LIBRARY OF THE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAICN 510.8*4 ho. 6,1-8)0 CO Digitized by the Internet Archive in 2013 http://archive.org/details/fundamentalsofju64popp UNIVERSITY OF ILLINOIS GRADUATE COLLEGE DIGITAL COMPUTER LABORATORY INTERNAL REPORT NO. 6h FUNDAMENTALS OF JUNCTION TRANSISTOR PHYSICS By W. J. Poppelbaum August 30, 1955 This work has been supported "by Contract N6ori-71 Task XXIV United States Navy ONR NR 048 094 TABLE OF CONTENTS 1. The Number of Possible States of a System of Particles 1 2. Number of Occupied States. Fermi Statistics 2 3. The Physical Meaning of the Fermi- level J4. k. Electrons and Holes. Energy Bands. Electron and Hole Densities 5 5. Origin of Energy. Intrinsic Densities. Equation for the Electro- static Potential 7 6. P- regions and n-regions. and y in an Equilibrium Junction. 9 Quasi Fermi- levels 7. Mobilities, Diffusion Constants, Einstein Relationship. Current Densities expressed as Gradients of Quasi Fermi-levels 11 8. Generation and Recombination of Holes and Electrons. Continuity Equation 13 9. Variation of , and "^ in a Non-equilibrium pn Junction lk 10. The Voltage- current Characteristic of a pn Junction 17 11. Variation of , and y/in a Non-equilibrium pnp Transistor 19 12. The Voltage- current Characteristics of a pnp Transistor 22 13. Behavior of the Base Region at High Injection Levels 2k lk. Carrier Delay in the Base Region as a Function of Injection Level 25 15. The Current Amplification a of a Transistor 28 16. The a Cutoff Frequency as a Function of Injection Level 31 17. The Transition Region Capacitance 55 18. The Collector Space Charge Layer 5° Ac know ledge me nt The deepest gratitude is due to Professor J. Bardeen for his helpful criticism and the suggestions he has made for the improvement of many sections. INTRODUCTION The following sections are an introduction to the theories used in the discussion of transistor "behavior. To simplify explanations -we only con- sider pnp transistors: all results can easily be transformed for npn units. The text is in no way original except for details of exposition and certain conclusions reached in the latter sections of the report. The exposition leans on diverse treatises on solid state physics and a number of articles which have become classic in the field. The principal sources used are: J. Bardeen: R. C. Tolman: F. Seitz: W. Shockley: W. Shockley, M. Sparks , G. K. Teal: W. Shockley: C . Herring 8 E. So Rittner: T. Misawa: J. M. Early: Fundamentals of Transistor Devices. Lectures at the Graduate Summer School in Semiconductors (1953) The Principles of Statistical Mechanics. Oxford University Press (1950). Modern Theory of Solids, McGraw-Hill (1940). Electrons and Holes in Semiconductors, Van Nostrand (1950). PN Junction Transistors, Physical Review, vol. 83 , No. 1, pp. 151-162 (July 1951). The Theory of pn Junctions in Semi-conductors and pn Junction Transistors, BSTJ, vol. 28, pp. 435-489, (July 19^9). Theory of Transient Phenomena in the Transport of Holes in an Excess Semiconductor, BSTJ, vol. 28 , pp. 402-469 (July 1949). Extension of the Theory of the Junction Transistor, Physical Review, vol. 94, No. 5, pp. Il6l-1171 (June 1954) Diffusion Capacitances and High Injection Level Operation of Junction Transistors, Proceedings I.R.E. p. 749 (June 1955)- Effects of Space- charge Layer Widening in Junction Transistors, Proceedings I.R.E. , p. l401 (November 1952). 1. The Number of Possible States of a System of Particles Let V be a volume in free space containing N particles (for example electrons). Let x be the radius vector from an arbitrary origin and —> — > let a- = mv be the momentum vector of a particle. It is convenient to represent the succession of states of movement (x,c-) in a 6 dimensional space called phase space and having coordinates x x x Figure 1 a ~ cr °z. The fundamental assumptions of quantum Coordinate Vector and mechanics can then be summarized as follows : Momentum Vector — » — > If by any apparatus we try to determine the components of x and o~ simultaneously, we will encounter uncertainties Ax. and A&~. such J ' 11 that for every value of i = 1, 2, 3 Ax .As". = h, where h is Planck's constant. /„ . , . „ . . , \ (Heisenberg's Principle; We deduce from Heisenberg's Principle that if we divide phase space into cells of size Ax 1 Ax_Ax : A < 3- Ao^ioo = h , we cannot distinguish the state of movement of two particles in the same cell. Statistics applied to electrons are based upon the following hypothesis: 3 Each cell of size h in phase space can only accomodate 2 particles (of opposite magnetic momenta or "spins") at the most. (Pauli's Principle) Remark : We can apply the above theory even in the case where the electrons are inside a crystal. Then the "crystal momentum"^ is no longer equal to mv but h -> -* still equals — r— k where k is the wave vector of the associated wave, 2rt Notice that each cell corresponds to a definite energy E = E(x,cr) but that for a given energy there may be several cells . Let us now choose a volume V = Ax..Ax^Ax small enough to neglect the variation of x; then E = E(s-) and the size of each cell in cr-space ("momentum space") becomes This cell contains again at most 2 particles . We can now calculate the number of possible states having energies between E and E + dE if we know E as a function of o-„ It turns out that very often 2 where E = potential energy of particle (zero of energy to be fixed later) j 2 %r- = kinetic energy. 2m Now the volume of a spherical shell between the radii (T and oE (2 ok) where g is a quantity independent of E and W. By comparing the result of statistical calculations to classical thermodynamics, it can be shown that ^-4r (2 - 5) where k is Boltzmann's constant and T the absolute temperature. Now consider a perturbation more in detail. Suppose it consists in the transfer of 5N particles from group r to group s « Then 8N = -5N and SE = fa - EjSN . Further (2.1) gives 5 InW = -5 ln(N i) - 5 ln[(S - N )l] ~S ln(N g O a. 5 ln[ (S - N )'.]. Since lnxi = x lnx, we have 5 InW = -8N InN + 5N ln(S - N ) - SN In N + 5N ln(S - N ) r r r N r r s s ss s 5(S - N ) = -5N etc.'.) and hy using (2.k) and (2.5) [-InN + ln(S - N ) + InN - ln(S - N )] r x r r s s s' E N r s s 1 o_ InW 1 kT 5E E * E r s gives E r N 1 "■ n ' + J1 S m N This being true for all groups r and s, we snail set each side equal to E_/lff and call EL, the Fermi-level. Then for any r N In s _ r N = (Eg, - E r )AT. (2.7) r r Now introduce the distribution function (value for group r) by N r Then f 1 r " (E -E )/KT 1 + e ' i.e., for any given energy range E, E + dE the quotient (possible states)/ (occupied states ) is (2.9) 1 + e (E-V/M ^. The Physical Meaning of the Fermi -Level As can be seen from (2-9) f = l/2 when E = E^, i.e. 50$ of the states are r occupied for an energy equal to the Fermi-level. We are now going to prove that if we transfer 5N electrons (of charge-q) from a conducter in equilibrium at level EL, and electrostatic potential T, to a conductor 2 in equilibrium at level E and potential V in a reversible J? 2 2 fashion, -*C*i - V 2 ) = E F1 - E F2 (3-D or, for any origin of E^ and V V = + constant (3-2) It is possible to show by irreversible thermodynamics that in the most general case of flow of electric and thermal currents (3'2) is still true. To prove (3-1), we transfer 5N electrons from conductor 1 to a conductor 2 at the same temperature in a reversible fashion (total 8S = and therefore 5F = 5E - T6S = 5E, where F is the free energy). Now 5F = 5F + 5F and by a well-known theorem of thermodynamics S = k InW (in the union of 2 systems we have to add entropies and multiply probabilities'.). Therefore 5F 2 = 5E 2 - T5S 2 = 5E 2 - kT(S lnW) 2 (3-3) Suppose that we add the electrons to states in group r in conductor 2 . Then by the preceding section (5 lnW) 2 = - lJ—{— It follows that / N 5? 2 = 5E 2 + kT 8N ln/g— §- ^ \ r ' r / 2 = E^ SN (3 A) by utilizing (2.6). We see that 8F is independent of the group of states to which the electrons are added. Using 8F = SF + SF we find that 5F = (E^ - E F1 )5N (3-5) while 5E = q.(V - V g )5N (3*6) As mentioned above, we have in this particular case BF = BE. Equation (3«l) thus follows immediately. h. Electrons and Soles. Energy Bands. Electron and Hole Density By a more detailed examination of the quantum mechanics involved in the movement of electrons inside a crystal, it can be shown that these electrons can only have energies in certain permitted ranges called bands. In the most common case there is a nearly empty band in which the electrons behave more or less like free particles % this is the so-called, conduction band. Then there is a nearly filled band; the valency band. Electrons in the valency band are better described by the behavior of the vacant energy levels or holes. Practically a hole behaves like a particle of charge +q having approximately the electronic mass m. We shall now calculate the electron density in the conduction hand. For this we set V = 1 in (l.k), obtaining the number of possible states per unit volume : N(E)dE = i| (2m) 3/2 (E - Ej 1 / 2 h The fraction f by which we must multiply to find the occupied states is given by (2.9): f _ 1 (E-1L, J/kT 1 + e * The number of occupied states between the energy limits EL and E of the con- duction band is thus •E 2 N(E)dE_ 1 1 + e (E-EpVl- (U - 1} Now this is obviously the electron density. We shall calculate (^.l) under the hypothesis that E > > E_ 3 then (E-E J7S 1 + e ^ 1 -(E-E F )/kT = e (^.2) and 'E, = e V«r M^t' 2 f 2 (E . a )i/2 ,-Vh dE E i Putting E* = E - E and taking account of the rapid convergence (which permits to extend the upper limit to +°°), we may write n - jEF-EnV^ M2m) 3 / 2 (^ ^&/^ & Y? Jo _ . e -(En-E F )/kT (k.3) n where o ^/p N = 2(2jtmkT/h ) ' = "effective density of electrons in the conduction hand." (^••^■) By reasoning on the holes in the valency band, the potential energy E^ r n€ density being necessarily measured from the same origin as E , we find for the hole where p = N e"^ "EpV^ (^5) P N = "effective density of holes in the valency band." 5. Origin of Energy. Intrinsic Densities. Equation for Electrostatic Potential We shall now proceed to choose the origin of the energies E, E , E , E_, n p s in such a way that there will be some simplification in the formalism. Let y= electrostatic potential (measured from infinity) E = energy gap between the top of the valency band and the bottom of the conduction band. The energy origin will then be chosen in such a way that ] /M E r E n = -qy+^KE Hr)'T- (5 ' 1} leading im mediately to /N \ E E = KV+ikT ln(Jl--^ . (5.3) V p/ Further let us introduce a Fermi -potential f> such that E F = -q0 (5-3) Then (4-3) and (4.5 ) can be written ; q(^-0)/kT ,q(0-yVfcr n = n. e 1 (5.10 n. e 1 where n i = / i n i p e " EG/kT = i^~' (5-5) It can be seen that f or V' = 0, p = n = n.: n. is called the intrinsic carrier density. (^.^0 shows that n. is independent of the location in the crystal as long as E is constant: This is usually assumed to be the case. Formulae (5«*0 are therefore perfectly general, n. having the same value every- where; only and "y vary with the location. We can now examine the case in which the crystal contains electrons, holes and ionized atoms. Let N, be the density of positively ionized atoms (called donors) and N the density of negatively ionized atoms (called acceptors). The 8 distribution of donors and acceptors is determined during the growing of the crystal. We thus obtain for the charge density f>= q(p - n + N d - N a ). (5-6) Substituting for n and p the values (5«*0 we have /> = q[N d (x) - N a (x) + 2n. sinhq^(x) -y(x)|/kT] , (5-7) If we specify 0(x) we can calculate ~yu from Poisson's equation *) -div E = lap y> = - ■£ (5.8) r o (MKS-system, e = dielectric constant of vacuum € = relative dielectric constant of crystal) In the treatment of transistor problems two cases are usually considered: as a first approximation a zero space charge solution (/°= 0) at low current densities and then a zero space charge solution at high current densities. Only in exceptional cases is one able to treat the case p f 0: we shall not attempt to do so until section 18. :2 *) In the one dimensional case lap = — gf 9x 6. p-regions and n-regions. and y/ in an Equilibrium Junction. Quasi Fermi -levels p n N N, ap dp N N, an dn P n P P P n n n Figure 2 Notation for p-regions and n-regions (5.6 ) gives Consider two adjacent regions distinguished by the indices p and n such that in the left hand region and in the right hand region respectively we have ap dp = P o > ° ) N - N^ = -n < an dn o (6.1) Let us consider the zero space charge case 5 then p = p + n P o p n = n n o (6.2) n We shall suppose that p > > n P P n > > p n . 3 (6-3) The p-region then practically only contains holes, while the n-region practically only contains electrons . Electrons in the p-region or holes in the n-region are called minority carriers . Let us try to gain some insight into the equilibrium behavior of and "y^when we go from a p-region into an n-region across a "pn junction" i.e. the transition region. The important thing to notice is that must have the same value throughout the sample: if it were not so, energy changes according to section 3 would accompany the transfer of electrons (or holes ) from one region to the other and this contradicts the hypothesis of equilibrium. This does not imply by any means that /-'is the same everywhere. But we can see that Y' must be constant at all locations where the concentration of carriers p n and p n is constant, i.e. inside the p-region and inside the n-region. *p p *n n ' * & to This is evident because carriers can only drift as a consequence of concentra- tion gradients and electric fields E = -grady<*)s if there is equilibrium and no concentration gradient, there can be no potential gradients either. *; In the one dimensional case grad Y* = — — a*- 10 fl- reg /0 5-7 r 72 - -r<2jt'er\ Figure 3 and V i n an Equilibrium pn Junction In the transistion zone however a big potential gradient must he present to offset the influence of the rapidly varying concentration of holes and electrons: and y> behave as indicated in Figure 3« In equilibrium the computation of p and n (called p and n to indicate * e n p n. p equilibrium'.) offers no difficulty when p and n are given. (5«^) and (5«5) are valid, therefore (p + n )n = n. and (n + p )p = n., giving by (6.3) o p p 1 o n n 1 J v o 1 P = n,/p 2 n./n (6A) Let us now take the non-equilibrium case. Then we do not know anything a priori about the Fermi-levels in the p- and n-regions . Worse stills when current flows across a pn junction we are evidently no longer entitled to use formulae (5«*0 which are a consequence of (^-3) and (^«5) and therefore valid only under equilibrium conditions . Nevertheless we can write relationships analogous to (5>*0 containing "quasi Fermi-levels" (sometimes called "imrefs") and to be defined in a suitable way: n n.e 1 q(/<-0 n )/kT (6.5) p = n.e * 1 q(0 p -y)/kT Solving (6.k) for and we then obtain formulae defining the quasi Fermi- levels : n r n ' q ' n i j w,/.** (6>6) In section 9 we shall look into the behavior of ys, and in a junction when current flows across it, i.e. the behavior in the non-equilibrium case. Mobilities, Diffusion Constants, Einstein Relationship and Current- Densities Expressed as Gradients of Quasi Fermi -levels The movement of electrically charged particles can come about in two ways; firstly they can attain a certain average speed due to the electric field present, the speed being limited because of collisions with the crystal lattice, It turns out that often the average speed is proportional to the electric field: one can then define a constant, called mobility u, which is the average speed per unit electric field. Secondly the movement can be caused by concentration gradients 2 the (random-) thermal Telocity then produces a general drift to the regions of lower carrier density. The number of carriers drifting across unit area in unit time often turns out to be proportional to the concentration gradient; one can then define a constant, called diffusion constant D, which is the particle current density per unit concentration gradient. Take for example holes having a mobility u and a diffusion constant D . P P Let p be their volume density and q their charge as before. Call E = - grad y* = - —*■ the electric field. Then the total particle current density due t,o E and grad p is u- pE - D grad p, i.e. the current density I for holes is given fey I = qG-yE - D p grad p) . (7-1) In the same way we would have for electrons I = q(jj. nE + D grad n) . (7-2) n n n *' ' Let us take an inhomogeneous p-region in equilibrium. Then we can write P = Ae-^/ M (7.3) q0 where A = n.e^ has the same value throughout the region as discussed in section 6 but Y= y(x). Using (7-3) we obtain arad T,-^-- B i^-H E grad P - 3^ - P Hdx - M ^ Since we have equilibrium, I =0 and (7*1 ) gives IT . pqEfc - D p £) or |i = rt" D (Einstein relationship) (7«^) In a similar way we can prove that We can now establish a relationship between the gradients of and and I and I in the general non-equilibrium case. We then have by equations (6.5) n = n e q(v-0n)/kT (7.6) p = n r «l(0p-/)/KP r i Let us now use (7-1) and (7'2), remembering (7^) and (7-5 )• Calculating' grad p from (7-6) we have grad p = p ^r (grad - grad 7^), so that (7'l) gives I = qp. pE - q — up^r- (grad + E) p p q p^ kT r -p = -wjp g rad 0-. (7*7) x^ Jr In the same way we can prove that I = -qu n grad (7 «8) n * p r n *) It is sufficient to consider the one -dimensional case in which grad p = 5g£ 13 8. Generation and Recombination of Holes and Electrons. Continuity Equation If we isolate unit volume inside a semiconductor (not necessarily in equili- brium), the number of holes (or electrons) in this volume varies for several causes. First of all holes and electrons recombine (i.e. electrons "fall" into the vacant energy levels called holes) at a rate r particle pairs/unit time. Then holes are generated due to thermal excitation, light, etc. at a rate g particles/unit time. Let us try to estimate the rate of change g-r. Take for instance holes in the n-region; then for equilibrium p = p given by (6.k). Let g and r be the values of g and r in the equilibrium cases g = r . Thus g-r = when p - p =0. We shall now make the funda- mental assumption that g is constantly equal to g , i.e. that the minority carrier concentration does not affect the creation of holes. Since the rate of recombination r must be proportional to p (for the probability of a re- combination is proportional to p ) we must have g - r = g o - X? n Taking the equilibrium case we find that g = X.p . Further we deduce from the definition that X has the dimensions (time) j we shall therefore introduce _, -l a constant c = \ called lifetime of holes in n-type material and write P ^ o P - P (g - r), , _ . . — s± • (8*1) vo hoxes m n- region 7~ -p In the same way we can treat electrons in p-type material by introducing a lifetime 2" and writing n o n - n ^° ' electrons in p-region _r* v • / n Notice that (8.1) and (8.2) have been established for the non-equilibrium case, i.e. the case where current flows. We shall now express that in this case the change of the number of holes (or electrons) per unit volume and unit time is due to g-r and the holes carried into the volume by the currents this is the continuity equation. Take again the case of holes in an n-regiom I , the current density of holes, is given by (7«1 )• The total charge of the holes going out of unit volume is given by the divergence of I ; the number of holes going into unit . _____ 3x *) In the one dimensional case div I = P 9x Ik volume in unit time is thus div I . The continuity equation thus "becomes q. p 3p P - P -, t=— ; = =: dlV I 5^ c q p (8.3) In exactly the same way we have for electrons in the p-region o > p and n i this is the so-called w ' o o n p "low" injection level case." We shall partially abandon this hypothesis in section 13' Using (7-7) ve find that T |grad I < — - — in the p-region i 1 p 1 < I m- p q. p 1 1 < 1 n ' u p q p n grad < — — — in the n-region p n Since p > > p by hypothesis 3 , I grad < < Igrad P P r p'n and it is reasonable to put |grad p | p = (9-3) i.e. is constant in the p-region. In the same way we can find that Igrad =0 (9.^) i.e. is constant in the n-region. r n As regards the transition region, we have to distinguish two cases. For forward and low reverse bias the carrier densities of either type will be sufficient to apply simultaneously (9 «3) aud (9.^). For high reverse bias there will be a very low carrier density and then and will vary in the transition region. Figure 5 gives the shape of and in the low bias case. Notice that the variation of in the p-region and that of in the n-region is arbitrary? it has no influence whatsoever on the other reasonings in the present section. By hypothesis (l) and must of course coincide at points 1 and 5s going n i back to (6.k) p(l) = and p(l) = p . Using (6.6) we obtain the announced result . 16 o o '/'S%>. r>o* Figure 5 and in a Non- equilibrium Junction Let us now turn our attention to )^. First of all (6.6) shows that"/' is constant between 1 and 2, for is constant and p ~ p everywhere. In the same way we can prove that ~v^ is constant between k and 5* The variation of y is therefore as represented in Figure 6: ■n S H\ X £?&/'/ y n ►> Cor>sfdr?7 v'j/i/e > c//3 7&/7CQ I -2 3 * Figure 6 ■y-in a Non-equilibrium Junction Lastly we can see that (l) - (5 ) = (l) - (5) = 5/* The reason is kT o again (6.6): -")£ = — In — at point 1 whether we consider equilibrium or not. If we therefore increase yby 5y$ (and = ) will increase by by. The same reasoning applied to point 5 shows that here and keep their initial equilibrium case value. Summing up Figure 7 gives 0,0 and"/-' for the equilibrium case in dashed lines while the heavy lines refer to a biased junction with 0,0 and")^: p. e/'-s7bnc^ y, and in a Non- equilibrium Junction 17 10. The Voltage Current Characteristic of a p-n Junction Let us again consider the junction of section 9- In point 4 (of coordinates x, ) we have from (6.5) v (k) - n e titj>M-Y(VVw (10-D y n K i n (k) = n elMV-tnMl/m (10-2) n i where p and n carry an index n to indicate the fact that we are in an n-region. Now by section 6 n = n = (N, - N ) in an n-region because we neglect the space charge. (10.2) therefore gives ^i e -qy*nOOAT m e -q)W/Kr . n o Substituting into (10. l) we find that (10 = !i JLtpfrhtrnfM/* p r n n o 2 But by Figure 7 (4) - (4) = by and by (6.9) n./n = p = normal density of minority carriers in the n-region. The last equation therefore becomes P n (*)=P° e* 8 tf ffi . (10.5) 3p n Let us now try to determine p (x) in the static case ( ~z-^ = 0) for n v ^ c x > x, . The continuity equation (8.3) then becomes p - p (x) . p - p (x) 31 n *n n 1 , . _ ^n *n 1 n /, A i, \ - — — div I = =s — 1 — ■* • (10.4) T q p IT q 3 x v .f x^ To this we must join the equation giving I for a given grad p (x) and a given E; Js being constant to the right of 4, E = and (7-1 ) gives 3p (x) I = -qD grad p (x) = -qD * (10. 5) Inserting (10.5) into (10.4) P - p (x) 9 p (x) = *V n + D —^— (10.6) P 9* 18 P being a constant, we can write the solution to (10.6) which satisfies the boundary condition p finite for x -»<*= in the form m o ( x k-x)/Lp ,,. rt * (x) = p + p (e 1 r/ - 1) e v 4 y/ * (10.9) p , r n n n As a last step we evaluate I = -qD =r — at point 4. (10.9) gives directly I (k) = I (e^/ M - 1) (10.10) p ps v ' where o^ qp D I = 5 P (10.11) ps L v P is the saturation hole current density in the junction. We see that for Sy big and negative I = -I P P s In a similar way we could discuss the electron current at point 2: I (2) = I (e* 5 ^ - 1) (10.12) ns where qn°D I = H 2 - 2 - (10.13) ns L v ' n is the saturation electron current density in the junction. Using hypothesis (9.2) we thus find that I . I (4) + I (4) = I (4) + I (2) p v ' n v P n = I (e^/^ - 1) (10.14) 1=1 +1 s ps ns 19 (10.15) "being the saturation current density. Formula (10.1^) is the fundamental relationship used in rectifier theory. We can of course multiply by the cross section A of the junction and call v the applied bias 5"^. Then the current i = IA is given by i = AI (e s ^/^ - 1) = i (e* v / H - 1) s (10.16) 11. Variation of 0_, and )o in a Non- equilibrium pnp Transistor It will be sufficient, as in the last section, to solve for p in the con- tinuity and the current density vs. field and concentration gradient equation, to obtain a complete picture. But this time the boundary conditions are slightly different. We shall deduce them by extending the arguments applied to a pn- junction. Figure 8 gives the aspect of a pnp transistors an n-region (very /'(/ric7/077 / /1//7C 7/omZ e § P- A n- | c ^ eroiTTc-r 6&SG 8 ee t/s-cTor ;5 / Schematic Di ?. z will still be valid when Figure 9b holds, the reason being that the minority carrier density at the collector junction becomes exceedingly small in either picture, i.e. the same boundary conditions hold. 21 V or ft c o//oc /or £/->) c//s7&ocq origin of x Figure 9a y*x p and for a Transistor with Low (Forward) Collector Voltage yorfi e m,7Ter Cp ) 6*se C-n ) Co//ecr r Cp) Or/gin af X Figure 9b y^ s and p for a Transistor with High (Reverse) Collector Voltage 22 12. The Voltage Current Characteristic of a pnp Transistor As in section 10 we can calculate p(2) and p(3) "by taking the difference between and . But (2 ) - (2 ) = v and (3) - (3) = v . Therefore r p r n P & e r p mi c P. n (2 ) = p° e qv e/ M qv c /kT (12.1) P n (5) =P n e The solution of the equation (10.6) (applicable because once again E = in the base region), i.e. = (12.2) o P- " Pr>> 2 f 9 P„(x :) n n + D n r P 3x^ p which satisfies the boundary conditions (12. l) at points 2 and 3 of coordinates x = and x = v is obviously [p n (5) - p°] sinh jL + [p n ( 2 )-p°]sin^ p(x)=p° + E E- (12.5) sinh =— P where L =,/ D r as before. P V p p 9P n (x) This allows us to calculate I (3) = -q.D — | by (12.3 )• P P v^ X = W Introducing j*( T ) = ^L£ (e^/M . 1} P L P = I (e^/^ - 1) (12 A) ps v ' we can then write y 3 ) - cseh/ft W(r e ) - cottv'^-W ). (12-5) Now the total amount of current across the junction in 3, going from left to right, is 1(3) = l p (3) + I n (3). (12-6) 23 The evaluation of I (3) can be done along the same lines as that of I (3) by considering the diffusion equation for electrons in the emitter and collector region together with the boundary conditions discussed in section 11. It is, however, easier to observe that the electron flow across each junction is independent of that across the other one. The reason for this is that the majority carrier density in the base, given by (6.5), is constant and equal to that in the equilibrium case (y> - y* and = in the base by section 111 )« Therefore the junction at 3 for example behaves like a pn junction biassed in the reverse direction. Therefore by section 10 I (3) = -I = - T * n (12.7) n x ns if we consider the reverse bias big enough. Calling A the cross section of the two junctions, the current i flowing into the collector is -l(3)A for continuity reasons (but I (3) is different from l(2) because of recombination in the base'.). We can therefore write i = -A cschf- T — |l*( v ) + A coth/=^\l*(v ) 4- AI (12.8) [ L v p e (pJ P and in a similar way we could establish that 1 = A > n . This is usually the cases the ns n p emitter and collector regions are more highly doped than the base. 2k 13 « Behavior of the Base Region at High Injection Levels The hypothesis 3 in section 9, i.e. n and p > > p and n . led to the o ^o r n p conclusion that = constant in a p-region and = constant in an n-region. This fact, together with the nearly constant concentrations p and n of holes 00 and electrons then allowed lis to take y as a constant except in transition regions: this means that we neglected electric fields. We shall now discuss certain features of transistors in which the minority carrier density p in the n-type base is actually much bigger than n . This can come about when the density of hole current across the emitter-base junction is high, i.e. for high injection levels. We can then still assume that the varia- tion of and is given by Figure 9a (This means that p > > p , which does not contradict p > > n l), while y/ would now vary in the base region on account of the very variable electron density n = n + p (zero space chargel). Thus an electric field appears in the base. To gain some oversight, let us go back to the general equations (8.5), (8.6), (7-1), (7.2), (5-6), (5-8) and I = I p + V & = 2lz^ - i div 1 (13.1) 5r ^ 1 p ^=^=^ + -divI (13.2) I - q(u p pE - D p grad p) (13-3) I n = q(u n nE + D n grad n) (13-^) I =I p + I n (13.5) div E = -S — (p - n + N, - N ) (13.6) C € ** da' v ' o r Let us recall that p and n are the equilibrium concentrations of carriers (in the p-region p = p in the n-region p = p etc.) We shall of course retain the zero space charge condition: p - n + N , - N =0 (13-7) da From (13.7) w e deduce 3 equations: 25 div E = o (13-8) £-!%. (13-9) grad p = grad n (13-10) Let us now define a dimensionless factor b by H = bn (13.11) n p By Einstein's relationship (see 7-^ and 7-5) we then have D = bD (13-12) n p Let us consider an n -region having an electron density of the form discussed in (6.2): n = n Q + p (13-13) (we leave out the subscript n to simplify the formulae). Adding (13-3) and (13 -M we then obtain I - qD (b - l) grad p E = r-7t n — C 1 — (13-1^) qu Lp(b + 1) + bn J P ° Therefore, by (13.15) pi - qbB (2p + n ) grad p ±p ■ P(/ + 1WL ^) which means that (13-1 ) can be written - o pi - qbD (2p + n ) grad p % + £-^- = - div 1 — Ij l-L— al . — (13.16) d? -T p q[p(b + 1) + bn Q ] This, then, is the differential equation for the concentration in an n-region in the most general case, assuming zero space charge. Before integration, one would have to discuss the boundary conditions. lh. Carrier Delay in the Base Region as a Function of Injection Level Lex us consider the mechanism by which signals are transmitted by a sistor. 2$ and (for 1 r V 'n * bias) are given by Figure 10, transistor. and (for high or low injection level and strong collector XT AA Vc err>///er fe e//'s7'c)/7C(? Figure 10 Quasi Fermi-levels in Transistor with High Reverse Collector Voltage Modifying v means — because e r p is nearly horizontal in all cases in the emitter region — modifying the hole concentration at the emitter-base junction; p (2) is given by (12.1). This modification is practically instantaneous and there is no signal delay in the emitter region. In an analogous way we could argue that the collector region causes no delay. So we are left with the time of pro- pagation of a perturbation of hole concentration from 2 to 3. This time dealy 5 is clearly the base width divided by the average velocity v of the carriers*: 5 = w V (l*.l) Now it is clear that in a fixed point we can define an average velocity v such that I = vpq (2A.2) P In the low injection case there is no electric field in the base and I = -<£D grad p (lA-3) P P It follows therefore that D v = - — ** grad p l .e. = m _ D /gradj> p V p , ~ _ D grad p P P Now for all practical purposed p (3) =0 and (Ik.k) *) v indicates an average over all the particles in a given point, averaged over the whole base width. 27 P n (3) - P_(2) -p (2) w while p=\p n (2) (14.6) where \ is a numerical factor of order unity. Thus it follows that D w and 5 X =xf (14.7) P where the upper index 1 refers to the low injection level. Let us now look at the high injection level case. For this let us go back to equation (13.16), remembering that the square parenthesis is simply I -2. . From (13-4) we obtain, using (13.IO), (13.11 ) and (13.13) I - qbD grad p E = n "P -_ (14.8) qbu (n Q + p) v which in the high injection level case p > > n becomes (14.9) Substituting this into (13 »3) we have I I = *p - 2qD grad p = - 2qDp grad p (l4.10) since I is negligible by the hypothesis p > > n°. We see that (l4.3) is replaced by I = -qD 1 grad p with P P D* = 2D . (14.11) P P V ' From there on all calculations remain the same and we obtain . 2 n ~ W p 28 Notice that (ik.'j), (1^.12) and (1^.13) are in perfect agreement with the results obtained in section l6 by more sophisticated raasonings . 15. The Current Amplification a of a Trans istor In section 12 we found expressions f< Positive Directions of Currents in a Transistor i and i as a function of the applied c e biases v and v . at least for the low c e' injection level case with zero electric field in the base. Let us call 1 the current flowing from the base electrode into the transistor; then i e ♦ ^ + i c = (15.1) i.e. by using (12.8) and (12.9) with I ~ ns 2 sinh' **-- = -i - i = 2/ w sinh/=^ ifl [I*(v ) + I*(v )] p v e' p v c' (15.2) which shows clearly that i, f 0, or |i | > |i |. This base current clearly corresponds (in the case where most current is due to holes) to recombination of holes and electrons in the base region or flow of electrons across the emitter junction. It is therefore sometimes called "recombination current." Its effect is to decrease the efficiency i /i of current transfer between * c' e emitter and collector. However it is not very useful to calculate i /i since in practice we are often interested in the response to small signals. In this case we define the current amplification factor a by di a - Si v = constant c (15.3) It may seem advantageous to derive a from (12.8) and (12.9) by differentiation. But this would only be valid for very slow variations. Practically we are interested in a at signal frequencies of considerable magnitude. We shall therefore take up the problem from the beginning, i.e. solve equation (l3»l6) with its time dependent 29 term S^» (left out in the steady-state discussion of section 12) and appropriate boundary conditions (v variable'.) e Let us treat the low injection level case n > > p first. Then (l3°l6) becomes o o ^£ + P " P = div D grad p = D lap p. (15-^) P^ tr P Introducing as before L = , / D "C and restricting ourselves to the one p V p p dimensional case, we have » P P - P 1 5P / ls o rf -—2" - r ?? (15 * 5) ax l p ^ p Suppose now that we apply an emitter bias V = V + V , e o e leaving v constant according to the definition of a in (15«3)« Let x = and x =w correspond to the position of the junctions as in section 12. Then the boundary conditions (12. l) are replaced by p( 2 )=p\'Me M )/ H *n v -^n Now assume that v > > v, then o 3 qve Jj/t /kT ~ . jafc/wn i ' = 1 + qve /KC and jo/C where P n (2) = p(0) = Pl + Pe J (15.6) P 1= p°e^oM (15 . 7) P = p^v/kT (15.8) The boundary condition at x -w is the same as the second condition in (12.1) because v is constant? c P n (5) = P(w) = p° e"c/H (15 . 9) 30 Notice that in the last equation v < 0. The solution of (15*5) satisfying (15.6) and (15«9) is ^ P l " P n^ S±uh ( \~ / + ^ P ^ " P n^ s±Ilh ( jT ) P = P. sinn / — L P Pe J > n . Thus i = -AI ( w ) c p 11 ' (15-12) where A is the cross section of the junctions. But since we are only interested in the alternating part i of i , it is sufficient to calculate (15.II) for the second term in (15.IO). We then obtain , J& \/l+3^ , i = APe c y } cB 4\/ 1+Jw p (sh]l In a similar way we can compute the alternating part i of i : i = APe c e -= * 1 coth \l 1+ ^ r v w p /—I and therefore (supposing everywhere v = constant) (15.13) (15.1*0 oC = di i c c 31 := 7 csch J 1+ J ■s ( i , )] coth \/l+j "• (s)J i .e 31 oC = Remark: For uj= Owe have 1 1 cosh H + a* It K P oL r w :osh ( j — ) (15-15) (15-16) Since w/L is quite small in general, a ' p o l6. The a Cutoff Frequency as a Function of the Injection Level From (15.15) we deduce two facts: first of all there is a phase shift between i and i at higher frequencies because 0. is complex. Secondly la] decreases with ou: the current amplification of a transistor drops at higher frequencies . It is useful to define a a -cutoff frequency f = 00 /2n such that the corresponding value |a | is "3db down." This means that a a 1 2 (16.1) We shall first examine the low injection case of the preceding section. (16.2) In order to calculate uo we shall suppose that uy V > > 1 c p 'hen \ fTT H ^=/5«-/?(w) (16.3) Let us introduce >? w_/t^5 L \ P = w 2D (16.4) by virtue of the definition of the diffusion length L . Then (15.15) gives 32 a 1 _ 1 c ■ cosh^ + ">ij ) costnO cos>) + j sinh^sinvi and la c 1 .2. 2 N . .2^ . 2, cos2n + cosh2>7 cosh n cos >7 + sinh ^9 sin V <. l V/ cos >i + sinh ^sin "^ Using (l6.l) this becomes 2 = cos2 ^ . ± cosh2 1 (16. 5 ) Solving this transcendental equation we obtain ■h = 1.103 which, using (l6.h) finally gives D ^ = 2.W-f (16.6) w where the upper index 1 draws attention to the fact that we are talking about the low injection level case. It seems at a first glance necessary to repeat the calculations of the preceding section for high injection levels. Happily enough this is not necessary; we shall show that everything behaves at high injection levels as if D had been replaced by D' = 2D . We only have to repeat the argument of section lk: using I p = -2qD p grad p (16.7) equation (13.16) becomes §£ + ^^ = 2D lap p ?r ^ P = D p lap p (16.8) This proves our contention for the differential equation. The boundary con- ditions remain unchanged, so all we have to prove is that (l5-ll) is unchanged if D' is substituted for D . But this is easy, for we have already shown at P P the beginning of this section that (130) — i.e. the equation for I taking ir account of electric fields --can be written in the form (16.7) in the high injection level case. (i5.ll) therefore becomes: 33 I = -qD' f*| X = V (16.9) All this leads to the conclusion that the reasonings at the beginning of 2D b this section are valid if D ->D' P P P frequency (divided by 2rt ) for the high injection level case, we therefore have Calling^ the value of the a cutoff uO 2A3I+ D' R = 4.868 -g = 2CQ. Figure 12 H> . and in a junction with Forward Bias Let us go back to the problem of a single pn junction treated in section 10. But this time we shall take account of the fact that the transition region between points 2 and h has a finite width, i.e. of the fact that n and p do not abruptly change at 3 but gradually between 2 and k. The width will however be taken small enough to allow us to assume that - in it is constant r p r n and equal to 5^ (see figure). We shall simplify the problem by assuming that relative to an origin of x situated between 2 and k (but not necessarily in the middle if n * p ! . ) we have o ' o n - p = N , - N da ax (17.1) a and the exact situation of the origin being such that n = ax (3) o v ' P. = ax (2) (17.2) *) The first half of the equation is a consequence of space charge neutrality. 3k Points 2 and h will "be chosen sufficiently far in the p- and n-regions res- pectively to warrant the absence of electron flow in 2 and the absence of hole flow in k. Further we shall neglect recombination between 2 and k and as already mentioned, assume space charge neutrality. Now modifying the applied bias by d5y =dv modifies 0,0 and the hole densities between 2 and^. Between 1 and 2 p ~ p , between 4 and 5 p ~ p ; i.e. the variation of the total charge of holes Q with a given variation of bias is produced in the transition region alone. The same reasoning leads to the conclusion that the change of the total electronic charge (-Q because of the neutrality condition) is produced in the transition region. This means then that dv produces a drift of charge to the right of dQ in 2 and a drift of electrons to the left (-dQ) in k: the transition region behaves like a capacitor of capacitance o - i CW.3) Using (6.5) we have in the general case 3 qfr-0 n )/kT n = n. e 1 p = n. e 1 q(0p-y)/kT (17.4) Now by figure 12 we have throughout the (narrow) transition region p - n = &/ = v. (IT- 5) Introducing the quantities n and by K + K = e qv/ 2k T (l7-?) we can then write q(>-0l)KT u n = n e^- v ' r ' = n e ^Wi-Wkt _ „ -u (17-8) p = n -, e = n e where u = q ^AV M (17.9) 35 We can then transform (17-1) into 2n.. coshu du = adx (17-10) Let P "be the total number of holes between 2 and 4; then P = A l pdx = A / n n e coshu du where A is once again the cross section of the junction and u. and u_ are the values of u at points 2 and 4 respectively. Integrating , 2 An, /u An, 1 / 4 f . -2ux, 1 P = / (1 + e )du = IX, ." 2u 4 . o' 2u 2 K " u 2 } (17.II) Now at points 2 and 4 we have from (17.8) U 2 In P(2) u ), = l 21 ~^ 4 ni (17.12) (i7.ll) can thus be written P = An, 2a~ n. "^2 *\-v+m -^h A 2 A 2 Ap An, *o 1 2a IT K - U 2 } (17.13) \} 2 n i where we neglect y— 7rr ( = ~k because we can assume that the n-region is ' o sufficiently doped to have n > > n y p(2) is replaced p . We can now calculate U 4 " V . p(2)n(4) . p(2)n(4) . n i u^ - u 2 = In *^ g * f = In ^ < 2 S ' + In — n l n i n l In n P n 00 2qy 2 " 2kT n. 1 (17-14) 36 by using (17.7). Replacing n and u, - u by their value in (17.13) 2 . 2 Q = qP = qAp qAn . /. _ r o I qv/kT — - — + e^ ' 2a a In n p 00 qv 2 ~ kT n. 1 - 1 .e. C = 2 A 2 q An. _ 1 kTa - n p , 00 qv , ln — " HP " 1 n. 1 (17.15) Let us suppose that the n- and p-regions are strongly doped (n and p > > n. ). Then at room temperature qv/kT is of order unity (v ~ 3Qniv'. ) and C = 2 « 2 q An. n p , 2 1 -, n 00 p av/kT kTa 2 n. (17.16) Now by (10.1*0 we have 1 + 1 s qv/kT Calling C the value of C for v = 0, we obtain 1 + 1 C = C o 1 (17-17) showing that the transition region capacitance increases with the junction current . It should be remarked that the reasonings in this section can only be applied to the emitter junction, firstly because we suppose and constant in the transition region (see section 9) and secondly because we assume space charge neutrality. As explained in the next section this latter assumption is not valid for the strongly reverse biased collector junction. 18. The Collector Space Charge Layer Let us now consider the collector transition region, extending from -A on the left of the junction 3 to +A at the right side of the junction, the exact position of which is given below. 37 - A + A 3 t Or/q/'n of X Figure 13 Variation of y- in a Transistor We have already explained that for sufficiently high reverse bias the space charge neutrality condition can no longer be satisfied, the reason being that the collector region forms a very good sink for the minority carriers in the base: they are drained out at such a rate that a "depletion layer" is formed in the transition region. In this depletion layer p ~ and n ~ 0, meaning that ^=q(N d -N a ) As in section 17 we can then use the second half of (l7»l), giving (18.1) /<>= ax (18.2) which fixes, as we have seen in section 17, the "midpoint" 3 (corresponding to x = 0) of the junction. We can now use (5 '8) in the one -dimensional case? 9_ 2L a* 2 ax € e r o (18-3) By integrating twice we obtain qax~ t be £ 12 r o 38 Now it is easy to find C, : 2 *#- m ^ ax . c dx 2e e 1 r o must be approximately zero for x = 4A when the "normal" p and n regions begin, for ^is constant in these regions (see section 9)« Therefore c i ■ frr ww> r o and if we put u =y(+A) -y(-A) (18.5) we obtain u = 2a qA 5 3e e r o Now u = v , therefore c z 3e e v a5 . -i£i (X8.6) It is thus evident that the depletion layer extends further and further into the base region as the collector voltage goes up, reducing the base width w to an effective width w eff = V-A. (18.7) When w „„ becomes zero we talk about punch-through: the corresponding collector voltage v is the punch-through voltage. Above this bias all transistor action is lost. Further (l8.6) and (18.7) show that the transistor equations (with w replaced by w „„ ) become more complicated, in particular the collector swing in an amplifier modifies V rr constantly and thereby introduces a feedback effect . Another consequence is a decrease of recombination of injected minority carriers in the base layer since the average carrier diffuses across the narrower base in a shorter time. This corresponds to an increase in a and in the a cutoff frequency.