The final power device to be considered is the pnpn thyristor. The thyristor is a four-layer, three-junction power switch that can be thought of as two merged bipolar transistors, as illustrated in Figure 9.27. The pnp BJT has a narrow base, and its emitter serves as the anode. The npn BJT has a wide, lightly-doped base, and its emitter acts as the cathode. The base of the pnp BJT is provided with an external contact called the gate, whose purpose will be described below. SiC thyristors are typically fabricated on substrates because of the high resistance of substrates.
As seen in Figure 9.27, the collector of the pnp BJT is the base of the npn BJT, and the collector of the npn is the base of the pnp. Consequently, when both BJTs are conducting there is no need to supply external base current, since the collector of each BJT supplies base current to the other BJT, and conduction through the four-layer structure is self-sustaining. If both BJTs are off and there is no external supply of base current, the structure is non-conducting. This gives rise to two distinct regimes for , as illustrated in Figure 9.28. These are the forward blocking mode and the forward conducting mode. In the blocking mode the thyristor holds off a large with very little current, while in the conducting mode the thyristor carries a large current with very low voltage drop. Forward operation is therefore bi-stable, since the device can be in either of two regimes at the same . The thin dotted lines depict a negative resistance regime where the anode–cathode voltage snaps back to a low value once a triggering threshold is reached. The triggering threshold can be varied by supplying external base current to the pnp BJT through the gate terminal. Once the BJTs begin to conduct and conduction becomes self-sustaining, the gate current can be removed and the device will remain in conducting state until the current falls below the holding current , or is reduced below the holding voltage . We will now consider the two forward modes in more detail.
In the forward conducting regime both BJTs are operating in saturation, and all three junctions in the thyristor are forward biased. As stated above, forward conduction is self-sustaining, but this requires that the increase in current around the loop formed by the interconnected collector and base regions is unity. The increase in current in each BJT is the ratio , which is the common-emitter current gain , so we require that . Since , this is equivalent to requiring that the sum of the alphas is unity, or .
To analyze the conduction in more detail, we refer to the one-dimensional slice shown in Figure 9.29. Here we illustrate the hole and electron densities as a function of position throughout the device. As noted, all three junctions are forward biased and injecting carriers. The p− drift region is in high-level injection, but we assume all the other regions are in low-level injection. This is reasonable, since their doping levels are orders of magnitude higher than the drift region. We also assume that the gate region is short compared to the minority-carrier diffusion length, that is, .
The total anode-cathode voltage is the difference in the quasi-Fermi levels from anode to cathode, as illustrated in the band diagram of Figure 9.30. With reference to the band diagram, we can write as
We first consider the potential drop across the drift region, . Since this region is in high-level injection, we can set and obtain an expression for by solving the ambipolar diffusion equation, subject to the appropriate boundary conditions. We assume unity injection efficiency at both ends of the drift region, which allows us to set and . Setting is equivalent to assuming the hole current diffusing across the n-type gate region from the anode is much larger than any electron current injected into the drift region from the gate. Under these assumptions, the development is the same as for the pin diode in Section 7.3, and the carrier densities are given by Equation 7.46, which can be written
Here , and refer to parameters in the lightly-doped drift region. To obtain the potential drop across the drift region, we again proceed as with the pin diode: We write an expression for the total current as a function of the electric field, solve for the electric field as a function of current, then integrate the electric field to obtain the potential as a function of position using Equation 9.112 for the carrier densities. We then evaluate this expression at the two ends of the drift region to calculate the potential drop. The result is the same as Equation 7.51, and can be written
where and . is the change in electrostatic potential, or the change in the intrinsic Fermi level , across the drift region.
We will next obtain an expression for the sum . Referring to the band diagram in Figure 9.30, the electron density at can be written
where is the difference between the electron quasi-Fermi level and the intrinsic level at , a quantity known as the chemical potential for electrons. Likewise, at the hole density can be written
where is the difference between the intrinsic level and the hole quasi-Fermi level at , or the chemical potential for holes. Solving Equations 9.114 and 9.115 for and and adding yields
When evaluated in equilibrium, Equation 9.116 gives the built-in potential of the junction. We can express Equation 9.116 in terms of current using Equation 9.112 to evaluate and , resulting in
where is
We note that this treatment parallels that of the pin diode, and Equation 9.118 is simply a re-statement of Equation 7.59.
The quasi-Fermi level splitting at the GD junction can be obtained from the law of the junction evaluated at ,
Solving for and employing Equation 9.112 for yields
Finally, we need to calculate the quasi-Fermi level splitting at the AG junction, . The anode and gate are assumed to be in low-level injection and the electric fields in the quasi-neutral regions are negligible, so current flow in these regions is due to diffusion. The total current crossing the AG junction is the sum of electron and hole diffusion currents, and can be written
where is the hole diffusion coefficient in the gate region and is the electron diffusion coefficient in the anode. The electron and hole densities in the anode and gate regions are obtained by solving the minority carrier diffusion equations in each region. If the anode is long compared to the electron diffusion length, the electron density in the anode can be written
From this it follows that
Since the gate region is short compared to the hole diffusion length, the hole density in the gate is a linear function of position, and we can write
Using the law of the junction, the product throughout the GD depletion region can be written
Thus
and we can write Equation 9.124 as
Inserting Equations 9.123 and 9.127 into Equation 9.121 and using Equation 9.112 for yields
Equation 9.128 is a quadratic equation in , and can be written more compactly as
where
We note that and are constants that depend only on device parameters. Solving Equation 9.129 for gives
We now express in Equation 9.132 in terms of using Equation 9.111,
Inserting Equation 9.120 for and Equation 9.117 for into Equation 9.133 allows us to write
where
Once again we note that the new parameter is a computable constant that depends only on device parameters. Inserting Equation 9.134 into Equation 9.132 gives the desired expression for the current–voltage characteristics in the forward conducting regime:
At the high current densities at which thyristors typically operate, the exponential term in the square root dominates all other terms, and we can write
where
Here and are constants that depend only on device parameters, and are given by Equations 9.118 and 9.113 respectively.
The result of this long derivation is perhaps somewhat anticlimactic. Equation 9.137 tells us that the thyristor current in the forward conduction regime increases exponentially with , with an exponential slope of 1/2. This is the same dependence as the pin diode, Equation 7.57. Moreover, the prefactor in Equation 9.138 is identical to the prefactor of the pin diode in Equation 7.58 except for the square root factor, which has a value close to unity. Thus we conclude that the forward characteristics of the thyristor are almost identical to those of a comparable pin diode.
The above analysis assumed that the anode is thick compared to an electron diffusion length, . If the anode is thinner than a diffusion length, we simply replace with in all the foregoing equations.
We turn now to the forward blocking regime and consider how the thyristor is switched from the forward blocking to the forward conducting regimes. Figure 9.28 contains a simple circuit with the thyristor connected to a resistive load. Kirchoff's current and voltage laws tell us there are two stable operating points, formed by the intersections of the resistive load line (dashed) with the thyristor I–V characteristics. If the supply voltage , the thyristor is in the forward blocking mode at the origin. As is increased with , the operating point (1) moves laterally along the forward blocking characteristics until becomes greater than the forward blocking voltage , when the circuit abruptly switches to operating point (2). As we shall see, the switching threshold can be reduced in a controllable manner by drawing current out of the gate.
The switching threshold is best understood by considering the internal currents in two cross-coupled BJTs, as illustrated in Figure 9.31, where the npn BJT is designated BJT-1 and the pnp BJT is designated BJT-2. The cathode current can be written
Solving for gives
Equation 9.140 shows that the cathode current becomes infinite when the sum of the alphas equals one. The alphas are evaluated in the forward blocking mode where very little current flows and all regions are in low-level injection. Equation 9.10 defines the common-base current gain as the product of the base transport factor and the emitter injection efficiency . For wide-base npn BJT-1 the base transport factor is given by Equation 9.11,
The emitter injection efficiency is the fraction of emitter current due to electrons injected into the base, so . Equation 9.10 specifies the emitter current as , but when the currents are small, as in the forward blocking regime, must also take into account the current due to recombination in the EB depletion region. With this in mind, can be written
Here is the recombination current in the depletion region of the EB junction, and due to the large doping asymmetry between the emitter and the base. For an one-sided step junction, the recombination current can be written
where A is the junction area and the generation lifetime is the same as the ambipolar lifetime . Note that is the total dopant concentration in the base, rather than the ionized concentration. Combining Equation 9.142 with Equation 9.7 and assuming the CB junction is strongly reverse biased, we can write Equation 9.141 in the form
Equation 9.143 can now be combined with Equations 9.10 and 9.11 to calculate .
Examining Equation 9.143, we see that in the low-current regime is no longer a constant, but instead depends on and hence on the current. When is small, approaches zero. As increases, increases, and at sufficiently high . Figure 9.32a is a plot of versus for several values of , using Equation 9.143 for and Equation 9.7 for . As seen, (and therefore ) increases monotonically with current. The same general arguments apply to the narrow-base pnp BJT-2.
The dependence of on current can also be illustrated using a Gummel plot, shown in Figure 9.32b. This presentation depicts and on a log scale as a function of , and we have also included and to aid the discussion. For is dominated by recombination in the depletion region, Equation 9.142, whereas for is dominated by hole injection into the emitter, Equation 9.2. Recombination in the neutral base is small, since the base is short compared to a diffusion length. At high currents the ratio is constant, independent of current, but at low currents and increase with increasing current.
The thyristor's triggering threshold is reached when the sum of the alphas equals one, as shown by Equation 9.140. Considering the thyristor in the forward blocking regime, we can now ask how the alphas of the internal BJTs depend on . As is increased, the alphas increase through two mechanisms. The first mechanism is base width modulation. Figure 9.33 illustrates the depletion regions in the device in the forward blocking mode. The middle junction is reverse biased, and its depletion region spreads mainly into the lightly-doped drift region, the base of BJT-1. As increases, expands, shrinks, and increases. The second mechanism is more subtle. As increases and widens, the thermal generation current within the depletion region of the middle junction increases. The generation current is identified in Figure 9.33, and can be written
where we have assumed all the applied voltage appears across the reverse-biased middle junction. As seen, the generation current increases as the square root of . Since , Figure 9.33 shows that this generation current must flow through the emitters of both BJTs. When the currents are small, as they are in the forward-blocking mode, is an increasing function of current, as shown in Figure 9.32a. Thus, increasing increases , which increases the current, which increases .
Although the above analysis of the triggering threshold appears reasonable, it turns out that triggering does not actually require to become infinite, as suggested by Equation 9.140. Instead, triggering only requires that becomes infinite. When this condition is met, any small fluctuation in gate current will produce an unlimited increase in cathode current, and the thyristor switches on. By reference to Figure 9.31 we can write
where and are small-signal alphas defined by
If , the small-signal will be greater than the DC , and this is usually the case. Returning to Figure 9.31, we can also write
Substituting Equation 9.147 into Equation 9.145 yields
Equation 9.148 indicates that the thyristor actually triggers when the sum of the small-signal alphas equals one. The small-signal alphas are defined by Equation 9.146, but how do we calculate ? We know from Equation 9.143 that the emitter injection efficiency increases with current, so alpha also increases with current. Thus , as postulated, and the small-signal alphas are indeed slightly higher than the DC alphas. The expressions for the small-signal alphas can be calculated by writing Equation 9.143 in terms of current and then differentiating. The resulting expressions have been derived by Yang and Voulgaris [5], and the reader is referred to this reference for the details.
Once the triggering threshold is reached, the thyristor switches quickly from the blocking to the conducting state, and the internal BJTs go from their forward-active regimes to their saturation regimes. This process can be understood by reference to Figure 9.34. Here we assume the thyristor is driving a resistive load, and the gate current is zero. As the supply voltage is increased from zero, the thyristor is initially in the blocking mode, and the operating point moves along the lower part of the characteristic, as illustrated by points (1) and (2). The internal BJTs likewise move along their lines from point (1) to point (2), as shown. When the supply voltage reaches the switching threshold, which happens just beyond point (2), the thyristor switches to the conducting state and comes to rest at point (5). Once at point (5), if is increased or decreased, the operating point moves along the conducting characteristic, so long as the current remains above the holding current. For reference, the carrier densities and band diagrams in the thyristor at points (1) and (5) are illustrated in Figure 9.35.
We will now examine the trajectory in more detail. While the thyristor is in the blocking state, the product (remember the npn BJT has a very wide base, and the emitter injection efficiencies are very low when the currents are low). As approaches the switching threshold, three effects occur. First, the current gain of the npn BJT increases due to base width modulation as shrinks. This increases the product . Second, the widening depletion region of the middle junction increases the leakage current , which increases the injection efficiencies of both BJTs according to Equation 9.143. This also increases the product . Third, avalanche multiplication in the reverse-biased middle junction adds more current, further increasing the injection efficiencies and increasing . At some point will exceed 1. When this happens, the current around the loop formed by the interconnected collectors and bases begins to increase. This corresponds to the BJT operating point moving from (2) to (3), (4) and (5) as the base currents increase, as shown in Figure 9.34b. The situation stabilizes when the internal BJTs enter saturation at point (5). Why is this point stable? In saturation the effective , the ratio of actual collector current to base current in a BJT, is lower than in the forward active region. This reduces the product to 1 and operation stabilizes at point (5). To see why the circuit is stable at point (5), imagine that the cathode current tries to increase. With a higher cathode current, the voltage drop across the load increases, the voltage drop across the thyristor therefore decreases, the depletion width decreases, the base width increases, and decreases, reducing the current. If the cathode current tries to decrease, the opposite happens. These forces maintain the circuit at point (5).
The switching threshold can be reduced by supplying an external gate current . This acts as base current to BJT-2, and the collector of BJT-2 then supplies base current to BJT-1. These higher currents increase the betas, allowing to reach 1 at a lower value of . This makes it possible to trigger the thyristor at a specific under the control of an external circuit.
The turn-on process occurs in three phases that can be considered sequentially, as illustrated in Figure 9.36. The three critical times are (i) the delay time, (ii) the rise time, and (iii) the spreading time. We will discuss each of these in turn.
Let us assume the thyristor is initially in the forward-blocking mode, and is triggered to switch on by a gate current pulse that begins at . Conditions inside the device prior to are as indicated in Figure 9.35a; both BJTs are in their forward-active modes but there is essentially no minority carrier injection into either base region. Referring to Figure 9.31, the gate current pulse that begins at supplies base current to pnp BJT-2. This causes injection of holes from the emitter into the narrow n-type base. The holes diffuse across the base, and the first holes arrive at the collector of BJT-2 in a base transit time given by
where is the width of the neutral base and is the diffusion coefficient for holes in the base. Once the holes reach the collector of BJT-2, they act as base current to npn BJT-1. This causes the injection of electrons from the emitter of BJT-1 into the wide base. These electrons diffuse toward the collector of BJT-1, and the first electrons reach the collector in a transit time given by
where is the width of the neutral base and is the diffusion coefficient for electrons in the wide base. At this point, the regenerative build-up of current within the thyristor begins. The delay time is then approximately given by the sum of the base transit times, or
The second phase is characterized by regenerative current feedback between the two BJTs, and the current rapidly builds up until a steady state is reached. The current rise time can be calculated by considering the build-up of stored charge in the base regions of the BJTs. We assume the build-up of stored charge is due only to carriers that flow in as base current, and we neglect recombination since the rise time is short compared to the minority carrier lifetimes in the base. Again referring to Figure 9.31, the build-up of charge in the base of BJT-1 can be written
Likewise, the build-up of charge in the base of BJT-2 can be written
Focusing on Equation 9.152, we can write as
where the collector current is expressed as the base minority carrier charge divided by the diffusion transit time across the base (note that in the forward-active mode, , so Equation 9.154 is equivalent to setting , which is the usual expression for diffusion current in the base). Inserting Equation 9.154 into Equation 9.152 and differentiating yields
where we have used Equation 9.153 for . However, we can also write
so Equation 9.155 becomes
Equation 9.157 is a second-order differential equation describing the build-up of minority carrier charge in the base of BJT-1, and the solution can be written
where is the steady-state value of the cathode current. Equation 9.158 tells us that the charge in the thyristor base builds up exponentially with a time constant given by the square root of the product of the base transit times: . The rise time can be found by setting the base charge to zero at , resulting in
In a real thyristor, two-dimensional effects play a major role, and the third phase of the turn-on process represents the lateral spread of regenerative action from the region near the gate contact into the interior of the device. When the gate pulse is applied, charge injection occurs initially in the region near the gate contact, and regenerative action begins there first. As the charge builds up, this conducting region supplies base current to adjacent regions, and regenerative action spreads across the device until nearly homogeneous current flow is established. The time required for this stabilization is the spreading time . It has been found that the spreading is characterized by a spreading velocity , and the spreading time can be written
where is the half-width of the anode, as shown in Figure 9.27. In silicon, the spreading velocity has been found to be proportional to , with a typical value in the range to . The spreading velocity therefore increases with anode current density and ambipolar lifetime, and is inversely proportional to the thickness of the wide p-base region. The base-width dependence means that the spreading velocity will be lower for devices designed for high blocking voltages.
SiC thyristors differ from silicon thyristors in several important respects. First, due to incomplete ionization of acceptors in the heavily-doped anode, the emitter injection efficiency of the pnp BJT, , is less than 1. This is different from silicon, where the injection efficiencies are routinely assumed to be 1. As temperature is increased, the acceptor ionization increases, as shown in Figure A.1, and increases. One result is that the rise time decreases with increasing temperature due to the increased injection efficiency. Secondly, as noted by Levinshtein et al. [6], the turn-on process in SiC thyristors is more homogeneous than in silicon devices. In silicon, the three phases of the turn-on transient have the relationship , and the turn-on transient is dominated by the spreading time , which is to for a 4 kV device. In SiC, the three phases of turn-on transient are nearly equal, for the same 4 kV device. Thus the SiC thyristor switches to the conducting state about 3 orders-of-magnitude faster than a comparable silicon thyristor.
When the thyristor is in the forward-blocking mode, a rapid rise in anode-cathode voltage can cause the device to switch on prematurely. This is known as triggering, and can be understood as follows. The middle junction is reverse biased, and its depletion region extends primarily into the lightly-doped drift region, as shown in Figure 9.33. As is increased, the depletion region widens and the neutral base of the npn BJT-1 shrinks, increasing . In addition, the majority carriers moving out of the widening depletion region represent an internal current of magnitude , where is the capacitance of the middle junction. This can be viewed as base current to both BJT-1 and BJT-2, which in turn induces additional emitter current, causing the alphas to rise, as shown in Figure 9.32a. The increasing alphas and increasing base currents can enable the coupled BJTs to reach a self-sustaining condition, whereupon the thyristor switches on. The effect increases with both the magnitude and the first derivative of .
The effect can be reduced by (i) reverse biasing the gate with respect to the anode to prevent injection from the emitter of the pnp BJT, (ii) reducing the lifetime in the base regions of the BJTs to reduce the alphas, although this degrades on-state performance, or (iii) providing anode shorts. Anode shorts have the desirable effect of reducing the effective current gain of the pnp BJT when the current is low, while allowing the gain to increase rapidly at higher currents. The basic concept is equivalent to placing a resistor across the EB junction of the BJT-2, as shown in Figure 9.37. The effective current gain of this composite structure can be written
As shown by the characteristics in Figure 9.37b, at low currents and . However as increases, increases more rapidly than , and . The use of anode shorts also has the desirable effect of increasing the forward blocking voltage, as will be discussed below.
Anode shorts can be implemented by simply extending the top metal layer across the entire cell area in Figure 9.27, thereby connecting the anode ohmic contacts to the gate ohmic contacts. The lateral spreading resistance of the n-type gate layer provides the desired resistance, which can be adjusted by the spacing of the anode and gate contacts. The anode shorts do not interfere with turn-on initiated by a gate current pulse, provided the shunt current is small compared to the gate current .
When the turn-on process is initiated by a gate current pulse, regenerative action occurs first in the region adjacent to the gate contacts, then spreads laterally into areas further removed from the gate contacts. The time required for the thyristor to stabilize is the spreading time . During the initial part of the spreading transient, conduction occurs mainly within a very small area near the gate contacts, and the high local current density can lead to extreme local heating that can destroy the device. While this is an important limitation in silicon thyristors, it is much less severe in SiC devices for several reasons: (i) the thermal conductivity of SiC is twice as high as in silicon, (ii) the maximum allowable temperature is higher than in silicon, due to the lower intrinsic carrier concentration and more robust material properties, and (iii) the plasma spreading time in SiC thyristors is orders-of-magnitude shorter than in silicon devices, leading to more homogeneous turn-on.
There are two principal ways in which the thyristor may be turned off. If the supply voltage is reversed, the thyristor naturally transitions into the reverse-blocking mode. The cathode current immediately reverses direction, and a large reverse current flows until all stored charge is extracted from the interior of the device. This form of turn-off is used in AC applications where reverses every half cycle, and is the only method of turn-off for two-terminal thyristors (semiconductor-controlled rectifiers, or SCRs). In DC applications the supply voltage remains positive, and to quench the current the thyristor must transition to the forward-blocking mode. This can be accomplished by a suitable negative pulse applied to the gate. This method of turn-off requires careful design so that the reverse gate current can break the self-sustaining current feedback of the internal cross-coupled BJTs. Such a device is known as a gate turn-off thyristor or GTO. We will first describe the turn-off process by reversal, and then consider gate-controlled turn-off in the GTO.
Figure 9.38 shows the current and voltage waveforms for turn-off by supply voltage reversal. The turn-off process can be described qualitatively as follows. The thyristor is initially in the forward-conducting mode and driving a resistive load, as illustrated in Figure 9.39. At time the supply voltage is suddenly switched from to . The cathode current will immediately change sign, as stored carriers are drawn out of the device. The reversal of current only requires a change in the slope of the carrier distributions at the edges of J1 and J3, as illustrated in Figure 9.40a, and this can happen very quickly. However both junctions remain forward biased until the carrier densities at the junction edges drop below their equilibrium values. This can be understood by recalling the law of the junction, Equation 7.25,
which shows that remains positive until the product at the junction edges falls below . As a result, the voltage across the thyristor remains positive, although the current switches immediately from to , as shown in Figures 9.39 and 9.40. The current and voltage then remain almost constant during the first phase of the turn-off, shown as in the figure, as carriers are removed from the base regions of the BJTs. Since the pnp BJT has a narrow base, it discharges more rapidly than the wide base npn BJT, and at the end of phase 1 the carrier density at the edge of J3 drops below its equilibrium value, and J3 becomes reverse biased, Figure 9.40a. This marks the beginning of phase 2.
Because the emitter and base are heavily doped, J3 enters avalanche breakdown at a low reverse voltage, and drops to approximately , a negative value that is small compared to , as shown in Figure 9.38. The cathode current drops slightly as the voltage drop across decreases. The current and voltage remain constant during phase 2, shown as in the figure, and the next major event comes at the end of phase 2 when the carrier density at J1 falls below its equilibrium value and J1 becomes reverse biased, Figure 9.40b. This marks the beginning of phase 3.
Since the drift region is wide and lightly doped, J1 can support a large reverse voltage, and becomes increasingly negative as the depletion region expands during phase three, Figure 9.40c,d. As becomes more negative, the thyristor supports more of the negative supply voltage, reducing the voltage drop across , and the current falls, as shown in Figures 9.39 and 9.40. During the entire turn-off transient, J2 remains forward biased and the npn BJT operates in the inverse-active mode, with J2 forward biased and J1 reverse biased. The n-type collector of the npn BJT (the gate layer of the thyristor) injects electrons into the base, while the reverse-biased J1 sweeps electrons out of the base. Because of the low injection efficiency for holes from the base into the collector, very few holes are injected into the collector, and the hole charge in the base decays mainly by recombination. Recombination is a relatively slow process, and this phase tends to be the longest portion of the turn-off transient.
An estimate of the recombination time can be obtained from the charge-control equation for base charge,
where is the total hole charge in the base and is the ambipolar lifetime in the drift region. This assumes that the majority of the base charge is removed by recombination, and not by diffusion. Equation 9.162 has the solution
where is the total charge stored in the conducting state, and can be written
We now define the critical charge as the base charge corresponding to the holding current in the forward-conducting mode,
The turn-off can be considered complete when the base charge falls below the critical charge, since at that point the regenerative feedback is insufficient to sustain forward operation, and the device is incapable of spontaneously turning on. Combining Equations 9.165 and setting equal to the time when , we can write
is the approximate duration of the third period of the thyristor turn-off, which typically dominates the transient.
In the above discussion, the supply voltage is suddenly switched from a positive to a negative value at . However, in most practical applications the thyristor is operated with an AC power source, and the voltage reversal occurs at the zero crossing of the sinusoidal waveform. This tends to obscure the first two periods of the turn-off transient, and a triangular reverse current waveform is observed, followed by a recombination tail, as illustrated in Figure 9.41 for two values of . To shorten the current tail, we could reduce the lifetime in the base region as shown by Equation 9.166, but this would increase the forward voltage drop in the conducting state and also increase the leakage current that triggers avalanche breakdown in the forward-blocking mode.
In DC applications the supply voltage remains positive, and the thyristor turns off by transitioning to the forward-blocking mode, moving from point (2) to point (1) in Figure 9.28. This is accomplished in the gate-turn-off (GTO) thyristor by a negative pulse to the gate that diverts base current from the pnp BJT and breaks the regenerative feedback that sustains forward conduction.
We will first consider the GTO turn-off process using a simplified one-dimensional model, then we will discuss two-dimensional effects. The first question is: “What magnitude of gate current is needed to turn the device off?” To answer, we note with reference to Figure 9.31 that when the thyristor is stabilized in forward conduction, the base current of the pnp BJT is given by
Since this is the base current required to sustain stable conduction, we require a negative gate pulse applied to the base of BJT-2 that brings below this value, namely,
Note that has the opposite polarity to in Figure 9.31. Solving for yields
The negative gate current required to turn off the thyristor is proportional to the cathode current , and this ratio can be expressed in the form of a turn-off gain,
A large turn-off gain is desirable to simplify operation, and this requires that be close to unity and be small. A high occurs naturally in the GTO, since the base of the pnp BJT is narrow and the emitter is heavily doped. However, to take advantage of the high we must avoid anode shorts. is naturally low due to the wide base of the npn BJT, and it can be reduced further by inserting a thin heavily-doped buffer between the drift region and the substrate, as shown in Figure 9.27. This layer reduces the injection efficiency of the npn emitter, thereby reducing .
Since the GTO is designed for DC operations under a positive supply voltage, a high reverse blocking capability is not needed. This allows us to optimize the forward blocking voltage without regard to the reverse blocking voltage. The buffer layer has a beneficial effect in this regard, since it permits a punch-through design, as illustrated in Figure 9.42b. Here the buffer layer prevents the depletion region of the J2 junction from reaching the substrate in the forward-blocking mode, allowing the desired to be achieved with a thinner base. This reduces the forward voltage drop of the thyristor in the conducting state, as shown by Equation 9.113 and Figure 7.13.
The turn-off process in the GTO can be described in three phases, as shown in Figure 9.43. Conditions inside the thyristor during the transient are illustrated in Figure 9.44. In the first phase, stored holes are removed from the base of the pnp BJT by the negative gate current. The process is as follows: A negative gate current corresponds to the flow of majority electrons out of the base. This reduces the injection of electrons from the base into the emitter, which reduces . The reduced reduces the injection of holes from the emitter, lowering the hole density in the base. At the end of the storage phase, junction J2 becomes reverse biased and both BJTs enter their forward-active regions. During the second phase, the depletion region of J2 spreads into the base layer, and the reverse voltage supported by J2 increases. As a result, the voltage drop across the thyristor rises, and the current, limited by the load resistance, falls. The final phase corresponds to the recombination of the remaining electrons in the base of the npn BJT.
We now consider each of the three phases in turn. During the first (storage) phase, the injection of holes from the pnp emitter is reduced until J2 becomes reverse biased and the BJTs enter their forward-active modes. This process is inherently two-dimensional, since injection is first reduced in the regions adjacent to the gate contacts, and the quenching process spreads laterally into areas further removed from the gate contacts. This is the inverse of the turn-on process, where the electron–hole plasma spreads laterally from the gate into the interior with a spreading velocity . Here we talk of a squeezing velocity, as the injection is squeezed into a smaller and smaller region under the center of the anode before being quenched entirely. The situation can be visualized by reference to Figure 9.27. Since gate current flows laterally through the n-type base, a lateral voltage drop develops due to the sheet resistivity of the base, given by Equation 9.78, and the voltage drop across the EB junction of the pnp BJT is reduced as we proceed toward the edges of the anode. This reduces the injection near the anode edges, squeezing the current into a small region under the center of the anode. The effect of base spreading resistance in BJTs was discussed in Section 9.1.9.
An approximate analysis of the storage time given by Wolley [7] shows that increases with turn-off gain , and can be approximated by
Here is the hole diffusion coefficient in the gate layer, is the hole diffusion length in the gate layer, and is the thickness of the gate layer, that is, the base of BJT-2. Equation 9.171 shows that there is a conflict between the desire for a high to reduce drive requirements, and a low to speed the turn-off process.
At the end of the storage phase, the entire area of junction J2 is reverse biased. During the second phase, the depletion region of J2 expands into the base of the npn BJT. As this occurs, the reverse voltage supported by J2 increases. With more voltage dropped across the thyristor, the current (limited by the load resistance) falls. An approximate expression for the time for the current to fall to 10% of its initial value can be derived by considering the rate of expansion of the depletion region [8]. The result is
where is the average electron density in the undepleted portion of the drift region. We note that the fall time is proportional to the square root of supply voltage and inversely proportional to the initial cathode current density, .
The reverse-blocking mode is important for thyristors that operate under AC conditions, where the supply voltage oscillates between positive and negative values. We will discuss avalanche breakdown and blocking voltage in Chapter 10.