An Electrochemical Study of the Crevice Corrosion of Titanium

Estudou-se a influência de diferentes fatores na corrosão por fresta em titânio comercialmente puro, em cloreto de sódio, a temperaturas intermediárias. Investigou-se as condições de pH, desarejamento do meio, temperatura e das caraterísticas geométricas da fresta, sob as quais este tipo de corrosão se apresenta. O tipo de íon de titânio que predomina na zona ativa foi determinado como sendo Ti. Foram usadas técnicas eletroquímicas e gravimétricas. Pites foram encontrados no titânio apenas a al tos potenciais (~5,3V) mas a corrosão por fresta pode aparecer ao potencial de corrosão. A utilização de uma célula de dois compartimentos mostrou que não se pode ultrapassar uma relação máxima en tre área catódica e anódica para poder modelar a ativação da superfície metálica dentro da fresta. O desenvolvimento de uma queda ôhmica através dos produtos de corrosão na entrada da fresta é proposto como um fator estabilizador da corrosão.

the crev ice as a func tion of time 7 .Nev er the less, the reproducibility of crev ice cor ro sion tests is poor, in duc tion times may be long and no sat is fac tory re la tion of these times to geo met ric fac tors ( i.e. crev ice width and depth, micro-and macro-roughness of the crev ice wall), nor to other vari ables such as tem per a ture and the na ture of the ma te rial in con tact with the metal, has been pro posed.More over, not with stand ing the fact that the at tain ment of a crit i cal con di tion of crev ice so lu tion is gen er ally ad mit ted, it is not clear how this con di tion is sta bi lized over time.
In this work, crev ice cor ro sion of com mer cially pure tita nium was stud ied us ing dif fer ent elec tro chem i cal techniques: potentiodynamic po lar iza tion curves, spon ta ne ous ac ti va tion mea sure ments, elec tro chem i cal sim u la tion of crev ice cor ro sion in a two-compartment cell, and the re dox po ten tial mea sure ment of the sim u lated crev ice so lu tion.Im mer sion tests in an au to clave were also per formed.Corro sion prod ucts were an a lyzed by Raman spectroscoy, X-ray dif frac tion, and scan ning elec tron mi cros copy (SEM).

Ex per i men tal
The test spec i mens of com mer cially pure ti ta nium were used in block or sheet forms of dif fer ent di men sions, with the sam ple ar eas ex posed to the so lu tion varyng from 1 to 12 cm 2 .The spec i mens were pol ished with sil i con car bide pa per (from 100 to 2400 grit).For the elec tro chem i cal tests, sam ples were pre pared with out or with crev ices ob tained by fas ten ing an acrylic sheet with a Ti screw on the sam ple sur face.For the im mer sion tests in the au to clave, crev ices were pro duced by ty ing sand wiches of in ert ma te rial and tita nium plates to gether with PTFE tape.In the elec tro chemi cal tests, the Py rex cell con tain ing the sam ples was placed in side a wa ter bath at room tem per a ture or at 50, 70, or 90 °C.The so lu tion used was 1 M so dium chlo ride pre pared with dou ble dis tilled wa ter, acid i fied with HCl, so as vary the pH be tween 0.87 and 6.72.Some tests were made in Na 2 SO 4 at pH 6.72.For the deaerated tests, pure ni tro gen was bub bled in the so lu tions.The pH within the crev ice was mea sured with pH-indicating pa per af ter the test.The ref er ence elec trode used a the sat u rated cal o mel elec trode (SCE), and all po ten tials in the text are re ferred to it.The equip ment used in the elec tro chem i cal tech niques in cluded a potentiostat (PAR model 174), a pro gram mer con trol ling the po ten tial scans (BESPW model SMP 72), and an x-t recorder (ECB).

Spon ta ne ous ac ti va tion curves
The ef fect of the en vi ron men tal vari ables (pH, so lu tion tem per a ture, and the pres ence or ab sence of O 2 ) on crev ice cor ro sion was stud ied by mea sur ing the spon ta ne ous ac tiva tion time, mon i tor ing the cor ro sion po ten tial, E corr , as a func tion of time, of uncreviced sam ples in NaCl so lu tions.This type of test eval u ates the en vi ron men tal con di tions un der which crev ice cor ro sion may take place.
Fig ure 1 shows the ef fect of pH on the E corr vs. time (t) curves in deaerated me dium.It is ob served that, for pH 2, for both NaCl and Na 2 SO 4 , the pre-existent ox ide film on the ti ta nium sam ple tends to dis solve ini tially.The po tential drops to ap prox i mately -550 mV, but these so lu tions seem un able to com pletely dis solve the film, since the poten tial soon in creases again to less neg a tive val ues.This increase in di cates a new for ma tion or trans for ma tion of a pas sive film, which was con firmed by a change in the super fi cial color to dark gray.This film for ma tion may be due to a re ac tion of the metal with wa ter.
At pH 2, dis so lu tion of the ox ide film is more ef fec tive.In the same fig ure it can be ob served that the po ten tial decay is faster for lower pHs, sta bi liz ing at a con stant value of around -720 mV.This kind of ex per i ment in very acid so lutions was also car ried out by Kelly 2 , who ob served a change in the slope of the de cay curve at trib uted to a tran sition from a thin ox ide phase to a monolayer of ad sorbed spe cies.
In the equi lib rium po ten tial-pH di a gram at 90 °C presented by Lee 4 , the po ten tial value of -720 mV is sit u ated in the Ti 2+ zone, al most on the bor der with the Ti 3+ zone, in dicat ing that Ti may dis solve with the for ma tion of one or both of these spe cies.
The ef fect of tem per a ture on spon ta ne ous ac ti va tion was also stud ied.It was ob served that the dis so lu tion of the ox ide film be came faster as the tem per a ture in creased.The sta tion ary po ten tials were also more neg a tive at higher tem per a tures, which means that cor ro sion is more likely as the tem per a ture in creases.

Potentiodynamic curves af ter spon ta ne ous ac ti va tion
Fig ure 2 shows potendiodynamic po lar iza tion curves ob tained af ter the spec i mens were pre vi ously ex posed to the so lu tions for 1000 min, as in Fig. 1.This fig ure shows the ef fect of pH on the an odic po lar iza tion curves.At pHs less than 2, there is al ways an ac tive cur rent peak at low poten tials.The peak cur rent den sity, i m , in creases and the corre spond ing po ten tial, E m, also in creases when the pH de creases, in agree ment with Kelly's data 2 .At pHs greater than 2, the ac tive peak is not vis i ble and a sec ond peak appears at higher po ten tials; this peak po ten tial in creases as the pH is raised.A sim i lar peak was ob served by Levy and Sklover 5 for com mer cial ti ta nium in 5% HCl, and was attrib uted to a sec ond ary ac tive-passive tran si tion.Shibata and Ameer 8 ob served a sim i lar peak for zir co nium in 1 N H 2 SO 4 at 30 °C, which it ap pears at 1.8 V vs. Ag/AgCl.These au thors at trib uted this peak to a phase change in the film that passes from a thin film be fore the peak to a thick golden film with struc tural de fects af ter the peak.The re actions that oc cur in zir co nium and ti ta nium may be sim i lar, as they both be long to the pe ri odic group 4.
The tem per a ture did not show any sig nif i cant in flu ence on E m , which re mained al most con stant at -650 mV; i m , on the other hand in creased when the tem per a ture in creased, in agree ment with re sults re ported by Griess 1 and Kelly 2 .Above 3 V, the cur rent in creases more rap idly at higher tem per a tures.Nev er the less, no pit ting was ob served in this zone.Thus, it was as sumed that the an odic re ac tion in this zone cor re sponds to the ox i da tion of wa ter.Using ex per imen tal data for tem per a tures be tween 50 and 90 °C, the acti va tion en ergy for the dis so lu tion of ti ta nium was ob tained from the slope of the ln i m vs. 1/T plot; the re sult ing value (5.3 kcal/mol) is quite low when com pared with the value ob tained by Kelly in a so lu tion of HCl 0.1 M + NaCl 0.9 M (11 kcal/mol) but his data were ob tained at lower tem per atures (be tween 30 and 75 °C) which may in di cate some differ ence in the dis so lu tion path.

Sim u la tion of the crev ice con di tions in a two-compartment cell
The ob jec tive of sim u lat ing crev ice cor ro sion con ditions in a two-compartment cell is to eval u ate the ef fect of the elec tro lyte com po si tion, the aer a tion of the ano lyte and catholyte, and the ra tio be tween the an odic area (A a ) and the cath odic area (A c ), on the cor ro sion pro cess at the anode, un der purely spon ta ne ous con di tions (with out the appli ca tion of an ex ter nal cur rent).
For this type of test, two phys i cally sep a rated metal sam ples (with out crev ices) were used, which were linked to each other by means of a potentiostat act ing as a zero resis tance am me ter which main tained the sam ples at a zero po ten tial dif fer ence and si mul ta neously read the short-circuit cur rent.Both com part ments were con nected through a salt bridge; one sam ple was im mersed in a NaCl so lu tion, acid i fied to pH 1 and deaerated with N 2, sim u lating the crev ice en vi ron ment (an odic sam ple).The other sam ple was im mersed in a neu tral aer ated NaCl so lu tion, rep re sent ing the ex ter nal sur face of the crev ice (cath odic sam ple).The po ten tials of both sam ples were in de pendently mon i tored with the help of high re sis tance voltammeters and sat u rated cal o mel ref er ence elec trodes.
Fig ure 3 pres ents the vari a tions of the an ode (E a ) and cath ode (E c ) po ten tials, and the an odic cur rent den sity as a func tion of time.Ini tially, both sam ples were pas sive and showed rel a tively high po ten tials, but soon these po ten tials dropped, show ing that an ode ac ti va tion oc curred.E c always re mained higher than E a .The cur rent den sity ac compa nied the po ten tial vari a tion.Ini tially, there was lit tle cur rent cir cu lat ing, but then the cur rent in creased with intense os cil la tion, which may be at trib uted to rup ture and ref or ma tion ep i sodes of the pas sive film on the an odic surface.
Fig ure 2. Potentiodynamic curves for dif fer ent pHs af ter spon ta ne ous Fig ure 3. Vari a tion of the an odic and cath odic po ten tial and the anodic cur rent den sity in elec tro chem i cal sim u la tion tests of crev ice cor ro sion in a two-compartment cell (T = 90 °C, 1 M NaCl, and A c/A a Kain and Lee 9 pro posed four stages of crev ice cor rosion.The first (ex haus tion of ox y gen) and the sec ond (increase of acid ity and chlo ride con tent) are as sumed to have al ready oc curred in the pres ent tech nique, be cause an acid and deaerated so lu tion is used in the an odic com part ment of the cell.So, only the last two stages are sim u lated, be ing the per ma nent rup ture of the pas sive film and the prop a gation of crev ice cor ro sion at the an ode.
The fol low ing area ra tios (A c /A a ) were tested: 8.0; 1.5; 1.0, and 0.6.It was found that for ti ta nium, the ra tio should be less than 1 in or der for cor ro sion to set in.Oth er wise oscil la tions be tween an odic and cath odic cur rent val ues occur.These re sults are con trary to those of Rosenfeld and Marshakov who sup posed that an in crease in the above-mentioned area ra tio would al ways in crease crev ice cor ro sion 10 .How ever, they are sim i lar to those ob tained by Kain and Lee for stain less steel 11 .The re sults may be explained by hy po thet i cal an odic and cath odic po lar iza tion curves, con sid er ing that re duc tion of ox y gen oc curs in the cath odic zone where the pH of the so lu tion is neu tral.When A c /A a is too high, the cath odic curve should in ter cept the an odic curve at a po ten tial in the pas sive zone, and this sit ua tion is not ex pected to lead to crev ice cor ro sion.When this ra tio is small, the in ter cep tion is ex pected in the ac tive zone of the an ode.An other pos si bil ity in this sec ond case is to as sume that there is a rea son able dif fer ence be tween E c and E a , which may cor re spond to an ohmic drop be tween the in side and the out side of the crev ice, as sug gested by Shaw et al. for the crev ice cor ro sion of a nickel al loy 12 .
It was also ob served that cor ro sion does not take place when the cell com part ment con tain ing the an ode is aer ated, de spite a pH of 1, in agree ment with the re sults of Kain and Lee for stain less steels 11 .This con firms that ox y gen de pletion of the crev ice is a nec es sary step for crev ice cor ro sion in ti ta nium.

Re dox po ten tial
Fig ure 4 pres ents the re dox po ten tial (E re dox ) of a 1 M NaCl so lu tion, with an ini tial pH = 1, at a tem per a ture of 90 °C in con tact with ti ta nium fil ings, in a deaerated N 2 medium.This so lu tion rep re sents the one that would be gen erated by crev ice cor ro sion.E re dox be comes more neg a tive with time, at tain ing a value of -450 mV af ter 30 h.It should be noted that af ter 15 h the so lu tion changed from col or less to vi o let blue, in di cat ing the pres ence of Ti 3+ , which should there fore be ex pected in the crev ice, in agree ment with the Pourbaix di a gram 4 and with the cor ro sion po ten tial val ues of Fig. 1, which also point to the pres ence of Ti 3+ in the crev ice so lu tion.

Dis con tin u ous and con tin u ous cy clic voltammetric curves
Fig ure 5 shows the dis con tin u ous and con tin u ous cy clic voltammetric curves ob tained in 1 M NaCl at 90 °C.Curves 1 and 2 are dis con tin u ous curves with a hold at +5300 mV (with a crev ice, a for ward scan at v = 1200 V/min and a reverse at v = 120 mV/min) and curve 3 is a con tin u ous one (with out a crev ice, and with scan ning in both di rec tions at v = 1200 mV/min).When the po ten tial was main tained at +5300 mV for 120 min (curve 1), the cur rent den sity, in spite of de creas ing with time, showed fairly high val ues (higher than 70 µA.cm - ).On re vers ing the po ten tial scan, the sam ple repassivated with an os cil lat ing cur rent.Af ter the test, a deep hole was ob served in the crev ice, shown in the SEM mi cro graph of Fig. 6.
Curve 2 was ob tained by hold ing the po ten tial at +5300 mV for 200 min.The cur rent in creased dur ing the hold.Fig ure 5. Dis con tin u ous (for cre viced sam ple) and con tin u ous (for uncreviced sam ple) cy clic voltammetric curves at 90 °C in neu tral and aer ated 1 M NaCl.The for ward scan was at v = 1200 mV/min, and the re verse at 120 mV/min for the dis con tin u ous curves; the for ward and re verse scans were at v = 1200 mV/min for the con tin u ous curve.Hold at +5300 mV for 120 min (curve 1) and 200 min (curve 2), and no hold (curve 3).
When the po ten tial scan was re versed, the sam ple repassivated slowly.At the end of the test there were several holes and some cor ro sion prod uct in the crev ice, and pit ting was also ob served on the free sur face.This po ten tial is there fore higher than or equal to the pit ting po ten tial, as pits were formed in the area ex posed to the bulk so lu tion.Uhlig and Revie 13 re ported a pit ting po ten tial of 6.74 V at 90 °C.
The con tin u ous curve 3 pres ents de creas ing cur rents on the re verse scan.No cor ro sion was ob served, only a sur face color change to dark blue, prob a bly as so ci ated with a mod ifi ca tion in the pas si vat ing film.These ex per i ments show that even ex cur sions to high po ten tials fail to pro mote instan ta neous lo cal ized cor ro sion of ti ta nium, in spite of the pres ence of a phys i cal crev ice.Nev er the less, crev ice cor rosion is nu cle ated sooner than pit ting at the same po ten tials.
When the spec i men was main tained at +5300 mV for 930 min in the same so lu tion, cat a strophic cor ro sion occurred through out the whole sam ple, be ing more pronounced within the crev ice.The pH in the crev ice was about 1, and great quan ti ties of cor ro sion prod uct pre cip itated.This prod uct was later dried and an a lyzed.Ti ta nium di ox ide in ores can be found in three forms: rutile (tetragonal), ana tase (tetragonal), and brookite (orthorhombic) 14 .Raman spec tros copy anal y sis as well as X-ray anal y sis of the cor ro sion prod uct showed fea tures of ana tase.Nev er the less, not all peaks could be iden ti fied, indi cat ing a more com plex na ture of the cor ro sion prod uct.De tails of the anal y sis are given else where 6 .Sev eral authors have tried to iden tify dif fer ent forms of TiO 2 by Raman spec tros copy, but the re sults do not al ways agree 14 .

Weight loss method (im mer sion tests)
For weight loss tests the spec i mens were placed in side a PTFE lined au to clave half filled with neu tral 1 M NaCl at 100 °C for 15, 30, and 47 days.Af ter the tests, in or der to de ter mine the weight loss, the sam ples were pick led in a 20% HNO 3 -2% HF so lu tion at 60 °C.
The area used to cal cu late the cor ro sion rates, V c, was only the crev ice sur face, as no vis i ble at tack was found on the ex ter nal sur faces.For each test, six sam ples were placed in the au to clave, two of each sand wich type: Ti/acrylic sheet, Ti/PTFE, and Ti/Ti.It was noted that in some cases cor ro sion did not take place be cause the crev ice was too tight.Mean cor ro sion rates of ca. 100 mg dm -2 day -1 (mdd), 200 mdd, and 300 mdd were found for the tests of 15, 30, and 47 days, re spec tively.The pH in the crev ice for times lon ger than one month was around 1. For the 15 day ex per i ment, the pH mea sured with pa per strips was higher than 2, but as the cor ro sion had not yet spread over en tire the crev ice sur face, the mea sured value was prob a bly only a mean value be tween higher and lower ones.Also the cor re spond ing lower cor ro sion rate must be at trib uted to the non-uniform at tack of the sur face.Note that the cor ro sion rate is higher as time in creases and the solu tion within the crev ice is more acid.For 47 days a very ad her ent cor ro sion prod uct ob structed the crev ice entrance, which can be at trib uted to the pre cip i ta tion of the tita nium ions com ing from the in te rior of the crev ice as they come in con tact with the less acidic so lu tion out side the crev ice.This cor ro sion prod uct rep re sents a bar rier between the an odic (in side) and the cath odic (out side) regions of the crev ice, and prob a bly leads to a sig nif i cant ohmic drop, thus sta bi liz ing a po ten tial dif fer ence be tween both re gions.It is worth not ing that the cor ro sion rate obtained from weight loss data (equiv a lent to some hun dreds of µA.cm -2 ) is much higher than the one ex tracted from the cor ro sion cur rent den si ties in the two-compartment cell sim u la tion.This may be as cribed not only to the 10 de gree dif fer ence in tem per a ture be tween the two ex per i ments, but also to the dif fer ent time scale, be sides the dis tinct phys i cal con di tions of the spec i mens.The dou ble cell exper i ments were run for only 12 h, and in the au to clave tests it was ob served that the mean at tack rate was still in creasing af ter 15 days.Yao et al. 7 mea sured cur rent den si ties in a ti ta nium crev ice at 100 °C and ob tained data of c.a. 30 µA.cm -2 , but their ex per i ments were not con ducted for more than 90 h.

Con clu sions
The re sults of the pres ent study con firm cer tain assump tions made in the lit er a ture about the crev ice cor rosion of ti ta nium in chlo ride so lu tions.This form of cor ro sion can arise at an open cir cuit po ten tial when Ti forms crev ices with acrylic, PTFE, or Ti ta nium it self, at around 100 °C.The prin ci pal me tal lic ion in the crev ice solu tion is Ti 3+ , the pH of this so lu tion is around 1, and very high cor ro sion cur rent den si ties can be found in it.The temper a ture not only fa vors the ki net ics of ac tive cor ro sion, but Fig ure 6.A scan ning elec tron mi cros copy im age of the in te rior of the crev ice of a sam ple sub mit ted to dis con tin u ous voltammetry, and held at +5300mV for 120 min.also ac cel er ates the dis so lu tion of the pre-existent pas sive film.
By means of the sim u la tion tech nique of the crev ice con di tions in a two-compartment cell, it was found that the ra tio of cath odic to an odic area (A c /A a ) should be less than 1 for ti ta nium in NaCl in or der for crev ice cor ro sion to become es tab lished.This re sult can be better in ter preted as the ne ces sity of an ap pro pri ate ohmic drop be tween an odic and cath odic ar eas to sta bi lize the crev ice cor ro sion.This ohmic drop, which may be as so ci ated with the thick cor rosion prod ucts pro trud ing from the ac tual Ti crev ices, cannot be prop erly mod eled in the two-compartment cell.

Ac knowl edg ments
The au thors wish to thank CNPq and FAPERGS for finan cial sup port in the de vel op ment of this re search, and Prof. Sonia R. Biaggio for many help ful com ments.

Fig ure 1 .
Fig ure1.Spon ta ne ous ac ti va tion curves for uncreviced Ti sam ples in a deaerated me dium for sev eral pHs; T = 90 °C, 1 M NaCl.

Fig ure 4 .
Fig ure4.Vari a tion of the re dox po ten tial of a deaerated 1 M NaCl solu tion, pH = 1, in the pres ence of Ti fil ings (T = 90 °C).