Experiment: MBE-growth of p-i-n laser/SOA’s structures

When a thin epi-layer which has an unstrained lattice constant a_u is grown on a thick substrate which has a different lattice constant a₀, the epi-layer will be coherently strained (i.e. pseudomorphic) as schematically demonstrated in Fig. 3.12 if the epi-layer thickness d is less than the critical layer thickness. In Fig. 2.4, we have already pointed out that our strained layers are controlled below which critical layer thickness. In such a pseudomorphic strained layer grown on InP, the growth rate in the perpendicular direction is expressed as

GR GR_{I A @I P} 1

where GRInAs@InP is the effective InAs growth rate on InP substrate, ε⊥ is the out-of-plane strain which is mentioned in appendix-A, and a⊥ is out-of-plane lattice constant. Depending on Eq. (3.7), Table 3-1 shows the relation of GR_InAs@InP vs. GR⊥ for possible pseudomorphic strained alloys based on 1eV bandgap InGaAlAs. According to the calculated results as GRInAs@InP is 0.4 μm/hr, those GR⊥ of strained epi-layer used in the design of MD3QW ( ● ) will be not exceeded 1μm/hr, except the layer of In¹_0.305Ga³_0.417Al¹_0.278As (k) for hole-stopping barrier. This is also the reason why we do not use a higher growth rate for lattice-matched alloys in the beginning.

In order to make sure if the MBE epitaxial sample is high quality and consistent with our original design, DCXRD rocking measurement and cross-section high resolution (HR) TEM have been applied in the final wafer calibration. Fig. 3.13 shows that the MBE sample is consistent with our original design according to the X-ray diffraction rocking measurement. In Fig. 3.14, the cross-section HR-TEM pictures separately caught from both crystal facet of (0-11) and (011) with three different zoom-in value further demonstrate that no point defect, clear dislocation, and quantum wire/dot are presented in our MBE sample.

Fig. 3.12. The schematic diagram for explaining mechanism of pseudomorphic layer (a) compressive strain, (b) tensile strain.

Table 3-1 Growth rate of possible pseudomorphic strained alloys.

Notations: GRInAs@InP: Effective InAs growth rate on InP; ●: To indicate which alloys are presented in laser/SOAs structure; superscript “3” means cell-1 and cell-2 are used simultaneously. Calibration: GRInAs@InP = GRIn1

Ga1

Fig. 3.13. The X-ray diffraction rocking measurement and simulation result of 1.55 μm SOA/laser structure (MD3QW).

Fig. 3.14. The cross-section HR-TEM pictures of TE-1.55 μm SOA/laser MBE sample at crystal facet of (0-11) and (011); (a), (b): 430,000X; (c), (d): 1450,000X; (e), (f): 2850,000X

Reference

[3.1] R. Sacks, R. Sieg, S Ringel, “Investigation of the Accuracy of Pyrometric Interferometry in Determining AlxGa1-xAs Growth Rates and Compositions,” J. Vac.

Sci. Tech. B, vol. 12, pp. 2157-2162, 1996.

[3.2] P. Pinsukanjana, A. Jackson, J. Tofte, K. Maranowski, S. Campbell, J. English, S.

Chalmers, L. Coldren, and A. Gossard, “Real-time Simultaneous Optical-Based Flux Monitoring of Al, Ga, and In using Atomic Absorption for Molecular Beam Epitaxy,”

J. Vac. Sci. Tech. B, vol. 14, pp.2147-2150, 1996.

[3.3] W. Gilmore III, D. Aspnes, “Performance Capabilities of Reflectometers and Ellipsometers for Compositional Analysis during AlxGa1-xAs Epitaxy,” Appl. Phys.

Lett., vol. 66, pp.1617-1619, 1995.

[3.4] C. Kuo, M Boonzaayer, D. Schreder, G. Maracas, B. Jons, “Real Time in-situ Thickness Control of Fabry-Perot Cavities in MBE by Wavelength Ellipsometry,”

Ninth International Conference on Molecular Beam Epitaxy, Malibu, Ca., 1996.

[3.5] W. Breiland, K. Kileen, “A Virtual Interface Method for Extracting Growth Rates and High Temperature Optical Constants from Thin Semiconductor Films using in situ Normal Incidence Reflectance,” J. Appl. Phys., vol. 78, pp. 6726-6736, 1995.

[3.6] G. Li, W. Yuen, K. Toh, L. Eng, S. Lim, C. Chang-Hasnain, “Accurate Molecular Beam Epitaxial growth of Vertical-Cavity Surface Emitting Laser Using Diode Laser Reflectometry,” IEEE Photonics Tech. Letters, vol. 7, pp. 971-973, 1995.

[3.7] J. Harris, B. Joyce, P. Dobson, “Oscillations in the Surface Structure of Sn-Doped GaAs during Growth by MBE,” Surf. Sci., vol. 103, pp.L90-L96, 1981.

[3.8] J. M. Hove, C. S. Lent, P. R. Pukite, and P. I. Cohen, “Damped oscillations in reflection high energy electron diffraction during GaAs MBE”, J. Vac. Sci. Technol.

B, vol. 1, pp. 741-746, 1983.

[3.9] Dieter K. Schroder, Semiconductor Material and Device Characterization (2nd, John Wiley and Sons, New York, 1998), chapter 8.

[3.10] A.Y. Cho, in The Technology and Physics of Molecular Beam Epitaxy, edited by E.

H. C. Parker, Plenum Press, New York, pp. 1-13, 1985.

[3.11] Y. G. Chai, “Effect of accelerated growth rate (1–5 μm/h) on molecular beam epitaxial GaAs using Si as a dopant,” Appl. Phys. Lett., vol. 37, pp. 379-382, 1980.

[3.12] M. Ilegems, “Beryllium doping and diffusion in molecular-beam epitaxy of GaAs and AlxGa1–xAs,” Journal of Applied Physics, vol. 48, pp. 1278-1287, 1977.

[3.13] C. Hilsum, “Simple empirical relationship between mobility and carrier concentration,” Electron Lett., vol. 10, pp. 259-260, 1974.

Chapter 4

InGaAlAs/InP Strain-Balanced MQWs Laser/SOA’s

In first, the 1.55-μm MD3QW SOA/laser MBE sample is characterized by PL spectra.

With helping by mesa diodes, their carrier confinement and interband transition energy are studied by room-temperature electroluminescence (EL) and photocurrent spectroscopy.

Besides, two kinds of ridge-waveguide lasers are fabricated. The first is Fabry-Perot (FP) lasers with as-cleaved facets, and the second is a tilted-end-facet (TEF) type with 7-degree tilted waveguide to reduce the reflections at two end facets [4.1].

Fig. 4.1. The top view picture of a 250-μm diameter mesa diode.

4.1 Mesa diode

As shown in Fig. 4.1, at the top of each mesa, a 250-μm diameter window containing a rectangular array of shallow trenches is formed on the p-type surface to facilitate optical transmission with reduced Fabry-Perot interference. The metal electrodes of Cr/Au are simultaneously deposited on the p-type InGaAs contact layer and the n⁺-InP substrate to form an ohmic contact for the mesa diode. The detailed process has been presented in [4.2].

4.2 Ridge waveguide fabrication

Because the thick p-cladding (InAlAs) layer is not easy attacked by ICP-RIE dry etching, hence, we develop a new multi-step wet-etching process to obtain smooth and nearly vertical sidewalls with controlled undercut, less surface damages, and roughness to minimize carrier recombination and optical scattering loss.

4.2.1 Multi-step wet-etching process

A 280-nm-thick SiO₂ film is first deposited on the wafer by PECVD. Then, a Ti/Ni (240-nm/45-nm) double layer is sequentially deposited and patterned by a liftoff process based on a double-layer photoresist (PR) method as given in section 4.2.2. The resulting SiO₂-Ti-Ni sandwich etch mask is schematically shown in Fig. 4.2(a). To obtain 2.2-μm wide ridge-waveguides, we use a 3.2-μm-wide mask to allow for undercutting during the wet etching process. The etching HBr-solution (HBr:HCl:H2O2:H2O = 5:4:1:70) used in our multi-step wet-etching process is essentially non-selective with respect to InAlAs, InGaAs and InP. The first step of etching reaches a depth of about 0.35-μm with an undercut of 0.35-μm as shown in Fig. 4.2(b). A new layer of PR is applied, flood exposed, and developed to leave the undercut region filled with PR as shown in Fig. 4.2(c). A repetition of the above procedure results in the second undercut step with a slight upward etch as shown in Fig. 4.2(d). This self-aligned process is repeated four times to reach a total

etching-depth of approximately 1.4-μm as is depicted in Fig. 4.2(e). A final dip of the sample in the same HBr-solution changes the scalloped sidewalls to a smoothed ridge profile as is depicted in Fig. 4.2(f). The etching times are adjusted to obtain a final ridge width of ~ 2.2-μm and a final etching-depth of ~ 1.79-μm. In Fig. 4.3(a) and (b), the cross-section SEM pictures separately show a sidewall profile of ridge waveguide after four times wet-etching process and after smoothing process. As see in Fig. 4.3, the new developed multi-step wet-etching process can handle the ridge shape having less undercut and keep the ridge sidewall smooth.

Fig. 4.2. Schematically diagram for explaining multi-step wet-etching process.

Fig. 4.3. The cross-section SEM pictures: (a) sidewall profile after four times etching, (b) after smoothing process.

4.2.2 Double-layer photoresist method

(1) To spin coat PR adhesive layer (HMDS) at speed 5000 rpm for 30s.

(2) To spin coat first PR layer (OCG 825) at speed 5000 rpm for 30s.

(3) Soft bake at hotplate (95^oC) for 1min.

(4) Flood UV exposing for 1s.

(5) To spin coat second PR layer (AZ 4210) at speed 5000 rpm for 30s.

(6) Soft bake at hotplate (95^oC) for 1min.

(7) Remove most outer PR by UV exposing for 90s, developing in (AZ 400k : H2O = 1 : 4) for 1min, and baking (95^oC) for 30s.

(Note: baking PR at 95^oC instead of higher than 100^oC is for avoiding bubble-blowing situation.)

Once the ridge waveguide is formed and corresponding to our design, the laser diode is then completely facilitated by the following sequent procedures.

(8) Lift-off sandwich etch mask (SiO₂-Ti-Ni) by solution of HF : H₂O = 1 : 10 in ultrasonic machine for 1 min.

(9) Deposit 250-nm SiO₂ by PECVD at 250^oC.

(10) Planarization process

Due to the ridge height is ~ 2 μm, five polymide layers are necessary to reduce the unflatness within 80-nm.

a. Spin coat VM651 (0.1%) at speed 4000rpm for 30s after the VM651 is dropped and stable for 20s.

b. Baking (120^oC) for 1 min.

c. Spin coat PI2562 at speed 4000rpm for 30s.

d. To solidify polymide by the following suggestive curling heating procedure,

20 200

15 mins

350

60 mins, 30 sccm N 20

e. Repeat a → d four times.

f. To bombard polymide until SiO₂ is exposed to the ridge top by RIE O₂ plasmain condition of O2 flow = 20 sccm and RF power = 60W as demonstrated in Fig. 4.4.

Fig. 4.4. The cross-section SEM picture after planarization process.

(11) Define the windows (width = 6 μm) for p-contact region and remove the left SiO2

inside those windows by solution of HF : H2O = 1 : 10.

(12) Define the metal pad windows for p-metal and probe.

(13) After cleaning the unwanted survived oxide with the solution of NH3 : H2O = 1 : 20 for 10s, p-type metal (Cr/Zn/Cr/Au = 3.5/5/3.5/600 nm) is deposited by thermal evaporator.

(14) After p-metal is completely lift-off by ACE, those polymide without p-metal protection are following removed by O2 plasma.

(15) After the wafer backside is polished to 100 μm thick, the n-type metal (Au-Ge/Au = 10/200 nm) is then deposited by thermal evaporator.

(16) Finally, the ohmic contact of samples are formed by RTA treatment with Forming gas (H2/N2 = 15%/85%) in 360^oC for 1 min.

The top-view pictures in Fig. 4.5 captured by optical microscope (OM) show the completive FP- and TEF-type ridge waveguide lasers.

Fig. 4.5. The top-view optical microscope pictures for ridge waveguides: (a) Fabry-Perot type, (b) Tilted-end-facet type.

4.3 Devices: experimental results and discussions

Room-temperature (RT) PL spectrum for the 1.55-μm sample of “MD3QW” is shown in Fig. 4.6(a). The PL emission peak is located at λ = 1540 nm with a full-width at half maximum (FWHM) of 100 nm (1468-1568, nm). The emission peak is close to the calculated e1-hh1 transition at λ = 1550 nm. In Fig. 4.6(b), the FWHM of EL spectra is increased from 118 nm to 178 nm (1410-1588, nm) as injection current increased from 10mA (current density ~ 20 A-cm^-2) to 100mA (200 A-cm^-2). The RT-EL spectra were measured from the top of the mesa diode. Therefore, the EL emissions are TE-polarized dominated by the transitions between electron subband and heavy-hole subband. According to our calculation, the EL emission peaks correspond to e1-hh2 (λ = 1460 nm) and e1-hh1 (λ = 1550 nm) transitions within the QWs. The spontaneous emission intensity from e1-hh2 transition becomes stronger than the e1-hh1 transition when injecting current density above 20 A-cm^-2. The higher optical gain of e1-hh2 transition is explained by the increase of spontaneous emission rate for band filling effect in modulation-doped QWs [4.3], [4.4].

On the other hand, once the ridge-waveguide laser diodes are fabricated, in first, they should be qualified by current vs. voltage (I-V) measurement as demonstrated in Fig. 4.7.

Comparing with the ideal diode, the real diode indeed has few leakage currents from the diode junction at low forward applied biases. Also, for large forward biases, there is a saturation effect as shown by the dashed line. This is due to ohmic losses from the finite resistance of the neutral n- and p-regions [4.5]. Generally speaking, in condition of serious resistance (R) < 10 Ω and slope factor (nf) < 2, diode performance is good enough to us for its further light emission test. The ridge-waveguide lasers are measured under continuous-wave (CW) operation at 20^oC controlled by a thermoelectric (TE) cooler. Fig.

4.8(a) and (b) separately shows the Light vs. current (L-I) characteristics and the lasing spectra for the FP-laser of waveguide width = 2.2 μm and length = 930 μm. The threshold current (I ) is ~ 60 mA, and the slope efficiency is ~ 47 μW/mA per facet. No kinks are

observed in the L-I curve. The laser spectra in Fig. 4.8(b) show an intriguing emission performance. At I = 70 mA, the laser spectrum shows a single lasing peak at λ = 1514 nm.

As we increase I = 100 mA, in addition to the main peak at λ = 1514 nm, a second lasing peak at λ = 1528 nm is observed. Further increasing the injection current up to I = 130 mA, the laser spectrum exhibits three peaks at 1514, 1528 and 1545 nm.

The laser performance of the TEF-laser is shown in Fig. 4.9. In comparison with the FP-laser ones, the I_th for the TEF laser is increased to 100 mA. And, the slope efficiency is ~ 42 μW/mA per facet. The increase of Ith and the decrease of slope efficiency are caused by the reduced mirror reflection for the tilted facet of θ = 7^o. The TEF-laser spectra at injection current above threshold are shown in Fig. 4.9(b). However, the TEF-laser maintains only single lasing peak operation at λ = 1511 nm, even that the injection current is up to 120 mA.

The FP laser performance has two notable signatures: (1) lasing starts at excited energy state of λ = 1514 nm; (2) lasing peaks sequentially appear from λ = 1514 nm, 1528 nm, to 1545 nm as injection current increases. These transitions are not either predicted by the subband calculation for the triple QWs structure, or identified in the RT-PL and RT-EL spectra.

To discuss how possibilities could cause this three unusual peaks emissions from FP-laser and why they start from the shorter wavelength, following discussions are given:

1) We use a new quantum mechanics tool, named as nextnano³ [4.6], to calculate whole laser structure in one-dimension with considering n-type doping inside the tensile-strained InGaAlAs barrier. Also, Schrödinger-Poisson equations, current continuity equations, and the current relations for electrons and holes are involved. When bias = 1.0 volt and current density = 3 A-cm^-2, the simulation results show the band diagram has already been in flat-band condition. It also shows the quasi-Fermi level of electron (Efn) is already 239meV above e1, however, the quasi-Fermi level of hole (E_fp) is still below the hh1-state about 3meV. Moreover, according to our voltage-current measurement of this

FP-laser diode, bias over 1.4 volt could be possible to offer the injection current density higher than threshold current density (Jth ~ 3K A-cm^-2). To say simply, the current density for lasing at J_th should be 1,000 times higher than that existing in flat-band condition. In order words, the electron-1 subband has been totally population inversions at its threshold.

It also means that optical gain will be dominated at how many holes are in condition of population inversions. Therefore, we deny to concern these multi-wavelength stimulated emissions could be possible caused by operating QWs under different and non-uniform built-in electrical field owing to n-type MD.

2) Then, we further consider the coupling effect between QW’s if possible to cause a splitting between electron- and heavy-hole subband higher than 7.5meV (1514-1528, nm) or 8.9meV (1528-1545, nm). In the case of two-period QW’s separated by 8.6-nm tensile-strained InGaAlAs barrier, the calculation results show they exist very weak coupling. And, it induces only 0.6 - 0.7meV separation between the transitions of e1^even-hh1^even and e1^odd-hh1^odd. Again, we don’t have to take coupling-effect into account.

3) When we further look back on the materials what we grew and the growth process, these thin compressive/tensile strain InGa(Al)As alloys could be possible to form and mix localized states like as quantum wires/dots due to the relative low growth temperature and In-Ga segregation/conjugation. To investigate the detailed electron-hole transitions in the epi-structure, photocurrent spectroscopy was carried out. The photocurrent spectrum at zero bias is shown in Fig. 4.10. The spectrum shows δ-like absorption peaks at λ = 1514 nm, 1528 nm, 1545 nm, and 1560 nm. Normally photocurrent spectrum features the joint density of states of the transition levels. Step-like signals are expected for absorption transitions of quantum well states, while the δ-like absorption peaks correspond to transition states of low dimensionality less than 2. The agreement of the lasing wavelengths with the δ-like absorption peaks suggests that the transition states are confined in quantum dots and/or quantum wires formed by the strain-field profile and the InGaAlAs alloy

segregation/migration within the QWs. Moreover, lasing operation at excited transitions of high modal gain have been reported for quantum-dots and quantum-wires lasers [4.7], [4.8].

The laser performances are consistent with the quantum dots/wires lasers. The energy separation for the lasing peaks is about 7-9 meV, which is also consistent with the calculated hole level separation for InGaAs quantum dots [4.7]. For the quantum structure of n-type modulation doping, the lasing peaks are controlled by the level occupation of excess holes.

The excess injected holes thermally relax to the higher level in the quantum dots. Since within the quantum dots, hole thermal relaxation process of the excited state is dominated by inelastic multi-phonon scattering, and presents a relaxation bottleneck to populate the underlying states [4.9], [4.10]. The relaxation time takes a few of 10 ps [4.10], which is longer than the stimulated radiative recombination time. At injection current above threshold condition, lasing transition at high energy state is observed. As the injection current increased, more excess holes relax to the next lower level, and the second lasing peak appears in addition to the main peak. Further increasing the injection current, the process goes on as long as there exists underlying hole levels. For our case, we have observed three peaks at I = 130 mA for the FP laser. From the photocurrent spectrum in Fig.

4.10, we would expect that the fourth lasing peak at λ = 1560 nm (7.7 meV separation to peak at λ = 1545 nm) will expose at higher injection current. However, the wavelength-dependent modal gain distribution of the diode laser can hinder the process as shown in Fig. 4.9 for the TEF-laser operating at the excited transition λ = 1511 nm.

Fig. 4.6. (a) PL spectra, and (b) EL spectra for sample “MD3QW” at room temperature.

Fig. 4.7. The characteristics of current density vs. applied forward biases (J-V).

(Website of nextnano³: http://www.wsi.tum.de/nextnano3/index.htm)

Fig. 4.8. FP-type ridge laser: (a) L-I characteristics, and (b) lasing spectra.

Fig. 4.9. TEF-type ridge laser: (a) L-I characteristics, and (b) lasing spectra.

Fig. 4.10. Photocurrent spectrum measured at V = 0.

Reference

[4.1] D. Marcuse, “Reflection loss of laser mode from tilted end mirror,” J. Lightw.

Technol., vol. 7, pp. 336-339, 1989.

[4.2] H. P. Fan, “Photocurrent and Electroabsorption spectroscopy for Semiconductor Quantum Well structures,” Master thesis, Inst. of Electro-optical Engineering, SYSU, 2001.

[4.3] A. Niwa, T. Ohtoshi, K. Uomi, and K. Nakahara, “Doping-type dependence of turn-on delay time in 1.3　m GaAsP-InP modulation-doped strained quantum-well lasers,” IEEE Photon. Technol. Lett., vol. 8, pp. 328- 330, 1996.

[4.4] V. D. Kulakovskii, E. Lach, and A. Forchel, “Band-gap renormalization and band-filling effects in a homogeneous electron-hole plasma in In0.53Ga0.47As/InP single quantum wells,” Phys. Rev. B, vol. 40, pp. 8087-8090, 1989.

[4.5] P. Bhattacharya, Semiconductor Optoelectronic Devices (2nd, Prentice-Hall, New Jersey, 1994), pp.175.

[4.6] (Website of nextnano3: http://www.wsi.tum.de/nextnano3/index.htm)

[4.7] G. Park, O. B. Shchekin, and D. G. Deppe, “Temperature dependence of gain saturation in multilevel quantum dotlasers,” IEEE J. Quantum Electron., vol. 36, pp.

1065-1071, 2000.

[4.8] E. Kapon, D.M. Hwang and R. Bhat, “Stimulated emission in semiconductor quantum wire heterostructures,” Phys. Rev. Lett. vol. 63, pp. 430-433, 1989.

[4.9] U. Bockelmann, and G. Bastard, “Phonon scattering and energy relaxation in two-, one-, and zero-dimensional electron gases,” Phys. Rev. B, vol. 42, pp. 8947-8951, 1990.

[4.10] R. Heitz, M. Veit, N. N. Ledentsov, A. Hoffman, D. Bimberg, V. M. Ustinov, P. S.

Kop’ev, and Zh. I. Alferov, “Energy relaxation by multiphonon processes in InAs/GaAs quantum dots,” Phys. Rev. B, vol. 56, pp. 10435-10445, 1997.

Chapter 5

Electro-Absorption Characteristics in N-type Modulation-Doped Strain-balanced MQWs

Because the QWI process is not really presented to blue shift our λ = 1.55 μm samples in this study, six additional 1.48 μm samples have been designed and grown to simulate a 70-nm blue-shifted material without actually carrying out the QWI process. In fact, these samples with different n-type modulation-doped (MD) distribution are designed with structure quite similar to that of 1.55 μm one but with reduced layer thickness inside the well to shift the transition wavelength (e1-hh1) to 1.48 μm. These wafers are used to study the relations between Δn, differential absorption (Δα), and MD distribution.

5.1 Six blue-shifted samples (λ = 1.48 μm)

According to the design concept of MD3QW, one period of the blue-shifted

According to the design concept of MD3QW, one period of the blue-shifted