136 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. PE-2, NO. 2, APRIL 1987

Turn-Off Failure of Power MOSFET'sDAVID L. BLACKBURN, MEMBER, IEEE

Abstract-Experimental results of the failure of power MOSFET's shown to be identical to that of a bipolar transistor expe-during inductive turn-off are discussed. The electrical characteristics riencing second breakdown. However, if the power MOS-of these devices during failure are shown to be identical to those of a

bipolar transistor undergoing second breakdown. Other comparisons Fetsa ache itsdain-source saigltg andof the power MOSFET failure and bipolar second breakdown are made. begins to sustain, it usually will turn off safely. This isA nondestructive measurement system is used that allows repeated contrary to what has been found for bipolar transistorsmeasurements of the failure characteristics as a function of various [10], [11]. It is demonstrated, as has been done for bi-parameters to be made on a single device. It is shown that commer- polar transistors [10]-[12], that power MOSFET's can becially available power MOSFET's do not fail as a result of dV/dt cur- stressed beyond the "normal" limits of failure withoutrents. Drain voltage slew rates up to 22 V/ns were studied. Other mea-surements show that the drain voltage at which failure occurs increases actually destroying or degrading the device. Examples ofwith temperature, the critical current above which failure occurs de- the failure characteristics for inductive load turn-off as a

creases with temperature, and the magnitude of the load inductance function of temperature, load inductance, and drain volt-has no effect on the failure. The results of this study are consistent with age rate of rise (dV/dt) are given. It is found that thethe theory that activation of the parasitic bipolar transistor initiates limits of safe operation depend upon temperature but arethe power MOSFET failure during turn-off,

independent of load inductance and dV/dt.

INTRODUCTION BACKGROUNDDURING the past several years, power MOSFET's At the time power MOSFET's were first introduced as

have become established in a wide variety of power practical commercial devices, it was usually noted that,control and conversion applications. Most of these appli- because the MOSFET is a majority carrier device, powercations require that the MOSFET be switched on and off MOSFET's were immune to second breakdown as ob-at a high frequency. The thermal and electrical stress that served in bipolar transistors. It was quickly recognized,the device undergoes during switching can be quite se- though, that the presence of a "parasitic bipolar transis-vere, particularly during turn-off with an inductive load tor" inherent in the power MOSFET structure (see Fig.at the drain terminal. 1) might allow a "second-breakdown-like" failure to oc-The possible failure of power MOSFET's under various cur [1]-[3]. It is now generally assumed that if the para-

operating conditionss has been discussed, with several dif- sitic bipolar transistor becomes active, particularly if theferent failure mechanisms proposed [1]-[6]. It is com- MOSFET drain-source voltage is greater than the collec-monly assumed that any mechanism that permits the tor-emitter sustaining voltage of the bipolar transistor, the"parasitic" bipolar transistor to become active will usu- power MOSFET may fail. Typically, the power MOS-ally lead to failure of the power MOSFET, particularly if FET drain-source breakdown voltage is about twice thethe drain voltage is greater than the collector-emitter sus- collector-emitter sustaining voltage of the bipolar transis-taining voltage of the bipolar transistor. It has also been tor. The failure would occur presumably because the de-proposed that avalanche injection [7], [8] in the drain re- vice drain voltage would "snap-back" to the sustaininggion of the MOSFET may lead to failure without the par- voltage of the parasitic device. This "negative resis-asitic bipolar transistor becoming active [9]. However, tance" characteristic would cause the total device currentthere has been little discussion of failures of state-of-the- to constrict to one or a few cells of the MOSFET struc-art commercially available devices, particularly in rela- ture.tion to how circuit parameters and device temperature af-fect limits of safe operation during high-speed inductive Bipolar Transistor Second Breakdownload switching.Thed purposeng. oth paeitdicst filef For bipolar transistors, it is well-known that the phe-The purpose of this paper iS to discuss the failure of

noeo fscn radw cue aatohcdvcpower MOSFET's during inductive load switching. The nomenon of second breakdown causes catastrophic device

failure. The term "second breakdown" refers to the factelecrica behviorof te poer M SFETat filur 1S that the collector voltage collapses very rapidly at the on-

Manuscript received September13,1985. This paper was originally pre- set of this failure (the "first breakdown'' is the self-sented at the Power Electronics Specialists Conference, Toulouse, France, clamping of the voltage due to normal avalanche or carrierJune 24-28, 1985. multiplication, i.e., sustaining). At second breakdown,

The author is with the Semiconductor Devices and Circuits Division, th otg a olpehnreso ot nlshn1National Bureau of Standards, Gaithersburg, MD 20899. tevlaemyclas udeso ot nls hn1

IEEE Log Number 8612103. ns. An example of the rapid collapse of the collector volt-

U.S. Government work not protected by U.S. copyright.

BLACKBURN: TURN-OFF FAILURE OF POWER MOSFET'S 137

-OxideSource metal Silicon gate

6= ~~~ ~~Chane- Body

Parasitic npn Drain-body diodebipolar transistor -

HIGH RESISTIVITY DRAIN REGION n-

DRAIN CONTACT REGION (n+)Fig. 1. Cross section of power MOSFET identifying parasitic bipolar transistor and diode.

SB [14], [151. Both are electrical and not thermal in natureand either may cause a focusing of the current to the cen-

400V ter of the emitters due to the presence of reverse base cur-

rent. It has also been shown experimentally that reverse-bias second breakdown in bipolar transistors is electri-cally initiated and is not related to thermal instability or

QOov energy dissipation within the device 110], [11lIn summary the precipitous voltage drop synonymous

time(5ns/div) with second breakdown is a result of avalanche injectionFig. 2. Example of collector voltage waveform for bipolar transistor ex- and any mechanism electrical or thermal, that can ca

periencing second breakdown. sthe current density to become large enough for avalancheinjection to occur can cause second breakdown. This

age at second breakdown is shown in Fig. 2.' Note that should be true for both bipolar transistors and powerthe voltage falls from over 400 V to about 100 V in less MOSFET's.than 10 ns. The "plateau" at 10-100 V after secondbreakdown has occurred is also characteristic of the phe- Power MOSFET Turn-Offnomenon. A cross-sectional view of a power MOSFET device isThe physical mechanism that causes the precipitous shown in Fig. 1. Identified in the figure are the parasitic

voltage drop associated with second breakdown is ava- bipolar transistor and a parasitic drain-body diode, bothlanche injection [7], [81. For avalanche injection to occur, of which are important for switching and failure consid-an electric field large enough to cause avalanche multi- erations. Different regions of the device become active orplication of carriers must coexist with a current density of important depending upon the operating conditions im-sufficient magnitude to totally modulate the fixed charge posed upon the device. During the on-state (positive gatein the normal depletion region. For a forward-biased bi- voltage above threshold for an n-channel device), elec-polar transistor, it has been shown that the only "practi- trons flow from the source, through the channel formedcal" mechanism that can cause the current density to be- by the inversion region at the body surface, and into thecome great enough to allow avalanche injection (second drain. Once in the drain, they are directed vertically downbreakdown) to occur is thermal instability [13]. Because to the drain contact. Reverse leakage current from the par-the power MOSFET is a majority carrier device, thermal asitic drain-body diode may contribute a small currentinstability of the nature that occurs in the bipolar transis- component in the saturation region, but for the devicetor cannot occur in the MOSFET. Thus forward-bias sec- turned on in steady-state operation, the parasitic diode andond breakdown as experienced in bipolar transistors is not bipolar transistor usually are not important.expected to occur in the power MOSFET. It is primarily Examples of current and voltage waveforms for afor this reason that power MOSFET's were originally MOSFET during safe turn-off with an inductive load atthought to be immune to second breakdown. the drain are shown in Fig. 3. Fig. 3(a) is for the drainTwo mechansims have been identified by which the voltage externally clamped and Fig. 3(b) is for the drain

current density may become large enough for avalanche voltage unclamped. To turn off the MOSFET, the inver-injection to occur in the reverse-biased bipolar transistor sion layer (channel) at the body surface must be removed.

This is accomplished by removing or reversing the gate'For all photographs of oscilloscope traces, a division is defined as the voltage. As charge in the inversion layer begins to be re-

small division markings on the photograph. For example, in Fig. 2, thetime axis covers loo ns (20 divisions), and the voltage axis covers 600 V moved, the channel current, and therefore the drain cur-(12 divisions). rent, attempts to decrease. The inductive load causes the

138 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. PE-2, NO. 2. APRIL 1987

Voltage Clamps third is avalanche current. All can theoretically cause theparasitic bipolar transistor to become active (and thus pos-

8 400V sibly cause failure), but as will be shown, only the latteris of concern in present day practical devices.

ov

_ 22 EXPERIMENTAL RESULTSA variety of measurements has been made to study the

o _____________ _ OA modes and possible mechanisms of power MOSFET fail-Q ure during inductive switching. A nondestructive test sys-

time (200ns/div) tem was used to stress the devices near to and beyond the(a) normal limits of failure. This permitted repeated and ex-

tensive measurements to be made on devices for condi-j I Voltage Sustains tions which normally would cause catastrophic failure.

Several electrical parameters, the values of which mightbe sensitive to electrical or thermal stress at or near thefailure limits, were measured to monitor possible device

OV degradation due to this testing.

Nondestructive Test SystemGOA A detailed description of the system, used to protect the

time (200ns/di~)power MOSFET from degradation during what would

me(200bs/div) normally be destructive operation, has previously been(b) published in reference to bipolar transistor second break-

Fig. 3. Current and voltage wavcforms for power MOSFET during tun-pdown [121. The system was modified for this work onlyoff. (a) Clamped inductive load. (b) Unclamped inductive load.

by providing an appropriate gate drive voltage to powerthe MOSFET in place of the base current required to drive

drain voltage to rise in order to maintain the drain current a bipolar transistor. A schematic of the system as used forconstant. During the drain voltage rise, the drain current MOSFET's is shown in Fig. 5. In the nondestructive testbecomes divided between the channel current and a dis- system, the rapid collector/drain voltage collapse is ca-placement current caused by the formation of the deple- pacitively sensed and all power rapidly removed from thetion region at the drain-body diode. This is demonstrated device (in less than 40 ns after the voltage collapse). Thisin Fig. 4(a). The displacement current is proportional to has been shown to save bipolar transistors from degrada-the time rate of drain voltage rise, or dVDs/dt. The rate tion [10], [111.at which the drain voltage can rise is limited both by howquickly the gate can be discharged as well as by how Power MOSFET Failurequickly the drain-body depletion region can be charged. Examples of the drain current and voltage waveformsThe latter is determined by both the drain-source capaci- for a power MOSFET undergoing safe turn-off have al-tance and the magnitude of the drain current. The voltage ready been discussed (see Fig. 3). An example of the volt-can rise more quickly with a large drain current than with age waveform for a device undergoing what would nor-a small drain current if this term dominates the turn-off. mally be failure during turn-off (but saved from failure

If the drain voltage is not clamped by an external circuit here by the nondestructive test circuit) is shown in Fig.as the drain voltage continues to rise, the drain-body diode 6. The rapid collapse of drain voltage observable in thiswill eventually begin to generate current carriers via av- figure causes the current to rapidly increase without con-alanche multiplication, and the device will go into a sus- trol and will contribute to failure of the device if it is nottaining mode (waveforms of Fig. 3(b)). During sustain- removed quickly. Additionally, the current has likely fo-ing, all of the drain current is passing through the drain- cused to a very small area of the chip as usually occursbody diode (as avalanche current), and the channel cur- during negative resistance operation.rent is zero. The current distribution for this mode of op- The drain voltage waveform shown in Fig. 6 and theeration is shown in Fig. 4(b). The gate turn-off wave- collector voltage waveform for a bipolar transistorforms are identical to those of an avalanching diode in undergoing second breakdown, shown in Fig. 2, are re-which the current is controlled by the inductive load. markably similar. Both show the sudden voltage collapseAs has been described previously [1], [3], the parasistic to a plateau of about 100 V. If the protective circuit is

bipolar transistor may become active if the current in the deactivated for either case, the device will fail. In eachbody region beneath the source becomes large enough. instance the failure physically manifests itself as a smallThree sources of current in this region, all originating at (if the total energy is limited) area of melted or alloyedthe drain-body diode, have been identified above. One is silicon through the device chip. On the basis of theseleakage current, another is displacement current (dVDS/dt comparisons, this failure mode for power MOSFET's willcurrent) due to the drain-body diode capacitance, and the be called second breakdown in this work.

BLACKBURN: TURN-OFF FAILURE OF POWER MOSFET'S 139

[dV/dt hole flow In body] annel current

\ 710 - Mo v i n g~~~~~~~~~~~Moin/ / \ / \ ~~~~~~~~~~~~depletion

--0 -I- - c boundaries

[dV/dt electron flow in drain] -\

(a)

Avalanche current In body

ireeglinonboundaries

Avalanche current In drain

(b)Fig. 4. Cross section of power MOSFET showing possible current paths during turn-off. (a) During drain voltage rise.

(b) During sustaining.

SB

|CLAMP SPPLY] -, H

LOAD ~~~~~~~~~~~~~~~~~~~400V2ilH 1 5H|| 4 I"ipDUCTOR| _ 40

3n .S_o v

SHU NT l time (Sns/div)

_Lf 'j- UNDEVICE Fig. 6. Drain voltage waveform for power MOSFET undergoing failure.

| GATE DRwIVE

Fig. 5. Schematic of nondestructive test system. 800V, 8A

It is also instructive to compare the voltage waveformsfor the power MOSFET and the bipolar transistor for con-ditions for which each device safely sustains for some pe- OV,OAriod of time. Sustaining occurs with no external clamp,but the device "self clamps" at its sustaining voltage, and time (1la/div)the entire energy stored in the inductive load is dissipated Fig. 7. Collector current and voltage for bipolar transistor during un-

within the device. The current and voltage waveforms for clamped, sustaining operation.a power MOSFET in sustaining operation have been dis-cussed (see Fig. 3(b)), and those for a bipolar transistor currence of second breakdown sometime during sustain-are shown in Fig. 7. Two important characteristics of the ing. It has been found that the bipolar transistor will nearlybipolar transistor sustaining operation are the increase in always fail during sustaining, and as Fig. 7 demonstrates,collector voltage near the end of sustaining and the oc- the failure can occur very near the end of turn-off [101,

140 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. PE-2, NO. 2, APRIL 1987

[11]. It has been found in this work that, unlike the bi- dVDS/dt) flowing laterally in the body (parasitic bipolarpolar transistor, if the power MOSFET safely sustains for transistor base) beneath the MOSFET source (parasitic bi-at least 100 ns or so, it will safely turn off, even if the polar transistor emitter) can generate a sufficient potentialtotal sustaining time is 10 ,us or more. Also, the voltage drop to turn on the parasitic bipolar transistor and causewaveforms during sustaining are different for the power MOSFET's to fail [3], [4]. To determine if this occurs forMOSFET and the bipolar transistor. This can be seen by commercially available power MOSFET's, the second-comparing the voltage waveforms of Fig. 3(b) and Fig. breakdown characteristics have been measured for a va-7. Although the features of each waveform are not under- riety of devices for different rates of rise of drain voltagestood in detail, the differences occur because during sus- (dVDs/dt). The dVDs/dt is varied by changing the mag-taining, the bipolar transistor still operates as a transistor nitude of the available gate turn-off current. This is ac-but the power MOSFET operates as a diode. complished by varying the magnitude of the external gateWhereas it has been found that the bipolar transistor can resistance Rg while keeping the gate turnoff voltage con-

fail by second breakdown either before significant sus- stant.taining occurs or at any time during sustaining [10], [11], Examples of the drain voltage waveform at secondthe power MOSFET has been found to fail by second breakdown for the same drain current but for different val-breakdown only at voltages before any significant sustain- ues of gate turn-off drive are shown in Fig. 8. For theseing occurs. This implies that the current focusing mech- results, the gate turn-off voltage is approximately -20 Vanism of the bipolar transistor is becoming stronger dur- and the external gate resistances are 10 9, 100 Q, and 1ing sustaining, but that the mechanism which causes kQ. The maximum rate of drain voltage rise varies fromsecond breakdown in the power MOSFET is ineffective approximately 22 V/ns for Rg = 10 Q to 5 V/ns forRg =once safe sustaining occurs. 1 kQ. If the failures shown are initiated by the dVDs/dt

current, then the voltage at which second breakdown oc-Measurements for Device Degradation curs should decrease with increasing dVDs /dt. This volt-

There are several electrical parameters that may be af- age is actually highest, however, for the largest dVDS/dtfected by second breakdown of the power MOSFET. For in this example, opposite to what would be expected ifbipolar transistors that have been stressed beyond the lim- the failure resulted from a dVDs/dt effect. The results pre-its of second breakdown, it has been found that the col- sented here are representative of the devices studied inlector-base leakage current ICBO, measured at a voltage that the rate of rise of drain voltage has been found to havenear to the collector-base breakdown voltage ( - 95 per- an insignificant or no effect on the second-breakdowncent BVCBO), is a very sensitive indicator of degradation characteristics of the power MOSFET.(ICBO increases if the device is degraded). The drain- The results presented in Fig. 8 are for unclamped op-source leakage current may be such an indicator for power eration. Extreme care must be taken in analyzing the re-MOSFET's similarly stressed. Other power MOSFET pa- sults if an external clamp circuit is used. Because clamprameters that may be affected by such stress are the gate- circuits cannot react instantaneously, there is usually asource leakage current (expected to increase if the gate voltage overshoot above the set clamp voltage, particu-oxide is degraded) and threshold voltage (expected to shift larly for large values of dVDS/dt. Although a research-in value if the gate oxide is charged or interface states are quality clamp circuit has been developed for the nonde-created). structive test system used in this work [16], a voltageNone of the parameters discussed above has been found overshoot of nearly 100 V occurs for the 22 V/ns wave-

to change for power MOSFET's tested using the nonde- form of Fig. 8 when the clamp is used. An example ofstructive test system. Single devices have been tested over the overshoot is shown in Fig. 9 for the same device for100 times with no measurable change in those parameters. Rg = 10 Q but with the clamp voltage set to 300 V. ForThe base line parameter for these comparisons is the ini- the 5 V/ns waveform, almost no overshoot has been ob-tial parameter measurement for the unstressed device, and served. Thus, for clamped operation, if only the set clampnot, for example, the device manufacturer's specification voltage at which second breakdown occurs is measuredsheet value. The actual occurrence of second breakdown (that is, no oscilloscope trace is recorded), erroneous con-does not appear to degrade or destroy the device, but it is clusions can be drawn. For example, for clamped opera-the localized power density resulting from second break- tion, but otherwise the same operating conditions as indown that eventually causes failure. Because of the non- Fig. 8, the clamp voltage at which second breakdown isdestructive measurement capability, it is possible to de- observed to occur is measured to be 480 V for Rg = 10termine the effects of different circuit parameters and 9, 500 V for Rg = 100 9, and 550 V for Rg = 1 kQ. Thetemperature on the second-breakdown characteristics of a actual voltage appearing at the device drain at which sec-single power MOSFET so that possible variations be- ond breakdown occurs, for either clamped or unclampedtween devices do not mask the data trends. operation, is the peak voltage for each waveform shown

in Fig. 8 (approximately 570 V, 560 V, and 550 V for RgThe Effect of dVDs/dt on Failure = 10 S2, 100 91, and 1 kQ, respectively). It might be con-

It has been shown, using specially fabricated devices, cluded from the clamped operation results that the sec-that the capacitive displacement charge (resulting from ond-breakdown characteristics of the power MOSFET do

BLACKBURN: TURN-OFF FAILURE OF POWER MOSFET'S 141

R,= 10l loon lkfn 25°C 5O°C 75°C 100°C 125°C

500OV W500V

o~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~o

SO S ov~~~~~~otime (20ns/div)

time (lOns/div) Fig. 10. Examples of drain voltage waveform at second breakdown for

Fig. 8. Examples of drain voltage waveform for power MOSFET undergo- power MOSFET at different temperatures.ing second breakdown for single value of drain current but for differentvalues of dV[)s/dt. dV0s/dt is varied by varying tum-off gate drive.

1.0

> 0.8

C 0-300V~ ~~~.i..0.4-

i0.2

time (IOns/div) ~25 50 75 100 125

Fig. 9. Example of voltage overshoot for Ry = 10 n waveform of Fig. 8. TEMPERATURE(OC)Clamp is set to 300 V.

Fig. 11. Examples for several devices of effect of temperature on criticalcuffent below which second breakdown does not occur.

depend upon the magnitude of DVDs/dt, but in fact it isonly the speed of the clamp circuit that is being charac- waveforms or device behavior prior to sustaining, it is notterized. surprising that the magnitude of the inductance is not im-

portant for the second-breakdown behavior. The only ef-The Effect of Temperature on Failure fect of changing the size of the load inductor is to changeThe effect of device temperature on the second-break- the amount of energy dissipated within the power MOS-

down characteristics of power MOSFET's has also been FET during sustaining.measured. An example of the drain voltage waveforms atthe occurrence of second breakdown for the same drain DisCUSSIONcurrent but for different temperatures are shown in Fig. The results that have been presented are consistent with10. A composite of the oscilloscope traces are shown in the theory that activation of the parasitic bipolar transistorFig. 10 because temperature does not normally affect the causes power MOSFET's to fail during turn-off. Theystorage or voltage rise times of the power MOSFET. As demonstrate, though, that activation of the bipolar tran-the temperature is increased, the voltage at which second sistor is not due to dVDs/dt current but is due to avalanchebreakdown occurs is increased. This behavior is the same current at the parasitic drain-body diode.that occurs for bipolar transistors [10], [11]. For all of the devices studied, the voltage at which sec-As mentioned above, the power MOSFET may often ond breakdown occurs is approximately equal to the power

be turned off in the sustaining mode without second MOSFET drain-source breakdown voltage BVDS. In nobreakdown occurring. For devices for which this is true, instance was a device observed to enter second break-there is a critical current Icrit below which the device safety down at a voltage near to the sustaining voltage of theturns off in the unclamped mode (sustaining) and above parasitic bipolar transistor ( -0.5 BVDS). These factswhich second breakdown occurs prior to sustaining. The strongly suggest that the drain-body diode begins to ava-effect of temperature on Icrit for several devices is shown lanche before second breakdown occurs.in Fig. 11. It has been found that in general, as temper- The variation of the second-breakdown characteristicsature increases, 'crit decreases. with temperature supports the above conclusions. It is

well-known that the avalanche voltage increases with in-The Effect ofLoad Inductance on Failure creasing temperature, just as the voltage at which secondAs with bipolar transistors, it has been found that the breakdown in the power MOSFET occurs has been found

magnitude of the load inductance has no effect on the sec- to increase. It is expected that the parasitic bipolar tran-ond-breakdown behavior of power MOSFET's. sistor should be more susceptible to becoming active asAs previously stated, if the power MOSFET begins to the temperature is increased. This is because the trans-

safely sustain, it usually will turn off safely. Because the verse potential drop beneath the MOSFET source re-magnitude of the inductor does not have any effect on the quired to "turn-on" the bipolar emitter decreases as the

142 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. PE-2, NO. 2, APRIL 1987

temperature increases ( -0.45 V at 100°C compared to [3] C. Hu and M. H. Chi, "Second breakdown of vertical power MOS--0.6 V at 250C) and the resistance through which the FET's," IEEE Trans. Electron Devices, vol. ED-29, pp. 1287-1293,

Aug. 1982.current passes beneath the source increases with increas- [4] D. S. Kuo, C. Hu, and M. H. Chi, "dV/dt breakdown in powering temperature. These facts suggest that the critical cur- MOSFET's," IEEE Electron Device Letters, vol. ED-4, pp. 1-2, Jan.

rent I shoud decease ith tmperatre, a is oserve. [5]1983.rent Icrit should decrease with temperature, as is observed. [5] W. J. Slusark, Jr., R. J. Laurie, and G. L. Schnable, "CatastrophicAll of the power MOSFET second-breakdown voltage burn out in power VDMOS field-effect transistors," in 1983 Inter-

waveforms are similar to those shown in Fig. 6. The pre- national Reliability Physics Symposium, 21st Annual Proceedings Re-liability Physics, pp. 173-177, Apr. 1983.cipitous voltage collapse begins at a voltage near to BVDS [6] R. Severns, "Avoiding dV/dt turn-on in power MOSFET's,"

and stops at about 10-100 V. In no instance does the volt- Electronic Products Magazine, pp. 89-94, Jan. 5, 1985.age collapse more slowly or stop when it reaches the sus- [7] P. L. Hower and V. G. K. Reddi, "Avalanche injection and secondtaining voltage of the parasitic bipolar transistor ( -0.5 breakdown in transistors," IEEE Trans. Electron Devices, vol. ED-

taming voltage ~~~~~~~~~~~~~17,pp. 320-335, Apr. 1970.BVDS). This implies that the parasitic bipolar transistors [8] B. A. Beatty, S. Krishna, and M. S. Adler, "Second breakdown inof only one or a few MOSFET cells become active and power transistors due to avalance injection," IEEE Trans. Electroncarry all of the device current, creating a current density Devices, vol. ED-23, pp. 851-857, Aug. 1976.

carry~~~ ~ ~ ~ ~ ~ ~~~~~~ty []P. L. Hower, private communication.great enough for avalanche injection (second breakdown) [10] D. L. Blackburn and D. W. Berning, "An experimental study of re-to occur immediately in those few parasitic bipolar tran- verse bias second breakdown," in Technical Digest, International

*istrs. +Electron Devices Meeting, pp. 297-301, Dec. 1980.sistors. [11] D. L. Blackburn and D. W. Berning, "Some effects of base currentNo evidence has been presented that second breakdown on transistor switching and reverse-bias second breakdown," Tech-

in the power MOSFET occurs without activation of the nical Digest, International Electron Devices Meeting, pp. 671-675,parasitic bipolar transistor. No measurements were made, [12] D. W. Beming, "Semiconductor measurement technology: A reversethough, at current levels great enough for the MOSFET bias safe operating area transistor tester," NBS Special Publicationitself (without activation of the bipolar) to experience sec- 400-54, Apr. 1979.

ond brakdownSuch urrentlevelsare usally wll be 13] P. L. Hower, D. L. Blackburn, F. F. Oettinger, and S. Rubin, "Sta-ond 11konSuhcretlsbe- ble hot spots and second breakdown in power transistors," in Poweryond the manufacturer's recommended safe levels. Also, Electronics Specialists Conf. Rec., June 1976, pp. 234-236.no measurements have been discussed for which the par- [141 S. Krishna and P. L. Hower, "Second breakdown of transistors dur-asitic drain-body diode is forward biased prior to the drain [1]ing inductive turnoff," Proc. IEEE, vol. 61, pp. 393-394, Mar. 1973..[5]p L. Hower, "A model for turn-off in a bipolar transistor," in Tech-voltage rising at turn-off. Failures in this operating mode nical Digest, International Electron Devices Meeting, pp. 289-292,have been discussed elsewhere [5], [6]. Dec. 1980.

116] D. W. Berning, "Use of vacuum tubes in test instrumentation for

CONCLUSION measuring characteristics of fast high-voltage semiconductor de-vices," IEEE Trans. Instrum. Meas., vol. IM-30, pp. 226-227, Sept.In conclusion, it has been found that the failure of power 1981.

MOSFET's during inductive turn-off:a) appears electrically identical to second breakdown

in bipolar transistors,b) does not depend upon dVDs/dt,c) has an increasing voltage at which failure occurs

with increasing temperature, David L. Blackburn (M'71) received the B.S.d) has a decreasing critical current (below which fail- degree in physics from Ohio University, Athens,

ure does not occur) with increasing temperature, in 1965 and the M.S. degree in physics from Caseand Western Reserve University, Cleveland, OH, in

e) does not depend upon the magnitude of the load in- Since 1968 he has been involved in research on

ductance. semiconductor materials and devices at the Na-tional Bureau of Standards in Gaithersburg, MD.

These results are all consistent with the concept that ac- Recent power device research has included ther-tivation of the parasitic bipolar transistor initiates the fail- mal and switching characterization, radiation ef-

fects characterization, and safe operating limitsure and that the activation is caused by avalanche current studies of both bipolar and MOSFET transistors. He is presently Groupgenerated at the drain-body diode. Leader for Device Technology in the Semiconductor Electronics Division

at NBS.REFERENCES Mr. Blackburn has received the U.S. Department of Commerce Silver

Medal Award for contributions in reducing electronic failures and increas-[I] 1. Yoshida, T. Okabe, M. Katsueda, S. Ochi, and M. Nagata, "Ther- ing the reliability of electronic systems. He has recently served on the pro-

mal stability and secondary breakdown in planar power MOSFET's," gram committees for the International Electron Devices Meeting and theIEEE Trans. Electron Devices, vol. ED-27, pp. 395-398, Feb. 1980. Power Electronics Specialists Conference, and served as Chairman of the

[21 M. H. Chi and C. Hu, "Some issues of power MOSFETs,'" in Power Planning Committee for the 1985 IEEE/NBS Power Semiconductor De-Electronics Specialists Conf. Rec., pp. 392-399, June 1982. vices Workship. He is a member of the American Physical Society.