CW And Mode-locked Integrated Extended Cavity ... - IEEE Xplore

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 3, NO. 3, JUNE 1997

885

CW and Mode-Locked Integrated Extended Cavity Lasers Fabricated Using Impurity Free Vacancy Disordering A. Catrina Bryce, Member, IEEE, Fernando Camacho, Pasquale Cusumano, and John H. Marsh, Senior Member, IEEE (Invited Paper) Abstract—A phosphorus-doped silica (P:SiO2 ) cap containing 5 wt% P has been demonstrated to inhibit the bandgap shifts of p-i-n and n-i-p GaAs–AlGaAs quantum-well laser structures during rapid thermal processing. Bandgap shift differences as large as 100 meV have been observed between samples capped with SiO2 and with P:SiO2 . The technique has been used to fabricate GaAs–AlGaAs ridge lasers with integrated transparent waveguides. With a selective differential blue-shift of 30 nm in the absorption edge, devices with 400 m/2.73-mm-long active/passive sections exhibited an average threshold current of 9 mA in continuous-wave (CW) operation, only 2.2 mA higher than that of discrete lasers of the same active length and from the same chip. Extended cavity modelocked lasers were also investigated and compared to all active devices. For the extended cavity device, the threshold current is a factor of 3–5 lower, the pulsewidth is reduced from 10.3 to 3.5 ps and there is a decrease in the free-running jitter level from 15 ps (measurement bandwidth 10 kHz–10 MHz) to 6 ps. In addition, the extended cavity lasers do not exhibit any self-pulsing modulation of the mode-locked pulse train, unlike the all-active lasers, and the optical spectra indicate that the pulses are more linearly chirped. Index Terms— Gallium arsenide, integrated optoelectronics, mode-locked lasers, quantum-well intermixing, quantum wells, semiconductor lasers.

I. INTRODUCTION

I

MPURITY-FREE vacancy disordering (IFVD) has been proven to be a useful technique for the intermixing of GaAs–AlGaAs quantum-well structures [1]. The technique involves depositing a SiO cap onto the surface of the GaAs and annealing at temperatures of 900 C or higher. At these temperatures, Ga has a very high diffusion coefficient in SiO , an effect which was has been reported as far back as 1957 [2], although the detailed mechanism for this diffusion is still unknown. A GaAs structure capped with SiO and annealed at these temperatures leads to the out diffusion of Ga and the Manuscript received April 4, 1997; revised July 11, 1997. This work was supported by the U.K. Ministry of Defence and EPSRC under Grant no. GR/K45968. The work of P. Cusumano was supported by the European Community, within the frame of the Human Capital and Mobility Program. A. C. Bryce, F. Camacho, and J. H. Marsh are with the Department of Electronics and Electrical Engineering, University of Glasgow, Glasgow G12 8QQ, Scotland, U.K. P. Cusumano is with the Department of Electronics and Electrical Engineering, University of Glasgow, Glasgow G12 8QQ, Scotland, U.K., on leave from the Dipartimento di Ingegneria Elettrica, Universita degli Studi di Palermo, Viale delle Scienze, I-90128 Palermo, Italy. Publisher Item Identifier S 1077-260X(97)07608-9.

vacancies that are created in the GaAs diffuse rapidly through the structure [3]. When the vacancies cross the quantum-well (QW) interface, QW intermixing occurs. One of the attractions of this process is that no impurities are introduced in the process, only vacancies, unlike impurity induced disordering. In order to prevent As desorption at the high annealing temperatures and, hence, the degradation of the material, it is necessary to cap the areas which are not to be intermixed. Several dielectric caps have been used as protective caps. Early studies made use of Si N [4] as a protective cap, however such films were found to have poor reproducibility in the suppression of intermixing. Furthermore, high-purity Si N films are difficult to obtain because of the systematic incorporation of O in the film, resulting in SiO N , which can be an effective cap for inducing Ga out-diffusion [4]. Another protective cap that has been used is SrF which has proven to be very effective at suppressing intermixing [5]. Photonic integrated devices such as extended cavity lasers [6], integrated passive waveguides for distributed Bragg reflectors [7], integrated laser, waveguide and detector [8], and a modulator integrated with a waveguide [9] have been fabricated using a combination of SrF and SiO . However, it has been found that there can be surface damage to the wafer after the thermal processing, particularly when the devices being fabricated require large areas of SrF . It is thought that the damage is due to stresses induced in the SrF , SiO film, and the GaAs substrate, because of the large differences in the coefficients of expansion between these three compounds which will lead to stresses during the annealing step. Despite this, SrF has been used to produce simple photonic integrated circuits, as referenced above, and a technique for one step selective intermixing in selected areas (SISA) has been developed [10]. The SISA effect is achieved by patterning the semiconductor, using electron beam lithography, with submicron to 1 m sized features of SrF to act as a bandgap control mask, followed by deposition of SiO over the sample to act as an intermixing source. The SrF mask pattern has to have dimensions smaller than, or comparable with, the diffusion length of the point defects to allow uniform intermixing at the quantum-well depth by overlapping of the vacancy diffusion fronts. As a result, spatial control of the bandgap shift can be achieved using a single annealing step. The degree of intermixing is dependent upon the area of sample in direct

1077–260X/97$10.00  1997 IEEE

886


contact with the SiO layer. SISA was used to fabricate lasers with five different emission wavelengths on a single chip [11], the characteristics of which showed that the material was still of good electrical and optical quality after intermixing. Four-channel wavelength demultiplexers or waveguide photodetectors were also fabricated [11]. Because of the small dimensions of the SrF capped regions, the SISA process is free from the surface damage effects referred to above. The use of other fluoride compounds as annealing caps has been investigated, namely BaF , CaF , and MgF [12]. It was found that these caps also suppressed intermixing, however it proved difficult to remove them without damaging the surface of the sample which put them at a disadvantage compared to SrF . Modification of the surface oxides using a hydrogen plasma treatment has recently been demonstrated as an effective mask for QW intermixing suppression in undoped structures [13]. The main limitation of this technique is that it is only effective at relatively low temperatures (900 C), and the possibility of transferring this technique to the more common doped p-i-n laser structures used in optoelectronics is still under investigation. The use of a SiO layer doped with phosphorus (P:SiO , also called phosphosilicate glass, PSG) was reported by Rao et al. [14] to be an “universal” intermixing source for III–V compounds. P:SiO with 1% by weight of P [15] was used to induce intermixing in nominally undoped GaAs–AlGaAs shallow multiple QW structures by furnace annealing at a relatively low temperature, (850 C), and, therefore, a masking dielectric cap to prevent As desorption was not required during annealing. Here, we report the effect of using a higher percentage of P in the SiO caps and discuss the underlying physical mechanisms. The technique is then used to fabricate CW and mode-locked monolithic extended cavity lasers, and results from these device are also presented. II. MATERIAL STRUCTURE The epitaxial structures used for the work were based on that of a double quantum-well (DQW) laser. For the intermixing experiments the epilayers were grown by molecular beam epitaxy (MBE) and the structures for the laser experiments were grown by metal organic chemical vapor deposition (MOCVD), although is not believed that the growth technique has a significant effect on the intermixing behavior. The basic structure was a p-i-n separate confinement heterostructure grown on a Si doped GaAs substrate with a 0.5- mthick n-type GaAs buffer layer grown initially. The lower Al Ga As cladding layer was 1.5- m-thick and doped to a concentration of 5 10 cm using Si. The DQW region was undoped and consisted of two 10-nm-wide GaAs quantum wells, separated by a 10 nm Al Ga As barrier. The top and bottom Al Ga As barriers were 0.1- m thick. The upper Al Ga As cladding layer was 0.8- m thick and doped to a concentration of 5 10 cm using Be for the MBE layers and C for the MOCVD layers. The top contact layer consisted of 0.1 m of GaAs doped with 5 10 cm of either Be (MBE) or Zn (MOCVD).

In addition, for the intermixing studies only, an n-i-p structure was grown by MBE. The structure of the n-i-p wafer was similar to that of the p-i-n, apart from the fact that a semi-insulating GaAs substrate was used and the dopants were interchanged, i.e., where Si was added in the p-i-n structure Be was added in the n-i-p structure and vice versa. III. INTERMIXING EXPERIMENTS Dielectric films of SiO and P:SiO , both having a thickness of 200 nm, were deposited at a temperature of 330 C using a conventional plasma enhanced chemical vapor deposition (PECVD) apparatus equipped with a separate PH flow line for P doping. The P content was measured by energy dispersive X-ray (EDX) spectroscopy on a 200-nm-thick layer of P:SiO deposited, using the same PECVD system and in the same experimental conditions, on a Si wafer. A low accelerating voltage (6 keV) and beam current (0.5 nA) were used to reduce the penetration depth and probe only the P:SiO film. The measured P content was 5% by weight, with a 1:2 ratio between Si and O confirming the accuracy of the method. Encapsulants with a P content higher than 5% have not been investigated due to their increasingly hygroscopic nature [16], which would affect the quality of the dielectric film and possibly reduce its reliability. After deposition of the oxides, the samples were cleaved into squares of area 2 2 mm and annealed in a rapid thermal annealer (RTA) for 60 s in N at temperatures between 800 to 950 C. Uncapped samples were annealed at the same time to study the thermal stability of the wafer. Samples were placed face-down between two fresh GaAs proximity caps. Photoluminescence (PL) measurements at 77 K were performed on the annealed samples to assess the degree of intermixing. The blue wavelength shifts in the 77-K PL peak, as compared with the as-grown material, versus the anneal temperature for the p-i-n and n-i-p structures are shown in Fig. 1. It can be seen that SiO doped with 5% P acts as a very effective mask for preventing Ga out-diffusion, limiting the bandgap widening to no more than about 5 nm for both the n-i-p and the p-i-n structures at annealing temperatures up to 950 C. Moreover, no surface damage was produced by this dielectric cap under the annealing conditions studied here. In contrast, the blue shifts for both p-i-n and n-i-p samples capped with SiO increase with temperature and are always greater than those for uncapped samples, demonstrating the enhanced disordering promoted by SiO capping as compared with uncapped samples. Secondary ion mass spectroscopy (SIMS) profling of the samples indicated that no diffusion of either Si or P into the semiconductor was taking place. There are at least two possible explanations for the masking properties of P:SiO films. Firstly, it is well known that P:SiO films are more dense and void-free [16] than SiO . Films of P:SiO with a weight ratio P O –SiO of 4% have been used [17], as a capping material for open-tube thermal activation of Si implants in GaAs and it has been found that the diffusion coefficient of implanted Si in semi-insulating GaAs is about one order of magnitude smaller for P:SiO than for SiO . This was attributed to the presence of a reduced number of group III vacancies due to less Ga out-diffusion. Secondly,

BRYCE et al.: CW AND MODE-LOCKED INTEGRATED EXTENDED CAVITY LASERS

(a)

(b) Fig. 1. Wavelength shifts of the 77-K PL peak for SiO2 capped, P:SiO2 capped and uncapped: (a) p-i-n and (b) n-i-p samples as a function of annealing temperature with a duration of 60 s.

due to the difference in thermal expansion coefficients at the annealing temperature, a strain effect exists at the interface between GaAs and SiO during the annealing stage. The thermal expansion coefficient of GaAs is about ten times larger than SiO and, as a consequence, the SiO film is under tensile strain and the GaAs surface layer is under compressive strain. Under this condition, because of the high-diffusion coefficient of Ga in SiO , the out-diffusion of Ga atoms into the SiO film is an energetically favorable process because it minimizes the strain in the system. The addition of P into the SiO film leads to an increase in the thermal expansion coefficient [16], and a decrease in the glass softening temperature [18]. Therefore, less compressive strain will be induced during the annealing step in the GaAs surface layer and, as a result, the number of Ga vacancies will be reduced due to less Ga out-diffusion. For the above reasons and from our experimental results we postulate that a P:SiO film with 5% wt P prevents Ga out-diffusion, and hence QW intermixing, during annealing whereas SiO , as is generally accepted and experimentally demonstrated, promotes it. Moreover, based on the above arguments, we propose that surface strain induced by dielectric caps plays an important role in the IFVD process. Our results do not necessarily conflict with those reported by Rao et al. [14] because the addition of small amounts of P (1% wt) does not drastically change [16] the properties of P:SiO as compared with SiO and hence both caps can have a promoting effect on intermixing. Also to be considered is that the structure used by Rao et al. had only a very thin layer of

887

GaAs (8 nm) between the dielectric cap and the AlGaAs. It has been shown that such a thin layer can allow Al to diffuse into the SiO in high enough concentrations as to cause a reaction between the O in the cap and the Al. This reaction leads to the creation of free Si and O which then diffuse through the epitaxial layers, resulting in intermixing through impurity induced disordering [19]. It is noted from Fig. 1(a) and (b) that p-i-n and n-i-p wafers have similar degrees of intrinsic thermal stability since the uncapped samples from these wafers shifted to about the same wavelength under similar annealing conditions. Comparing Fig. 1(a) and (b) shows that SiO capped n-i-p samples exhibited larger degrees of QW intermixing than the p-in samples. This effect has been observed before [20] and is attributed to the crystal Fermi level effect [21], through which the equilibrium Ga vacancy concentration is larger in n-type material than in p-type material due to a reduction in the formation energy of group III vacancies. In the n-i-p sample, the n-doped contact and top cladding layers support Ga vacancies generated by the SiO layer, hence, larger degrees of intermixing were observed. From the integrated photonic devices point of view, the above results suggest that the growth of n-i-p structures would give a higher degree of intermixing than conventional p-i-n structures. Large differential shifts between regions masked with SiO and P:SiO have been observed with excellent surface morphology and a high degree of reproducibility. Transmission spectra of both disordered and undisordered ridge waveguides cleaved from the same sample were measured using a tuneable Ti:sapphire laser. The fabrication process involved first selective IFVD and then the waveguide formation. A 200nm-thick P:SiO film containing 5% wt P was deposited by PECVD and removed from selected areas by photolithography and wet etching in buffered HF solution. The samples were then completely capped with a 200-nm-thick SiO film deposited by PECVD and rapid thermal annealed for 60 s at 940 C, which produced a differential shift between the masked and the intermixed area of 40 nm. After RTA, the sample showed excellent surface morphology under optical microscope inspection, even at the edge interface between P:SiO and SiO . 3- m single-mode ridge waveguides were dry etched using SiCl , after which the samples were cleaved to produce guides, some of which had been intermixed and others that had not. The transmission spectra are shown in Fig. 2. The absorption edge of the intermixed waveguide is about 27 nm shorter than that of the waveguide from the area masked with the P:SiO , in good agreement with the results from 77-K PL measurement. IV. EXTENDED CAVITY LASERS Extended cavity lasers (ECL’s) were fabricated from samples 7 9 mm in size with a differential shift, between the region capped with P:SiO and SiO , of 30 nm after RTA at 940 C for 60 s, using a simple self-aligned technique. Single-mode ridge waveguides (3- m wide, 0.7- m deep) were formed by dry etching in SiCl using photoresist as a dry etching mask. After e-gun evaporation of 250 nm of SiO at room temperature, liftoff in acetone was used to

888


(measured into an NA of 0.65) to 0.19 W/A per facet. The losses in similar bandgap widened passive waveguides have been measured directly as a function of the wavelength and for TE polarization, using the Fabry–Perot resonance method, giving a propagation loss of about 1.9 cm throughout the wavelength range from 850–900 nm [22]. Below 850 nm, the loss increases due to resonant absorption in the partially intermixed QW. Assuming a logarithmic dependence [23] of the gain versus current density , (1) Fig. 2. Transmission spectra of ridge waveguides fabricated on intermixed and masked regions of the same sample after RTA at 940 C for 60 s.

and from the balance between gain and losses at threshold, the following formulas are obtained: (2)

(3)

Fig. 3. L–I curves and spectra of normal and extended cavity ridge lasers.

remove the silica from the top of ridges. Ohmic p-contact pads (NiCr–Au) were deposited on the active sections and the top GaAs layer was selectively (with respect to Al Ga As) dry etched in SiCl to prevent current from spreading to the integrated transparent waveguides. After thinning, the ncontact (AuNiGe) was deposited on the back of the samples and the contacts were alloyed at 360 C for 60 s. Finally discrete and extended cavity lasers were cleaved and soldered on Cu mounts for testing in CW operation at 20 C. Representative light versus current ( – ) characteristics and related spectra for a discrete 400- m-long laser and for an integrated device with 400- m/2.73-mm-long active/passive sections are shown in Fig. 3. The extended cavity lasers showed equal power from both active and passive ends with emission wavelengths between 853 and 855 nm. The average threshold current of a batch of discrete lasers was 6.7 0.5 mA and for a batch of integrated devices was 9.0 1.0 mA. Due to the losses introduced by the integrated waveguide, the mean threshold current increases by 33%, from 6.8 mA for the discrete laser to 9 mA for the extended cavity laser, and the slope efficiency decreases by 40%, from 0.32 W/A per facet

where and , are the losses and length, respectively, of the active and passive sections, is the optical overlap factor per well, the QW gain constant, the transparency current density, the number of wells, and is the internal quantum efficiency. In deriving (2) and (3), coupling losses between the active and passive waveguide have been assumed to be zero because no mode mismatch due to refractive index change is introduced by the IFVD process. The change in refractive index produced by intermixing the QW’s [24], will be at most 3%, and the optical overlap of the guided mode with the QW’s is 2.75% per well. The effective refractive index step will therefore be 10 , giving an interface reflection coefficient of only 10 . In order to determine the material parameters, broad-area lasers of width 75 m and with cavity lengths in the range 400–1000 m were fabricated from the same wafer as the extended cavity devices. From the usual plots of reciprocal of the external slope efficiency against cavity length and log of threshold current against reciprocal cavity length, the following material parameters were deduced: 5.2 cm , 865 cm , 68 A cm , and 70%, the values of and being in close agreement with theoretical ˚ QW’s [23]. The threshold current density for values for 100-A infinite cavity length was 216 A cm . The efficiency of the ridge devices is reduced because of current spreading. From (2), we obtain a value for of 2 cm , in close agreement with the linear loss measurement of 1.9 cm discussed above [22]. Assuming a modal reflectivity of 0.32, the mirror losses are 28.5 cm and the losses introduced by the integrated waveguide amount to 13.6 cm due to the quite high ratio of 6.87. From (3), we then predict a drop in the slope efficiency due to the integrated waveguide of 30%, which is reasonably close to the observed drop of 40% given the approximate nature of (1).


889

(a)

(b) Fig. 4. Diagram of (a) the all-active mode-locked semiconductor laser and (b) extended cavity mode-locked semiconductor laser.

Transparency could be also achieved by current injection in a nonintermixed QW but this, for long integrated devices, would require a second contact, electrically insulated from the active region, and the subsequent increase in the total injected current not always is compatible with CW operation. The behavior of a long all-active laser (AAL) is compared with an ECL in Section V, and it is seen that the threshold current is a factor of around 5 larger for the AAL. Losses in the integrated waveguides are limited mainly by p-type free carrier absorption in the upper cladding layer with a possible contribution from acceptor-like Ga vacancies [25] that have diffused into active region. A simple way to reduce the losses could be by tailoring the doping profiles together with the structural parameters of the laser structure [26]. Although no systematic laser lifetime measurements were carried out, no discernible degradation of laser performance was observed during the measurements. V. MODE-LOCKED LASERS In order to demonstrate the use of IFVD in a demanding application, we have fabricated and characterized both AAL’s and ECL’s for use as mode-locked (ML) sources. The QW intermixing was carried out as described above with the annealing conditions being 925 C for 60 s. 77-K PL measurements showed that the intermixed regions, had been blue shifted by 33 nm with respect to the as-grown material, and the active areas, under the P:SiO , had shifted by 6 nm. The differential shift between active and passive regions was, therefore, 27 nm. Single mode ridge waveguide lasers were fabricated as described earlier, except the p-type contact to the active section of the device was split, using liftoff, dividing the active region into different sections, namely a gain section and a saturable absorber section. The contacts were separated by a 10- m gap, which gave an isolation resistance of 5–6 k when the highly p-doped GaAs contact layer was removed from between them. The lasers were then cleaved. A 5-mmlong AAL and a 4-mm-long ECL were studied, the former working at around 8 GHz and the latter at around 10 GHz. The

Fig. 5. L–I curves for the ECL, from both the active and passive ends of the laser, and for the AAL.

AAL devices had a 4910- m-long gain section and an 80- mlong saturable absorber and the ECL’s had a 440- m-long gain section, a 50- m-long saturable absorber and a 3500- m-long extended cavity, as shown in Fig. 4. – characteristics for the AAL and for the ECL, shown in Fig. 5, were taken under pulsed excitation with the saturable absorber floating. The threshold current for the ECL was 32 mA, less than a third of that of the AAL which was 105 mA. A greater improvement in the threshold current was measured when the saturable absorber was short circuited with the gain section. The threshold current for the ECL, with both sections forward biased, was reduced to 18 mA, while the AAL threshold current was only reduced to 100 mA, more than a factor of five larger than that of the ECL. Apart from the improvement in threshold current, the ECL also has a major advantage compared to the AAL, in the absence of self-pulsating regimes accompanying mode locking. Depending on the negative bias applied to the saturable absorber and the current supplied to the gain section, the AAL operated in several distinct dynamic regimes, which included pure ML, pure self-pulsation, and ML combined with a deep self-pulsing envelope [27]. For all device conditions investigated, the ECL showed no self-pulsing modulation of the ML pulse train. Fig. 6 shows temporal measurements taken with a highspeed streak camera together with autocorrelation traces. In Fig. 6(a) and (c), 500-ps-long streak camera profiles are depicted. Fig. 6(a) and (b) measured from the 5-mm-long AAL, show a ML pulse train at approximately 8 GHz with 10.3-ps pulses, while Fig. 6(c) and (d), measured from the 4-mmlong ECL, shows a ML pulse train at around 10 GHz with 3.5-ps pulses. The pulse lengths were measured using a twophoton absorption semiconductor waveguide autocorrelator [28], which is sufficiently sensitive to allow us to acquire autocorrelation measurements even under pulsed excitation. These measurements show that the ECL pulse width varied from 3.5 to 6.5 ps (FWHM assuming hyperbolic secant square shape), depending on the biasing conditions, while the AAL pulse width varied from 10.2 to 13.8 ps, again depending on the biasing conditions. The increase in the pulse width in

890


(a)

(b)

(c)

(d)

Fig. 6. Temporal (streak camera) measurements from (a) the AAL and (c) the ECL, and autocorrelation data and fitted curves for (b) the AAL and (d) the ECL.

the AAL is most probably due to gain dispersion [29]. Gain dispersion effects tend to broaden the pulses more effectively when the optical pulse propagates through long sections of active waveguide. Fluctuations in mode-locked lasers include variations in both pulse intensity and pulse timing [30]. Carrier density fluctuations modulate the round trip time for the optical pulses inside the laser cavity, and cause jitter in ML devices. The root-mean-square (rms) timing jitter, under CW operation, was estimated by the frequency-domain technique [31]. The AAL jitter (10 kHz–10 MHz) was found to be around 15 ps, while the ECL jitter (10 kHz–10 MHz) was just 6 ps. The ECL shows the expected reduction in jitter levels predicted by Derickson et al. The jitter levels are larger for all-active waveguide configurations than for extended cavity configurations due to the fact that, for similar carrier density levels in each active waveguide, the phase noise level will be larger in the AAL than in the ECL roughly by the ratio of the active waveguide lengths, the former being 5-mm long and the latter being 500- m long. Optical spectra were measured using an optical spectrum analyzer with a resolution of 0.1 nm. Fig. 7 shows spectra taken from the AAL and ECL when a) the saturable absorber was not biased and b) the devices were mode-locked. In both cases, the optical spectra suffer a shift of more than 2 nm to longer wavelengths when the laser is mode-locked, they are very asymmetric and the spectral width increases. For the AAL, the spectral width increases from 0.4 nm at zero absorber bias to almost 2 nm when the saturable absorber is reverse biased and for the ECL from 0.2 to 2 nm. The optical spectrum of light pulses travelling in the laser cavity is distorted considerably during the amplification process if the refractive index becomes nonlinear, even when the pulse shape remains unchanged. Gain and absorption saturation produce a shift and distortion in the optical spectrum. The physical mechanism responsible for this shift and distortion is selfphase modulation (SPM) [32]. The time dependence of the saturated gain leads to a temporal modulation of the phase, i.e.,

Fig. 7. Spectral measurements taken with (a) the saturable absorber floating and (b) the saturable absorber reversed biased from the AAL (upper graph) and ECL (lower graph).

the pulse modulates its own phase as a result of gain saturation. The multi-shouldered structure shown in the AAL optical spectrum has been observed previously in semiconductor laser amplifiers [32], and it was shown that SPM was responsible. The spectrum from the ECL is more symmetric than that from the AAL suggesting that the pulse chirp is more linear from the ECL, a very important factor since linearly chirped pulses lend themselves well to pulse compression techniques [29]. The time-bandwidth product for the AAL was calculated to be 7, while for the ECL it was just 2.5. Previously reported extended cavity mode-locked lasers, made using regrowth techniques [33], show similarly large time-bandwidth products to that of the AAL, around 20 times larger than the theoretical value, most probably due to reflections between the active and passive section. In our case, because of the negligible reflection at the interface active/passive waveguide of the ECL, the time-bandwidth product is reduced to 2.5, and the overall performance is, to best of our knowledge, the best achieved so far with a monolithic Fabry–Perot diode laser at repetition frequencies around 10 GHz [34]. VI. CONCLUSION The use of 5 wt% P-doped SiO as an effective cap to suppress quantum-well intermixing in p-i-n and n-i-p GaAs–AlGaAs structures has been demonstrated. This cap, along with SiO , has been used to selectively intermix areas of laser samples by IFVD. The processed samples were then used to fabricate extended cavity lasers with a differential shift of 30 nm between the active and passive sections. Devices with 400- m/2.73-mm-long active/passive sections were compared to discrete lasers of length 400 m, and the performance of the ECL’s was consistent with a passive section loss of around 2 cm . The intermixing technique has also been used to fabricate extended cavity modelocked semiconductor lasers with active and passive waveguides. Comparison of the extended cavity devices with all active modelocked lasers show improved threshold current, pulsewidth, timing jitter, reduction of the time-bandwidth product, more linear chirp and absence of self-pulsations accompanying the modelocking.


REFERENCES [1] D. G. Deppe, L. J. Guido, N. Holonyak, Jr., K. C. Hsieh, R. D. Burnham, R. L. Thornton, and T. L. Paoli, “Stripe-geometry quantum well heterostructure Alx Ga10x As–GaAs lasers defined by defect diffusion,” Appl. Phys. Lett., vol. 49, pp. 510–512, 1986. [2] C. J. Frosch and L. Derick, “Surface protection and selective masking during diffusion in silicon,” J. Electrochem. Soc., vol. 104, pp. 547–552, 1957. [3] L. J. Guido, N. Holonyak, Jr., K. C. Hsieh, R. W. Kaliski, W. E. Plano, R. D. Burnham, R. L. Thornthon, J. E. Epler, and T. L. Paoli, “Effect of dielectric encapsulation and As overpressure on Al–Ga interdiffusion in Alx Ga10x As quantum-well heterostructures,” J. Appl. Phys., vol. 55, no. 6, pp. 540–542, 1989. [4] M. Kuzuhara, T. Nozaki, and T. Kamejima, “Characterization of Ga out-diffusion from GaAs into SiOx Ny films during thermal annealing,” J. Appl. Phys., vol. 66, pp. 5833–5836, 1989. [5] J. Beauvais, J. H. Marsh, A. H. Kean, A. C. Bryce, and C. Button, “Suppression of bandgap shifts in GaAs/AlGaAs quantum wells using strontium fluoride caps,” Electron. Lett., vol. 28, pp. 1670–1672, 1992. [6] J. H. Marsh, P. Cusumano, A. C. Bryce, B. S. Ooi, and S. G. Ayling, in Proc. SPIE, 1995, vol. 2401, p. 74. [7] J. H. Marsh and A. C. Bryce, “Fabrication of photonic integrated circuits using quantum well intermixing,” Mater. Sci. Eng. B, vol. B28, pp. 272–278, 1994. [8] D. Hofstetter, H. P. Zappe, J. E. Epler, and P. Riel, “Monolithically integrated DBR laser, detector, and transparent waveguide fabricated in a single growth step” IEEE Photon. Technol. Lett., vol. 7, pp. 1022–1024, Sept. 1995. [9] P. Cusumano, T. F. Krauss, and J. H. Marsh, “High extinction ratio GaAs/AlGaAs electroabsorption modulators integrated with passive waveguides using impurity-free vacancy diffusion,” Electron. Lett., vol. 31, no. 4, pp. 315–317, 1995. [10] S. G. Ayling, J. Beauvais, and J. H. Marsh, “Spatial control of quantumwell intermixing in GaAs/AlGaAs using a one-step process,” Electron. Lett., vol. 28, pp. 2240–2241, 1992. [11] B. S. Ooi, M. W. Street, S. G. Ayling, A. C. Bryce, J. H. Marsh, and J. S. Roberts, “The application of the selective intermixing in selected area (SISA) technique to the fabrication of photonic devices in GaAs/AlGaAs structures,” Int. J. Optoelectron., vol. 10, pp. 257–263, 1995. [12] I. Gontijo, T. Krauss, J. H. Marsh, and R. M. De La Rue, “Postgrowth control of GaAs/AlGaAs quantum well shapes by impurity free vacancy disordering,” IEEE J. Quantum Electron., vol. 30, pp. 1189–1195, 1994. [13] C. J. Hamilton, S. E. Hicks, B. Vogele, J. H. Marsh, and J .S. Aitchison, “Suppression of bandgap shifts in GaAs/AlGaAs multiquantum wells using hydrogen plasma processing,” Electron. Lett., vol. 31, pp. 1393–1394, 1995. [14] E. V. K. Rao, A. Hamoudi, Ph. Krauz, M. Juhel, and H. Thibierge, “New encapsulant source for III–V quantum well disordering,” Appl. Phys. Lett., vol. 66, pp. 472–474, 1995. [15] E. V. K. Rao, private communication. [16] S. K. Ghandhi, VLSI Fabrication Principles. New York: Wiley, 1994, pp. 530–532 and references therein. [17] S. Singh, F. Baiocchi, A. D. Butherus, W. H. Grodkiewicz, B. Schwartz, L. G. Van Uitert, L. Yesis, and G. J. Zydzik, “Electron-beam evaporated phosphosilicate glass encapsulant for post-implant annealing of GaAs,” J. Appl. Phys., vol. 64, pp. 4194–1918, 1988. [18] T. Y. Tien and F. A. Hummel, J. Amer. Ceramics Soc., vol. 45, p. 422, 1962. [19] L. J. Guido, J. S. Major, Jr., J. E. Baker, W. E. Plano, N. Holonyak, K. C. Hsieh, and R. D. Burnham, “Column III vacancy- and impurity-induced layer disordering of Alx Ga10x As–GaAs heterostructures with SiO2 or Si3 N4 diffusion sources,” J. Appl. Phys., vol. 67, pp. 6813–6818, 1990. [20] B. S. Ooi, A. C. Bryce, J. H. Marsh, and J. S. Roberts, “Effect of p and n doping on neutral impurity and SiO2 dielectric cap induced quantum well intermixing in GaAs/AlGaAs structures,” Semiconduct. Sci. Technol., vol. 12, pp. 121–127, 1997. [21] D. G. Deppe and N. Holonyak, Jr., “Atom diffusion and impurityinduced layer disordering in quantum well III–V semiconductor heterostructures,” J. Appl. Phys., vol. 64, pp. R93–R113, 1988. [22] B. S. Ooi, K. McIvlaney, M. W. Street, A. Saher Helmy, S. G. Ayling, A. C. Bryce, and J. H. Marsh, “Selective quantum well intermixing in GaAs/AlGaAs structures using impurity free vacancy diffusion,” IEEE J. Quantum Electron, vol. 33, pp. 1784–1793, Oct. 1997. [23] P. W. A. McIlroy, A. Kurobe, and Y. Uematsu, “Analysis and application of theoretical gain curves to the design of multiquantum well lasers,” IEEE J. Quantum Electron., vol. QE-21, pp. 1958–1963, 1985. [24] S. I. Hansen, J. H. Marsh, J. S. Roberts, and R. Gwilliam, “Refractive

[25] [26] [27] [28] [29]

[30]

[31] [32] [33]

[34]

891

index changes in a GaAs multiple quantum well structure produced by impurity-induced disordering using boron and fluorine,” Appl. Phys. Lett., vol. 58, pp. 1398–1400, 1991. S. O’Brien, J. R. Shealy, F. A. Chambers, and G. Devane, “Tunable (Al)GaAs lasers using impurity-free partial interdiffusion,” J. Appl. Phys., vol. 71, no. 2, pp. 1067–1069, 1992. R. G. Waters, D. S. Hill, and S. L. Yellen, “Efficiency enhancement in quantum well lasers via tailored doping profiles,” Appl. Phys. Lett., vol. 52, no. 24, pp. 2017–2018, 1988. K. Lau and J. Paslaski, “Condition for short pulse generation in ultrahigh frequency mode-locking of semiconductor lasers,” IEEE Photon. Technol. Lett., vol. 3, pp. 974–976, Nov. 1991. F. R. Laughton, J. H. Marsh, D. A. Barrow, and E. L. Portnoi, “The two-photon absorption semiconductor waveguide autocorrelator,” IEEE J. Quantum Electron., vol. 30, pp. 838–845, 1994. D. J. Derickson, R. J. Helkey, A. Mar, J. R. Karin, J. G. Wasserbauer, and J. E. Bowers, “Short pulse generation using multisegment modelocked semiconductor lasers,” IEEE J. Quantum Electron., vol. 28, pp. 2186–2202, 1992. D. J. Derickson, P. A. Morton, J. E. Bowers, and R. L. Thornton, “Comparison of timing jitter in external and monolithic cavity modelocked semiconductor lasers,” Appl. Phys Lett., vol. 59, pp. 3372–3374, 1991. D. von der Linde, “Characterization of the noise in continuously operating mode-locked lasers,” Appl. Phys. B, vol. 39, pp. 201–217, 1986. G. P. Agrawal and N. A. Olsson, “Self-phase modulation and spectral broadening of optical pulses in semiconductor laser amplifiers,” IEEE J. Quantum Electron., vol. 25, pp. 2297–2306, 1989. P. B. Hansen, G. Raybon, U. Koren, P. P. Iannone, B. I. Miller, G. M. Young, M. A. Newkirk, and C. A. Burrus, “InGaAsP monolithic extended-cavity lasers with integrated saturable absorber for active, passive, and hybrid mode locking at 8.6 GHz,” Appl. Phys. Lett., vol. 62, pp. 1445–1447, 1993. F. Camacho, E. A. Avrutin, P. Cusumano, A. Saher Helmy, A. C. Bryce, and J. H. Marsh, “Improvements in mode-locked semiconductor diode lasers using monolithically integrated passive waveguides made by quantum well intermixing,” IEEE Photon. Technol. Lett., vol. 9, pp. 1208–1210, Sept. 1997.

A. Catrina Bryce (M’91) was born in Glasgow, Scotland, U.K., in 1956. She received the degree of B.Sc. degree in physics from Glasgow University, Glasgow, Scotland, U.K., in 1978, the M.Sc. degree in amorphous materials from Dundee University, Scotland, U.K., in 1979, and the Ph.D. degree in phonon scattering in thin-film glasses from Glasgow University. She joined the Department of Electronics and Electrical Engineering at the University of Glasgow in 1985 as a Research Assistant in MBE. In 1987, she joined the optoelectronics group to work on nonlinear optical properties of GaInAs quantum-well structures at 1.5 m. Since then, her research work has included GaInAs–InP electrooptic modulators, quantum-well intermixing particularly in 1.55-m and 980-nm material systems and lasers at both 980 nm and 1.55 m. Her main research interests are optoelectronic integration and short pulse semiconductor lasers.

Fernando Camacho was born in Madrid, Spain, in 1970. He received the degree in telecommunication engineering from the University of Malaga, Spain, after completing his final year project at the University of Kassel, Germany, and is working towards the Ph.D. degree at the University of Glasgow, Scotland, with his thesis on investigating ultrafast optical pulse generation using modelocking semiconductor lasers. His Post-Doctoral work in Glasgow includes research into monolithic integration of lasers and waveguides. His Ph.D. work was supported by the Department of Electronics and Electrical Engineering of the University of Glasgow and by Hamamatsu Photonics Ltd.

892


Pasquale Cusumano was born in Lucca Sicula (AG), Italy, on October 11, 1964. He received a first class honors degree and Ph.D. degree in electronic engineering from the University of Palermo, Italy, in 1990 and 1995, respectively. His final year project dealt with the design and fabrication of Mach–Zehnder electrooptic modulators in LiNbO3 using Ti indiffusion technology. The subject of his Ph.D. (primarily carried out at the Department of Electronics and Electrical Engineering at the University of Glasgow, Scotland, U.K.) was the study of selective intermixing processes in GaAs–AlGaAs quantum-well laser structures for photonic integration. He was awarded an EU grant to work for two years as a Visiting Research Fellow at the University of Glasgow, where he investigated GaAs–AlGaAs laser diode amplifiers and their integration with multimode interference couplers to obtain lossless optical switches. He is currently working within the Institute of Photonics, University of Strathclyde, Scotland, where his main interests are VCSEL’s and InP-based quantum-well lasers.

John H. Marsh (M’91–SM’91) was born in Edinburgh, Scotland, U.K., in 1956. He received the B.A. degree in engineering and electrical sciences from the University of Cambridge, Cambridge, U.K., in 1977, and the M.Eng. degree in solid-state electronics from the University of Liverpool, Liverpool, U.K., in 1978, and Ph.D. degree in the LPE growth and electrical transport properties of InGaAsP alloys from Sheffield University in 1982. He joined the Department of Electronics and Electrical Engineering at the University of Glasgow in 1986, where he currently Professor of Optoelectronic Systems. His research interests are particularly concerned with linear and nonlinear integrated optoelectronic devices in III–V semiconductors. He has developed new integration technologies for photonic integrated circuits based on quantumwell devices and quantum-well intermixing. He is author or coauthor of more than 200 journal and conference papers. He was a committee member of the IEE Professional Group concerned with Optical Devices and Systems from 1988 until 1994 and a corresponding editor of the IEE Electronics and Communication Journal. He was Director of the NATO Advanced Study Institute on Waveguide Optoelectronics held in Glasgow in 1990 and is coeditor of the book with the same name. He is currently a member of the editorial board of the International Journal of Optoelectronics and Chair of the Scottish Chapter of LEOS. Dr. Marsh is a Fellow of the Institution of Electrical Engineers (FIEE) and of the Royal Society of Arts, and a member of the British Association for Crystal Growth.