Electrical characterization of engineered ZnSeGaAs heterojunction diodes

Electrical characterization of engineered ZnSeGaAs heterojunction diodes

Journal ELSEVIER of Crystal Growth 1751176 (1997) 603-607 Electrical characterization of engineered ZnSe-GaAs heterojunction diodes Marco Michel...

463KB Sizes 2 Downloads 98 Views

Journal

ELSEVIER

of Crystal

Growth

1751176 (1997) 603-607

Electrical characterization of engineered ZnSe-GaAs heterojunction diodes Marco

Michele Lazzer?*, Vittorio Pellegrini”, Fabio BeltramaT1, Lazzarinob3C, Jens J. Paggelb,c32, Lucia Sorbab’c’3, Silvia Rubinibgc, Alberta Bonannib,“, Alfonso FranciosibTc,4

a Scuola Normale Superiore and Istituto Nazionale per la Fisica della Materia, I-56126 Piss, Italy b Laboratorio Nazionale TASC-INFM, Area di Ricerca, I-34012 Trieste. Italy ‘Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, Minnesota 55455, USA

Abstract Electrical characterization of ZnSe/GaAs n-p heterodiodes grown by molecular beam epitaxy under different Zn/Se flux ratios is reported. Large tunability of the band discontinuity at the heterojunction is shown by photocurrent measurements at low temperature with conduction-band offsets in the range 0.26-0.75 eV. Achievement of device-grade heterostructures with engineered offsets is shown under appropriate growth conditions.

PACS: 73.40. - c; 73.4O.Lq; Keywords:

79.6O.J~

Diodes; Heterojuctions

1. Introduction One of the most important tools of band-gap engineering is the tuning of energy-band discontinuities at heterojunctions. Several authors [ 141 * Corresponding author. ’Also with Laboratorio Nazionale TASC-INFM, I-34012 Trieste, Italy. ’ Present address: Philipps-Universitit-Marburg, Fachbereich Physik, Renthof 6, D-35032 Marburg, Germany. ‘Also with Instituto ICMAT de1 CNR, Monterotondo, I00016 Rome, Italy. 4Also with Dipartimento di Fisica, Universita di Trieste, I-34127 Trieste, Italy. 0022-0248/97/$17.00 Copyright PII SOO22-0248(96)00987-6

attempted to achieve such tuning through the controlled introduction of impurities in the junction regionThese efforts, however, have had limited impact on device applications. This is mostly because few results have been verified for heterojunctions incorporated in functional devices, and because the impurities used to change the local interface electrostatics [Z] are poorly suited to device applications, since they are known to give rise to reliability problems (alkali metals), or deep levels (noble metals). Recently, however, owing to the growth of high-quality heterovalent heterostructures for optoelectronic applications, new avenues to achieve such tunability have become available

(3 1997 Elsevier Science B.V. All rights reserved

M. Lazzeri

604

et al. ~Journal

of‘Ctyta1

[S-S]. In heterovalent heterojunctions with polar orientation the band alignment should depend on the detail of the local atomic configuration achieved at the interface during growth [S-S]. Offset changes are therefore possible even in the absence of foreign impurities, provided that the growth kinetics can be exploited to obtain different interface configurations. A few applications of this technique have been recently demonstrated in heterojunctions ~ including ZnSe/GaAs - fabricated with heterovalent overlayers thin enough (223 nm) for the offset to be probed by photoemission spectroscopy [7, 81. In this article, we report on the tunability of ZnSe/GaAs(O 0 1) heterovalent heterojunctions, which are essential elements of all recently demonstrated II-VI based solid-state blue-green lasers [9], in fully functional heterojunction diodes. We analyzed the current-voltage (1-I’) and photo 1-V characteristics of two sets of n-ZnSe/p-GaAs diodes fabricated by molecular beam epitaxy (MBE) with different interface compositions. Our results allowed us to determine the band discontinuity and the applicability of this technique to the fabrication of complete devices.

2. Experimental

procedure

Samples were grown by solid-source MBE in a system that includes interconnected chambers for the growth of III-V and II-VI materials. We used GaAs(0 0 1) wafers, on which ZnSe layers were grown with a Zn/Se beam pressure ratio (BPR) of 0.1, 1, or 10. We employed two different growth methods to modify the local interface composition. In the first method, the whole II-VI overlayer was grown in Se-rich or Zn-rich conditions. Alternatively, the Se-rich or Zn-rich growth conditions were limited to a 2 nm-thick composition-control interface layer (CIL), the rest of the II-VI overlayer being grown with BPR = 1. The two methods have been shown to yield Se-rich (for low BPRs) or Zn-rich (for high BPRs) interface compositions under an essentially stoichiometric II-VI overlayer [X, lo]. Two sets of samples were grown. One set was used for the determination of the conduction-band offset AE, via measurements of the tunneling cur-

Growth

175:176

(l997J

603- 607

rent from photoinjected carriers, using a technique illustrated elsewhere in detail [ll], and one set was used for 1-V characterization. The latter samples were grown on p-type wafers and consisted of a 500 nm-thick GaAs layer grown at 580°C and doped p = 3-4 x 1016 cmm3, followed by a 500 nmthick ZnSe overlayer (including the CIL) doped with chlorine at y1= 334 x lOi cm-j. A graded n+ Zn, _,Cd,Se region concluded the growth. The latter region was fabricated to promote the formation of ohmic contacts [12] which were realized by e-beam deposition of Al with no thermal treatment. Devices were fabricated by defining circular mesas (75 urn diameter) by standard photolithographic and wet etching techniques with removal of the entire II-VI epilayer around each mesa. Samples used for photo 1-V measurements differ in that the GaAs region adjacent to the ZnSe overlayer was limited to 10 nm and was grown on a A1,.,GaO,,As layer doped p = 334 x lO”‘j cme3. The compositional profile was chosen so that carriers could be photogenerated only in the thin GaAs layer near the interface using a low-power laser diode tuned at 1570 meV, i.e., above the GaAs band gap, but below the A&GaO.sAs band gap. This allowed us to avoid carrier generation far from the interface and possible transfer into GaAs satellite valleys before tunneling. All photo 1-V measurements below 35 K were performed in a closed-cycle refrigerator in order to suppress thermally activated transport. More detail on the sample structure and experimental protocol can be found elsewhere [ 131.

3. Results We measured the photo 1-V characteristics at applied bias such that tunneling through the ZnSerelated triangular barrier is the dominant mechanism (i.e., in the range 12-20 V [13]). In the bias and temperature ranges of interest, electrons populate only the ground-state subband EO. The functional dependence of the current on the applied bias can then be described by [14] I = xFi exp[ - B/F,,],

(1)

where B = $(@/eh)@i/2, rn: = O.l4m, is the ZnSe electron effective mass [15], x is a constant,

M. Lazzeri et al.

i Journal

ofCyta1

& is the activation energy and F, is the relevant electric field in the ZnSe layer. Since very low carrier densities were photogenerated, AE, = &a + EO. E, is a function of the electric field in the GaAs interface layer. The spatial dependence of the electric field across the entire structure was calculated solving Poisson’s equation in the depletion-layer approximation and imposing the continuity of the displacement vector at the heterojunction interface [ 131. In Fig. 1 we show representative semilog experimental plots of I/F: versus l/F, for diodes with ZnSe overlayers grown with BPR = 0.1 (top left) or 10 (bottom left) throughout, as well as diodes incorporating Se-rich (top right) or Zn-rich (bottom right) CIL. The solid lines are linear fits of the experimental data in the applied bias range where Eq. (1) holds [13, 141. The slope of the fits yields B and, therefore, AE,.

Growth I75 JI 76 (1997) 603-607

In Table 1 we summarize our findings for the conduction band discontinuity in all samples at 35 K (first column). The quoted experimental errors include both the uncertainty in the fitting

Table I Column one: Zn/Se beam pressure ratio (BPR) used for fabrication of ZnSe overlayers on GaAs, or for the composition control interface layer (CIL) in ZnSe/GaAs heterojunctions; column two: Measured conduction-band discontinuities; column three: Valence-band d&continuities BE, as deduced from the experimental AE, (this work) and the ZnSejGaAs bandgap difference of 1301 meV; column four: AE, as deduced from X-ray photoemission (XPS) data on thin overlayer samples; the quoted uncertainty in the XPS data correspond to the data scatter among several dozen thin overlayer samples [S, IO] BPR 0.1

I 10 0.1 CIL 10 GIL

BPR = 0.1

605

AE, (meV)

AE, (meV)

AE, (XPS) (meV)

716 361 251 751 261

585 940 1050 550 1040

620 850 1120 580 1050

f + i k k

45 20 10 25 10

f * * + &

45 20 10 25 IO

+ k k * f

70 100 80 50 50

-BPR=l -----BPR=O.l

upn= ,u

‘FItI BPR=IO

3

1

0

4

Bias & I

------.--...

XlOd

2.7~10'

BPR= 1 BPR= 0.1 GIL BP& 10 GIL

3.0x10-~ i

Fig. 1. FowlerrNordheim plot of the photocurrent at low temperature and high applied bias for ZnSe/GaAs diodes incorporating Se-rich (top) or Zn-rich (bottom) interfaces. These were obtained by growing the ZnSe overlayer with Zn/Se beam pressure ratio (BPR) of 0.1 (top left) or 10 (bottom left) throughout, or by incorporating a thin composition control interfaces layer (CIL) grown in nonstoichiometric conditions (rightmost sections). Also shown is a least-squares fit of the data according to Eq. (1).

IE-14

.I .\ I’ .,I : ,,.,........ ‘r 0

I

. ...’

1

.:., I

I

2

3

4

Bias (V)

Fig. 2. Current&voltage characteristics at room temperature for representative diodes grown with different Zn/Se beam-pressure ratio BPRs. A given BPR were kept at the value indicated for the whole overlayer (upper panel) or for a 2 nm-thick interface layer (lower panel).

606

M. Lazzeri et al. / Journal ofCcyysta1 Growth 175/l 76 (1997) 603-607

procedure and the scatter among the different individual devices measured. Samples grown with BPR = 1 yield AE, = 361 f 20 meV in agreement with values reported by other authors [16]. Conduction-band offsets in the range from 0.26 to 0.75 eV are observed in going from Zn-rich to Se-rich interfaces, proving both the efficacy of the offset modulation technique here exploited and the validity of the CIL concept in functional devices. To study further the validity of the present approach for device fabrication, we also examined the I&I/ characteristics of a set of GaAs/ZnSe heterodiodes at room temperature (for the case of GaAs extending to the whole p-doped side). Fig. 2 shows our data for samples grown with the same BPR throughout the overlayer (upper panel) or incorporating an appropriate CIL (lower panel).

4. Discussion Recent XPS studies which laid a background for the present work had indicated that the Zn/Se BPR employed during the early stages of interface fabrication determine the local interface composition and the band offset. Specifically, Nicolini et al. [S] presented the first X-ray photoemission spectroscopy (XPS) evidence of a valence band offset change in ZnSe-GaAs(0 0 1) heterojunctions involving 2-3 nm overlayers, discussed the variations in interface composition that correlate with the change, and examined theoretically the interface configurations that may account for offset tunability. Bonanni et al. [lo] showed by XPS and photoluminescence spectroscopy that the growth conditions responsible for the change in interface composition and valence band offset can be restricted to the very early stages of interface formation, with a corresponding improvement in II-VI optical material quality. All of the above results on the band alignment [S, lo] were obtained using heterostructures with ZnSe overlayers thin enough (223 nm) for the offset to be probed by XPS. The results in Fig. 1 are instead, to our knowledge, the first verification of a locally engineered heterojunction band offset in a fully functional device by transport methods. A quantitative comparison with the early XPS re-

sults in thin-overlayer samples can be made using the literature value of the ZnSe band gap to derive the valence band offsets from the conduction band offsets determined in Fig. 1. The resulting valence band discontinuities are listed in column two of Table 1, while the results obtained earlier by XPS [S, lo] are listed in column three. The agreement between the two sets of values is very satisfactory and suggests that the interface configurations responsible for band-offset tuning are sufficiently stable to withstand the fabrication stages required to go from the thin-overlayer samples to fully functional devices. The results of Fig. 2 further support the possibility of exploiting the tunability of heterovalent heterojunctions in actual devices while pointing out some of the related limitations. The poorer structural quality of BPR = 10 overlayers, which contain a comparatively high density of stacking faults and dislocations [13, 171 manifests itself in Fig. 2 with the large leakage current (upper panel, dotted line), and is only marginally improved by the CIL technique (lower panel, dotted line). On the contrary, diodes fabricated with BPR = 0.1 (dashed line) show acceptable behavior and display good rectifying properties, particularly in the CIL case. The observed reduced reverse current as compared to the BPR = 1 control samples can be linked to the increased conduction-band discontinuity, which hinders the transfer of thermally generated electrons from the GaAs side into the ZnSe layer. Diodes with engineered interfaces (dashed and dotted lines) do exhibit a somewhat degraded forward characteristics. Since heterostructures incorporating Se-rich CILs show extended defect densities substantially lower than those observed in the control samples [17], the most likely explanation is to be found in the presence of new point defects in the interface region leading to enhanced recombination at room temperature. Recent cathodoluminescence studies of ZnSe/GaAs heterojunctions [ 181 have shown that interfaces grown in Se-rich conditions are less stable against atomic interdiffusion and the formation of defect complexes involving substitutional Ga atoms. Systematic current-voltage studies in progress will address the relation between interfacial engineering and the final interface state density.

M. Lazzeri et al. /Journal

of Crystal Growth 1751176 (1997) 603-607

5. Conclusions We have demonstrated that engineered band offsets can be observed and exploited in practical devices. In particular, reduction of the large valence-band offsets in ZnSe/GaAs heterojunctions can be obtained without degradation of the electrical properties of the heterojunction. This result may find an important applications in II-VI on III-V blue-green lasers since it should yield a reduced series resistance as a result of the improved hole injection in the active layer.

Acknowledgements This work was supported in part by the US Army Research Office under Grants Nos. DAAH04-93-G-0206 and DAAH04-93-G-0319, and by CNR under the GaAsNET project.

References 111 F. Capasso,

A.Y. Cho, K. Mohammed and P.W. Foy, Appl. Phys. Lett. 46 (1985) 664. and P. Perfetti, in: Heterojunction Band PI G. Margaritondo Discontinuities: Physics and Device Applications, Eds. G. Margaritondo and F. Capasso (North-Holland, Amsterdam, 1987) 2. on Semiconductors, Vol. I, Ed. c31 L.J. Brillson, in: Handbook P.T. Landsberg (North-Holland, Amsterdam, 1992) p. 28 1. 141 A. Franciosi and CC. Van de Walle, Surf. Sci. Rep. 214 (1996) 1.

607

[51 W.A. Harrison, E.A. Kraut, J.R. Waldrop and R.W. Grant, Phys. Rev. B 18 (1978) 4402. I61 S. Baroni, R. Resta, A. Baldereschi and M. Peressi, in: Spectroscopy of Semiconductor Microstructures, Eds. G. Faso], A. Fasolino and P. Lugli (Plenum, London, 1989) p. 251. c71 G. Biasiol, L. Sorba, G. Bratina, R. Nicolini, A. Franciosi, M. Peressi, S. Baroni, R. Resta and A. Baldereschi, Phys. Rev. Lett. 69 (1992) 1283. PI R. Nicolini, L. Vanzetti, Guido Mula, G. Bratina, L. Sorba, A. Franciosi, M. Peressi, S. Baroni, R. Resta, A. Baldereschi, J.E. Angelo and W. Gerberich, Phys. Rev. Lett. 72 (1994) 294. 191 See, for example: R.L. Gunshor and A.V. Nurmikko, Eds., ll_Vl Blue/Green Laser Diodes, Proc. SPIE 2346 (1994) 1. L. Vanzetti, L. Sorba, A. Franciosi, M. [lOI A. Bonanni, Lomascolo, P. Prete and R. Cingolani, Appl. Phys. Lett. 66 (1995) 1092. Cl 11 V. Pellegrini, A. Tredicucci, F. Beltram, L. Vanzetti, M. Lazzarino and A. Franciosi, J. Appl. Phys. 79 (1996) 929. Cl21 M. Lazzarino, T. Ozzello, G. Bratina, L. Sorba and A. Franciosi, Appl. Phys. Lett. 68 (1996) 370. 1131V. Pellegrini, M. Boerger, M. Lazzeri, F. Beltram, J.J. Paggels, L. Sorba, S. Rubini, M. Lazzarino, A. Franciosi, J.M. Bonard and J.D. Ganiere, Appl. Phys. Len. 69 (1996) 3233. Cl41 S.V. Meshkov. Sov. Phys. JETP 64 (1986) 1337. Cl51 V. Pellegrini, R. Atanasov, A. Tredicucci, F. Beltram. C. Amzulini, L. Sorba, L. Vanzetti and A. Franciosi, Phys. Rev. B 51 (1995) 5171. Cl61 S. Colak. T. Marshall and D. Cammack, Solid State Electron. 32 (1989) 647; D.J. Olego, Phys. Rev. B 39 (1989) 12743; Q.-D. Qian, J. Qiu, M. Kobayashi, R.L. Gunshor, M.R. Melloch and J.A. Cooper, Jr., J. Vat. Sci. Technol. B 7 (1989) 793. [I71 G. Bratina, L. Vanzetti, A. Bonanni. L. Sorba, J.J. Paggel, A. Franciosi, T. Peluso and L. Tapfer, J. Crystal Growth 159 (1996) 703. llS1 A.D. Raisanen, L.J. Brillson, L. Vanzetti, A. Bonanni and A. Franciosi, Appl. Phys. Lett. 66 (1995) 3301.