* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Contact and Space-Charge Effects in Quantum Well Infrared
Survey
Document related concepts
Quantum electrodynamics wikipedia , lookup
Superconductivity wikipedia , lookup
Lorentz force wikipedia , lookup
Maxwell's equations wikipedia , lookup
History of electromagnetic theory wikipedia , lookup
Electron mobility wikipedia , lookup
History of quantum field theory wikipedia , lookup
Aharonov–Bohm effect wikipedia , lookup
Electromagnetism wikipedia , lookup
Electrical resistivity and conductivity wikipedia , lookup
Introduction to gauge theory wikipedia , lookup
Electric charge wikipedia , lookup
Mathematical formulation of the Standard Model wikipedia , lookup
Transcript
Jpn. J. Appl. Phys. Vol. 38 (1999) pp. 5815–5822 Part 1, No. 10, October 1999 c °1999 Publication Board, Japanese Journal of Applied Physics Contact and Space-Charge Effects in Quantum Well Infrared Photodetectors Victor RYZHII and H. C. L IU1 Department of Computer Hardware, University of Aizu, Aizu-Wakamatsu 965-8580, Japan 1 Institute for Microstructural Sciences, National Research Council, Ottawa, Ontario K1A 0R6, Canada (Received April 19, 1999; accepted for publication July 5, 1999) We present the results of a new self-consistent analytical model for the calculation of the dark current and the electric-field and space-charge distributions in quantum well infrared photodetectors (QWIPs). This model takes into account thermionic emission from the QWs, tunneling injection of electrons from the emitter contact, transport of the excited and injected electrons across an active region of a QWIP and their capture into the QWs. It is shown that the electric-field and space-charge distributions in the QW structure are nonuniform in general. The character of their nonuniformity is determined by the relationships between the structural parameters, parameters of elementary processes, and bias voltage. KEYWORDS: quantum well infrared photodetector, electric-field distribution, space charge, dark current 1. Introduction Quantum well infrared photodetectors (QWIPs) utilizing intersubband electron or hole transitions have attracted much attention over the recent years. The basic physics of QWIPs has been well documented.1–3) QWIP operation is associated with the photoexcitation of electrons (holes) from the bound state in the QWs to the continuum states above the barriers, the electron injection from the emitter contact, and the transport and capture of photoexcited and injected electrons. Thermionic emission of electrons from QWs determines the dark current. The collector contact enables the extraction of electrons from an active region of a QWIP. Different models have been proposed for steady-state and transient processes in QWIPs.4–19) Early models of QWIPs with multiple QWs are based on the assumption that the electric–field distribution in the active region is uniform.4, 5) Such simple QWIP models assume a perfectly injecting emitter contact,5) i.e., the contact injects as many electrons as needed by the QW structure. Thus, the injected current density should be that which yields the required rate of electron capture into the QWs to compensate the escape of electrons due to their thermoemission and photoexcitation from the QWs. A real emitter contact provides such an injected current density if the contact electric field has a proper value which can differ from the average electric field in a QWIP. An appropriate value of the electric field at the emitter contact is ensured by a space charge in the QW structure, which is created due to the difference in the electron and donor sheet concentrations in the QWs.8, 10–12) The space charge in the QW structure can be either positive or negative depending on bias voltage. Under certain bias voltages, determined by the parameters of the emitter contact, the space charge disappears, hence, the electric field becomes constant and coincides with the average field in the QW structure.12) The existence of a space charge in the QW structure results, in general, in the nonuniformity of the electric field. This effect has been considered previously for the cases of single20, 21) and multiple22) QWIPs. It has been demonstrated, using numerical simulations,8, 10, 11, 15) that in QWIPs with a large number of QWs, the region of nonuniform electric field can be relatively narrow and located near the injecting contact, while the electric field in the other part of the QWIP active region is nearly constant. The nonuniformity in photoexcitation due to photon flux attenuation adds complexity to the electric-field distribution.6, 9) It is instructive that in QWIPs in which the electric field weakly affects the emission of electrons from the QWs and their capture into the QWs, the electric field can be essentially nonuniform but rather smooth with a scale of nonuniformity comparable to the QW structure thickness.12) In such a case, the electron concentrations in all QWs are almost the same. The latter leads to a near-parabolic distribution of the electric potential and, consequently, a linear distribution of the electric field.12) This indicates that strongly nonuniform distributions, like distributions with high electric field domain near the emitter contact, are associated with relatively strong dependences of the electron excitation and capture rates upon the local value of the electric field. Properties of the emitter contact can also influence the space charge in the QW structure and, hence, the distribution of the electric field in it. Thus, the electric-field distribution in a QWIP and, consequently, its characteristics, are determined by both the emitter contact parameters and the field dependences of the rates of electron excitation from and capture into the QWs. The overall characteristics of QWIPs with multiple QWs can be rather insensitive to the emitter contact parameters in a wide range of applied voltages.12) Recently, a weak effect of the shape of the emitter barrier in QWIPs with a large number of QWs was confirmed experimentally.23) In contrast, the contact and space-charge effects in QWIPs with a moderate number of QWs can be essential in allowing some flexibility for the optimization of such devices. Apart from this, properties of the emitter contact can manifest themselves in nonlinear effects in QWIPs, in particular, at a high power of incident infrared radiation12, 15) (see also experimental results presented in refs. 24 and 25). Phenomena in QWIPs can be well described by numerical simulations based on either drift-diffusion or Monte Carlo models. However, QWIPs are characterized by numerous parameters. This hinders the use of numerical models for the extraction of physical information on the details of QWIP operation. From this standpoint, analytical models leading to explicit formulas for different characteristics of QWIPs can be beneficial. In this paper we present a study of contact and space-charge effects using a self-consistent analytical model for QWIPs with an arbitrary number of QWs and tunneling injection of electrons. We derive explicit analytical expressions as functions of bias voltage and structural parameters, taking into ac- 5815 5816 V. RYZHII and H. C. L IU Jpn. J. Appl. Phys. Vol. 38 (1999) Pt. 1, No. 10 count both the nonideality of the emitter contact and the influence of the electric field on the processes of electron thermoemission and capture. This paper is a significant generalization of a previous paper by one of the authors.12) 2. Device Model and Main Equations The QWIP under consideration consists of a multiple QW structure, sandwiched between the emitter and collector barriers, followed by heavily doped contact layers. The QW structure comprises thin doped narrow-gap QWs separated by thick undoped wide-gap barriers. An example of the conduction band diagrams of a QWIP is shown in Fig. 1. It is assumed that all the QWs, the barriers between the QWs, and the barrier between the extreme QW and the collector layer are identical. The barrier separating the emitter contact layer and the first QW (the top-most barrier or the emitter barrier) can differ from the interwell barriers in height. The thickness of the QWs (L w ) is much smaller than the thickness of the barriers (L b ). The thickness and the depth of the QWs are chosen in such a way that the QW has a single bound state while the first excited level lies near the top of the barriers.26) It is assumed that the electron injection from the emitter contact into the QW structure is due to tunneling. Thermionic emission is considered to be the main mechanism of electron escape from the QWs if the QWIP is not illuminated. The model of the QWIP is described by the Poisson equation and equations governing the electron balance in the QWs:12) N d 2ϕ 4π e X = (6n − 6d ) · δ(x − nL), dx 2 æ n=1 Gn = j pn . e (1) (2) Here ϕ = ϕ(x) is the self-consistent electric potential, 6n is the electron sheet concentration in the nth QW (n = 1, 2, 3, . . . , N , where N is the number of QWs), 6d is the g sheet concentration of donors in each QW, G n = G(E n , 6n ) is the rate of the electron escape from the nth QW, pn = p(E nc ) is the capture parameter (probability5) ) for an electron g traversing the nth QW, E n and E nc are the effective electric fields which determine the excitation rate from the nth QW and the capture rate into the latter (as seen below, these fields can differ from the average electric field and actual electric Emitter QW structure Collector Bound state L Fig. 1. Schematic view of the QWIP conduction band diagram under bias voltage. fields at the left-hand and right-hand sides of the QW), j is the current density, e is the electron charge, æ is the dielectric constant, L = L w + L b ' L b is the QW structure period, δ(x) is the Dirac δ function, and x is the coordinate in the direction perpendicular to the QW plane (in the growth direction). At temperatures near the temperature of liquid nitrogen, electron escape from the QWs is associated with thermoemission.4) The rate of electron thermoemission from the QWs increases exponentially with increasing local electric field, primarily due to the lowering of the barrier height with respect to the Fermi energy of electrons in the QWs, which leads to the decrease of the activation energy. The increase of the electron sheet concentration resulting in the elevation of the Fermi level also leads to the decrease of the activation energy. Hence, when thermo-emission is the dominant mechanism,14, 27–29) the rate of electron escape from the QW can be presented as · w g ¸ ²act (6n ) − e E n l g . (3) G(E n , 6n ) = g · exp − kB T Here, g is the preexponential factor, kB is the Boltzmann conw (6n ) = ²0w − ²Fw (6n ), ²0w and stant, T is temperature, and ²act w ²F are the energy of the bound state in the QW with respect to the barrier top (the QW ionization energy) and the Fermi energy of the two-dimensional electron gas in the QW, respectively, and l ' L w /2. For the degenerate two-dimensional electron gas (K = π h̄ 2 6d /mkT À 1) one has ²Fw = π h̄ 2 6n /m, where h̄ is the Planck constant and m is the electron effective mass. Taking this into account, the rate of thermionic emission from the QWs can be presented as · g ¸ En π h̄ 2 (6n − 6d ) , + (4) G(E ng , 6n ) = G · exp Eg mkB T where G = g · exp(π h̄ 2 6d /nkB T ) is the thermoemission rate when 6n = 6d and E nw = 0, and E g = kB T /el is the characteristic field of electron thermoemission from the QWs. At liquid nitrogen temperatures the inequality K À 1 occurs for GaAs QWs with 6d ≥ 5 × 1011 cm−2 . At lower temperatures, the two-dimensional electron gas is degenerate even at smaller concentrations. The term on the right-hand side of eq. (2) describes the capture of electrons into the QWs under the assumption of drift transport of electrons across the QW structure when the electron diffusion can be neglected. In such a case, the concentration of electrons in the continuum states is proportional to the current density. This is valid if the electric field in all parts of the QW structure is sufficiently strong. The QWIP model under consideration differs from that used previously12) to derive analytical expressions via the inclusion of the dependences of the excitation and capture rates on the electric field. To obtain the QWIP characteristics in closed analytical form some model field-dependence for the capture parameter p(E nc ) should be assumed. The mechanisms of the electron capture into QWs have been extensively studied (see, for example, ref. 30). However, the nature of the field dependence of the capture probability p(E n ) is not yet fully understood. In n-QWIPs made, for example, of A3 B5 heterostructures, the fraction of electrons in the 0-valley having the energy lower than the optical phonon energy can be small in comparison with the fractions of hot 0 Jpn. J. Appl. Phys. Vol. 38 (1999) Pt. 1, No. 10 V. RYZHII and H. C. L IU electrons and electrons in the L- and X-valleys, and markedly decreases with increasing electric field. Because the capture processes are mainly associated with the emission of optical phonons by low energy 0− electrons,31–33) the capture probability drops when the electric field increases due to the decrease of the number of such electrons. As shown by Monte Carlo modeling,18, 34) the drop of the capture probability can be notably strong. Taking this into account, we will use the following field dependence of the capture probability: µ ¶ E nc c , (5) p(E n ) = exp − Ec where E c is the characteristic electric field associated with the capture of electrons. The “exponential” can fit real dependences and is in good agreement with recent Monte Carlo simulations.34) It is necessary to stress that the rates of thermoemission g and capture depend on the effective electric fields E n and E nc , which are given by the local electric field averaged over some area in the vicinity of a particular QW. In the simplest case, these effective fields are determined by the real electric fields in the left (E n ) and right (E n+1 ) barriers. If the electric field is uniform, the effective fields coincide with the average electric field in the QWIP E = V /W , where V is the bias voltage and W = (N + 1) L is the thickness of the QW structure. g Generally, the effective electric fields E n and E nc are different from the average electric field and from each other. To take into account the nonlocality effects in question we assume that g,c E ng,c = E + αl (E n − E) + αrg,c (E n+1 − E). g,c αl g,c αr (6) and are coefficients. These coefficients deHere, termine the relative contributions of the real electric fields in the left barrier (with the subscript “l”) and in the right one g (with the subscript “r”) to the effective electric fields E n and c E n . In the case of uniform electric field in the QWIP when g,c E n = E n+1 = E, one obtains E n = E, i.e., the effective fields are equal to the average field. It is obvious that if the electric field in the left or right barriers becomes larger than the average value, the rate of thermoemission increases, g,c g,c while the capture rate drops. Thus, αl , αr ≥ 0. It is necessary to emphasize that the electron gas in the continuum states is strongly heated under conditions typical of QWIP operation (strong electric field and relatively low temperature). In such a situation, the characteristic length L ² of hot electron relaxation can significantly exceed the QW structure period L and be comparable to the QW structure thickness W . It is conceivable that in AlGaAs/GaAs QWIPs at liquid nitrogen temperature, the energy relaxation length is of L ² ' 102 –103 nm. Hence, for the usual period of the QW structure of L = 30–50 nm, L ² /L ' 2–30. This indicates that the association between the rates and the electric field can be g,c quite nonlocal. Strong nonlocality corresponds to small αl g,c and αr when the balance between thermoemission and capture is determined primarily by the average electric field. This case has been studied analytically in ref. 12. The model used g in numerical simulations11) conforms, in fact, to αlc = αr = 1 g c and αl = αr = 0 (see also ref. 15). It is reasonable to assume the boundary conditions in the form ¯ ¯ ϕ ¯¯ = 0, and x=0 ¯ ¯ ϕ ¯¯ = V. 5817 (7) x=W Because the current density j is the same throughout the QW structure, it coincides with the density of the tunneling current through the emitter barrier. Hence, taking into account the triangular shape of the emitter barrier in the presence of the electric field, one can write12, 16, 35) µ ¶ Et j = jm exp − , (8) Ee where E e = (dϕ/dx)|x=0 is the electric field at the emitter contact (in the emitter barrier, hence, E e = E 1 ), E t is the characteristic tunneling electric field, and jm is the maximum current density from the emitter layer. The quantity jm depends mainly on the doping level of the emitter layer. The characteristic tunneling field is determined by the activation energy of electrons in the emitter contact ²act = ²0 −²F , where ²0 is the height of the emitter barrier for electrons in the emitter layer, and ²F is the Fermi energy√of electrons in this layer. 3/2 This field is expressed as36) E t = 4 2m²act /3eh̄. If the electron tunneling is assisted by thermoexcitation, E t can depend on the temperature as well. The model under consideration yields a self-consistent description of the QWIP under dark conditions. Using this model, one can express the dark current in a QWIP as a function of the bias voltage V (the average electric field E = V /W ) and the characteristic fields related to the elementary mechanisms: E t (tunneling through the emitter barrier), E g and E c (thermoemission and capture, respectively). Similar models have been previously used in numerical simulations.8, 10, 11, 15) In contrast to those works, below we obtain the exact analytical solution of the equations of the model. 3. Dark Current Equations (2) and (4) yield 6n − 6d = ¤ æ £ · γ E gc ln i dark + γl (E − E n ) + γr (E − E n+1 ) . (9) 4π e Here, i dark = jdark /j0 (E) is the dark current reduction factor, jdark is the dark current density, j0 (E) = e G(E, 6d )/ p(E) = e G exp(E/E gc ), 4kB T , eaB E gc µ g ¶ 4kB T αl αc + l , γl = eaB Eg Ec µ g ¶ 4kB T αr αc + r , γr = eaB Eg Ec γ = where E gc = E g E c /(E g + E c ) and aB = æh̄ 2 /me2 is the Bohr radius. The quantity j0 (E) is the dark current density in a QWIP at the bias voltage V = (N + 1)E L À kT /e in the case of a homogeneous electric field. The deviation of the dimensionless dark current density i dark from unity indicates the deviation of the dark current density derived using our model from that obtained assuming a uniform electric field. The QWIP characteristics and, in particular, the form of the potential and space-charge distributions depend on the parameters γ , γl , γr and N . 5818 Jpn. J. Appl. Phys. Vol. 38 (1999) Pt. 1, No. 10 V. RYZHII and H. C. L IU Substituting 6n − 6d from eq. (9) into eq. (1) and integrating over dx, we obtain the following equation for the selfconsistent electric field in the nth barrier: · ¸ γ E gc ln i dark , E n = a n−1 E e + (1 − a n−1 ) E + (γl + γr ) (10) where a = (1 − γl )/(1 + γr ). Taking into account that the electric field in the nth barrier E n does not depend on the coordinate, boundary conditions (7) can be reformulated as L N +1 X E n = V. (11) n=1 Substituting E n from eq. (10) into eq. (11), one can obtain the equation which relates the emitter electric field E e , the average electric field E, and the quantity i dark as: Ee = E + γ (1 − b N ) E gc ln i dark , (γl + γr ) (12) where b N = (N + 1)(1 − a)/(1 − a N +1 ). Substituting E e from eq. (12) into eq. (8), we obtain an equation for the dark current reduction factor: γ (1 − b N ) Et E gc ln i dark , =E+ (13) ln α − ln i dark (γl + γr ) where ln α = ln[ jm /j0 (E)]. The latter equation can be rewritten as ln α = ln α0 − E/E gc , where α0 = jm /eG. Taking into account that usually ln α À 1 (due to the very large value of jm /eG), we obtain the following approximate solution for the dark current reduction factor: ¶ µ (γl + γr ) Et . (14) E− ln i dark = γ (b N − 1) E gc ln α From eq. (14) for the dark current density as a function of the average electric field (i.e., the bias voltage) and structural parameters one obtains ¶¸ · µ (γl + γr ) Et . (15) E− jdark = j0 (E) · exp γ (b N − 1)E gc ln α If (γl + γr )N ¿ 1, eq. (15) yields the same formula for the dark current as that obtained previously.12) This special case corresponds to the rather weak dependences of the thermoemission and capture rates on the local value of the electric field. Equation (15) can be simplified if N = 1 or N À 1. For QWIPs with N = 1, one has b1 (γ ) = 2/(1 + a), and from eq. (15) we obtain ¶¸ · µ (2 + γr − γl ) Et . (16) E− jdark ' j0 (E) · exp γ E gc ln α When N À 1 and |a| < 1, eq. (15) leads to ¶¸ · µ (1 + γr ) Et . E− jdark ' j0 (E) · exp γ N E gc ln α (17) 4. Derivation of the Electric-Field and Space-Charge Distributions Using eqs. (12) and (14), one can obtain the value of the self-consistent electric field in the nth barrier as: En = Et + ln α µ Et −E ln α ¶ b N (a n−1 − 1) . (b N − 1) (18) In particular, from eq. (18) one can obtain the following formula for the electric field E e = E 1 in the emitter barrier: Et . (19) Ee = ln α The electric fields in the collector barriers (E c = E N +1 ) in QWIPs with N = 1 in the limiting case of (γl + γr )N À 1 are given, respectively, by Et , E c = 2E − ln α µ ¶ (20) Et 1 + γr E− ' E. Ec ' E + (γl + γr )N ln α The obtained formulas result in the following equation for the electron sheet concentration in the nth QW: µ ¶ E t (a n−1 − a n ), b N æ E− . (21) 6n − 6d = 4π e ln α (b N − 1) In the limiting case of (γl + γr )N ¿ 1, eq. (21) results in µ ¶ Et æ E− , (22) 6n − 6d ' 2π eN ln α i.e., the different QWs are charged equally. Only a few QWs near the emitter contact are actually charged, forming a charged domain, while the bulk of the QW structure is quasineutral. The total charge accumulated in the QWs (per unit area) is given by µ ¶ Et æc N −E , (23) Q= 4π ln α where the factor c N = b N (1−a N )/(b N −1) varies from 2 to 1 with increasing (γl + γr )N . Hence, the magnitude of the total space charge Q is halved when (γl + γr )N varies from a small value to a large one. According to eq. (23), the total charge Q > 0 if E < E t / ln α, and Q < 0 when E > E t / ln α. If E ln α = E t , the total charge becomes zero, and the electricfield distribution becomes strictly uniform. The latter can be clearly seen from eq. (18). In such a case, as seen from eqs. (15)–(17), one obtains jdark = j0 (E). These conclusions were reached previously12) using a simplified QWIP model. 5. Examples The comparison of eqs. (15)–(17) reveals that in all limiting cases the dependences of the dark current upon the average electric field are very similar. The influence of the contact and space-charge effects on the dark current is described by the dark current reduction factor i dark = jdark /j0 (E). The latter is determined by the exponential factors in the righthand sides of eqs. (15)–(17). These factors describe the deviation of the current-voltage characteristic from that calculated under the assumption that the self-consistent electric field is uniform, i.e., without the inclusion of the effects in question. The dark current reduction factor also describes the dependence of the dark current on the number of QWs because j0 is independent of N . If the number of QWs is large enough, so that (γl + γr )N À 1 one obtains jdark ' j0 (E) ∝ exp(E/E gc ). In this limiting case, the dark current does not depend on the number of QWs or any parameter related to V. RYZHII and H. C. L IU Jpn. J. Appl. Phys. Vol. 38 (1999) Pt. 1, No. 10 This is in agreement with experimental results.23) In contrast to QWIPs with a large number of QWs, in QWIPs with a moderate number of QWs, the contact and space charge-effects markedly influence the dark current. If N is small, the dark current jdark can be markedly lower than j0 (see curves for N = 1 and 4 in Fig. 2). However, the contact and space-charge effects resulting in the change of the dark current value essentially do not modify its dependence on the average electric field. This is apparent from Fig. 2 where the ratio jdark /j0 (E) is almost independent of the electric field for QWIPs with the number of QWs varying in the range of N = 1–32. As shown in Fig. 3, the dark current is weakly dependent on the emitter parameters (which determine the tunneling field E t ) when N = 32, and is fairly sensitive to these parameters for a moderate N (see curves in Fig. 3 for N = 8). Figures 4–6 show the spatial distributions of the self-consistent electric field in the QWIPs (electric field versus normalized distance x/L) with different N , E t for different average electric fields E. These figures reveal that, generally, the electric field steeply decreases from high values in several barriers ad- 1.0 0.8 N=32 0.6 Reduction factor the properties of the contacts. The current-voltage characteristic is determined only by the “combined” characteristic field E gc = E g E c /(E g + E c ). The electric-field dependences of the dark current obtained experimentally for the n-type QWIPs made of AlGaAs/GaAs23, 37, 38) and InGaAlAs/InP39) heterostructures at liquid nitrogen temperatures in the range of the average electric fields E > 5 kV/cm correspond to E gc ' 5 kV/cm. Assuming that the thermoemission characteristic electric field is given by27) E g ' kT /el ' 8kT /5e L w , for L w = 5 nm and T = 77 K, we obtain E g ' 20 kV/cm. Thus, E gs is markedly smaller than E g . This indicates that the electron capture processes affect the current-voltage characteristics more strongly than thermoemission does (E c < E g ). Using the above data we obtain E c ' 6.7 kV/cm. This is in excellent agreement with the results of Monte Carlo modeling of AlGaAs/GaAs QWIPs.34) Figures 2–9 show examples of the dark current reduction factor as a function of the average electric field (i.e., as a function of the bias voltage) and the spatial distributions of the self-consistent electric field and charges in QWIPs with different numbers of QWs, N , and different tunneling fields E t at T = 80 K. The plots were calculated using the formulas obtained in the previous section. It was assumed that E gc = 5 kV/cm, ln α0 = 20, γl = 0.5–1.5, and γr = 0.5. The characteristic tunneling field E t was chosen to be in the range of 300–2000 kV/cm. The span of the variation of E t corresponds to the range of activation energies ²act for electrons in the emitter contact layer, ²act ' 70–240 meV. The minimum value of E t corresponds to a QWIP with a relatively low emitter barrier for electrons in the contact layer. The maximum value of E t corresponds to a QWIP with a rather high emitter barrier which can differ from the interwell barriers. The parameter ln α0 can be estimated as w /kB T . For AlGaAs/GaAs QWIPs ln α0 = ln( jm /eG) ' ²act 11 −2 with 6d = 5 × 10 cm designed for the wavelength range λ = 8–12 mµ, ln α0 ' 15–20. It is seen from Figs. 2 and 3 that in QWIPs with a large number of QWs, the dark current reduction factor, i.e., the ratio jdark /j0 (E), is on the order of unity in a wide range of electric fields. In QWIPs with N = 32 and quite different E t , the values of the dark current are very close to each other. 5819 Et=300kV/cm 1000kV/cm 2000kV/cm 0.4 1.0 N=8 0.8 0.6 0.4 10 20 30 Electric field (kV/cm) 40 50 Fig. 3. Dark current reduction factor as a function of electric field for QWIPs with different characteristic tunneling fields. 1.0 80 70 E=30kV/cm E=20kV/cm E=10kV/cm 60 Electric field (kV/cm) Reduction factor 0.8 0.6 0.4 Et=1000kV/cm N=32 8 4 1 N=8 50 Et=1000kV/cm 40 30 20 0.2 10 0.0 10 20 30 Electric field (kV/cm) 40 50 Fig. 2. Dark current reduction factor as a function of electric field for QWIPs with different numbers of QWs. 0 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 Normalized distance 1 2 3 4 5 6 7 8 Fig. 4. Electric-field spatial distributions in QWIP with eight QWs for different average electric fields (different biases). 5820 V. RYZHII and H. C. L IU Jpn. J. Appl. Phys. Vol. 38 (1999) Pt. 1, No. 10 80 80 70 E=30kV/cm E=20kV/cm 70 E=10kV/cm γl=1.5 60 Electric field (kV/cm) Electric field (kV/cm) 60 γl=1.0 γl=0.5 N=32 50 Et=1000kV/cm 40 30 30 20 10 10 2 8 14 20 26 32 2 8 14 20 26 32 2 8 14 20 26 0 32 Et=1000kV/cm E=30kV/cm 40 20 0 N=8 50 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 Normalized distance Normalized distance Fig. 5. The same as Fig. 4 but for QWIP with 32 QWs. 1 2 3 4 5 6 7 8 Fig. 7. Electric-field spatial distributions in QWIPs for different values of parameter γl . 140 Et=2000kV/cm Et=1000kV/cm 35 Et=300kV/cm Et=2000kV/cm 1000kV/cm 300kV/cm 30 100 N=8 E=30kV/cm Relative QW charge (%) Electric field (kV/cm) 120 80 60 40 20 0 25 20 N=8 E=30kV/cm 15 10 5 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 Normalized distance 1 2 3 4 5 6 7 8 Fig. 6. Electric-field spatial distributions in QWIPs with different characteristic tunneling fields. jacent to the emitter to relatively low values in the QWIP bulk and near-collector region. When E < E t / ln α and a > 0 (i.e., γl < 1), the electron sheet concentrations 6n in all the QWs are lower than the donor sheet concentration 6d . This follows directly from eq. (21). According to eqs. (21)–(23), such a situation corresponds to positive charges in QWs and in the QW structure as a whole. The QWs adjacent to the emitter are depleted more strongly than those in the QW structure bulk, hence, the space charge is located near the emitter contact (see Fig. 8). This is consistent with the results of previous numerical simulations.8, 10, 11, 15) However, in QWIPs with a small E t and average electric fields satisfying the inequality E > E t / ln α, the electricfield distributions can be quite different. As seen from Fig. 6, the distribution for E t = 300 kV/cm at E = 30 kV/cm corresponds to the low electric-field domain localized near the emitter and the relatively high electric field in the bulk. In this case, the QWs are charged negatively due to their enrichment with electrons. Figure 7 shows the spatial electric-field distributions for different values of the parameter γl and the fixed parameter γr (γr = 0.5). When γl = 1, the electric field in all the barriers, except the emitter barrier, is constant, and it is significantly smaller than that in the emitter barrier. It is interesting to note 0 5 1 2 3 4 5 QW index 6 7 8 Fig. 8. Space charge distributions for QWIPs with different characteristic tunneling fields. that if γl > 1, the electric-field distribution exhibits an oscillatory behavior. Such a distribution is shown in Fig. 7 (see curve for γl = 1.5). The relative charges of the QWs (6d − 6n )/6d for E t = 300–2000 kV/cm are shown in Fig. 8. It is seen that the QW charges are positive for E = 30 kV/cm and E t = 2000 and 1000 kV/cm when E < E t / ln α. In contrast, when E > E t / ln α, the QWs are charged negatively, as shown in Fig. 8 (see curve for E t = 300 kV/cm). In the latter case, the dark current reduction factor i dark exceeds unity. The average electric field corresponding to the change of the sign of the total charge, and consequently, to the transition from the high-field domain near the emitter to the lowfield domain, satisfies the equation E ln α = E t . Taking into account the explicit dependence of α on E, the latter equation can be presented as E ln α0 − E 2 /E gc = E t . This equation has a solution if the characteristic field E gc is large enough, namely, E gc ≥ 4E t / ln2 α0 . Figure 9 shows the charges of the QWs for different values of the parameter γl with γr = 0.5. These plots correspond to the electric-field distributions shown in Fig. 7. As shown in Fig. 9, when γl = 1, only the QW closest to the emitter V. RYZHII and H. C. L IU Jpn. J. Appl. Phys. Vol. 38 (1999) Pt. 1, No. 10 25 γl=0.5 1.0 1.5 Relative QW charge (%) 20 15 3) N=8 Et=1000kV/cm E=30kV/cm 10 4) 5 0 5) 5 10 Fig. 9. 1 2 3 4 5 QW index 6 7 8 The same as Fig. 8 but for different values of parameter γl . 6) contact is charged. This is also evident from eq. (21). It is interesting to note that if γl > 1 when a < 0, the sign of the QW charge oscillates as a function of the QW index. As seen from Fig. 9, the odd-numbered QWs are charged positively, whereas the charges of the even-numbered QW are negative. Such a charge distribution results in oscillatory electric-field distribution damping in the direction from the emitter to the collector. Large values of the parameter γl correspond to a strong influence of the electric field in the preceding barrier (left-hand-side barrier) on the electron balance in the QW. The damping oscillations of the self-consistent electric field and space charge can be explained as follows. A strong electric field in the emitter barrier leads to a small probability of electron capture into the first QW. As a result, the electron concentration in this QW significantly decreases and the QW is charged positively. The positive charge of the first QW results in a substantially smaller electric field in the next barrier compared with that in the first one. This, in turn, is the reason for the rather low capture rate into the second QW leading to the excess of electrons in the latter, i.e., to the negative QW charge. The negative charge of the QW increases the electric field in the next barrier and decreases the capture rate into the successive QW. 6. Conclusions We have proposed a self-consistent analytical model for QWIPs with tunneling injection of electrons through the extreme barrier. The QWIPs under dark conditions have been investigated for the temperature range where thermionic emission is the main mechanism of electron escape from QWs. Explicit analytical expressions have been derived for the dark current as well as for the electric-field and spacecharge distributions as functions of the bias voltage (average electric field), and for the QWIP structural parameters, in particular, the number of QWs. The following has been shown using the obtained analytical formulas. 1) The contact and space-charge effects weakly influence the dark current in QWIPs with a large number of QWs, whereas these effects can be essential in QWIPs with a moderate number of QWs. 2) The dark current-voltage characteristics of QWIPs with 5821 a large number of QWs weakly depend on the number of QWs and are primarily determined by the dependences of the thermoemission and capture rates on the average electric field in the QW structure. The dark currents in QWIPs with moderate and large numbers of QWs can be markedly different. The electric-field distributions in the QW structure are, in general, nonuniform regardless of the strength of the dependences of the rates of the elementary processes (thermoemission and capture) on the local electric field. The electric-field distributions in QWIPs with relatively strong dependence of the thermoemission and capture rates on the local electric field are partitioned into two regions with markedly different values of the electric field with the formation of a charged domain (positive or negative) adjacent to the emitter contact and quasi-neutral region in the bulk. The spatial distributions of the self-consistent electricfield and space charge can be oscillatory. Acknowledgments The authors thank M. Ryzhii and I. Khmyrova for essential assistance and useful comments on the manuscript. 1) B. F. Levine: J. Appl. Phys. 74 (1993) R1. 2) H. C. Liu: Long Wavelength Infrared Photodetectors, ed. M. Razeghi (Gordon and Breach, Amsterdam, 1996) p. 1. 3) K. K. Choi: The Physics of Quantum Well Infrared Photodetectors (World Scientific, Singapore, 1997). 4) S. R. Andrews and B. A. Miller: J. Appl. Phys. 70 (1991) 993. 5) H. C. Liu: Appl. Phys. Lett. 60 (1992) 1501. 6) V. D. Shadrin, V. V. Mitin, V. A. Kochelap and K. K Choi: J. Appl. Phys. 77 (1995) 1771. 7) V. Ryzhii and M. Ershov: J. Appl. Phys. 78 (1995) 1214. 8) M. Ershov, V. Ryzhii and C. Hamaguchi: Appl. Phys. Lett. 67 (1995) 3147. 9) V. D. Shadrin, V. V. Mitin, K. K. Choi and V. A. Kochelap: J. Appl. Phys. 78 (1995) 5765. 10) M. Ershov, C. Hamaguchi and V. Ryzhii: Jpn. J. Appl. Phys. 35 (1996) 1395. 11) L. Thibaudeau, P. Bois and J. Y. Duboz: J. Appl. Phys. 79 (1996) 446. 12) V. Ryzhii: J. Appl. Phys. 81 (1997) 6442. 13) V. Ryzhii, H. C. Liu, I. Khmyrova and M. Ryzhii: IEEE J. Quantum Electron. 33 (1997) 1527. 14) H. Schneider, P. Koidl, C. Schönbein, S. Ehret, E. C. Larkins and G. Bihlmann: Superlattices & Microstruct. 19 (1996) 347. 15) A. Sa‘ar, C. Mermelstein, H. Schneider, C. Schönbein and M. Walther: Intersubband Transitions in Quantum Wells, eds. S. Li and Y.-K. Su (Kluwer, Boston, 1998) p. 60. 16) V. Ryzhii, I. Khmyrova and M. Ryzhii: Jpn. J. Appl. Phys. 36 (1997) 2596. 17) M. Ryzhii and V. Ryzhii: Appl. Phys. Lett. 72 (1998) 842. 18) M. Ryzhii, I. Khmyrova and V. Ryzhii: Jpn. J. Appl. Phys. 37 (1998) 78. 19) M. Ryzhii, V. Ryzhii and M. Willander: J. Appl. Phys. 84 (1998) 3403. 20) E. Rosencher, F. Luc, Ph. Bois and S. Delaitre: Appl. Phys. Lett. 61 (1992) 468. 21) K. M. S. V. Bandara, B. F. Levine, R. F. Leibenguth and M. T. Asom: J. Appl. Phys. 74 (1993) 1826. 22) V. Ryzhii and M. Ershov: Jpn. J. Appl. Phys. 34 (1995) 1257. 23) H. C. Liu, L. Li, M. Buchanan and Z. R. Wasilewski: J. Appl. Phys. 82 (1997) 889. 24) M. Ershov, H. C. Liu, M. Buchanan, Z. R. Wasilewski and V. Ryzhii: Appl. Phys. Lett. 70 (1997) 414. 25) C. Mermelstein, H. Schneider, A. Sa‘ar, C. Schönbein, M. Walter and G. Bihlmann: Appl. Phys. Lett. 71 (1997) 2011. 26) A. G. Steele, H. C. Liu, M. Buchanan and Z. R. Wasilewski: Appl. Phys. Lett. 59 (1991) 3625. 27) A. G. Petrov and A. Ya. Shik: Semicond. Sci. Technol. 6 (1991) 1163. 5822 Jpn. J. Appl. Phys. Vol. 38 (1999) Pt. 1, No. 10 28) M. J. Kane, S. Millidge, M. T. Emeny, D. Lee, D. R. P. Guy and C. R. Whitehouse: Intersubband Transitions in Quantum Wells, eds. E. Rosencher, B. Vinter and B. Levine (Plenum Press, New York, 1992) p. 31. 29) H. C. Liu, A. G. Steele, M. Buchanan and Z. R. Wasilewski: J. Appl. Phys. 73 (1993) 2029. 30) E. Rosencher, B. Vinter, F. Luc, L. Thibaudeau, P. Bois and J. Nagle: IEEE J. Quantum Electron. 30 (1994) 2875. 31) J. A. Brum and G. Bastard: Phys. Rev. B 33 (1986) 1420. 32) J. M. Gerard, E. Deveaud and A. Regeny: Appl. Phys. Lett. 63 (1993) 240. V. RYZHII and H. C. L IU 33) 34) 35) 36) L. Thibaudeau and B. Vinter: Appl. Phys. Lett. 65 (1994) 2039. M. Ryzhii and V. Ryzhii: submitted to Jpn. J. Appl. Phys. V. Ryzhii: IEEE Trans. Electron Devices 45 (1998) 1797. S. M. Sze: Physics of Semiconductor Devices (John Wiley & Sons, New York, 1981). 37) J.-C. Chiang, S. S. Li, M. Z. Tidrow, P. Ho, M. Tsai and C. P. Lee: Appl. Phys. Lett. 69 (1996) 2412. 38) A. G. U. Perera, W. Z. Shen, S. G. Matsik, H. C. Liu, M. Buchanan and W. J. Schaff: Appl. Phys. Lett. 72 (1998) 1596. 39) C. Jelen, S. Slivken, V. Guzman, M. Razeghi and G. J. Brown: IEEE J. Quantum Electron. 34 (1998) 1973.