- Email: [email protected]
X-ray studies and time-resolved photoluminescence on optically pumped antimonide-based midinfrared type-II lasers
Spectrochimica Acta Part A 60 (2004) 3387–3392
X-ray studies and time-resolved photoluminescence on optically pumped antimonide-based midinfrared type-II lasers Carsten Schwender, Jan Oliver Drumm, Goetz Hoffmann, Birgit Vogelgesang1 , Henning Fouckhardt∗ Integrated Optics and Microoptics Research Group, Physics Department, Kaiserslautern University of Technology, Erwin-Schroedinger-Str. 46, D-67663 Kaiserslautern, Germany Received 26 September 2003; accepted 26 November 2003
Abstract We report on high-resolution X-ray diffraction and time-resolved photoluminescence (TR-PL) studies of antimonide-based midinfrared (MIR) type-II laser samples. A structural characterization taking into account asymmetrical strain, layer tilting, and relaxation enables an accurate determination of the average lattice constant of the active region and the composition of the cladding layers. By designing the antimonide-to-arsenide interfaces, we achieve exact lattice matching of the active region to the substrate. Non-radiative recombination processes are investigated with time-resolved photoluminescence. The samples are also characterized under optically pumped laser operation. By an examination of the time-integrated and time-resolved amplified spontaneous emission (TR-ASE), we investigate the modal gain and gain dynamics. The variable stripe length method is combined with the TR-PL approach. Compared to the time-integrated gain spectra the spectral dependence of the maximum and minimum time-resolved gain shows a broad plateau. The full width half maximum (FWHM) of the TR-ASE pulse is 5.5 ± 0.5 ps. Thus, short pulses in this range should be achievable upon laser operation. The active regions of the laser structures investigated here are promising subunits of type-II quantum cascade lasers. © 2004 Elsevier B.V. All rights reserved. Keywords: Midinfrared semiconductor lasers; Antimonide lasers; Type-II heterostructures; Non-radiative recombination; Shockley–Read–Hall; Auger
1. Introduction In the last decade there has been considerable progress in midinfrared semiconductor lasers (MIR, 2–5 m) [1–8]. Such devices are very useful for detection of solvents and trace gases such as hydrocarbons. One approach uses so-called “W-structures”, consisting of a symmetric type-II broken-gap quantum well double-heterostructure, e.g. InAs/Ga(In)Sb/InAs bounded by Al(As)Sb barrier layers [2,4,6,7]. Different suggestions have been made to improve these lasers. Designing the strain and the mixed anion interfacial bond stoichiometry between the single W-structure layers gives an opportunity to reduce non-radiative recombination processes [9–12]. For an excellent crystal quality, it is also necessary to avoid unintentional layer relaxation. ∗
Corresponding author. Tel.: +49-631-205-4145; fax: +49-631-205-4147. E-mail address: [email protected] (H. Fouckhardt). 1 Meanwhile with Robert Bosch GmbH, Stuttgart, Germany. 1386-1425/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.saa.2003.11.040
To evaluate the influence of the suggested changes on the laser properties, a detailed knowledge of the structural parameters of the samples is necessary. Our samples have a design similar to the lasers in [6] but with different W-period numbers and cladding layer types. In the present paper, an analysis of non-radiative recombination processes (Shockley–Read–Hall (SRH) and Auger (AR) with holes) as well as a detailed structural characterization of the laser structure is performed. Adjusting the composition of the mixed anion interfacial bonds allows for strain compensation of the active region. The laser properties, modal gain, and gain dynamics are examined.
2. Experiments and results The samples are grown by molecular beam epitaxy (MBE) on n-type GaSb substrates using a DCA 450 MBE-system. The active region of the laser samples consists of 70 periods of InAs/GaSb/InAs/AlSb (6 ML/7 ML/6 ML/11 ML)
3388
C. Schwender et al. / Spectrochimica Acta Part A 60 (2004) 3387–3392 Table 1 Thicknesses of single W-structures tW in laser samples determined by RHEED and HRXRD, FWHM bs of the first order HRXRD satellite reflexes
Fig. 1. (0 0 4) HRXRD rocking curves of sample 296 and 306.
surrounded by (each) 1.4 m thick AlAsSb cladding layers, which are intended to be lattice matched to the GaSb substrate. All samples are capped by a 5 nm thick GaSb layer. The layer thicknesses are monitored by using a reflective high-energy electron diffraction (RHEED) system equipped with an image processing tool. In order to use this in situ measurement technique the samples are not rotated during growth. The samples differ in interface composition only, which is controlled by the shutter sequences. The As for Sb exchange on Ga(Al)Sb surfaces can be greatly reduced by using In to tie Sb to the surface [13]. This also gives the opportunity to control the As for Sb exchange at the interface by varying the amount of deposited In. The sequence for the antimonide-to-arsenide interfaces is 10.0 s Sb soak for all interfaces and 4.0 s In deposition for samples 276, 277, 279, 290, and 296 and 2.8 s In deposition for sample 306 to adjust the average lattice constant of the active region with a mixed-interface stoichiometry. Detailed control of the interface stoichiometry is essential to obtain precise values for the average lattice constant as desired and thus to achieve strain compensation. Detailed information about the growth and the calibration of the growth parameters can be found in [14]. For a precise structural characterization of the samples, high-resolution X-ray diffractometry (HRXRD) measurements are performed by using a four crystal, high-resolution X-ray diffractometer. To probe growth quality and to determine the thickness of the W-structure periods in the active region, a HRXRD diffraction profile is measured at the (0 0 4) plane. As an example, Fig. 1 gives the rocking curves of samples 296 and 306 which show sharp satellite peaks up to the seventh order. This indication of high crystal quality is similar for all samples. The low full width half maximum (FWHM) bs of the first order satellite peaks, which is 167 arcsec for samples 296 and 140 arcsec for sample 306 also confirms high crystal quality. The values for all samples are shown in Table 1. The angular spacing of the satellite peaks gives the period thickness tW of the W-structures in the active region, which is intended to be 32 ML for all samples including 4 × 0.5 ML for the four interfacial bonds between
Sample
tW (ML) RHEED
tW (ML) HRXRD
bs (arcsec)
276 277 279 290 296 306
31 30.5 30.5 32 32 33
31.1 31.2 31.6 33.9 32.8 32.4
117 118 110 108 140 167
antimonide and arsenide layers. Table 1 shows a comparison of the thickness values of all samples determined with RHEED and with HRXRD. They are in good agreement with the target value. Besides the standard rocking curve measured at the (0 0 4) plane, eight rocking curves are taken from the four inclined crystal planes (1 1 5), (1¯ 1¯ 5), (1 1¯ 5), and (1¯ 1 5) in grazing incidence (+) and grazing exit (−) geometry. Due to the detailed analysis of the crystal structure it is possible to eliminate effects such as layer tilting and asymmetrical strain relaxation [15,16]. The HRXRD results for the inclined planes are examplified in Fig. 2 with the rocking curves measured at the (1 1 5) plane in grazing incidence (1 1 5(−) ) and grazing exit (1 1 5(+) ) geometry. From the measurements at the inclined planes the lattice parameters perpendicular to the surface d⊥ = d001 and the in-plane lattice parameters d|| of the different layers are extracted [15] whereby the value of d⊥ for the active region is only a mean value over the different layers of the periodic structure. With d⊥ and d|| the lattice constants of the layers are calculated taking the elastic constants into account. Applying Vegard’s law, the As fraction for the different AlAsSb layers varies from 10.3 to 12.0% for samples 276, 277, and 279 instead of the 8.4% as desired
Fig. 2. (1 1 5) HRXRD rocking curves (Θ is the diffraction angle relative to the angle of the substrate peak).
C. Schwender et al. / Spectrochimica Acta Part A 60 (2004) 3387–3392
to achieve lattice matching. After an optimization of the As/Sb flux ratio values between 7.7 and 10.0% were reached whereby the lattice mismatch of the cladding layers relative to GaSb is in the range of −0.13 to 0.05%. For the mean lattice constant of the active region of samples 276–296 finally values from 6.1061 to 6.1077 Å are achieved reproducibly. The active region of sample 306 has a mean lattice constant of 6.0959 Å (lattice match to GaSb) and so strain compensation of the active region by varying the mixed anion interfaces is clearly achieved. By comparing the expected mean lattice constants for arsenide- and antimonide-like interfaces with the values extracted from HRXRD measurements, the interface composition can be deduced. With an In deposition time of 4.0 s, the interface composition is 70–75% InSb-like. The reduced In deposition time of 2.8 s for sample 306 results in a reduction of the Sb-fraction in the interface to 53%. But the 20% higher As-fraction in these interfaces does not noticeably effect the crystal structure. To examine the non-radiative recombination processes in the laser devices, for samples 296 and 306 the coefficients of SRH and AR are investigated with a time-resolved photoluminescence up-conversion setup (TR-PL) [17,18]. This setup provides 1 ps temporal resolution in the midinfrared wavelength range with high sensitivity [19]. To determine the SRH and AR coefficients with TR-PL, the rate equation −
∂N = AN + BN2 + CN3 ∂t
(1)
is used, whereas A, B, and C are the SRH, radiative and AR coefficients, respectively. In our measurements, the carrier densities are in the range of (5–15)×1018 cm−3 where the AR term for high carrier densities is more complicated and has to be written as CN3 → C1 N + C2 N 2 + C3 N 3
(2)
which results in a modified rate equation: −
∂N = (A + C1 )N + (B + C2 )N 2 + C3 N 3 . ∂t
(3)
This means that even the coefficients of the linear and the quadratic term contain parts due to Auger recombination. The measurements for determination of the SRH coefficients are performed at 10 K and for the AR coefficient at 80, 120, and 200 K. The temperature and/or carrier density dependence of Auger and SRH coefficients are investigated for sample 306. The excellent growth quality of the devices is indicated by SRH coefficients at 10 K of (1.8 ± 0.6) × 108 s−1 for sample 296 and (2.2 ± 0.7) × 108 s−1 for sample 306. Fig. 3 shows that the cubic AR coefficient C3 for sample 306 decreases in a characteristic manner with increasing initially excited carrier density. To determine the AR coefficient at low carrier densities, i.e. C30 , we use a convergence equation according to [19] C3 (N0 ) =
C30 , 1 + (N0 /NC )p
(4)
3389
Fig. 3. Experimental cubic AR coefficient versus carrier density at temperature of 200 K. The decay of C3 (N0 ) was fitted by a convergence equation [19].
where C30 , NC , and p are fit parameters to data points C3 (N0 ). We achieved very low cubic AR coeffients C30 (valid at low carrier densities) of (2.3 ± 0.1) × 10−29 cm6 s−1 at 80 K, (2.5 ± 0.3) × 10−29 cm6 s−1 at 120 K, and (1.2 ± 2.6) × 10−28 cm6 s−1 at 200 K for this sample [19]. The experimentally determined AR coefficients are mainly due to CHHS Auger recombination. A more detailed report on these TR-PL studies is published elsewhere [19]. To test the laser properties of the grown structures 500 m long edge emitting laser devices are prepared and characterized. The devices are mounted into a heatable liquid nitrogen dewar and are optically pumped by a Nd:YAG laser operating either in continuous wave or pulsed radiation emission mode (200 ns, 50 kHz). The width of the pump stripe is 60 m. For the two different interface stoichiometries the laser results of the two samples 296 and 306 are presented. Fig. 4 shows the wavelength spectra of Fabry–Perot laser devices of samples 296 and 306 at a temperature of 83 K. For applications in MIR spectroscopy the lasers have to be operated with an external dispersive resonator in order to provide for single-longitudinal-mode operation. By optical pumping with a Nd:YAG laser source, the samples show cw laser emission at 83 K at wavelengths of 3.40 and 3.32 m, respectively. At this temperature, the highest maximum output power is measured for sample 306 with 14 mW per facet. The slope efficiency is 3.3%. The maximum working temperature for cw laser operation is 170 K. The characteristic temperatures at cw laser operation are T0 = 57.6 ± 2.9 K for sample 296 and T0 = 54.2 ± 2.3 K for sample 306. In pulsed laser operation sample 306 emits laser radiation up to 320 K; under these conditions we determined a characteristic temperature of T0 = 73.5 ± 6.9 K. These T0 values are among the best for MIR lasers with similar structures and similar emission wavelengths [6]. The laser emission results confirm the low non-radiative recombination coefficients and reveal excellent growth quality.
3390
C. Schwender et al. / Spectrochimica Acta Part A 60 (2004) 3387–3392
Fig. 5. TI-gain spectra for different carrier densities.
In contrast Beer’s more general law I(L) ∝ exp(gL)
Fig. 4. Spectra of laser devices made out of samples 296 and 306 at T = 83 K optically pumped with the pumping power p.
On sample 296 the modal gain and gain dynamics of a type-II W-structure laser device are investigated both time as well as spectrally resolved. The modal gain is determined with the commonly used variable stripe length method (VSLM) [20]. To examine gain dynamics of the laser device the VSLM is combined with the TR-PL approach. Our setup has a temporal resolution of 1 ps. All measurements are carried out at a device temperature of 90 K under pulsed excitation (150 fs) with carrier densities between 1.68×1017 and 5.6 × 1017 cm−3 . The amplified spontaneous emission (ASE) intensity I of a medium with a gain coefficient g depends on the stripe length L according to I(L) ∝
exp(gL) − 1 . g
(7)
is applicable on the whole time scale. Fig. 5 shows the TI modal gain for three different carrier densities. In this low excitation regime a blue-shift of the gain maximum is observed which turns back into a red-shift due to heating effects for carrier densities higher than 2.8 × 1017 cm−3 (Fig. 6 left hand axis). Fig. 6 also shows that the maximum gain gmax saturates in this range (right hand axis). For 700 m pump stripe length, the FWHM of the TR-ASE pulse is determined to be 5.5 ± 0.5 ps for all investigated carrier densities and transition energies, respectively. Fig. 7a shows the evolution of the TR-ASE pulse after excitation with a 150 fs pulse for two different stripe length of L1 = 700 m and L2 = 500 m at a transition energy of 390 meV and a carrier density of 2.8 × 1017 cm−3 . The corresponding TR-gain g(t) evaluated with the Eqs. (6) and (7) is shown in Fig. 7b. Comparing the results shows that in its validity range the ‘ASE model’ (Eqs. (5) and (6)) agrees with Beer’s law. The spectral behaviour of the TR-gain is also examined. Fig. 8 shows the gain spectra taken at the time of maximum TR-gain at carrier densities between 2.8 × 1017 and
(5)
From the relation I(L1 )/I(L2 ) measured at two different stripe lengths L1 and L2 the spectrally resolved time integrated (TI) gain g is calculated. To determine the time-resolved (TR) gain g(t) we use two macroscopic equations. Due to non-linear gain effects the equation described above is not generally valid but can be used to determine the TR-gain in a certain time interval around the gain maximum only. Here the following equation is used: I(L1 ) exp(gL1 ) − 1 = . I(L2 ) exp(gL2 ) − 1
(6)
Fig. 6. TI-gain maximum (right hand axis) and corresponding energy (left hand axis) versus carrier density.
C. Schwender et al. / Spectrochimica Acta Part A 60 (2004) 3387–3392
Fig. 7. (a) Evolution of the TR-ASE pulse after excitation with 150 fs pulse, (b) corresponding evolution of the TR-gain g(t).
3391
the achieved compositions of the AlAsSb cladding layers from sample 290 only have revealed a small lattice mismatch to the substrate. The samples had been designed with different mixed anion interfaces in the active regions. The interfaces in sample 306 were intended to yield exact lattice matching of the active region (mean value) to the substrate, which has clearly been achieved. As expected, sample 306 shows a more arsenide-like interface stoichiometry compared to the other samples, this has no negative influence on the laser properties. TR-PL measurements show low SRH coefficients for samples 296 and 306 and an excellent low AR coefficient for sample 306. This, together with the high values of the characteristic temperatures in cw laser operation, is again a confirmation of the good crystal quality of the samples. Sample 306 even exhibits pulsed laser emission above room temperature (320 K). TR-ASE studies on III-V compound semiconductors were performed for the first time. The spectral dependence of the maximum TR-gain shows a broad plateau compared to the TI-gain spectra. According to our results, short pulses in the range around 5.5 ps should also be reachable in laser operation.
Acknowledgements We gratefully acknowledge support by the German Research Foundation (DFG) under contracts Fo157/7, Fo157/13, Fo157/18, and Fo157/20 as well as by the ‘Stiftung Rheinland-Pfalz fuer Innovation’ under contract 8312-386261/300. Fig. 8. Gain spectra at times of maximum TR-gain for different carrier densities.
5.6×1017 cm−3 . Compared to the time integrated modal gain (Fig. 5) the spectral dependence of the maximum TR-gain shows a broad plateau, which remains to be explained in the future.
3. Conclusions We have presented a detailed study of the structural properties and non-radiative recombination processes of optically pumped midinfrared laser samples. The modal gain and gain dynamics have also been investigated. The samples have undergone a complete HRXRD analysis. Further characterization methods used have been RHEED, TR-PL, and laser characterization during optical pumping. To examine gain dynamics, TR-PL has been combined with the variable stripe length method. HRXRD results reveal very good crystal quality for all samples. For all laser samples, the active regions were grown pseudomorphically between the cladding layers, whereas the lower cladding layers were partially relaxed with respect to the GaSb substrate. Growth of AlAsSb has been optimized and
References [1] J. Faist, F. Capasso, D.L. Sivco, A.L. Hutchinson, C. Sirtori, S.N.G. Chu, A.Y. Cho, Appl. Phys. Lett. 65 (1994) 2901. [2] R.Q. Yang, S.-S. Pei, J. Appl. Phys. 79 (1996) 8197. [3] D.Z. Garbuzov, R.U. Martinelli, P.K. York, R.J. Menna, J.C. Connolly, Appl. Phys. Lett. 69 (1996) 2006. [4] W.W. Bewley, C.L. Felix, I. Vurgaftman, E.H. Aifer, J.R. Meyer, L. Goldberg, J.R. Lindle, D.H. Chow, E. Selvig, IEEE Photonic Technol. Lett. 10 (1998) 660. [5] N. Bertru, A. Baranov, Y. Cuminal, G. Almuneau, F. Genty, A. Joullié, O. Brandt, A. Mazuelas, K.H. Ploog, Semicond. Sci. Technol. 13 (1998) 936. [6] W.W. Bewley, C.L. Felix, E.H. Aifer, I. Vurgaftman, L.J. Olafsen, J.R. Meyer, H. Lee, R.U. Martinelli, J.C. Connolly, A.R. Sugg, G.H. Olsen, M.J. Yang, B.R. Bennett, B.V. Shanabrook, Appl. Phys. Lett. 73 (1998) 3833. [7] W.W. Bewley, H. Lee, I. Vurgaftman, R.J. Menna, C.L. Felix, R.U. Martinelli, D.W. Stokes, D.Z. Garbuzov, J.R. Meyer, M. Maiorov, J.C. Connolly, A.R. Sugg, G.H. Olsen, Appl. Phys. Lett. 76 (2000) 256. [8] C. Becker, I. Prevot, X. Marcadet, B. Vinter, S. Sirtori, Appl. Phys. Lett. 78 (2001) 1029. [9] H.Q. Le, C.H. Lin, S.J. Murray, R.Q. Yang, S.-S. Pei, IEEE J. Quantum Electron 34 (1998) 1016. [10] I. Vurgaftman, J.R. Meyer, F.H. Julien, L.R. Ram-Mohan, Appl. Phys. Lett. 73 (1998) 711.
3392
C. Schwender et al. / Spectrochimica Acta Part A 60 (2004) 3387–3392
[11] M.J. Yang, W.J. Moore, B.R. Bennett, B.V. Shanabrook, J.O. Cross, W.W. Bewley, C.L. Felix, I. Vurgaftman, J.R. Meyer, J. Appl. Phys. 86 (1999) 1796. [12] W. Barvosa-Carter, M.E. Twigg, M.J. Yang, L.J. Whitman, Phys. Rev. B 63 (2001) 245311. [13] Q. Xie, J.E. van Nostrand, J.L. Brown, C.E. Stutz, J. Appl. Phys. 86 (1999) 329. [14] G. Hoffmann, H.J. Schimper, C. Schwender, N. Herhammer, G.F. West, B. Vogelgesang, J.O. Drumm, H. Fouckhardt, M. Scheib, J. Cryst. Growth 257 (2003) 42. [15] M.A.G. Halliwell, J. Cryst. Growth 170 (1997) 47.
[16] M.A.G. Halliwell, Measurement of relaxation in strained layer semiconductor structures, in: C.S. Barrett (Ed.), Advances in X-ray Analysis, vol. 33, New York, 1990, p. 61. [17] D.-J. Jang, J.T. Olesberg, M.E. Flatté, T.F. Boggess, T.C. Hasenberg, Appl. Phys. Lett. 70 (1997) 1125. [18] W.T. Cooley, R.L. Hengegold, Y.K. Yeo, Appl. Phys. Lett. 73 (1998) 2890. [19] J.O. Drumm, B. Vogelgesang, G. Hoffmann, C. Schwender, N. Herhammer, H. Fouckhardt, Semicond. Sci. Technol. 17 (2002) 1115. [20] K.L. Shaklee, R.F. Leheny, Appl. Phys. Lett. 18 (1971) 475.
X-ray studies and time-resolved photoluminescence on optically pumped antimonide-based midinfrared type-II lasers Carsten Schwender, Jan Oliver Drumm, Goetz Hoffmann, Birgit Vogelgesang1 , Henning Fouckhardt∗ Integrated Optics and Microoptics Research Group, Physics Department, Kaiserslautern University of Technology, Erwin-Schroedinger-Str. 46, D-67663 Kaiserslautern, Germany Received 26 September 2003; accepted 26 November 2003
Abstract We report on high-resolution X-ray diffraction and time-resolved photoluminescence (TR-PL) studies of antimonide-based midinfrared (MIR) type-II laser samples. A structural characterization taking into account asymmetrical strain, layer tilting, and relaxation enables an accurate determination of the average lattice constant of the active region and the composition of the cladding layers. By designing the antimonide-to-arsenide interfaces, we achieve exact lattice matching of the active region to the substrate. Non-radiative recombination processes are investigated with time-resolved photoluminescence. The samples are also characterized under optically pumped laser operation. By an examination of the time-integrated and time-resolved amplified spontaneous emission (TR-ASE), we investigate the modal gain and gain dynamics. The variable stripe length method is combined with the TR-PL approach. Compared to the time-integrated gain spectra the spectral dependence of the maximum and minimum time-resolved gain shows a broad plateau. The full width half maximum (FWHM) of the TR-ASE pulse is 5.5 ± 0.5 ps. Thus, short pulses in this range should be achievable upon laser operation. The active regions of the laser structures investigated here are promising subunits of type-II quantum cascade lasers. © 2004 Elsevier B.V. All rights reserved. Keywords: Midinfrared semiconductor lasers; Antimonide lasers; Type-II heterostructures; Non-radiative recombination; Shockley–Read–Hall; Auger
1. Introduction In the last decade there has been considerable progress in midinfrared semiconductor lasers (MIR, 2–5 m) [1–8]. Such devices are very useful for detection of solvents and trace gases such as hydrocarbons. One approach uses so-called “W-structures”, consisting of a symmetric type-II broken-gap quantum well double-heterostructure, e.g. InAs/Ga(In)Sb/InAs bounded by Al(As)Sb barrier layers [2,4,6,7]. Different suggestions have been made to improve these lasers. Designing the strain and the mixed anion interfacial bond stoichiometry between the single W-structure layers gives an opportunity to reduce non-radiative recombination processes [9–12]. For an excellent crystal quality, it is also necessary to avoid unintentional layer relaxation. ∗
Corresponding author. Tel.: +49-631-205-4145; fax: +49-631-205-4147. E-mail address: [email protected] (H. Fouckhardt). 1 Meanwhile with Robert Bosch GmbH, Stuttgart, Germany. 1386-1425/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.saa.2003.11.040
To evaluate the influence of the suggested changes on the laser properties, a detailed knowledge of the structural parameters of the samples is necessary. Our samples have a design similar to the lasers in [6] but with different W-period numbers and cladding layer types. In the present paper, an analysis of non-radiative recombination processes (Shockley–Read–Hall (SRH) and Auger (AR) with holes) as well as a detailed structural characterization of the laser structure is performed. Adjusting the composition of the mixed anion interfacial bonds allows for strain compensation of the active region. The laser properties, modal gain, and gain dynamics are examined.
2. Experiments and results The samples are grown by molecular beam epitaxy (MBE) on n-type GaSb substrates using a DCA 450 MBE-system. The active region of the laser samples consists of 70 periods of InAs/GaSb/InAs/AlSb (6 ML/7 ML/6 ML/11 ML)
3388
C. Schwender et al. / Spectrochimica Acta Part A 60 (2004) 3387–3392 Table 1 Thicknesses of single W-structures tW in laser samples determined by RHEED and HRXRD, FWHM bs of the first order HRXRD satellite reflexes
Fig. 1. (0 0 4) HRXRD rocking curves of sample 296 and 306.
surrounded by (each) 1.4 m thick AlAsSb cladding layers, which are intended to be lattice matched to the GaSb substrate. All samples are capped by a 5 nm thick GaSb layer. The layer thicknesses are monitored by using a reflective high-energy electron diffraction (RHEED) system equipped with an image processing tool. In order to use this in situ measurement technique the samples are not rotated during growth. The samples differ in interface composition only, which is controlled by the shutter sequences. The As for Sb exchange on Ga(Al)Sb surfaces can be greatly reduced by using In to tie Sb to the surface [13]. This also gives the opportunity to control the As for Sb exchange at the interface by varying the amount of deposited In. The sequence for the antimonide-to-arsenide interfaces is 10.0 s Sb soak for all interfaces and 4.0 s In deposition for samples 276, 277, 279, 290, and 296 and 2.8 s In deposition for sample 306 to adjust the average lattice constant of the active region with a mixed-interface stoichiometry. Detailed control of the interface stoichiometry is essential to obtain precise values for the average lattice constant as desired and thus to achieve strain compensation. Detailed information about the growth and the calibration of the growth parameters can be found in [14]. For a precise structural characterization of the samples, high-resolution X-ray diffractometry (HRXRD) measurements are performed by using a four crystal, high-resolution X-ray diffractometer. To probe growth quality and to determine the thickness of the W-structure periods in the active region, a HRXRD diffraction profile is measured at the (0 0 4) plane. As an example, Fig. 1 gives the rocking curves of samples 296 and 306 which show sharp satellite peaks up to the seventh order. This indication of high crystal quality is similar for all samples. The low full width half maximum (FWHM) bs of the first order satellite peaks, which is 167 arcsec for samples 296 and 140 arcsec for sample 306 also confirms high crystal quality. The values for all samples are shown in Table 1. The angular spacing of the satellite peaks gives the period thickness tW of the W-structures in the active region, which is intended to be 32 ML for all samples including 4 × 0.5 ML for the four interfacial bonds between
Sample
tW (ML) RHEED
tW (ML) HRXRD
bs (arcsec)
276 277 279 290 296 306
31 30.5 30.5 32 32 33
31.1 31.2 31.6 33.9 32.8 32.4
117 118 110 108 140 167
antimonide and arsenide layers. Table 1 shows a comparison of the thickness values of all samples determined with RHEED and with HRXRD. They are in good agreement with the target value. Besides the standard rocking curve measured at the (0 0 4) plane, eight rocking curves are taken from the four inclined crystal planes (1 1 5), (1¯ 1¯ 5), (1 1¯ 5), and (1¯ 1 5) in grazing incidence (+) and grazing exit (−) geometry. Due to the detailed analysis of the crystal structure it is possible to eliminate effects such as layer tilting and asymmetrical strain relaxation [15,16]. The HRXRD results for the inclined planes are examplified in Fig. 2 with the rocking curves measured at the (1 1 5) plane in grazing incidence (1 1 5(−) ) and grazing exit (1 1 5(+) ) geometry. From the measurements at the inclined planes the lattice parameters perpendicular to the surface d⊥ = d001 and the in-plane lattice parameters d|| of the different layers are extracted [15] whereby the value of d⊥ for the active region is only a mean value over the different layers of the periodic structure. With d⊥ and d|| the lattice constants of the layers are calculated taking the elastic constants into account. Applying Vegard’s law, the As fraction for the different AlAsSb layers varies from 10.3 to 12.0% for samples 276, 277, and 279 instead of the 8.4% as desired
Fig. 2. (1 1 5) HRXRD rocking curves (Θ is the diffraction angle relative to the angle of the substrate peak).
C. Schwender et al. / Spectrochimica Acta Part A 60 (2004) 3387–3392
to achieve lattice matching. After an optimization of the As/Sb flux ratio values between 7.7 and 10.0% were reached whereby the lattice mismatch of the cladding layers relative to GaSb is in the range of −0.13 to 0.05%. For the mean lattice constant of the active region of samples 276–296 finally values from 6.1061 to 6.1077 Å are achieved reproducibly. The active region of sample 306 has a mean lattice constant of 6.0959 Å (lattice match to GaSb) and so strain compensation of the active region by varying the mixed anion interfaces is clearly achieved. By comparing the expected mean lattice constants for arsenide- and antimonide-like interfaces with the values extracted from HRXRD measurements, the interface composition can be deduced. With an In deposition time of 4.0 s, the interface composition is 70–75% InSb-like. The reduced In deposition time of 2.8 s for sample 306 results in a reduction of the Sb-fraction in the interface to 53%. But the 20% higher As-fraction in these interfaces does not noticeably effect the crystal structure. To examine the non-radiative recombination processes in the laser devices, for samples 296 and 306 the coefficients of SRH and AR are investigated with a time-resolved photoluminescence up-conversion setup (TR-PL) [17,18]. This setup provides 1 ps temporal resolution in the midinfrared wavelength range with high sensitivity [19]. To determine the SRH and AR coefficients with TR-PL, the rate equation −
∂N = AN + BN2 + CN3 ∂t
(1)
is used, whereas A, B, and C are the SRH, radiative and AR coefficients, respectively. In our measurements, the carrier densities are in the range of (5–15)×1018 cm−3 where the AR term for high carrier densities is more complicated and has to be written as CN3 → C1 N + C2 N 2 + C3 N 3
(2)
which results in a modified rate equation: −
∂N = (A + C1 )N + (B + C2 )N 2 + C3 N 3 . ∂t
(3)
This means that even the coefficients of the linear and the quadratic term contain parts due to Auger recombination. The measurements for determination of the SRH coefficients are performed at 10 K and for the AR coefficient at 80, 120, and 200 K. The temperature and/or carrier density dependence of Auger and SRH coefficients are investigated for sample 306. The excellent growth quality of the devices is indicated by SRH coefficients at 10 K of (1.8 ± 0.6) × 108 s−1 for sample 296 and (2.2 ± 0.7) × 108 s−1 for sample 306. Fig. 3 shows that the cubic AR coefficient C3 for sample 306 decreases in a characteristic manner with increasing initially excited carrier density. To determine the AR coefficient at low carrier densities, i.e. C30 , we use a convergence equation according to [19] C3 (N0 ) =
C30 , 1 + (N0 /NC )p
(4)
3389
Fig. 3. Experimental cubic AR coefficient versus carrier density at temperature of 200 K. The decay of C3 (N0 ) was fitted by a convergence equation [19].
where C30 , NC , and p are fit parameters to data points C3 (N0 ). We achieved very low cubic AR coeffients C30 (valid at low carrier densities) of (2.3 ± 0.1) × 10−29 cm6 s−1 at 80 K, (2.5 ± 0.3) × 10−29 cm6 s−1 at 120 K, and (1.2 ± 2.6) × 10−28 cm6 s−1 at 200 K for this sample [19]. The experimentally determined AR coefficients are mainly due to CHHS Auger recombination. A more detailed report on these TR-PL studies is published elsewhere [19]. To test the laser properties of the grown structures 500 m long edge emitting laser devices are prepared and characterized. The devices are mounted into a heatable liquid nitrogen dewar and are optically pumped by a Nd:YAG laser operating either in continuous wave or pulsed radiation emission mode (200 ns, 50 kHz). The width of the pump stripe is 60 m. For the two different interface stoichiometries the laser results of the two samples 296 and 306 are presented. Fig. 4 shows the wavelength spectra of Fabry–Perot laser devices of samples 296 and 306 at a temperature of 83 K. For applications in MIR spectroscopy the lasers have to be operated with an external dispersive resonator in order to provide for single-longitudinal-mode operation. By optical pumping with a Nd:YAG laser source, the samples show cw laser emission at 83 K at wavelengths of 3.40 and 3.32 m, respectively. At this temperature, the highest maximum output power is measured for sample 306 with 14 mW per facet. The slope efficiency is 3.3%. The maximum working temperature for cw laser operation is 170 K. The characteristic temperatures at cw laser operation are T0 = 57.6 ± 2.9 K for sample 296 and T0 = 54.2 ± 2.3 K for sample 306. In pulsed laser operation sample 306 emits laser radiation up to 320 K; under these conditions we determined a characteristic temperature of T0 = 73.5 ± 6.9 K. These T0 values are among the best for MIR lasers with similar structures and similar emission wavelengths [6]. The laser emission results confirm the low non-radiative recombination coefficients and reveal excellent growth quality.
3390
C. Schwender et al. / Spectrochimica Acta Part A 60 (2004) 3387–3392
Fig. 5. TI-gain spectra for different carrier densities.
In contrast Beer’s more general law I(L) ∝ exp(gL)
Fig. 4. Spectra of laser devices made out of samples 296 and 306 at T = 83 K optically pumped with the pumping power p.
On sample 296 the modal gain and gain dynamics of a type-II W-structure laser device are investigated both time as well as spectrally resolved. The modal gain is determined with the commonly used variable stripe length method (VSLM) [20]. To examine gain dynamics of the laser device the VSLM is combined with the TR-PL approach. Our setup has a temporal resolution of 1 ps. All measurements are carried out at a device temperature of 90 K under pulsed excitation (150 fs) with carrier densities between 1.68×1017 and 5.6 × 1017 cm−3 . The amplified spontaneous emission (ASE) intensity I of a medium with a gain coefficient g depends on the stripe length L according to I(L) ∝
exp(gL) − 1 . g
(7)
is applicable on the whole time scale. Fig. 5 shows the TI modal gain for three different carrier densities. In this low excitation regime a blue-shift of the gain maximum is observed which turns back into a red-shift due to heating effects for carrier densities higher than 2.8 × 1017 cm−3 (Fig. 6 left hand axis). Fig. 6 also shows that the maximum gain gmax saturates in this range (right hand axis). For 700 m pump stripe length, the FWHM of the TR-ASE pulse is determined to be 5.5 ± 0.5 ps for all investigated carrier densities and transition energies, respectively. Fig. 7a shows the evolution of the TR-ASE pulse after excitation with a 150 fs pulse for two different stripe length of L1 = 700 m and L2 = 500 m at a transition energy of 390 meV and a carrier density of 2.8 × 1017 cm−3 . The corresponding TR-gain g(t) evaluated with the Eqs. (6) and (7) is shown in Fig. 7b. Comparing the results shows that in its validity range the ‘ASE model’ (Eqs. (5) and (6)) agrees with Beer’s law. The spectral behaviour of the TR-gain is also examined. Fig. 8 shows the gain spectra taken at the time of maximum TR-gain at carrier densities between 2.8 × 1017 and
(5)
From the relation I(L1 )/I(L2 ) measured at two different stripe lengths L1 and L2 the spectrally resolved time integrated (TI) gain g is calculated. To determine the time-resolved (TR) gain g(t) we use two macroscopic equations. Due to non-linear gain effects the equation described above is not generally valid but can be used to determine the TR-gain in a certain time interval around the gain maximum only. Here the following equation is used: I(L1 ) exp(gL1 ) − 1 = . I(L2 ) exp(gL2 ) − 1
(6)
Fig. 6. TI-gain maximum (right hand axis) and corresponding energy (left hand axis) versus carrier density.
C. Schwender et al. / Spectrochimica Acta Part A 60 (2004) 3387–3392
Fig. 7. (a) Evolution of the TR-ASE pulse after excitation with 150 fs pulse, (b) corresponding evolution of the TR-gain g(t).
3391
the achieved compositions of the AlAsSb cladding layers from sample 290 only have revealed a small lattice mismatch to the substrate. The samples had been designed with different mixed anion interfaces in the active regions. The interfaces in sample 306 were intended to yield exact lattice matching of the active region (mean value) to the substrate, which has clearly been achieved. As expected, sample 306 shows a more arsenide-like interface stoichiometry compared to the other samples, this has no negative influence on the laser properties. TR-PL measurements show low SRH coefficients for samples 296 and 306 and an excellent low AR coefficient for sample 306. This, together with the high values of the characteristic temperatures in cw laser operation, is again a confirmation of the good crystal quality of the samples. Sample 306 even exhibits pulsed laser emission above room temperature (320 K). TR-ASE studies on III-V compound semiconductors were performed for the first time. The spectral dependence of the maximum TR-gain shows a broad plateau compared to the TI-gain spectra. According to our results, short pulses in the range around 5.5 ps should also be reachable in laser operation.
Acknowledgements We gratefully acknowledge support by the German Research Foundation (DFG) under contracts Fo157/7, Fo157/13, Fo157/18, and Fo157/20 as well as by the ‘Stiftung Rheinland-Pfalz fuer Innovation’ under contract 8312-386261/300. Fig. 8. Gain spectra at times of maximum TR-gain for different carrier densities.
5.6×1017 cm−3 . Compared to the time integrated modal gain (Fig. 5) the spectral dependence of the maximum TR-gain shows a broad plateau, which remains to be explained in the future.
3. Conclusions We have presented a detailed study of the structural properties and non-radiative recombination processes of optically pumped midinfrared laser samples. The modal gain and gain dynamics have also been investigated. The samples have undergone a complete HRXRD analysis. Further characterization methods used have been RHEED, TR-PL, and laser characterization during optical pumping. To examine gain dynamics, TR-PL has been combined with the variable stripe length method. HRXRD results reveal very good crystal quality for all samples. For all laser samples, the active regions were grown pseudomorphically between the cladding layers, whereas the lower cladding layers were partially relaxed with respect to the GaSb substrate. Growth of AlAsSb has been optimized and
References [1] J. Faist, F. Capasso, D.L. Sivco, A.L. Hutchinson, C. Sirtori, S.N.G. Chu, A.Y. Cho, Appl. Phys. Lett. 65 (1994) 2901. [2] R.Q. Yang, S.-S. Pei, J. Appl. Phys. 79 (1996) 8197. [3] D.Z. Garbuzov, R.U. Martinelli, P.K. York, R.J. Menna, J.C. Connolly, Appl. Phys. Lett. 69 (1996) 2006. [4] W.W. Bewley, C.L. Felix, I. Vurgaftman, E.H. Aifer, J.R. Meyer, L. Goldberg, J.R. Lindle, D.H. Chow, E. Selvig, IEEE Photonic Technol. Lett. 10 (1998) 660. [5] N. Bertru, A. Baranov, Y. Cuminal, G. Almuneau, F. Genty, A. Joullié, O. Brandt, A. Mazuelas, K.H. Ploog, Semicond. Sci. Technol. 13 (1998) 936. [6] W.W. Bewley, C.L. Felix, E.H. Aifer, I. Vurgaftman, L.J. Olafsen, J.R. Meyer, H. Lee, R.U. Martinelli, J.C. Connolly, A.R. Sugg, G.H. Olsen, M.J. Yang, B.R. Bennett, B.V. Shanabrook, Appl. Phys. Lett. 73 (1998) 3833. [7] W.W. Bewley, H. Lee, I. Vurgaftman, R.J. Menna, C.L. Felix, R.U. Martinelli, D.W. Stokes, D.Z. Garbuzov, J.R. Meyer, M. Maiorov, J.C. Connolly, A.R. Sugg, G.H. Olsen, Appl. Phys. Lett. 76 (2000) 256. [8] C. Becker, I. Prevot, X. Marcadet, B. Vinter, S. Sirtori, Appl. Phys. Lett. 78 (2001) 1029. [9] H.Q. Le, C.H. Lin, S.J. Murray, R.Q. Yang, S.-S. Pei, IEEE J. Quantum Electron 34 (1998) 1016. [10] I. Vurgaftman, J.R. Meyer, F.H. Julien, L.R. Ram-Mohan, Appl. Phys. Lett. 73 (1998) 711.
3392
C. Schwender et al. / Spectrochimica Acta Part A 60 (2004) 3387–3392
[11] M.J. Yang, W.J. Moore, B.R. Bennett, B.V. Shanabrook, J.O. Cross, W.W. Bewley, C.L. Felix, I. Vurgaftman, J.R. Meyer, J. Appl. Phys. 86 (1999) 1796. [12] W. Barvosa-Carter, M.E. Twigg, M.J. Yang, L.J. Whitman, Phys. Rev. B 63 (2001) 245311. [13] Q. Xie, J.E. van Nostrand, J.L. Brown, C.E. Stutz, J. Appl. Phys. 86 (1999) 329. [14] G. Hoffmann, H.J. Schimper, C. Schwender, N. Herhammer, G.F. West, B. Vogelgesang, J.O. Drumm, H. Fouckhardt, M. Scheib, J. Cryst. Growth 257 (2003) 42. [15] M.A.G. Halliwell, J. Cryst. Growth 170 (1997) 47.
[16] M.A.G. Halliwell, Measurement of relaxation in strained layer semiconductor structures, in: C.S. Barrett (Ed.), Advances in X-ray Analysis, vol. 33, New York, 1990, p. 61. [17] D.-J. Jang, J.T. Olesberg, M.E. Flatté, T.F. Boggess, T.C. Hasenberg, Appl. Phys. Lett. 70 (1997) 1125. [18] W.T. Cooley, R.L. Hengegold, Y.K. Yeo, Appl. Phys. Lett. 73 (1998) 2890. [19] J.O. Drumm, B. Vogelgesang, G. Hoffmann, C. Schwender, N. Herhammer, H. Fouckhardt, Semicond. Sci. Technol. 17 (2002) 1115. [20] K.L. Shaklee, R.F. Leheny, Appl. Phys. Lett. 18 (1971) 475.