FAILURE MECHANISMS SOLID ELECTROLYTIC CAPACITORS

Electrocomponent Science and Technology1977, Vol. 3, pp. 233-246

(C) Gordon and Breach Science Publishers, Ltd., 1977Printed in Great Britain

FAILURE MECHANISMS IN SOLID ELECTROLYTICCAPACITORS

J. BRETTLE and N. F. JACKSON

The Allen Clark Research Centre, The Plessey Co. Ltd., Caswell, Towcester, Northants, England

(Reeeived January 19, 19 7 7)

The failure mechanism of solid tantalum and aluminium capacitors have been investigated using a combination ofelectrical measurements and electron microscopy. The capacitor dielectric was examined before and after life testingand changes correlated with electrical measurements.

The basic mechanism of failure of solid tantalum capacitors is found to be field crystallisation of the essentiallyamorphous dielectric oxide. The growth of higher conductivity crystalline oxide during operation of the capacitorscauses an increase in leakage current and may result in catastrophic failure. The effect of field crystallisation can beminimised by using high purity tantalum to reduce the number of crystallisation nucleation sites. Since crystallinegrowth is primarily dependant on applied voltage, high voltage capacitors are much more susceptible to failure thanlow voltage units.

There appears to be no long term failure mechanism ir solid aluminium capacitors. However, a particular problemwith these units is that they are difficult to make. This is because the anodic layer is chemically less stable in the caseof aluminium than in the case of tantalum. The attack is initiated during the deposition of manganese oxide bypyrolysis from manganese nitrate solution and developed by the reform process. Solid aluminium capacitors oftenhave a lower capacitance and higher initial leakage current than comparable solid tantalum units; however, theleakage current decreases on life tests and their reliability is high.

INTRODUCTION

In solid electrolytic capacitors the liquid electrolyteof conventional liquid filled types is replaced by asolid conductor of manganese oxide. 1’5 The higherconductivity of manganese oxide imparts significantimprovements in performance, especially under con-ditions of low temperature and high frequency.Tantalum capacitors constitute the majority of solidtypes, only a small proportion being made fromaluminium.

Although liquid filled tantalum capacitors have areputation for high reliability,2 it is general experiencethat solid tantalum units have a tendency to undergoa sudden and catastrophic failure. When the capacitoris used in a low impedance circuit, failure rate may beparticularly high. Earlier studies in these laboratoriesof field crystallisation of anodised tantalum inaqueous electrolytes3 prompted a belief that the sameprocess might be the principal cause of failure in solidcapacitors, although it has been shown that at leastfor aqueous electrolytes, other failure mechanisms arepossible.4

The significant features of field crystallisationrelevant to this reasoning are:

a) The crystalline form of the oxide is more con-ductive than the amorphous form.

233

b) The crystalline form grows within the amor-phous film and may rupture the dielectric.

c) In contact with a liquid electrolyte sufficientcrystalline material may eventually grow by anodisa-tion to limit the leakage current;no such mechanismmay be present with a solid electrolyte.

The present work attempts to determine whetherfield crystallisation is responsible for the observedfailures in solid tantalum capacitors and themechanism by which such failures occur.

Failure processes in solid aluminium capacitors, thedielectric of which is known not to undergo fieldcrystallisation, have also been examined. The workhas principally involved a study of the changes in thephysical, chemical and electrical properties of thedielectric of experimental capacitors during storage atelevated temperature with voltage applied.

2 EXPERIMENTAL CAPACITOR CONSTRUCTION

In order to facilitate the examination of the dielectriclayers, wound foil electrodes were used. Since fieldcrystallisation is known to be promoted by impuritiesin the foil3 most of the work was performed withrelatively low purity anode foil. In the case of

234 J. BRETTLE AND N. F. JACKSON

tantalum, degreased capacitor grade foil of thickness0.0013 cm was used, previous work having establishedthat foils cleaned in this way had a surface impuritylevel of approximately 107 particles cm -2 .aAluminium anodes were prepared from degreased99.8% foil of thickness 0.005 cm. A brief descriptionof the constructional features of the capacitors isgiven in this section; specific details of processing aregiven in paragraph 3.1 and 4.1.

Tantalum and aluminium foils were cut to a widthof cm and a length of 7 cm. Riser wires of the samematerial as the anode were attached at one end. Aftercleaning, the foils were anodised, washed and dried,and then wound tightly with an etched aluminiumcathode and a fibreglass spacer, using the riser wire ofthe aluminium cathode as a mandrel. The woundstructure was secured with chloride free adhesive tapeand then reanodised to heal damage caused to thedielectric during assembly.

The unit was coated with manganese oxide usingstandard impregnation procedures with manganesenitrate followed by high temperature pyrolysis.Anodisation (reforming) in a liquid electrolyte wasapplied at intervals during coating. This process isusually known as reforming, since the nature of theelectrochemical reaction is not fully understood andit is useful to distinguish it from normal anodisation.No encapsulation was employed. In most casescapacitors were tested using a low impedance lead-acid battery as a voltage source.

3 TANTALUM CAPACITORS

Capacitors were constructed using anodes prepared inthree different ways"

1) anodised to 150 V,2) anodised to 24 V,3) anodised to 150 V on specially purified

tantalum.

Foil for 1) and 2) was degreased before anodising byimmersion in warm propan-2-ol. For 3), the standardfoil was vacuum annealed at 1950C for 30 minutesat a pressure of 10 -4 torr.

3.1 Preparation

The anode foil was formed to the appropriate voltagein 1% phosphoric acid at 90C using a current densityof 0.1 mA cm-2 followed by a period of 60 minutesat constant voltage. Reanodisation after winding waseffected under the same conditions, but the period at

constant voltage was limited to 15 minutes. The coilswere impregnated in manganese nitrage solution(s.g.-- 1.73 g cm-a) at 50C and pyrolysed at 300Cfor 10 minutes. After three coatings had been applied,the units were reformed in 0.01% phosphoric acid at85C. The reform voltage, approximately 40% of theoriginaI voltage, was attained by applying a current of0.1 mA cm-: and then holding at constant voltage for30 minutes. After final washing the capacitors weredried for hour at 125C. Dry capacitance valueswere approximately 80% of the initial wet values.Typical unit characteristics at room temperature wereas follows: High voltage capacitors (150 V), capaci-tance 0.8/.tF, tan i 2.5% (50 Hz), leakage current<1 A (10 V); Low voltage capacitors (24 V), capaci-tance <3.3/aF, tan 6% (50 Hz), leakage current15 uA (6 V).

The thickness of oxide would have been approxi-mately 240 nm and 38 nm respectively.

3.2 Life Test Results

Foils anodised under the conditions described in thepreceding section were made up into capacitors andtested at 125 C. Units anodised at 150 V were initiallytested at 40 V; because of rapid failures, this wasreduced to 30 V. Units anodised at 24 V were testedat 4.8 V, i.e. at the same field strength (--1.25 x10a V/m)as the higher voltage units at the reducedlevel. Figure shows the form of the leakage current/time plot for the three types of capacitor. It can beseen that low voltage capacitors and capacitors usingpurified foil have low and constant leakage currents,whereas high voltage capacitors exhibited a sharplyrising leakage current which peaked early in the lifetest and slowly declines thereafter. In addition to thenormal high leakage currents, current surges oftenoccurred in this type of capacitor which could blowO0 mA safety fuses placed in series with each

capacitor. The capacitor often recovered on replace-ment of the fuse, but the latter was sometimes blownagain at a later date.

3.3 Structure of the Anodic Oxides

3.3.1 High voltage capacitor (degreased foil)a) Before Life Test Figure 2(a)shows the generalstructure of the oxide. Defects in the film lie in linesalong the rolling direction of the foil suggesting thatthey are initiated from damage induced by rolling orimpurities introduced by the rolls. The defects are ofthree main types: bumps, cracked bumps and pits.

Figure 2(b) shows a cracked bump in more detail.

ELECTROLYTIC CAPACITORS 235

Leakage CurrentA

5oo

7oo

6oo

5oo

4oo

300

2oo

lO0

1200 2bOO 3600

High Voltage (anodised at 150V; 30V applied)

Degreased Foil

High Voltage (anodised at 150V; 30V applied)

Purified Foil

Low Voltage (anodised at 2hV; h.8V appli)Degreaed Foil

800 6000 Time, Hrs.

FIGURE Leakage current vs time for solid tantalum capacitors (125 C)

The structure is 1.5/am across and consists of apyramidal growth of crystalline oxide cracking openthe amorphous oxide to produce a characteristicthree branched crack.

The uncracked bumps are somewhat smaller indiameter, generally from #m downwards without avisible crystalline body. Low angle examination ofthese features shows that they are approximatelyhemispherical in shape. Using transmission electromicroscopy the cracked sites appear as dark triangularfeatures and the larger number of uncracked sites as

small circular dark areas with a wide distribution ofsizes. Rather indefinite pits lying in the rolling direc-tion can also be seen. When decreasing the microscopeaccelerating voltage the number of visible pits slowlydiminished indicating that they have varying depths.At higher magnification (Figure 3(a)) the trefoil shapeof the cracked sites can be clearly seen and small porescan be observed within uncracked bumps. These smallpores (<0.1 #m diameter) show much more clearly ifthe tantalum oxide is thinned slightly by argon ionbombardment or a few seconds etching in HF; under

FIGURE 2 Structure of 150 V film prepared on degreased tantalum foil: (a) general structure, (b) cracked bump.


FIGURE 3 Structure of 150 V film prepared on degreased tantalum foil:(a) transmission micrograph(b) after argon beam thinning, (the white line across the micrograph is an artefact produced by the film rupturing).


these conditions every bump can be shown to have acentral pore. Tilting of the specimen shows that thesepores are not spherical but are in fact elongated in thedirection of the film thickness. Examination of theunder surface of a stripped film shows that bumps areagain present lying along the rolling direction; thesebumps often have dimples in their surface.

To examine the structures below bumps andcracked bumps argon beam thinning of the strippedoxide film was attempted. A thinned bump is shownin Figure 3(b). The centre of the bump is still darkdue to thickness contrast but a fringe of crystallineoxide showing diffraction contrast can be seen.

Selected area electron diffraction shows the crystallineoxide to be 13-Ta O

From the above observations it can be seen that avariety of structures can be observed in the essentiallyamorphous oxide produced on nominally puretantalum. From this and earlier worka it appears thatbumps and pits can form at impurity sites in theunderlying metal, a number of the bumps becomeparticularly large, crack and reveal a crystallinenucleus below. During prolonged anodisation it hasbeen showna that this nucleus can grow and displacethe amorphous oxide film with a crystalline film.

FIGURE 4 Structure of 150 V film prepared on degreased tantalum foil: (a) before life test, (b) after life test(1000 hours), (c) large breakdown site, (d) after extended life test (9000 hours).


b) After Life Test Standard, high voltage capacitorswhich had been on life test for several thousand hoursand which were therefore showing large leakagecurrents were removed from life test, the manganesedioxide removed with HNO3/H.O2 reagent and theanodic oxide re-examined using the microscopicalmethods described earlier. On stripping film for trans-mission electron microscopy it was immediatelyobvious that the film was very much more fragile thanthe untested film. The reason for this was apparent onexamination: many large cracked bump sites werepresent after testing, the centre of the defect oftenbeing lost, so leaving a hole in the film. All stages ofdefect growth could be seen, in contrast to untestedfilm where either uncracked bumps or bumps in alate stage of crack development were more usual.

To obtain more quantitative information on themorphological changes occurring during life testingrandom areas of life tested and untested foil wereexamined by scanning electron microscopy and thenumbers of cracked and uncracked bumps compared.Figures 4(a) (untested) and 4(b) (lifetested) are acomparison of typical areas. It was found that thenumber of large cracked bump sites approximately

doubled on life testing. A much larger change, 1-2order of magnitude, was seen in the number ofuncracked bumps. Generally these bumps werelarger, more distinct and more difficult to distinguishfrom the cracked bumps in the lifetested case than inthe untested case; a less strict definition of what wastaken as a cracked site would have produced a muchlarger ratio than the doubling reported above. Inaddition to the defect sites much larger failure sites(Figure 4(c)) were observed. The probable sequenceof events is that bumps grow in size during life testand finally crack open; a number of these crackedsites than suffer electrical breakdown resulting in thestructure shown.

Capacitors which had been on test for much longerperiods were also examined. These had suffered a slowgradual decline in leakage current from the early highvalue. Figure 4(d) shows a typical area of oxide filmin such a unit; it appears that the cracked bump siteshad become damaged. All stages of damage could beseen, from the lip of the site being partly brokenaway, through the "cap" of the site being removedexposing the underlying crystal to the crystal beingremoved. It appears possible, therefore, that the

FIGURE 5 Structure of 24 V film prepared on degreased tantalum foil.


decrease in leakage current later in the life test maybe due to conducting crystallisation sites mechanicallydetaching from either the manganese oxide or thetantalum, thus lowering the current.

3.3.2 Low voltage capacitors In contrast to thehigh voltage oxide, no differences in structure wereapparent between films examined before and afterlife test.

Scanning electron microscopy showed that noneof the pit, bump or cracked bump features previouslyobserved could now be seen. However because thefilm had consumed such a thin layer of tantalumduring anodisation (’-’20 nm)6 the original surface ofthe tantalum was effectively replicated resulting in arough looking oxide. The close replication of thesurface allowed the extraction of surface impuritiesby the oxide (Figure 5). The dark areas are tantalummetal retained in the folds of the rough oxide, thebright particles being crystalline impurities shown upby the dark field conditions employed in the electronmicroscope. It has proved difficult to analyse theseparticles sufficiently well by selected area diffraction

to identify them unequivocally; however it can beshown that they are not Ta2Os.

Examination of low voltage anodic oxide hastherefore shown that the structures present in highervoltage oxides are absent but that impurities from theunderlying tantalum are incorporated in the growingoxide.

3.3.3 Oxide formed on specially purified tantalumAs with the low voltage oxide, life testing had noeffect upon the structure. On examination the oxidewas found to be smooth and largely defect free. Anoccasional line of impurities could be discernedfollowing a prior grain boundary in the underlyingmetal and crystallisation sites were present at a fewof these impurities (Figure 6). It can be seen thereforethat a purification treatment greatly inhibits theoccurrence of defects in the anodic oxide by reducingthe number of initiating impurities. Although thestarting foil was of 99.99% purity spark mass spectro-metry has shown that most impurity elements haveconcentrations 1-2 orders of magnitude higher at the

FIGURE 6 Structure of 150 V film prepared on purified tantalum foil.


surface than the bulk. Therefore nominally 99.99%pure material may have only a 99% pure surface layer.

3.4 Environmental Effects on Solid TantalumCapacitors

Early in the life test it became apparent that theleakage current of the capacitors was dependant onenvironment; large initial changes occurred on "dryingout" of the capacitors in the first few hours on testand admitting fresh moist air by opening the ovendoor later in the test perturbed the leakage currentvalues. It appeared possible that the environment,particularly water vapour, could affect the degrada-tion mechanism. Therefore, the dependance ofcapacitor characteristics on environment was investi-gated more systematically.

An all glass vacuum system capable of vacuumdrying capacitors at 125C and <10-3 torr was used.Capacitors were held in capsules, dried underdynamic vacuum and backfilled as required withgaseous environments. The capsules could be sealedon the vacuum line, removed, and the capacitors lifetested in the sealed capsules. Testing voltages wereapplied via contacts sealed through the capsule walls.The environments examined were 13 torr of air freewater vapour, dry air at 760 torr, dry oxygen at760 torr and perfluoro-N-Hexane at 234 torr. Thislast gas is known to inhibit electrical dischargesbecause of its high breakdown strength and couldpossible inhibit the formation of large electricalbreakdown sites if breakdown through air wasinvolved in their formation. The dry gases had aresidual water content estimated at <5 x 10-s torr.

Of the environments examined only water vapourwas found to have any effect on the leakage current/time plots of the capacitors on life test. The effectof water vapour was examined more closely by lifetesting capacitors in capsules attached to the vacuumline so that known pressures of water vapour could beintroduced to, or removed from, the capacitor at willduring the test.

It appeared from the observations that two inde-pendent processes were occurring. Firstly there wasthe basic degradation mechanism which caused theleakage current to rise steadily and irreversibly withtime, and which was unaffected by water vapour.

Secondly, superimposed on this steady increase incurrent, short term reversible decreases could be madeto occur by introducing water vapour. This effect wasin contrast to the effect of liquid water where leakagecurrent rises sharply because water makes electricalcontact with damaged areas of oxide previously

uncontacted by manganese dioxide. It appears thatwater vapour at 125C probably enters the manganeseoxide structure and increases its resistivity, thuslowering leakage current.

3.5 Circuit Impedance Effects on SolidTantalum Capacitors

Commercial experience suggests that solid tantalumcapacitors are more stable in service if they have animpedance in series with them than if they havevoltages applied from a low impedance source. Tocheck this and investigate the mechanism of thecircuit impedance effect, life tests were performedwith an added circuit impedance.A four terminal power supply was adapted with ak2 series resistor to provide up to 20 mA without

voltage drop to the capacitor terminals; from 20 mAto 50 mA the capacitor terminal voltage fell linearlyfrom 30 to 0 V. In this way variations in appliedvoltage due to variations in steady state leakagecurrent could be avoided while maintaining theability of the circuit impedance to inhibit largecurrent surges. A series of standard capacitors withvoltages applied from this power supply were com-pared with a series with voltages applied from lowimpedance lead-acid batteries. The power supplyvoltage was adjusted to exactly the battery voltage(nominally 30 V)and the tests performed simul-taneously. The addition of a circuit impedanceresulted in a much lower and more stable leakagecurrent which increased gradually from about 20to 45/.tA over a period of 1300 hours. The mor-phology of the oxide surface before and after lifetest is shown in Figure 7. With a high circuit impe-dance, bumps in the film, although increasing innumber during life test as in the low circuit impe-dance case, do not generally crack, in contrast to theobservations reported in Section 3.3.1. A furtherimportant feature was the initial absence of largeelectrical discharge sites, only one of these beingfound.

It appears that an added circuit impedance notonly prevents the development of large electricaldischarge sites but also the growth of uncrackedbumps into somewhat larger cracked bumps. Thesmoothness of the leakage current/time plot in thelow impedance case suggests a gradual increase in thenumber of "leaky" cracked bump sites rather thanthe sudden appearance of the occasional large elec-trical discharge site. To investigate the mechanism ofthe circuit impedance effect more closely some orderofmagnitude calculations were performed to see


FIGURE 7 Structure of 150 V film prepared on degreased tantalum foil after life testing: (a) high impedance circuit(b) low impedance circuit

what effect current surges could have on the bumpand cracked bump sites.A model for a defect site may be developed by

considering it as a disc shaped area of low resistancesurrounded by an area of high resistance. The rate ofheat input from the power supply minus the rate ofheat loss of the disc by conduction to its surroundingsgives the net rate of change of heat content of thedisc in the form of a differential equation. Approxi-mating the situation by assuming heat is only con-ducted from the disc through the metal tantalum, asolution can be obtained:

and

where

kA IV kAT/d

IVd

t time required to reach temperatureT.

d distance from disc to unheatedtantalum substrate.

rn mass of discc specific heat of disck thermal conductivity of tantalumA peripheral area of discI power supply currentV power supply voltage

Tmax maximum temperature attainable.

Substituting suitable values]" of these variables forsmall crystallisation sites and large damage sitesallows the response of these sites to current surgesfrom the power supply to be evaluated.

Because of their very small size, surge currentsthrough crystallisation sites lead to extremely hightemperatures in extremely short times e.g.104 C in 3 x 10-s sec. More realistic results areobtained in applying the calculation to large damagesites e.g. 104 C is reached in 3 x 10-s sec. suggestingthat a circuit impedance which would limit the currentpassed to much less than the value of 100 mA used inthe calculation may act by preventing an excessivetemperature rise at such a defect site.

]’Values used in the calculations were as follows:-

(i) Small Crystal Sites I 10 -s ampsc 0.4 joules/gmA 7.85 x 10 -3 mrn 1.93 x 10 -1 gmd=10-6 mK 5 x 10 -3 watts m-/C-(ii) Large Damaged Sites As for above, except:I-- 10 -1 ampsA 7.85 x 10 -9 mrn 1.93 x 10 -s gm


4. ALUMINIUM CAPACITORS4.1. Preparation

Plain anode foils were formed to 80 V in a borax-phosphoric acid electrolyte at 95C, using a constantcurrent density of--’6 mA cm-2 The anodes wereheld at the forming voltage for 30 minutes. Reanodi-sation after winding, was accomplished at 80 V for30 minutes in a saturated aqueous solution ofammonium pentaborate (A.P.B.) at room temperature,using the internal cathode of the coil as the counterelectrode. Impregnation was performed in manganesenitrate solution (s.g. 1.61 cm-3) at 50 C, followedby pyrolysis at 350C for 15 minutes. Reformationwas then carried out using the identical conditions tothose for reanodisation. Current densities used toattain the reform voltage were chosen to be as low aspossible, while still maintaining a reasonable rate ofincrease of voltage with time up to 80 V. Afterwashing and drying, two further pyrolyses andreformations were given; the level of the final currentafter reform increased with the number of coatingsapplied. Three coatings were found to be the maxi-mum possible, since additional applications resultedin little observed reform action and very high leakagecurrents in the finished capacitors. In general, unitsafter three coatings had dry capacitance values of70-80% of the initial wet values. Typical unitcharacteristics at room temperature were:capacitance 0.5/aF, tan 8 4% (50 Hz), leakagecurrent 1/aA to 40/aA at 10 V. The film thicknesswas approximately 11 nm.

4.2. Electrical Measurements

Capacitors were life tested at 125C with 30 V d.c.applied. Initial leakage currents were similar to thetantalum units but subsequent behaviour wasmarkedly different. In general the leakage currentdecreased with time up to "--1000 hours and thenremained essentially constant for test periods up to9000 hours. The initial decrease in current was foundnot be a "drying-out" effect, dependant upon tempera-tures alone. This was established by storing capacitorsat 125C for 2,500 hours prior to applying the voltage.As shown in Figure 8, the leakage currents decreasedin the same way as the control capacitors. It was alsoobserved that if the applied voltage was interruptedfor a short period of time when the leakage-currenthad fallen to the steady value, a significantly lowercurrent was apparent on reconnection. However, agradual recovery to the steady value occurred over atime period of several tens of hours.

56o

,8o

.00

320

2kO

160

Time (Hours)

FIGURE 8 Leakage current vs time for solid aluminiumcapacitors on test at 125C with 30 V applied.Nos. and 2, testing voltage applied immediatelyspecimens had reached temperature. Nos. 3 and 4 testingvoltage applied after 2400 hours at temperature.

4.3. Structure ofAnodic Oxide

Examination of the anodised foil revealed thepresence of nodules of non-coherent oxide distributedrandomly over the surface (Figure 9(a) and (b).) Thenodules were in the region of 1/am to 3/am indiameter and had a population density of-’-3 x 106 cm-2 Electron probe microanalysisindicated a high concentration of iron at the nodularsites. After processing and immediately prior to lifetesting, the surface showed signs of severe corrosion,a particularly bad area containing massive damagebeing shown in Figure 9(c). After life testing, manyof the less damaged sites had become plugged withnew oxide growth (Figure 9(d)).

Studies of the changes occurring during processingof the capacitor indicated that little attack wasapparent until after the first pyrolysis/reform. Itappeared that pyrolysis initiated damage to the oxidelayer and these sites then underwent dissolutionduring the reform process. The nature of the damageto the oxide caused by the pyrolysis could not beestablished. For example, X-ray diffraction failed todetect the presence of any hydrated oxide, e.g.Boehmite, and there was no apparent correlationwith surface impurities. However, examination afterpyrolysis revealed a corrosion pattern characteristicof the warp and weft structure of the fibreglass spacer


FIGURE 9 Structure of aluminium oxide"(a) & (b) after anodisation, (c) after final reform showing a severely corroded area, (d) after life testing.

and indicated that damage to the anodic oxide waslargely confined to those areas where the spacer was inclose contact with the surface. This was probablybecause the decomposition products of pyrolysis,largely oxides of nitrogen and water, could not readilyescape from these regions. It was observed that the’dry’ leakage current was always much lower than the’wet’, suggesting that much of the reform effect wasdue to electrical isolation of the damaged sitesrather than a reanodisation of the base aluminium.

4. 4. Evaluation ofPyrolysis DamageThe dissolution rate of anodised aluminium in

bromine-methanol may be used for indirectlymeasuring the damage caused by pyrolysis. Thebromine-methanol reagent has been used extensivelyin this work to strip oxides from the underlyingmetal. The basis of this technique depends upon theselective solubility of the metal in the reagent. Thedissolution of the metal proceeds via defects in theoxide which allow access to the reagent. Thus ameasure of the extent of damage after pyrolysis canbe readily obtained by determining the weight afterimmersion under standard conditions. Results forspecimens pyrolysed three times without reformationare given in Table I. Pyrolysed specimens were foundto show a greater weight loss than untreated specimens.


TABLEWeight loss of anodised aluminium specimens in bromine-methanol reagent

Specimen

PyrolysedUntreated

PyrolysedUntreated

PyrolysedUntreated

Experimentno.

Dissolutiontime(hr)

1.751.75

3.03.0

wt. AIlost(%)

8.94.7

36.33.9

50.411.2

Ratioof wt.loss

1.9

9.3

4.5

The spread in the results is not unexpected, sincethere will be an inherent variability in the sizes andthe numbers of defects. In addition, problems arisedue to differences in the activity of the reagentcaused by small variations in the amount of water.

5 DISCUSSION

This work has provided evidence that the major causeof failure in solid tantalum capacitors is the growth ofcrystalline oxide in the amorphous oxide dielectric atlocalised sites during capacitor operation, so creatinglow resistance paths through the dielectric. Highcurrents through these sites can lead to catastrophicthermal breakdown of the capacitor. The crystallisa-tion sites lie in longitudinal lines suggesting that theyare initiated by mechanical damage or chemicalimpurities introduced by the foil rolling process. Ithas been shown that impurities exist at the surface invery much higher concentrations than the bulk but itis difficult to determine the composition of theimpurities unequivocally because of their very smallsize (<0.1/m). The probable dependance ofcrystallisation site initiation on impurities ormechanical damage immediately suggests thatimprovements in capacitor reliability may be made byusing vacuum heat treated material. This was foundto be the case; vacuum purification of tantalum foilat 1950C for 30 minutes gave anodic films whichwere largely crystallisation site free and which werestable in capacitors on life test.

Field crystallisation may also be controlled bylimiting the anodising voltage. Low voltagecapacitors utilising low voltage film do not appear tobe subject to failure by field crystallisation. It isknown from earlier work that the incidence of fieldcrystallisation decreases as the anodising voltage is

reduced. There does not appear to be a simplerelationship between voltage and degree ofcrystallisation; neither is it possible to define veryeasily a voltage threshold below which crystallisationwill not occur. In the present work however nocrystallisation sites could be seen in film formed at24 V and low voltage capacitors behaved well on lifetest. It can be seen from this work that the importanteffect which causes increasing leakage current andfailure of solid tantalum capacitors is the presence ofcrystalline oxide. Relatively more defective, thinneroxide, is much more stable as it does not containdeveloping crystallisation sites. The best behaviourhowever is observed with purified foil, where theoxide is largely defect free and smooth.

On life test, capacitors constructed using standardconditions showed a sharply rising leakage currentwhich levelled out after "--1000 hours and fell slowlythereafter. Current surges occurred which producedthermal damage sites --0.1 mm diameter in thedielectric, presumably at particularly conductivecrystallisation sites. The fact that not all of thesethermal damage sites led to fuse failure and that thecapacitor was stable after a discharge suggests thatthe manganese oxide over such a site is eitherdispersed by the discharge or converted to a lowerconductivity oxide. It may be that when this isolationdoes not take place, current sufficiently large toproduce fuse failure can occur.

Examination of oxide before and after life test ofstandard capacitors showed that the number of defectbumps in the oxide increased considerably and thatmany uncracked bumps became cracked. Nosignificant changes were seen with low voltage orpurified foil. Transmission microscopy showed thatcrystals under cracked bumps grew to several times thediameter of the cracked site, opening the crack andallowing the crystalline body to be seen. From thecalculation it can be seen that if crystalline sitesbecome sufficiently conductive, high temperaturescan be obtained allowing growth by purely thermalmeans. This may be important in the later stages ofdevelopment of a cracked site but it is likely that, inthe earlier stages of uncracked bump development,growth occurs by ion transport in the electric fieldleading to a continued slow anodisation. It isbelieved that the slow decline in leakage current laterin the life test is due to a loss of adhesion betweenthe high conductivity points in this dielectric and theoverlying manganese oxide.

Environmental effects on solid tantalum capacitorswere only found to be important in the case ofwater vapour. The effects were reversible however


and were thought to occur through changes in themanganese oxide. As they were not relevant to theirreversible capacitor degradation mechanism theywere not investigated in detail.

It was observed that the addition of a circuitimpedance of the order of k2 prevented thedevelopment of large thermal runaway sites. A simplecalculation of the temperature developed at suchsites during electrical discharge indicated that theinhibition of current surges by the added circuitimpedance was a reasonable explanation for theseobservations. It was also observed however that ahigh circuit impedance inhibited the development ofbumps into cracked bumps, while leaving the increasein number of uncracked bump sites unaffected. Thisis consistent with the high impedance life testsresults where leakage currents stayed relatively low;increases in leakage current are thought to correlatewith the development of cracked bumps. Themechanism by which circuit impedance affects bumpdevelopment is still unclear however as the currentspassed at such sites are too small to be significantlyaffected by the circuit impedance.A major part of the work on solid aluminium

capacitors was concerned with developing andexamining the capacitor construction process.Control of the reform stages was the most difficult.Aluminium oxide was found to be much moresusceptible to damage than tantalum pentoxide andthis made reform more difficult and limited thenumber of reforms to three. It was often observedthat a capacitor which would not initially reformsatisfactorily could be "retrieved" by prolongedsoaking in ammonium pentaborate solutionfollowed by a second attempt at reform. Theseobservations suggested that diffusion of productsfrom the reform site was an important factor inachieving a successful reform. The current passed isnormally much larger than that required to reformdamage sites in the oxide; most of the current isconsumed in dissolving aluminium to form ions. Thelow pH environment produced by hydrolysis of theseions persists where washing is inadequate or in crevicesituations, i.e. at the contact of the spacer tape andoxide surface. It was found that prolonged washingcould be used to mitigate this effect without fear ofhydrating the anodic oxide.

Attempts were made to minimise pyrolysisdamage while increasing the manganese oxide fillfactor by using progressively lower reform voltagesand reforming after multiple pyrolyses. Thesevariations on the standard process had limitedsuccess however.

Examination of capacitors during the construc-tion process showed that the occurrence of damagewas highly variable but mainly developed subsequentto the first pyrolysis and reform treatments. Theconstruction of the capacitors was important, damagebeing largely confined to areas where the space tapewas in close contact with the surface. The bromine-methanol test showed that damage to the oxide wasindeed caused by pyrolysis although this wasdifficult to see by scanning electron microscopy. Itshould be noted that tests were performed on flatpieces of foil; in the case of a coiled capacitor withspacer tape pyrolysis damage may be expected to bemore severe because of the difficulty of pyrolysisproducts escaping from this construction.

The situation appears to be, therefore, thatpyrolysis initiates damage to the sites in the dielectricwhich subsequently suffer dissolution during thereform. The fact that the "dry" leakage currents ofsuch units are always much lower than "wet" leakagecurrents suggests that much of the reform effect isdue to electrical isolation of damage sites bydissolution of the overlying manganese oxide ratherthan reanodising of bare aluminium.

Life test results on solid aluminium capacitorsshowed that they recovered much more slowly frominterruption in polarisation than solid tantalum units.In general however their behaviour on life test wasmuch more stable than solid tantalum; no failuresoccurred and leakage currents tended to fall. The fallwas rapid initially and dependant on polarising voltage.Examination of capacitors after test showed why thiswas so; the damage sites in the dielectric had a plug ofoxide over them. It is probable that early in the lifetest continued anodisation occurs at the defect sites,plugging them with poorly formed oxide, and soreducing the leakage current.

Solid aluminium capacitors appear to be stable onlife test compared to solid tantalum units becausethey have no inherent degradation mechanism such asfield crystallisation. They are difficult to manufacturebut may be expected to become slightly better in useand relatively immune to catastrophic failure.

6 CONCLUSION

The basic mechanism of failure of solid tantalumcapacitors has been shown to be field crystallisationof the dielectric oxide. However, in the case of solidaluminium capacitor it appears that problems arecaused by the damage initiated in the dielectric by the


pyrolysis process which is subsequently aggravated bythe reformation.

ACKNOWLEDGEMENTS

This work has been carried out with the support of theProcurement Executive, Ministry of Defence, sponsored byD.C.V.D. The authors are also grateful to The PlesseyCompany Limited for permission to publish.

REFERENCES

1. R. L. Taylor and H. E. Haring, J. Electrochem. Soc.,103, p. 611, 1956.

2. D. Hayward, Electrocomp. ScL & Tech., 2, p. 249, 1976.3. N. F. Jackson, J. Appl. Electrochem., 3, p. 91, 1973.4. N. F. Jackson,and D. S. Campbell, Thin Solid Films, 36,

p. 331, 1976.5. D. S. Campbell, Rad. & Elec. Eng., 41, p. 5, 1971.6. A. R. Morley and D. S. Campbell, Rad. & Elec. Eng., 43,

p. 421, 1973.

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttp://www.hindawi.com Volume 2010

RoboticsJournal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation http://www.hindawi.com

Journal ofEngineeringVolume 2014

Submit your manuscripts athttp://www.hindawi.com

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014

SensorsJournal of


Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks


FAILURE MECHANISMS SOLID ELECTROLYTIC CAPACITORS

Documents

Transcript of FAILURE MECHANISMS SOLID ELECTROLYTIC CAPACITORS