10

Click here to load reader

Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

Embed Size (px)

Citation preview

Page 1: Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

Fc

BMa

b

c

a

ARRAA

KCTCAETF

1

bfcAioegsfstd

S

M(

0d

Electrochimica Acta 56 (2011) 9609– 9618

Contents lists available at ScienceDirect

Electrochimica Acta

jou rn al hom epa ge: www.elsev ier .com/ locate /e lec tac ta

ailure mechanism of thin Al2O3 coatings grown by atomic layer deposition fororrosion protection of carbon steel

elén Díaza,b, Emma Härkönenc, Vincent Mauricea,b,∗, Jolanta Swiatowskaa,b, Antoine Seyeuxa,b,ikko Ritalac,∗∗, Philippe Marcusa,b,∗

Chimie ParisTech, Laboratoire de Physico-Chimie des Surfaces (LPCS), 11 rue Pierre et Marie Curie, F-75005 Paris, FranceCNRS UMR 7045, 11 rue Pierre et Marie Curie, F-75005 Paris, FranceLaboratory of Inorganic Chemistry, University of Helsinki, P.O. Box 55, FIN-00014 Helsinki, Finland

r t i c l e i n f o

rticle history:eceived 25 May 2011eceived in revised form 25 July 2011ccepted 26 July 2011vailable online 4 August 2011

eywords:arbon steel

a b s t r a c t

Combined analysis by electrochemical impedance spectroscopy (EIS), time-of-flight secondary ion massspectrometry (ToF-SIMS) and field emission scanning electron microscopy (FESEM) of the corrosion pro-tection provided to carbon steel by thin (50 nm) Al2O3 coatings grown by atomic layer deposition (ALD)and its failure mechanism is reported. In spite of excellent sealing properties, the results show an averagedissolution rate of the alumina coating of ∼7 nm h−1 in neutral 0.2 M NaCl and increasing porosity of theremaining layers with increasing immersion time. Alumina dissolution is triggered by the penetrationof the solution via cracks/pinholes through the coating to the substrate surface where oxygen reduction

hin oxide coatingsorrosion protectionLDISoF-SIMSESEM

takes place, raising the pH. At defective substrate surface sites of high aspect ratio and concentratedresidual mechanical stress (along scratches) presumably exposing a higher steel surface fraction, local-ized dissolution of the coating is promoted by a more facile access of the solution to the substrate surfaceenhancing oxygen reduction. De-adhesion of the coating is also promoted in these sites by the ingress ofthe anodic dissolution trenching the steel surface. Localized corrosion of the alloy (i.e. pitting) is triggered

tion o

prior to complete dissolu

. Introduction

Atomic layer deposition (ALD), a CVD derived method with aroad spectrum of applications [1,2], is an ideal candidate for theabrication of thin coatings with excellent sealing properties fororrosion protection of high-precision metallic parts or systems. InLD the introduction of precursors is alternated and separated with

nert gas purging to ensure reaction between the precursors onlyn the substrate surface. The growth proceeds through cyclic rep-tition of saturative chemical reactions allowing a (sub)monolayerrowth per cycle. The same amount of material is deposited on allurface sites, and thin films with high density, conformality, uni-ormity and low overall defect density are produced even at large

cale and on high aspect ratio substrate surfaces [1,3]. An accuratehickness control is obtained simply by counting the number ofeposition cycles.

∗ Corresponding authors at: Chimie ParisTech, Laboratoire de Physico-Chimie desurfaces (LPCS), 11 rue Pierre et Marie Curie, F-75005 Paris, France.∗∗ Corresponding author.

E-mail addresses: [email protected] (V. Maurice),[email protected] (M. Ritala), [email protected]

P. Marcus).

013-4686/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.oi:10.1016/j.electacta.2011.07.104

f the alumina film on the elsewhere still coated surface matrix.© 2011 Elsevier Ltd. All rights reserved.

ALD oxide coatings for corrosion protection were first con-sidered in 1999 with Al2O3, TiO2 and Ta2O5 single layer andAl2O3–TiO2 nanolaminate coatings ranging from 170 to 500 nm inthickness on stainless steel [4]. Later, ALD TiO2 and Al2O3 films werealso grown on PVD coatings to block pinholes and other defectsleft in the structure [5,6]. An increase of corrosion potential and aone order of magnitude decrease of corrosion current density wereachieved. Similar results were also obtained with bare 50 nm ALDTiO2 coatings deposited directly on stainless steel [7]. One advan-tage of ceramic oxide coatings is that they can also offer a betterwear protection in comparison to metallic or organic counterpartsdue to higher hardness and strength and with the advantage oflower thickness [8].

Quite recently, we have reported on 316L stainless steel coatedwith ALD Al2O3 and Ta2O5 single layers [9] and carbon steel(AISI 52100) coated with ALD Al2O3, Ta2O5 single layers andAl2O3–Ta2O5 nanolaminates [10,11]. It was shown that 50 nmAl2O3 deposited at 250 ◦C could decrease the corrosion current den-sity of stainless steel by four orders of magnitude, a more moderateperformance being achieved with 50 nm Ta2O5 owing to higher

residual carbon contamination from the Ta organometallic precur-sor [9]. Al2O3 coatings were also grown at 160 ◦C on stainless steel,but the sealing performance was not as good as with the coatingsprepared at 250 ◦C also because of higher residual contamination
Page 2: Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

9 ica Ac

fdtsoctmdArd

AbAwtbsAsm[tfiIoesAn

2

2

pewtf

Rwdtowa

2

(aeat

TC

610 B. Díaz et al. / Electrochim

rom the organometallic and water precursors. On carbon steel theeposition temperature was limited to 160 ◦C in order to maintainhe properties achieved through hardening and tempering the sub-trate material. A larger than three orders of magnitude decreasef the corrosion current density was achieved with a 100 nm Al2O3oating [10]. Ta2O5 films again had slightly inferior sealing proper-ies as on stainless steel, and Al2O3–Ta2O5 nanolaminates had inter-

ediate properties [11]. Plasma-enhanced ALD is an alternative foreposition at low temperature on heat sensitive engineering alloys.

decrease of over three orders of magnitude of the corrosion cur-ent density was also reported recently with 50 nm Al2O3 coatingseposited at 150 ◦C on the same carbon steel [12].

The present article aims to assess the stability of thin (50 nm)l2O3 coatings grown by ALD for corrosion protection of car-on steel (AISI 52100) and to discuss their failure mechanism.lumina was chosen as a coating material because it nucleatesell on challenging engineering substrate materials when using

rimethyl aluminium as a precursor [13–16] and because it cane deposited at the low temperatures demanded by the carbonteel with excellent sealing properties [9–12]. The structure ofLD Al2O3 films is amorphous [12,17–19], and thus free of defectsuch as grain boundaries that can limit the barrier property. Aain problem of Al2O3 is its solubility in acidic and basic media

20]. However, it is expected to be stable at neutral pH accordingo the Pourbaix diagram [21], which is confirmed for ALD Al2O3lms grown on silicon substrates as will be reported separately.

n this study electrochemical impedance spectroscopy (EIS), time-f-flight secondary ion mass spectrometry (ToF-SIMS) and fieldmission scanning electron microscopy/energy dispersive X-raypectroscopy (FESEM/EDS) were combined in order to study theLD Al2O3 film deposited on carbon steel in a chloride-containingeutral environment.

. Experimental

.1. Sample preparation

The nominal chemical composition of the hardened and tem-ered low alloy carbon steel (AISI 52100, DIN 100Cr6) substratemployed in this study is given in Table 1. The surface was lappedith a water based diamond suspension (∼6 �m). Prior to coating

he substrates were carefully wiped with acetone, ultrasonicatedor 5 min in isopropanol, and blow-dried with compressed air.

The Al2O3 coating process was carried out in a Picosun SUNALE-150 ALD reactor using trimethyl aluminium (TMA, Al(CH3)3) andater as precursors. The deposition procedure, described in moreetails previously [9,10], was performed at 160 ◦C. The coatinghicknesses were measured from a silicon wafer coated simultane-usly with the steel substrates. The measurements were conductedith X-ray reflectance spectroscopy (XRR, Bruker AXS D8 Advance)

nd modelled with Leptos 7.05.

.2. Corrosion tests and electrochemical impedance spectroscopy

The corrosion tests were conducted at open circuit potentialOCP) in a conventional three-electrode cell. The working electrode

rea was 0.44 cm2 delimited by a Viton O-ring. A saturated calomellectrode (SCE) and a platinum wire were employed as referencend counter electrodes, respectively. A 0.2 M NaCl aqueous solu-ion (pH 7) prepared with ultra pure water (resistivity > 18 M� cm)

able 1hemical composition (wt%) of the carbon steel substrate (DIN 100Cr6, AISI 5210) used in

C Si Mn P S

0.9–1.05 0.15–0.35 0.25–0.45 Max. 0.03 Max. 0

ta 56 (2011) 9609– 9618

and reagent grade chemicals (NaCl Analar Normapur analyticalreagent, VWR® BDH Prolabo®) was used as electrolyte. The exper-iments were conducted at room temperature and the electrolytewas bubbled with Ar for 30 min before the test and during theexperiment.

The impedance measurements were conducted during the cor-rosion tests. The OCP value was registered for 50 min and then theimpedance spectra were measured for the next 10 min to completethe first hour of immersion. The procedure was repeated 5 timesduring a total immersion period of 6 h. A potentiostat/galvanostatAutolab 30 was used for the measurements. Frequencies between100 kHz and 10 mHz were used with an amplitude signal set to10 mV to guarantee a linear response. All coated samples werecleaned with ethanol for 10 min in an ultrasonic bath and driedwith compressed air prior to the tests. For ToF-SIMS and FESEM/EDSanalysis, the corrosion tests were stopped after 1, 3 and 6 h. Beforesurface analysis samples were rinsed in ethanol and blow-dried.

2.3. Time-of-flight secondary ion mass spectrometry

A ToF-SIMS 5 spectrometer (IonTof) operating at a pressureof 10−9 mbar was used for depth profiling and chemical map-ping. For elemental depth profiling the spectrometer was run inthe HC-BUNCHED mode with optimum mass resolution but poorlateral space resolution. A pulsed 25 keV Bi+ primary ion sourcewas employed for the analysis, delivering ∼1 pA of current over a100 �m × 100 �m area. It was interlaced with a 2 keV sputtering Cs+

beam giving a ∼82 nA target current over a 400 �m × 400 �m area.The profiles were recorded with negative secondary ions, moresensitive to fragments originating from oxide matrices.

For surface chemical mapping, the procedure included a fewseconds of sputtering in order to remove the contamination layerexisting at the outermost surface. Negative ion chemical mapswere then recorded in low primary ion current conditions (non-etching static SIMS conditions) at optimal lateral space resolution(∼150 nm) but with poorer mass resolution than for depth pro-filing. A pulsed 25 keV Bi+ primary ion source was employed foranalysis, delivering ∼0.2 pA of current over an analyzed area of20 �m × 20 �m or 100 �m × 100 �m.

For both depth profiling and chemical mapping, the opticalvideo camera of the spectrometer was used to select analysis fieldsof view that appeared homogeneous. Data acquisition and post-processing analyses were performed using the Ion-Spec software.

2.4. Field emission scanning electron microscopy and energydispersive X-ray spectroscopy

A FESEM microscope (Hitachi S-4800) was used for microstruc-tural imaging. The microscope was operated in combinedsecondary and backscattered electron mode at accelerating volt-ages of the primary beam of 20 kV. For EDS analysis, multiple spotanalysis of areas less than 2 �m in diameter was performed with aprimary beam of 20 kV.

3. Results and discussion

3.1. Pristine coated sample

Fig. 1 shows the ToF-SIMS depth profile for the carbon steelalloy covered by the pristine 50 nm ALD Al2O3 coating. The selected

this study.

Cr Ni Cu Fe

.025 1.35–1.65 Max. 0.3 Max. 0.3 Balance

Page 3: Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

B. Díaz et al. / Electrochimica Ac

0 100 20 0 30 0 400 50 0 60 0 700 80 0100

101

102

103

104

105

Inte

nsity

(cou

nts)

Sputt eri ng ti me / s

C-

O-

OH

18O-Al-

CAl-

Cr-

Fe-

AlO2-

CrO2-

FeO2-

Cl-

coat ing interface substrate

Fi

i5

osoCd

mavStdsficitmactco

bTc0te

Fciadeeti

ig. 1. ToF-SIMS negative ions depth profiles for the pristine 50 nm ALD Al2O3 coat-ng grown at 160 ◦C on the 100Cr6 carbon steel substrate.

ons were 16O−, 59AlO2−, 17OH−, 27Al−, 18O−, 35Cl−, 12C−, 39CAl−,

6Fe−, 52Cr−, 88FeO2− and 84CrO2

−. 18O is the naturally occurringxygen isotope also recorded since the 16O− signal was close toaturation. Their intensities are presented in logarithmic scale inrder to emphasize the low intensity signals and plotted versuss+ sputtering time. Three regions, marked on the profile, are wellistinguished.

Starting from the outermost surface, the coating region is wellarked by the stable and parallel intensities of the AlO2

−, Al−, 16O−

nd 18O− ions. This confirms the growth of a film with no in-depthariation of stoichiometry in agreement with previous results [10].ome organic and OH contamination is evidenced in this region byhe C−, CAl− and OH− ions profiles, also confirming our previousata. Peaking at the outermost coating surface because of expo-ure to air, this contamination decreases to its residual level after aew seconds of sputtering. Afterwards, the OH contamination, orig-nating from the water precursor, is stable in the bulk coating. Inontrast, the C− ion intensity continuously increases before reach-ng the interfacial region. The CAl− ion profile is parallel showinghat carbon contamination results from incomplete removal of the

ethyl ligands of the TMA precursor [9,10,22]. Peaks in the OH−

nd CAl− ion profiles are well-marked in the interfacial region indi-ating that unreacted precursor fragments are more prominent inhe initial stages of growth. A marked influence of environmentalontamination present at the substrate surface in the initial stagesf growth cannot be excluded.

Chloride residual contamination of the coating is also evidencedy Fig. 1. Its origin is mostly related to trace contamination of theMA precursor (0–0.1 % content in Cl). The Cl concentration in theoating was found previously to be below the detection limit of.5 at% of XPS measurements [10]. Cl contamination also peaks athe outermost surface due to additional contamination from ambi-nt air.

After about 310 s of sputtering, the intensities of the substratee− and Cr− ions start to increase defining the onset of the interfa-ial region where the intensities of the AlO2

−, Al−, 16O− and 18O−

ons decrease. As previously discussed [9,10,12], the presence ofn interfacial spurious oxide containing iron and chromium is evi-enced by the peaks in the FeO2

− and CrO2− ions profiles. The

xposure of the uncoated metallic substrate to water during thearly stages of the ALD process, i.e. before complete covering byhe first alumina layers, possibly contributes to the growth of thisnterfacial layer in addition to the native oxide previously formed

ta 56 (2011) 9609– 9618 9611

in air. Although the chemical composition cannot be extracted fromthe ToF-SIMS data, one expects this interfacial oxide to contain amixture of iron, chromium and aluminium ions [22,23]. Our previ-ous XPS study has shown the presence of Fe(II) and Fe(III) speciesunderneath a 10 nm alumina film grown in the same conditions butchromium, most likely present as Cr(III), was below the detectionlimit of the measurement [10]. The end of the interfacial region, setat the intersection point of the AlO2

− and the Fe− ions profiles, isreached after 700 s of sputtering for this sample. The large widthof the interface region is an artefact effect of the rough substratesurface causing shadowing effect of sputtering.

3.2. Chemical modifications of the immersed coated surface

Fig. 2 shows the ToF-SIMS depth profiles measured after 1, 3 and6 h of immersion in the neutral NaCl electrolyte. Besides a strikingdecrease of the width of the coating region discussed below, thedata only show slight chemical modifications of the remaining alu-mina layers and of the interfacial region after immersion in the testsolution. The Cl− ions intensity significantly increases (factor of 2)in the bulk of the oxide coating already after the first hour of immer-sion whereas the composition of the oxide matrix is preserved asjudged from the unchanged intensities of the AlO2

−, Al−, 16O− and18O− ions. This is consistent with the presence of defect channels(cracks, pinholes) in the coating through which the solution andthe aggressive ions can penetrate. A significant increase of the Cl−

peak at the interface, reaching the intensity level of the 18O− ions,is also observed after the first hour of immersion and confirms thatthe electrolyte enters the coating and reaches the interface with thealloy substrate after 1 h of immersion. However no further accumu-lation of chlorides is observed in the bulk coating or at the interfaceafter longer immersion times (3 and 6 h).

The OH− ion intensity significantly decreases (factor of 2) inthe bulk of the alumina coating after the first immersion step (1 h)but no further afterwards. This variation combined with the chlo-ride entry suggests a substitution of hydroxide groups by chloridesin the bulk coating, a typical mechanism proposed for the passiveoxide film breakdown and repair poisoning on passivable metallicsubstrates [24–27]. Substitution is thought to preferentially takeplace at the surface of the walls in the cracks/pinholes throughwhich the solution penetrates. It appears homogeneous in-depthsince no variation of the OH− and Cl− ion profiles is observed in thebulk coating region.

The C− ion profiles are superimposed in Fig. 3. The effect of thecoating thickness decrease has been corrected by shifting the timeaxis. The intensities in the bulk of the remaining coating layers arealmost invariable during immersion, showing the absence of disso-lution of the organic contaminants of the bulk coating despite thepenetration of the electrolyte. This is possibly because most of theC contamination of the bulk coating is trapped in the oxide matrixnot exposed to the electrolyte. At the onset of the coating/substrateinterface, the increase of intensity between the oxide matrix leveland the substrate level becomes steeper after immersion and theinterfacial region appears narrower. This variation is not real butassigned to the artefact caused by the roughness related shadow-ing effects of sputtering. This artefact progressively vanishes withdecreasing thickness of the coating.

The interfacial oxide is only slightly modified after the immer-sion test. The intensities of the FeO2

− and the CrO2− ions

barely increase after immersion. The FeO2−:CrO2

− intensity ratioincreases from ≈8.5 for the pristine sample to ≈9.6 after 6 h ofimmersion showing a slight variation in the composition of the

interfacial oxide. This slight iron enrichment of the oxide is anindication of some accumulation of iron corrosion products in theinterfacial region, presumably at the defective sites of the coatingexposing the substrate.
Page 4: Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

9612 B. Díaz et al. / Electrochimica Acta 56 (2011) 9609– 9618

0 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0100

101

102

103

104

105

Inte

nsity

(cou

nts)

Sputt ering t ime / s

C-

O-

OH-18O-

Al-

CAl-

Cr-

Fe-

AlO2-

CrO2-

FeO2-

coa ting interface sub strate

Cl-

1 hou r

0 10 0 20 0 30 0 40 0 50 0 60 0 70 0100

101

102

103

104

105

3 hours

Inte

nsity

(cou

nts)

Sputt ering t ime / s

C-

O-

OH-18O-

Al-

CAl-

Cr-

Fe-

AlO2-

CrO2-

FeO2-

coa ting interface sub strate

Cl-

0 50 10 0 15 0 20 0 25 0 30 0 35 0 40 0 45 0100

101

102

103

104

105

6 hours

Inte

nsity

(cou

nts)

Sputt ering t ime / s

interface sub stratecoa ting

C-

O-

OH-

18O-Al-

CAl-

Cr-

Fe-

AlO2-

CrO2-

FeO2-

Cl-

FaN

3

3sih

0 100 200 300 400 500 600

102

103

104

6 h

3 h

0 h

Inte

nsity

(cou

nts)

Sputtering time / s

1 h

resistance (Re) and the charge transfer resistance (Rct), respec-tively. Constant phase elements (CPEs) were used instead of purecapacitances for an improved data fit. CPEs take into account the

ig. 2. ToF-SIMS negative ions depth profiles for the 50 nm ALD Al2O3 coating grownt 160 ◦C on the 100Cr6 carbon steel substrate after 1, 3 and 6 h of immersion in 0.2 MaCl.

.3. Porosity increase and coating dissolution

Fig. 4(A) presents the EIS data (Bode plots) recorded at t = 1, 2,

, 4, 5 and 6 h of immersion of the 50 nm ALD Al2O3 coated carbonteel at OCP in the neutral NaCl solution. The uncoated substrates also presented for comparison. The impedance at t = 1 h is veryigh in comparison to that of the bare substrate confirming the

Fig. 3. ToF-SIMS depth profiles for the C− ions before and after immersion in 0.2 MNaCl.

improved corrosion resistance provided by the 50 nm ALD Al2O3film [10,12]. For the coated samples a continuous decrease of theglobal resistance, more marked at lower frequencies, is observedwith increasing immersion time as well as a drop of the phase anglein the high frequency region.

The equivalent circuits presented in Fig. 4(B) were used for fit-ting the spectra. For the coated samples, it is an adaptation of theequivalent circuit previously discussed by Bonnel et al. [28]. Thelimit at high and low frequencies corresponds to the electrolyte

Fig. 4. (A) Bode plots obtained during immersion in 0.2 M NaCl; (B) equivalentcircuits employed for fitting the impedance data.

Page 5: Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

ica Acta 56 (2011) 9609– 9618 9613

dNC

Z

wftbdt(atoot

iiif(pctc

csucof[tswFtfeva

CAcaeantt

A

6543210,1

1

10

Por

osity

(%)

Immersion time / h

porosity (Rct) porosity (CPE-Cdl)

B

10 5432 60

10

20

30

40

50

thickness (ToF-SIMS ) linear fit (y=55.67-8.06x) thickness (EIS) linear fit (y=48.77-6.17x)

Thic

knes

s / n

m

Immersion time / h

Fig. 5. Porosity calculated from the EIS data (A) and thickness calculated from theEIS (discs) and ToF-SIMS (squares) data (B) versus immersion time in 0.2 M NaCl of

TE

B. Díaz et al. / Electrochim

eviations from an ideal capacitance behaviour manifested on theyquist plot by a depressed semicircle [29]. The impedance of thePE is defined by Eq. (1),

CPE = 1Q (jω)n (1)

here Q is a value, independent of the frequency, obtained directlyrom the fitting, and the factor n is the CPE power, which is related tohe angle ((1 − n) × 90) that evaluates the Nyquist plot depressionelow the x-axis. The CPE-Ccoat and the CPE-Cdl elements in Fig. 4(B)escribe the coating capacitance and double layer capacitance athe electrolyte/substrate interface, respectively. A third resistanceRredox) has been included to account for the deviation of the phasengle observed in the high frequency domain. It can be assignedo the reactions, oxygen cathodic reduction coupled to iron anodicxidation [28], taking place on the substrate surface at the bottomf pinholes as a result of the penetration of the electrolyte throughhe coating.

For the bare substrate, a different equivalent circuit, also shownn Fig. 4(B), was considered. Besides the electrolyte resistance, itncludes the charge transfer resistance and the double layer capac-tance. A second time constant could be distinguished in the lowestrequency range. The fitting values for this second time constantnot presented here) correspond to the air-formed native oxideresent on the surface as reported in the literature [30]. For theoated substrates, this time constant could not be resolved due tohe small uncoated surface fraction exposed to the electrolyte asonfirmed below.

The values of the fitting parameters are presented in Table 2. Theharge transfer resistance (Rct) decreases with increasing immer-ion time. As previously discussed [10], the coating porosity P (orncoated substrate surface fraction) can be reliably measured byomparing the Rct values of the coated samples with the Rct valuef the bare substrate. The coating porosity can also be extractedrom the double layer capacitance values as previously described10]. This previous work showed the good agreement obtained withhe two procedures for as-deposited coatings after 0.5 h of immer-ion necessary to reach stable potential. Here the P values obtainedith the two procedures are plotted versus the immersion time in

ig. 5(A). A very good agreement is obtained and the data are quan-itatively consistent to show an increase of the uncoated surfaceraction with increasing immersion time of the carbon steel cov-red by the ALD Al2O3 layer. Consistently, the variation the Rredoxalues in Table 2 indicates a decreasing hindrance of the reactionst the uncoated sites with increasing immersion time.

The real coating capacitance values were estimated from thePE values reported in Table 2 using the Brug equation [9,10,31,32].ssuming a flat parallel capacitor, the coating thickness can be cal-ulated from the real capacitance if one knows the coated surfacerea and permittivity. The uncoated and coated surface areas werextracted from the porosity P knowing the electrode working area

nd the values are presented in Table 3 together with the thick-ess values deduced from the real capacitances. Fig. 5(B) shows thehickness variation with immersion time. It shows the decrease ofhe coating thickness during the immersion test in the neutral 0.2 M

able 2IS fitting parameters obtained using the equivalent circuits in Fig. 4(B).

OCP (mV) Re (�) Rct (�) Z(CPEcoat) (�−1

Bare −672 91.9 3.31E+04

1 h −635 85.1 9.84E+06 8.37E−08

2 h −634 85.1 7.15E+06 1.17E−07

3 h −645 84.3 6.37E+06 1.43E−07

4 h −650 80.8 4.56E+06 2.53E−07

5 h −667 81.6 2.23E+06 5.75E−07

6 h −730 77.45 1.85E+05 4.58E−06

the 50 nm ALD Al2O3 coating grown at 160 ◦C on the 100Cr6 carbon steel substrate.The linear fits in (B), from which the dissolution rate is extracted, is indicated by thedashed lines.

NaCl solution and thus the dissolution of the coating covering thecarbon steel.

The ToF-SIMS depth profiles in Fig. 2 also show a continuousdecrease of the sputtering time (i.e. depth) of the coating regionwith increasing immersion time, confirming the decrease of thecoating thickness measured by EIS and thus the dissolution of theALD Al2O3 layer. Using the 50 nm nominal thickness value of thepristine coating and assuming no change of the sputtering yieldafter immersion, the coating sputtering time was converted intothickness (Table 3), and the values measured by ToF-SIMS were

superimposed to those obtained by EIS in Fig. 5(B). The good agree-ment between the two sets of values indicates that the equivalentcircuit selected for the EIS data fit is appropriate to describe thevariation of the coating thickness with immersion time. A dissolu-

sn) n Rredox (�) Z(CPEdl) (�−1 sn) n

2.96E−05 0.791.00 2.27E+03 5.46E−08 0.790.99 1.56E+03 7.35E−08 0.810.98 8.86E+02 8.92E−08 0.840.94 3.87E+02 7.09E−08 0.890.90 1.26E+02 4.77E−07 0.850.75 5.47E+01 7.68E−06 0.94

Page 6: Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

9614 B. Díaz et al. / Electrochimica Acta 56 (2011) 9609– 9618

Table 3Coated area (sEIS

coat), uncoated area (sEISuncoat), coating thickness (dEIS

coat and dToF-SIMScoat ) and coating sputtering time (s · tToF-SIMS

coat ) for the 50 nm ALD Al2O3/100Cr6 sample as obtainedby EIS during immersion in 0.2 M NaCl at OCP and by ToF-SIMS after immersion.

Immersion time (h) sEISuncoat × 103 (cm2) sEIS

coat × 103 (cm2) dEIScoat (nm) s · tToF-SIMS

coat (s) dToF-SIMScoat (nm)

0 – – – 310 501 1.48 440.1 42.33 297 47.92 2.04 439.6 34.05 – –3 2.29 439.3 31.20 192 31

27.78 – –18.45 – –

9.17 46 7.5

tsi

toaamtn

tiiFtiim[reTsvmt

0 10 20 30 40 50

1

10 ToF-S IMSEIS

Por

osity

(%)

Al2O

3 thickness / nm

4 3.20 438.4

5 6.56 435.0

6 7.89 462.6

ion rate of 7 ± 1 nm h−1 is obtained from a linear fit of the two dataets and the lifetime of the 50 nm alumina coating on carbon steels ∼7 h in these test conditions.

The data in Fig. 5(B) also indicate a good agreement betweenhe EIS analysis performed over the macroscopic working areaf the electrode (44 mm2) and ToF-SIMS analysis performed over

microscopic area (0.01 mm2). This shows that the microscopicrea selected for ToF-SIMS analysis is well-representative of theacroscopic electrode. However, possible local effects of preferen-

ial failure of the coating cannot be excluded as discussed in theext section.

The variation of the coating porosity as a function of the coatinghickness was calculated by combining the porosity data shownn Fig. 5(A) and the linear fits of the thickness variation withmmersion time shown in Fig. 5(B). The result is presented inig. 6. The graph shows that the remaining layers, those closero the alloy substrate, are more porous as demonstrated by thencrease of the coating porosity when thickness is reduced. Thiss in agreement with the previously discussed formation of a

ore defective and porous layer in the initial stages of growth9,10,12,33]. It is also possible that defects are induced in theemaining film by the chloride containing solution and/or thexisting pores are enlarged due to some preferential dissolution.he more defective character of the remaining layers is also

upported by the CPE-Ccoat power value, n, reported Table 2. Thisalue, associated with the homogeneity of the coated surface, isarkedly lower than 1 after 6 h of immersion in agreement with

he formation of a more heterogeneous surface when the coating

Fig. 7. ToF-SIMS negative ion chemical maps for the ALD Al2O3 coating grown at 16

Fig. 6. ALD Al2O3 coating porosity versus coating thickness obtained from the EISand ToF-SIMS data.

is markedly reduced in thickness. With increasing thickness theALD layers become more and more compact and uniform as shown

by Fig. 6 and the CPE-Ccoat n power value in Table 2 [33].

Observing the dissolution of the alumina coating was somehowunexpected since the dissolution is usually not predicted at neutralpH [20,21]. Besides it is in contrast to the long term stability (over

0 ◦C on the 100Cr6 carbon steel substrate after 6 h of immersion in 0.2 M NaCl.

Page 7: Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

ica Ac

1iroe

f

A

toehe

O

Ouas[al

ctBosrbso[cpts

tbidoTtufo

twcf

3

owsroma

intensity assigned to scratches resulting from the surface prepara-tion. These scratches are still covered by the ALD layer as shown bythe AlO− and AlO2

− maps, showing no coating defect in these sitesat the space resolution of the measurement. However, it cannot

B. Díaz et al. / Electrochim

800 h) observed in a neutral 0.5 M NaCl solution of ALD Al2O3 coat-ngs grown in the same conditions on a silicon substrate (detailedesults will be reported separately). This difference points to a rolef the carbon steel substrate on the failure mechanism despite thexcellent sealing property of the coating.

Alumina dissolution in neutral water proceeds according to theollowing reaction:

l2O3 + 2OH− + H2O → 2AlO2− + 2H2O (2)

hat requires hydroxide groups and thus a pH increase. The stabilitybserved in the case of the alumina coating on silicon in 0.5 M NaClxcludes a major effect of the chloride ions. An obvious origin forydroxide ions is the cathodic reduction of oxygen dissolved in thelectrolyte according to:

2 + 2H2O + 4e− → 4OH− (3)

n carbon steel, this reaction most likely takes places on thencoated surface exposed to the electrolyte by the coating defects,s supported by the quite similar cathodic branches of the polari-ation curves previously reported for coated and uncoated samples10]. On silicon blocking of the cathodic oxygen reduction by thebsence of n-type doping may be a major factor preventing disso-ution.

The anodic reactions that balance oxygen reduction can pro-eed at the outermost surface of the coating but this is unlikely inhe present case because of the insulating properties of alumina.esides the alumina film thickness is far beyond the limit of therder of a few nanometres for tunnelling of electrons from the filmurface to the substrate. A much more likely location for anodiceactions is then the uncoated surface exposed to the electrolytey the coating defects. There, anodic dissolution of the carbon steelurface can take place, which is supported by the similar shapef the anodic branches recorded for coated and uncoated samples10]. The absence of a significant accumulation of iron or chromiumorrosion products at the coating/alloy observed by ToF-SIMS depthrofiling is consistent with their release in the electrolyte and withhe i–E curves that show no sign of passivation of the uncoatedurface.

Another requirement for the proposed mechanism is related tohe generalized character of the alumina dissolution as observedy ToF-SIMS and EIS. If produced locally at the bottom of the coat-

ng defects exposing the substrate surface, hydroxide ions mustiffuse in the electrolyte to the coating surface and access theverall alumina surface to trigger an overall dissolution reaction.his supposes a relatively high density of coating defects exposinghe substrate and thus small dimensions of the defects since thencoated surface fraction is low. The increase of the exposed sur-ace observed with ongoing dissolution would then promote theverall character of the reaction.

This proposed failure mechanism does not exclude that the reac-ion may proceed locally faster, in particular on substrate siteshere the coating defects exposing the carbon steel can be more

oncentrated and/or larger (i.e. locally higher uncoated surfaceraction).

.4. Local failure

Fig. 7 shows the ToF-SIMS chemical maps obtained with a fieldf view of 20 �m × 20 �m after 6 h of immersion. The same resultsere obtained with larger fields of view (100 �m × 100 �m) and

imilar observations were made after 1 and 3 h of immersion. As a

esult of the static SIMS conditions of the measurement, in whichnly secondary ions originating from the outermost surface areeasured, the most intense maps are those of the O−, OH−, AlO−

nd AlO2− ions of the coating. The distribution in these maps is

ta 56 (2011) 9609– 9618 9615

very homogeneous with no defects identified at the space resolu-tion limit of the measurement (∼150 nm). This shows within thislimit that the substrate surface is still uniformly covered by theremaining Al2O3 layer in the areas selected for this analysis. Theabsence of pinholes exposing the interfacial region is confirmed bythe absence of intensity in the FeO− and CrO− ions maps. The Cl−

ion map shows a low but non zero background intensity. Its uni-formity indicates that the chloride contamination is homogeneousat the space resolution limit of the measurement.

The hydroxide ion maps reveal some lines of slightly higher

Fig. 8. FESEM images of the 50 nm ALD Al2O3 coated carbon steel after (A, C) 1 and(B) 3 h of immersion in 0.2 M NaCl. Lighter and darker areas in (A, B) show coatinglift up and coating removal, respectively. Coating removal is magnified in (C).

Page 8: Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

9616 B. Díaz et al. / Electrochimica Acta 56 (2011) 9609– 9618

FAg

bi

spima(eicicfttcep1afdpsucitdsp

rm

ig. 9. FESEM/EDS analysis of coating removal at surface scratch of the 50 nm ALDl2O3 coated carbon steel after 3 h of immersion in 0.2 M NaCl. Concentration isiven in wt%.

e excluded that failure preferentially occurs at these sites duringmmersion as shown below.

FESEM analysis after immersion revealed three types of defectshown in Figs. 8–10 that were not present on the as-deposited sam-les. One type, consistent with coating lift up (i.e. de-adhesion), is

llustrated by the lighter grey areas in Fig. 8(A) and (B). Its develop-ent occurs preferentially along the scratches after 1 h (Fig. 8(A))

nd then extends over the surface with increasing immersion timeFig. 8(B)). After 3 h of immersion (Fig. 8(B)) this type of defect cov-rs a major fraction of the surface, so that it is necessarily includedn the areas selected for ToF-SIMS analysis. A possible reason foroating lift up is the accumulation of hydroxide products and/orron corrosion products causing de-adhesion as observed in filiformorrosion [34–41]. This would take place at the coating/alloy inter-ace following the penetration of the solution at coating defects andhe ensuing coupling of the cathodic and anodic reactions. Howeverhe ToF-SIMS data do not show any significant increase of the OH−

ontent at the interface after 3 h of immersion nor a pronouncednrichment in iron corrosion products, thus not supporting thisossibility. Cl− penetration down to the interface is observed after

h but it does not increase further after 3 h of immersion. Thisbsence of significant chemical changes of the coating/alloy inter-ace with increasing immersion time allows excluding a coatinge-adhesion caused by the accumulation and ingress of corrosionroducts. However the observed active behaviour of the uncoatedurface [10] is consistent with the release of the corrosion prod-cts. This would lead to trenching of the alloy surface below theoating with the possible consequence of the lift up of the remain-ng coating disbonded from the steel. The FESEM data shows thathis would occur preferentially at surface scratches where coatingefects can be expected to expose a higher surface fraction of theteel because of the high aspect ratio of the topography and the

resence of residual mechanical stress at the substrate surface.

Another type of defect observed after immersion is coatingemoval illustrated by the darker areas in Fig. 8(A) and (B) andagnified in Fig. 8(C). The EDS analysis in Fig. 9 evidences the

Fig. 10. FESEM secondary electron images showing local sites of corrosion on the50 nm ALD Al2O3 coated carbon steel surface after (A) 1 and (B) 3 and (C) 6 h ofimmersion in 0.2 M NaCl, respectively.

substantial decrease of the Al and O content in these darker areascompared to the coated matrix. The concentration of aluminiumis at the detection limit of EDS suggesting that some alumina lay-ers may remain. Fig. 8 also shows that, like lift up, coating removalpreferentially occurs at surface scratches and develops with time,although to a much lower extent since only covering a small surfacefraction after 3 h of immersion. Unlike for lift up, this developmentis not sufficiently large to be included in significant proportion inthe areas selected for ToF-SIMS analysis after 3 h of immersion. Ifso, the ToF-SIMS depth profiles would have included, in the coating

regions, a plateau of the Fe− and Cr− ions intensities, which was notobserved (Fig. 2). However, it is possible that the defects revealed inthe ToF-SIMS OH− maps correspond to these coating removal sites.It is proposed that these areas of coating removal correspond to
Page 9: Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

ica Ac

spcfitt

cs6or∼iAtisai(ibtwtrti

4

5si

aaboDcdsiictoiheitia

oAftmbf

[

[

[

[[

[[

[

[

[[

[

[[

[[[[

[

[

[

[

B. Díaz et al. / Electrochim

ites of preferential dissolution of the coating. The reaction wouldropagate faster along and in the vicinity of surface scratches whereracks/pinholes of the coating are expected to expose a higher sur-ace fraction thus increasing locally the production of hydroxideons required to trigger alumina dissolution. It cannot be excludedhat coating removal also results from the rupture and release ofhe coating after disbonding from the trenched alloy surface.

The third type of defects identified by FESEM after immersionorresponds to corrosion spots (Fig. 10). These corrosion spots werecarce after 1 h of immersion and became more frequent after

h. However, large areas without corrosion spots could still bebserved after 6 h of immersion. The lateral dimensions of the cor-osion spots increased from ∼4 �m after 1 h of immersion up to20 �m after 6 h of immersion. Fig. 10(A) and (B) shows two typ-

cal examples observed after 1 and 3 h of immersion, respectively.lthough the corrosion of the alloy has been initiated (it marks

he corrosion spot), propagation in the alloy appears quite limitedn both cases since the polishing grooves marking the substrateurface can be seen still crossing the corrosion spot. In contrast,fter 6 h of immersion (Fig. 10(C)), the alloy is clearly pitted show-ng in-depth propagation of the corrosion process. EDS analysisnot shown) confirms coating removal combined with a markedncrease of the C and O content inside the pit. Each pit is surroundedy a circular region where the steel surface subsists as marked byhe polishing scratches but where the coating has been removedithout extensive corrosion of the substrate. Most likely the reac-

ion of cathodic reduction of oxygen takes place in this surroundingegion (balancing the anodic reactions of oxidation of the alloy inhe pit). Further away from the pit, the surface matrix is still coatedn agreement with the EIS and ToF-SIMS data.

. Conclusions

This study shows that, in spite of excellent sealing properties,0 nm thick pure Al2O3 layers grown by ALD at 160 ◦C on carbonteel fail to provide durable corrosion protection to the substraten neutral solution containing chloride.

Combining EIS and ToF-SIMS analysis, it was observed that thelumina coating dissolves in 0.2 M NaCl aqueous solution at an aver-ge rate of 7 nm h−1 at room temperature. The remaining layersecome increasingly porous as a result of the detrimental effectf the thickness decrease on the sealing property of the coating.efects are possibly induced in the remaining film by the chlorideontaining solution and/or the existing pores could be enlargedue to some preferential dissolution. Penetration of the chlorideolution was evidenced by ToF-SIMS depth profiling that showedn-depth homogeneous substitution of hydroxides by chloridesn the coating after immersion. It is assigned to the presence ofracks/pinholes connecting the substrate surface to the bulk elec-rolyte through the coating. The cathodic reduction of dissolvedxygen can take place at the steel surface exposed by the coat-ng defects, raising the pH and triggering alumina dissolution. Aigh density of coating defects appears necessary to generate gen-ralized dissolution of the coating. The balancing anodic reactions concluded to be the dissolution of the steel also taking place athe surface exposed by the coating defects. ToF-SIMS depth profil-ng confirms no marked accumulation of corrosion products of thelloy undergoing active dissolution at the coating defects.

Preferential failure of the coating was observed by FESEM toccur along and in the vicinity of substrate polishing scratches.t such sites cracks/pinholes would expose locally a larger sur-

ace fraction of the steel because of the higher aspect ratio of the

opography and of localized residual mechanical stress, thus pro-

oting the mechanism of failure. Local failure was characterizedy coating lift up assigned to de-adhesion of the coating disbondedrom the steel surface by the ingress of anodic dissolution from

[

[[

ta 56 (2011) 9609– 9618 9617

the coating cracks/pinholes along the interface. Local failure alsoresulted in the loss of the coating by faster dissolution and/or rup-ture after disbonding from the trenched steel surface. Localizedcorrosion of the steel was also observed by FESEM to be triggeredwith increasing immersion leading to pitting prior to complete dis-solution of the alumina film on the elsewhere still coated surfacematrix. The pits were surrounded by uncoated circular areas wherecathodic reactions are thought to proceed to balance anodic pitgrowth.

Acknowledgements

The research leading to these results has received fundingfrom the European Community’s Seventh Framework Programme(FP7/2007-2013) under grant agreement no. CP-FP 213996-1 (COR-RAL). Region Ile-de-France is acknowledged for partial support forthe ToF-SIMS equipment.

References

[1] M. Ritala, J. Niinistö, Atomic layer deposition, in: A.C. Jones, M.L. Hitchman(Eds.), Chemical Vapor Deposition, Precursors, Processes and Applications,Royal Society of Chemistry, 2009, p. 158 (Chapter 4).

[2] K.L. Choy, ECS Trans. 25 (2009) 59.[3] M. Ritala, M. Leskelä, in: H.S. Nalwa (Ed.), Handbook of Thin Film Materials, vol.

1, Academic Press, San Diego, 2001, p. 103 (Chapter 2).[4] R. Matero, M. Ritala, M. Leskelä, T. Salo, J. Aromaa, O. Forsén, J. Phys. IV Fr. 9

(1999) 493.[5] C.X. Shan, X. Hou, K.L. Choy, P. Choquet, Surf. Coat. Technol. 202 (2008) 2147.[6] E. Marin, L. Guzman, A. Lanzutti, L. Fedrizzi, M. Saikkonen, Electrochem. Com-

mun. 11 (2009) 2060.[7] C.X. Shan, X. Hou, K.L. Choy, Surf. Coat. Technol. 202 (2008) 2399.[8] J.B. Watchman, R.A. Haber, in: J.B. Watchman, R.A. Haber (Eds.), Ceramic Films

and Coatings – An Overview, Noyes Publications, 1993, p. 1 (Chapter 1).[9] B. Díaz, J. Swiatowska, V. Maurice, A. Seyeux, B. Normand, E. Härkönen, M. Ritala,

P. Marcus, Electrochim. Acta (2011), doi:10.1016/j.electacta.2011.02.074.10] B. Díaz, E. Härkönen, J. Swiatowska, A. Seyeux, V. Maurice, P. Marcus, M. Ritala,

Corros. Sci. 53 (2011) 2168.11] E. Härkönen, B. Díaz, A. Seyeux, M. Vehkamäki, T. Sajavaara, M. Fenker, J.

Swiatowska, V. Maurice, P. Marcus, Mikko Ritala, J. Electrochem. Soc., submittedfor publication (MS #JES-11-1682).

12] S.E. Potts, L. Schmalz, M. Fenker, B. Díaz, J. Swiatowska, V. Maurice, A. Seyeux,P. Marcus, G. Radnóczi, L. Tóth, M.C.M. van de Sanden, W.M.M. Kessels, J. Elec-trochem. Soc. 158 (2011) C132.

13] M. Kemell, E. Färm, M. Ritala, M. Leskelä, Eur. Polym. J. 44 (2008) 3564.14] M. Kemell, M. Ritala, M. Leskelä, R. Groenen, S. Lindfors, Chem. Vap. Depos. 14

(2008) 347.15] C.A. Wilson, R.K. Grubbs, S.M. George, Chem. Mater. 17 (2005) 5625.16] M.D. Groner, F.H. Fabreguette, J.W. Elam, S.M. George, Chem. Mater. 16 (2004)

639.17] S.J. Yun, K.H. Lee, J. Skarp, H.R. Kim, K.S. Nam, J. Vac. Sci. Technol. A 15 (1997)

2993.18] S. Jakschik, U. Schroeder, T. Hecht, M. Gutsche, H. Seidl, J. Bartha, Thin Solid

Films 425 (2003) 216.19] K. Kukli, J. Ihanus, M. Ritala, M. Leskelä, J. Electrochem. Soc. 144 (1997) 300.20] R.C. Weast, M.J. Astle, W.H. Beyer, CRC Handbook of Chemistry and Physics,

67th edition, CRC Press Inc., U.S.A., 1987, p. B-67.21] M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, Perga-

mon, New York, 1966.22] W.S. Jeon, S. Yang, C.S. Lee, S.W. Kang, J. Electrochem. Soc. 149 (2002) C306.23] K. Tapily, J.E. Jakes, D.S. Stone, P. Shrestha, D. Gu, H. Baumgart, A.A. Elmustafa,

J. Electrochem. Soc. 155 (2008) H545.24] G.S. Frankel, J. Electrochem. Soc. 145 (1998) 2186.25] D.D. Macdonald, Pure Appl. Chem. 71 (1999) 951.26] P. Marcus, H.-H. Strehblow, V. Maurice, Corrosion 50 (2008) 2698.27] H.-H. Strehblow, P. Marcus, Mechanisms of pitting corrosion, in: P. Marcus (Ed.),

Corrosion Mechanisms in Theory and Practice, 3rd edition, CRC Press, Taylorand Francis, 2011.

28] A. Bonnel, F. Dabosi, C. Deslouis, M. Duprat, M. Keddam, B. Tribollet, J. Elec-trochem. Soc. 130 (1983) 753.

29] E. Barsoukov, J.R. Macdonald (Eds.), Impedance Spectroscopy Theory, Experi-ment, and Applications, 2nd edition, John Wiley & Sons, 2005, p. 494.

30] L. Freire, X.R. Nóvoa, F. Montemor, M.J. Carmenzin, Mater. Chem. Phys. 114(2009) 962.

31] G.J. Brug, A.L.G. van den Eeden, M. Sluyters-Rehbach, J.H. Sluyters, J. Electroanal.

Chem. Interf. Electrochem. 176 (1984) 275.

32] V.M.-W. Huang, V. Vivier, M.E. Orazem, N. Pébère, B. Tribollet, J. Electrochem.Soc. 154 (2007) C99.

33] C. Liu, Q. Bi, A. Matthews, Corros. Sci. 43 (2001) 1953.34] R.T. Ruggieri, T.R. Beck, Corros. Sci. 39 (1989) 452.

Page 10: Failure mechanism of thin Al2O3 coatings grown by atomic layer deposition for corrosion protection of carbon steel

9 ica Ac

[[[[

618 B. Díaz et al. / Electrochim

35] A. Bauista, Prog. Org. Coat. 28 (1996) 49.36] W. Schmidt, M. Stratmann, Corros. Sci. 40 (1998) 1441.37] J.H.W. de Wit, Electrochim. Acta 46 (2001) 3641.38] G. Williams, H.N. McMurray, D. Hayman, P.C. Morgan, Phys. Chem. Commun. 6

(2001) 1.

[

[[

ta 56 (2011) 9609– 9618

39] N.Le. Bozec, D. Persson, A. Nazarov, D. Thierry, J. Electrochem. Soc. 149 (2002)B403.

40] P.P. Leblanc, G.S. Frankel, J. Electrochem. Soc. 151 (2004) 105.41] A. Afseth, J.H. Nordlien, G.M. Scamans, K. Nisancioglu, Corros. Sci. 44 (2002)

2529.