tantalumcapfailure

8/7/2019 TantalumCapFailure

1/9

Electrocomponent Science and Technology1976, Vol. 3, pp. 171-179 (C ) Gordon and Breach Science Publishers Ltd., 1976Printed in Great Britain

FAILURE MECHANISM OF SOLID TANTALUMCAPACITORSB. GOUDSWAARD and F. J. J. DRIESENS

N. V. Philips, Gloeilampenfabrieken, P. D. Elcoma/Development Dept. Eindhoven, Netherlands(Received June 3, 19 76 , in final form September 30, 19 76)

It has been suggested that failure of solid tantalum capacitors is due to thermal migration of impurities from thetantalum anode to flaws in the oxide layer. This implies, however, that leakage current gradually increases undernormal operating conditions, an effect which has not been observed. An alternative hypothesis advanced here is thatfailure is due to crystallization of tantalum oxide under the influence of the electric field. Scanning electronmicroscopy of specially cleaned anodized tantalum sheet on which thin gold electrodes have been deposited clearlyshows the occurrence of crystallization after 17 hours at an applied voltage of 75 V and a temperature of 65 C.Results of accelerated life tests on solid tantalum capacitors at temperatures of 85C and 125C, and at up to 2, 5times rated voltage also accord better with a field crystallization hypothesis than with a thermal migration failurehypothesis.

FAILURE ANALYSISSudden breakdown failures of solid tantalumcapacitors undergoing test are well known. Somebreakdowns are followed by healing but result inhighly increased leakage current; these are calleddegradation failures. Others result in a short-circuit;these are called catastrophic failures. No significantincrease in leakage current can be observed beforemost instances of either type of failure.

Predicting the life of solid tantalum capacitorsrequires an understanding of the failure mechanism.Much work has already been done on breakdowninvestigation and hypotheses have been advanced, butquantitative treatments to support them are lacking.By collating and analysing this work, we have come tothe conclusion that failure in solid tantalumcapacitors is due to the field crystallization process,and that the life of the capacitor is determined by thegrowth rate of the crystal. This paper sets out ourreasoning.

1. 1 Failure RateBy observing the number of failures in a given numberof components over a period of time, it is possible togain an overall picture of component reliability. Thefailure rate at any instant is expressed as the numberof failures at that instant divided by the product ofthe number of components and the elapsed time.Several such calculations produce the familiar "bathtub" shape life curve shown in Figure 1.

171

failurerate

FIGUREtub" curve.

timeComponent failure rate with time. Typical "bath

The early failure period of this curve is partly dueto manufacturing anomalies and partly to accidentaldamage during installation and early operation. Thefinal upturn of the curve is the wear-out failureperiod, in which deterioration processes in thecomponent cause rapidly increasing failure rate. Thecentral constant failure rate period is the useful lifeof the component, where failures occur at an approxi-mately uniform rate. Our work is concerned withmeasurements at the wear-out failure period.1.2 Accelerated TestingUnder normal stress conditions the useful life of solidtantalum capacitors is several years; thus the normal


2/9

172 B. GOUDSWAARD AND F. J. J. DRIESENSlife-test method is of little help in establishing thefailure mechanism. A better method is acceleratedtesting. At higher stress levels the failure rate increasesand the test results lend themselves to more objectiveanalysis.

Catastrophic failure rates show a dependence onapplied voltage, temperature and series impedance.Acceleration factors regarding applied voltage andtemperature for Kemet solid tantalum capacitors havebeen investigated by Didinger.: ,a He suggests thatthermal breakdown is initiated by the migration ofimpurities in the sintered tantalum anode to flaws inthe oxide layer, but a quantitative treatment of thisprocess is not given.

1.3 BreakdownBreakdown of a capacitor is characterized by a suddenincrease of the leakage current under even circum-stances, for instance during a slow increase of voltage.The oxide layer is ruptured when a spark or ionizationoccurs, after which a higher leakage current prevails.Breakdown investigations have been conducted byHoward and Smith4 an d by Burnham. Their signifi-cant findings are:

The mechanism of breakdown appears to bethermal.When healing occurs after a breakdown, an increasein leakage current is observed fo r all ratings.Maximum breakdown voltage is only slightlytemperature dependent.Maximum breakdown voltage varies between 50%and 80% of the forming voltage and is independentof the manufacturer of the capacitor.Maximum breakdown voltage decreases withincreasing time in step stress tests.The duration of breakdown is several milliseconds.There is no premonitory noise or significantincrease in leakage current prior to breakdown.Electron micrographs of breakdown spots, takenafter the removal of manganese dioxide, showradial cracks in the oxide layer.

1.4 Failure MechanismThe problem of the failure mechanism, however, isnot solved either by failure rate results or breakdownstudies. An understanding is therefore necessary ofthe formation process by which a conducting channelis produced in the oxide layer. In view of the facts,this process has to be a function of time, applied fieldstrength, and temperature.

Didingers suggestion a that impurities of the

anode, impelled by locally higher field strength, willbe concentrated at flaws in the oxide film implies acontinuously increasing leakage current while a locali-zed heating of the flaw initiates the thermal break-down. Experimental results do not confirm this idea. 6

Here we propose that the failure mechanism insolid tantalum capacitors is the field crystallization oftantalum oxide. A quantitative analysis of the pub-lished experimental data lends support to thissupposition.

2 FIELD CRYSTALLIZATIONField crystallization of anodic tantalum oxide filmsunder wet conditions has been extensively investi-gated by Vermilyea. 8, 9 Thermal crystallization ofamorphous anodic oxide can be achieved at hightemperatures. Crystallization can also be achievedbelow 100C if a strong electric field is applied acrossthe anodic oxide film; this produces circular areasconsisting of wedge-shaped polycrystalline segments.The material is an entirely new crystalline film, with-out consumption of the amorphous phase, pushingback the amorphous film during its growth (Figure 2).Capacitors in Figures 2 to 9 are anodized at 200 V.

FIGURE 2 Micrograph of the start of crystallization ofwet tantalum system (Vermilyea (8)).

Field crystallization occurs only at certain sites onthe metal-oxide interface favourable to the formationof crystal nuclei (see Section 3). These favourablesites may be areas of high impurity content in themetal. Vermilyea has analysed the crystal growth as afunction of field strength and temperature, and hascalculated its activation energy and the distancebetween the equilibrium positions of the ions.To confirm the suggestion that field crystallizationoccurs even in solid capacitors during life test, anattempt was made to detect these crystals.


3/9

SOLID TANTALUM CAPACITORS 173A chemically cleaned, sintered tantalum sheet wa s

anodized at 200 V and thin gold film electrodes ofabout 2 mm 2 were evaporated onto it. Theseelectrodes were only approximately 20 0 A thick andnot homogenous and therefore it is supposed, wouldnot impede the transport of oxygen or moisture fromthe ai r during the test period. The capacitors soformed were first examined with a scanning electronmicroscope while noting their x and y co-ordinatesThey were then stressed at 70 V and 65C for aperiod of 17 hours. Subsequent electron-microscopyrevealed that breakdowns had occurred in most of thecapacitors an d that the resulting damage madedetailed examination of the surface impossible. How-ever, on some of those remaining, white spots(possibly crystals of Ta20s) were discernible afterthe tests (Figures 3, 4 an d 5).

While scanning the surface of the capacitors out-side the initially scanned areas, some very clearcrystals were observed (Figures 6 and 7). A chanceobservation of one of these crystals showed that theamorphous film was already cracked, but without

FIGURE 3 Micrograph of oxide surface before life test.

FIGURE 4 Micrograph of oxide surface after life test:incipient crystallization.

FIGURE 5 Micrograph showing the start of crystallization.

FIGURE 6 Micrograph showing rystalline areas notscanned before life test.

FIGURE 7 Micrograph of the crystalline area A ofFigure 6.

breakdown because the gold film on this crystalbecame insulated (electrically) from the main part ofthe electrode. This micrograph (Figure 8) is similar inappearance to those by Vermilyea 8 in wet systems(typical of which is Figure 2) an d to those obtained in


4/9

174 B. GOUDSWAARD AND F. J. J. DRIESENS

FIGURE 8 Micrograph of the crystalline area B ofFigure 6. FIGURE 9 Micrograph showing the breakdown area ofsolid tantalum capacitor with MnO2 electrode (Burnham (5)).

TABLEAccelerated test matrixfor 10 F, 30 V capacitors

Temp. 85CTemp. 125C

Applied voltage/Working voltage1,i7" ! 1,33 1,67 1,83 2,024 24 14

2,1727

24 24 16 21 10 17 17Sample size

2,33 [. 2, 57 5

percentfailures

99.999,0-90,0?0,050,030.0

10,05,0

V= V/Vw Vw working voltage

V= 2, 5 2,33 2,0 2,17

life tested at 85C

1.010-3 10-2 I0 -1

FIGURE 10

1,331,17

/

10 102 103 10 z 105fQilure-age (hours)

Results of G. H. Didinger2 for 10 F, 30 V tantalum capacitors at 85C.


5/9

SOLID TANTALUM CAPACITORS 175solid tantalum capacitors with manganese dioxide ascathode (Figure 9), (see Burnham).

These experiments support the idea that fieldcrystallization occurs even in solid tantalum capaci-tors. It is now proposed that this process accounts forthe failures during life test and that, because somecritical size of crystal is needed to crack the film, therate of crystal growth is the factor that determinesthe failure time of the capacitor.

3 EXPERIMENTAL DATAExamination of experimental data supports this hypo-thesis. The data given here for mean failure times arefrom Kemet solid tantalum capacitor reliabilitystudies by G. H. Didinger.2 The accelerated tests weremade on 10/aF, 30 V capacitors, style CS14MIL-C26655/A. Tests were made at 85C and125C, at voltages ranging from 0,833 to 2,5 timesthe rated value. The accelerated test matrix is given inTable I. Any capacitor whose leakage resistancebecame equal to its capacitive reactance at 120 Hzwas considered to be a catastrophic failure.

The cumulative percentages of failures against timewere plotted on Weibull graph paper (Figures 10 and11). The slope and intercept parameters define theWeibull distribution function:

F(t) exp(-t/to)# (1)

where: F(t) is the cumulative percentage of failuresat time t

/3 is the shape factorto is the scale parameter.This function can be transformed into a linearequation"

F(t) exp(-t/to)a( 1In 1-F(t) --(t/t)In In F(t) 3(ln In to) (2)

The data are plotted on Weibu11 paper withln(1/(1 F(t)} as ordinate an d In t as abscissa.The shape factor or slope of the straight line

changes with the shape of the probability densityfunction fit), the first derivative of the cumulativedistribution function F(t). This variation is given inFigure 12.

Failure due to field crystallization usually takesplace at flaws in the oxide layer, where field strengthis greatest. Each capacitor has many flaws of differentsize and can be considered as consisting of an equalnumber of capacitive elements connected in parallel,each having a single flaw. The characteristic lifepattern of the capacitor as a whole is then equivalentto that of the weakest element, which is the one with

percentfailures

V=V/Vw Vw=working voltagelife tested at 125C

99,999,0 V=1,33 1,1790,0 V=2 1,8370,0"

o0,833

50,030,0-

10,05,0- tx

_2 102 103 10 10failure-age (hours)

FIGURE 11 Results of G. H. Didinger for 10 F, 30 V tantalum capacitors at 125C.


6/9


7/9

SOLID TANTALUM CAPACITORS 177l_.0pF- 30V

life 125C(s) 10a_ DidingerGoudswaard, Driesens107- 3_3NF-20V106. Burnham eA LotLot EI 85oClO5-

lO4-

!03102 electric field E (106V/cm]

5 6 ? 8 9 t0 11 12I0,8 1,0 1,2 1,Z 1,6 1,8 2, 0 Z, 2 2,4applied voltage/working voltage V/Vw

FIGURE 13 Mean-life time vs. applied voltage/working voltage for Kemet 10 #F, 30V capactiors at 85C, 125C; forKemet 3,3 #F, 20V chip pacitors, from Burnham.time is determined by the rate of growth of a crystalto a critical size at which the film cracks. This growthrate can now be calculated:

A1 =Aexp [ O-qaE]where: A is a constant

Q is the activation energy at zero fieldq is the charge of the iona is the distance from the equilibrium ion

position to the top of the potential barrierE is the electric field, determined by the

applied voltage and the thickness of theoxide layer.

A1 is the critical size of the crystalzt is the time of growth to the critical size.

Since zl is constant:0 qaEIn At In A + In AkT

This equation shows that the slope of In t against E isequal to qa/kT.The problem now is to transform the appliedvoltage into electric field strength in the mean-life

curve (Figure 13). Vermilyea9 has shown thatelectrical breakdown of anodic tantalum oxide filmstarts at a flaw in the oxide layer. These flaws occur

(3)

(4)

at places where the growth of the oxide film isimpeded by impurities, e.g. carbide or oxide particleswhich, as preAously mentioned, are the points atwhich field crystallization begins. Figure 14 showshow a typical flaw might appear in cross-section in a2400 A film (as given by Vermilyea). This drawingshows how the film thickness at the flaw (dflaw)about a quarter of the original thickness (dox). Inthis way, the field strength is determined by theoxide film thickness at the site of the flaw.The working voltage Vw of solid capacitors canvary between 1/3 and 1/5 of the forming voltage V.According to measurements made by Burnhams forMIL Style hermetically sealed capacitors, this ratio isapproximately 1/4.The field strength Ew for working voltage Vw at aflaw can now be written as:

Ew EF (5)SinceVv 4

anddo x

dflawEF 5.106 V/cm.


8/9

178 B. GOUDSWAARD AND F. J. J. DRIESENS= 3000,&

surface of film

dfl(wdox

5000,8,

T2AO0,FIGURE 14 Diagram of a cross-section of a flaw in a 2400 A Ta O film, after Vermilyea.9

Ths equation enables the ratio V/Vw in the curveof mean-life time to be transformed into electric fieldstrength. The transformation function was checked byusing results given by Burnham6 of recent acceleratedlife test measurements on Kemet solid tantalumcapacitors of 3,3 #F, 20 V. Figure 15 gives theWeibull plots for two different test lots, one lothaving a V/Vw ratio of 2,45 and the other lo t having

the ratio 2,35. Taking the 63% value for each lot fromFigure 15, using electric field strength values calculatedfrom Eq . (5) (with Vw/Vr 1/5, 4), an d trans-ferring the two points thus obtained to Figure 13 showsreasonable agreement with the 85C line.By measuring the maximum breakdown voltage ofa capacitor, and knowing the forming voltage andoxide thickness, the flaw thickness can be obtained

percentfailures99,999,090,070,050,030,0,

10,0-S,O

Burnham Lot ! V/Vw=2,h5Lot r V/Vw 2,35 life tested at 85C

1,0failure-age (hours)

FIGURE 15 Weibull plots of .3,3 #F, 20V tantalum chip capacitors of different lots, life-tested at 85C with V/Vw of2,35, as measured by Burnham.


9/9

SOLID TANTALUM CAPACITORS 179from Eq . (5). This, in conjunction with Figure 13 ,enables the mean-life time of solid tantalumcapacitors to be determined.

Using Eq . (4) to obtain values fo r a (thedistance from the equilibrium position to the top ofthe potential barrier), calculated from the slope of theline and with q 5e, gives:

for 85C: a 1,36 A,for 125 C: a 1,28 A

and, for E 7, 5 x 106 V/cm, the activation energy Qis 1,2 eV. From results of field crystallization growthrate of wet tantalum systems, Vermilyea a reportedvalues for a and Q of 1,6 A and 1,2 eV respectively;Gulbransen and Andrew o give an activation energyof 1,2 eV for oxidation of tantalum at temperaturesat which crystalline oxide is formed.

with Au counter-electrodes, there is evidence thatfield crystallization can start in commercial MnO2counter-electrode capacitors because of thefollowing-

1) Capacitors with gold or MnO2 cathodes areboth so-called "dry" capacitors and in both there ishumidity access from the air.

2) Photographs of crystals on anodized tantalumsmade by Vermilyea (wet conditions), Burnham(MnO: cathode) and (gold-electrode) are very similar.

3) Calculations of the activation energy of thefield-crystallization on tantalum, made by Vermilyea,are very much in line with the results obtained in thepresent paper for commercial capacitors.

4 CONCLUSIONSThe results of accelerated life test measurements,together with the electron micrographs, support thepropositions that:

Failure of solid tantalum capacitors is due to fieldcrystallization;the life time of the capacitor is determined by thecrystal growth rate;the time required fo r crystal growth to the criticalsize at which the film cracks is an exponentialfunction of field strength.There will be some influence on field crystallization

due to the anodization conditions, but any errorintroduced is small an d would be negligible in thesecalculations. Although the electron microscopyevidence presented is mainly in terms of structures

REFERENCES1. B. Kleindienst, lnternationale Elektronische Rundschau8,205 (1970).2. G. H. Didinger, Kemet Engineering Bulletin, (January1961).3. G. H. Didinger, Evaluation Engineering 6, (Sept/Oct.1964).4. L. F. Howard and A. W. H. Smith, Proc. ElectronicsComponents Conference 14, 187 (1964).5. J. Burnham, Proc. Electronics Components Conference21,314(1971).6. J. Burnham, Proc. Electronics Components Conference23, 111 (1973).7. D. A. Vermilyea, J. Electrochemical Society 102, 207(1955).8. D. A. Vermilyea, J. Electrochemical Society 104, 542(1957).9. D.A. Vermilyea, J. Electrochemical Society 110, 250(1963).

10. E. A. Gulbransen and K. F. Andrew, J. ElectrochemicalSociety 96, 364 (1949).

tantalumcapfailure

Documents