1. Introduction

A transistor laser (TL) has the structure of a transistor with multi-quantum wells (MQWs) near its base region [1,2]. From a TL, an electrical signal can be outputted simultaneously with a light signal by inputting one electrical signal, making it suitable for future optoelectronic integrated device applications [2]. In TLs, the distribution of the injected carriers is tilted and the carriers not recombined in the MQWs are swept out of the base region, leading to high modulation bandwidths of the TLs [3,4]. Thus, TLs are promising light sources for high speed optical data transmission systems, such as interconnections between boards and chips. Because of the transistor structure, TLs have many interesting characters, including gain compress when base current is larger than threshold [1], voltage controlled mode of operation [5,6] and abnormal temperature performance [7].

Room temperature continuous wave (CW) operations of GaAs based shallow ridge TLs with short wavelength emissions (~1μm), in which MQWs are embedded in the base region, have been realized successfully [1,5]. The fabrication of long wavelength TLs base on InP material that are fitful for low loss fiber systems, however, is more difficult because of the relatively low material gain of the InP material system and the strong free carrier absorption of p type base material at 1.3 or 1.5μm. An InP based shallow ridge TL with 1.5μm wavelength has been reported and can work only at about −190 °C [8]. AlGaInAs/InP npn TLs with a buried emitter ridge structure and 1.3μm emissions have been fabricated [9]. Though room temperature CW workings have been achieved, a complex fabrication process, which includes at least three epitaxy runs and several precise etching steps, is needed and the current gain is only 0.02 [9].

InP based TLs with a deep ridge structure have also been presented [10–12]. With the structure, the effects of p dopant in the base layer on the optical properties of the device can be decreased noticeably compared with the shallow ridge structure [10]. Relative to the TLs with buried ridge structure [9], a much simpler fabrication procedure, which includes only one epitaxy run, is needed for the device. However, CW working was achieved only below −40 °C up to now [10]. In this paper, we report the fabrication of deep-ridge npn-InGaAsP/InP transistor lasers with several processes specifically designed to improve the device performance. CW operations up to 20 °C of deep-ridge npn-InGaAsP/InP transistor lasers emitting at 1.5-μm wavelength are achieved. It is found that at low base current, the device lasers on the first excited state of the MQWs. At high base current, however, the device shows stimulated emissions on the ground state transition. The trend is contrary to what has been observed in the GaAs based TLs and is beneficial for the applications of the TLs.

2. Experimental procedure

An AIXTRON 3 × 2 in. close-coupled showerhead metalorganic chemical vapor deposition (MOCVD) system was used for the growth of the InGaAsP/InP TLs. The substrates used for the devices are (001) oriented epiready s doped InP (n = 2 × 10¹⁸ cm⁻³). The source materials are Trimethylindium (TMIn), Trimethylgallium (TMGa), arsine (AsH₃) and phosphine (PH₃). Disilane (Si₂H₆) and diethylzinc (DEZn) were used for n-type and p-type doping, respectively. The schematic structure of the deep ridge TLs is shown in Fig. 1. The device materials were grown in a single MOCVD run, including a 200 nm Si-doped buffer layer (n = 0.9 × 10¹⁸ cm⁻³), a 200 nm undoped InP collector, a 150 nm undoped quaternary InGaAsP collector with a 1.2 μm emission wavelength (1.2Q), a 300 nm Zn-doped 1.2Q InGaAsP base layer (p = 1.6 × 10¹⁸ cm⁻³), a 90 nm undoped 1.2Q InGaAsP setback layer, an undoped active layer which consists of 5 InGaAsP QWs with 1.5 μm emission wavelength, a 80 nm undoped 1.2Q InGaAsP upper waveguide layer, a 800 nm n-doped InP emitter/cladding layer (n = 2.8 × 10¹⁸ cm⁻³), and a 100 nm n + 1.2Q InGaAsP emitter contact layer (n = 3 × 10¹⁸ cm⁻³). A 15 minute growth interruption was introduced before the growth of the base layer to reduce the group III interstitials in the Si doped buffer layer, thus reducing the Zn diffusion [11]. The detailed fabrication process of the deep ridge edge emitting TLs can be found in [10]. The etching of the 3 μm wide emitter ridge terminates in the upper part of the base layer, as shown schematically in Fig. 1.

 

Fig. 1 The schematic structure of the deep ridge TL.

Download Full Size | PDF

The relatively poor quality of our previous devices [10] can be attributed to two aspects. First, a large amount of p dopant Zn still diffuses from the base layer into the MQWs, which degrades the quality of the MQWs. Then, there are a large number of non-radiative recombination centers at the exposed side walls of the emitter ridge. To improve the performance of the TLs, several methods were used during the fabrication process of the device. To further decrease the Zn concentration in the MQWs, the growth temperature of the InP emitter was set to be 600 °C, which is lower than the 640 °C of the rest of the structure. It was shown that the diffusion of Zn can be suppressed greatly by lowering the temperature [13]. Additionally, the growth rate of the InP emitter was increased to be 1.5 times of that of the InP buffer layer. The growth time of the emitter, during which Zn diffuses, can thus be reduced for a given emitter thickness. Then the thickness of the undoped set back layer is increased to 90nm. With a thicker set back layer, both the Zn diffusion into the MQWs and the optical absorption of the heavily p-doped base material and are smaller. Figure 2 shows the depth and concentration profile of Si and Zn with Ga, As and P as markers. As can be seen, as a result of the above measures, the Zn level in the MQWs is as low as 1.5 × 10¹⁷ cm⁻³. To reduce the non-radiative recombination centers on the exposed MQW side walls, (NH₄)₂S aqueous solution were used for the passivation of the deeply-etched base-emitter junction. The passivation procedure is as follows: first, samples are dipped in the (NH₄)₂S solution for 600 seconds after base-emitter junction etching. Then, the samples are rinsed in acetone and methanol successively. Finally, the samples are blown dry by Nitrogen after washing, before transferred rapidly for further processing.

 

Fig. 2 The depth and concentration profile of Si and Zn with Ga, As and P as markers.

Download Full Size | PDF

The devices have a length of 1500 μm and were alloyed onto Cu heat sinks for testing. The facets of the devices were left uncoated. The optical output power of the devices was measured by an integrating sphere. An Agilent B2902A power source was used to provide current and bias voltage. All the measurements were conducted under CW conditions.

3. Device characterizations and discussions

Figure 3 shows the light intensity–base current (L-I_B) characteristics of the device under common emitter (CE) configuration. As can be seen CW operation of the device at up to 20 °C is achieved. When the collector emitter voltage (V_CE) is 0 V, the threshold currents of the 1500μm long device are about 180mA, 230mA and 340mA, for the temperatures of 10 °C, 15 °C and 20 °C, respectively. As V_CE is increased to about 1.4V, the threshold currents decrease dramatically and saturate at about 75mA, 84mA, and 110mA, for the temperatures of 10 °C, 15 °C and 20 °C, respectively. At low V_CE, the large threshold currents are a result of the forward bias of the base collector junction. Part of I_B follows to the collector and dose not contribute to the light emission in such a case. As V_CE increases further to be above 1.4V, while the threshold currents increase slightly, the maximum output power decrease noticeably, which are caused by the self heating effect.

 

Fig. 3 CW L-I_B characteristics of the device under CE configuration and different temperatures when V_CE = 0 V (a), different V_CE when T = 10 °C (b), 15 °C (c), and 20 °C (d). The increase step of V_CE is 0.2V.

Download Full Size | PDF

The achievement of up to 20 °C CW working of the TLs can be attributed to several factors related to the structure and the fabrication of the device. As can be seen from Fig. 1, in the deep ridge TLs, the MQWs are placed above the heavily Zn-doped base layer, and a thick undoped setback layer is used to further separate the MQWs from the base layer. Thus, the damage of the material quality coming from the diffusion of Zn into the MQWs can be released. The optical absorption of the heavily p base material can also be decreased. By lowering both the growth temperature and time of the thick InP emitter, the Zn diffusion into the MQWs can be further suppressed. The photoluminescence (PL) spectrum of the MQWs (not shown here) was obtained after the InP emitter layer and the InGaAsP emitter contact layer were removed. The full width at half maximum (FWHM) of the PL spectrum is 60nm, which is significantly smaller than the 89nm of the MQWs in our previous device [11], indicating a better material quality. Finally, surface passivation is used during the fabrication process of the device, reducing the non-radiative recombination of carriers on the exposed MQW side walls greatly.

The collector current versus V_CE (I-V) curves of the TLs under CE mode are shown in Fig. 4. Under the CW operation condition, I_B is increased in a 10mA interval from 0 to 140 mA with V_CE swept from 0 to 2 V. When I_B = 100mA which is larger than the threshold currents and V_CE = 2V, the CW current gain are 4.9 and 3.5, respectively, for the temperatures of 15 °C and 10 °C. The current gain is significantly larger than the 0.14 of the shallow ridge InP TL as reported in [8] and the 0.02 of the InP TLs with buried emitter ridge as in [9], indicating a good quality of our devices. With a lot larger out put power and a thicker set back layer, the current gain of the TLs is still larger than those of the deep ridge light emitting transistors [11], showing the effectiveness of the surface passivations. The higher current gain at 15 °C than at 10 °C reflects the fact that at a higher temperature, fewer carriers are recombined radiatively in the MQWs and more are collected by the collector.

 

Fig. 4 The CW I-V curves of the TLs under CE mode at 10 °C (a) and 15 °C (b). I_B is increased from 0 mA to 140mA with a 10 mA step.

Download Full Size | PDF

The output spectra of the TLs are obtained by an optical spectrum analyzer (OSA). The emission from a TL is coupled into a single mode fiber and fed into the OSA. For the measurements, the TLs are in CE configuration. The spectra at different I_B and when V_CE = 1.5V and T = 10 °C is shown in Fig. 5. At I_B of 110 mA, the laser operation occurs on the first excited state of the MQWs at λ ≈1529 nm. As I_B = 130mA, the TL lasers on both the excited state and the ground state. As I_B is larger than 150 mA, the device emits only at the ground state at λ ≈1539nm. This trend is contrary to what have been observed from GaAs based TLs, where ground state lasing occurs at low base current and excited state lasing at high base current [14–16].

 

Fig. 5 The spectra at different I_B under CE mode and when V_CE = 1.5V and T = 10 °C.

Download Full Size | PDF

That the TLs emission first at the excited state wavelength might be attributed to the deep ridge structure, in which the carriers are better confined laterally than in the shallow ridge TLs, making it easier for the carriers to occupy the excited state. It has been shown that for TLs the excited state emissions have advantages, such as higher modulation band width [14] and lower relative intensity noise (RIN) [16]. However, for the reported GaAs based TLs, the excited state emissions were realized at relatively high base current (several times of the threshold current) or device temperature [14–16]. Comparatively, the excited state emissions can be obtained at the beginning of lasing of our TLs, helping to reduce the power consumption of the devices.

The shift of emission wavelength from the excited state at low I_B to the ground state can be explained by the change of carrier flow. At high I_B, the base emitter voltage (V_BE) is larger than V_CE (1.5V) and the base collector junction is forward biased. Because the carrier pulling effect of the collector is weakened compared with when V_BE is smaller than V_CE, the electrons that pass through the MQWs diffuse laterally in the base layer, which leads to the recombination of holes with electrons in materials outside of the MQWs. Thus the population density of holes occupying the excited state is decreased, reducing the optical gain. When the gain is lower than a threshold value, the operation of the laser is changed to the ground state.

Finally, we would like to point out that several other techniques can be used to further enhance the performance of the device. Though the Zn concentration in the TLs is relatively low, its effect on the optical quality of the MQWs is still obvious. The FWHM of the MQWs of the TLs (60nm) is apparently larger than that (50nm) of the MQWs with the same growth parameters in a conventional diode laser, indicating a worse material quality. To this problem, a Si doped setback layer can be used. It is shown that in material with Si doping, the mobility of Zn can be reduced greatly by forming a donor-acceptor pair [17]. The diffusion of Zn toward the MQWs can then be decreased. A thicker setback layer than that used in the present device can also be used to alleviate the effects from the p base layer. Besides, the MQWs and the upper waveguide layer can also be doped with Si, so that the emitter resistance can be reduced. The doping level needs to be optimized by considering its effects on the properties of the devices. With a 1500 µm cavity length, the forward base-collector resistance is larger than 10Ω. The related self heating effects can harm the device performance seriously. The resistance can be decreased by reducing the distance between the base metal and the emitter ridge, which is about 2µm in our present device. The self align technique which is widely used for the fabrication of conventional transistor can be used to decrease the distance to be less than 1µm [18]. A much smaller resistance can thus be obtained. It should be noted that the short wavelength emission as in Fig. 5 is not resulted from a mode shift due to a suboptimal mode confinement because high resolution (0.02nm) optical spectrum measurements show that the two emissions at different wavelengths such as when I_B = 130mA belong to the same Fabry–Pérot mode comb. Because of having different effective refractive index, different lateral modes possess clearly different mode combs [19].

4. Conclusion

In summary, InP based deep ridge TLs with 1530nm emissions have been fabricated. CW operation of the device up to 20 °C is achieved. Besides light emissions, the devices have CE mode CW current gains, which are larger those of the previously reported long wavelength TLs. The TLs show stimulated emissions on the ground state transition at the beginning of lasing, which is beneficial for the applications of the devices. These results indicate that the deep ridge structure is promising for the fabrication of low cost high quality long wavelength TLs.