Open Access
Add to BibSonomyAbstract
We report on the design, fabrication and analysis of vertical-cavity surface-emitting transistor-lasers (T-VCSELs) based on the homogeneous integration of an InGaAs/GaAs VCSEL and an AlGaAs/GaAs pnp-heterojunction bipolar transistor (HBT). Epitaxial regrowth confinement, modulation doping, intracavity contacting and non-conducting mirrors are used to ensure a low-loss structure, and a variety of design variations are investigated for a proper internal biasing and current injection to ensure a wide operating range. Optimized devices show mW-range output power, mA-range base threshold current and high-temperature operation to at least 60°C with the transistor in its active mode of operation for base currents well beyond threshold. Current confinement schemes based on pnp-blocking layers or a buried tunnel junction are investigated as well as asymmetric current injection for reduced extrinsic resistances.
© 2015 Optical Society of America
1. Introduction
By combining the functionality of a high-speed transistor and a diode laser, the transistor laser (T-laser) represents a new device paradigm with potential major implications in terms of high-capacity data processing and communication [1]. Different from a conventional diode laser, the T-laser relies on minority carrier recombination in the quantum-well (QW) region which modifies the carrier dynamics. In particular, the zero-charge boundary condition at the base-collector junction imposes a carrier concentration gradient across the base/cavity region that leads to a reduced carrier lifetime. This is of significance, since a reduced carrier lifetime is one mechanism for increasing the direct-modulation bandwidth of the device [2–6]. In addition, it allows for resonance-free frequency response (and thereby suppression of the relaxation oscillations and reduced turn-on delay) [7] and reduction of the relative intensity noise (RIN) [8]. Other potential advantages attributed to the transistor laser include the option for collector-current feedback control allowing for simplified power monitoring [9], enhanced modulation bandwidth [10] or improved linearity [11], as well as novel functionalities by virtue of the three-terminal configuration [12–14].
While most of these T-laser advantages so far have been discussed or demonstrated with respect to edge-emitting laser structures, the present technology of choice for many applications, especially related to short-reach data communication, is the vertical-cavity surface-emitting laser (VCSEL). This is primarily motivated by a quest for cost- and power-efficient emitters suitable for high-speed direct modulation. Driven by emerging applications and increasing demands for high-bandwidth interconnects, there have been substantial efforts related to the development of high-speed VCSELs during recent years. By a systematic optimization of the device structure to increase the intrinsic bandwidth and overcome limitations imposed by thermal effects and electrical parasitics, the maximum bandwidth for a directly modulated 850-nm VCSEL has gradually been extended to a present record value (at any wavelength) of 28 GHz [15]. These VCSELs were also used to demonstrate error-free data transmission up to 57 Gbit/s [15] and even higher data rates up to 71 Gbit/s was achieved using electronic equalization [16]. While these record-values have been laboratory-scale demonstrations, several vendors are presently in the process of introducing VCSEL products capable of 25 Gbit/s data transmission [17]. However, the requirements on bandwidths are steadily increasing and present predictions estimate that single-channel data rates as high as 100 Gbit/s will be required around year 2020 [18]. To reach such high and even higher modulation rates over an extended temperature range, with sufficient output power and with stringent requirements on power-efficiency (measured in fJ/bit) using directly modulated VCSELs, can be anticipated to be an extremely challenging task that calls for radically new design concepts.
To further increase the modulation bandwidth and data rate of VCSELs, alternative concepts have been suggested in the literature, including external modulation [19], self-injection-locked lasers [20], transverse-mode coupled-cavity VCSELs [21], minimized modal volume using high-index-contrast grating mirrors [22], and/or new modulation formats [23], but these schemes are difficult to implement in real-world systems, and direct modulation remains attractive to preserve a simplified system interface. During recent years it has also been suggested that transistor-VCSELs (T-VCSELs) may have significantly increased bandwidth as compared to conventional diode-VCSELs [24].
Transistor VCSELs are presently at an early stage of development but from revised and optimized designs, the same performance benefits, e.g. in terms of bandwidth, noise and linearity, as predicted for edge-emitting transistor lasers may be expected. Faraji and associates derived an analytical model for the small signal modulation of a transistor laser and estimated the 3dB-bandwidth for a typical T-VCSEL in the common-base configuration to 48 GHz, or greatly in excess of what would be expected from a standard diode-type but otherwise similar VCSEL structure [25]. The major difficulties in the design of a T-VCSEL as compared to an edge-emitting transistor laser lies in the three-dimensional character of the device structure which puts higher demands on the design optimization to control spatial potential variations within the device but also making the contacting of the different terminals more demanding. However, numerical simulations have shown that these are practical rather than fundamental problems and that T-VCSELs indeed should have a great potential for high-frequency operation that may break the bandwidth bottleneck of conventional laser diodes [26].
The first experimental demonstration of T-VCSELs were reported 2012 by Wu, Feng and Holonyak [27]. These devices were oxidation-confined npn-type 980-nm T-VCSELs with InGaP emitter, InGaAs/GaAs multiple-quantum-well (MQW) active region and AlGaAs/GaAs p- and n-type conducting distributed Bragg reflectors (DBRs), respectively and typical transistor laser features such as gain compression due to onset of stimulated emission and voltage-controlled operation could be demonstrated [28]. In subsequent work, they optimized the device structure for improved power efficiency using asymmetric contact layout and a hybrid dielectric/semiconductor top DBR [24,29]. However, these devices were only designed for low-temperature operation (−75°C) and showed a modest output power below 100 µW [24].
In the present paper, we review our development of pnp-type AlGaAs/GaAs/InGaAs T-VCSELs for emission around 980 nm. A low-loss cavity structure is realized using modulation doping, triple-intracavity contacting and non-conducting DBRs, and current confinement is due to epitaxial regrowth, either in the form of pnp-type blocking layers or a buried tunnel-junction (BTJ) structure, in a symmetric or asymmetric electrical contact configuration. Continuous-wave operation with an emission power above 1 mW is demonstrated for operating temperatures up to 60°C. It is demonstrated how the device can be optimized to maintain the transistor in its active mode of operation well beyond threshold and the static performance is analysed in some detail, including the breakdown mechanism in the limit of high base current and/or high collector-emitter voltage. Some of these results have been published elsewhere [30].
2. T-VCSEL design
The overall design and fabrication sequence of the T-VCSELs is based on our previous fabrication of 1.3-µm InGaAs/GaAs VCSELs [31]. As illustrated in Fig. 1, three different device designs have been considered. They are all based on epitaxial regrowth current confinement, using npn-blocking layer with symmetric (Fig. 1(a); “Design A”) or asymmetric (Fig. 1(b); “Design B”) current injection or a BTJ injection (Fig. 1(c); “Design C”). Top-view optical micrographs of the fabricated symmetric and asymmetric devices are shown in Fig. 2, while the detailed epitaxial layer configurations are given in Tables 1 and 2. This includes three different design variations of the base with varying thickness of the doped layer below the MQW region of 10, 100 and 200 nm, below referred to as 10-, 100- and 200-nm designs. The BTJ-T-VCSEL (Design C) only has the 200-nm design. The modulation doping as well as positioning of the active layer have been adjusted with respect to the standing wave of the electromagnetic field. The rationale behind design B is to reduce the extrinsic collector resistance. The distance from the collector contact to the centre of the device is in this case reduced from 28 to 11 µm.
Fig. 1 Schematic illustration of the three basic T-VCSEL designs: (a) Symmetric and (b) asymmetric designs with pnp blocking layer confinement; (c) Buried-tunnel junction blocking layer confinement. Blue and red colors indicate n- and p-doping, respectively while undoped or non-conducting regions are indicated using grey scale. Letters C, B and E, indicated the collector, base and emitter contacts (yellow color), respectively.
Fig. 2 Optical micrographs of the (a) symmetric and (b) asymmetric device designs. The sidewalls of the etched square-shaped mesas are oriented along the crystallographic <011>-type directions.
Table 1. Epitaxial layer structure for the pnp-blocking layer designs (Design A and Design B)
Table 2. Epitaxial layer structure for the BTJ current confinement design (Design C)
3. T-VCSEL fabrication
The epitaxial base structures according to Fig. 1 and Tables 1 and 2 were grown by metal-organic vapor-phase epitaxy (MOVPE) on Si-doped GaAs (001)-oriented substrates using standard precursors (AsH3, TMGa, TMAl and TMIn for the bulk material, AsH3, TMGa and TMIn for the QWs and BTJs, and SiH4, DEZn and CBr4 as dopant sources) in an Aixtron 200/4 low-pressure horizontal reactor. The process pressure was 100 mbar with H2 as carrier gas, the growth temperature was 680°C (calibrated using an aluminum-silicon eutectic) and the growth rate was approximately 4.5 µm/h. A reduced growth rate of approximately 0.1 nm/s and a reduced temperature of 510°C were used for the growth of the InGaAs/GaAs QWs and BTJ structures.
The basic fabrication sequence of the T-VCSELs proceeds as follows: The T-VCSEL base structure including the undoped bottom DBR and part of the cavity is grown by MOVPE. This includes the collector, base (including the MQW active layer) and most of the emitter regions according to Table 1. Square-shaped mesas oriented along the <100> directions and 200 nm in height are defined in the topmost GaAs layer using a two-step lithographic process.
The latter is of importance to realize sharp 90° corners and thereby a well-controlled regrowth process [32]. The 120-nm thick n-type blocking layer according to Table 1 is then regrown with the mesas still covered by a SiO2 mask. Directly after this first regrowth step, the SiO2 mask is removed and the cavity is completed by the second regrowth step. Larger mesas are then etched outside the active region to allow the contacting of the collector, base and emitter area according to Fig. 1(a), and electrical contacts are evaporated on the n- and p-type areas, respectively. A three-period α-Si (refractive index 3.6)/SiO2 (refractive index 1.45) DBR is deposited by plasma-enhanced chemical vapor deposition, and finally via holes are etched through the top DBR and probe metal pads are electroplated on the structure. The fabrication of the BTJ T-VCSEL in Fig. 1(c) proceeds in a similar fashion except that there is only one regrowth step (Table 2) and the emitter has n-type conductivity. A detailed account discussion of the processing steps as applied to a similarly-designed InGaAs/GaAs diode-VCSEL can be found in [32].
4. Results and discussion
4.1 Base region design
Due to the three-port configuration, a T-VCSEL requires a careful design-optimization for a correct internal biasing when the terminal voltages are varied. As demonstrated using numerical simulations by Shi et al. [33], the lateral voltage drop along the base and collector may result in a collector-base junction with different states at different positions, resulting in that the transistor is in different operational regimes (active or saturation) as function of the position within the device. This is an undesirable situation since the potential performance advantages of a T-laser relies on the transistor being in its active regime of operation but also that the collector current resembles the output power for feedback operation or power monitoring. In addition, Liu et al pointed out the importance of a reduced base resistance for high-frequency operation and therefor suggested an asymmetric device layout to reduce the base contact-to-active region distance [29].
In our previous work on T-VCSELs we noted a premature saturation prior to threshold when increasing the base current of the device [34], which could be directly linked to internal potential variations [35]. While the device locally operated in the active mode in the center region, it was already in the saturation regime as reflected by the measured current-voltage characteristics. There are essentially two ways to improve the situation; either to reduce the base and/or collector resistances or to decrease the associated currents to reduce the voltage drops. An extension of the base width should combine these effects, leading to a reduced baseresistance as well as a reduction of the collector current due to a reduced current gain, as demonstrated using numerical simulations [36]. To verify these predictions, light-current-voltage (LIV) characteristics were measured for 10-µm devices with the 10, 100 and 200-nmdesigns. The results are displayed in Fig. 3. Here, the collector current (IC), emitter-base voltage (VBE) and output power (Pout) are measured as function of base current (IB) for a specific collector-emitter voltage (VCE). A clear trend is observed for increasing base width regarding both IC and VBE where VBE as well as the current gain (β = ΔIC/ΔIB), and thereby IC, is decreasing with increasing base width, both of which is of significance for a delay in saturation current towards higher IB; a reduced increase in VBE when increasing IB extends theregion over which VBC = VCE-VBE>0, and a reduction in IC decreases the voltage drops in the emitter and the collector. While the base saturation current (IBsat) is below 1 mA and in case of the 10-nm design, it is extended to 4.5 mA for the 100-nm device and to 9 mA for the 200-nm device. Considering that the base threshold current (IBth) is around 2 mA in all cases, it is clear that the devices only are in the forward active mode of transistor operation during lasing in the latter two cases.
Fig. 3 Measured dependency of IC, VBE and Pout on IB for VCE = 3.5 V. The device designs are A-10, 100 and 200 nm with increasing base width as indicated by arrows for each curve set and the device size is 10 µm in all cases.
4.2 Cavity effects
To further examine the effect of the optical cavity on the T-VCSEL electrical characteristics we compare in Fig. 4 the electrical and optical characteristics of a 10-µm device according to design A-200-nm before and after etching off the topmost dielectric DBR structure. The removal of the topmost mirror leads to a reduction of the out-coupling mirror reflectance from >99% to approximately 30% (defined by the GaAs-air interface) and the mirror loss becomes far too high for the laser to reach threshold. Consequently, no lasing occurs in this case and there is only a linear increase in the emitted power with increasing base current due to spontaneous emission. The general features of the LIV characteristics of the T-VCSEL will be discussed below. Here we only note that a removal of the cavity and thereby the onset of lasing results in that the transistor runs into saturation significantly earlier than is the case of the full T-VCSEL structure. In the latter case, the carrier recombination becomes significant in the base at threshold due to the decreased recombination lifetime. This leads to a reduction in the base transport factor (αT = IC/IEh; IEh being the hole part of the emitter current) and thereby the current gain so that IC flattens out. This is also accompanied by a flattening out of VBE due to the increase in recombination current. However, one may also note that the collector current in saturation is higher and the corresponding base-emitter voltage is lower for the T-VCSEL as compared to its counterpart without topmost DBR. This is presumably related to photon-assisted carrier transport (free-carrier absorption) in the former case where the full cavity laser provides a cavity-enhanced spontaneous emission even beyond complete roll-over.
Fig. 4 Dependency of IC, VBE and Pout on IB for a 10-µm, A-200-nm device before and after removal of the top dielectric DBR. VCE = 3.5 V.
4.3 Details of the light-current-voltage characteristics
In Fig. 5 we investigate the dependency of IC, VBE and Pout on IB for different values of VCE and different device sizes between 4 and 10 µm. The different features in these LIV characteristics are now discussed with respect to the simplified current flow diagram in Fig. 6. Starting with the 10-µm device in Fig. 5(a) and VCE = 3.5 V we first note that several operational regimes as labelled with Roman numbers I-IV can be identified. In Region I (the sub-threshold region), IC is increasing superlinearly with IB. The differential current gain (β = dIC/dIB) is here limited by base recombination current which is mainly due to spontaneous recombination in the QWs (current component 2 in Fig. 6). When reaching Region II, at threshold, the recombination rate (and thereby dIB/dVBE) increases drastically due to the onset of stimulated emission. This is reflected by pronounced kinks and flattening of IC and VBE. As IB is further increased through Region II, IC is again increasing superlinearly with a sharp and breakdown-like character at the boundary to Region III which drives the transistor into saturation. The collector current is then decreasing throughout Regions III and IV until it eventually becomes negative, implying that the base-collector junction is fully forward biased and the device works like a double-injection diode. The slight kink in IC at the boundary between Regions III and IV correlates to the complete roll-over and total quenching of the optical output power, presumably indicating a photon-assisted current component that is active up to Region III.
Fig. 5 Dependency of IC, VBE and Pout on IB for A-200-nm devices of size (a) 10 µm, (b) 8 µm, (c) 6 µm and (d) 4 µm for different VCE as indicated. The dashed lines are included as guides to the eye. The regions I-IV applies to VCE = 3.5 V.
Fig. 6 Illustration of the current components in a T-VCSEL biased in the common-emitter configuration. Blue and red colors indicate electron and hole currents, respectively. The different contributions are: 1) Hole-injection over the forward-biased base-emitter junction (IEh); 2) Base recombination current (IBr); 3) Part of the emitter current that is swept into the collector (B•IEh); 4) Majority-carrier contributions due to tunneling over the base-collector junction; and 5) Minority carrier contributions due to electron and hole injection over the base-collector junction (only effective in the saturation regime).
The reduction in VCE below 3.5 V has quite pronounced effects on the LIV characteristics: The breakthrough in IC close to saturation is less pronounced and even disappears as VCE is reduced below 1.5 V; the signature in VBE get correspondingly less pronounced; and the roll-over of the optical output power is significantly delayed. One noteworthy observation is that the total quenching of the emission appears to occur at a specific value of VBE = VBEQ ≈3.5 V, but at increasing IB for decreasing VCE. This is indicated by dashed lines in Figs. 5(a)-5(d).
Devices with smaller aperture sizes show essentially the same pattern but besides the obvious reduction in threshold current, output power and electrical conductance, there are some distinct differences in that the breakdown in IC and VBE gets more pronounced with decreasing device dimension and that Region III becomes narrower; see Figs. 5(b)-5(d). A direct consequence of the latter effect is that the quenching of the light after breakdown in IC becomes more abrupt for smaller devices; for the 4-µm device it is virtually instantaneous. VBEQ remains approximately the same (≈3.5 V) also for smaller device sizes, although it may be noted that VBE not is well-defined for the 4-µm device at VCE = 3.5 V due to its rapid increase at breakdown.
Figures 7(a) and 7(b) shows electrical and optical collector diagrams (IC and Pout as function of VCE for different values of IB) for a 4-µm device. Similar to Fig. 5(d) the breakdown has a very abrupt character and there is a strong correlation between IC and Pout. It can also be noted that the breakdown in IC occurs at reduced VCE for increasing IB, which corresponds to the situation in Fig. 5 where the breakdown in the IC-vs-IB characteristics occurs at reduced IB for increasing VCE. An interesting observation regards the reduction in threshold current at high VCE≈5 V also manifested as an increase in Pout for constant IB less than approximately 1.5 mA. This is most clearly viewed in the magnified view in Fig. 7(c), and is believed to be due to an additional supply of electron to the base and active region via direct tunneling from the collector as schematically illustrated in Fig. 8(a). The power consumption at threshold, defined as P = IC·VCE + IB·VBE at threshold voltage VCE≈0.55 V in Fig. 7(b) for IB = IBth≈0.8 mA, is calculated to 1.95 mW using IC≈0.7 mA and VBE≈1.61 V which is significantly lower than previously published values for T-VCSELs [24].
Fig. 7 (a) Electrical and (b) optical collector diagrams for a 4-µm T-VCSEL according to design A-200 nm. The gain compression at threshold is clearly manifested as a narrowing of the collector current curves for increasing base current in equidistant steps. Red color indicates stimulated emission whereas black color indicates spontaneous emission. The reduction in threshold current around VCE = 5V corresponds to the direct-tunneling regime as schematically illustrated in Fig. 8(a). (c) A magnification of the optical collector diagram in (b) showing the evolution of the optical power for base currents close to threshold.
Fig. 8 Schematic band diagrams illustrating (a) the direct-tunneling event between base and collector at sufficiently large band-bending in the base-collector junction and b) the photon-assisted tunneling effect proposed in [27] to be responsible for the breakdown in collector current in an npn-type T-VCSEL.
Figure 9 shows the corresponding collector diagrams for a 10-µm device. Apart from higher threshold and output power the features from the smaller-sized device in Fig. 5 are essentially retained, including the abrupt character of the breakdown in IC, Pout and VBE. In this case, the power consumption at lasing threshold is approximately 5.15 mW (using VCE = 0.7 V, IC = 2.5 mA, IBth = 2.0 mA and VBE = 1.7 V). Here we have also included lines that indicate the condition VBC = VBE-VCE = 0, nominally corresponding to the transition between the active and saturation modes. For a clearer presentation we show IC-vs-VCE and Pout-vs- VCE characteristics for a single value of IB in Fig. 10. The shift between the kink in IC and the point VBC = 0 of around one volt reflects the difference in potential variations along the base and collector; while the measured value of IC reflects the condition in the interior of the device the terminal voltages VB and VC only defines the potential close to the metal contacts [35].
Fig. 9 Electrical and optical collector diagrams for a 10•10-µm2-device, including (a) IC, (b) Pout and (c) VBE as function of VCE for different values of IB as indicated. The dashed lines indicate VBC = 0 and hence the transition between saturation and active range of transistor operation as measured on the metal contacts.
Fig. 10 Extraction of IC, VBE and Pout as function of VCE from Fig. 9 for a single base current of 4.2 mA.
4.4 Temperature-dependent operation
Figure 11 shows Pout, VBE, β and IC, as function of IB for a 4·4-µm2 device for different stage temperatures of 20-60°C. Due to the small device aperture size, the breakdown is well defined with an virtually instant quenching of the optical power at all temperatures, and it is noted that the devices works continuous-wave with preserved transistor laser-like operational characteristics over the entire temperature range. The threshold current decreases slightly with increasing temperature, reflecting a negative gain-cavity offset at room temperature of approximately 40 nm (emission wavelength: 1005 nm, peak photoluminescence wavelength for QW reference sample: 965 nm) while the peak optical power is reduced from 0.45 to 0.15 mW over the temperature interval. The temperature-driven effect on breakdown current is quite significant, corresponding to a reduction from 5.4 mA at 20°C to 2.3 mA at 60°C. Larger-area devices of 10-µm deliver slightly more than 1 mW at room temperature (and up to 1.3 mW at 10°C) and still 0.1 mW at 80°C (not shown). However, transistor laser-like operational characteristics such as gain compression at threshold were only observed up to 70°C. The observation of a critical base-emitter voltage (VBEQ) for total quenching of the optical output power also holds over the temperature range but with decreasing value with increasing temperature. Figure 12 summarizes the observed trend in VBEQ as function of temperature.
Fig. 11 Temperature-dependent operation of a 4·4-µm2 T-VCSEL showing Pout, β VBE and IC, for VCE = 3.5 V and different temperatures as indicated.
Fig. 12 Base-emitter voltage corresponding to collector-current breakdown as function of temperature between 10 and 80°C. The data is acquired on different device sizes ranging from 4 to 10 µm.
4.5. Breakdown mechanism
A similar breakdown in IC and corresponding quenching of Pout in the limit of high IB and VCE as reported here was also observed for npn-type T-VCSELs by Wu et al. [28], who attributed it to a Franz-Keldysh-type photon absorption process in the reverse-biased base-collector junction that adds carriers to IC and also generates a resupply of majority carriersinto the base. This process is illustrated in Fig. 6 as current component number 5 and using a simplified band-diagram in Fig. 8(a). An important aspect of this proposed mechanism is that it is self-supporting in that the back-injection of electrons to the base will, due to charge neutrality reasons, trig the injection of yet another hole over the emitter-base junction. This mechanism has been extensively discussed in the case of avalanche breakdown and is usually referred to as the “hole feedback effect” [37].
The assumption of a Franz-Keldysh-governed breakdown mechanism for IC similar to what is proposed in [28] would require the combination of sufficiently high band-bending in the base-collector junction and a sufficiently high photon density. The photon density increases with increasing IB but is in turn also connected to an increasing VBE that decreasesrather than increases the band-bending. This can be argued to be consistent with the observed trends in Fig. 5 inasmuch as the run-off in IC is concerned; the breakdown occurs at reduced IB for increasing VCE, and in the limit of low IB and low VCE there is no significant increase in IC. However, it cannot explain the quenching of the output power. Any increase in IB beyond the breakdown (at IB,BD) will decrease the band-bending and would thereby reduce the absorption. The same thing holds in the optical collector diagram in Figs. 7 and 9. While an increasing VCE would be expected to lead to increased absorption through band-to-band absorption in the base-collector junction and thereby further reduction in the output power as observed, the dependency on IB is not consistent in this picture. For constant VCE (vertical displacement in the Pout-vs-VCE diagram), increasing IB leads to increasing absorption and more rapid quenching while the expectation rather would be the opposite based on the assumption of a reduced band bending.
Even more doubts on a Franz-Keldysh absorption process comes from the temperature-dependent measurements. As observed in Fig. 11, increasing temperature leads to increasing VBE and decreasing Pout, i.e. decreasing band-bending and decreasing photon density. Still, the breakdown occurs at strongly reduced IB with increasing temperature. For instance, at the point of breakdown at 60°C (IB = 2.3 mA), VBE = 2.64 V and Pout = 0.13 mW while the corresponding values at 20°C for VBE and Pout are 2.06 V and 0.18 mW, respectively. This is clearly inconsistent with the requirement on the combination of a high photon density and strong band-bending in the base-collector junction for the absorption process to be effective.
Instead, we suggest that the breakdown in IC rather is due to QW band filling, where the saturation of the available states as IE is increased will lead to increasing base transport factor that increases IC. Due to the saturation in recombination current this also increases VBE which thereby leads to a further increase in IE and IC until the device is forced into saturation. Furthermore, the observation of a critical base-emitter voltage (VBEQ) indicates that the quenching of the light presumably is related to the band structure and carrier distribution in the base region. It is likely that the increase in VBE will alter the supply of carriers to the active MQW active region. Injected minority holes will be subject to lateral diffusion and recombination remote from the active region, eventually leading to insufficient supply of electrons to maintain threshold carrier density. The instant but partial drop in intensity at the point of breakdown is presumably an effect of local heating due to the rapid increase in both IC and VBE. The decline in optical intensity beyond that point is strongly dependent on the device dimension, reflecting the higher series resistance and steeper increase in VBE in a smaller device; see Figs. 5(a)-5(d).
4.6 Asymmetric current injection
Fig. 13 Dependency of (a) VBE, (b) IC and (c) Pout on IB and different VCE as indicated for T-VCSELs with asymmetric (solid lines) and symmetric (dashed lines) current injection according to design B-200-nm and A-200-nm, respectively.
An important element of the T-VCSEL design regards the lateral current injection from the metal contacts at the peripheral region of the device. As evident from the present results the reduction of extrinsic emitter, base and collector resistances is of critical importance to the device performance. Transistor laser-like characteristics could e.g. only be observed after which the base width was significantly increased. However, such designs will come into conflicts with the requirements of high-speed operation, which e.g. will require short base and cavity regions to keep down the base transit time as well as modal volume. A differentapproach of reducing the lateral resistances and potential drops would be to modify the lateral device design to move the different contact closer to the central parts of the device. Such a design was already demonstrated by Lie et al. for the case of npn-type T-VCSELs where an asymmetric design was used to move the base contact significantly closer to the active region in order to reduce the extrinsic base resistance [29]. In the present work, we have evaluated an asymmetric device design with the objective of reducing the length of the collector channel; see Figs. 1 and 2. The motivation for this was that the collector resistance was identified as amajor source of performance limitation [35,36] but it is also clear that other similar design modifications can be considered for a very compact device layout and it is of interest to see how the overall device performance is affected by asymmetric current injection.
Fig. 14 Dependency of (a) VBE, (b) IC and (c) Pout on IB for a T-VCSEL with buried tunnel current injection according to design C-200-nm (solid lines) compared to the A-200-nm design based on pn-blocking layer confinement (dashed lines). The vertical dashed lines spanning over all Figs. indicate the threshold currents.
Figure 14 shows LIV characteristics for the asymmetric design (Fig. 1(b); B-200-nm) as compared to the symmetric design evaluated above (Fig. 1(a); A-200-nm). It should be noted that these are neighboring devices on the same wafer, so it is only the lateral device design which is different. The epitaxial layer structures are identical and the processing was done in parallel. Somewhat surprisingly, the asymmetric design leads to quite dramatic performance degradation. While the threshold currents seem to be identical between the two wafers, the asymmetric design requires a higher VBE, has a narrower region of active mode operation and provides less optical output power. A possible interpretation is that the asymmetric current injection leads to a less favorable gain-mode overlap as the bias is increased. This would then result in less output power but also reduced gain saturation and increased VBE due to the reduction in recombination current, in all indicating that any designs for altered current injection must be handled with great care.
4.7 Buried-tunnel junction current injection
Buried-tunnel junction (BTJ) current confinement are expected to have several advantages compared to the pn-blocking layer based one applied in the T-VCSELs discussed above; it only involves one regrowth step which makes the fabrication more straightforward; the emitter region is n- rather than p-doped which reduces the optical loss; and the electrical parasitics can be reduced. We have previously compared BTJ and pn-blocking layer current injection for 1.3-µm InGaAs/GaAs diode VCSELs with overall similar design features as the present T-VCSELs, and could demonstrated reduced series resistance and enhanced modulation bandwidth [38]. In Fig. 13 we show the measured LIV characteristics for the 10-µm BTJ-T-VCSEL, based on an n+p+-InGaAs/InGaAs tunnel junction. From a comparison with Fig. 5, of which the data is re-plotted as dashed lines in Fig. 13 , it is clear that the performance of this BTJ-T-VCSEL does not match that of the pn-blocking layer type in terms of optical output power or range of active mode operation.
The reason for the lower output power of the BTJ-T-VCSEL is presumably a less favorable alignment between the active region gain curve and the cavity resonance, the measured detuning between QW PL emission and cavity being −15 nm (λPL = 970 nm, λcav = 985 nm) as compared to −40 nm (λPL = 965 nm, λcav = 1005 nm) for the pn-blocking layer VCSELs, and it is conceivable that the self-heating during operation pushes the gain curve to the long-wavelength side of the cavity resonance. In addition, due to a mistake in the growth sequence there is mismatch between the electromagnetic standing wave and the modulation doped layers (layers 10 in Table 2), leading to enhanced absorption loss. The tunnel junction itself can also be optimized for higher conductance, see e.g [39], although it should be realized that the application of low-bandgap materials for effective tunnel junctions based on type-II heterojunctions will set a lower limit for the T-VCSEL emission wavelength in order to avoid band-to-band absorption of the fundamental laser light.
These devices also run into saturation at an earlier stage so that while the general T-laser-specific features are present for both device types, the active mode of operation is much narrower for the BTJ-T-VCSEL. The differential resistance is lower at high IB in the saturation regime as seen in Fig. 13(a) but it is higher in the sub-threshold and sub-saturation regimes, thereby displaying a more pronounced increase in VBE. As a combined consequence of a lower recombination current, leading to less gain saturation upon threshold, and higher VBE the device goes more rapidly into saturation.
5. Summary
The design optimization of T-VCSELs is an intricate business that needs to take into account a rich interplay between electric fields, carrier dynamics and optical field. It has here been demonstrated how epitaxially regrown pnp-type T-VCSELs can be optimized for an extended range of active mode operation, providing power efficiency, threshold current, output power and high temperature operation approaching those of conventional diode-type VCSELs. These results thereby define state-of-the-art for T-VCSELs. Special attention was also paid to the mechanism of collector current breakdown in the limit of high base current and/or collector-emitter voltage, and it was argued that this, in contrast to previously models for npn-type T-VCSELs based on Franz-Keldysh absorption in the base-collector junction [28], presumably rather is an effect of quantum-well band filling. Finally, alternative biasing schemes were examined for reduced emitter, base and collector resistances, including buried-tunnel junction injection and asymmetric contact layout. Somewhat unexpectedly, these devices showed degraded rather than improved performance as compared to the symmetric pn-blocking layer devices, indicating some inherent problems with these approaches and the need for further optimizations.
Acknowledgments
This work was supported by the Swedish Research Council under Grant 2010-4386.
References and links
1. H. W. Then, M. Feng, and N. Holonyak, “The transistor laser: Theory and Experiment,” Proc. IEEE 101(10), 2271–2298 (2013). [CrossRef]
2. H. W. Then, M. Feng, and N. Holonyak, “Physics of base charge dynamics in the three port transistor laser,” Appl. Phys. Lett. 96(11), 113509 (2010). [CrossRef]
3. B. Faraji, W. Shi, D. L. Pulfrey, and L. Chrostowski, “Analytical modeling of the transistor laser,” IEEE J. Sel. Top. Quantum Electron. 15(3), 594–603 (2009). [CrossRef]
4. T. Tan, R. Bambery, M. Feng, and N. Holonyak, “Transistor laser with simultaneous electrical and optical output at 20 and 40 Gb/s data rate modulation,” Appl. Phys. Lett. 99(6), 061105 (2011). [CrossRef]
5. M. Shirao, S. H. Lee, N. Nishiyama, and S. Arai, “Large-signal analysis of a transistor laser,” IEEE J. Quantum Electron. 47(3), 359–367 (2011). [CrossRef]
6. I. Taghavi, H. Kaatuzian, and J. P. Leburton, “Bandwidth enhancement and optical performances of multiple quantum well transistor lasers,” Appl. Phys. Lett. 100(23), 231114 (2012). [CrossRef]
7. M. Feng, H. W. Then, N. Holonyak, G. Walter, and A. James, “Resonance-free frequency response of a semiconductor laser,” Appl. Phys. Lett. 95(3), 033509 (2009). [CrossRef]
8. F. Tan, R. Bambery, M. Feng, and N. Holonyak, “Relative intensity noise of a quantum well transistor laser,” Appl. Phys. Lett. 101(15), 151118 (2012). [CrossRef]
9. E. W. Iverson and M. Feng, “Transistor laser power stabilization using direct collector current feedback control,” Photon. Technol. Lett. 24(1), 4–6 (2012). [CrossRef]
10. B. Faraji, N. A. F. Jaeger, and L. Chrostowski, “Modelling the effect of the feedback on the small signal modulation of the transistor laser,” 23rd annual meeting of the IEEE Photonics Society, p. WX4, Denver CO, Nov 7–11 (2010). [CrossRef]
11. H. W. Then, F. Tan, M. Feng, and N. Holonyak, “Transistor laser optical and electrical linearity enhancement with collector current feedback,” Appl. Phys. Lett. 100(22), 221104 (2012). [CrossRef]
12. W. Shi, L. Chrostowski, and B. Faraji, “Numerical study of the optical saturation and voltage control of a transistor vertical-cavity surface-emitting laser,” IEEE Photon. Technol. Lett. 20(24), 2141–2143 (2008). [CrossRef]
13. B. Faraji, W. Shi, D. L. Pulfrey, and L. Chrostowski, “Common-emitter and common-base small-signal operation of the transistor laser,” Appl. Phys. Lett. 93(14), 143503 (2008). [CrossRef]
14. H. W. Then, C. H. Wu, G. Walter, M. Feng, and N. Holonyak, “Electrical-optical signal mixing and multiplication (2 → 22 GHz) with a tunnel-junction transistor laser,” Appl. Phys. Lett. 94(10), 101114 (2009). [CrossRef]
15. P. Westbergh, R. Safaisini, E. Haglund, J. S. Gustavsson, A. Larsson, and A. Joel, “High-speed 850 nm VCSELs with 28 GHz modulation bandwidth for short reach communication,” Proc. SPIE 8639, 86390X (2013). [CrossRef]
16. D. M. Kuchta, A. V. Rylyakov, F. E. Doany, C. L. Schow, J. Proesel, C. Baks, P. Westbergh, J. S. Gustavsson, and A. Larsson, “A 71-Gb/s NRZ modulated 850-nm VCSEL-based optical link,” Photonics Technol. Lett. 27(6), 577–580 (2015). [CrossRef]
17. See, e.g., Proceedings of the Optical Fiber Communication Conference, Optical Society of America (2014).
18. A. Larsson, “Advances in VCSELs for communication and sensing,” IEEE J. Sel. Top. Quantum Electron. 17(6), 1552–1567 (2011). [CrossRef]
19. A. Paraskevopoulos, H.-J. Hensel, W.-D. Molzow, H. Klein, N. Grote, N. Ledentsov, V. Shchukin, C. Möller, A. Kovsh, D. Livshits, I. Krestnikov, S. Mikhrin, P. Matthijsse, and G. Kuyt, “Ultra-High-Bandwidth (>35 GHz) Electrooptically-Modulated VCSEL,” Optical Fiber Communication (OFC) conference 2006, PDP 22 (2006). [CrossRef]
20. L. Chrostowski, X. Zhao, C. J. Chang-Hasnain, R. Shau, M. Ortsiefer, and M. C. Amann, “50-GHz optically injection-locked 1.55-µm VCSELs,” IEEE Photon. Technol. Lett. 18(2), 367–369 (2006). [CrossRef]
21. H. Dalir and F. Koyma, “High-speed operation of bow-tie-shaped oxide aperture VCSELs with photon–photon resonance,” Appl. Phys. Express 7(2), 022102 (2014). [CrossRef]
22. I. S. Chung and J. Mørk, “Speed enhancement in VCSELs employing grating mirrors,” Proc. SPIE 8633, 863308 (2013). [CrossRef]
23. K. Szczerba, B.-E. Olsson, P. Westbergh, A. Rhodin, J. S. Gustavsson, A. Haglund, M. Karlsson, A. Larsson, and P. A. Andrekson, “37 Gbps transmission over 200 m of MMF using single cycle subcarrier modulation and a VCSEL with a 20 GHz modulation bandwidth,” Proc. 36th Eur. Conf. Opt. Commun., Torino, Italy, 2010. [CrossRef]
24. M. K. Wu, M. Liu, R. Bambery, M. Feng, and N. Holonyak, “Low power operation of a vertical cavity transistor laser via the reduction of collector offset voltage,” IEEE Photon. Technol. Lett. 26(10), 1003–1006 (2014). [CrossRef]
25. B. Faraji, W. Shi, D. L. Pulfrey, and L. Chrostowski, “Common-emitter and common-base small-signal operation of the transistor laser,” Appl. Phys. Lett. 93(14), 143503 (2008). [CrossRef]
26. W. Shi, B. Faraji, M. Greenberg, J. Berggren, Y. Xiang, M. Hammar, M. Lestrade, Z. Li, Z. M. Simon Li, and L. Chrostowski, “Design and modeling of a transistor vertical-cavity surface-emitting laser,” Opt. Quantum Electron. 42(11–13), 659–666 (2011). [CrossRef]
27. M. K. Wu, M. Feng, and N. Holonyak, “Surface emission vertical cavity transistor laser,” Photon. Technol. Lett. 24(15), 1346–1348 (2012). [CrossRef]
28. M. K. Wu, M. Feng, and N. Holonyak, “Voltage modulation of a vertical cavity transistor laser via intra-cavity photon-assisted tunnelling,” Appl. Phys. Lett. 101, 088102 (2012).
29. M. Liu, M. K. Wu, M. Feng, and N. Holonyak, “Lateral feeding design and selective oxidation process in vertical cavity transistor laser,” J. Appl. Phys. 114(16), 163104 (2013). [CrossRef]
30. Y. Xiang, C. Reuterskiöld-Hedlund, X. Yu, C. Yang, T. Zabel, M. Hammar, and M. N. Akram, “Performance Optimization of GaAs-based vertical-cavity surface-emitting transistor-lasers,” Photonics Technol. Lett. 27(7), 721–724 (2015). [CrossRef]
31. R. Marcks von Würtemberg, X. Yu, J. Berggren, and M. Hammar, “Performance optimization of epitaxially regrown 1.3-µm VCSELs,” IET Optoelectronics 3(2), 112 (2009). [CrossRef]
32. R. Marcks von Würtemberg, Z. Zhang, J. Berggren, and M. Hammar, “A novel electrical and optical confinement scheme for surface emitting optoelectronic devices,” Proc. SPIE 6350, 63500J(2006). [CrossRef]
33. W. Shi, L. Chrostowski, and B. Faraji, “Numerical study of the optical saturation and voltage control of a transistor vertical-cavity surface-emitting laser,” IEEE Photon. Technol. Lett. 20(24), 2141–2143 (2008). [CrossRef]
34. X. Yu, Y. Xiang, J. Berggren, T. Zabel, M. Hammar, and N. Akram, “Room-temperature operation of transistor vertical-cavity surface-emitting laser,” Electron. Lett. 49(3), 208–210 (2013). [CrossRef]
35. Y. Xiang, X. Yu, J. Berggren, T. Zabel, M. Hammar, and N. Akram, “Minority current distribution in InGaAs/GaAs transistor-vertical-cavity surface-emitting laser,” Appl. Phys. Lett. 102(19), 191101 (2013). [CrossRef]
36. M. Nadeem Akram, Y. Xiang, X. Yu, T. Zabel, and M. Hammar, “Influence of base-region thickness on the performance of pnp transistor-VCSEL,” Opt. Express 22(22), 27398–27414 (2014). [CrossRef] [PubMed]
37. J. J. Chen, G.-B. Gao, J.-I. Chyi, and H. Morkoc, “Breakdown behaviour of GaAs/AlGaAs HBT’s,” IEEE Trans. Delectron. Dev. 36(10), 2165–2172 (1989). [CrossRef]
38. X. Yu, Y. Xiang, T. Zabel, J. Berggren, and M. Hammar, “1.3 μm Buried Tunnel junction InGaAs/GaAs VCSELs,” 37th Workshop on Compound Semiconductor Devices and Integrated Circuits held in Europe (WOCSDICE 2013), May 26th to 29th, 2013, Warnemünde, Germany.
39. N. Suzuki, T. Anan, H. Hatakeyama, and M. Tsuji, “Low resistance tunnel junctions with type-II heterostructures,” Appl. Phys. Lett. 88(23), 231103 (2006). [CrossRef]
