Anda di halaman 1dari 11

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 40, NO.

6, JUNE 2004

629

Modeling of the Threshold Operation of 1.3-m GaAs-Based Oxide-Confined (InGa)AsGaAs Quantum-Dot Vertical-Cavity Surface-Emitting Lasers
Robert P. Sarzala

AbstractIn the paper, the self-consistent optical-electrical-thermal-gain threshold model of the oxide-confined (OC) quantum-dot (QD) (InGa)AsGaAs vertical-cavity surface-emitting diode laser (VCSEL) is demonstrated. The model has been developed to enable better understanding of physics of an operation of GaAs-based OC QD VCSELs in a full complexity of many interactions in its volume between individual physical phenomena. In addition, the model has been applied to design and optimize the low-threshold long-wavelength 1.3- m GaAs-based OC QD VCSELs for the second-generation optical-fiber communication systems and to examine their anticipated room-temperature (RT) performance. An influence of many construction parameters on device RT lasing thresholds and mode selectivity has been investigated. Some essential design guidelines have been proposed to support efforts of technological centers in producing low-threshold single-mode RT devices. Index TermsOptical fiber communication, quantum dots, semiconductor device modeling, semiconductor lasers, simulation.

I. INTRODUCTION

NTIL VERY recently, phosphide diode lasers have been generally considered to be the best laser structures for the new-generation 1.3- m optical-fiber communication systems. However, their technology has not essentially been improved upon in a long time, which means that there is still a number of acute problems with manufacturing highly reflective phosphide distributed Bragg reflector (DBR) mirrors, such as confinements of both current spreading and optical fields within phosphide structures, a high-temperature device operation, and device reliability, to name the most important ones. The above disadvantages are especially influential in the case of phosphide vertical-cavity surface-emitting lasers (VCSELs), because their technology (especially DBR mirrors) requires a very complex procedure, their production is relatively expensive, and their performance characteristics are still far from general expectations.

Manuscript received October 6, 2003; revised February 4, 2004. This work was supported by the Polish State Committee for Scientific Research (KBN) under Grant 7-T11B-073-21, Grant 4-T11B-014-25, and Grant 4-T11B-060-25. R. P. Sarzala is with the Laboratory of Computer Physics, Institute of Physics, Technical University of Ldz, 93-005 Ldz, Poland (e-mail: rpsarzal@ck-sg.p.lodz.pl). Digital Object Identifier 10.1109/JQE.2004.828228

At the same time, arsenide technology remains relatively cheap and easy to manufacture and offers manageable ways to create efficient radial electrical and optical confinements with Al O native-oxide apertures as well as much better DBR arsenide mirrors [1]. In addition, arsenide VCSELs are proven to exhibit very high reliability. The only problem now is to switch the wavelength of their generated radiation from the obvious 0.85 m to 1.3 m. Currently, it may be done using two active-region structures: (GaIn)(NAs)GaAs quantum wells (QWs) or (InGa)AsGaAs quantum dots (QDs) (other possible 1.3- m VCSEL structures are discussed in [2]). Each of these solutions offers some advantages over its counterpart, but definitely lower lasing thresholds of the latter may turn out to become the most important factor. The technology of (InGa)AsGaAs QD lasers is relatively simple: self-organized QDs (working as artificial atoms of strictly defined energy levels) are created in the strained epitaxial layer in the so-called StranskyKrastanov process. When a layer thickness exceeds a critical value for dislocation formation, the growth front spontaneously starts forming three-dimensional (3-D) islands. Afterwards, the next epitaxial layer is deposited to cover the islands and finally the lens-shaped QD structures are formed [3], [4]. (InGa)AsGaAs QD edge-emitting (FabryPerot) diode lasers have already been proven to exhibit excellent performance characteristics, in particular an extremely low threshold current of 1.5 mA [5] and a threshold current density of 38 kA/cm [6], a reasonably high slope efficiency of 0.26 W/A [7], exceptionally good high-temperature operation (an inherent property of lasers operating on discrete energy levels) with as high as 196 K [8], threshold currents less than 7 mA at 85 C and 10.5 mA at 100 C as well as 3-mW output power at 100 C [7], and 9-GHz 3-dB bandwidth under small-signal modulation [7], to name the most important features. Unfortunately, some unexpected drawbacks of these lasers have been also reported, e.g., strong Auger recombination deteriorating their high-temperature performance [9], relatively low density of states leading to gain saturation even at low excitation levels [10], relatively low internal quantum efficiency of 21.5% [11], and simultaneous two-state lasing [12]. Nevertheless, using improved -doped QD active regions [13][15], low-threshold temperature-insensitive 1.3- m QD DFB lasers with 3-dB

0018-9197/04$20.00 2004 IEEE

630

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 40, NO. 6, JUNE 2004

modulation bandwidth well above 10 GHz can be expected in the near future [7]. Threshold simulation of 1.3- m oxide-confined (OC) edge-emitting QD (InGa)AsGaAs diode lasers has been recently reported in our previous paper [16]. However, because of well-known advantages of VCSEL structures [inherent dynamic single-longitudinal-mode operation (owing to their short optical cavities), low-divergence nonastigmatic circular output beams, device geometry suitable for integration into two-dimensional (2-D) laser arrays or for monolithic integration with electronic devices, compatibility with vertical-stacking architecture, in situ testing possibility, to name the most important VCSEL properties], VCSELs are believed to be the most suited structures for fiber optical communication. The first 1.3- m GaAs-based (InGa)AsGaAs QD VCSELs have been already manufactured and their first impressive performance characteristics have just been reported (see e.g. [17][20]). However, there is still some room for the technology and structure improvements at this relatively immature development stage. The physics of QD lasers is often different from physics of conventional diode lasers. Therefore, structures regarded as optimal for hitherto existing diode lasers do not have to remain optimal for QD ones. Hence, the present technology of these lasers needs some theoretical support to optimize conventional diode-laser structures or even to discover completely new laser structures specially designed for this purpose. Accordingly, in this paper, the threshold operation of GaAs-based (InGa)AsGaAs OC QD VCSELs is accurately simulated and carefully analyzed to enable better understanding of their physics, to determine their weak and strong points, to optimize their device structures for a low-threshold operation, and to propose some guidelines for their successful design. II. PRELIMINARY CALCULATIONS Threshold optical gain necessary for VCSEL lasing may be expressed as (1) is the cumulative active-medium thickness (including where and stand for reflectivity coefficients of all QD layers), is the resonator length, and stands both DBR mirrors, for averaged internal cavity losses. Then, for a typical stack of 6-nm QD layers and neglecting for a moment three internal cavity losses, room-temperature (RT) threshold gain is and to at least equal to at least 2800 cm for . 4200 cm for Optical gain achievable in QD layers depends on their sur, their size uniformity expressed with the aid face density of the full-width at half-maximum (FWHM) of their gain spectrum, and the active-region carrier concentration . While typical QD densities achieved presently for the (InGa)AsGaAs cm , much higher values, exQDs are cm , have also been reported [7], [21][24]. ceeding even FWHM, typically equal currently to 3050 meV, may be reduced to the 2025-meV area with the best achievements even below 20 meV [25][28].

Quite recently, available technology has ensured the best meV. Then, for a reasonable achievable value of cm (because active-region carrier concentration of of gain saturation, its further increase is not recommended, cf. Section V-A), optical gain of 3000 cm would require excm . Fortunately, recent tremely high QD densities of technology progress has been followed by more uniform QD-size distributions, which means lower FWHM values. In particular, cm to achieve the above optical gain, for the same carrier concentration, a standard FWHM value of 30 meV enables cm , and reduction of the required QD density to a still achievable FWHM value of 20 meV makes it possible to cm for reduce the required carrier concentration to cm . The last two quite a reasonable QD density of and meV parameters together with are used in our simulation as a standard data set. III. MODEL General rules of 3-D modeling of a VCSEL operation are well known and have been already described in many papers. Their review may be found in [29], where an interested reader may find a more detailed description of interesting physical phenomena crucial for VCSEL operation. Therefore, here the electrical, optical, and thermal VCSEL models used in the simulation are described very briefly. Besides, it should be stressed that, because of very lowRTlasingthresholdsofQDlasers,determinedtemperatureincreases within their volumes are very low. Therefore, the thermal VCSEL model is added here only to complete all important interactions between individual physical phenomena taking place within VCSEL volumes. Gain phenomena within QD active regions, on the other hand, are relatively new, especially when applied to VCSEL structures. Therefore, they are described below in more detail. A. Electrical, Thermal, and Optical VCSEL Models The finite-element (FE) self-consistent electrical VCSEL model is based on solving the Laplace equation with the 3-D profile of the electrical conductivity. The conductivity is determined in each algorithm loop taking into account not only the layered VCSEL structure but also the current 3-D profiles of carrier concentration and temperature. Effective conductivity of the active-region area is calculated from experimental characteristics reported in [30]. Afterwards, the 3-D current-density distribution is found from the differential Ohm law and the analogous carrier-concentration profilefrom the below-threshold diffusion equation [31]. From experimental results reported by Shchekin and Deppe [8], who have found in 1.3- m InAs QD edge-emitting diode lasers that the carrier leakage over the barriers is insignificant even at high temperatures, we can neglect this effect and assume that all carriers injected into the active region are in fact injected into its QDs. RT values of the model parameters are listed in Table I. In the FE thermal model, the following heat sources are included: the nonradiative recombination and the reabsorption of spontaneous radiation within the active region, the volume Joule heating in all structure layers, and the barrier Joule heating in the contacts. The top and side crystal walls are assumed to be ther-

SARZAA: MODELING OF THE THRESHOLD OPERATION OF 1.3- m GaAs-BASED OC QD VCSELs

631

TABLE I VALUES OF ELECTRICAL AND RECOMBINATION MODEL PARAMETERS

Supposing that potentials in the conduction and the valence bands are identical, the following equality holds: (3) where and are effective masses of electrons in the conduction band and heavy holes in the valence band, respectively. In unbiased QD lasers, carrier distributions on successive QD energy levels are properly described by the FermiDirac statistics. This statistics also remain valid for relatively long radiative recombination times (c.f. [4, p. 156]). However, in strongly biased QD lasers, some saturation effects in population distributions of carriers on the QD levels are observed as a result of slow relaxation processes. Those effects have been found to be proportional to the laser excitation level, i.e., to its output power [40], [41]. Nevertheless, for the threshold laser operation considered here, nonequilibrium distributions of carriers in QDs may be neglected [41]. and of the above carriers are deterQuasi-Fermi levels mined by the following equations: (4a)

mally isolated whereas all walls of the heat sink not adjoining the laser crystal are assumed to be kept at the RT of the ambient. Thermal conductivities [36] are dependent not only on layer compositions, but also on temperature. The optical VCSEL model is based on the effective frequency method (EFM) proposed by Wenzel and Wnsche [37]. Optical properties of each structure layer and possible gain or absorption processes within it are described by 3-D profiles of both parts (real and imaginary) of its complex index of refraction. The above profiles are strictly related not only to a device layered structure but also to the 3-D profiles of temperature and carrier concentrations. RT values of refractive indices and absorption coefficients are listed in Table II. In the EFM method, complex optical fields in VCSEL resonators are governed by two mutually interrelated nearly one-dimensional (1-D) wave equations along both the axial and radial directions. The equations are solved using a self-consistent algorithm, because the effective frequency is present in both of them. As their boundary conditions, outgoing plane waves at the bottom and at the top surfaces of the laser cavity and outgoing cylindrical wave at the sufficiently large radial distance are assumed. For each radiation mode, its lasing threshold is determined from the condition of the real propagation constant associated with this mode. B. Gain Model The gain is postulated to be an effect of electrons and heavy holes recombination. The model is based on the assumption that energetic states associated with a quantum dot may be approximated by: the 2-D harmonic oscillator [38], [39], i.e., the 2-D parabolic potential, in the pn junction plane; the finite (InGa)AsGaAs QW of a width equal to the dot height in the direction. The first assumption follows from the observation that successive level pairs (in the conduction and the valence bands) are approximately energetically located at equal distances. It is also assumed in the gain calculations that all carriers injected into the active layer are in fact injected into its QDs (see [4, p. 185]). Validity of this assumption has been recently confirmed by Shchekin and Deppe [8], who have found experimentally the carrier leakage in a similar laser practically insignificant even at high temperatures. and denote the first energy levels in respective Let ) are bands. Hence successive energy levels (for given by (2a) (2b)

(4b) where in a nominator is associated with a level degeneration, is the Boltzmann constant, is the active-region electron (hole) concentration, and is the volume density of dots rein the following way: lated to the analogous surface density , with the dot height. Since broadening related to various dot sizes is much stronger than analogous broadening associated with finite carrier lifetime, we can neglect the latter. Thus we avoid integration in the gain formula

(5) where stands for the light velocity in vacuum, is the index is the electron rest mass, is the vacuum of refraction, is the momentum matrix element (asdielectric constant, and sumed to be of the form known for standard QWs), are the FermiDirac functions for the recombining carriers, and is assumed to be the following sum of the density of states the Gaussian distribution density functions: (6) Values of the dispersions energies (related to FWHM) and the (7) are taken from the plot of photoluminescence intensity versus wavelength (photoluminescence spectra). In the above equais the energy gap and is the number of photolutions, minescence peaks taken into consideration. During the fitting, and . (3) is providing an additional relation between

632

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 40, NO. 6, JUNE 2004

TABLE II LAYER THICKNESSES AND RT VALUES OF REFRACTIVE INDICES AND ABSORPTION COEFFICIENTS IN THE 1.3-m GaAs-BASED OC QD (InGa)AsGaAs VCSEL UNDER CONSIDERATION. THERE ARE 28 PAIRS OF THE =4 LAYERS IN THE p-SIDE DBR AND 34 PAIRS OF THE ANALOGOUS LAYERS IN THE n-SIDE DBR. THE ACTIVE REGION CONTAINS THREE QD LAYERS SEPARATED BY TWO BARRIERS. TWO VALUES OF THE ABSORPTION COEFFICIENT ARE CONSIDERED FOR THE p-TYPE SPACER, WHICH ARE ANALYZED IN SECTIONS V-B AND V-C

C. Interactions Between Individual Physical Phenomena In the simulation, all important interactions between individual physical phenomena are considered, including: 1) thermal focusing, i.e., the temperature dependence of refractive indices; 2) self-focusing, i.e., the carrier-concentration dependence of the above indices; 3) temperature- and carrier-concentration dependences of thermal conductivities; 4) temperature-, dopant-, and carrier-concentration dependences of electrical conductivities; 5) temperature- and carrier-concentration dependences of the active-region energy gap. Accordingly, 3-D profiles of all model parameters are determined not only on the basis of various chemical compositions of structure layers but also taking into account (using the self-consistent calculation algorithm with the internal electrical-thermal self-consistency loop and the external gain-optical one) 3-D profiles of temperature, current density, radiation intensity, and carrier concentration within a whole device volume. In the above, temperature and carrier concentration dependences of electrical resistivities have been taken from [42, Appendix B] and analogous dependences of refractive indices from [43] to [46]. IV. STRUCTURE Let us consider a typical structure of the 1.3- m OC GaAsbased (InGa)AsGaAs QD VCSEL shown in Fig. 1. Thicknesses and compositions of all of its structure layers are given in Table I. The laser active region is similar (but not identical) to the one reported by Ledentsov et al. [17], [18], [20], [30]. It is composed of a stack of three QD 6-nm (InGa)As layers cm separated by the 25-nm GaAs barriers, both doped to cm [30]. This solution is applied instead of a single QD layer to overcome, by increasing an effective QD density, problems with QD gain saturation. The active region is sandwiched by the p-type and n-type radial-current-spreading

cm [30]. Twenty-eight GaAs layers doped to at least periods of quarter-wave GaAsAl Ga As layers and analogous 34 periods of GaAsAlAs layers are assumed to be the upper and bottom DBR resonator mirrors, respectively. Their diameters are equal to 50 and 100 m, respectively. Two (n-side and p-side) annular contacts are deposited on both GaAs spacer layers, working also as radial-current-spreading layers. Internal contact diameters are postulated to be equal to 54 (p-side) and 74 m (n-side), respectively, whereas external onesto 70 and 100 m. Typical values of their contact resistances are equal to cm (n-side) and cm (p-side) (the reported values, see, e.g. [47] and [48], are much lower). A radial selective oxidation (transforming AlAs into Al O native oxides) is proposed to create both the electrical (to funnel current spreading from annular contact toward the central active region) and the optical (to confine an optical field in a radial direc. A nominal tion) oxide apertures of an assumed diameter of thickness of oxide apertures is assumed to be equal to 15 nm. V. RESULTS Some results of our simulation and proposed structure optistructure of the OC GaAs-based mizations of the nominal 1.3- m (InGa)AsGaAs QD VCSELs (see Fig. 1 and Table II) are described in successive subsections. A. Optical Gain Gain spectra of QD active regions depend on QD densities , uniformity of their size distribution (expressed by . FWHM), and free-carrier active-region concentrations Dependence of their RT maximal values on carrier concenare plotted in Fig. 2 for three different QD densities, trations cm , cm , and cm and i.e., three FWHM values [37 meV [Fig. 2(a)], 30 meV [Fig. 2(b)], and 20 meV [Fig. 2(c))]. Two recombination mechanisms are considered: solid lines correspond to a desired recombination between fundamental QD levels (followed by an emission of m), whereas dashed radiation of the wavelength

SARZAA: MODELING OF THE THRESHOLD OPERATION OF 1.3- m GaAs-BASED OC QD VCSELs

633

Fig. 1. Typical structure of the 1.3-m GaAs-based OC (InGa)AsGaAs QD VCSEL.

ones correspond to a recombination between excited levels m . An increase in QD densities is followed [for by a small increase in the transparency concentration ] and saturation effects occurring which at much higher gain. Both of these effects are associated with an increased number of QDs, which need more carriers to be excited to a level necessary to compensate for optical losses, but, after reaching the material transparency condition, any carrier surplus is then producing gain more effectively. Transparency concentrations do not depend on QD-size uniformity, values are considerably increased for but the saturated more uniform (less FWHM) QD sizes because, for the same carrier concentrations, less scattered QD sizes are followed by values. Recombination narrower gain spectra and higher between pairs of excited QD levels needs somewhat more carriers to reach transparency, but their saturated gain values are distinctly higher. Therefore, higher carrier concentrations are not recommended in QD active regions of in-plane QD diode lasers not only because of distinct saturation effects but also because of possible emission of radiation of an unwanted wavelength or even a simultaneous two-wavelength lasing [12]. In QD VCSELs, however, this problem is unimportant, because strong wavelength selectivity of their resonators guarantees blocking of radiation of a wavelength distinctly different from the one for which the resonator has been designed. modes are influenced by a given Various radiation distribution, where in a difoptical gain distributions match ferent way because their intensity differently. In addition, spatial distribution of optical losses and its overlapping with the mode intensity are also of importance. Therefore, it is worthwhile to introduce an effective describing 3-D correlation between both the modal gain and the distributions as follows:

is the structure radius and where resonator length. B. Optimization of a VCSEL Resonator

stands for the

(8)

Oxide apertures may play in the structure under consideration two important roles: they may be, as electrical apertures, funneling current flows from both ring contacts into the central active region and they may be confining unwanted spreading of an optical field in a radial direction (optical apertures), the latter providing they are located close to the anti-node position of the optical standing wave. Usually, for example, in a VCSEL design proposed in [30], both oxide apertures are at the same time electrical and optical. Then the length of the standard VCSEL resonator, understood as the distance between both resonator mircavity. A rors, is equal to . We call such a resonator the cavity is possible when one of the apertures is shifted to the node position, so it is working as an electrical aperture only. To optimize the cavity, let us consider first an impact on a VCSEL lasing threshold of a localization [within the resonator] of the bottom oxide aperture. For the fundamental transverse mode, dependences of its RT effective modal and its RT threshold material gain on a disgain tance of the bottom oxide aperture from the bottom DBR are plotted in Fig. 3(a). As one can see, the threshold gain is nearly constant for the oxide aperture localized close to the anti-node m of the transverse mode standing position wave. On the other hand, a shift of the oxide aperture toward m is followed by a considerable the node position threshold increase. plot [Fig. 3(a)] is not symmetSurprisingly, the rical with respect to the anti-node standing-wave position nm and the highest modal gain is achieved for the bottom oxide aperture slightly shifted from its seemingly opnm. The explanatimized anti-node position to tion may be found in Fig. 3(b). As one can see, the wave intensity has a much higher amplitude at the active-region posinm tion for the bottom oxide aperture shifted from nm (dashed curve) than for this aperture just to

634

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 40, NO. 6, JUNE 2004

Influence of the localization of the bottom oxide aperture within the resonator [cf. Fig. 4(c)]. (a) RT effective modal gain G of the fundamental LP mode and the threshold material gain g versus the distance z of the bottom oxide aperture from the bottom DBR plotted for the cavity length equal to 3=2 and the active-region radius r = 4 m. The internal cavity losses ( ) are temporarily neglected. (b) Optical-intensity standing wave (shown on the refractive-index profile) for two localizations of the bottom oxide aperture symmetrically placed on both sides of the anti-node = 184 nm): the dashed lines correspond to standing-wave position (z z = 164 nm, whereas the solid ones correspond to z = 204 nm.
3

Fig. 3.

=2(p)

Fig. 2. Maximal RT values g of an optical gain in QD (InGa)AsGaAs versus the active-region carrier active regions with indicated QD densities  concentration n determined for FWHM of (a) 37 meV, (b) 30 meV, and (c) 20 meV. Solid and dashed lines are plotted for recombination of carriers captured at fundamental QD levels ( = 1303 nm) and at excited QD levels ( = 1223 nm), respectively.

shifted symmetrically toward the opposite direction, i.e., from nm to nm (solid curve). The above effect follows from different DBR resonator mirrors. Therefore, it may be concluded that there are two cavity optimization conditions influencing low-threshold performance of the laser: the oxide position should be chosen in order to introduce an efficient

radial confinement mechanism for an optical field (so it should be close to the standing-wave anti-node position) and, at the same time, to maximize an intensity amplitude at the active-region position, so the really optimal position of the oxide aperture should be chosen as a compromise between both of the above conditions. The standard structure of the QD VCSEL reported in [30] is resonator [Fig. 4(a)] with two oxide aperequipped with the tures positioned exactly at the anti-node positions of the radiation standing wave (Fig. 5), so they are working as both optical and electrical apertures. However, our simulation revealed that an optical field is quite effectively confined within the GaAsbased OC VCSEL resonator even with only one oxide aperture, e.g., the upper one, located at the anti-node position. Then the bottom aperture may be shifted to the node position (where it is working as the electrical aperture only) without a significant increase in lasing threshold. As a result, instead of the standard

SARZAA: MODELING OF THE THRESHOLD OPERATION OF 1.3- m GaAs-BASED OC QD VCSELs

635

Fig. 4. Possible resonator configurations under consideration (a) Traditional 3 resonator (b) The 3=2(n) resonator with thinner bottom (n-type) spacer. (c) The 3=2(p) resonator with thinner upper (p-type) spacer. TABLE III

COMPARISON

OF THE 3=2(p) AND THE 3 RT WITH 1.5 V

VCSELS BIASED

AT

resonator, we may consider QD 1.3- m GaAs-based OC VCresonator [Fig. 4(b)]. SELs the design of the It is interesting to consider another modification of the VCSEL resonator [Fig. 4(c)] with the upper electrical aperture and the bottom one working as both optical and electrical ones. There is one essential disadvantage of this solution connected with Joule heating within the upper p-type GaAs current-spreading layer. Its thickness here is considerably reduced, which is followed by increasing the Joule-heat generation during a radial current flow between the upper annular contact and the central active region. Fortunately, GaAs-based OC QD VCSELs exhibit such low RT lasing thresholds that this effect turns out to be nearly negligible. However, decreasing the thickness of the p-type layer also has a profitable impact on lasing thresholds because of decreasing optical absorption (Table II). It happens to be much higher in p-type layers than in n-type ones, so, reducing thicknesses of p-type layers, we are and are reducing lasing thresholds [cf. (1)] because both decreased. Layer thicknesses as well as RT values of refractive indices and optical absorption coefficients are listed in Table II. Table III enables comparison of some operation parameters of biased at RT with 1.5-V oxide-confined GaAs-based 1.3- m resonator or the QD VCSELs equipped with the

Fig. 5. Standing wave of the LP mode in the 3 VCSEL resonator. The profile of an index of refraction is also shown.

resonator. An operation current of the latter one is equal to 1.22 mA in agreement with the value reported in [30]. An analoVCSEL (0.85 mA) is noticeable lower. gous current of the It is followed by decreasing maximal active-region temperadespite increasing Joule-heat generation ture increase within the upper p-type layer. The pn junction voltage drop is practically unchanged. Lower operation current in the VCSEL is followed by distinctly lower active-region current densities, changing from to (see Table III), but, surprisingly, by only somewhat lower active-region concentrato , which is mostly a consequence tions, varied from of an intense Auger recombination.

636

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 40, NO. 6, JUNE 2004

Fig. 6. RT effective modal gain G [see (8)] of the fundamental LP mode versus a number of stacked QD layers, each wth a density of 5110 cm , in the 8-m-diameter VCSEL active region plotted for various injection efficiencies  and the 3 resonator.

To enhance radial current spreading between the top annular contact and the central active region (which means to obtain more uniform current injection into the active region), the p-type spacer should be relatively highly doped. However, this is followed by relatively high intervalence absorption and free-carrier absorption within the p-type layers. While for cm reported theoretical values of the intervalence absorption coefficient are exceeding 100 cm [49], experimental ones are equal to about 80 cm [50]. For extremely highly doped p-type InP layers, this value may be as high as even 1000 cm [51]. On the contrary, much lower p-type doping cm (which means optical losses of about 20 of only cm ) was reported in [30]. Therefore, in our calculation, excm and treme two values of the possible p-type doping ( cm ) are considered, and the corresponding two values cm and cm ) have been chosen for the ( total absorption coefficient in the p-type layers. Their impact on lasing threshold will be discussed in Section V-C. C. Optimization of an Active Region of stacked QD layers in a VCSEL active reThe number gion is another important construction parameter of the 1.3- m GaAs-based OC QD VCSEL. As was stated earlier, a stack of QD layers instead of a single one is used to increase an effective [c.f. (8)] of QD density. In Fig. 6, the effective modal gain the mode versus a number of the stacked QD layers, each cm , in the 8- m-diamwith an assumed density of eter VCSEL active region, is plotted for various injection efand the same supply current, enabling reaching ficiencies an RT lasing threshold for . is defined as the ratio of a carrier concentration injected into QDs of a given layer and an analogous concentration injected into QDs of a previous value for equal to layer. All curves exhibit a maximal 4 6. The higher the injection efficiency , the higher ensures the best VCSEL performance. Therefore, it is clearly seen in the plot that the stack of only three QD layers is definitely not optimal for the VCSEL devices under consideration.

Fig. 7. QD density  , necessary to reach the RT LP -mode lasing ) in the 8-m-diameter thresholds (of given estimated values of g 3-QD-layer VCSEL active region, plotted versus FWHM for the 3 and 3=2 resonators and for two values of the absorption coefficient in the p-type spacer. The dashed line corresponds to the 6-QD-layer active region. 3=2(n) and 3=2(p) indicate resonators with thinner bottom (n-type) and upper (p-type) spacers, respectively.

Much better results should be achieved for equal to 56. A slight gain decrease observed for perfect carrier injection, i.e., and , is a result of a gradual carfor rier-concentration reduction with an increase in a total active-re, where is the averaged QD gion thickness , this decrease is more pronounced because size. For lower of a decreasing carrier concentration in successive QD layers. is followed by the Additionally, an excessive increase in gradually reduced efficiency of coupling of the radiation field and recombination carriers. Then the stack of QD layers should be divided into segments located close to successive anti-node positions of the radiation standing wave. In Fig. 7, the impact of FWHM on RT lasing thresholds in the 8- m-diameter VCSEL is examined. Two values of the absorp) tion coefficient within the p-type spacer and two ( and resonators are considered. In addition, an increase in the number of the QD layers is analyzed. As one can see, VCSELs with shorter resonators exhibit distinctly lower lasing thresholds and much lower QD densities may enable their lasing. An increase (from 3 to 6) is followed by a distinct reduction in in necessary to reach threshold. In addition, VCSELs with a thinner upper (p-type) spacer exhibit distinctly lower lasing VCSELs with a thinner bottom (n-type) thresholds than one. Besides, doping of the p-type spacer layers has been found to be crucial for farther threshold reductions. Its decrease is followed by more nonuniform current injection into the active region which, however, is less critical for the lasing threshold of mode than a corresponding decrease in the fundamental

SARZAA: MODELING OF THE THRESHOLD OPERATION OF 1.3- m GaAs-BASED OC QD VCSELs

637

Fig. 8. RT threshold active-region current-density radial profiles determined for the 3 resonator and the 3-QD-layer VCSEL active regions of given radii r .

Fig. 9. RT effective modal gain G of the lowest threshold LP modes excited in the 3-QD-layer VCSEL active regions of given radius r versus a thickness d of the Al O oxide layer. The nominal 3 resonator is assumed.

the optical absorption within the p-type spacer. Therefore, lower cm [30] has been confirmed p-type doping of about (Fig. 7) to be more optimal than its higher values. In Fig. 8, radial profiles of the RT active-region threshold curare plotted. As rent densities for various active-region radii one can see, the threshold current density is very high for small active regions, but its radial distribution within the active region , is is quite uniform. In addition, its edge value, i.e., surprisingly high which means that passive regions surrounding the active region are also to some extent supplied with this current. With an increase in the active-region radius, current-density distributions are becoming more and more nonuniform: the current of steadily reduced density is reaching the broad active-region central part and the density maximum comes nearer the active-region edge. D. Optimization of the Oxide Aperture of the Al O oxide aperture An impact of the thickness of the lowest threshold on the effective modal gain transverse modes is illustrated in Fig. 9. Generally, the effective modal gains are higher for thicker oxide layer. However, this innm. crease is insignificant for larger active regions and Quite a different situation is the case of relatively small active regions, for which thicker oxide layers enable receiving definitely higher effective modal gains, which is followed by lower lasing thresholds. This effect is associated with diffraction losses higher for both smaller and thinner oxide apertures. E. Mode Selectivity of various Fig. 10 presents the effective modal gains transverse modes as a function of the active-region ravalue, all effective modal gains are deterdius . For each mined for the current slightly higher than the threshold current value. As one can see, of the lowest threshold mode for this is unavoidably folan increase in the active-region radius lowed by a decrease in the effective modal gain values of lower modes and an increase in analogous gain values of order higher order ones. For m, single fundamental-mode
Fig. 10. RT values of effective modal gains G determined for various LP transverse modes for the currents slightly higher than the threshold current of the lowest threshold mode as a function of the radius r of the 3-QD-layer VCSEL active region and the nominal 3 resonator.

RT operation is not possible any longer. It would require additional methods of enhancement of radial current spreading in both spacer layers and/or improving radial optical-field confinements. Besides, in larger active regions, multimode operation is unavoidable, because effective modal gains of adjacent transverse higher order modes are very similar. VI. CONCLUSION In this paper, the self-consistent threshold model of the oxideconfined quantum-dot GaAs-based 1300-nm (InGa)AsGaAs VCSELs is presented. The model enables a detailed physical analysis of the operation of the considered lasing devices to better understand their performance, to simulate their operation characteristics and finally to optimize their known structures as well as to design their completely new structures corresponding to special features of 1.3- m OC QD GaAs-based diode lasers. On the basis of our simulations, some guidelines for manufacturing low-threshold devices have been proposed. To reduce

638

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 40, NO. 6, JUNE 2004

VCSEL lasing thresholds, two oxide apertures should be created within its cavity. However, the standard VCSEL equipped resonator with both apertures located at anti-node with the positions of an optical standing wave have been found to exhibit distinctly higher RT lasing threshold than properly designed resonator. In this new design, the upper VCSEL with the oxide aperture is shifted to the node position, so it is working as electrical aperture only. The optimal position of the bottom aperture is also somewhat shifted toward the bottom DBR to increase field intensity within the active region. Besides, the best anticipated CW threshold characteristics have been obtained for m m -diameter active regions, the six-layer-QD stacks, cm p-type spacers, and thicker relatively low doped oxide layers. Single fundamental-mode operation is possible for active-region diameters lower than 12 m. The model presented in this paper has been developed for the RT threshold operation of QD lasers. For higher excitation levels, it should be supplementedseveral new physical phenomena should be addressed. The FermiDirac carrier statistics assumed here should be replaced by more exact modeling of carrier energy distribution, e.g., by approaches suggested in [40] or [41]. Carrier injection into QDs should include their possible migration toward nonradiative centers and possible leakage effects. Stress-related effects around QDs should be taken into account [52]. Realistic QD shape and composition profile within its area, i.e., its authentic potential distribution, should be included in the gain spectra calculations. The hole-burning effect should also be considered in mode gain determination. Currently, however, many numerical values of material and device parameters necessary in such a modeling of QD devices are not available. Nevertheless, our slightly simplified simulation seems to remain quite exact for threshold analysis of QD lasers and should be useful in anticipation of their RT operation characteristics and optimization of their structures. ACKNOWLEDGMENT The author would like to thank M. Bugajski, W. Nakwaski, M. kowiak, and P. Mendla for their kind and fruitful Wasiak, P. Mac assistance. REFERENCES
[1] M. Grundmann, The present status of quantum dot lasers, Physica E, vol. 5, pp. 167184, Dec. 1999. [2] L. A. Coldren, Advances in vertical-cavity and widely-tunable lasers using InP-based PIC technology, IEEE LEOS Newsletter, vol. 17, pp. 1620, Feb. 2003. [3] D. G. Deppe, D. L. Huffaker, G. Park, and O. B. Shchekin, Quantum dots: a new generation of semiconductor lasers?, IEEE LEOS Newsletter, vol. 14, pp. 36, June 2000. [4] D. Bimberg, M. Grudmann, and N. N. Ledentsov, Quantum Dot Heterostructures. Chichester, U.K.: Wiley, 1999. [5] G. T. Liu, A. Stintz, H. Li, K. J. Malloy, and L. F. Lester, Extremely low room-temperature threshold current density diode lasers using InAs dot in In Ga As quantum-well, Electron. Lett., vol. 35, pp. 11631165, 1999. [6] O. B. Shchekin and D. G. Deppe, 1.3 m InAs quantum dot laser with T = 161 K from 0 to 80 C, Appl. Phys. Lett., vol. 80, pp. 32773279, 2002. [7] L. Zhang, R. Wang, Z. Zou, A. L. Gray, L. Olona, T. C. Newell, D. Webb, P. Varangis, and L. F. Lester, InAs quantum dot DFB lasers on GaAs for uncooled 1310 nm fiber communication, presented at the Optical Fibre Communication Conf., Atlanta, GA, Mar. 2527, 2003.

[8] O. B. Shchekin and D. G. Deppe, Low-threshold high-T 1.3-m InAs quantum-dot lasers due to p-type modulation doping of the active region, IEEE Photon. Technol. Lett., vol. 14, pp. 12311233, Sept. 2002. [9] I. P. Marko, A. D. Andreev, A. R. Adams, S. Krebs, J. P. Reithmaier, and A. Forchel, Importance of Auger recombination in InAs 1.3 m quantum dot lasers, Electron. Lett., vol. 39, pp. 5859, 2003. [10] A. Markus, A. Fiore, J. D. Ganiere, U. Oesterle, J. X. Chen, B. Deveaud, M. Ilegems, and H. Riechert, Comparison of radiative properties of InAs quantum dots and GaInNAs quantum wells emitting around 1.3 m, Appl. Phys. Lett., vol. 80, pp. 911913, 2002. [11] J. X. Chen, U. Oesterle, A. Fiore, R. P. Stanley, M. Ilegems, and T. Todaro, Matrix effects on the structural and optical properties of InAs quantum dots, Appl. Phys. Lett., vol. 79, pp. 36813683, 2001. [12] A. Markus, J. X. Chen, C. Paranthon, A. Fiore, C. Platz, and O. Gauthier-Lafaye, Simultaneous two-state lasing in quantum-dot lasers, Appl. Phys. Lett., vol. 82, pp. 18181820, 2003. [13] O. B. Shchekin and D. G. Deppe, The role of p-type doping and the density of states on the modulation response of quantum dot lasers, Appl. Phys. Lett., vol. 80, pp. 27582760, 2002. [14] D. G. Deppe, H. Huang, and O. B. Shchekin, Modulation characteristics of quantum-dot lasers: the influence of p-type doping and the electronic density of states on obtaining high speed, IEEE J. Quantum Electron., vol. 38, pp. 15871593, Dec. 2002. [15] P. Bhattacharya and S. Ghosh, Tunnel junction In Ga As/GaAs quantum dot lasers with 15 GHz modulation bandwidth at room temperature, Appl. Phys. Lett., vol. 80, pp. 34823484, 2002. [16] R. P. Sarzaa, M. Wasiak, T. Czyszanowski, M. Bugajski, and W. Nakwaski, Threshold simulation of 1.3-m oxide-confined in-plane quantum-dot (InGa)As/GaAs lasers, Opt. Quantum Electron., vol. 35, pp. 675692, 2003. [17] D. Bimberg, N. N. Ledentsov, M. Grundmann, F. Heinrichsdorff, V. M. Ustinov, P. S. Kopev, M. V. Maximov, Z. I.Zh. I. Alferov, and J. A. Lott, Application of self-organized quantum dots to edge emitting and vertical cavity lasers, Physica E, vol. 3, pp. 129136, 1998. [18] J. A. Lott, N. N. Ledentsov, V. M. Ustinov, N. A. Maleev, A. E. Zhukov, A. R. Kovsh, M. V. Maximov, B. V. Volovik, Z. I.Zh. I. Alferov, and D. Bimberg, InAs-InGaAs quantum dot VCSELs on GaAs substrates emitting at 1.3-m, Electron. Lett., vol. 36, pp. 13841385, 2000. [19] N. N. Ledentsov, M. Grundmann, F. Heinrichsdorff, D. Bimberg, V. M. Ustinov, A. E. Zhukov, M. V. Maximov, Z. I.Zh. I. Alferov, and J. A. Lott, Continuous-wave low-threshold performance of 1.3-m InGaAs-GaAs quantum-dot lasers, IEEE J. Select. Topics Quantum Electron., vol. 6, pp. 439451, May/June 2000. [20] V. M. Ustinov, A. E. Zhukov, N. A. Maleev, A. R. Kovsh, S. S. Mikhrin, B. V. Volvik, Y. G.Yu. G. Musikhin, Y. M.Yu. M. Shernyakov, M. V. Maximov, A. F. Tsatsulnikov, N. N. Ledentsov, Z. I.Zh. I. Alferov, J. A. Lott, and D. Bimberg, 1.3-m InAs/GaAs quantum dot lasers and VCSELs grown by molecular beam epitaxy, J. Cryst. Growth, vol. 227228, pp. 11551161, 2001. [21] D. Bimberg, M. Grudmann, F. Heinrichsdorff, N. N. Ledentsov, V. M. Ustinov, A. E. Zhukov, A. R. Kovsh, M. V. Maximov, Y. M. Shernyakov, B. V. Volovik, A. F. Tsatsulnikov, P. S. Kopiev, and Z. I.Zh. I. Alferov, Quantum dot lasers: breakthrough in optoelectronics, Thin Solid Films, vol. 367, pp. 235249, 2000. [22] A. Polimeni, M. Henini, A. Patan, L. Eaves, P. C. Main, and G. Hill, Optical properties and device applications of (InGa)As self-assembled quantum dots grown on (311) B GaAs substrates, Appl. Phys. Lett., vol. 73, pp. 14151417, 1998. [23] R. L. Sellin, Ch. Ribbat, M. Grundmann, N. N. Ledentsov, and D. Bimberg, Close-to-ideal device characteristics of high-power InGaAs/GaAs quantum dot lasers, Appl. Phys. Lett., vol. 78, pp. 12071209, 2001. [24] T. Suzuki, Y. Temko, and K. Jacobi, Shape of InAs quantum dots grown on the GaAs (113) B surface, Appl. Phys. Lett., vol. 80, pp. 47444746, 2002. [25] P. M. Smowton, E. J. Pearce, H. C. Schneider, W. W. Chow, and M. Hopkinson, Filamentation and linewidth enhancement factor in InGaAs quantum dot lasers, Appl. Phys. Lett., vol. 81, pp. 32513253, 2002. [26] P. Gilet, L. Grenouillet, P. Grosse, P. Ballet, P. Duvaut, N. Dunoyer, A. Million, V. Celibert, and C. B.C. Bru Chevallier, GaInAsN/GaAs multiple quantum wells and InAs/GaAs quantum dots: the effect of temperature on the light emission wavelength, in Proc. Int. Workshop GaAs Based Lasers for 1.31.5 m Wavelength Range, Wroclaw, Poland, Apr., 2426 2003. [27] E. C. Le Ru, P. Howe, T. S. Jones, and R. Murray, Strain engineered InAs/GaAs quantum dots for 1.5 m emitters, in Proc. Int. Workshop GaAs Based Lasers for 1.31.5 m Wavelength Range, Wroclaw, Poland, Apr., 2426 2003.

SARZAA: MODELING OF THE THRESHOLD OPERATION OF 1.3- m GaAs-BASED OC QD VCSELs

639

[28] J. He, X.-D. Wang, B. Xu, Zh.-G. Wang, and Sh.-Ch. Qu, Structure and photoluminescence study of InGaAs/GaAs quantum dots grown via cycled (InAs) =(GaAs) monolayer deposition, Jpn. J. Appl. Phys., pt. 1, vol. 42, pp. 11541157, 2003. ski and W. Nakwaski, Three-dimensional simulation of ver[29] M. Osin tical-cavity surface-emitting semiconductor lasers, in Vertical-Cavity Surface-Emitting Laser Devices. Berlin, Germany: Springer, 2003, ch. 5, pp. 135192. [30] N. Ledentsov, D. Bimberg, V. M. Ustinov, Zh. I. Alferov, and J. A. Lott, Quantum dots for VCSEL applications at 1.3 m, Physica E, vol. 13, pp. 871875, 2002. [31] R. P. Sarzaa and W. Nakwaski, Carrier diffusion inside active regions of gain-guided vertical-cavity surface-emitting lasers, Proc. Inst. Elect. Eng. Eng., pt. J, vol. 144, pp. 421426, 1997. [32] A. Fiore, 1300 nm quantum dot lasers: the role of intraband relaxation and carrier confinement, in Proc. Int. Workshop GaAs Based Lasers for 1.31.5 m Wavelength Range, Wroclaw, Poland, Apr., 2426 2003. [33] S. Ghosh, P. Bhattacharya, E. Stoner, J. Singh, H. Jiang, S. Nutticnck, and J. Laskar, Temperature-dependent measurement of Auger recombination in self-organized In Ga As=GaAs quantum dots, Appl. Phys. Lett., vol. 79, pp. 722724, 2001. [34] C. Walther, J. Bollmann, H. Kissel, H. Kirmse, W. Neumann, and T. Masselink, Characterization of electron trap states due to InAs quantum dots in GaAs, Appl. Phys. Lett., vol. 76, pp. 29162918, 2000. [35] J. Bloch, J. Shah, L. N. Pfeiffer, K. W. West, and S. N. G. Chu, Optical properties of multiple layers of self-organized InAs quantum dots, Appl. Phys. Lett., vol. 77, pp. 24452547, 2000. [36] W. Nakwaski, Thermal conductivity of binary, ternary, and quaternary III-V compounds, J. Appl. Phys., vol. 64, pp. 159166, 1988. [37] H. Wenzel and H.-J. Wnsche, The effective frequency method in the analysis of vertical-cavity surface-emitting lasers, IEEE J. Quantum Electron., vol. 33, pp. 11561162, July 1997. [38] M. Grundmann and D. Bimberg, Theory of random population for quantum dots, Phys. Rev. B, vol. 55, pp. 97409745, 1997. [39] D. G. Deppe, D. L. Huffaker, S. Csutak, Z. Zou, G. Park, and O. B. Shchekin, Spontaneous emission and threshold characteristics of 1.3-m InGaAs-GaAs quantum-dot GaAs-based lasers, IEEE J. Quantum Electron., vol. 35, pp. 12381246, Aug. 1999. [40] M. Wasiak, M. Bugajski, E. Machowska-Podsiado, T. Ochalski, J. kowiak, T. Czyszanowski, W. Nakwaski, J. Katcki, R. P. Sarzaa, P. Mac X. Chen, U. Oesterle, A. Fiore, and M. Ilegems, Optical gain saturation effects in InAs/GaAs self-assembled quantum dots, Optica Applicata, vol. 22, pp. 291299, 2002. kowiak, T. Czyszanowski, [41] M. Wasiak, M. Bugajski, R. P. Sarzaa, P. Mac and W. Nakwaski, Output power saturation in InGaN/GaAs quantum dot lasers, Phys. Stat. Sol. C, vol. 0, pp. 13511354, 2003. ski, Thermal analysis of GaAs-AlGaAs [42] W. Nakwaski and M. Osin etched-well surface-emitting double-heterostructure lasers with dielectric mirrors, IEEE J. Quantum Electron., vol. 29, pp. 19811995, June 1993. [43] R. J. Deria and M. A. Emanuel, Consistent formula for the refractive index of Al Ga1 xAs below band edge, Appl. Phys. Lett., vol. 77, pp. 46674672, 1995.

[44] F. A. Kish, S. J. Caracci, N. Holonyak Jr., J. M. Dallesasse, K. C. Hsieh, M. J. Flies, S. C. Smith, and F. D. Burnham, Planar native-oxide As index-guided Al Ga GaAs quantum well heterostructure lasers, Appl. Phys. Lett., vol. 59, pp. 17551757, 1991. [45] C. Tanguy, Temperature dependence of the refractive index of direct band gap semiconductors near the absorption threshold: application to GaAs, J. Appl. Phys., vol. 80, pp. 46264631, 1996. [46] H. G. Grimmeiss and B. Monemar, Temperature dependence of the refractive index of AlAs and AlP, Phys. Stat. Sol. A, vol. 5, pp. 109114, 1971. [47] M. O. Aboelfotoh, M. A. Borak, and J. Narayan, Ohmic contact to p-type GaAs using Cu Ge, Appl. Phys. Lett., vol. 75, pp. 39533955. [48] U. J. Ueng, N.-P. Chen, D. B. Janes, K. J. Webb, D. T. McInturff, and M. R. Melloch, Temperature-dependent behavior of low-temperaturegrown GaAs nonalloyed ohmic contacts, J. Appl. Phys., vol. 90, pp. 56375641, 2001. [49] S. Kakimoto and H. Watanabe, Intervalence band absorption in InP, J. Appl. Phys., vol. 85, pp. 18221824, 1999. [50] H. C. Casey Jr. and P. L. Carter, Variation of intervalence band absorption with hole concentration in p-type InP, Appl. Phys. Lett., vol. 44, pp. 8283, 1984. [51] J. Piprek, D. I. Babic, and J. E. Bowers, Numerical analysis of 1.54 m double-fused vertical-cavity lasers operating continuous-wave up to 33 C, Appl. Phys. Lett., vol. 68, pp. 26302632, 1996. [52] A. Fiore, U. Oesterle, R. P. Stanley, R. Houdr, F. Lelarge, M. Ilegems, P. Borri, W. Langbein, D. Birkedal, J. M. Hvam, M. Cantoni, and F. Bobard, Structural and electrooptical characteristics of quantum dots emitting at 1.3 m on gallium arsenide, IEEE J. Quantum Electron., vol. 37, pp. 10501058, Aug. 2001.

Robert P. Sarzaa received the M.Sc. degree in technical physics from the Technical University of Lodz, Lodz, Poland, in 1989 and the Ph.D. degree in electrical engineering from the Institute of Electron Technology, Warsaw, Poland, in 1998. In 1990, he joined the Laboratory of Computer Physics, Institute of Physics, Technical University of Lodz, and the teaching staff of the Faculty of Technical Physics, Computer Science and Applied Mathematics, Technical University of Lodz. His research interests include self-consistent computer modeling of physical phenomena crucial for operation of diode lasers. In particular, he is interested in simulation of optical, electrical, thermal, and mechanical phenomena taking place during operation of in-plane diode lasers and vertical-cavity surface-emitting diode lasers as well as of diode-laser arrays. Recently, he has been involved in an intense investigation of laser devices which may be used in modern optical-fiber communication systems taking advantage of successive optical windows (0.85 m, 1.3 m, and 1.55 m). In particular, he is working on modeling of the operation of GaAs-based diode lasers with both the (GaIn)(NA5)GaAs quantum-well and the InAsGaAs quantum-dot active regions. He is the author or coauthor of almost 80 scientific papers and conference contributions and two patents, all devoted to physics of semiconductor lasers.

Anda mungkin juga menyukai