Microscopic observations of voids in anodic oxide films on aluminum

Article abstract of DOI:10.1149/1.1753582

The relationship was explored between nanoscale voids in anodic aluminum oxide films and the surface condition of aluminum samples prior to anodizing. Transmission electron microscopy (TEM) detected voids on the order of 10 nm in anodic films. Atomic forc

Full text of DOI:10.1149/1.1753582

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Journal of The Electrochemical Society, 151 ͑7͒ B379-B386 ͑2004͒
B379
0013-4651/2004/151͑7͒/B379/8/$7.00 © The Electrochemical Society, Inc.
Microscopic Observations of Voids in Anodic Oxide Films
on Aluminum
R. Huang,a, K. R. Hebert,
*
a,**,z
and L. S. Chumbleyb
b
aDepartment of Chemical Engineering and Ames Laboratory-U.S. Department of Energy, and
Department of Materials Science and Engineering, Iowa State University, Ames, Iowa 50011, USA
The relationship was explored between nanoscale voids in anodic aluminum oxide films and the surface condition of aluminum
samples prior to anodizing. Transmission electron microscopy ͑TEM͒ detected voids on the order of 10 nm in anodic films. Atomic
force microscopy ͑AFM͒ of these films, obtained after partial oxide dissolution, revealed surface cavities; comparison of TEM and
AFM suggested that the cavities were the oxide voids. AFM images after variable extents of oxide dissolution showed that the
voids were distributed evenly through the inner 60% of the film thickness, indicating that they were formed at the metal-oxide
interface during film growth. Both AFM and TEM showed that the void concentration in the film was sensitive to the extent of
dissolution of the aluminum samples in NaOH prior to anodizing. Positron annihilation spectroscopy had previously detected
voids in samples without anodic films, located in the metal near the oxide-metal interface; the quantity of these interfacial voids
was controlled by NaOH dissolution. The void concentration in the inner part of the anodic films was proportional to the quantity
of these pre-existing interfacial voids. It was inferred that the oxide voids were formed by incorporation, during anodizing, of
interfacial metal voids into the oxide film. The uniform concentration of oxide voids in the inner film suggested that interfacial
metal voids formed continuously during anodizing and that metal voids were generated repeatedly at specific interfacial sites
during film growth.
© 2004 The Electrochemical Society. ͓DOI: 10.1149/1.1753582͔ All rights reserved.
Manuscript submitted March 28, 2003; revised manuscript received January 25, 2004. Available electronically May 19, 2004.
Various surface defects have been proposed to act as sites for
corrosion pit initiation on pure metals, including dislocations, mi-
crosegregated impurities, and flaws in the surface oxide film.1 How-
ever, as yet there is no general agreement on the nature of pit pre-
cursor sites. Recently, positron annihilation spectroscopy ͑PAS͒ has
been used to detect nanometer-scale voids in aluminum metal, lo-
cated within 100 nm of the metal-oxide film interface.2-4 The mea-
surements indicated that the void surfaces were free of oxide, sug-
gesting a high reactivity if exposed during uniform corrosion.
Therefore, the possibility that the interfacial voids act as initiation
sites for pitting corrosion was explored. Atomic force microscopy
͑AFM͒, after chemical stripping of the oxide film in CrO3-H3PO4
solution, revealed surface cavities, the depth and area coverage of
which agreed with PAS measurements of buried voids. Further, the
sites of these cavities corresponded to those of corrosion pits, sug-
gesting that the voids may act as pit precursor sites.3,4 Surface treat-
ments such as NaOH dissolution, which are used in etching appli-
cations to enhance the pit number density, also generated increased
numbers of voids.3 The void number density estimated from AFM
was comparable to that of pits formed during anodic etching.4
Because PAS indicated that the voids are at least of nanometer
size, an attempt was made here to view them by high-resolution
microscopic techniques. The visualization of voids can yield infor-
mation not obtainable with PAS about their size, morphology, and
location relative to surface features such as chemical or structural
inhomogeneities. Direct observations of interfacial voids in the
metal using cross-sectional TEM may be possible.5 However, given
the typical number densities of pitting sites, the relatively small
interface area sampled by these cross sections may not permit pit
precursor defects such as voids to be effectively viewed. However,
one can expect the interfacial voids in the metal to be incorporated
into the oxide during anodic film growth as the metal layer contain-
ing voids is oxidized. Assuming that the volume change accompa-
nying incorporation is minor, inferences about void geometry and
location would be possible from the microscopic observations of
oxide voids. Plan-view microscopic images of oxide films sample a
larger interface area compared to cross-sectional images, increasing
the probability of finding voids. Nanoscale voids formed during an-
odic alumina formation have been found in prior transmission elec-
tron microscopy ͑TEM͒ studies, but were thought to derive from gas
evolution.6,7 In another study, TEM-detected voids were due to local
densification during oxide crystallization after anodizing.8
In this work, microscopic observations of voids in anodic oxide
films were carried out using both TEM and AFM. The experimental
protocol is illustrated in Fig. 1. Aluminum foil samples were used in
either the as-received condition or after various times of NaOH im-
mersion at open circuit. The different samples contained variable
quantities of interfacial voids in the metal, as determined by the
NaOH treatment time.3 Because anodic oxide growth occurs at both
metal-oxide and oxide-solution interfaces,9 the interfacial voids
should be incorporated into the interior of the film, as shown. In
some experiments, the metal underlying the anodic film was dis-
solved, and the remaining oxide was examined in plan view with
TEM. Otherwise, the anodic film was partially dissolved in a hot
acidic solution and its surface viewed with AFM; the uniformity of
oxide dissolution by such procedures was established earlier.10 As
illustrated in Fig. 1, an appropriate depth of dissolution would ex-
pose the voids as surface cavities. AFM images obtained after
different extents of oxide dissolution provided an assessment of
the void concentration as a function of depth. This quantitative in-
formation complemented the nondestructive through-thickness TEM
images.
Recently, a PAS study of anodic oxidation demonstrated that new
interfacial voids in the metal are created during anodizing.11 It was
speculated that void formation involves the agglomeration of inter-
facial metal vacancies formed by oxidation. After the formation of
such a void, continued anodic film growth would likely result in its
incorporation into the oxide. The microscopic observations would
then reveal voids occupying a range of depths, as opposed to the
single layer of voids shown in Fig. 1 resulting from only pre-
existing interfacial voids. It is seen here that microscopic character-
ization of the void depth profiles in anodic films yielded insight into
the process of interfacial void formation during anodic oxide
growth. The void depth profiles obtained for samples with different
NaOH pretreatment times suggested that voids are not formed at
random locations on the metal-oxide interface but instead at certain
favored sites. With regard to the possible function of voids as pitting
sites, their association with other types of surface defects may be
consistent with the empirically known correspondence between pits
and surface impurities12 or topographic asperities.13-15
* Electrochemical Society Student Member.
** Electrochemical Society Active Member.
z E-mail: krhebert@iastate.edu
Downloaded on 2014-06-27 to IP 128.252.67.66 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
B380
Journal of The Electrochemical Society, 151 ͑7͒ B379-B386 ͑2004͒
potentiostat/galvanostat ͑EG&G PAR-273͒, and the counter elec-
trode was a Pt wire. Anodic oxidation continued until voltages of 27,
53, 80, and 106 V were attained vs. the counter electrode. After
anodizing, the samples were rinsed thoroughly with DI water.
To prepare samples for TEM examination, the anodized foils
were immersed in solutions of 10 vol % bromine and 90 vol %
methanol until the aluminum metal was completely dissolved. The
remaining oxide films were placed on copper grid holders for TEM.
The films as mounted on the grids were electron transparent, so no
further sample preparation was necessary. The samples were exam-
ined in a Philips CM30 scanning/transmission microscope operated
at various voltages from 100 to 300 keV.
For AFM imaging, the anodized foils were cut into small pieces,
on which the surface oxide films were partially stripped by immer-
sion for different times up to 4 min in 5% H3PO4 at 80°C. The
stripped samples were washed in DI water and dried in air before
AFM imaging. To quantitatively determine the rate of film dissolu-
tion during stripping, the partly stripped foils were reanodized in
borate buffer, during which the potential transients were recorded.
As discussed in the results and discussion section, the fraction of the
oxide thickness removed during stripping could be inferred from
these potential transients. AFM imaging was carried out in direct
contact mode, using a 14 ␮m scanner with Si cantilevers and a
Si3N4 tip ͑Digital Instruments Nanoscope III͒. The photodiode volt-
age was set to 4.60 V; assuming a cantilever spring constant of 0.06
N/m, the estimated applied force is 1.5 nN. Each sample was imaged
four to five times at different positions, using a typical scan area of
5 ϫ 5 ␮m. The top view images presented here were further mag-
nified by cropping and enlarging, and a ‘‘sharpen filter’’ ͑Adobe
Photoshop͒ was used to highlight the small features of interest.
Figure 1. Schematic of the experimental protocol leading to preparation of
TEM and AFM samples. Voids are represented as circles, and the metal and
oxide films by shaded and white areas, respectively. The drawings illustrate
the formation of oxide voids by incorporation of pre-existing interfacial
voids during anodic film growth; creation of interfacial voids during anod-
izing and their subsequent incorporation in the film are not depicted.
Experimental
The aluminum foils used in this work were manufactured for use
in aluminum electrolytic capacitors ͑Toyo Aluminum͒. The foils
were 100 ␮m thick with a typical grain size of 100 ␮m, and their
nominal purity was 99.98%. The large grain size is due to extensive
annealing after rolling, e.g., for 5-6 h at 600°C.16 Impurities include
Cr, Cu, Fe, Ga, Mg, Si, and Zn with bulk concentrations of the order
10 wt-ppm.17
Results and Discussion
TEM.—During anodizing both NaOH-treated and untreated foils,
the potential increased at a constant rate of 1.1 Ϯ 0.05 V/s. From
Faraday’s law and the forming voltage-thickness ratio of 1.3-1.4
nm/V,18 a 100% current efficiency for oxide growth would imply a
rate of potential increase of 1.15 V/s at 2.5 mA/cm2. Hence, the
potential-time slopes indicate current efficiencies of approximately
100%.
TEM images of 106 V anodic films after dissolution of the alu-
minum substrate are presented in Fig. 2 and 3. The film in Fig. 2 was
formed on a foil that had been treated in NaOH for 5 min, whereas
that in Fig. 3 had received no NaOH treatment. Selected area elec-
tron diffraction confirmed that the anodic films in both cases were
amorphous. White features, many of which are roughly circular, are
apparent at higher magnification. Examples of these features are
Caustic treatment was carried out by immersion of foils in 1 N
NaOH solution for different lengths of time at room temperature,
after which they were washed thoroughly with deionized ͑DI͒ water.
Prior to anodic oxidation, the samples were dipped in 1 N HNO3 at
ambient temperature for 1 min to avoid the erratic anodizing behav-
ior sometimes encountered on foils anodized directly after NaOH
treatment. The nitric acid dip may eliminate a gel-like surface layer
formed in NaOH. For consistency, foils that had not received NaOH
treatment were also dipped in nitric acid prior to anodizing. Anodic
oxidation of the pretreated samples was carried out in a borate buffer
solution ͑pH 8.5-8.7͒ at room temperature, at a constant applied
current density of 2.5 mA/cm2. The current source was
a
Figure 2. TEM images of anodic alumi-
num oxide film formed to 106 V on alu-
minum foil treated in NaOH for 5 min. ͑a͒
Low magnification and ͑b͒ high magnifi-
cation.
Downloaded on 2014-06-27 to IP 128.252.67.66 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Journal of The Electrochemical Society, 151 ͑7͒ B379-B386 ͑2004͒
B381
Figure 3. TEM images of anodic alumi-
num oxide film formed to 106 V on alu-
minum foil with no NaOH treatment. ͑a͒
Low magnification and ͑b͒ high magnifi-
cation.
indicated by arrows in Fig. 2a,b and Fig. 3b. No diffraction could be
obtained from these features, or from the surrounding material. Ste-
reopair imaging was carried out and confirmed that the circular fea-
tures in Fig. 2 are voids within the film. However, some white fea-
tures in the as-received sample were found to be voids contained in
the film ͑e.g., indicated by the black arrow in Fig. 3b͒, and others are
pores through the film thickness ͑e.g., indicated by the white arrow
in Fig. 3b͒.
Clear differences in void size and distribution are apparent be-
tween Fig. 2 and 3. The voids in Fig. 2 appeared scattered through-
out the sample. Because of their small size and low associated con-
trast, they were not detected except at high magnification, and their
number density could not be estimated reliably. Linear strings of
voids were often seen, indicating that some common feature has
determined their position. For example, in Fig. 2a, the voids are
distributed along a slightly thicker region of the film implied by the
band of darker contrast running from the upper right corner to the
lower left. Such regions of thicker film may have been found along
the ubiquitous surface ridges apparent in AFM images, as discussed
later. In a sample of 18 voids, the diameter was distributed about a
mean of 42 nm with a standard deviation of 13 nm; the largest and
smallest voids were 67 and 25 nm diam.
of 1.3-1.4 nm/V.18 Takahashi et al. confirmed this relationship be-
tween the reanodizing potential and film thickness, using film thick-
ness from capacitance measurements and TEM, as well as obtaining
the mass of oxide dissolved from chemical analysis of the stripping
solution.10
Potential-time curves during reanodizing in borate buffer solu-
tions are shown in Fig. 4 for several samples with different NaOH
treatment times. In all cases, the anodic film had been formed to 106
V and then stripped for 60 s in phosphoric acid. In each potential
transient, the potential at first increased at a constant rate up to 80 V,
at which point the slope abruptly decreased. The initial slope in-
creased with NaOH treatment time, whereas the slope above 80 V
was the same for all samples. The final voltage of 110 V in the figure
is the voltage limit of the power supply.
Potential transients with rapid slope changes like those in Fig. 4
were observed previously during the anodic oxidation of aluminum
substrates that were covered by alumina films of known
porosity.20,21 Analysis of these experiments showed that, at the out-
set of anodizing, the applied current flowed only to the pore bot-
toms, where the oxide thickness was smallest; hence, the local film
growth current density was larger than the applied current density.
In contrast to Fig. 2, voids and pores in the film on the as-
received foil were readily visible at relatively low magnification.
Figure 3a is representative of a randomly chosen area of this size.
Bands of voids and pores, such as that indicated by the arrow, run
along bands of thicker film. These bands are roughly parallel to one
another and correspond to rolling lines on the foil surface. Preferen-
tial etch pit formation along rolling lines on as-annealed foils has
been noted earlier, suggesting that pit precursor defects are located
there.15,19 From a sample of 36 voids and pores, the average diam
was 76 nm with a standard deviation of 52 nm; the largest and
smallest diam were 208 and 22 nm. From the clear differences of
size, appearance, and distribution of voids in films on treated and
untreated samples, void formation during anodizing is clearly sensi-
tive to the foil’s pre-existing surface condition.
Chemical stripping of the anodic oxide film.—To detect the oxide
voids with AFM, the anodic films were partly dissolved in a phos-
phoric acid stripping bath, and then their surfaces were imaged.
Knowledge of the rate and uniformity of oxide dissolution during
stripping was required to interpret the AFM images in terms of the
depth distribution of voids in the film. This information was ob-
tained from potential measurements during constant current reanod-
izing of the samples with partially dissolved oxide films. The oxide
film thickness after stripping is revealed by these measurements, as
it is related to the measured anodizing voltage by a constant factor
Figure 4. Potential-time curves during reanodizing aluminum after partial
stripping of the anodic oxide film. Films were originally formed to 106 V
and then chemically stripped for 1 min in H3PO4 . The parameter is the
NaOH treatment time prior to the first anodization.
Downloaded on 2014-06-27 to IP 128.252.67.66 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
B382
Journal of The Electrochemical Society, 151 ͑7͒ B379-B386 ͑2004͒
Figure 5. Potential at the slope change during reanodizing ͑see Fig. 4͒ vs.
stripping time in H3PO4 . Aluminum foils were treated in NaOH for 5 min,
and then anodic films were formed to the indicated voltages.
This rapid film growth was associated with a high rate of potential
rise, which decreased with increasing porosity. When the pores were
completely filled with oxide, the slope of the potential transient
decreased suddenly; at this time, the film began to grow uniformly
across the surface, at a reduced rate equivalent to the applied current
density.
Therefore, the abrupt slope changes in Fig. 4 indicate that the
partially stripped anodic films were porous. According to this inter-
pretation, the porosity was largest in the anodic film on the as-
received foil and decreased to low values as the NaOH treatment
time was increased to 5 min. This trend is consistent with the TEM
observations, which revealed through-pores in the film on the as-
received foil, which are likely responsible for the small potential-
time slope on this sample. However, the voids in the film on the 5
min NaOH-treated sample were smaller and isolated and did not
appear to form continuous pores. This relatively nonporous film
morphology is consistent with the high initial potential-time slope
for this sample. Finally, note that the common slope of the potential
transients above 80 V is ϳ0.5 V/s, significantly smaller than the
value of 1.1 V/s expected for 100% current efficiency. This suggests
a reduced current efficiency after stripping, possibly due to a surface
layer left by the phosphoric acid stripping procedure.
AFM observations were confined to foils with at least 1 min
NaOH treatment, on which the film porosity was small and stripping
was expected to proceed uniformly. Because the film thickness and
forming voltage are proportional, the progress of stripping could be
inferred from the potential at the slope change during reanodizing.
For example, in Fig. 4, the fraction of the anodized film thickness
that remains after stripping is ͑80 V͒/͑106 V͒, or 0.75. Figure 5
shows the potential at the slope change plotted as a function of
stripping time. Data for two anodic film thicknesses are given, for
the same NaOH treatment time of 5 min. The linear decrease of
potential shows that film thinning occurred uniformly and at a con-
stant rate. Assuming a film thickness/voltage ratio of 1.4 nm/V, the
stripping rate was approximately 35 nm/min, independent of initial
film thickness. The reanodizing voltage decreased linearly to ϳ1.5
V, after which it remained constant, indicating complete removal of
the anodically formed oxide. The constant dissolution rate permitted
the stripping time to be calibrated in terms of film thickness; thus,
the fraction of the film thickness removed is the ratio of the stripping
time to the time for complete dissolution of the anodic film.
The relatively slow, uniform oxide dissolution during stripping
indicated that it could be used in conjunction with AFM surface
imaging to determine the depth profiles of voids or other defects in
Figure 6. AFM images after partial stripping of 106 V anodic films on foils
with 5 min NaOH treatment, showing effect of stripping time. Stripping
times: ͑a͒ 40 s ͑17% of oxide film thickness removed͒, height contrast 128
nm; ͑b͒ 90 s ͑38% of oxide removed͒, 128 nm; ͑c͒ 120 s ͑50% of oxide
removed͒, 116 nm; ͑d͒ 150 s ͑63% of oxide removed͒, 156 nm; ͑e͒ 180 s
͑75% of oxide removed͒, 156 nm; ͑f͒ 210 s ͑88% of oxide removed͒,
147 nm.
the film. Because of the slow dissolution, enlargement of void size
or distortion of shape by dissolution should be minimized. This ex-
pectation was examined further by comparison of TEM and AFM
results, as described in detail in the following paragraphs.
AFM observations of partially stripped anodic films.—Images of
oxide film surfaces after various stages of oxide stripping are shown
in Fig. 6 for a foil treated in NaOH for 5 min and then anodized to
106 V. When less than 40% of the anodic film had been dissolved,
the surface morphology was unchanged by stripping ͑see Fig. 6a,b͒.
However, open surface cavities with approximately circular shapes
were revealed after longer stripping times ͑see Fig. 6c-f͒. The term
‘‘cavity’’ is applied to these pit-like features revealed by oxide
stripping to distinguish them from corrosion pits. The average diam
of a population of about 100 cavities was 66 nm, with a standard
deviation of 18 nm. Surface height profiles are shown in Fig. 7
which correspond to the lines drawn in Fig. 6c and f; each profile
spans two surface cavities. As illustrated by Fig. 7, the typical cavity
depth was about 10-20 nm, suggesting a shallow, disk-like shape.
Figure 8 directly compares AFM and TEM images at the same
magnification. The dark circular areas in the AFM image ͑cavities͒
correspond to the white circular spots in the TEM image ͑voids͒.
Examples of both types of feature are indicated by arrows. The
Downloaded on 2014-06-27 to IP 128.252.67.66 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).