InP-Based Antimony-Free MQW Lasers in 2-3 µm Band

Mid-infrared semiconductor lasers in the wavelength range of 2-3 µm have aroused increasing interests as they are highly desired for a wide range of applications ranging from medical diagnostics to environmental sensing. Access to this wavelength range was mainly achieved by antimony-containing compound semiconductor structures on GaSb substrates. Besides, InP-based In x Ga 1-x As (x>0.53) type-I multiple quantum well laser is a promising antimony-free approach in this band. The emission wave‐ length can be tailored to the 2-3 µm band by increasing the indium composition in the quantum wells. During the demonstration of this kind of lasers, controlling the strain and keeping fair structural quality is the main obstacle. In this chapter, the route for developing this kind of lasers is reviewed. The schemes of pseudomorphic and metamorphic structures are discussed for the 2-2.5 µm and 2.5-3 µm range, respectively. In the pseudomorphic scheme, triangular quantum wells grown by digital alloy technology are applied to restrict the strain and increase the lasing wavelength. Lasers at 2.43 µm were demonstrated under continuous wave operation at room temperature. To extend the emission wavelength longer, an InP-based metamorphic template with larger lattice constant was produced and InAs quantum wells were then grown. The lasing wavelength was further increased up to 2.71 µm. The details on the gas source molecular beam epitaxial growth, device processing as well as performance characterization are presented.


Introduction
The 2-3 µm wavelength range covers both the short side of the mid-infrared (2.5-25 µm) band and the long side of the near-infrared (0.7-2.5 µm) band. Semiconductor laser diodes emitting in this spectral range are very attractive for free space communications, ultra-low loss fluoride fiber communication, solid or fiber lasers (amplifiers) pumping and seeding, light detection and ranging (LIDAR), tunable diode laser absorption spectroscopy (TDLAS), etc., due to the abundant spectroscopic features in this wavelength range. For example, the 2.1-2.4 µm band is free of water and carbon dioxide absorption as shown in Figure 1, which forms a clear atmosphere window; on the other hand, the strong and wide absorption band of water (mainly around 2.7 µm) and the moderate absorption band of carbon dioxide (around 2.0-2.1µm and 2.65-2.85 µm) forms a high contrast area, which is usable for humidity and carbon dioxide monitoring, active spectroscopic imaging of water containing subjects, etc.  Furthermore, as shown in Figure 2, the fingerprint absorption lines of several molecules also exist in th therefore, semiconductor lasers in this band are of great interest in atmospheric pollution monitoring diagnostics by using TDLAS [1,2]. As a monochromatic light source, the laser diode in this wavelength desired for the evaluation of optoelectronic materials and devices such as photodetectors and focal plane arr in this wavelength range, because in the measurement schemes wavelength match is very important to actual features [3,4]. The choices and developments of semiconductor lasers in the 2-3 μm band will be reviewed at first. T introduce the recent progress for the design, growth and demonstration of InP-based antimony-free mult well (MQW) lasers in this band. Schemes of pseudomorphic and metamorphic structures will be discussed μm and 2.5-3 μm ranges, respectively. Furthermore, as shown in Figure 2, the fingerprint absorption lines of several molecules also exist in this band, and therefore, semiconductor lasers in this band are of great interest in atmospheric pollution monitoring and medical diagnostics by using TDLAS [1,2]. As a monochromatic light source, the laser diode in this wavelength band is also desired for the evaluation of optoelectronic materials and devices such as photodetectors and focal plane arrays operated in this wavelength range, because in the measurement schemes wavelength match is very important to acquire their actual features [3,4].
The choices and developments of semiconductor lasers in the 2-3 µm band will be reviewed at first. Then we will introduce the recent progress for the design, growth and demonstration of InP-based antimony-free multiple quantum well (MQW) lasers in this band. Schemes of pseudomorphic and metamorphic structures will be discussed for the 2-2.5 µm and 2.5-3 µm ranges, respectively.

Development of semiconductor lasers in the 2-3 µm band
There are several options of semiconductor lasers in the 2-3 µm wavelength band. Traditionally, such semiconductor lasers are developed in an antimony (Sb)-containing material system on GaSb or InAs substrates. InGaAsSb/AlGa(In)AsSb type-I quantum well (QW) lasers are the primary approach, and high power and single-mode lasers have been achieved and the wavelength range has covered 2-3.8 µm [5][6][7][8][9][10], and continuous wave (CW) operation above 80°C has been reached for 2.1 µm lasers [10]. The laser diodes have been applied in TDLAS demonstrations [11,12]. There were also some sporadic reports on InGaAsSb/GaSb type-II QW lasers covering this band [13,14]. GaSb-based interband cascade lasers combining type-II heterostructure and cascade concept have been developed adequately, but it is still a big challenge to shift the lasing wavelength below 3 µm, which requires new material systems [15].
Compared to GaSb, InP substrates display superior quality and are easier to acquire. Sbcontaining InGaAs/GaAsSb type-II structures on InP substrates have been researched while room temperature (RT) photoluminescence (PL) and sporadic device results have been reported [16,17]. InAsSb quantum dots (QDs) on InP are another potential scheme to demonstrate light sources in 2-3 µm, but the improvements of material quality are still needed to achieve more experimental progress [18]. On the other hand, the growth and processing technologies of the Sb-containing materials are still less mature than those of InP-based Sbfree materials. Besides, the thermal characteristics of antimonide are much inferior to the phosphides and arsenide [19,20]. Many impressive results on mid-infrared intersubband emission of InP-based Sb-free InGaAs/InAlAs quantum cascade lasers have been reported in the recent two decades, whereas the emission wavelengths are mainly longer than 3 µm [21].
Besides the aforementioned options, InP-based Sb-free type-I In x Ga 1-x As (x>0.53) QW laser is a promising approach in this band. As the bandgap wavelength of InAs at RT is as long as 3.5 µm, the emission wavelength of In(Ga)As QW can be tailored to the 2-3 µm range [22]. For this InP-based Sb-free In(Ga)As system, mature growth and processing technologies can be relied on, and a simple structure can be applied for the type-I QW lasers using interband emission, so the laser diodes with better performance could be expected. The QW lasers applying InP-lattice-matched InGaAsP as QW layers have been well developed with the aim of telecom lasers [23]. However, the lasing wavelength is shorter than 1.7 µm. To extend the wavelength beyond 2 µm, a higher InAs composition should be applied and thus strain would be introduced into the QW layers.
The InP-based InGaAs strained QW lasers were first demonstrated by Forouhar et al. in the early 1990s [24]. The lasing wavelengths are around 2 µm at RT, and In 0.75 Ga 0.25 As layers grown by metal-organic vapor phase epitaxy (MOVPE) were applied as the QW layers and other layers were all lattice matched to InP [25][26][27]. As shown in Figure 3( [35][36][37]. Lasers up to 2.4 µm were recently demonstrated by using triangular QWs grown by MBE digital alloy technology [38]. As mentioned earlier, further increase of well width is restrained by the strain between the QW part and InP substrate, and thus the longest emission wavelength is limited. To extend the emission wavelength to longer wavelengths, a metamorphic "virtual substrate" can be produced with a larger lattice constant than InP, e.g. metamorphic In 0.8 Ga 0.2 As layer as shown in Figure 3(b). In this case, the critical thicknesses can be larger than those of pseudomorphic structures, therefore the QW width can be increased, which would increase the emission wavelength. Recently, step-graded InAs x P 1-x buffers on InP have been applied to achieve 2.8-3 µm PL at RT [39][40][41], where 16.5 nm InAs or 22 nm InAs 0.94 P 0.06 were used as the well layers. By using 15 nm InAs QWs on continuously-graded metamorphic In x Al 1-x As buffers, RT PL of QWs at 3.05 µm [42] and low temperature lasing of laser diodes at 2.7 µm have been achieved [43]. This scheme can even be explored to mid-infrared metamorphic InAs QWs on GaAs substrate, which is even more attractive than on InP substrate [44].

Theoretical evaluation of InP-based Sb-free type-I QW lasers
At first, we theoretically evaluate the type-I transition energy and emission wavelength of rectangular pseudomorphic

Theoretical evaluation of InP-based Sb-free type-I QW lasers
At first, we theoretically evaluate the type-I transition energy and emission wavelength of rectangular pseudomorphic In x Ga 1-x As/In 0.53 Ga 0.47 As (x>0.53) and metamorphic In x Ga 1-x As/ In 0.8 Ga 0.2 As (x>0.8) QWs. The subband energies of the QWs were calculated by numerically solving the Schrödinger equation with strain-contained 4×4 Luttinger-Kohn Hamiltonian, and the transition energy from the first electron energy level to the first heavy hole energy level was calculated. The emission wavelength versus the well width of In x Ga 1-x As/In 0.53 Ga 0.47 As (x>0.53) and In x Ga 1-x As/In 0.8 Ga 0.2 As (x>0.8) QWs are shown in Figure 4(a) and Figure 4(b), respectively. The well width is limited by the critical thickness calculated from the force balance model as shown in Figure 4. It is seen that the critical thickness can be extended by using the metamorphic scheme, and therefore, a thicker well width can be used, leading to a longer wavelength.  Furthermore, calculation shows that the lasing wavelength can be extended dramatically if the energy band shape is changed from a rectangular to a triangular one [45,46]. Figure 5(a) shows the schematic band structure of the InAs/In 0.53 Ga 0.47 As triangular QW (TQW) and In 0.765 Ga 0.235 As/In 0.53 Ga 0.47 As rectangular QW (RQW). The two QW structures are considered to have the same strain extent in the QW part. From Figure 5(b), we can see that changing the energy band from the rectangular shape to a triangular one redshifts the lasing wavelength markedly. For the well width of 16 nm, the calculated wavelength of RQW is only about 2.1 µm, but it is beyond 2.5 µm for TQW. The theoretical estimations show the promising potential of the TQW to extend the wavelength and improve the performances.

Pseudomorphic scheme for lasing wavelength below 2.5 μm
The pseudomorphic scheme is based on the lattice-mismatched material system except the active the main concern of such lasers is the control of strain in the QW and to keep fair structural q molecular beam epitaxy (GSMBE) growth parameters and the structures of the QW parts need to b 48]. Then the whole laser structures were grown and the lasers were demonstrated and analyzed.

GSMBE growth and optimization of pseudomorphic triangular QWs
To construct the QWs with triangular shape in practice, usually two methods can be applied. The alloy where the emission source flux changes gradually with an extremely low rate [49], and so-called digital alloy, where short period superlattice composed of two materials are grow thicknesses [50]. The growth rate is normal for the latter one, and this method is especially muc molecular beam epitaxy (MBE). In MBE, the time for opening and closing a shutter is far below 1 s, of the short period superlattice can be high enough. It is also found that digital alloy technology thickness and restrain the three-dimensional material growth [51,52].
A VG Semicon V80H GSMBE system was applied for the growth. The best background vacuum was about 1×10 -11 Torr. The elemental indium and gallium cells with two heaters as well as alumin used as group III sources. The fluxes of these group III emission sources were controlled by the ce V sources were As2 and P2 cracked from Arsine (AsH3) and phosphine (PH3) cracking cells at arou of As2 and P2 were controlled by pressure. Standard beryllium and silicon cells were used as p-and

Pseudomorphic scheme for lasing wavelength below 2.5 µm
The pseudomorphic scheme is based on the lattice-mismatched material system except the active QW region. Therefore, the main concern of such lasers is the control of strain in the QW and to keep fair structural quality. The gas source molecular beam epitaxy (GSMBE) growth parameters and the structures of the QW parts need to be optimized at first [47,48]. Then the whole laser structures were grown and the lasers were demonstrated and analyzed.

GSMBE growth and optimization of pseudomorphic triangular QWs
To construct the QWs with triangular shape in practice, usually two methods can be applied. The first one uses analogue alloy where the emission source flux changes gradually with an extremely low rate [49], and the other one uses the so-called digital alloy, where short period superlattice composed of two materials are grown at digitally setting thicknesses [50]. The growth rate is normal for the latter one, and this method is especially much more convenient for molecular beam epitaxy (MBE). In MBE, the time for opening and closing a shutter is far below 1 s, so the control precision of the short period superlattice can be high enough. It is also found that digital alloy technology can extend the critical thickness and restrain the threedimensional material growth [51,52].
A VG Semicon V80H GSMBE system was applied for the growth. The best background vacuum achieved in this system was about 1×10 -11 Torr. The elemental indium and gallium cells with two heaters as well as aluminum standard cell were used as group III sources. The fluxes of these group III emission sources were controlled by the cell temperatures. Group V sources were As 2 and P 2 cracked from Arsine (AsH 3 ) and phosphine (PH 3 ) cracking cells at around 1000 °C. The fluxes of As 2 and P 2 were controlled by pressure. Standard beryllium and silicon cells were used as p-and n-type doping sources, respectively. The doping levels were controlled by the temperatures of the emission cells. The growth rates of InP-lattice-matched InGaAs(P) and InP layers were all controlled to be around 1 µm/h.

T g (°C)
Parameters from

HRXRD simulation PL wavelengths (µm) PL FWHM (meV)
λ 300  (1) where α is the expected average indium composition in each short period, and d is the period of the short period setting to be 1 nm here. The barrier layer is AlInGaAs grown using the growth parameters of InP-lattice-matched In 0.52 Al 0.48 As and In 0.53 Ga 0.47 As, thus the composition is around Al 0.33 In 0.36 Ga 0.31 As. Table 1 lists the structure parameters, growth conditions, high resolution X-ray diffraction (HRXRD) and PL results of the grown QW samples. The HRXRD (004) scanning curves of all samples are shown in Figure 6. For all the samples, the upper and lower curves show the measured and simulated results, respectively. The good agreement between the measurement and simulation in samples 1-5 indicates the pseudomorphic growth. By simulation, the average indium composition in the well and barrier layers were about 0.765 and 0.355 in samples 1-5, the deduced well widths d w and barrier widths d B for all the samples are also listed in Table  1, agreeing well with the designed values. Figure 7 presents the PL spectra of samples 1-6 at RT and 12 K, respectively. The peak wavelengths and full-width at half-maximum (FWHM) of the PL spectra are listed in Table 1. The growth temperature Tg is a crucial growth parameter for high crystalline and optical quality during the For the growth of high indium materials, generally Tg needs to be lowered. However, for the In0.53Ga0.47As ba Al-containing AlInGaAs barriers, a relatively higher growth temperature is preferred to improve the optical is really unpractical to increase and decrease Tg in the interface of well and barrier layers, because the inte would be significantly deteriorated by the growth interruption. Therefore, Tg is needed to be optimized, espe QW growth. Samples 1 and 2 are with different growth temperatures at 500 °C and 530 °C, respectively. Fi the sharp satellite peaks in the HRXRD scanning curves of samples 1 and 2, denoting the nice crystalline qu samples. Similarly, PL spectra of the two samples have been observed at 300 K as shown in Figure 7, where PL intensity of sample 2 is stronger than that of sample 1. The FWHM at 12 K is 19 meV for sample 2, an sample 1. It is shown that the sample grown at a relatively higher temperature shows better optical quality. The growth temperature T g is a crucial growth parameter for high crystalline and optical quality during the MBE growth. For the growth of high indium materials, generally T g needs to be lowered. However, for the In 0.53 Ga 0.47 As barriers and the Al-containing AlInGaAs barriers, a relatively higher growth temperature is preferred to improve the optical properties. It is really unpractical to increase and decrease T g in the interface of well and barrier layers, because the interface quality would be significantly deteriorated by the growth interruption. Therefore, T g is needed to be optimized, especially for the QW growth. Samples 1 and 2 are with different growth temperatures at 500 °C and 530 °C, respectively. Figure 6 shows the sharp satellite peaks in the HRXRD scanning curves of samples 1 and 2, denoting the nice crystalline quality of both samples. Similarly, PL spectra of the two samples have been observed at 300 K as shown in Figure 7, whereas at 12 K the PL intensity of sample 2 is stronger than that of sample 1. The FWHM at 12 K is 19 meV for sample 2, and 24 meV for sample 1. It is shown that the sample grown at a relatively higher temperature shows better optical quality.  By increasing the well width from 9 nm of sample 2 to 13 nm of sample 3 and 19 nm of sample 4 while keeping the same barrier width, the HRXRD curve becomes indistinct as shown in Figure 6, indicating the deteriorated crystalline quality of sample 4. As the QW width increases, more defects may be generated due to the larger lattice mismatch. The wider QW also weakens the quantum restriction, and the carriers have more probability to occupy the second subband, which induces the larger PL FWHM and the much shorter peak wavelength than designed for sample 4. For sample 5, by increasing the barrier width on the basis of sample 4, the satellite peaks of HRXRD curves recover to be distinct and the PL intensity is enhanced. The PL FWHM is reduced to 17 meV at 12 K and 33 meV at 300 K for sample 5, comparing to 24 meV at 12 K and 36 meV at 300 K for sample 4. The well width of sample 6 is further enlarged to 25 nm in order to extend the wavelength further. It can be seen from Figure 6 that the satellite peaks of the HRXRD curve become indistinct, and two PL peaks at 12 K exist as shown in Figure 6(b), whereas no PL signal is observed at 300 K. It suggests that 2.4 μm is around the ceiling wavelength for the In(Ga)As strained QW.

Demonstration of CW-operated lasers above room temperature
In our preliminary work, InP layers were applied as waveguides for the convenience of the growth, but the lasing could only be achieved below 150 K [53,54]. Subsequently, InGaAsP quaternary alloys with bandgap around 1.1 eV and lattice-matched to InP were used as the waveguide layers, and the laser performances were dramatically improved. In this section, the performances of RQW and TQW lasers at 2.2 μm will be presented. And the demonstration of TQW lasers beyond 2.4 μm with increased QW width will also be introduced.
For the lasers with designed lasing wavelength at 2.2 μm, two structures with RQW and TQW were grown [52]. The active QW regions of both samples were sandwiched between 120-nm-thick InGaAsP waveguide layers, as well as 1000-nm-thick n-type bottom and 1700-nm-thick p-type upper InP cladding layers, and 300-nm-thick p-type In0.53Ga0.47As layers were grown as a top contact layer. The active QW region was formed by four QW layers sandwiched between By increasing the well width from 9 nm of sample 2 to 13 nm of sample 3 and 19 nm of sample 4 while keeping the same barrier width, the HRXRD curve becomes indistinct as shown in Figure 6, indicating the deteriorated crystalline quality of sample 4. As the QW width increases, more defects may be generated due to the larger lattice mismatch. The wider QW also weakens the quantum restriction, and the carriers have more probability to occupy the second subband, which induces the larger PL FWHM and the much shorter peak wavelength than designed for sample 4. For sample 5, by increasing the barrier width on the basis of sample 4, the satellite peaks of HRXRD curves recover to be distinct and the PL intensity is enhanced. The PL FWHM is reduced to 17 meV at 12 K and 33 meV at 300 K for sample 5, comparing to 24 meV at 12 K and 36 meV at 300 K for sample 4. The well width of sample 6 is further enlarged to 25 nm in order to extend the wavelength further. It can be seen from Figure 6 that the satellite peaks of the HRXRD curve become indistinct, and two PL peaks at 12 K exist as shown in Figure 6(b), whereas no PL signal is observed at 300 K. It suggests that 2.4 µm is around the ceiling wavelength for the In(Ga)As strained QW.

Demonstration of CW-operated lasers above room temperature
In our preliminary work, InP layers were applied as waveguides for the convenience of the growth, but the lasing could only be achieved below 150 K [53,54]. Subsequently, InGaAsP quaternary alloys with bandgap around 1.1 eV and lattice-matched to InP were used as the waveguide layers, and the laser performances were dramatically improved. In this section, the performances of RQW and TQW lasers at 2.2 µm will be presented. And the demonstration of TQW lasers beyond 2.4 µm with increased QW width will also be introduced.
For the lasers with designed lasing wavelength at 2.2 µm, two structures with RQW and TQW were grown [52]. The active QW regions of both samples were sandwiched between 120-nm-thick InGaAsP waveguide layers, as well as 1000-nm-thick n-type bottom and 1700-nm-thick p-type upper InP cladding layers, and 300-nm-thick p-type In 0.53 Ga 0.47 As layers were grown as a top contact layer. The active QW region was formed by four QW layers sandwiched between 20-nm-thick In 0.53 Ga 0.47 As barriers for both samples. The QW layers were 3-nm-thick compressive InAs layers in the RQW structure, and 10-nm-thick triangular In 0.53 Ga 0.47 As/InAs/ In 0.53 Ga 0.47 As layers in the TQW structure. The total strain content of the TQW was still larger than that of the RQW due to the much thicker well width.
Ridge waveguide lasers with a strip width of 6 µm were fabricated by standard lithography and wet chemical etching process. Then, a 300-nm-thick Si 3 N 4 layer was deposited by using plasma enhanced chemical vapor deposition (PECVD). On the top of the ridge a 4-µm-wide window was opened. The top and bottom metallic contacts were sputtered Ti/Pt/Au and evaporated Ge/Au/Ni/Au, respectively. The chips were cleaved into 0.8-mm-cavity-length bars with uncoated facet, soldered on copper heat sinks and wire bonded. 20-nm-thick In0.53Ga0.47As barriers for both samples. The QW layers were 3-nm-thick compressive InAs layers in the RQ structure, and 10-nm-thick triangular In0.53Ga0.47As/InAs/In0.53Ga0.47As layers in the TQW structure. The total strain cont of the TQW was still larger than that of the RQW due to the much thicker well width.
Ridge waveguide lasers with a strip width of 6 μm were fabricated by standard lithography and wet chemical etch process. Then, a 300-nm-thick Si3N4 layer was deposited by using plasma enhanced chemical vapor deposition (PECV On the top of the ridge a 4-μm-wide window was opened. The top and bottom metallic contacts were sputtered Ti/Pt/ and evaporated Ge/Au/Ni/Au, respectively. The chips were cleaved into 0.8-mm-cavity-length bars with uncoated fa soldered on copper heat sinks and wire bonded.    The maximum operation temperature of the RQW laser is about 310 K and the lasing peak is about 2.16 μm at 310 whereas the TQW laser shows an increased maximum operation temperature up to 330 K. Over the whole temperat range, the lasing peaks of the TQW laser are about 60 nm longer than those of the RQW laser, probably due to the mo error in theoretical design as well as due to the growth error. These results show that the TQW laser can operate at hig temperatures even with a longer lasing wavelength, proving the benefit of TQW. The average temperature coefficien the lasing wavelength Δλ/ΔT is about 0.8 nm/K for the RQW laser and 1 nm/K for the TQW laser. The temperature-dependent I-P characteristics and the I-V curve at 300 K of the lasers are shown in Figure 9. T threshold current of the RQW laser is 124 mA at 300 K, corresponding to a threshold current density of 2.58 kA/cm 2 , a the output power is 3.6 mW/facet at an injection current of 400 mA. In contrast, the TQW laser shows a much low threshold current of 68 mA (1.42 kA/cm 2 ) at 300 K. Also, a much higher output power of 10.4 mW/facet is obtained at K, almost three times that of the RQW laser. The turn-on voltage and the differential resistance for both lasers are alm the same at about 1.25 V and 1.87 Ω at 300 K.  Figure 8 shows the CW lasing spectra at a driving current of 1.2 times the threshold current in a temperature step of 10 K. The maximum operation temperature of the RQW laser is about 310 K and the lasing peak is about 2.16 µm at 310 K, whereas the TQW laser shows an increased maximum operation temperature up to 330 K. Over the whole temperature range, the lasing peaks of the TQW laser are about 60 nm longer than those of the RQW laser, probably due to the model error in theoretical design as well as due to the growth error. These results show that the TQW laser can operate at higher temperatures even with a longer lasing wavelength, proving the benefit of TQW. The average temperature coefficient of the lasing wavelength Δλ/ ΔT is about 0.8 nm/K for the RQW laser and 1 nm/K for the TQW laser.
The temperature-dependent I-P characteristics and the I-V curve at 300 K of the lasers are shown in Figure 9. The threshold current of the RQW laser is 124 mA at 300 K, corresponding to a threshold current density of 2.58 kA/cm 2 , and the output power is 3.6 mW/facet at an injection current of 400 mA. In contrast, the TQW laser shows a much lower threshold current of 68 mA (1.42 kA/cm 2 ) at 300 K. Also, a much higher output power of 10.4 mW/facet is obtained at 300 K, almost three times that of the RQW laser. The turn-on voltage and the differential resistance for both lasers are almost the same at about 1.25 V and 1.87 Ω at 300 K.   The threshold current as a function of temperature is shown in Figure 10, covering the temperature range from 200 310 K and 330 K for the RQW and TQW lasers, respectively. The characteristic temperatures were achieved from temperature-dependent threshold current density as shown in the inset of Figure 10. In the low temperature r 200-280 K, a characteristic temperature of T0=82.1 K was obtained for the TQW laser, higher than the value of T0=65.2 the RQW laser. In the high temperature range beyond 290 K, T0 is decreased for both lasers, but the TQW laser also higher value of T0=40.5 K than the T0=32.1 K for the RQW laser.  The threshold current as a function of temperature is shown in Figure 10, covering th 310 K and 330 K for the RQW and TQW lasers, respectively. The characteristic tem temperature-dependent threshold current density as shown in the inset of Figure   Figure 10. Temperature-dependent threshold current of rectangular QW laser and triangular QW laser around 2.2 µm at various temperatures under CW driving condition. The inset shows the temperature-dependent threshold current densities and characteristics temperatures. The threshold current as a function of temperature is shown in Figure 10, covering the temperature range from 200 K to 310 K and 330 K for the RQW and TQW lasers, respectively. The characteristic temperatures were achieved from the temperature-dependent threshold current density as shown in the inset of Figure 10. In the low temperature range 200-280 K, a characteristic temperature of T 0 =82.1 K was obtained for the TQW laser, higher than the value of T 0 =65.2 K for the RQW laser. In the high temperature range beyond 290 K, T 0 is decreased for both lasers, but the TQW laser also has a higher value of T 0 =40.5 K than the T 0 =32.1 K for the RQW laser. By further increasing the well width of TQW laser to 16 nm, the lasing wavelength of the device has been extended. Figure 11(a) shows the CW lasing spectra in the temperature range of 200 K to 340 K. The lasing wavelength is 2.37 µm at 300 K and red-shifts to 2.43 µm at 340 K. The average temperature coefficient of the wavelength is 1.2 nm/K in 200-340 K as shown in the inset of Figure 10(a). As shown in Figure 11(b), the output power at 200 K in CW mode is 23 mW/facet injected by a current of 350 mA and at 300 K the power dropped to around 11 mW/ facet. At 200 K the threshold current is 24 mA and at 300 K the threshold current is increased to 62 mA. The corresponding threshold current density is 1.3 kA/cm 2 for the laser with four QWs and 325 mA/cm 2 for each QW, respectively. At 340 K the maximum output power is decreased to 1.4 mW/facet with a threshold current of 186 mA.
The laser threshold current density and external differential quantum efficiency η d as a function of temperature are shown in Figure 12(a). The characteristic temperature T 0 is derived to be about 99 K in the temperature range of 200-300 K, and decreases to 35 K in 300-340 K range. The external differential quantum efficiency at 200 K is 41% and is decreased to 23% at 300 K and is further dropped to 7% at 340 K. The quantum efficiency characteristic temperature of T 1 =142 K is obtained by fitting the external differential quantum efficiency in 200-320 K temperature range. The reciprocal external differential quantum efficiency η d as a function of the cavity length is shown in figure 12(b). The internal quantum efficiency η i and internal loss α i are calculated by fitting the data following 1/η d =(1/η i )[1+α i L/ln(1+R)], where L is the cavity length and R is the facet reflectivity. Taking R as 0.35 for the as-cleaved facet in the calculation, the internal quantum efficiency and internal loss is calculated to be 58% and 19 cm -1 , respectively. Note that the internal quantum efficiency value is favorable but the loss is relatively high, which limits the laser performances. The intervalence band absorption in the p-type InP up cladding is one main cause of the large internal loss. The inset of Figure 11(b) illustrates the threshold current density versus the reciprocal cavity length. By extrapolation, the threshold current density at infinite cavity length is 841 A/cm 2 (210 A/cm 2 per QW).

Metamorphic scheme for lasing wavelength beyond 2.5 µm
For the metamorphic scheme, InAlAs graded buffers were grown on InP substrate to form the metamorphic template with larger lattice constant than InP. The InAs QW structures were then grown on the template with thicker QW width, and the wavelength could be beyond 2.5 µm. For this kind of metamorphic laser structures, the material quality is very crucial for the device performance, therefore both of the buffer layers and growth parameters were optimized to improve the buffer quality. On the basis of the buffer optimization, the lasers were grown, demonstrated and characterized.

GSMBE growth and optimization of metamorphic buffers and QWs
As a preliminary experiment, the metamorphic buffer composed of a 2.5-µm-thick In x Al 1-x As continuously graded buffer and a 1-µm-thick In 0.8 Ga 0.2 As virtual substrate layer were grown on InP substrate after the growth of a 100-nm-thick InP buffer layer. In the graded buffer, the indium composition x was graded from 0.52 to 0.8 through the simultaneously linear increase of indium source temperature and decrease of aluminum source temperature. After that, two periods of InAs/In 0.53 Ga 0.47 As strained QWs were grown. The In 0.53 Ga 0.47 As barrier is used to compensate the compressive strain of InAs well comparing to the virtual substrate layer. The lattice mismatch of InAs well and In 0.53 Ga 0.47 As barrier with respect to In 0.8 Ga 0.2 As virtual substrate layer is +1.3% and -1.8%, respectively. Two samples were grown and the widths of InAs well and In 0.53 Ga 0.47 As barrier were 10 nm/7 nm and 15 nm/10 nm, respectively to form a strain compensated QW structure with minimal residual strain. The In 0.53 Ga 0.47 As barrier can also stop the diffusion of exited carriers between the QW region and the rest part of the sample.
The grown samples show mirror-like surface morphology without haziness under optical microscopy. Regular cross-hatch patterns are observed on the surfaces with ridges and troughs along the [110] and [1-10] crystal directions. The root mean square (RMS) roughness measured by atomic force microscope (AFM) over 40×40 µm 2 is 3.9 nm and 6.1 nm for the 10 nm and 15 nm QW lasers, respectively. From the cross-sectional transmission electron microscope (XTEM) images of 15 nm QW laser measured at 160 kV as shown in Figure 13, the dislocations are mainly localized within the In x Al 1-x As graded buffer and do not propagate into the upper structures. The In 0.8 Ga 0.2 As virtual substrate layer is free of dislocations in the XTEM measurements, which means the threading dislocation density is under the detection limit (about 10 7 cm -2 ). In the amplificatory image of the InAs/In 0.53 Ga 0.47 As QWs region, distinct interfaces between the well and barrier layers are observed. The thickness fluctuations of wells and barriers are revealed around ±1 nm by a careful inspection.

GSMBE growth and optimization of metamorphic buffers and QWs
As a preliminary experiment, the metamorphic buffer composed of a 2.5-μm-thick InxAl1-xAs continu and a 1-μm-thick In0.8Ga0.2As virtual substrate layer were grown on InP substrate after the growth of buffer layer. In the graded buffer, the indium composition x was graded from 0.52 to 0.8 through linear increase of indium source temperature and decrease of aluminum source temperature. After InAs/In0.53Ga0.47As strained QWs were grown. The In0.53Ga0.47As barrier is used to compensate the co InAs well comparing to the virtual substrate layer. The lattice mismatch of InAs well and In0.53Ga0.47As to In0.8Ga0.2As virtual substrate layer is +1.3% and -1.8%, respectively. Two samples were grown and well and In0.53Ga0.47As barrier were 10 nm/7 nm and 15 nm/10 nm, respectively to form a strain compen with minimal residual strain. The In0.53Ga0.47As barrier can also stop the diffusion of exited carrie region and the rest part of the sample.   The HRXRD ω/2θ (004) scanning curves were measured and shown in Figure 14 (thic intensity corresponds to InP substrate, and the wide peak with a relaxed mismat corresponds to In0.8Ga0.2As virtual substrate layer. The HRXRD signals from the virtual some difference, but become the same after etching away the QW regions at surface. R InAs/In0.53Ga0.47As QWs, are observed symmetrically on the both sides of the In0.8G compensation effect with respect to the In0.8Ga0.2As virtual substrate layer. The lattice performed and the simulated curves are shown in Figure 14 (thin lines). In the s wells/barriers were set the same as the design. The InxAl1-xAs buffer was assumed g virtual substrate layer fully relaxed with respect to InP substrate, meanwhile InAs we fully strained with respect to In0.8Ga0.2As virtual substrate layer. The HRXRD ω/2θ (004) scanning curves were measured and shown in Figure 14 (thick lines). The peak with the largest intensity corresponds to InP substrate, and the wide peak with a relaxed mismatch of 1.8×10 -2 with respect to InP corresponds to In 0.8 Ga 0.2 As virtual substrate layer. The HRXRD signals from the virtual substrate of the two samples show some difference, but become the same after etching away the QW regions at surface. Rough satellite peaks, related to the InAs/In 0.53 Ga 0.47 As QWs, are observed symmetrically on the both sides of the In 0.8 Ga 0.2 As peak, indicating the strain compensation effect with respect to the In 0.8 Ga 0.2 As virtual substrate layer. The lattice dynamical simulations were also performed and the simulated curves are shown in Figure 14 (thin lines). In the simulation, the thicknesses of the wells/barriers were set the same as the design. The In x Al 1-x As buffer was assumed graded relaxed and the In 0.8 Ga 0.2 As virtual substrate layer fully relaxed with respect to InP substrate, meanwhile InAs wells and In 0.53 Ga 0.47 As barriers were fully strained with respect to In 0.8 Ga 0.2 As virtual substrate layer.
The PL measurements were performed as shown in Figure 15. The photo-excited carriers are very sensitive to threading dislocations due to the diffusion length of photo-excited carriers in micrometers. Therefore, the PL intensity can reflect the material quality. The strong PL intensity at 300 K in Figure 15(a) indicates few threading dislocations in QW structures. For each sample, two emission peaks can be observed at 300 K, where the peak at about 2.4 µm corresponds to In 0.8 Ga 0.2 As virtual substrate layer, the peaks at 2.90 µm and 3.05 µm for the two samples correspond to the active QWs. In Figure 15(b) the peak corresponds to In 0.8 Ga 0.2 As buffer blue-shifts to about 2.3 µm at 77 K. The PL signals of active QWs blue-shift to about 2.6 µm and 2.7 µm at 77 K for the 10 nm and 15 nm QW lasers, respectively, and the peak shapes are affected significantly by the water absorption. The relatively wide PL peaks even at low temperatures are possibly due to the presents of lattice defects and the poor carrier confinement on the top side of the QW. The metamorphic QWs with 15 nm thick InAs layer still show moderate optical quality. This structure is promising for the lasers around 3 µm.

Demonstration of metamorphic CW-operated lasers at low temperatures
To demonstrate the metamorphic lasers, the effects of strain-compensated QWs and separate confinement heterostructure (SCH) were studied. The growth of the samples started with a 200-nm-thick n + InP buffer layer grown, followed by a 1800-nm n + In x Al 1-x As continuously graded buffer layer, both highly doped with Si to about 2×10 18 cm -3 . In the graded buffer the indium composition was graded from 0.52 to 0.84 introducing an indium composition nm QWs, respectively. The absorption bands of water are also indicated [42]. (Reprinted with pe The PL measurements were performed as shown in Figure 15. The photo-excited carr dislocations due to the diffusion length of photo-excited carriers in micrometers. There material quality. The strong PL intensity at 300 K in Figure 15 thick InAs layer still show moderate optical quality. This structure is promising for th

Demonstration of metamorphic CW-operated lasers at low temp
To demonstrate the metamorphic lasers, the effects of strain-compensated QWs and se (SCH) were studied. The growth of the samples started with a 200-nm-thick n + InP 1800-nm n + InxAl1-xAs continuously graded buffer layer, both highly doped with Si overshoot of 0.04 to achieve the full relaxation of the metamorphic layer. Then an 800-nm n + In 0.8 Al 0.2 As cladding layer with carrier density of about 2×10 18 cm -3 was grown to form a template. Then a 150-nm-thick nwaveguide layer, undoped active double QWs and a 150nm-thick pwaveguide layer were grown [55]. The schematic band energy of the active regions for samples A, B and C can be found in Figure 16 Figure 17 shows the HRXRD scanning curves and PL spectra of the samples after etchi and p + In0.8Al0.2As cladding layer. Rough satellite peaks, related to the QWs, can be In0.8Al0.2As/In0.8Ga0.2As peaks for the samples, although they are not so distinct  Figure 17 shows the HRXRD scanning curves and PL spectra of the samples after etching away the p-type contact layer and p + In 0.8 Al 0.2 As cladding layer. Rough satellite peaks, related to the QWs, can be observed on the left side of the In 0.8 Al 0.2 As/In 0.8 Ga 0.2 As peaks for the samples, although they are not so distinct. Comparing the intensities of In 0.8 Al 0.2 As/In 0.8 Ga 0.2 As and QW peaks for the three samples, it is observed that the layer peak intensities of samples A and C with strain-compensated QWs are stronger than those of sample B without strain compensation. The lattice dynamical simulations were also performed and the HRXRD intensities of layer peaks on the left side of InP substrate were very close for all the samples in the simulations. Therefore, the measured stronger layer peak intensities of samples A and C indicate their better material quality than sample B due to the strain compensation of QW regions. As shown in Figure 17(b), the PL signals between 2.6-2.8 µm are affected by the absorption of water vapor. Two PL peaks are observed for sample A, the relatively weak peak at 1.55 µm corresponds to the In 0.8 Al 0.2 As cladding and waveguide layers, and the peak at 2.88 µm corresponds to the QWs. For sample B, there are also two PL peaks, and the PL signal of In 0.8 Al 0.2 As is not clear in linear scale. The two PL peaks at 2.51 µm and 3.02 µm correspond to In 0.8 Ga 0.2 As waveguide layers and the QWs, respectively. The PL intensities of the QWs for samples A and C are several times of that of sample B at both 300 K and 77 K, confirming the better material quality of samples A and C. The phenomenon of PL measurements is consistent with the results of HRXRD.
The ridge waveguide lasers were then demonstrated and the cross-section of the ridge waveguide structure was shown in the upper inset of Figure 18(a), where the upside of the ridge was narrowed due to the lateral etching. For the devices fabricated by sample A, no lasing was observed even at 77 K and injected by current up to 1 A. For sample B, around 2.3 µm lasing was observed at low temperatures. This implies that the SCH is very crucial for the device performances, even more important than the crystalline quality of the QWs. For sample B the optical confinement factor is calculated about 0.22, but for sample A the confinement factor is only 0.12 because only the QWs provide a very limited optical confinement in sample A. Figure 18 shows the typical 77 K CW spectrum of the device fabricated by sample B injected by the current of 195 mA. The lasing was declined markedly as the temperature increases. The maximum operation temperature was about 100 K in CW mode and 170 K in pulsed mode with a 5% duty cycle. The lasing transition energy is close to the transition from In 0.8 Ga 0.2 As waveguide layers, indicating that excessive band filling exists and higher gain from the QW transition is required. Quite a few carriers may escape from the QWs to the In 0.8 Ga 0.2 As layers, due to the small offsets of conductive and valence bands between InAs and In 0.8 Ga 0.2 As layers, especially when injected by relatively high currents. The 77 K I-V-P curves of sample B are shown in the lower inset of Figure 18(a). The output power was smaller than 1 mW/facet and the turn-on voltage was about 0.5 V. The temperature-dependent spectra of sample C. The upper inset shows the spectra at different injection currents at 120 K. The lower inset shows the I-V-P spectra at 77 K [55]. (Reprinted with permission from IOP) For sample C, the CW lasing as long as 2.7 µm was observed, and the temperature-dependent lasing spectra are shown in Figure 18(b). The laser was operated in CW mode and injected by the current of 1.2 times the threshold current. As shown in the lower inset of Figure 18(b), the output power at 77 K was 5.7 mW/facet injected by the current of 400 mA and the turn-on voltage was about 0.65 V. The lasing wavelength moved from 2.70 µm at 77 K to above 2.71 µm at 110 K. When the temperature increased to 120 K, the lasing bounded to around 2.34 µm. The luminescence spectra at different injection currents at 120 K are shown in the upper inset of Figure 18(b). Two luminescence envelops at about 2.34 µm and 2.7 µm can be observed under the injection current of 150 mA. As the injection current increased the 2.34 µm luminescence was enhanced whereas the 2.7 µm luminescence was declined, and only the lasing at 2.34 µm could be observed when the injection current was increased to 170 mA.
Sample C combines with SCH and strain-compensated QWs and shows the best performances in these three metamorphic QW laser structures. The calculated optical confinement factor of sample C is about 0.21, close to that of sample B with SCH. The larger bandgap of In 0.6 Ga 0.4 As barrier layers compared to that of In 0.8 Ga 0.2 As barriers in sample B can also enhance the carrier confinement and gain of QWs. On the other hand, the strain-compensated QWs ensure the feasible material quality. Nevertheless, the confinements in this laser structure are still unfavorable, as the bandgap of In 0.8 Ga 0.2 As waveguide layers is smaller than that of In 0.6 Ga 0.4 As barriers, and the refraction index is larger in In 0.8 Ga 0.2 As waveguide. The excessive band filling carrier overflow still exists and the lasing at 2.34 µm is possibly due to the transition from the waveguide layers. Generally, increasing the QW number is an effective approach to improve the optical gain and confinement, but more QWs may introduce more strain in the QW region. The growth process needs to be optimized and a tradeoff needs to be considered between the material quality and the QW number.

Conclusion
In conclusion, we have demonstrated InP-based Sb-free QW lasers in 2-3 µm wavelength band by optimizing the structural design, GSMBE growth and device processing. To control the strain in the pseudomorphic QW region, the QWs with triangular shape were grown by using digital alloy technology. RT-CW lasers up to 2.43 µm have been successfully achieved by this approach and the performances are competitive to those well-developed GaSb-based QW lasers. Metamorphic InAs QWs were grown on InP-based In 0.8 Al 0.2 As template to extend the lasing wavelength even longer. The lasers with CW lasing up to 2.71 µm have been demonstrated by applying 15 nm thick InAs QWs, although they still can only work at a low temperature of 110 K. Our efforts make clear that the extending of the lasing wavelength of InPbased Sb-free QW lasers into 2-3 µm band is not only attractive, but also practical. For CW operation of the laser at RT, the lasing wavelength up to about 2.5 µm should be possible through the fine design of the epitaxial structure, as well as optimization of the growth and processing technologies, adopting pseudomorphic or metamorphic schemes. For longer wavelengths, limited by the strain in pseudomorphic scheme or poor material quality and confinements of both photons and carriers in metamorphic scheme, the reachable performances of the lasers are degraded remarkably.