuous-wave, singlemode 1300 nm vertical cavity lasers from 0 to 70 C,” Electron Lett., vol. 34, pp. 1405–1407, 1998. [4] D. Vakhshoori, J.-H. Zhou, M. Jiang, M. Azimi, K. McCallion, C.-C. Lu, K. J. Knopp, J. Cai, P. D. Wang, P. Tayebati, H. Zhu, and P. Chen, “C-band tunable 6 mW vertical-cavity surface-emitting lasers,” in OFC 2000 Tech. Dig., 2000, Paper PD13, postconference edition, pp. PD131–PD13-3. [5] R. Lewén, K. Streubel, A. Karlsson, and S. Rapp, “Experimental demonstration of a multifunctional long-wavelength vertical cavity laser amplifier-detector,” IEEE Photon. Technol. Lett., vol. 10, pp. 1067–1069, 1998. [6] D. Wiedenmann, C. Jung, M. Grabherr, R. Jäger, U. Martin, R. Michalzik, and K. J. Ebeling, “Oxide-confined vertical-cavity semiconductor optical amplifier for 980 nm wavelength,” in CLEO 98 Tech. Dig., 1998, Paper CThM5, p. 378. [7] C. Tombling, T. Saito, and T. Mukai, “Performance predictions for vertical-cavity semiconductor laser amplifiers,” IEEE J. Quantum Electron., vol. 30, pp. 2491–2499, Nov. 1994. [8] H. Ghafouri-Shiraz, Fundamentals of Laser Diode Amplifiers. West Sussex, U.K.: Wiley, 1996. [9] J. Piprek, S. Bjorlin, B. Riou, P. Abraham, Y.-J. Ciu, and J. E. Bowers, “Performance optimization of 1.3 m vertical-cavity semiconductor optical amplifiers,” in Proc. Eur. Conf. Optical Comunication (ECOC), Munich, Germany, 2000. [10] D. Derickson, Fiber Optic Test and Measurement. Englewood Cliffs, NJ: Prentice-Hall, 1998. [11] A. Black, A. R. Hawkins, N. M. Margalit, D. I. Babic, A. L. Holmes Jr., Y.-L. Chang, P. Abraham, J. E. Bowers, and E. L. Hu, “Wafer fusion: Materials issues and device results,” IEEE J. Select. Topics Quantum Electron., vol. 3, pp. 943–951, 1997. [12] H. W. Song, C. K. Kim, and Y. H. Lee, “Photo-pumped 1.3 m verticalcavity surface-emitting lasers,” J. Opt. Soc. Korea, vol. 8, pp. 111–115, Apr. 1997. [13] S. F. Lim and C. J. Chang-Hasnian, “A proposal of broad-bandwidth vertical-cavity laser amplifier,” IEEE Photon. Technol. Lett., vol. 7, pp. 1240–1242, 1995. [14] S. R. Mallinson and J. H. Jerman, “Miniature micromachined Fabry–Perot interferometers in silicon,” Electron Lett., vol. 23, pp. 1041–1043, 1987. [15] T. Durhuus, B. Mikkelsen, C. Joergensen, S. L. Danielsen, and K. E. Stubkjaer, “All-optical wavelength conversion by semiconductor optical amplifiers,” J. Lightwave Technol., vol. 14, pp. 942–954, Jun. 1996.

BJÖRLIN et al.: LONG WAVELENGTH VCSOAs

[16] G. Sherlock, J. D. Burton, P. J. Fiddyment, P. C. Sully, A. E. Kelly, and M. J. Robertson, “Integrated 2 2 optical switch with gain,” Electron Lett., vol. 30, pp. 137–138, 1994. [17] N. Bouché, B. Corbett, R. Kuszelewicz, and R. Ray, “Vertical-cavity amplifying photonic switch at 1.5 m,” IEEE Photon. Technol. Lett., vol. 8, pp. 1035–1037, 1996. [18] K. L. Lear, K. D. Choquette, R. P. Schneider Jr., and S. P. Kilcoyne, “Modal analysis of a small surface emitting laser with a selectively oxidized waveguide,” Appl. Phys. Lett., vol. 66, pp. 2616–2618, 1995. [19] R. Michalzik and K. J. Ebeling, “Generalized BV diagrams for higher order transverse modes in planar vertical-cavity laser diodes,” IEEE J. Quantum Electron., vol. 31, pp. 1371–1379, Aug. 1995. [20] B. Ma, M. Saitoh, and Y. Nakano, “All-optical wavelength conversion based on Fabry–Perot semiconductor optical amplifier,” in CLEO 2000 Tech. Dig., 2000, Paper CThG5, pp. 422–423. [21] N. Suzuki, M. Ohashi, and M. Nakamura, “A proposed vertical-cavity optical repeater for optical inter-board connections,” IEEE Photon. Technol. Lett., vol. 9, pp. 1149–1151, 1997. [22] R.-H. Yan, R. J. Simes, and L. A. Coldren, “Surface-normal electroabsorbtion reflection modulators using asymmetric Fabry–Perot structures,” IEEE J. Quantum Electron., vol. 27, pp. 1922–1931, July 1991. [23] D. I. Babic, V. Jayaraman, N. M. Margalit, K. Streubel, M. E. Heimbuch, R. P. Mirin, B. J. Thibeault, J. E. Bowers, E. L. Hu, and S. Denbaars, “Long-wavelength vertical-cavity surface-emitting laser diodes,” in Proc. Materials Research Society Symp., vol. 421, 1996, pp. 63–74.

2

Staffan Björlin received the M.S. degree in engineering physics from the Royal Institute of Technology, Stockholm, Sweden, in 2000, and is currently working toward the Ph.D. degree in electrical engineering at the University of California at Santa Barbara. His research interests include design and characterization of vertical cavity semiconductor optical amplifiers and vertical cavity lasers.

B. Riou, photograph and biography not available at the time of publication.

281

Patrick Abraham (SM’99) received the M.S. and Ph.D. degrees in material science from the University C. Bernard Lyon 1, France, in 1984 and 1987, respectively. He was a Researcher with the Laboratoire de Physico-Chimie Minérale, Centre National de la Recherche Scientifique, France, until 1988. He was then a Research Engineer at the University of California at Santa Barbara until spring 2000. He is currently with Agility Communications, Goleta, CA.

Joachim Piprek (SM’98) received the Ph.D. degree in solid-state physics from Humboldt University, Berlin, Germany, in 1986. He worked in industry and academia on design and analysis of optoelectronic devices. He is currently an Adjunct Associate Professor at the University of California at Santa Barbara. His research interests include vertical-cavity lasers, novel semiconductor materials, and advanced computer simulation.

Y.-J. Chiu (M’99), photograph and biography not available at the time of publication.

K. Alexis Black received the B.S. degree from Massachusetts Institute of Technology, Cambridge, MA, in 1995, and the Ph.D. degree in material science from the University of California at Santa Barbara in 2000. She is currently with Zaffire, San Jose, CA. Her research interests include long-wavelength VCSELs, components for DWDM, and optical networking.

A. Keating (M’95), photograph and biography not available at the time of publication.

John E. Bowers (F’93) received the Ph.D. degree in applied physics from Stanford University, Standford, CA, in 1981. He was with Honeywell and AT&T Bell Laboratories before joining the University of California at Santa Barbara in 1987, where he currently is a Professor of Electrical Engineering, Director of the Multidisciplinary Optical Switching Technology Center, and a member of the Heterogeneous Optoelectronics Technology Center and the National Science Foundation Center on Quantized Electronic Structures His main research interests are in the development of novel optoelectronic devices for the next generation of optical networks.