Semi-insulating InPFe for buried-heterostructure

Journal of Crystal Growth ] (]]]]) ]]]–]]]

Contents lists available at SciVerse ScienceDirect

Journal of Crystal Growth journal homepage: www.elsevier.com/locate/jcrysgro

Semi-insulating InP:Fe for buried-heterostructure strain-compensated quantum-cascade lasers grown by gas-source molecular-beam epitaxy M.P. Semtsiv n, A. Aleksandrova, M. Elagin, G. Monastyrskyi, J.-F. Kischkat, Y.V. Flores, W.T. Masselink Department of Physics, Humboldt University Berlin, Newtonstrasse 15, D-12489 Berlin, Germany

a r t i c l e i n f o

Keywords: A3. Molecular-beam epitaxy B2. Semiconducting III–V materials B3. Heterojunction semiconductor devices B3. Infrared devices

abstract We describe the realization of buried-heterostructure strain-compensated quantum-cascade lasers that incorporate a very high degree of internal strain and are grown on InP substrates using gas-source molecular-beam epitaxy (GSMBE). The active region of the lasers contains AlAs layers up to 1.6 nm thick with 3.7% tensile strain; restricting any post-growth processing to temperatures below 600 1C to avoid relaxation. We demonstrate that buried-heterostructure devices can be realized by using GSMBE to over-grow the etched laser ridge with insulating InP:Fe at temperatures low enough to preserve the crystal quality of the strain-compensated active region. Two distinct growth techniques are described, both leading to successful device realization: selective regrowth at 550 1C and non-selective regrowth at 470 1C. The resulting buried-heterostructure lasers are compared to a reference laser from the same wafer, but with SiO2 insulation; all three have very similar threshold current densities, operational thermal stability, and waveguide losses. & 2013 Elsevier B.V. All rights reserved.

1. Introduction Since their invention in 1994, quantum-cascade lasers (QCLs), [1] have become the ultimate semiconductor laser sources from near-infrared [2] to middle THz [3] spectral range. QCLs are widely used now as sources of infrared laser emission for spectroscopy and chemical sensing. They also have the potential for application in military countermeasures and free-space communication. Several applications, for example photo-acoustic spectroscopy, require high average output power, which in turn requires efficient heat sinking out of the laser ridge. There has been enormous progress in QCL technology towards the optimization of the heat extraction and the best results are obtained by means of buriedheterostructure (BH) laser fabrication using semi-insulating InP:Fe overgrown by metal-organic vapor phase epitaxy (MOVPE) [4–8]. A successful QCL design for wavelengths between 3 and 5 mm is based on strain compensation with very high degrees of internal strain, including pure 1–2 nm AlAs barriers with 3.7% tensile strain to the InP substrate [9–12]; these structures, however do not withstand the typical MOVPE regrowth temperature of about 650 1C [13]. Thus, the growth of high-resistivity InP at low growth temperatures is required for these strain-compensated QCLs. Gas-source molecular-beam epitaxy (GSMBE) has already been used before to fabricate BH telecom lasers [14] and also BH QCLs with a lattice-matched active zone [13]. However, extending the GSMBE overgrowth technique to strain-compensated QCLs with

n

Corresponding author. Tel.: þ49 30 20937919. E-mail address: [email protected] (M.P. Semtsiv).

highly strained alternating In0.73Ga0.27As and AlAs layers is not trivial and has not been demonstrated yet. Contrary to the MOVPE technique, GSMBE leaves the etched side walls of the laser ridge half a time without the group-V flux at elevated temperatures of 500–550 1C before the regrowth process is initiated. The situation arises because with a single gas injector and conventional sample rotation, the ridge shades each etched sidewall from the sidecoming surface-stabilizing group-V flux during the half of the sample rotation. This is problematic because it can lead to material intermixing on the sidewalls due to strain-driven surface migration, which can cause electrical shorts. In this paper we describe the successful InP:Fe regrowth of strain-compensated QCLs that include highly strained and relatively thick (up to 1.6 nm) AlAs barriers in the active zone. The regrowth can be done selectively around a dielectric mask, similar to the technique typically used with MOVPE, but at a significantly lower growth temperature of 550 1C. These BH QCLs, when compared to the reference QCLs with SiO2 insulation of the ridges, show comparable threshold current densities, Jth, and T0; the confinement factor, G, is somewhat lower due to the smaller index discontinuity, and the waveguide losses, aw , are somewhat lower due to a superior interface.

2. QCL active region design The conduction band edge profile of the single active region cascade and the moduli square of the confined states, wavefunctions are shown in Fig. 1. The active region is designed to emit at the wavelength close to 4 mm at an operation voltage of approximately 13 V for 30 cascades. High values of T0 and T1 require the use of high

0022-0248/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jcrysgro.2012.12.079

Please cite this article as: M.P. Semtsiv, et al., Journal of Crystal Growth (2013), http://dx.doi.org/10.1016/j.jcrysgro.2012.12.079i

2

M.P. Semtsiv et al. / Journal of Crystal Growth ] (]]]]) ]]]–]]]

Wavelength (μm)

1

5

1

0

40

50 z (nm)

60

4

3.5

Intensity (arb.u.)

Energy (eV)

2

4.5

70

Fig. 1. Conduction band diagram and probability functions calculated within a single period of the active region with a 100 kV/cm electric field. The moduli square of the wavefunctions (1 and 2) responsible for the laser transition is drawn with the thick lines. The design emission wavelength is approximately 4 mm at an operation voltage of 13 V for 30 cascades.

barriers within the active region to minimize the thermally activated carrier leakage from the upper laser state 2. For that purpose, we use pure AlAs with very high tensile strain (3.7%) that are grown thinner than the critical thickness for coherent growth on InP. The net strain across the single cascade is compensated by compressively strained In0.7Ga0.3As wells. The lattice-matched InAlAs layers add to the AlAs barriers and allow to tune the tunneling probability across the structure without affecting the strain. The layer thicknesses in nm from left to right, beginning with the thickest injection barrier are 3:0=0:9=1:1=0:2=4:0=1:2=3:3=1:6=3:0=1:4=2:7=1:2=2:4=1:0=2:3=0:9 =2:0=1:0=0:9=1:9=2:0=0:9=1:6. The AlAs barriers are in bold, the In0.52Al0.48As barrier material layers are in bold italic, and the In0.73Ga0.27As well layers are in roman. Underlined layers are Si-doped to 4 1018 cm 3, which results in an electron sheet density of approximately 2 1012 cm 2 per cascade or 5 1017 cm 3 in terms of average doping across the cascade. This particular design uses the thick InAlAs barrier next to the thinnest quantum well instead of the thin and high AlAs barrier, which lowers the interface scattering [12] in the transition region, evidenced by narrow electroluminescence spectrum (Fig. 2).

3. Sample fabrication Both the primary growth of the laser structure and the second growth (regrowth) for the BH part of the process are carried out in a Riber Compact 21 T GSMBE system, using solid sources for In, Ga, and Al; As and P are supplied by arsine and phosphine that are precracked at 920 1C. Both the Si doping for the active region as well as the Fe doping for the regrown insulator layers are supplied from solid sources. The growth chamber is pumped using a 1600-l/s turbo pump. The laser structure was grown on a low-doped (n ¼ 2 1017 cm3 ) InP:S substrate, which serves as the lower cladding layer. The epitaxy sequence consists of 100 nm InP:Si (n ¼ 1 1017 cm3 ); 200 nm of lattice matched InGaAs:Si spacer (n ¼ 7 1016 cm3 ); then the 1:3 mm (total thickness), 30-period active region; 200 nm of lattice matched InGaAs:Si spacer (n ¼ 7 1016 cm3 ); then a 2:5 mm (n ¼ 1 1017 cm3 ) InP:Si plus 1 mm (n ¼ 3 1018 cm3 ) InP:Si top cladding; and 100 nm of lattice matched InGaAs:Si (n ¼ 8 1018 cm3 ) contact layer.

2000

3

x0.05

2500

3000

Wavenumber (cm-1) Fig. 2. Development of the emission spectrum of the BH QCL between 1.8 and 3.0 A in pulsed mode at room temperature. The half-width of the low-current electroluminescence spectrum is equal to 29 meV.

One reference laser and two BH lasers were fabricated from the same wafer. The reference QCL was processed into approximately 25-mm-wide deep-etched ridges using optical lithography and wet-chemical etching. The side-wall insulation of the reference QCLs was accomplished by sputtering of a 500 nm thick SiO2 layer. The two BH QCLs were first wet-etched into stripes of various widths (approximately 10, 14, 20, 29, and 49 mm) then insulated by overgrowth of nominally 6 mm InP:Fe. One BH QCL (BH470) was overgrown non-selectively at 470 1C with subsequent removal of the overgrown material on the top of the stripes to facilitate electrical contacts. The second BH QCL (BH550) was overgrown selectively at 550 1C, using a dielectric mask to define the growth area and with subsequent lift-off of the mask in buffered HF. Thermally evaporated Au/Cr is used for ohmic contacts. Selective regrowth at 550 1C significantly simplifies the further fabrication process and represents an advantageous way to produce buried-heterostructure QCLs compared to the non-selective regrowth at 470 1C. In order to avoid electrical shorts due to surface migration of the atoms at the etched laser side walls, regrowth in sample BH550 was initiated as soon as the temperature had reached the set-point of 550 1C, minimizing the time of detrimental exposure. This step ended up being successful. Fig. 3 shows the scanning electron microscope image of the cleaved facet of the 14 mm-wide laser ridge of sample BH550. Sample BH550 was processed with metallic contacts evaporated directly onto the InP:Fe and the top of the laser ridge. Sample BH470 includes a 500 nm-thick top SiO2 insulating layer with the contact windows opened above the laser ridge. Including this top electrical insulation has significantly improved the yield of defect-free lasers across the wafer. The absence of the insulating SiO2 layer in sample BH550 allows a direct characterization of the insulating properties of the overgrown InP:Fe material. The electrical resistance of the InP:Fe insulation layer (sample BH550) cleaved next to the laser ridge (the bond pad itself) was measured as a function of temperature. Fig. 4 shows the resistance of the 0.2 mm2 pad as a function of reciprocal temperature. The room-temperature resistance gives a resistivity of roughly 0:2 108 O cm, similar to what was measured in Ref. [13]. A fit of the data gives a thermal activation energy of 680 meV. The value of the extracted activation energy is about half of the InP band gap, consistent with the Fe deep acceptor level that results in the insulating properties of the



3

ref. 1.0mm ref. 4.3mm BH470 BH550

Jth (kA/cm2)

10

1

Fig. 3. Scanning electron microscope image of the laser ridge facet (sample BH550), showing the overgrown InP:Fe layers with the gap close to the top of the laser stripe due to the shadowing effect of dielectric mask. Dashes highlight the interfaces between the various layers as indicated.

0

50

αw=

fit

2.9cm

-1

gΓ= 4.0cm/kA

3.0cm-1 4.2cm/kA

3 Jth (kA/cm2)

Resistance (Ω)

300

BH550

BH470

exp.

Cr/Au InP:Fe 10

250

Fig. 5. Temperature dependence of the threshold current density for two reference stripes (4.3 mm-long and 1 mm-long) and two buried stripes overgrown at different temperatures (2.9 mm-long stripe of BH470 and 2.1 mm-long stripe of BH550).

4

106

100 150 200 Temperature (K)

ref. αw=3.5cm-1

2

5

gΓ=4.7cm/kA

InP:S substr.

2.4

2.6

2.8

3.0

1000/T (K-1) Fig. 4. Resistance of the insulating layer (sample BH550) cleaved from near the laser ridge as a function of the reciprocal measurement temperature. The straight line is an exponential fit with the activation energy of 680 meV.

regrown InP:Fe material. The lower laser yield on the BH550 wafer appears to be due to local growth defects, the nature of which needs some further clarification.

4. Laser performance Fig. 5 shows the temperature dependence of the threshold current density in logarithmic scale for two reference laser stripes and two BH laser stripes that were overgrown at different temperatures. The data show a comparable slope in logarithmic scale. The data fit with J th ¼ J 0 expðT=T 0 Þ below 6 kA/cm2 gives comparable T0 parameter in the range of 170–220 K. The 6 kA/ cm2 current density is also the approximate position of the rollover in power–current characteristics and is defined by the active zone design and its doping level. Above 6 kA/cm2 the threshold grows super-exponentially due to misalignment of the confined states (Fig. 1). Altogether, all three structures (reference, BH470, and BH550) show approximately the same dependence of the threshold current density on temperature. We do not observe any

1

0

2

4

6

8

10

1/L (cm-1) Fig. 6. Pulsed threshold current density measured at 80 K vs. reciprocal laser length compared for the two BH and the reference QCLs. A linear fit of the data (dashed lines) indicates a slightly lower optical confinement factor for the BH lasers, but also somewhat lower waveguide losses.

additional temperature-activated processes due, for example, to leakage current through the overgrown InP:Fe material in either BH lasers, regardless of overgrowth technique and temperature. The values of threshold current density scale as usual with reciprocal stripe length. A more detailed comparison of the BH and reference QCLs is made in terms of waveguide losses and the mode confinement. Fig. 6 depicts the threshold current density as a function of reciprocal laser length, comparing the two BH QCLs and the reference QCL. The value of the threshold current density is taken to be Jth ¼ aw =g G þ am =g G, where aw is the waveguide loss coefficient, am is the mirror loss coefficient, and g G is the modal gain coefficient. The mirror loss coefficient for the laser ridge with both uncoated facets aw ¼ ð1=LÞlnðRÞ, where L is the laser length, R ¼ ððneff 1Þ= ðneff þ1ÞÞ2 is the facet reflectivity, and neff 3:25 is the averaged refractive index of the waveguide giving lnðRÞ 1:27. First, both BH and reference lasers are extremely similar for a given stripe length. A closer look indicates that the modal gain


4


coefficient of the BH lasers is about 15% lower than the reference lasers. Of course, a somewhat lower confinement factor G is expected in BH laser due to the smaller lateral-direction discontinuity in refractive index. The gain coefficient g should be the same in all of the samples investigated since the active regions of the BH and reference QCLs were all from the same growth run on the same wafer. The measured result suggests that there is no degradation of g during regrowth and the small difference in g G is due entirely to a somewhat smaller value of G. The waveguide loss aw is also about 13% lower in the BH lasers compared to the reference lasers. This result is also expected, as the thick InP:Fe insulator separates the optical mode very well from the lossy metal–dielectric interface. Altogether, the buriedheterostructure overgrowth of InP:Fe using low-temperature GSMBE does not appear to deteriorate either the active region in terms of gain or the waveguide quality, and actually reduces waveguide losses.

5. Conclusions To conclude, we have demonstrated successful fabrication of BH QCLs, in which the InP:Fe is regrown using GSMBE at temperatures well below 600 1C. The QCL active region is based on strain compensation and contains highly strained AlAs layers up to 1.6 nm thick. The semi-insulating InP:Fe material has resistivity in the range of 107 2108 O cm at room temperature, with a measured thermal activation energy characteristic of InP with mid-gap acceptor levels such as Fe. Selective regrowth of InP:Fe at 550 1C using GSMBE allows a convenient facilitation of the contact windows, as with MOVPE, but at much lower growth temperatures. The BH QCLs operate with comparable threshold current densities and T0, and somewhat lower waveguide losses compared to the reference QCL with dielectric insulation of the sidewalls. The possibility of low-temperature fabrication of the

BH QCLs is expected to improve the high-average-power results of short-wavelength (l 325 mm) QCLs using highly strained AlAs barriers in the active region.

Acknowledgments The research was supported by the German Research Society (DFG) and by the European Union through the Erasmus Mundus External Cooperation Window (Lot 7) program. References [1] J. Faist, F. Capasso, D.L. Sivco, C. Sirtori, A.L. Hutchinson, A.Y. Cho, Science 264 (1994) 553. [2] O. Cathabard, R. Teissier, J. Devenson, J.C. Moreno, A.N. Baranov, Applied Physics Letters 96 (2010) 141110. [3] C. Walther, M. Fischer, G. Scalari, R. Terazzi, N. Hoyler, J. Faist, Applied Physics Letters 91 (2007) 131122. ¨ [4] M. Troccoli, S. Corzine, D. Bour, J. Zhu, O. Assayag, L. Diehl, B.G. Lee, G. Hofler, F. Capasso, Electronics Letters 41 (2005) 1059. ¨ [5] L. Diehl, D. Bour, S. Corzine, J. Zhu, G. Hofler, M. Loncˇar, M. Troccoli, F. Capasso, Applied Physics Letters 88 (2006) 201115. ¨ [6] L. Diehl, D. Bour, S. Corzine, J. Zhu, G. Hofler, M. Loncˇar, M. Troccoli, F. Capasso, Applied Physics Letters 89 (2006) 081101. [7] A. Evans, S.R. Darvish, S. Slivken, J. Nguyen, Y. Bai, M. Razeghi, Applied Physics Letters 91 (2007) 071101. [8] Y. Bai, N. Bandyopadhyay, S. Tsao, S. Slivken, M. Razeghi, Applied Physics Letters 98 (2011) 181102. [9] M.P. Semtsiv, M. Wienold, S. Dressler, W.T. Masselink, Applied Physics Letters 90 (2007) 051111. [10] D.G. Revin, J.W. Cockburn, M.J. Steer, R.J. Airey, M. Hopkinson, A.B. Krysa, L.R. Wilson, S. Menzel, Applied Physics Letters 91 (2007) 051123. [11] A. Bismuto, M. Beck, J. Faist, Applied Physics Letters 98 (2011) 191104. [12] M.P. Semtsiv, Y. Flores, M. Chashnikova, G. Monastyrsk, W.T. Masselink, Applied Physics Letters 100 (2012) 163502. ¨ [13] M. Chashnikova, G. Monastyrskyi, A. Aleksandrova, M. Klinkmuller, M.P. Semtsiv, W.T. Masselink, Applied Physics Letters 100 (2012) 213504. [14] Ph. Pagnod-Rossiaux, M. Lambert, F. Gaborit, F. Brillouet, P. Garabedian, L. Le Gouezigou, Journal of Crystal Growth 120 (1992) 317.
