Sunteți pe pagina 1din 8

solid-Staft Electronics Vol. 24. pp. 195-202 Pergmon Press Ltd.. 1961.

Printed in Great Britain

TRANSPORT EQUATIONS FOR THE ANALYSIS OF HEAVILY DOPED SEMICONDUCTOR DEVICES


M. S. LUNDSTROM, J. SCHWARTZ J. L. GRAY R. and
School of Electrical Engineering, Purdue Univer$ty,
West Lafayette, IN 47907. U.S.A. (Received 12 June 1980;in revised form19 Sepfember 1980) AbstracGTransport equations for use in analyzing heavily doped semiconductor devices are considered. These transport equations describe the effects of the nonuniform band structure and the influence of Fermi-JIirac statistics, which are important in heavily doped semiconductors. Previous workers[ I.21 have derived transport equations in terms of the nonuniform band structure. These equations, however, are not convenient for use in semiconductor device analysis because the band structure of heavily doped semiconductors is not well known. In

this paper, the transport equations of Marshak and van Vliet[l, 31are recast into a simple, Boltzmann-like form in which the effects associated with the nonuniform band structure and degenerate carrier concentrations are described by two parameters, the effective gap shrinkage, Ao, and the effective asymmetry factor, y. The experimental determination of both of these parameters is also discussed. Finally, Adlers contention[4], that some important features of semiconductor device operation can be modeled accurately by using an electrically measured Ao with an arbitrarily chosen Y.is considered. The validity of this procedure, under certain simplifying assumptions, is established.
NOTATION

asymmetry factor effective asymmetry factor bandgap bandgap shrinkage effective bandgap shrinkage hole (electron) diiusion coefficient valence (conduction) band density of staies valence (conduction) energy band mobility edge hole (electron) quasi-Fermi level hole (electron) current density Boltzmanns constant valence (conduction) band effective density of states for a semiconductor with an unperturbed band structure intrinsic carrier concentration in a semiconductor with an unperturbed band structure effective carrier concentration hole (electron) carrier concentration magnitude of the electronic charge absolute temperature (Kelvin) electrostatic potential built-in potential applied voltage drop across the p-n junction reverse bias voltage hole (electron) mobility electron affinity the gradient operator with 7.. (nC) held constant Equilibrium values of the parameters listed above are denoted by a superscript or subscript o.
1.

INTRODUCTION

The analysis of heavily doped semiconductor devices is complicated by changes in the energy bands which occur in the regions of high doping density[S]. The resulting nonuniform band structure must be considered in order to model p-n junction devices accurately[5-8]. The purpose of this paper is to discuss the transport equations for materials with a position-dependent band structure. The emphasis is on the development of transport equations in a form suitable for use in analyzing heavily doped semiconductor devices.. Transport equations for materials with a
195

positiondependent band structure have been derived by van Overstraeten d al.[2], Mock[9], and Marshak and van Vliet[l, 3,101. However, since the band structure of heavily doped semiconductors is not well known, these equations are not convenient for use in semiconductor device analysis. In a recent paper[lO], Marshak and van Vliet recast their equations into a form containing the effective intrinsic carrier concentration, a parameter which has been determined experimentally. Even in this form, however, their equations are restricted to the nondegenerate case and involve the intrinsic level (which depends on the unknown band structure). Because of the limited knowledge of the band structure of heavily doped semiconductors, it has been necessary to use a simplified approach in order to semiconductor analyze devices. A common approach [7,8] assumes no change in the shape of the density of states function (i.e. the rigid band approximation), Boltzmann statistics, and equal distribution of the change in bandgap between the conduction and valence bands. In the resulting equations, the effects associated with the heavy doping are described by the effective intrinsic carrier concentration, which has been determined experimentally[l l-141. Although this approach has resulted in accurate modeling of heavily doped semiconductor devices[7, 81, its validity can be questioned. The assumption that the band edges shift equally, for example, is unnecessary if an asymmetry factor is introduced [ 151.The resulting transport equations can be written as J,=-pqc~p and J.=-n~~[V(V+A~)]+kTlr.Vn.
(2)

V V-(1-A)? H

- kTw,VP

(1)

1%

M. S. LUNDSTROM a/. ef

The bandgap shrinkage,

Dirac statistics are correctly accounted for without assuming a particular band shape.

AEG =-(AE,-AEv),

(3)
2. ENERGY BANDS IN HEAVILY DOPED SEMICONDUCTORS

is related to the effective intrinsic carrier concentration, ni,. by


2 (A&/kT) 2_ ni, - ni, e

The asymmetry factor.


A=$-,
G

where ,y is the electron afinity, measures the fraction of the reduction in bandgap that occurs in the conduction band. The terms in (1) and (2) that involve the bandgap shrinkage can be regarded as quasi-electric fields [ 161that affect the transport of holes and electrons. Although Boltzmann statistics were assumed to derive (1) and (21, Adler [4,7] argues that the use of an experimentally determined AE, in these equations correctly accounts for Fermi-Dirac statistics. In addition, he argues that accurate models of heavily doped semiconductor devices can be obtained by choosing the asymmetry factor arbitrarily. An alternative approach has been taken by Marshak ef al. [IS] and by Shibib et al. [17] who retained the rigid band approximation but modified (1) and (2) to incorporate Fermi-Dirac statistics. According to these authors, the asymmetry factor must be known in order to model devices in which one of the carriers is degenerate. Here, the equations developed by Marshak and van Vliet[l, 3, lo] are recast into a simple form in which the effects of heavy doping, the modified band structure and influence of Fermi-Dirac statistics are completely described by two parameters. The first parameter, the effective gap shrinkage, has been inferred from electrical measurements, and a method to determine the second parameter, the effective asymmetry factor, is proposed. The form of these equations is identical to that of (1) and (21, but the definitions of the two parameters have been generalized. In the approach advocated here, no assumption concerning the shape of the density of states needs to be made, and since the effects of carrier degeneracy are included in the two experimentally determined heavy doping parameters, no explicit correction for Fermi-Dirac statistics is required. In addition, arguments are presented to show that some aspects of typical semiconductor devices can be modeled accurately without knowing the effective asymmetry factor. The conclusion of this paper, therefore, is that Adlers approach[4,7] to the modeling of heavily doped semiconductor devices is justified. The most important result of this work, however, is that rigorous transport equations for materials with position-dependent band structure can be written in terms of two parameters which can be experimentally determined. These equations are written in a simple form which permits their use in semiconductor device analysis. Both the modified band structure and the influence of Fermi-

In a lightly doped semicondutctor, the impurities introduce localized states within the bandgap and do not perturb the conduction or valence bands. If the doping density is increased, however, the wave functions of neighboring doping atoms overlap and cause the impurity states to become extended and to broaden into a band. At room temperature the impurity band does not play an important role in the conductivity of lightly or moderately doped semiconductors because the impurity states have low mobility and are mostly unoccupied[ 181. For semiconductors with a high doping density the influence of the impurity band cannot be ignored. In fact, when the doping density exceeds about 10 cmm3, the impurity band overlaps the conduction band and the concept of separate bands loses meaning[l8]. The randomly fluctuating potential of the impurities produces tails in both the conduction and valence bands, but deep within the bands the shape is unchanged. It is also important to note that even in the absence of dopant atoms free carriers modify the bands of a semiconductor. The semiconductor bandgap and effective masses are altered by electron exchange interactions and screening of minority carriers [18,19]. Little is known about the band structure of heavily doped semiconductors, but the existence of mobility edges, analogous to those found in amorphous materials, is expected. The mobility edges, EC and E,, are indicated in Fig. 1, a model band diagram which will be employed for this analysis. States located below EC and above E, (i.e. deep within the band tails) are localized and do not

E co

Ec E
E

vo

Fig.

1. Model

energy

band

diagram conductor.

for a heavily

doped

semi-

Transport equations for the analysis of heavily doped semiconductor

devices

197

participate in conduction at room temperature [ 181.Since band structure calculations indicate that the number of these localized states is very small[l8], the impurities in a heavily doped semiconductor are almost fully ionized. The transition from localized to extended states is assumed to occur abruptly at the mobility edge. Although several calculations of the energy bands in a heavily doped semiconductor have been performed, many uncertainties remain [ 181.Since experimental work is also quite limited, the band structure of a heavily doped semiconductor must be regarded as essentially unknown. In this paper, the existence of sharp mobility edges is assumed, but the shape of the band is not specified. It will be shown that the modifications to the transport equations due to heavy impurity doping can be expressed in terms of two parameters and that by making approximations suitable for heavily doped semiconductors, these parameters can be determined experimentally. The practice of assuming a particular band shape, as is done in the frequently-used rigid band approximation, will be shown to be unnecessary.
3. TRANSPORTEQUATIONS

The mobility edges are written as[l] E,=E,-x-EG-qV and E,=E,-x-qV, (13) (12)

where E, is the field-free vacuum level and V is the electrostatic potential. The relationship between the gradient of the band edges and the gradient of the electrostatic potential V, is given by VE,=-qVV-V(xtE,) and VE, = -qVV-Vx. (15) (14)

The hole and electron current densities for isothermal conditions are given by:

J, = PILP~&
and J. = np.VF,.

(6)

Note that the electrostatic potential in a material with a position-dependent band structure is not parallel to the intrinsic level [lo]. By using (6)-W), transport equations for materials with positiondependent band structure can be written in a form similar to that obtained by Marshak and van Vliet [ 1,3,10]. These equations, derived in the Appendix, are

(7) and

JP =-pqcLI, v

[(vty-$ )I d?oVp
(16)
(17)

F, and F. are the quasi-Fermi levels for holes and electrons, and cup and CL.are the carrier mobilities. The validity of these equations for materials with a positiondependent band structure has been established by Marshak and van Vliet[l]. The hole and electron concentrations, p and n, are evaluated from the densities of states in the valence and conduction bands, D, and DC,and the Fermi function, f. For the band structure of Fig. 1, the result is L&E -E, xl it1 -fE and
n=

where (18)

F,ME

(8)

and (19)

&op

D,(E - E,, x)f(E F.)dE.

(9) are the Einstein relations, and . (20)

EC

Because the temperature is constant, the hole concentration can be expressed in terms of two independent variables, position, x and ny, where q =(E.-F,) kT

(10)

and VI, =+V% ( a9c > (21)

Similarly, the electron concentration is a function of x and nC, where

(11) are the so-called density-of-states effects[l, 3, lo].

198

M. S. LUNDSTROM al. et metry factor have been generalized in (26) and (27) to account for an arbitrary band shape and the influence of degenerate carrier statistics. Electrical measurements to determine these parameters in heavily doped semiconductors will be discussed in Section 5. 4. THE EQUILIBRIUM
p PRODUCT

Marshak and van Vliet have related these transport equations to those obtained by others[lO]. Since it is difficult to experimentally separate the effects caused by the modified band structure from those due to the influence of Fermi-Dirac statistics, it is convenient to move the effects of degeneracy in the diffusion terms of (16) and (17) to the drift terms. In the Appendix, two parameters, 8, and 8., are defined in order to write the transport equations as

x_E,
4 4 )I

and J.=-nq~.[V(V+$+~)]+kT~.Vn.

The equilibrium product of carrier densities, nope. is of interest because it can be inferred from experimental measurements and can be related to the band structure theoretically. In equilibrium the current densities, J, and k%pVp (22) J., are zero and (22) and (23) can be solved for the equilibrium carrier concentrations, n, and po. The resulting nope product is (23) nope = ni, exp
2

AEG t @,Oi 6. kT I

P-4

In the above formulation of the transport equations, the terms involving x and E. account for the positiondependence of the band edge (i.e. the rigid band effect). The terms, 8, and 8., account for both the modified band shape (i.e. the density-of-states effect) and the influence of Fermi-Dirac statistics. In a lightly doped material with standard band structure, 8, and 8. are defined to be zero. If Boltzmann statistics are used, then VB, = VT, and V8. = Vr, (see the Appendix). For use in semiconductor device analysis, it is convenient to recast the transport equations in the form: J, = -pq/& and J.=-nq~.[O(V+y~)]~kT~Vn, where AG = (AEo + 8, + 8,) is the effective bandgap shrinkage and y-Ax=@
AC

For lightly doped semiconductors, the equilibrium np product is constant, equal to the square of the intrinsic carrier concentration, ni,,, but for high doping densities the product becomes doping dependent. The concept of an intrinsic carrier concentration has been generalized by defining an effective intrinsic carrier concentration[2.3,51 ni,* = n,p,. (29)

v v-(1+

VP )I-kTwp
(24) (25)

(26)

The effective intrinsic carrier concentration is equal to ni, for low doping, but when the doping density is large, the effects of band tailing and gap shrinkage increase n;,. The influence of Fermi-Dirac statistics is included in ni, and acts to oppose the influence of band tailing and gap shrinkage at high doping densities. When the semiconductor is nondegenerate, Boltzmann statistics can be used to replace the 8 terms in (28) with the corresponding T terms. As a result, for nondegenerate samples, (28) reduces to the expression obtained by Marshak and van Vliet [3]. The effective intrinsic carrier concentration defined by (28) however, is not restricted to the nondegenerate case. Finally, using (26), the effective intrinsic carrier concentration can be related to the effective gap shrinkage by 2= ni,
nio2 eA&kT,

(27)

(30)

is an effective asymmetry factor. If Boltzmann statistics and the rigid band approximation are assumed, A0 reduces to the actual bandgap shrinkage, AEo. and y reduces to the actual asymmetry factor, A. Although these equations are expressed in a simple form, they correctly describe carrier transport in materials with a position-dependent band structure. The existence of mobility edges was assumed in the derivation of these equations, but no particular band shape was specified. The two parameters, A0 and y, describe the effects of the nonuniform band structure and the influence of Fermi-Dirac statistics. Note that the transport, (24) and (2% are identical in form to (1) and (2) but the definitions of the gap shrinkage and asym-

which is a generalization of (4). The assumptions of a rigid shift of the bands and Boltzmann statistics that were used to obtain (4) have been eliminated in (30) by replacing the bandgap shrinkage with an effective bandgap shrinkage.
5. ELRCTRICAL CHARACTRRIZATION OF HEAVY DOPING EFFECTS

Electrical measurements of the parameters AG and y, which describe the position-dependent band structure and the influence of Fermi-Dirac statistics, are considered in this section. Only semiconductors for which the nonuniform band structure is due to the presence of heavy impurity doping are considered. The dopants are assumed to be fully ionized. The heavily doped regions

Transport equations for the analysis of heavily doped semiconductor

devices

199

are also assumed to be quasi-neutral and in low injection so that the majority carrier concentration is nearly equal to the impurity doping density. We further assume that the band parameters, A0 and y, remain fixed at their values in equilibrium and do not change when the device is not in equilibrium. The assumption of complete ionization of impurities is justified at high doping densities because the overlap of wave functions prevents localized levels from occurring. Although localized states do exist below the mobility edge (deep within the band tails), calculations show that the number of these states is small[l8]. Experimental evidence also suggests that in heavily doped silicon, most dopants are ionized. For example, measurements of the Hall effect[20] and specific heat[21] of heavily doped silicon show that the free carrier concentration is roughly equal to the doping density. A discussion of other evidence supporting the view that impurities in heavily doped silicon are mostly ionized can be found in Ref. 1221. A. Electrical determination of AG Bandgap narrowing has been invoked to explain why conventional semiconductor device theory overestimates the gain of bipolar transistors[6]. Measurements of the collector current of bipolar transistors have subsequently been used to study the effects of heavy doping in silicon [ 1l-141. In this section, these electrical measurements are related to the band structure of a heavily _ doped semiconductor. The collector current of a bipolar transistor is related to the injected minority carrier concentration in the base. If the base is thin, recombination can be neglected, and the excess minority carrier at the emitter-base junction can be determined. Using the transport, (24) and (29, a law of the junction(231 can be derived to relate the excess minority carrier concentration to the equilibrium minority carrier concentration in the base. Therefore, this electrical measurement determines the equilibrium minority carrier concentration. Since the majority carrier concentration is assumed to be equal to the doping density, measurement of the equilibrium minority carrier concentration permits n,p, to be determined. The equilibrium effective gap shrinkage can then be determined from the inferred &,p, and (30). The effective gap is the experimental result usually shrinkage quoted[ll, 121. In practice, several complications may occur. For example, if the doping is nonuniform, only an integral that involves ni. is determined (see [3] eqn (4.11)). However, an analysis of measured currents in bipolar transistors that is based on (24) and (25) should permit at least limited characterization of AG. It should also be possible to characterize AG by using (24) and (25) to interpret other experimental measurements (e.g. [ 121). The effective gap shrinkage is related to the band structure of the heavily doped semiconductor by (26) which shows that the experimentally determined parameter, AG, is not equal to the actual reduction in bandgap, AEGo. The assumption, AGo= AEGo, is valid only if no change in the shape of the bands occurs (i.e. a

rigid band shift) and if the use of Boltzmann statistics is justified. Since the effect of band tailing may be viewed as an increase in the effective densities of states of the conduction and valence bands, 8, t 8, is expected to be positive for heavily doped semiconductors. Consequently, the parameter inferred from electrical measurements, AGo, should be larger than the actual reduction in bandgap, AEGo. It has been pointed out recently that electrical measurements of AEG yield values which are too large to be consistent with theoretical predictions [24]. An explanation that is consistent with this observation is that significant band tailing occurs and thereby invalidates the assumption that AGO = AEGO. Although Boltzmann statistics and the rigid band approximation are frequently used to analyze experimental results[ll-141, an analysis which uses the transport, (24) and (29, is formally identical, but makes neither assumption. The experimentally inferred parameter, AGo, however, must not be interpreted as the actual change in bandgap. B. Electrical determination of y The electrical measurement of AGo does not provide a complete description of carrier transport. Fortunately, as will be discussed in Section 6, even the limited information available in AGo is sufficient to accurately mode1 many features of semiconductor devices. However, a complete description of carrier transport would be desirable, and in this section a technique to measure the equilibrium value of y is proposed. Since the built-in potential, Vbi. of a p-n junction occurs in order to align the Fermi levels in the p and n regions, a measurement of Vbi provides additional information about the band structure of heavily doped semiconductors. If the current densities are set to zero, the equilibrium carrier concentrations PO=aiOeXp -4 VO-(l-y)? [ ( and
n, =ni, exp q V, + 7.y

/kT >

(31)

[(

(32)

are obtained from (24) and (25). The built-in potential is found by assuming pO= N.,, at one contact and n, = ND at the other. The result is

Vbi = 4
kT

NAND In 7 - ye [ 10

1 .a,,o
4

where the subscripts P and N refer to the p and n contacts respectively. Equations (33) shows that Vbi is reduced by effects of heavy doping. This reduction in Vbi has been observed experimentally [2]. The built-in potential of the p-n junction can be determined .using reverse biased capacitance measurements. As an example, for an abrupt junction the voltage

200

M. S. LUNDSTROM al. et

intercept of (I/C) vs V,, can be related to Vbi[25). If a low doped p-type material is used. AGpOin (33) is zero. As a result, measurement of Vbi allows yNoAGNoto be determined. Since AoN may be determined from independent measurements of ni,, y0 can be determined in the n-doped region. Similarly, use of a low doped n-type material permits determination of y0 in heavily doped p-type regions. These measurements can be performed as a function of the doping density on the heavily doped side of the junction to measure y0 as a function of doping density. The measurement of Vbi, when combined with that of nopor permits the equilibrium values of the terms in the transport equations which arise from the effects of heavy doping to be determined. These experimentally determined parameters account for the modified band structure and the influence of Fermi-Dirac statistics. If the values of the heavy doping parameters do not change significantly out of equilibrium, then these experimentally determined parameters may be used in the transport, (24) and (2.5). to model heavily doped semiconductor devices. 6. APPLICATION DEVICE TO ANALYSIS The authors purpose has been to recast the semiconductor transport equations into a form that is convenient for use in modeling heavily doped semiconductor devices. The use of these equations will now be discussed. A semiconductor device is modeled by solving Poissons equation and the hole and electron continuity equations in conjunction with the transport eqns (24) and (25)[5]. The solution to these equations gives the elect;ostatic potential, V, and the hole and electron carrier densities p and n, from which the current through the device may be evaluated. In order to obtain this solution, both of the position dependent bandgap narrowing terms, AG and y, must be known. Because these parameters are not well-characterized, several simplifying assumptions are typically employed in order to model heavily doped semiconductor devices. An important simplification is the use of an arbitrarily chosen y[4]. In this section we argue that the use of an arbitrary y results in accurate modeling of some features of typical heavily doped semiconductor devices. A complete model, however, must await the accurate experimental determination of the two heavy doping parameters. Complete ionization of the dopants is the first assumption that will be made. The validity of the assumption has been discussed in Section 5. The heavily doped regions of the device are also assumed to be quasi-neutral and in low injection so that the majority carrier concentration is nearly equal to the doping density. Because they include the effects of degeneracy and since the band structure itself depends on the free carrier densities[l8,19], the heavy doping parameters, AG and y, are functions of the carrier concentrations. Techniques to determine the equilibrium values of these parameters were discussed in Section 5. We now assume that ihe values of these parameters do not change when the device is not in equilibrium. This is a reasonable assumption because we are considering only heavily

doped regions that are in low injection and therefore, in some sense, near equilibrium. The assumptions employed here are those that were used in Section 5 in order to permit the determination of the heavy doping parameters by electrical measurements. . If the assumptions made above are valid, heavily doped semiconductor devices can be modeled in terms of the two equilibrium heavy doping parameters, AoO and yO. Unfortunately, no measurements of the effective asymmetry factor have been reported. When the semiconductor equations are solved numerically, however, it is observed that the resulting carrier concentrations and current voltage characteristics are not sensitive to the choice of y [7]. This observation has led Adler to contend that semiconductor devices can be modeled accurately without knowing y[41. We now consider the effect of y on the solution by discussing simplified, analytic solutions to the semiconductor equations. In a heavily doped region in low injection, the majority carrier densities and the effective fields are near their values in equilibrium. Consequently, in a heavily doped n-type region, we may assume n = hrD and set J. = 0 and solve (25) for -vv~-v(v+y~)s-~~, (34)

which is the effective field (electric plus quasi-electric) for electrons. The effective field for holes is simply -

v VPx-v

(v-(l-y)!$)~-vv+v (A$.>
(35)

Equation (35) can then be used to write the minority carrier current density, (24), as

(36)
Subject to the assumptions made above, the quasi-fields for both carriers, and the minority carrier current density are correctly evaluated independently of how y is chosen. In order to solve for the steady-state-minority carrier concentration, (36) is inserted into the hole continuity equation: V-J,=q(G-R), (37)

where G, the generation rate of carriers and R, the rate at which they recombine, are assumed to be independent of the effective asymmetry factor. When boundary conditions on p are specified, (37) can be solved for the hole concentration. The relationship between the minority carrier concentration at the edge of the space-charge region and the applied voltage under conditions of low injection,

Transport equations for the analysis of heavily doped semiconductor devices

201

where Vi. is the applied junction voltage, is obtained by integrating (24) across the space-charge region[23]. Equation (38) demonstrates that the correct relationship between the junction potential and the minority carrier concentration at the edge of the space-charge region is obtained regardless of the choice for y in the transport equations. When the minority carrier concentration or its normal derivative is specified along the other boundaries of the heavily doped region, (37) may be solved for the minority carrier concentration. The contribution of the heavily doped region to the device current is then obtained by evaluating (36) at the edge of the spacecharge region. The discussion presented above shows that the carrier concentrations, minority carrier current density and the current-voltage characteristic of typical heavily doped semiconductor devices are correctly modeled even when the choice for y in (24) and (25) is made arbitrarily. This result is also observed when the semiconductor equations are solved numerically[7]. Our work, therefore, supports Adlers contention that some features of semiconductor device operation can be modeled accurately without knowing the effective asymmetry factor[4]. However, although the effective fields are correctly modeled, y must be known in order to evaluate the electrostatic potential. The common practice of using an arbitrary 7, therefore, does not result in a complete model for semiconductor devices. For example, the builtin potential of a p-n junction cannot be evaluated unless y is known. In addition, the accuracy of this simplified approach may be questioned if the junction voltage approaches Vbi or if quasineutrality is violated. Finally, it should be noted that the validity of some key assumptions (e.g. that the heavy doping parameters do not change significantly from their values in equilibrium) have not been established.
7.CONCLUSlONS

shown that the carrier concentrations and current densities in semiconductor devices with quasi-neutral, heavily doped regions operating at voltages well below the built-in potential, can be modeled accurately, independently of the choice of 7.
Acknowledgements-This work was supported by the United States Department of Energy through Sandia National Laborstories, contract number U-2304. We would like to thank Dr. H. T. Weaver of Sandia National Laboratories for several clarifying discussions. The authors also benefited from discussions with Dr. A. H. Marshak concerning the relationship between the electrostatic potential and the band edges and concerning the approximations used for device analysis.

REFERJTNCES

I.

A. H. Marshak and K. M. van Vliet, Solid-St. Electron. 21,

Transport equations for materials with a positiondependent band structure have been discussed. In order to derive these equations, the assumption was made that mobility edges for the conduction and valence bands existed, but no specific band shape was specified. These general transport equations, which are indentical in form to those derived by Marshak and van Vliet[3], were then recast into a simple form for use in semiconductor device analysis. In this form, the effects of the nonuniform band structure and the influence of Fermi-Dirac statistics a;e described by two parameters, the effective gap shrinkage, AC, and the effective asymmetry factor, y. Simplifying assumptions were made, and the experimental determination of both of these parameters in heavily doped semiconductor devices was discussed. These parameters correctly account for the modified energy
bands and influence of Fermi-Dirac statistics which

417427 (1978). 2. R. J. van Overstraeten, H. J. DeMan and R. P. Mertens, IEEE Trans. Eltcrron Dem.ED-20.290-298 (1973). 3. A. H. Marshak and K. M. van Vliet, Solid-k klectron. 21, 429-434 (1978). 4. M. S. Adler, Proc. NASACODE I Conf.. pp. 3-30, Trinity College, Dublin, lreiand (1979). 5. F. A. Lindholm and C. T. Sah, IEEE Trans. Electron Dee. ED-24,299-304 (1977) 6. H. J. J. De Man, IEEE Trans. Electron Deu. ED-II, 833-835 (1971). 7. M. S. Adler, Tech. Digest, pp. 550-555 IEDM, Washington, D.C. (1978). 8. P. Lawers, J. van Meerbergen, P. Bulteel, R. Mertens and R. van Overstraeten. Solid-St. Electron. 21.741-752 (1978). (19?3). 9. M. S. Mock, Solid-St. Electron. 16, 1251-1259 IO. K. M. van Vliet and A. H. Marshak, Solid& Electron. 23, 49-53 (1980). II. J. W. Slotboom and H. C. De Graaff, Solid-St. Electron. 19, 857-862(1976). 12. F. A. Lindholm, A. Neugroschel, C. T. Sah, M. P. Godlewski and H. W. Brandhorst, J., IEEE Trans. Eelecrron Dev. ED24,402-410 (1977). 13. J. van Meerbergen, J. Nijs, R. Mertens and R. van Overstraeten, Rec. 13th IEEE Photovokaic Specialists Conf. (1978). 563-570 14. D. D. Tang, IEEE Trans. Electron Dev. -27, (1980). 15. A. H. Marshak, M. A. Shibib, J. G. Fossum and F. A. Lindholm, IEEE Trans. Electron Dev. March (1981). 16. H. Kroemer, RCA Review 28,332-342 (1957). 17. M. A. Shibib, F. A. Lindholm and F. Therez, IEEE Trans. Electron Dev. ED-X, 959-965 (1979). 18. R. A. Abram, G. J. Rees and B. L. H. Wilson, Advan. Physics 27,799-892 (1978). 19. H. P. D. Lanyon and R. A. Tuft, IEE Trans. Electron Dev. ED-%, 1014-1018 (1979). 20. V. I. Fistul, Heavily Doped Semiconductors. Plenum Press, New York (1%9). 21. N. Kobayashi, S. Ikehata, S. Kobayashi and W. Sasaki, Solid St. Commun. 24.67 (1977). 22. M. A. Shibib and F. A. Lindholm, IEEE Trans. Electron Deu. (July 1980). 23. A. H. Marshak and R. Shrivastava, Solid-St Electron. 22, 567571 (1979). 24. D. Redfield, Appl. Phys. Letl. 33.531-533 (1978). 25. D. P. Kennedy, Solid-St Electron u), 311-319(1977).

occur in regions that are heavily doped. When experimentally determined heavy doping parameters are used in simple, Boltzmann-like transport equations, no explicit correction for degenerate statistics should be made. Finally, since information concerning y is not presently available, the use of an arbitrary y was discussed. It was

APPENDLX

Derivations of the transport equations for heavily doped semiconductors are presented in this appendix. Only the electron transport equation is considered here because the derivation of the hole transport equation is similar.

202

M. S. LUNDSTROM et al. is defined in order to write the electron current density in the form J. = np.Q[E, - 8.1 t kTp,Qn.

For isothermal conditions, the electron concentration is a function of the two parameters, n< and position. The position of the electron quasi-Fermi level with respect to the mobility edge is measured by nr. In addition, the position must be known in order to evaluate n when the density of states is position-dependent. As a result. the chain rule may be applied to write
Qn =

649)

Q'h +$ Q,
c

If the definition of nc, (II), is used, (Al) can be used to express the gradient of the electron quasi-Fermi level as (A2)

Equation (23) results by relating E, to the electrostatic potential. Both the position-dependence of the density of states and the influence of Fermi-Dirac statistics are accounted for by the parameter 0.. The hole transport equation is derived in a similar manner. For holes the 0 term is defined by

Q0, VT, -

so, - k%a PCLP

1
Qp

MO)
(All)

The madient of the auasi-Fermi level is related to the current density-by (7) which is used with (A2) to obtain

Expressions for 0. and 0, can be obtained by using the definitions of F, and Fp in (A8) and (AlO). The result for 0. is

Q0. = kT ;-

kTVnc,

n.

(A3) which can be integrated to write

The generalized Einstein relation is given by

n = C. exp[@,

+ F. - Ec)/kT].

W-0

4=kT
CL*

4 dn (G

n
>

(A4)

If we define 0. = 0 for a nondegenerate sample with unperturbed band structure, then it follows that C. = N,. A corresponding development for holes leads to P = C, expW, - FP f EdlkTl, (Af3)

A parameter, F,, where VT, = -$ (G > and

Qncn

(0

where C, = N,.

By defining effective densities of states as 7;s; = N, e(Wkr) (Af4)

is defined to account for the term in (A3) which occurs because of the position-dependence of the density of states. Using these definitions, (A3) becomes J. =nP.V[E,-F,ltqD,Vn, (A6)

K=

N W e@dkr) form as

(Al9

(Al2) and (A13) can be. written in Boltzmann-like

which is the result obtained by Marshak and van Vliet[4]. Equation (17) results by relating the gradient of E, to the gradient of the electrostatic potential through (15). The effects of the modified band structure and the influence of Fermi-Dirac statistics are difficult to separate in a degenerately doped semiconductor. Therefore, the influence of degeneracy in the diffusion term is moved to the drift term and added to the term which involves F.. The diffusion term can be written as (A7)

and

p =x

ey

(A17)

When Boltzmann statistics apply, K and x are independent of r), and n.. Substitution of (A16) and (A17) into (A8) and (AlO) shows that VF, = V0. and VF, = V0, in a nondegenerate semiconductor. Furthermore, (AS) can be used to evaluate QF, as

The first term in (A7) is moved to the drift term in (A6). and a parameter, Q0. = VI', -

VI- =kTViij; I x

WE)

(A@

in a nondegenerate semiconductor. The terms in the transport equations that involve QF represent additional diffusion introduced by the position-dependence of the density of states.

S-ar putea să vă placă și