Modeling of tunnel junctions for current injection in Vertical Cavity Surface Emitting Lasers (VCSELs) Grigore Adrian IORDACHESCU, Joël JACQUET1 SUPELEC, Laboratoire Matériaux Optiques Photonique et Systèmes, Unité Mixe de Recherche - 7132, 2 rue Edouard Belin, 57070 METZ France Unité Mixte internationale Georgia Institute of Technology – Centre National de la Recherche Scientifique (CNRS), UMI 2958 ABSTRACT Tunnel junctions have been widely used in the fabrication of Vertical Cavity Surface Emitting Lasers since it allows fabrication of low electrical resistance as well as low optical absorption Bragg mirrors. The basic idea is to inject holes through a highly doped reverse biased n+/p+ tunnel junction. We present in this paper a review of the different materials that can be used for various wavelength applications ranging from UV (GaN) to IR (GaSb). We have elaborated a new modelling tool that has been validated for homo-junctions. The results show that the injection efficiency is directly linked to the energy gap of the material and to the effective mass of the electrons and light holes. The first important discussion is related to the condition on doping levels to get the material degenerate. Low band gap materials such as InAs or GaSb semiconductors are well appropriate to realise tunnel junctions with moderate doping levels. At the opposite large band-gap materials as GaN or AlN require very high doping levels to reach the tunnelling condition. GaSb based VCSELs emitting in the infrared region (2-3µm) can use very efficiently such electrical injection scheme. On the other side, it will be much less beneficial to use it for surface emitting laser emitting in the ultra violet wavelength range. Comparison with published papers will be discussed as well as preliminary work done in the case of hetero junctions. Keywords : Tunnel junction, Esaki Junction, semiconductor material, semiconductor lasers, Vertical Cavity Surface Emitting Lasers (VCSELs), 1. INTRODUCTION A better understanding of the tunneling conduction and its mathematical model has become increasingly important with the advent of VCSELs in the laser industry. This is due to the tunneling junctions’ capacity to enhance the VCSEL’s performances by significantly reducing its series resistance and, as a consequence, its Joule heating. The specific sheet resistence of the tunnel junction can become as small as 3⋅10-6 Ωcm2 for the case of InP-based semiconductor materials[1]. The exact employment of such a method will not be discussed here, but it could be found in References[1]. The purpose of this article is to present a review of two theoretical methods used to describe the tunneling conduction and to show some new interesting computer simulations which rely on them. One such method for tunneling current estimation is the Esaki integral, which sadly needs for its solution the value of the tunnel effect probability for all the energies in its domain of integration. The benefit for choosing it is that it works for both heterojunctions and homojunctions alike. Due to the ever-increasing importance of heterojunctions in the VCSEL industry, we’ll be more and more constrained to using this method in future simulations. As a result, we’ll give an account of it towards the end of this article. On the other hand, the method on which many of the simulations from this paper were based upon is the one by Demassa and Knott[2]. According to this method, the I-V characteristic of a tunnel diode is obtained by using a very good algebraic approximation of the experimental results (eq 1), in which the two unknown parameters of the tunneling current can be computed theoretically. 1 Joel.Jacquet@Supelec.fr Physics and Simulation of Optoelectronic Devices XVII, edited by Marek Osinski, Bernd Witzigmann, Fritz Henneberger, Yasuhiko Arakawa, Proc. of SPIE Vol. 7211, 721113 · © 2009 SPIE · CCC code: 0277-786X/09/$18 · doi: 10.1117/12.808468 Proc. of SPIE Vol. 7211 721113-1 2. SIMULATION TOOL For better grasping the signification of the two parameters, please refer to figure 1. (1) In order to widen the applicability domain of this method we have removed all the calculus appoximations from the original paper when computing Vp. Another thing one must pay attention to when using this method is some printing errors found in [2], and later on spread in [3, 4] regarding the use of Kane method for calculating Jp[5] (eq 2). J* Tuiiiiel '' 4 6/ /&Ex -'ICu Figure 1 : The J-V characteristic of a tunnel diode and the representation of Jp and Vp V p (2) For the signification of physical quantities in (2) please turn to the original paper by Demassa and Knott[2], where you’ll also notice the erroneous lack of the minus sign in the exponential. Sadly, the big limitation of this method when comparing to the Esaki integral is that its use is restricted to homojunctions. Nevertheless, its embodiment as a Matlab application yielded very interesting qualitative and even exact quantitative results (where the verification allowed) for such materials as GaSbyAs1-y, AlxGa1-xAs, GaxIn1-xAsyP1-y, Ge, InP, InSb, InN, InAs, GaSb, GaAs, AlN, GaN. We’ll make a run through all of them in the next section. Besides the universal constants, the specific physical quantities needed for each of those materials in order to apply this method are presented in Table 1. Their values have been taken or calculated from data made available by Ioffe Institute[6]. For ternary and quaternary compounds, the respective physical quantities have been written as functions of relative concentrations x and y. Proc. of SPIE Vol. 7211 721113-2 Table 1 : definition of parameters ε dielectric permittivity Eg band gap Mn electron effective mass for density-of-states calculations Mn,c electron effective mass for conductivity calculations Mp hole effective mass for density-of-states calculations Mlp light hole effective mass The validity of this method has been quantitatively verified in the case of Ge by Demassa and Knott[2] for all the positive values of the bias applied on the junction. A simpler way to check the results returned by this method is to do it for only the peak value of the direct tunneling current (Jp). This does not assure us that our plot is quantitatively correct for all bias voltages, but predicting the correct value for Jp is of most importance in realizing the strength of the tunneling effect. As an example of such practical confirmation, we’ll take the case of a GaAs tunnel diode[7] with a high density of deep-states of 1020cm-3. In this experiment a current density of 16kA/cm2 and a negative-conductance of (226Ωμm2)-1 are obtained[7]. If we look at the simulation results in figure 2, we’ll notice a peak tunnel current density of 15kA/cm2 and a negativeconductance of (330Ωμm2)-1. The orders of magnitude are similar but in general we would expect the experimental values be less performant than the theoretical ones because of the damage we inflict to the crystalline structure when we use such high levels of impurities. We don’t see such a phenomenon here though, partly because at high doping levels, even a slight deviation from set values may produce dramatic changes in the measured tunnel effect. So what should be followed next should be the order of magnitude and not the exact values of the simulation results. GaAs GaAs Na=1 .00e+020 cm3 Nd=1 .00e+020 cm3 Na=1.00e+020 cm3 Nd=1.00e+020 cm3 1 N E 0 0 500 1000 1500 V(mV) Proc. of SPIE Vol. 7211 721113-3 -40 -20 V(mV) -30 -10 0 Figure 2 – Matlab simulation of a GaAs tunnel diode with a high density of deep-states of 1020cm-3 3. DISCUSSION We have investigated various semiconductors materials commonly used in for laser fabrication. Table 2 summarizes the data we have used for these materials. The last two columns give the limit of the degeneracy (minimum doping level to get the Fermi level in the valence or conduction band for p and n doping respectively). Eg Mn Mlh (eV) Nd_min Na_min cm-3 cm-3 GaSb 0,726 0,04 0,05 3E+17 1E+19 GaN 3,45 0,2 0,26 3 E+18 1E+20 AlN 6,2 0,4 0,24 1E+19 1E+21 InAs 0,354 0,023 0,026 1E+17 1E+19 InP 1,344 0,073 0,089 1E+18 1E+19 InN 2 0,11 0,27 1E+18 1E+20 InSb 0,17 0,013 0,015 1E+17 1E+19 GaAlAs 2 0,1 0,11 1E+18 1E+19 GaAsSb 0,78 0,05 0,066 3E+17 1E+19 Table 2 : Energy gap and effective masses of electrons and light holes for the materials we have used. 3.1 Influence of the Energy gap on Tunnel condition We can observe that the degeneracy is reached with lower doping levels for n as compared to p material. The second observation is that the larger the energy bandgap, the higher the doping level needs to be to for the degeneracy to be reached. Doping levels as high as 1019 cm-3 are necessary for AlN n material (Eg=6.2eV) to have an efficient tunneling capacity. On the contrary, InSb can be used at a doping concentration around 2 orders of magnitude lower. For p doping, the concentration limit rises to 1019 cm-3 for AlN ; this means that this material will absolutely not be a good candidate to be used in optical devices. The minimum doping concentrations for the other materials are at least 1019 cm-3. All these observations are illustrated in figure 3. Proc. of SPIE Vol. 7211 721113-4 1,E+21 Nd or Na (cm-3) 1,E+20 1,E+19 1,E+18 Nd Na 1,E+17 0 1 2 3 4 5 6 7 Eg (eV) Figure 3 : Minimum doping level to reach degeneracy for p and n material with various bang gap energies 3.2 Influence of doping level on Tunnel characteristics. We have then investigated the performance of tunnel homojunctions made with these materials. We first calculated the values of Vp as a function of p doping concentration (figure 4) for a constant n doping of 1020 cm-3. Vp increases with this doping concentration and is lower for the higher band gap materials. We have after that taken the case of an InSb junction separately and calculated Vp as a function of p doping concentration for different n doping levels. The results can be seen in figure 5. 250 200 Vp (mV) 150 100 GaAlAs GaInAsP GaN 50 GaSb InP 0 0,0E+00 InSbSb 5,0E+19 1,0E+20 Na 1,5E+20 2,0E+20 (cm-3) Figure 4 : Calculated Vp as a function of p doping concentration for tunnel homojunctions made with the materials investigated in this work. Proc. of SPIE Vol. 7211 721113-5 300 250 Vp(mV) 200 150 100 Nd=1e+18 Nd=5e+18 Nd=1e+19 Nd=5e+19 Nd=1e+20 50 0 1E+19 1E+20 1E+21 Na(cm-3) Figure 5 : Calculated Vp as a function of p doping concentration for InSb tunnel homojunction at different n doping concentrations. 3.3 Influence of doping level on Jp Very often in the literature one often encounters a physical measure which comprises both the effects of n and p doping. This quantity is named the reduced concentration of impurities and it is written as n*. It is defined as the product of donor and acceptor impurity concentrations divided by their sum. In figure 6 we have reported exactly the influence of this measure on the maximum current density Jp, considering all impurities active (n=Nd and p=Na). Because this type of graphic is usually hard to follow, we have also calculated Jp as a function of p doping level (figure 7), the n doping level beeing kept at a constant value of 1020 cm-3, for the same materials as in figure 4. Proc. of SPIE Vol. 7211 721113-6 Jp (A/cm²) 1E+06 1E+05 1E+04 1E+03 1E+02 1E+01 1E+00 1E-01 1E-02 1E-03 1E-04 1E-05 1E-06 1E-07 1E-08 1E-09 1E-10 1,0E-10 AlGaAs GaInAsP GaN GaSb InP InSb 1,5E-10 2,0E-10 2,5E-10 3,0E-10 n*-1/2 (cm-3/2) Figure 6 : Calculated Jp as a function of n* Nd=1e+20cm-3 1E+06 1E+04 Jp(A/cm2) 1E+02 1E+00 1E-02 AlGaAs GaInAsP GaN GaSb InP InSb 1E-04 1E-06 1E-08 0E+00 1E+20 Na(cm-3) 2E+20 Figure 7: Calculated Jp as a function of p doping concentration for tunnel homojunctions made with the materials investigated in this work Proc. of SPIE Vol. 7211 721113-7 3.4 Influence of Eg on Jp and Vp We have plotted Vp and Jp for Ga(1-x) AlxAs n+p+ homojunction as a function of the relative concentration x. The impurities concentrations of the junction in this example are Na=Nd=5e+19cm-3. When x varies from 0 to 1, the energy gap covers the 1.4 to 3 eV range – figure 8. We observe that Vp is not strongly affected over the whole range. Opposing it, Jp decreases from 500 mV for GaAs material down to nearly 0 for AlAs composition (the exact value is in fact 0,25A/cm2, but because we are using a linear scale, we can’t differentiate it from 0). 600 Jp 500 Vp Jp (A/cm²) or Vp (mV) 400 300 200 100 0 1,4 1,6 1,8 2 2,2 2,4 2,6 2,8 3 Eg (eV) Figure 8 : Calculated Vp and Jp as a function of the relative concentration x for a homojunction of Ga1-xAlxAs doped with n=p=5e+19cm-3 3.5 Injection current in reverse biased junctions Finally, we arrived to the most important feature of this work: the tunnel current that can be injected in reverse biased tunnel junctions realised out of the different materials investigated. In this section, we calculate the I(V) characteristics. An example of such calculation is shown in figure 9. The result is as expected for a tunnel junction. When watching the next three figures, we have to be aware of the fact that we are neglecting the resistive effects given by the lengths of the two sides of the junction or by the resistive contacts we are Proc. of SPIE Vol. 7211 721113-8 using to mount the junction into a circuit. Those effects tend to become increasingly powerful as we increase the injected current into the junction. Ga(O.47)In(0.53)As(0.46)P(0.54) Na=5.00e+019 cm3 Nd=5.00e+019 cm3 -0.5 -00 0 200 400 V(mV) 600 800 Figure 9 : J(V) characteristic for forward and reverse biased InGaAsP tunnel homojunction. The region of interest is on the reverse bias side. We record the voltage value which is necessary to reach 10kA/cm² in the junction, V10. The best materials are the ones with the lowest V10 values; in that case the electrical consumption will be the lowest and the thermal heating reduced as well. The V10 values are reported in table 3 for six different homojunctions with doping level of 5⋅1019 cm-3 for both p and n type materials. V10 varies from 2 mV for InSb to 1.34 Volts for GaN based materials respectively. A clear correlation between V10 and the Energy gap of the material is observed. This is illustrated in figure 10. V @10kA/cm² (mV) Eg (eV) GaN InP InSb GaAlAs GaAsSb GaInAsP 1344 3,45 218,9 1,344 2 0,17 631,8 2 19,44 0,78 76,17 0,9 Table 3 : Necessary voltage values to reach 10 kA/cm² in the reverse biased 19 -3 tunnel homojunction of different materials doped with n=p=5⋅10 cm . In figure 11, we have plotted the values of the current injected in reverse biased homojunctions polarized at -1V for different materials with varying composition. Considering the high value of the calculated current density, it’s obvious that we are dealing with an ideal junction and that in a real experiment the thickness of the junction induces resistive effects that should be accounted for. Proc. of SPIE Vol. 7211 721113-9 1400 '4 1200 V (mV) @ 10kA/cm² 1000 800 600 400 200 0 0 1 2 3 4 Eg (eV) Figure 10 : Graphic showing the correlation between the necessary reverse voltage needed to reach a current density of 10kA/cm2 and the bangap of the base material on which the homojunction is realised. 1E+09 1E+08 J (A/cm²) @ -1 Volt 1E+07 1E+06 Al(x)Ga(1-x)As 1E+05 Ga(0.47)In(0.53)As(x)P(1-x) GaAs(1-x)Sb(x) 1E+04 0 0,2 0,4 0,6 0,8 1 X or 1-X Figure 11 : Value of the current injected in reverse biased homojunctions polarized at -1V for different materials with varying composition. Proc. of SPIE Vol. 7211 721113-10 4. PROSPECTS As we have seen in the previous section, a large number of possible simulations may be realised when we have a theoretical model that can be easily numerically implemented. The increasing role the heterojunctions are playing in this industry compel us to realize in their case the same kind of work we have done for the homojunctions. In this respect, we need a broader theory that can account for the effective mass and bandgap discontinuities between the two sides of those junctions. As we have already stated in the Introduction, the starting point of this endeavor is the original Esaki integral. We will use an improved form of it[8]: (3) where we prefer using β=(kT)-1 instead of the temperature for avoiding any confusion between the usual temperature notation and T(E), which represents the transmission probability (the probability for the electric carrier of energy E to surmount the potential barrier). This probability doesn’t depend on only the energy of the carrier, but also on the geometry of the potential barrier. Thus, in order to find T(E) we’ll need the solution of the Schrödinger equation (eq 4) in which V(x) coincides with the potential energy profile across the junction. (4) The method for solving eq. 4 by expanding it in a series of algebraic equations is to be found in [8]. At this stage the problem of not having a coherent theory for calculating the discontinuities in the energy levels of conduction and valence bands forces us to introduce a junction specific parameter (ΔEc – the discontinuity of the conduction band at the interface between the two materials). Thus, for all types of heterojunctions for which this parameter is measured, we’ll be soon able to obtain the same kind of graphics as for the homojunctions in the previous section. 5. ACKNOWLEDGMENTS This work has been realized in the frame of the ANR MIREV project. The authors would like to thank the Conseil Regional de la Lorraine (France) for financial support of this project. 6. REFERENCES: [1] Markus ORTSIEFER - Fabricating buried tunnel junctions for InP-based VCSELs. Solid State Technology, February 2006 [2] Thomass DEMASSA and David KNOTT – The Prediction of Tunnel Diode Voltage-Current Characteristics. Vol. 13 pp 131-138 Pergamon Press 1970 [3] Manish MEHTA, Danny FEEZELL, David BUELL, Andrew JACKSON, Larry COLDREN and John BOWERS – Electrical Design Optimization of Single-Mode Tunnel-Junction-Based Long-Wavelength VCSELs. IEEE Journal of Quantum Electronics Vol. 42, No. 7, July 2006 [4] Manish MEHTA – High-Power, High-Bandwidth, High-Temperature Long-Wavelength Vertical-Cavity SurfaceEmitting Lasers. PhD Dissertation – University of California Santa Barbara, June 2006 [5] Evan O. KANE – Zener Tunneling in Semiconductors. Vol 12 pp 181-188 Pergamon Press 1959 [6] Institut IOFFE. New Semiconductor Materials. Characteristics and Properties. 1998-2001 [7] Janet L. PAN, J.E. McMANIS, L. GROBER and J.M. WOODALL – Gallium-arsenide deep-level tunnel diode with record negative conductance and record peak current density. Solid-State Electronics, Volume 48, Issues 10-11, October-November 2004 Proc. of SPIE Vol. 7211 721113-11 [8] William R. FRENSLEY. Heterostructure and Quantum Well Physics. http://www.utdallas.edu/~frensley/technical/ 1995 Proc. of SPIE Vol. 7211 721113-12