221901-2
Carcia, McLean, and Reilly
Appl. Phys. Lett. 97, 221901 ͑2010͒
dividual cells. Each cell represents barrier performance for a
specific ALD growth temperature ͑50, 75, or 100 °C͒ and
thickness, i.e., number of ALD cycles ͑N=75, 125, or 250͒.
In these transmitted-light images, the white spots are regions
where metallic Ca fully oxidized to optically transparent
Ca͑OH͒2. The recorded aging time ͑in days͒, when these
oxidation defects first appeared, is indicated in an individual
cell. For ALD growth conditions, for which the Ca-pixels
retained a featureless, defect-free metallic appearance7 ͑no
apparent defects͒, and damp heat aging only produced a very
thin, uniform surface oxidation layer, we report only the
value of WVTR in that cell.
The key observations were the following. ͑1͒ For all
thicknesses: 75, 125, and 250 cycles, films grown at 100 °C
had a WVTR equal to the glass control limit ͑Ͻ5
ϫ10−5 g/m2 day at 38 °C/85% RH͒. ͑2͒ When ALD films
were thick ͑250 cycles͒, independent of growth temperature,
they also had WVTR at the glass control limit. ͑3͒ For the
intermediate thickness, corresponding to 125 cycles, most
Ca-pixels developed oxidation defects after 51 days, except
for the isolated Ca-pixels, without oxidation defects. WVTR
determined from pixels without defects was also at the mea-
surement limit. Finally, ͑4͒ for the thinnest film ͑75 cycles͒,
Ca-pixels became fully transparent in only 7 days, corre-
sponding to the first measurement, for growth at 50 °C,
while Ca-pixels developed defects sooner ͑34 days͒ than
thicker films for growth at 75 °C.
After an additional ϳ500 h at 60 °C/85% RH, ALD
films grown at 100 °C for all thicknesses and thick Al2O3
films ͑250 cycles͒ grown at all temperatures continued to
show no visible oxidation defects and maintained a WVTR
equal to the glass control sample ͑ϳ3ϫ10−4 g/m2 day at
60 °C/85% RH͒. After 12 days at 60 °C/85% RH, the 75
cycle film grown at 75 °C now also became fully oxidized.
Moreover, for 125 cycles, Ca-pixels protected with a barrier
made at 50 °C showed more degradation—greater pixel
shrinkage—compared to a 125 cycle film made at 75 °C.
These data support that growing barrier films at higher tem-
perature or that are thicker reduces moisture permeation.
In comparison with the Ca results, we measured WVTR
at 38 °C/85% RH, using the MOCON Aquatran-1, versus
the thickness of ALD Al2O3 films grown at 50, 75, and
100 °C. The results are summarized in Fig. 2 with the pub-
lished instrument sensitivity limit, 0.5 mg H2O/m2 day,
indicated as a horizontal dotted line. For all growth tempera-
tures, the WVTR falls as the ALD film thickness increases.
For growth at 100 °C, a dashed line fitting these data inter-
sects the instrument measurement limit at ϳ7.5 nm ͑81
cycles͒. We define this thickness as the “threshold”
creases rapidly and above which permeation is at the limit of
the measurement technique. Because of the scatter of the
data for ALD growth at 50 and 75 °C, we fitted these com-
bined data with a single dashed line and assigned one thresh-
old thickness for the two temperatures, ϳ9.6 nm ͑113
cycles͒.
FIG. 2. ͑Color online͒ WVTR measured with commercial instrument at
38 °C/85% RH for Al2O3 barrier films grown on PET by ALD at 50, 75,
and 100 °C vs the Al2O3 thickness. The horizontal dotted line corresponds
to the published instrument measurement limit of 0.5 mg H2O/m2 day.
oxide thin films. In ALD Al2O3 films, the defects are
H-atoms that form OH bonds to Al, whereas in electrical
breakdown the defects are charge traps. At ALD growth tem-
perature of 50 °C, the H-defect concentration is quite
large,11 ϳ21 at. % or 2.5ϫ1022 H atoms/cm3, decreasing
as the ALD growth temperature increases. We postulate that
water permeates the film by an exchange mechanism, such as
described by the reaction: AlO͑OH͒+H2O→Al͑OH͒3. For
any ALD Al2O3 film thickness, there is a corresponding criti-
cal defect density, at which the onset of water permeation
occurs. Below this threshold thickness, permeation is facile.
Above the threshold thickness, negligible permeation occurs.
To quantitatively describe defect related permeation, we
use the analytical expression of Sune19 for analyzing electri-
cal breakdown in oxide films,
ao2
a /d
o
d
ao3
N =
.
͑1͒
ͩ ͪ
A
Applied to our permeation case, N ͑1/cm2͒ is the critical
surface H- or OH-defect concentration responsible for per-
meation through the thickness d, which corresponds to the
threshold Al2O3 thickness for permeation, along defect clus-
ao=1.25 nm, an efficiency factor18,19 of 1/200 that an indi-
vidual percolating chain of defects contributes to permeation,
and the reference area A=1 cm2, Fig. 3 plots N versus d.
ALD at 50 and 100 °C, plotted at the corresponding thresh-
old thickness, determined in this work to be 9.6 and 7.5 nm,
respectively. At d=7.5 nm, the model predicts a critical de-
fect concentration of N=0.75ϫ1022 cm−3 about half of
1.47ϫ1022 cm−3, the H-defect concentration measured in
ALD films grown at 100 °C, and for d=9.6 nm N=6.2
ϫ1022 cm−3, which is larger than 2.5ϫ1022 cm−3, the mea-
sured H-defect concentration in ALD Al2O3 grown at 50 °C.
In general, the model predicts the observed trend that a film
with lower defect density, corresponding to a higher ALD
Al2O3 growth temperature, will have a lower threshold thick-
ness for permeation. Moreover, the predicted values of N
agree reasonably well with the measured H-defect content.
To explain these results for water vapor permeation in
ALD thin films, we propose a new model that does not rely
on pinholes in our ALD films, which we contend are pinhole-
free. In our model, gas permeation proceeds along chains of
chemical defect clusters that percolate the thickness of the
film. This is related to models17–20 of electrical breakdown in
130.18.123.11 On: Mon, 22 Dec 2014 11:01:30
Carcia
