A TEN STEP MECHANISM FOR THE PI'ITING CORROSION OF ALUMINIUM ALLOYS
M.C. Reboull.2,T.J. Warner, H. Mayer, and B. Barour Ipechiney Centre de Recherches de Voreppe, BP 27-38340 Voreppe, France 2Institute National Polytechnique de Grenoble (LTPCM/GSERIP) 38402 St Martin d'Hères, France
ABSTRACT Among the passive metals, aluminium is one of the valve metals. ln severe corrosive environments, aluminium, like other passive metals, is prone to localised corrosion which results from local breakdown of the oxide film. This paper presents a survey of the literature on this subject which results in a new ten-step mechanism for the initiation, the development, or the repassivation of pits, which takes ioto account bath the well-known galvanic model and the more receot capacitance mode!. Proposed ten step mechanism: 1. cr adsorption, io the oxide film's defects (micro-flaws), assisted by the high electric field (106 107V/cm) resulting from the Al-air corrosion cell (emf - 2.9 V) through the barrier oxide film, enhanced by the surface charge. 2. Slow oxygen reductioo on the cathodic area, charging the double layer capacitance (-50 J1F/cm2). 3. Breakdown of the oxide film at weak points corresponding to the micro-flaws. 4. Fast aluminium oxidation of bare aluminium at the break points
-
producing soluble chloride complexes. 5. Dissolution of chloride complexes and repassivation ofpits. These first five steps produce a large number (_106/cm2) ofmicro-pits (0.1-1 J11tl) . 6. The cathodic reactioo limits the propagation to a few pits, where the chloride build up is high eoough to form a "stable" chloridel
471
Vol. 15, Nos. 3-4, 1997
A Ten Step Mechanism for the Pitting Corrosion of Aluminium Alloys
oxychloride layer at the bottom of propagating pits. This layer must be renewed !aster than it dissolves, which implies a large enough cathodic area for each active pit. 7. Hydrolysis of soluble chloridesloxychlorides resulting in acidification (down 10pH = 3) of the solution within propagating pits, which is also a necessary condition 10maintain pit activity. 8. Al3+diffusion from the concentrated acidic solution inside the pit and aluminium hydroxide precipitation outside the pit in the neutral bulk solution. 9. Aluminium auto-conosion in the aggressive solution (AlCI], 3M, pH = 3) within the pit. producing H2 bubbles, which limit cr build up and acidification inside propagating pits. 10. Repassivation and pit death when the required conditions of stability of the chloride layer (JpnIrpit>10-2A/cm) are no longer met. The chloridel oxychloride film is dissolved and replaced by a passive oxide film. The solution within the pit is diluted and will revert to the composition of the bulk solution.
The new capacitance model, as opposed 10 the previonsly accepted continuonsand monotonicmicro-galvanicmodel, showsthat pitting conosion is a discontinuons phenomenon. The double tayer capacitance temporarilystoresthe electrochemicalconosionenergy.ln the abovemechanism, the essential features of both descriptions of the corrosion phenomenaare included.
KEY WORDS aluminium, corrosion, pitting corrosion, mechanism, review
INTRODUCTION Pitting corrosion is an important form of corrosion for passive metals. It reflects the quality of passivation and it is the common departure for ail the forms of corrosion for passive metals. This literature review is organised in nine sections:
472
M.c. Reboul, T.J. Warner, et al.
.
.
. . . . .
. .
Corrosion Reviews
aluminiumpassivity pitting initiation pitting propagation pitting corrosionas a discontinuousphenomenon metaIlurgicaleffects the influenceof surfacetreatment pitting potentialand repassivationpotential maximumpit depth measurementwithextremevalue statistics proposedmechanism
ALUMINIUM PASSIVITY Pure Aluminium (99.99%) The air formed oxide film is amorphous alumina (Al2Û3)/1.21. to 4 nm thick at roorn temperature. Thickness increases with temperature; typica1ly 10 nm after heat treatment at 400°C. ln contrast with humid environments. the external side of the oxide film hydrolysesto produce hydrated oxides such as bayerite (Al203. 3H20 or Al(OHh) formed under 70°C and bohmite (Al203. H20 or AlOOH) formed above 100°C /2/. Because hydrated oxides are less protective than anhydrous oxides. aluminium will oxidise and the film thickness will increase. The aluminium oxide film is usually a multi-layer oxide film: an anhydrous and amorphous barrier layer in contact with the metaI and a hydrated layer resulting from the superficial hydrolysis of the former. This oxide film is only stable in the pH range 4 to 9. Among the passive metals. aluminium is one of the valve metals. along with Hf.Nb. Ta, Ti and Zr 13/;an electriccurrentcan pass the metalelectrolyte interface in the reduction direction only. Cath99.99%). ln these capacitors the barrier oxide film is the dielectric, artificia1ly thickened by anodising in boric or tartaric acid electro1ytesabove the service voltage. Electrons can cross an oxide film a few nm thick by tunnel effects, but the aluminium oxide film is mainly an ionic conductor (through the vacancies
-
473
Vol. 15, Nos. 3-4, 1997
A Ten Step Mechanismfor
the Pitting Corrosion of Aluminium Alloys
wbich need an activation energy, wbich increases with distance between vacancies). Within the aluminium oxide film, O. anion and Al* cation are both mobile; their transport numbers are nearly 0.5 so the aluminium oxide film simultaneously grows at both interfaces 151. This oxide formation process produces an oxide film that is very adhesive 10 the aluminium metal (no voids are produced between metal and oxide, during oxide formation). ln addition, this oxide film is very compact, VoxN Al= 1.4/3/. Aluminium oxide contains many (104/cm2) microflaws (-10 nm) wbich have been observed by Thompson et al. 15-7/.
Aluminium Alloys The barrier oxide film formed on pure aluminium is a pedect insulator wbich is no longer true with impurities or alloying additions 131. More oxidisable clements than aluminium, Le., Li, will oxidise first and form poorly protective oxide layers at the extreme surface. On the other hand, more noble elements than aluminium, i.e., Cu, present in solid solution or in the form of small (5 10 50 nm) coherent precipitates, will accumulate at the metal oxide interface 151.From the experience of the workers at Pechiney, less mobile cations than Al*, i.e. M~, Be2+,despite being more active than aluminium, also accumulate at the metal oxide intedace. Large noncoherent precipitates (1 to 10 J1l1l)formed with more noble elements than aluminium generally form less reactive intermetallics than the aluminium matrix. For as-rolled aluminium, Lunder & Nisancioglu 181think tbat hard intermetallics are more or less coveredby soft aluminium matrix and oxides. A schematic view of aluminium surface is shown in Figure 1. Like other passive metals protected by a thin oxide film (Fe-Ni-Cr; Ti; Zr), aluminium is prone to pitting corrosion 191in severe environments. It is usually considered that pitting corrosion is a two-step phenomenon: first step initiation and ~nd step propagation. We will use tbis two-step partition for this review, even though it will rapidly appear too simplistic to descn'be the phenomena involved.
Pitting Initiation Pitting corrosionresultsfrom the corrosionreaction: 2Al + 3H20 + 3/2Û2 -. 2Al(OHh
474
MC. Reboul, T.J. Warner, et al.
Corrosion Reviews
AI203,3H20 CH2)n.1120 superfIclalcontamination lI2 C03. alkallne surfacesegregation
AI2 03 amorphous oxide barrierlayer 3nm
noble &Iarge elements segregation Fig. 1:
Schematic view of aluminium alloys passivity on roUed products.
This reaction is the sum oftwo electrochemical half-reactions: - at anodic sites: 2A1 ~ 2Al3++ 6e- at cathodic sites: 3H20 + 3/202 + 6e- ~ 60fr .
ln the galvanic model, these electrochemicalreactions occur simultaneously on different sites. This electrochemicalmechanism was first demonstrated by Brown & Mears /10/; confirmedrecently by Isaacs & Isbikawa/111with a scanningvibratingmicro-electrode. For aluminium, pitting corrosionoccurs most of the time in aerated cbloride containingenvironments(CIl. Accordingto Sugimotoet al. /121, the pitting densityincreaseswith the cbloridecontentand passes through a maximumfor NaCI SN. ln practicechlorideions are not suflicient:pitting corrosion does not occur in deaerated chloride solutions on isolated aluminium, a stronger oxidant than water is necessaryfor the cathodic reaction; dissolved oxygen in aerated solutionsis generally involved.The cathodic reaction is the limiting factor for pitting corrosion of aluminium /9/. Because of this, noble cations present in the solution (i.e. cU~ will plate on the aluminium surfaceand form a very activecathode producing pits in chloride containing solutions.Anodic or galvanic polarisation of aluminiumwill also producepits in (CIl containingenvironments,with or withoutoxygen.
475
Vol. 15, Nos. 3-4,1997
A Ten Step Mechanismfor
the Pitting Corrosion of Aluminium Alloys
Bogar & Foley /13/, Foroulis & Thubrikar /14/. Foley /lS/ apd Augustynski/161think that cbloridesare first adsoIbedon the oxide film then form soluble oxycbloridesallowing the propagaûon of pits. Augustynski /16/ verified cbloride adsorpûon on the oxide-soluûoninterfaceby XPS analysis.Bargeron& Givens/17/ propose six steps for the iniûaûon process: [1] CI- adsorpûon on the oxide film. through oxide defects, assisted by the electric field [2] CI- build up; cr is involved in the fondamental reacûon. The pitûng potenûa1 is connected to the cr content by Nernst's law: Ep = E~ +2.3n RT 3F 10g(CI-) presented in the Pitûng Potenûal section ofthis review. [3] Cbloride hydrolysis with gas formaûon H2 and/or HCI at the metal oxide interface AlCh + H20 ~ AlChOH + HCI (gas) [4] The gas pressure (H2 and/or HCI) increases and the blister begins to grow. [5] Micro cracks will fonD, allowing the reactants to reach the metal. [6] Finally the blister breaks grossiy and localised corrosion begins on a larger scale. Bargeron & Givens /17/ have observed blister formaûon under the oxide film. then blister cracking; the first three steps are sti1l hypotheûca1. McCafferty /18/ recently used the Bargeron & Givens /17/ iniûaûon process for modelling the pitûng iniûaûon on aluminium. They specified that adsorpûon of ions is most favoured when the surface charge is posiûve. This occurs when the pH is 1essthan PHnc. Le. 9 for AI. Chloride ions will then migrate in the oxide film through oxide vacancies assisted by the electric field and will form soluble chlorides snch as AlClt. at the metal-oxide interface, where pits will initlate /lS/. Videm's /19/ auto-radiographic studies with 36Cl conclude that no detectab1e chloride concentraûon may be seen before passivity breakdown, but high chloride concentraûons are detected after breakdown and formaûon of pits. This observaûon has been confirmed by Abd Rabbo et al. /20/. who immersed an aluminium specimen covered with a 72 nm oxide film for 24h in KCI lM solution. SIMS analysis shows CI concentration at the oxide-electrolyte interface, but no chloride within the
476
MC. Reboul, T.J. Warner, et al.
Corrosion Reviews
oxide film. For Strehblow /21/ contradictory data on cr adsorption result from different preparation of specimens (whether with a water rinse or not) and different sensitivities of the various analysis techniques (xpS, AES, SIMS). ln addition the oxide film can include an amorphous and anhydrous barrier oxide layer in contact with the aluminium metal and an external hydroxide layer. The presence of chloride within the external hydroxide layer does not mean passivity breakdown; cr must pass the barrier layer and reach the metal really to break passivity. Chlorides are not the only anions which may form pits on aluminium. Galvelle et al. /22/ have obtained pitting by polarising pure aluminium in the following solutions: NaCI, KBr, NaCIÛ2,NaNÛ) and NaSCN. The large size of some of these anions seems incompatible with their adsorption in the oxide film. For Augustynsky /16/ and Bargeron & Benson /23/, pitting corrosion in these environments includes side or preliminary reactions with aluminium: CI04"~ CI- ,N03" ~ NH4, SCN" ~ SH"",producing less voluminous species which could be the true anions responsible for aluminium pitting. On the other band we know that aluminium anodising, in different mineraI acids (H2S04, H~04, H2Cr04) produces porous oxides, the size of porosities varying with the anion /24/. Richardson & Wood /25/, Johnson /26/, Wood et al. /27/, Nguyen & Foley /28/, Thompson et al. /5-7/, Janik-Czachor et al. /29/ and Stehblow /21/ think that pits initiate at flaws (-10 nm) very numerous in aluminium oxide film: 104/cm2in pure aluminium (99.99%), 1010/cm2for Al-Cu alloys /5,30/. The presence of pre-existing defects within the oxide film explains why electrochemical noise is obtained as soon as aluminium is immersed in a pitting solution. There is no incubation period, corresponding to the slow degradation of passivity by CI- adsorption then migration within the oxide film. ln these conditions how can we explain the induction time t observed by BroU et al. /31/ when polarising aluminium anodicaIiy to E at different chloride contents: y~
E_E ) = 2.9x 1O-sC-o.87 CC ( P
This incubation time could be due to the necessary cr build up and acidification within defects necessary for propagation. A pH of roughly 3 and a 3M AlCh content have been measured within active pits /32/. incubation time decreases when cr concentration increases /33/.
477
Vol. 15, Nos. 3-4,1997
A Ten Step Mechanism for the Pitting Corrosion of Aluminium Alloys
ln a recent AFM pitting study of mecbanically polished 2024-T4 (Al-CU) in an aerated chloride solution, Wamer & Schmidt /34/ have observed the formation of a large number of micro-pits ( 100 nm) within the first 2 minutes in the depleted Cu zone near the large Al2Cu precipitates. Pitting initiation results in the formation of a large number ( 10%m2) of micro-pits (0.1 - 1 J1l1l).
Pitting Propagation
.
Stabilisationof pits requires a minimumcr concentrationwithin pits 135/. Within the electrolyte,the current producedby the micro-galvaniccell is transportedby ions.Becausechlorideions movefasterthan most other ions, the current will be mainlytransportedby cr ions. This causesthe formation of a conœntrated AlCh solutionwithin activepits (Fig. 2). AlCh solutions are saturatedat 3M and the 90% saturatedsolutionpH is 3 /32/. Successive polarisations show that when polarisation stops, pits passivate.If the specimenis polarisedagain the passivationis so perfectthat new pits are formedat other sites/36/. Zahavi & Metzger/37/ have studied pitting of aluminiumin H2SO..(2M)solutionwith anodicpolarisation.They obtained a large number of micro-pits but repassivation was very fast. Stability of the oxide in the absence of cr does not result from lack of breakdown, but from the quality of repairs. The influence of cr is a determining factor above the minimum co~centration;it prevents repairs
Fig. 2:
478
Pit propagation
Corrosion Reviews
M.c. Reboul, T.J. Warner, et al.
and allows pit propagation. One of the solutions used 10 prevent pitting fro alUminium alloys is 10 increase the oxide film thickness by anodising in H2S04 solution. The growth of the oxide film thickness does not occur in chloride polluted solutions. The stabilisation of propagating pits requires a critica1 cr concentration within pits 135/. Pride, Scu11y& Hudson /38/ estimate 10 3M the necessary concentrations within an open pit for it to survive. Impressed current tests show that a minimum current is needed (>0.1 mAlcm2) for the necessary cr build up. This acidified and concentrated chloride solution prevents pit repassivation. So a propagating pit will continue to grow due to this autocatalytic mechanism, as long as the micro-galvanic cell can renew the solution within the pit more rapid1y than diffusion and hydrogen bubbles tend 10dilute it. According to Nguyen & Foley /28/, the determining step is the formation of soluble aluminium chlorideloxychloride complexes, in place of the passive oxide film at active site areas inside pits. This is the reason why the anion involved is so important. Cblorides adsorbed at flaws /27/ form soluble chloridesloxychlorides during Al oxidation /13,14/ allowing pit propagation: _
(n-3)-
Al + nCI ~ Al Cln
+ ne-
According to Bogar & Foley /13/ the number of chloride ions involved in complexes varies with the chloride content of the bulk solution: 3.0 < n < 4.8 for high chloride content 8 < n < 12 in diluted solutions.
This chlorideloxychloridelayer must be present for pit survival. To maintain this soluble layer, it is necessmyto form it morerapid1ythan it dissolves,whichrequiresa largeenoughcurrent. For Pride et al. /38/, stablepits needto meet the criterion:(ipJrpit> 10-2 A/cm)betweenthe anodiccurrent (in the pit) and the radius of the pil For isolated specimens, cr build up depends on the correspondingcathode activity, the bulk chloridecontent,the pit size, and competitionwitll other pits /39/. This pit solution acidifies due to hydroxide precipitation in nentraI solutions:
479
Vol. 15. Nos. 3-4.1997
A Ten Step Mechanism for the Pitting Corrosion of Aluminium A//oys
For Pryor /40/, this reaction occurs as soon as the A13+content is greater than 10-9ions g/1, and produces a minimum pH = 2.8. pH values of 3 to 4 have been measured within active pits /19,32,39/. Additional self-corrosion within the pit will occur, due to aeidification and high chloride content This self-corrosion reaction: AJ+3HP ~AJ(OH)3+3/2 ~ t produces H2 gas within pits, as observed by many workers. The contribution of this "side" reaction to the pit propagation is not significant according 10 Isaacs & Ishikawa /11/. Hydrogen bubbles and diffusion limit cr build up and acidification within pits, and help homogenisation of the pit solution with the bulk solution, resulting in pit death. There is general agreement that there is a constant anodic current within active pits, but the estimated current density varies with authors, i > 0.1 A/cm2 for 8ato /35/, 1 for 8trehblow /21/ and 3 to 30 A/cm2 for Baumgartner & Kaesche /41/. For Wong & Alkire /32/ the corrosion products migration limits the oxychloride dissolution, which controls pitting corrosion. For 8zlarska8mialowska /42/ migration within the pit is not the controIling factor. Pit propagation is controlled by alloying elements' solubility in the aeid solution within pits. But the pitting potential measurements used to support this view are not convincing as far as aluminium is concemed, as discussed in the pitting potential section of this review. For Aziz /43/ avery large number of pits initiate rapidly, but most of them stop after a few days. Ishikawa & Isaacs /44,45/ observed an aluminium specimen during pitting with a scanning vibrating electrode; from their studies, pit activity varies with time. A pit appears, its activity increases, pass through a maximum, then decreases and disappears and finally appears again. With anodic polarisation pit propagation results from the formation of discontinuous micro-tunnels /41,46/. Polarisation changes pit distribution; anodic current is constant within pits if the impressed current increases, the pit number increases /25/. On isolated specimens the higher the pit density, the lower the penetration. For Mattson /39/ pitting corrosion of aluminium alloys is controlled by the cathodic reaction; pit formation depends on the swface area and on
480
MC. Reboul, T.J. Warner, et al.
.
Corrosion Reviews
cathode activity.Aziz & Godard/47/ bave studiedthe pitting probabilityof aluminium in Kingston water in the 0.06 to 243 cm2 surface range. Pitting probability increases with the specimen surface area from 0.06 to 60 cm2; it is constant and equal to 1 above 60 cm2 (in Kingston water).
Pitting Corrosion as a Discontinuous Phenomenon It is well known that the corrosion potential of a passive metal shows fast fluctuations, during pitting corrosion, towards more active values (active state). Present studies of these fluctuations, called ElectroChemica1 Noise Analysis (ECNA), link this noise to individual activation repassivation events. ln his ECNA study on stainless steels Isaacs /48/ indicates that metastable pitting transients result from the slow charge of the oxide surface capacity which "periodically" discharges very rapidly. Around an active pit, the oxide film is charged slowly like a capacitance (20 to 80 pF/cm2) by the slow cathodic reaction (generally dissolved oxygen reduction) then discharged rapidly and perhaps stochastically. We can apply this model to aluminium alloys considering both a -2.9 Volt emffrom the Al/air corrosion cells and a -3 nm barrier oxide film. We may suggest that the cathodic reaction charges the double layer capacitance until the dielectric breakdown of the oxide film (in cr containing environments) at the weakest points the micro-cracks. The density of these micro-cracks was found to be of the order of lOs to 107 cm2 /5,30/, which is also the density of micro-pits observed during the first minutes of Al-Cu corrosion in aerated NaCI solution /34/ or within the first milliseconds of an anodic polarisation /4/. Time fluctuations: The time fluctuations generated by the pitting corrosion of aluminium alloys was investigated by two of us /49,50/ in an aerated chloride solution and in the ASSET solution (NH.CIIM + ~N03 O.25M + H202 O.09M + (NH.)2CJI20 O.OIM) which is used at PechineyCRV to predict the pitting corrosion behaviour of aluminium alloys in a marine atmosphere. It forms large pits in susceptible Al alloys within 48h. Two aluminium specimens were immersed (up to 15 days) in the ASSET solution without H202 to decrease its aggressivity. [The ASSET test (ASTM066) bas been designed to predict exfoliation corrosion bebaviour of 5XXX aluminium alloys in marine environments. We also use it at Pechiney-CRV at room temperature for 24h, to predict pitting bebaviour of aluminium
481
Vol. 15,Nos.3-4, 1997
A Ten Step Mechanismfor
the Pitting Corrosion of Aluminium Alloys
alloys in marine environments. We get fairly good correlaûons with lxxx, 3XXX and SXXX series aluminium alloys.] The corrosion potenûal V(t) and the disequilibrium current I(t) (between the two specimens) were simultaneous1y recorded. Fluctuaûons of these t'Woquanûûes are related either to the birth and death of metastable pits or to mulûple events (type 1 and type 2) are observed, alternaûng with ûme in the same experiment (Fig. 3). Type 1 can be rougbly analysed as being mainly a stocbasûc process. However, type 2 seems to be quasi-periodic, with strong coupling between I(t) and V(t) signals.
Type 1
-V(Q M 1 .1.4e-003 .
Type 2
1
1
1
1
1" 1 1 r
II-v(QM
1 'k
Il
1.0e-003
1\11/11111111111 III
1..
'mnvrnvmr
.Ue-003
.I.Oe-003
.2.0e.OO3 1 0
1 "V' 1 1 2
1(Q(A) IL. LI
4 .t(SIl
1
1
'.0e-007
1
1
2
3
4
I(t)lA)
Il
1
0
S.Oe-007
1
O.Oe+OOO
8.0e.007
.
811Jft' IIln
1
n
1.. '.. il
7.0e.Q07 1 2
Fig. 3:
482
l' 3
1 4
t (S)
,
-S.Oe.007
.
-\.Oe.OO6 tls)
.1.Se.OO6 .
o
2
3
4
V(t) and let) noise Type 1 (initiaûon) and Type 2 (propagation) during pitting corrosion of aluminium alloys (from Ref./49,SO/).
Corrosion Reviews
MC. Reboul, T.J. Warner, et al.
The autocorrelation function Cvv(t) allows comparison of the signal at a time t with the signal just before. ln other words, it makes it possible 10 know if the event which generates the signal is dependent or not on the events occurring just before. For the corrosion potential the autocorrelation curves are definitely different in the two cases (Fig. 4). For type 1, Cvv(t) decreases rapidly and monotonical1y to null, which is characteristic of a stochastic phenomenon; each event is independent and occurs unevenly. On the other band, for the type 2 signal the autocorrelation curve exhibits several extremes, which is characteristic of a deterministic phenomenon which could resuit from a multi-step mechanism Relevant information can also be drawn from the phase portraits which are a 3D time derivation of the signal in the [I, dYdt, d2Ydt1 phase diagram (Fig. 5). No more information is obtained for type 1 signais; the trajectol)' fi1ls the whole space and only confirms a stochastic phenomenon. For type 2 the situation is more complex: an attractor is evidenced whose dimension
CVV(t)
Hm
= t-++OO
t
1
[0
V(t) *V(t + t)dt
]
1 0.9 0.8 0.7 0.6 O.S 004 0.3 0.2 0.1
o -0.1 -0.2 0.5
Fig. 4:
1.5
2
tau (s)
Potential autocorrelation function (from Ref./49,SOI).
483
~ 00 ~
us X'Ct)
us XCt)
T
xcu
8011 11_1 .DAT
Fig. 5:
.....
................
1 ~..oii.:..
le
,91'li..
'~~;:':"""""""""""".""."'1 ..'
.
r.?r:~~~~ ~:.? :..t~
