PN Junctions PN junctions form the basis of most semiconductor devices. Understanding their operation is basic to understanding most devices. Fabrication PN junctions ere normally fabricated by solid state diffusion. The two “simple” impurity profiles that result from this process are the complementary error function (erfe) and Gaussian, These result from the boundary conditions of constant surface concentration (No) and constant dose (N' or Q) as showa below. os) Fis £31 Commemenity err furcton profi Fie s Often, to simpuify the math, these profiles are approximated by the “ideal” pro! below. : Shallow diffusions: “Predeps” + short drive-in + STEP JUNCTION Deep diffusions: D¢(Drive-in) > Dt(Predep) => LINEARLY GRADED JUNCTION (Cross over between the two occurs at 2; = 2um as a rule of thumb). We shall consider these two cases in our analysis although real junctions are often in between the two.Band Diagrams at Equilibrium IN and P type materials are brought into contact, establishment of equilibrium is analagous to the Schottky diode we previously considered with the following fundamental differences: © Both sides of the junction will generally be depleted. © “Surface States” are usually not important (single crystal). © Both electrons and holes usually move to establish equilibrium. x = clectron affinity = Ey — E, , = semiconductor work function eae ae ap = Eo ~ Esp en = Eq — Ejn electrons flow: N + P holes flow: P + N © At Equilibrium: E, = CONSTANT 1 He Alternatively, a electrons are in high concentration on N side Initially holes are in high concentration on P side ‘Therefore, since diffusion of carriers occurs away from regions of high concentration, diffusion leaves behind uncompensated Nj and Nz. Resulting space charge sets up clectric field. Blectric field tends to pull electrons and holes back to their original positions. Equilibrium when drift = diffusion. Depletion ow ©‘The general relationship between the charge distribution and the potential is given by Poisson equation: Gat = Keagllt—P)— (No - Nal] @ a sae (P5") @ he regien p= nex (BF) @) ‘These equations are applicable in the neutral, de- neutral even pletion and transition regions under equilibrium (no Ly applied bias, etc.). Neutral regions: N region: n (Nf — Nz), p= n/m n—p—Np+N4=0 (ao net charge) a eal 2 Gaz = 9 (the potential is constant in the N region) Ey - E; kT Na = én = —ln— (from (2) 4 sts te= Bin Ze (rom (2)) @) Similarly in the neutral P region, aT, ON, =I 5) t= Ta ) ‘Thus we have: uilt in potential = $5 + dn dex EE ig Nee 6) @ ‘This is the built-in potential which exists across a PN ane junction with no applied bias (thermal equilibrium), Polarity: N(+), P(—). Notes: An impurity doping change N+ to N~ or P+ to P~ results in a built in potential. © The built in potential across a PN junction increases as Nz or Ng increases. 3Depletion Regions If this region is nearly completely depleted of carriers, then Fee Leen. oO Solution of this equation requires some knowledge about the spatial variation of Na and N,. We will consider this shortly. ‘Transition Regions In the transition regions between the depletion and neutral regions, neither our depletion nor neutrality approximations are valid and we have By = po llm—2)— (Me Ma, we (2p), rome (Ea), General solutions across the transition regions require computer techniques. Using such techniques, the transition region is found to be © 3Lp where Lp is the extrinsic Debye length and is given by oe 8) Na — ®) Lp may be thought of physically as the distance over which a fixed charge (Nf or Nz) exhibits [ Kek?_]? an influence over mobile carrier concentrations. In this instance it is a measure of the abruptness of the transition region between the depletion regions and the neutral regions. Example: Ng or N, = 10'cm~ results in Lp * 300A and 3Lp ~ 900A. As long as the depletion region is wide compared to this length, the abrupt transition approximation will be approximately valid. In summary, the charge distribution in the semi- conductor is given approximately by p © 0 outside depletion region. p= q(Nj — Nz) within depletion region. Boundary layer distance © 3LpStep Junctions ‘To proceed, we make some assumptions: 1, Step Junction (abrupt transition): Ng constant in N material. oN, constant in P material. 2. Depletion Approximation (ignores boundary or transition regions): —Na, = {Na O<2K 25 Emax ‘Therefore, av oF (20) From neutrality, Naty = Natn = Q/¢ (1) where the total depletion layer width is given by 2g = ant zy. ‘Taking the N side of the junction as an example, we have Vp aNe --Bil gee € mm) 0= Fit (mata) | en Ba ween [Mea(h+d)]”. - ‘The depletion region width depends most strongly on the doping on the lightly doped side and varies approximately as the inverse square root of that doping. More generally, to calculate the potential drop across the junction (or built-in voltage ¢,), Equations (12) and (13) are integrated. ei “ede (19) L if (e+2,)dz - Mf ~2,)4s| (Foon) -mEomaf] -aebe()-m(3) We already know ¢; from Equation (7). Since the charges on either side of the junction must be balanced, Nyzp = Nutn. Therefore, we have two equations ((15) and (16)) in two unknowns and can solve for z, and z,, and therefore zy with the same results as above.g ~*, Na Note that the zero of the potential curve is at the Xn Junction only if N, = Ng. In fact for a one sided step junction (Ny >> Ne or Ng >> Ng), we have: K,| i * The depletion region is primarily in the lightly doped side of the junction. + Most of the potential change also occurs across the lightly doped side. In the limit for this example (Nz —» 00), we approach a metal-semiconductor connection, Notes: * 24 is most strongly dependent on the doping in the lightly doping side of the junction. © 24 varies inversely as the square root of the doping on the lightly doped side. Summary for Step Junctions n and 4, are on the order of 0.2 to 0.5V. Therefore ¢; is on the order of 0.4 to LOV. zg is typically 0.1-1.0ym. Example: Given a PN step junction with Ng = 10"cm~* and Ny = 4 x 10!cm-,Linearly Graded Junctions eo Me IN, — Ng = —a2 (typically used as an approxima- tion for Gaussian profiles). a=slope (cm) (21) (22) (23) ‘The built-in voltage is determined by the doping con- centrations at the edges of the depletion region, ° KD are | oi Fr In car (24) 1 q - Qe, = 2a = [Bkee (25) Note the 1/3 power dependence of 2, on voltage. ‘The built-in voltage may be caleu- lated from (18) and (19) and is shown at the right. Again note that ¢j is in the range of 0.5 to 0.9 volts. Thisis approx. °*- Gp imately twice the value for the Schottky 07} 09; diodes considered earlier. os ose oat Roe ooo oe em)Conclusions The first order theory that we have developed for the PN junction allows calculation of: from fundamental material parameters (K, Eo, q, n, etc.) and practical considerations (Na, Nu). ¢ In general this theory is much better at predicting junction behavior than the first order Schottky theory we previously considered. The main reason for this is that the PN junction does not suffer from surface state problems (a perfect crystal forms the boundary).PN Junction Capacitance Under Reverse Bias Under equilibrium conditions we have, Wath reverse bate ae (v.20) barrier to majority carrier current (i.e. e: NAP, ht: PN). Ive A built in potential ¢; provides a bias the junction externally with the NV zie / mee side +, Pside —, the barrier to majority a enere carrier current flow is increased. avr eye vd Results 1. Applied voltage appears essentially entirely across the depletion region, 2, 2g increases, mobile carriers pulled further away from junction, 3. Minority carrier flow is increased but J is small because they are low in number. : 4, E field actoss depletion region increases. Step Junction a (26) = 2(6i— Ve) _ 4 ( potential across depletion region (en) peer tetera ‘width of depletion region 4: is replaced by (¢; — Va) (Note that Vq is negative, i.e. sign conventions are the same as with Schottky diodes, i Reverse bias > ¢i,Va add Forward bias => ¢;,V, subtract 10The small signal capacitance of the junction is given by 90" _ Keo OV, ~ ae or (28) (A derivation in the Muller and Kamins text shows that this latter result holds for an arbitrary doping profile. Given any profile, if we can calculate 2g, we can calculate the small signal capacitance C.) By small signal capacitance we mean: (1) Apply a D.C. potential Vics to establish 2g. (2) ‘Measure the capacitance by superimposing a small AC signal (Vac View) 50 24 is not changed significantly. Example: Consider a step junction yi 2 w=% 1 8Q' 82, 825 Os By = Neaye = ONS But from (23), 2p = (Ne/Na)2q and zy = 2, +25, Therefore, QNatn = qNaxy Bn = Bg an and yr 2K eo (1 1 _« FEG+a@-») 8 TENN. Na ca " lt Therefe fore ey | we L Va Na |2a(1/Ne + 1/2) (i — Va) — lea aKeo |e © (2C/N.+1/Na) (6: — Va). Just as would have been calculated based on Equation (28). (29) (30) (31) (32) In the reverse direction C falls as the square root Cc diode case, measurement of C vs V, cE of Va. In the forward direction, C —+ co as Vg extract information on the doping profile. bi. (Actually in forward bias our analysis breaks down be- ‘cause mobile carriers in depletion region causes (29) to be invalid.) Equivalent expressions may be calculated for @ for other impurity profiles. As in the Schottky can be used toPN Junction Breakdown ‘As the voltage across a pn junction is increased, the depletion region widens and the & field increases. Eventually, a limit must be reached. When this happens, the diode begins to conduct large currents in the reverse direction. Further increase in the voltage will cause currents to be limited only by the external circuit. ‘Three principal mechanisms are responsible for this breakdown, 1. Avalanche Breakdown Electrons enter the depletion region (or are gen- erated there by G-R centers), If€ is high enough, ne the free electrons can gain sufficient energy that inated when they collide with valence electrons they aon can break the covalent bonds and excite the va- prem yells nt lence electron to the conduction band, creating pol an electron-hole pair. This process is repeated as those electrons and holes in turn gain energy due to € and collide with other bound electrons and results in large currents. 2. Zener Breakdown If the € field is large enough, it can directly pull an electron away from a covalent bond, creating an electron-hole pair. This process also results in large currents. 3. Punch-Through In a narrow diode, as the depletion region increases, it can reach all the way through to the metal contact which can sup- ply a very large number of carriers. Large currents can result since no more voltage can be dropped across the depletion re- poi] cin tt gion. When z, = Wa or 2p = W,, punch-through occurs. } Punch-through almost always occurs on the lightly-doped side Ky of the junction since almost all of the depletion region is on that side. 12Avalanche Breakdown In avalanche breakdown, if the field is not large enough, free carriers will scatter without generating new carriers. Therefore, a critical field is required for multiplication and avalanche breakdown. 108 Sueksown FeV emt) ws 1 10 10 10 100 CConcenration W fem?) Figure 3412 The eriueal electric fields for avalanche and Zener breakdown silica as funcuons of dopant concentration.'** Note that the critical field increases for avalanche breakdown as the doping increases. This occurs because ionized impurity scattering becomes dominant with heavy doping and scattering becomes more frequent so carriers are unable to gain as much energy before being scattered. ‘The critical field for zener breakdown is approximately constant and is the limiting factor for Junctions which are heavily doped on both sides (p and n). To analyze avalanche breakdown, we can make the following observations: © Near the edges of the depletion or space charge region, £ < Eq and multiplication will not © In the central region, near Emac, if € > Eq avalanche multiplication can occur. For electrons and holes within the region where £ > Eq, we define ionization coefficient of e~ and = probability e~ creates an e~-h+ pair within dz ‘p = ionization coefficient of h* a,dz = probability h+ creates an e~-h* pair within de 1BAssuming a, = a = a, which is only approxi- mately true, then for both holes and electrons (33) where M = avalanche multiplication factor. Therefore, avalanche multiplication occurs when 1+ M0 (34) Evaluation of this integral can be difficult because a is a function of € and therefore =. We can consider a simplified case. Suppose we have a one-sided junction in which Ny > Ne. Most of the depletion region occurs on the lightly doped side. Emer = 20%=VA) from (22) 2a If we define the breakdown to occur when the maximum field reaches the critical field (this is a conservative estimate since a field greater than the critical field has to exist over a distance for breakdown to occur) then KeEx? aye 2qNo (38) 14Since the eritical field varies slowly with doping, the breakdown voltage varies approx- imately inversely as the doping on the lightly doped side of the junction. Meesdown voli, ¥ 1 1 1 1 1 17 ro ro To sawtreciion (NY « For BV greater than a few volts, the inverse relationship holds reasonably well, © At high doping levels, the BV flattens out because a new mechanism (Zener breakdown, to be discussed next) becomes dominant. © BV increases for a given doping as the band gap of the material increases. This is to be expected since it is more difficult to create hole-electron pairs in wider band gap materials. # Similar calculations for linearly graded junctions give similar results with larger gradient producing lower BV. 1090 Linen graded junctions 300K : = 10) 4 0” 1 ro 10 to Ireprity diene © For avalanche breakdown, BV increases with temperature since scattering lifetimes r and therefore a decrease with temperature. 15Zener Breakdown (Tunneling Breakdown) For avalanche breakdown a narrow depletion re- gion results in JZ: a dz being smaller because the field is near its maximum over only a short distance. There- fore, the breakdown voltage due to avalanche is in- creased. Zener breakdown occurs when both sides of a june- tion are so highly doped that the depletion region is narrow enough to allow tunneling (< 100A = 10nm) and there is a very large electric field across the deple- tion region. Zener breakdown occurs when both sides ate so heavily doped to at least approach degencracy. ‘The situation is analogous to ohmic contacts in metal- semiconductor contacts. With an applied reverse bias, an electron can tunnel directly from the valence band on the p side to the conduction band on the n side. © For Zener breakdown, BV < 5V. Tunneling Figure 9.7. Energy bands of a junction in Zener breahdowr © BV decreases with temperature since B, (the height of the barrier to be tunneled through) decreases with temperature. © Can differentiate between Zener and avalanche breakdown by observing effect of T on BV. + = avalanche — = Zener ~ 0 = mixed (4.6V) 1 $ T LD Fiance 94. The mal dd: 1-7 eae) sand ste a be ae reference (Zener) diode. 16Summary All diodes operated in reverse breakdown are called Zener (or reference) diodes although most actually break down by avalanche. After breakdown, the amount of current which flows through the diode is limited by the external circuit. Most breakdown is reversible unless the external circuit allows enough current to flow that ohmic heating becomes destructive. Avalanche: Most usual, BV « (1/N<). BV > 5V. Zener: High doping (nearly degenerate) on both sides, 2¢ < 100A, BV < 5V. Punch-through: Narrow region on lightly-doped side, zy = W. Practical Considerations In any real junction made with planar technology, the calculations above which relate to plane (flat) junctions do not strictly apply. “Planac" Tunckon “Plane” Tunction Curvature of the planar junction under the oxide edge due to lateral diffusion increases the E field in these regions resulting in lower BV than in the plane junction. Computer calculations show how BV is reducedas " CSS=xay r B the radius of curvature or 2; is reduced, with shallow eS}; Junctions showing the greatest reduction. The use of guard rings such as discussed for Schottky diodes can * ae help BV for planar junctions approach that for the » [~~ . ideal plane junction, s S ate 10 10% 10 Tot 7 Macieo?Current-Voltage Characteristics of PN Junctions So far, we have described the behavior of PN junctions in equilibrium and under reverse bias, conditions where the current is small. We will now consider forward bias where significant currents can flow. In thermal equilibrium 4; = dn + » = built-in potential. Under forward bias, the barrier to majority carrier flow is reduced. electrons are injected: N > P holes are injected: P + NV © Applied voltages will still be dropped across depletion region so ¢i => (4: — Va) # The barzier to majority carriers is thus reduced by V, and we would expect, based on the Boltzmann approximation to Fermi-Dirac statistics, that the current would depend exponentially on Vq. (‘The number of majority carriers with sufficient energy to surmount barrier depends exponentially on energy). © Once injected, the majority carriers become minority carriers on the other side. The behavior of minority carriers is of fundamental importance to the behavior of PN junctions. * The injected minority carriers will recombine with time, while at the same time they diffuse away from the junction. 18‘The number of injected carriers can be related to Vq. To begin we know that (36) We can define: no = number of electrons in P region in equilibrium Pro = number of holes in N region in equilibrium tino = number of electrons isi N region in equilibrium Pyo = number of holes in P region in equilibrium mo Nav Pao = 5 (37) ne po & (38) n, KE é= 2m Ete (39) q Mp0 QI Pno ‘Therefore, ai "ana = Mo exp ot (40) is abi Pea = Pod &=P fo (41) ? N In other words, the minority carrier concentrations f ; on one side of the depletion region are related to the bo yr [Fai ~My = Mn majority cartier concentrations on the other side by _—_ J p Ve | + a i. Just as for reverse bias, we replace ; by (4; — Va) : 1 |e! and with the result that: pa ‘ 1 oh P, hoo . n c a(¢i~ Ve) T~ TetRy a(n) = Me(—ap)exp ME *), (42) ieseee lia - t cas a(¢i= Ve) Ke h Pe(—Zp) = Pa(ta)exp“T =, (43) (the subscript 0 has been eliminated because we are no longer in equilibrium.) 19Low Level Injection ‘We will frst assume that ty = rno and pp = 7,0, the low level injection condition, as for light generated carriers, is that the majority carrier concentrations are unaffected by the injection process. Using the above equations then, bi ad: — Vo nyo exp 40 = ng(—29) exp we), (44) Va tp) = Myo ex G (45) Similarly for holes: Va Pa(2n) = Pro exp Tt (48) ‘These equations provide boundary conditions for the minozity carrier concentrations at the edge of the depletion region which can be used in calculating current densities. Summarizing, under low level injection: nj and pp are unaffected by injection. @ np and pp, are related exponentially to myo and pao as exp(qVa/kT). # nyo and pyo are related exponentially to nao and pyo as exp(—a¢i/kT). Due to the injected carriers: 1. The injected carriers set up an & field in the both the N and P regions. 2. This field causes majority carriers, which are present in much larger numbers and thus respond more strongly to electric fields, to move in order to re-establish neutrality and eliminate the € field. 3. At the same time, the excess injected minority carriers diffuse away from the junction into the neutral regions. 4, These minority carriers recombine with majority carriers with the recombination process taking place over some distance. 20xt At right is pictured the carrier concentrations in | m0) 10" en? | the N region as a function of distance. The majority carrier concentration is essentially unchanged (nao = = 8 Pa)- 4 Ng = 10em~3 > 10%em: ‘The injected hole concentration decays to its equi- “| librium value over some distance. x08 10 om 2 egomen pare ‘The € field which exists in the regions where there is an excess of injected carriers causes @ drift current which in the N zegion is given by Gintiy€ for majority carriers (e~) Taste = Tene (47) pPn€ for minority carriers (h*) Since m, > Pa; the drift of minority carriers is neglible. ‘Therefore, the minority carriers move primarily by diffusion (proportional to gradient not concentration) while for majority carsiers the drift and diffusion terms nearly balance. ‘The minority carriers control the behavior of PN junctions. This is in contrast to Schottky diodes where the majority carriers dominate the behavior. We will assume that the injected minority carriers move away from the junction by diffusion alone. This is known as the Diffusion Approximation and is only valid for low level injection. In the quasi-neutral N region: a dn ~qAD eT (48) 8pm _ py OPm Pa Pro Paen otaatiricees (49) ‘The second equation is just the continuity equation for holes. In the quasineutral P region: (50) (51) a1Ideal Diode I-V Characterisics By solving equations (48) - (51) with appropriate boundary conditions, we can find the current flowing in a diode with an applied V,. Long Base Diode Recombination reduces pp and n, to equilibrium values before ohmic contacts. In steady state: Oat, dt (52) (53) ‘The appropriate boundary conditions are: We pax {POXP Er FH Fn Pro, z= 00 We ty = 5 OOP Ep p05 22‘Thus we have > mq (54) <2, (58) where L, = \/Dyr, = hole diffusion length /Datm = electron diffusion length The hole and electron currents are given by substitution into (48) and (50) ) exp 22, z>2, (58) z+z, ony 2<-2, (57) Note that: pore ! type ‘The minority carriers decrease with distance From continuity of I, the ma- Jotity currents must therefore in- crease with distance (i.e. minor ity current is continuously trans- formed into majority carrier cur- rent). 23‘The total current flowing across the junction is thus given by the sum of I, and Jy at the edges of the depletion region (2 = 2, and respectively). Trot = Iplz=en + Int; Dypnc I=¢A (2e + (58) I= (exp (59) where Iy=9A ( - (60) oe a(_Dp =9An; ( Wal; * Wok © This last result illustrates particularly well that it is the doping level on the lightly doped side of the junction that determines the current. Example: N side is lightly doped, therefore therefore py >> ny and N side controls amount of current. Current is mostly due to injection into lightly doped side of junction. Short Base Diode A second limiting case of considerable practical importance occurs when the widths of the neutral regions are « Ly and/or Ln. A particularly important example is the base region of a bipolar transistor.The surface recombination velocity of the ohmic contacts is ~ oo because of the large number of traps available there. Therefore pa(z = Wn) = Pao and np(z = —W,) = nyo. If we assume that no recombination occurs within W,, then the minority carrier current densities (due to diffusion) are given by 4 Jp = Dye (61) a= Dae (62) But these currents must be constant (no change in majority carrier current flows because no recombination). Therefore ap dn & and © = constant ge and G, = constants and hole and electrons concentrations decrease linearly with distance as shown, ‘Under these conditions, dp =gAD, > GAD, AD, Fe (oe -1) (63) Similarly (64) The total current is again given by the sum of the two minority carzier currents, so that D, NW, T= qAn? [ ra] (ex iG = 1) (65) * Comparing the form of (65) with the equivalent expression for the long base diode (58), we see that the two equations are identical except for the characteristic length appropriate to the geometry (pn ot Way). 25Summarizing we have: Dy, D, Nel,” Nola An? [ | Long Base sane | Pos. Pi] Short Base 1= h(a %—1) Actual diodes may be intermediate cases, in either case, it is usually the lightly doped and/or narrow side of the junction which mostly determines I. Assumptions ‘© Ohmic drops in neutral region are neglible. Applied voltage is dropped across de- pletion region. Generally true except under high level injection. © Quasi-equilibrium assumes that injected currents are small relative to the normal compensating drift and diffusion currents across the depletion region. We required this assumption to calculate pa(2,) and n,(—z)). This assumption is also generally valid except under high level injection. 26Space Charge Recombination Currents We have assumed to this point that all the minority carriers which enter the depletion region cross to the other side. This is not true in practice since some recombine through trapping centers in the depletion region. From previous notes for + = 7, = 7p! pan; [ptr rincon FE] a Under reverse bias, these centers act as generation sites since np < n,? and increase the Us Teverse saturation current over the ideal value calculated earlier. Under forward bias (np > n;2) the centers act as recombination sites increasing the forward current. In the depletion region, both n and p are small. The generation or recombination rate is maximized when n = p. ‘The the G-R process is dominated by that portion of the depletion region where that relation approximately holds. With an applied voltage, monte so for maximum |U|, (07) and if = B;, (68) Va 2m (ex G + 1) where 7y is the lifetime of cazsiers in the depletion region. (As we saw in the homework, 2ro could be replaced by 7» + 7p in the depletion region if the lifetimes are different.) Forward Bias: If V, > 6kT'/q then mo We Unen © exp Se (69) ‘The recombination current will increase as the exponential of gV,/2kT. Tree = QA A " U(e) de (70) 27If we assume that the recombination rate is near its maximum throughout the depletion region, we tnt Ve Tee or PP OEE i) The total current is the sum of the diffusion current already calculated and the re- combination current. Ve ge oe oer: Trorwara © In exp Fe (72) ‘Therefore, the recombination current will dominate for small forward biases, while for larger forward biases the ideal diode current will dominate. Reverse Bias: Under reverse bias, Iyec generally dominates the ideal reverse leakage current Io. Under reverse bias pn < nj so (73) (74) ‘This current increases as the reverse bias increases since the depletion region width in- Region Region Region wor par ow Series resistance due to bulk ohmic drops recomb inet on anreut ot awall Va Practical Implications toa fo © Recombination currents at low levels result in f roll-off in bipolar transistors at low current levels. © Adding gold to switching diodes or transistors to reduce 7 (improve speed) will cause larger reverse currents (higher leakage currents). 28Charge Storage in PN Diodes Consider the long base diode shown below (one sided) alnost no Unjection on pt Side sumuce pro> N If the forward bias is removed at t = 0 and replaced by a reverse bias, the stored holes will flow back across the depletion region resulting in a reverse current until they are all removed. ‘The driving force (V;) for injected holes is removed at t=0, The concentration de- creases to equilibrium by flow backwards across the barrier. 4 rate Tine ° ae ii ae 4 ® ee 7 Je j oy lls Yat & . ” o 29‘The total charge stored in the N side of the diode is given by =0A |" apes 7 Q.= 04 ["" apde (75) The continuity equation for holes is given by 10J, Ap__ op ont. (76) Integrating from z = 0 to z = Wa, we obtain ph Geet [ Paem— [Bae Using (75), we obtain 4 Pala) — FO) + ce Ts0) ~ Iy(¥7q) = 28+ 4 & (r) ‘This is the charge control equation and relates current to stored charge. In the long base diode, J,(Wp) = 0 and under steady state, forward bias conditions, dQ,/dt = 0, so that Ty = Ip(0) Qe (78) te © The stored charge is directly proportional to I; and to >. If we reverse the current direction, equation (77) becomes — 421 wha te (79) ‘This yields as a solution Q(t) = [- + (Is + In) exp (80) 30The stored charge goes to 0 when Q,(t) = 0, so that rpln (+2) (81) Tr. 4, 9 if | a Hee ‘Cons tan 3 “1 i bee ¢ w Note: 1, The storage time is proportional to 7p. Therefore, Au doping or damage to reduce Tp reduces storage time. 2. Diode supports full reverse voltage after t, and an RC decay takes place due to diode capacitance and circuit-diode R. 3, Increasing reverse current decreases t,. 4, Higher forward bias leads to longer t, since more charge is stored. 31Circuit Models for Diodes A simple circuit model for the PN junction is as follows: Ks Ww Note that this is a small signal model because all parameters are f(V). BR, = diode series resistance (ohmic contacts, bulk resistance) ge=diode conductance (1/diode resistance) (82) aH Migexp Mt we Ly dv kT AT "kT Gadepletion region capacitance a (83) qKeo ‘ 5 @) = : step junction a fusion capacitance (due to stored minority charge) 84) alot», Wa O EP? er All of the parameters except R, depend on V;. For large signal calculations, piecewise linear models can be used or computer calculations. 32Heterojunctions Heterojunctions are interface between two different semiconductors. They are grown epitax- ially and require an approximate lattice fit between the layers in order to grow single crystal materials. They can be isotype (n-N or p-P) or anisotype (p-N or n-P) where the capital letters signify the material with the larger bandgap. AE. = X1—X2 AB, = AE, ~ AE, ‘The amount of band bending in each material can be calculated using Poisson’s Equation with different dielectric constants. © The barrier for hole # the barrier for electrons. —— hs. __ =i) oa N2GeAs | ye AlGaAs bey 33Light Interactions with PN Junctions Absorption and Emission of Photons vp, sega ee TAREE CMO MINA ali rmiatetil Tons (4) eK) Zs 300 cus 300 Gat! 300 Gass os 30 s 12 300 Ge Ls 300 Inas x0 7 ns 3S n He CeTe sia n Po'sate ia n Gece 30 1 Absorption Let ¢ = photon flux. The generation rate of hole-electron pairs then is ae) = ~$ = oe) Cc) where a(A) is the absorption coefficient. #2) = nbexo(—a(0)e) (68) where $o = P/E, = P/hv = PX/he is the incident photon flux and 1 is the percent of incident photons which are not reflected. 9(2) = 70d exp(—az) (87) 34sen) For\ > Ames, @—> 0. For A < Ames Goas NS @ becomes very large for direct band 7 * gap materials. For indirect band gap materials like silicon however, for Eyy ~ i E, phonon interactions are required so @ remains small until Ep, ~ Exivect Oar tae Figure 10... Absorption coefficient a(2) for Si, Ge and GaAs at 300'K. For a photoconductor, what is important is the total amount of generation. e nett Th a= [ o(2) de = 160 [1 ~ exp(—ad)] (88) For d—+ 0, g + 0. For d+ co, g: + no. The optimum thickness is d= while for a thick material, too much of the material is unaffected by -25/a since for a very thin material, very little light is absorbed, (i the light so the percentage change in conduction is small. If light shines on a diode, the carriers created in the depletion region result in Jon. Minority carriers generated within a diffusion length of the depletion region also have a good chance ending up as current by diffusing to the interface before recombining p(cartier + current) ‘Therefore, the I-V characteristic is shifted by the photocurrent due to generation in or near the depletion region. In= [aA b g(2,A) dz dd =04 J o(A)n() exp (—a(A)21) — exp (—a(A)es)] €d (89) Ligh ~1)- Jaton (90) 35Hy Aniieetection Contact eaating PIN Photodetector f° ew owe — @ e Rot DA oF Ei te 6 A diode can be used as a photodetector by operating it under reverse bias. In order to get maximum current for a given incident flux, 21 should be made as small as possible and 2y as Tange as possible. 2; can be reduced by making the top region narrow (Wx < L,) so that only a small percentage of the photons are absorbed in that region and most of thote that do diffuse to the interface before recombining. If, however, the light is absorbed too strongly near the suzface, the large recombination velocity of the interface will cause a loss of carriers. In order to make 22 large, the depletion region should be made as wide as possible. This can be accomplished by having a very lightly doped region (nearly intrinsic) between the P and N regions (pin structure). For these limiting cases Jun 94 f do(X)n(2) [0 — exp(—a(A)W7)] ad (1) where W = W, + 24+ Dn. ‘The response time of a diode photodetector is limited by the transit time across the depletion region, which for the pin structure is almost all in the intrinsic region. né pV A photodetector can also be operated at the edge of avalanche, causing avalanche multiplication to magnify the current but such diodes are also much more noisy and behave nonlinearly. The minimum detectable signal is approximately the reverse leakage current (Jp) plus current due to background radiation. An alternative method by which to increase sensitivity is by using a phototransistor, light shining on the base creates hole elecron pairs which as as base current and become multiplied by the gain of the transistor. 36Photovoltaic Effect A photodiode can also be operated in the fourth quadrant. In this case the illuminated diode acts as a source rather than a sink. The short circuit current is —Ipn. Under open circuit conditions given by Isc I=050 Io (ox Tee - 1) 25, and the open circuit voltage is given by Voo = HP in (B43). q Solar Cell A photodiode can also be operated with a load to produce power from the incident radiation. big bf yet See Vp = -IRp = Vy + IR; —~ WA-em p= TYPE si oeD /iuckt ils Ya "p(T +2). we Ry Jon +1 venga) 0 ORR htt % cee _ In +I “tod. 4d + gn -IV; = -CIin (H+ 1). (92) = where a AT Ry c=? r q Ret Ry, t* 7, i a: Y ta :In practice, Ip +1 > Io so (93) The power vutput is maximized when dP/dI = 0 or Tatty 7 1a( Z)= Tatt (94) Equation (94) must be solved iteratively. If the solution is Jm then : ppt petted (35) Tint Im ¢ Ret Roly + In Tigh power zequires low Jp (larger Vj) and low R; (lower losses in neutral regions). Heavy doping reduces both Jo and R;, but also reduces depletion region width and eventually reduces minority cartier lifetime, mobility and diffusion length and thus Jpn. Practical solar cells use shallow, heavily doped n-regions (~ 10") on top of wide, more lightly doped p-regions (~ 10"). “Comb” contact is used to reduce R; due to shallow n-region. Antire- flective coatings are use to absorb maximum amount of light. Maximum achievable efficiencies are on the order of 15% with Voc ~ 0.6V. Parallel and series connections are used to produce higher current and voltage. a Boron “contact 1 —Top o 38Light Emitting Diodes For light emission we need excited carriers. These can be provided by two different injection mechanisms: 1, Generation by absorbing photons or high energy electrons 2. Injection of carriers from nearby regions (ie. forward biased diode) Recombination will only result in radiation if radiative (direct) recombination dominates. i 2 (sloo) m mothe; hosphors dx. Win, visible ot (Lamps) ey Optical Coupling A combination of a LED and a photodetector can provide transfer of information while keeping the source and detector electrically isolated (no unwanted feedback). Optical communication links using optical fibers can transmit large quantities of information due to the small wavelength of light. co) ) Figure 10.8. Optical coupling: (a) photon-coupled isolator; (6) use of two isolators to obsain a signal-chopping circuit. 39Semiconductor Laser There are several requirements which much be fulfilled for lasing to occur. 1. Need direct band gap for recombination to occur predominately by photon emission, 2. Need population inversion. There are three interactions that can occur between light and a semiconductor. hv = Ey — By & hy (a) Absorption E Ee, (b) Spontaneous Emission hy &, yo 2 (c) Stimulated Emission ne, Jw phe cd é, Based on quantum mechanics, it is known that the rate constant for stimulated emission is the same as for absorption. Which process dominates depends on the relative numbers of full and empty states. P(absorption) « p(full state at E1)p(empty state at Hy = By + hv) p(stimulated emission) oc p(empty state at B;)p(full state at E2) Warren = Kf NB) = f(B)] No(B ~ hv) f(E ~ hv) Wenission = K / N.(B)f-(E)N,(E = hv) [1 — f.(B — hv)] ab where _ 1 Tre ES) i Tee Ear) 40In a normal semiconductor, the valence band is mostly full and the conduction band is mostly empty and most photon interactions will result in absorption. Under conditions of very high level injection the quasi-Fermi levels can move into the bands so that the bottom of the conduc- tion band is mostly full while the top of the va- lence band is mostly empty. In that case, known as population inversion, most photon interac- tions will result in stimulated emission rather than absorption. For this to occur Ey < hv < (Em — Ejp). Therefore, similar to the case of avalanche breakdown, each incident photon can create more photons all with the same phase. ‘The gain is given by g(v) o Werissioa—Wabsorption- starts turco NY t 8, Li RI ‘The final requirement is confinement. In other words, the number of photons created in the active region must be more than the number lost to the rest of the semiconductor where absorption dominates. Lasers are made with cavities that confine the light. The most common type of cavity is reflective on two sides and roughened on the other two, in order to get only a single operating mode. The light is magnified in the direction with the reflective sides. tte Sa ericatty roar ayo, ranaliet Bae FABaY-PEROT caviTY 41In order for lasing to occur, light must completely traverse cavity without attenuation. Rexp((g—@)L) >1 where Ris the reflectivity of the end facets, a is the attenuation factor due to free carrier absorption and scattering and L is the length of the cavity. The minimum gain required then is =attin Gmin = OT 1 e ‘The size of the cavity determines the exact wavelength at which the laser will operate since reflection can only occur for constructive interference of the incident and reflected waves, ‘Therefore, the length of the cavity must be an integral number of half-wavelengths. where n, is the index of refraction, Heterojunction Lasers ‘The most efficient lasers are made using heterojunctions. The varying com- position can both concentrate the re- combination process in a small region while also containing the photons by the differences in refractive indices of the layers. These lasers can be modulated very quickly (up to 2GHz) and by altering the composition of the layers, the out- put wavelength can be controlled. Since the active region is very small relatively small current levels can sustain the Ias- ing action (10-20mA). N= AlgaGu024s roa(e) ad ersssvon Si02 a 9 =n P= RioaGra3 8 porns n-GaAs 42