Atomic layer deposition of alumina onto carbon fibers

Article abstract of DOI:10.1111/j.1551-2916.2010.04340.x

Bundles of carbon fibers were successfully coated with alumina by atomic layer deposition via sequential exposures to trimethylaluminum and water at 77°C. Fibers were not damaged by this procedure. Scanning electron microscopy (SEM) images revealed that individual filaments were coated separately with a smooth layer; no bridging of fibers was observed. Transmission electron microscopy (TEM) and SEM images indicate that the coating was conformal and adhered well to the fiber surface. The average deposition rate was 0.25±0.02 nm/cycle, calculated from SEM images obtained at various positions of the fibers and of fibers coated with various numbers of cycles. In addition to image analysis, the coating thickness was as well estimated from the elemental composition of selected parts of a fiber bundle that was coated by 450 cycles. Along the fiber axis, the weight fraction of aluminum of the coated fibers varied from 5.3 to 6.5 wt% and perpendicular to the bundle axis it varied from 4.8 to 7.4 wt%. This translates into a variation of the estimated coating thickness along the fiber from 166 to 126 nm and perpendicular to it from 114 to 183 nm. One can estimate independently an average coating thickness from the analysis of SEM images of this specific bundle of 126 nm. The alumina coating improved oxidation resistance of the carbon fiber significantly. The oxidation onset temperature was 600°C for fibers coated with a 30-nm-thick layer of alumina and increased gradually with increasing coating thickness up to 660°C at a thickness of 120 nm. On the other hand, uncoated fibers started to oxidize already at 300°C.

Full text of DOI:10.1111/j.1551-2916.2010.04340.x

J. Am. Ceram. Soc., 94 [7] 2014–2022 (2011)
DOI: 10.1111/j.1551-2916.2010.04340.x
r 2011 The American Ceramic Society
ournal
J
Atomic Layer Deposition of Alumina onto Carbon Fibers
Amit K. Roy,z Wolfgang Baumann,z,J Steffen Schulze,y Michael Hietschold,y Thomas Mader,z
¨
Daisy J. Nestler,z Bernhard Wielage,z and Werner A. Goedelw,z
zPhysical Chemistry, Institute of Chemistry, Chemnitz University of Technology, 09111 Chemnitz, Germany
ySolid Surfaces Analysis Group, Institute of Physics, Chemnitz University of Technology, 09126 Chemnitz, Germany
zInstitute of Material Science and Engineering, Chemnitz University of Technology, 09125 Chemnitz, Germany
Bundles of carbon fibers were successfully coated with alumina
by atomic layer deposition via sequential exposures to trimethyl-
aluminum and water at 771C. Fibers were not damaged by this
procedure. Scanning electron microscopy (SEM) images re-
vealed that individual filaments were coated separately with a
smooth layer; no bridging of fibers was observed. Transmission
electron microscopy (TEM) and SEM images indicate that the
coating was conformal and adhered well to the fiber surface. The
average deposition rate was 0.2570.02 nm/cycle, calculated
from SEM images obtained at various positions of the fibers and
of fibers coated with various numbers of cycles. In addition to
image analysis, the coating thickness was as well estimated from
the elemental composition of selected parts of a fiber bundle that
was coated by 450 cycles. Along the fiber axis, the weight frac-
tion of aluminum of the coated fibers varied from 5.3 to 6.5 wt%
and perpendicular to the bundle axis it varied from 4.8 to
7.4 wt%. This translates into a variation of the estimated
coating thickness along the fiber from 166 to 126 nm and per-
pendicular to it from 114 to 183 nm. One can estimate inde-
pendently an average coating thickness from the analysis of
SEM images of this specific bundle of 126 nm. The alumina
coating improved oxidation resistance of the carbon fiber sig-
nificantly. The oxidation onset temperature was 6001C for fibers
coated with a 30-nm-thick layer of alumina and increased grad-
ually with increasing coating thickness up to 6601C at a thick-
ness of 120 nm. On the other hand, uncoated fibers started to
oxidize already at 3001C.
metallic5–7 or ceramic matrices.8 However, carbon fibers are
sensitive to oxidation and at least in the case of metallic alumi-
num (Al) composites have a tendency to form aluminum car-
bide. In presence of oxidizing humid conditions, tensile
properties start to deteriorate already at temperatures exceed-
ing 4001C.9
Thus, the protection of carbon fibers from reaction with
oxygen or with the matrix has been a big challenge for the last
few decades. One successful protection strategy is to coat the
fibers by a diffusion barrier made from silica, alumina, titania,
etc. In this case, the thickness of the coating plays an important
role; it should be thick enough to protect the fiber but still thin
enough to leave the tensile strength unaffected.7
An oxide coating can be deposited onto carbon fibers by
conventional techniques such as sol–gel10,11 and chemical vapor
deposition (CVD).12 Both techniques have the disadvantage that
the carbon fiber is damaged by oxygen containing precursors or
moisture, especially at temperatures exceeding 4001C. However,
there are interesting results of oxide deposition by nonhydrolytic
sol–gel methods,13–16 where organometalic precursors are used
as an oxygen source instead of direct oxygen sources like mo-
lecular oxygen and water. As well, there are some procedures of
controlled CVD that avoid direct oxygen sources (such as O3,
H2O, HOOH) and instead use oxygen-containing chemicals
(such as alkoxides or esters) as precursor.17–19 Another tech-
nique to reduce oxidative damage is to guide oxygen or oxygen-
containing precursors into the deposition chamber at some delay
after the other precursors.12 There is also a report20 of silica de-
position by CVD pursued at room temperature using silicon
tetrachloride, water as precursors, and ammonia as catalyst. In
addition, sol–gel methods require a careful optimization of a
plethora of process parameters, such as the concentration of the
coating liquid, the particle surface charge, densities, viscosity of
the sol, the wetting relationship between the liquid and the fiber,
and the coating rate. Liquid-phase precursors often form bridges
between filaments in a bundle. In these cases, bridges can impede
infiltration of a fiber bundle by the matrix and increase the ma-
trix porosity at the time of composite preparation. As a result,
the final strength of the ceramic matrix composite decreases sig-
nificantly. In case of CVD, the stoichiometry of the coating can
vary from filament to filament between the center and the outer
regions of the fiber bundle because of differences in the depo-
sition kinetics of the individual gaseous reactants. Therefore,
there is a demand of a smart coating technique that can deposit
uniform coatings onto each individual filament at low temper-
atures (o4001C) and which allows to control the thickness in the
range of nanometers.
I. Introduction
N fiber-reinforced composites, the mechanical properties of
structural materials are very successfully improved by incor-
I
porating fibers into a matrix. Especially demanding is the prep-
aration of such composites either with ceramic or metallic
matrices that are intended for use at temperatures exceeding
12001C.1,2 At such a high temperature, the composite might fail
due to an onset of creep of the fiber, chemical reaction between
fiber and matrix, or oxidation and/or hydrolysis due to oxygen
and moisture in the environment. Polycrystalline ceramic
fibers,3,4 such as SiC or various oxide materials, have the dis-
advantage of creep at the desired high temperatures usually
caused by changes in the polycrystalline microstructure. In ad-
dition, SiC is sensitive to oxygen.
On the other hand, carbon fibers have a significantly lower
tendency to creep and have been successfully incorporated into
W. Mullins—contributing editor
One technique to obtain conformal coating of well-controlled
thickness, penetrating deep into trenches, is atomic layer depo-
sition (ALD).21–24 ALD is a sequential cyclic vapor deposition
process in which substrates are exposed to two (or more) pre-
cursors in a sequential manner. The principle of ALD is shown
in Fig. 1, with an example of alumina deposition using
trimethylaluminum and water as precursors onto a substrate
Manuscript No. 28375. Received July 28, 2010; approved November 16, 2010.
JAuthor deeply involved in this project, unfortunately he is no more with us, and did not
witness completion of the project.
wAuthor to whom correspondence should be addressed. e-mail: werner.goedel@
chemie.tu-chemnitz.de
2014

July 2011
Deposition of Alumina onto Carbon Fibers
2015
tool, one may expect that processing conditions are not going to
damage the carbon fibers.
ALD was initially developed as a coating technique in the
preparation of flat panel displays and in microelectronics to coat
‘‘trenches in hard and inorganic’’ surfaces.21,22 However, now-
adays, ALD becomes more and more used for the coating and
modification of organic, even delicate, materials at rather mild
conditions, e.g. synthetic polymers such as polyethylene
terephthalate, polyethylene or polyvinylalcohol30,33,35 biopoly-
mers such as celluloses,39,40 spider silk,41 and carbon nano-
tubes.42,43 Whereas biopolymers usually comprise reactive
groups that easily react with inorganic precursors, in the case
of synthetic polymers or carbon, one might face the problem,
that initially there are no reactive groups that might covalently
bind an monolayer of reactants in a first cycle. Nevertheless,
conformal coating might be achieved, if at least one of the pre-
cursors is absorbed at the surface or absorbed within a region of
the substrate near to the surface. Thus, one may expect that
ALD-processing conditions are possible that allow to coat car-
bon fibers with alumina without damaging them.
Alumina often is deposited via ALD using water and
trimethylaluminum as precursors.30,33,35,41,42 Thus, in this
article, we apply ALD to coat carbon fiber bundles with alumina
using trimethylaluminum and water as precursors extending a
first report of the idea.44,45
II. Experimental Section
A coating process was carried out using a ‘‘home built’’ ALD
reactor and gas-flow system. A reaction tube, 1 m long and 4 cm
in diameter, made out of stainless steel is placed inside a 70-cm-
long tube furnace. The reaction tube contained a 30-cm-long
fiber holder with 12 channels made out of stainless steel. As
shown in Fig. 2, there are two separate precursor chambers sep-
arately connected with the reactor via tubes, made out of stain-
less steel of 1/400 diameter. Precursor doses are controlled by
automatic valves that can switch within seconds. Each separate
precursor delivery tube is connected to an inert gas supply that
acts as a carrier gas during precursor flow, otherwise, as a purge
gas. Inert gas flow was controlled by mass flow controllers. The
pressure inside the reactor is controlled by valve 3 and the pump.
The pressure inside the reactor is measured by various pressure
measurement gauges. The temperature of the main reactor is
controlled by the furnace. All valves are controlled by a home-
Fig. 1. Scheme of atomic layer deposition of alumina via sequential
exposure of a substrate containing hydroxyl groups to pulses of
trimethylaluminum and water. The symbols
hydrogen, carbon, oxygen, and aluminum atoms, respectively.
represent
already bearing hydroxyl groups. The substrate is exposed to
trimethylaluminum followed by purging. In the next step, it is
exposed to water followed by another purging. These four steps
deposit a layer of certain thickness onto the substrate. In be-
tween the precursors, the reaction chamber is purged by an inert
gas. Each of the precursors reacts with the surface, ideally form-
ing a chemisorbed monolayer that exposes functional groups,
which can react with the subsequent precursor. Purging by inert
gas in between the reaction steps not only prevents mixing of
precursors in the gas phase but also in addition removes by-
products of the surface reaction and secures that there is a min-
imum of impurities in the deposited coating. As a result, at ideal
conditions, homogeneous and uniform thin films are formed on
the substrate.25 Thus, under saturation reaction condition, the
thickness of the additionally deposited layer is constant for each
cycle. Ideally, one expects one monolayer for each cycle. In
practice, however, often only a distinct fraction of monolayer
(e.g., 1/2 monolayer)26 forms due to steric hindrance of the pre-
cursors. Depending on the process and reactor being used, one
complete cycle takes from 0.5 s to a few seconds and deposits a
0.01- to 0.35-nm-thick layer.27 ALD is well known for the depo-
sition of various oxides at low temperatures (below 1001C)28–35
and also without any direct oxygen source like molecular oxygen
or water.26,36–38 Thus, if ALD is used as an oxide deposition
Fig. 2. (a) ALD flow reactor designed for fiber coating. There are two
separate precursor chambers connected to the main reactor separately by
two valves, valve 1 and valve 2. Each precursor chamber is connected to
an inert gas supply that is controlled by a mass flow controller. There are
two pressure measurement gauges for monitoring the pressure inside the
reactor. Valve 3 and the pump are controlling the pressure inside the
reactor. (b) Scheme of the sample holder used to place the fibers within
the furnace.

2016
Journal of the American Ceramic Society—Roy et al.
Vol. 94, No. 7
Fig. 3. Scanning electron microscopic images of (a) carbon fibers desized, but uncoated, (b), (c), and (e) coated with layers of alumina obtained after 350
deposition cycles, and (d) after 100 deposition cycles. (b) Overview of a part of a fiber bundle coated with alumina. (c) A single fiber coated with alumina,
dark gray part in front is the carbon fiber and bright part is the alumina coating. (d) and (e) Magnified images with an axis of view almost perpendicular
to the cut end. One can see uniform and conformal coating and good adhesion of the coating onto the fiber surface at two coating thicknesses. Small
fractions of the corrugated perimeter and of the cross section are visible in the left part of (d) and the lower left part of (e).
written program with the help of the software ‘‘labview’’ from
National Instruments.
longitudinally oriented within the reaction tube. At the two ends
of the fiber holder (one facing the precursor flow, the other one
facing the pump), the fibers were gently bent by 1801. Fibers
were coated by alumina at a temperature of 771C and a pressure
of 0.1 mbar using trimethylaluminum (98% from ABCR GmbH
& Co. KG, Karlsruhe, Germany) and deionised water as pre-
cursors. Trimethylaluminum was heated to 701C and the tem-
perature of the water was maintained at 501C. Nitrogen was
used as an inert gas. Both mass flow controllers were set at 20
standard cubic centimeters per minute (sccm). Thus, in purging
Tenax HTA 5331 6K, PAN-based carbon fibers with 6000
filaments (of 7 mm in diameter) were purchased from Toho Tenax
(Europe GmbH, Wuppertal, Germany). As obtained, the fiber
bundles bear a polymeric sizing. This sizing was removed by
thermal treatment at 7001C in N2 atmosphere. Besides this ther-
mal desizing, no further chemical or physical treatment was
performed before coating. The 3.6-m-long segments of this fiber
were mounted onto the fiber holder in such a way that they were

July 2011
Deposition of Alumina onto Carbon Fibers
2017
Fig. 4. Scanning electron microscopic images of carbon fibers coated with a thickness of 50 nm that were bent and subsequently straightened again.
Bending was done around cylinders of decreasing radius of (a) 12.5 mm, (b) 7.5 mm, (c) 0.75 mm, and (d) around a razor blade.
cycles, there was a flow of 40 sccm purging the reactor. In the
first half cycle, the fiber bundle was exposed to trimethylalumi-
num for 20 s followed by purging the reactor with nitrogen for
30 s. In the second half cycle, it was exposed to water vapor for
20 s followed by purging the reactor with nitrogen for 30 s. One
complete cycle consisted of 100 s and was repeated for appro-
priate times to obtain systematically varied coating thicknesses.
Purging times were increased up to 100 s. In a separate set of
deposition experiments, the flow of purging gas was increased
up to 100 sccm, while keeping the purging time constant at 100 s.
Thermogravimetry was performed using a TGA7 from Perkin
Elmer (Waltham, MA). The 5-mm-long coated fiber specimens
(weight varying between 3.0 and 4.0 mg) were heated from 301
to 9001C with a heating rate of 51C/min and subsequently held
for 30 min at 9001C in 20 mL/min synthetic air flow.
Elemental flash analysis was performed via flash pyrolysis/
oxidation of the sample, coupled to a chromatographic separa-
tion and detection of the gaseous oxidation products using a
FLASHAE 1112 of Thermo Scientific (Waltham, MA).
The amount of Al deposited onto the carbon fibers was
obtained by complete degradation of a known length of coated
fiber in nitric acid, followed by dilution to a standardized
volume and quantitative measurement of the Al content of
this solution via atomic absorption spectroscopy (AAS). From
this value, the amount of deposited alumina was calculated as-
suming that aluminum was present in the form of Al2O3. The
thickness of the coating, d, was estimated from the mass of the
alumina coating ma, the density of the alumina, ra 5 2.6 g/cm3
taken from the literature,30 the radius of a single filament,
r 5 3.5 ꢀ 10ꢁ4 cm, the number of filaments per bundle, N, and
the length of the bundle, L:
!
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ma=ra
1 þ
d ¼ r
ꢁ 1
(1)
pr2NL
As a check for consistency, the thickness was as well calcu-
lated using the total mass, mt of a fiber bundle, instead of the
latter three parameters and the density of the carbon, rc (using
the value of rc 5 1.76 g/cm3 provided by the manufacturer):
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!
ma=ra
1 þ
d ¼ r
ꢁ 1
(2)
ðmt ꢁ maÞ=rc
Both methods give almost identical results.
Samples for imaging with transmission electron microscopy
(TEM) (Fig. 5) were prepared using a focused ion beam (FIB).
The details of this method were reported elsewhere.46
III. Results and Discussion
Bundles of 6000 filaments of carbon fiber were coated with
alumina by sequential exposure to trimethylaluminum and

2018
Journal of the American Ceramic Society—Roy et al.
Vol. 94, No. 7
Fig. 5. (a) High-resolution transmission electron microscopic image of a cross section perpendicular to the fiber axis through a near surface region of a
single fiber coated with alumina (and embedded into an auxiliary resin). (b) Enlarged image of the interface between the fiber and the coating (arrow). (c)
and (d) Dark-field images. The variations in gray shades in the images indicate some degree of crystallinity, albeit there seems to be coexistence of several
polymorphs. The degree of crystallinity seems not to be a function of the distance from the interface.
water vapor at reduced pressure and a temperature of 771C; the
coated carbon fibers were characterized by scanning electron
microscopy (SEM), TEM, energy-dispersive X-ray spectroscopy
(EDXS), AAS, and thermogravimetric analysis (TGA). SEM
images of uncoated and coated fibers and fiber ends cut after the
coating process are shown in Fig. 3. Figure 3(a) shows that the
uncoated fibers have a corrugated surface bearing grooves par-
allel to the fiber axis and irregular protuberances; this morphol-
ogy varies from fiber to fiber. The overview image (Fig. 3(b))
reveals that the fibers within the bundle are still separated from
each other, no bridging has been formed, and that coatings have
been deposited evenly on all filaments. Figure 3(c) shows the end
of one individual fiber. There is an obvious contrast in the in-
tensity of secondary electrons between the 7 mm of the fiber close
to the end and the final part of the fiber that appears to be less
bright in the SEM image. This indicates that the coating, which
comprises the heavier element Al and thus gives rise to a
stronger secondary electron emission, was removed from the
front part of the filament in the cutting process. Besides this in-
tensity contrast, the surface of the coated region has a structure
very similar to the surface of the noncoated regions. This fact is
as well visible in the close up images Figs. 3(d) and (e). Thus, the
coatings are smooth, uniform, and conformal. There are almost
no regions with the exceptions of the freshly cut ends that in-
dicate delamination of the coating. Thus, there seems to be good
adhesion of the coating to the fiber. In places where the cutting
induced delamination, one observes clear steps (see Figs. 3(d)
and (e)). These indicate that cohesion within the coating is
stronger than the adhesion. In addition, the occurrence of these
well-defined steps allows to estimate the layer thickness. To get a
bit more insight into the adhesion strength, fibers were bent
around cylinders of decreasing radius (see Fig. 4). If the fibers
are bent around cylinders of radii down to 12.5 mm, no signifi-
cant delamination was observed. Slight delamination (o1% of
the total area) occurred if the bending radius was in between
12.5 and 0.75 mm. Severe delamination of more than half of the
total area, however, occurred only if the fibers were bent around
the edge of a razor blade. This bending around a razor blade
caused most fibers to break and delamination occurred in the
proximity of the break points. Adhesion is a function of residual
mechanical stress and of defects within the material. Residual
stress might arise from mismatch of thermal expansion coeffi-
cients, if deposition was done at an elevated temperature and the
fibers subsequently cooled down. Defects might arise from ther-
mally induced recrystallization after deposition that causes fault
lines and grain borders within the material. The process depicted
here is conducted at a comparatively low temperature; thus, it is
likely that there is a low amount of residual stress and defects

July 2011
Deposition of Alumina onto Carbon Fibers
2019
(e.g., the TEM in Fig. 5 indicates some degree of crystallinity),
which favors the adhesion of the coatings to the fibers.
of the fiber might be due to the comparatively low process tem-
perature of 771C.
A TEM image of a cross section of the interface between fiber
and coating (prepared using a FIB) shows that the coating is
conformal; the distance between the two interfaces is almost in-
dependent on the position despite the corrugated surface (see
Fig. 5). The magnified image (Fig. 5(b)) reveals that the interface
between fiber and coating is sharp; if there is a zone of inter-
diffusion it has a thickness of o3 nm.
To determine whether all of the 6000 filaments of a fiber
bundle were coated evenly and separately regardless of the po-
sition within or along the bundle, a bundle (coated at 771C using
450 deposition cycles) was split into parts originating from re-
gions close to the center of the bundle and close to the perimeter
of the bundle and cut into segments originating from the part
that faced the gas supply, a middle part, and a part that faced
the pump during the coating process. For each of these seg-
ments, the amount of Al was measured via elemental analysis by
AAS as described above; the Al weight fraction varies from
5.3 wt% at the part facing the pump to 6.5 wt% at the part
facing the gas supply and from 7.4 wt% at the fraction taken
from the perimeter to 4.8 wt% at the fraction taken out of the
center of the fiber. This translates into an estimated coating
thickness (respectively deposition rate), which varied along the
fiber from the end facing the gas supply to the end facing the
pump from 166 to 126 nm (0.37–0.28 nm/cycle) and from center
to perimeter from 114 to 183 nm (0.25–0.40 nm/cycle).
The coating thickness of the same fiber bundle was measured
from SEM images at various positions along the length and also
along the cross section of the filament bundle. SEM images re-
vealed the general trend but the average coating thickness was
126 nm. Discrepancies between the results in these two methods
can be attributed to experimental error and the uncertainties of
the parameters used for calculating the coating thickness (e.g.,
the value taken for the density). The variations in coating thick-
ness observed here indicate that conditions for ALD might not
have been fully achieved. The coating thickness increases with
decreasing distance of the fiber parts to the precursor source.
This observation, combined with an observed deposition rate
above the usual value, might indicate that there is in addition to
chemical reaction of the precursors some degree of physisorp-
tion taking place in a nonequilibrium manner and being influ-
enced by the flow characteristics of the reactor.
Elemental analysis of the coated fiber by EDXS (Fig. 6) re-
vealed a dominating carbon content as well as Al and oxygen.
The dominating amount of carbon was expected because of the
carbon fiber, the presence of Al and oxygen confirms the
expected alumina coating.
Because EDXS does not always yield accurate quantitative
chemical compositions, coated fibers were burned selectively by
thermal oxidation in air. The carbon, nitrogen, and hydrogen
content of the residue were measured via elemental flash anal-
ysis. Results of this analysis indicate that the amount of carbon,
hydrogen, and nitrogen within the residue of thermal oxidation
were 1.6%, 0.06%, and o0.01%, respectively. Possibly, this
small amount of hydrogen and a part of the carbon can be at-
tributed to decomposition products of the trimethylaluminum,
and the rest of the carbon to an irremovable part of the fiber.
The ratios of mass-to-length of uncoated and coated fibers
were obtained by gravimetry. The ratio of the mass of the de-
posited alumina to the length of the fiber was obtained inde-
pendently from elemental analysis by AAS. If this value is
subtracted from the weight/length of the coated fiber, one ob-
tains a weight/length that differs from the original weight/length
of the uncoated fiber by only 0.7%. Thus, one can conclude that
the carbon fibers did not significantly lose weight even though
the process involved the presence of moisture. This preservation
The coating deposition rate was measured by the analysis of
SEM images of ends of fibers that were exposed to various
numbers of coating cycles and cut in the middle after the process
such as the one shown in Fig. 3. As can be seen from Fig. 7,
there is a linear relationship between the coating thickness and
the number of cycles with an average deposition rate of
0.2570.02 nm/cycle. The deposition rate is lower than the value
that can be estimated theoretically for a monomolecular layer of
alumina but larger than the rates usually reported in the liter-
ature.47 Purging times were increased up to 100 s (keeping the
flow of purging gas constant at 40 sccm) and subsequently the
flow of purging was increased to 100 sccm (keeping the purging
time at 100 s). Both variations had no significant effect on the
amount deposited per cycle. The growth per cycle decreased to
0.1770.03 nm/cycle, however, if the temperature was raised to
1501C. The explanation for the growth per cycle observed here,
Fig. 6. Elemental analysis of a coated carbon fiber by energy-dispersive
X-ray spectroscopy (EDXS). (a) The area selected for EDXS analysis.
(b) EDXS spectrum indicating that there are four elements present in
that selected area. There is a signal for copper that is due to the sample
holder used for this analysis. There is a dominating signal for carbon
originating from the fiber, as well as aluminum and oxygen as expected
from the alumina coating.
Fig. 7. Coating thickness as a function of the number of cycles. The
coating thickness increases almost linearly with the number of cycles at
an average deposition rate of 0.2570.02 nm/cycle.

2020
Journal of the American Ceramic Society—Roy et al.
Vol. 94, No. 7
thus, might be the comparatively low deposition temperature
used in our experiments. In most reports in the literature, the
deposition temperature is significantly higher than the 771C used
here. Thus, part of the deposition may be due to physisorbed
water that is present in a larger amount if the temperature is low,
but becomes less if the temperature is increased. Because growth
per layer is independent of purging time in our case, one might
conclude that purging of an excess of gaseous precursors was
complete in each cycle and that physisorbed water was either
completely removed as well or desorbed with such a low rate
that further purging did not change the residual water content.
There are indeed reports showing that outgassing of physisorbed
water in low-temperature ALD processes applied to substrates
of high surface area like powders and nanotubes may need times
even longer than half an hour.48–50 In Fig. 7, the data can be
represented by a line passing through the origin of the graph.
This indicates that the coating starts spontaneously from the
first cycle without a nucleation period, or in other words that a
continuous layer of limited growth or a dense array of tiny seed
particles are formed already at a very early stage of the coating
process. This interpretation is further supported by the smooth
and conformal appearance of the coating. To some extent, this
might be unexpected because (i) carbon fibers presumably have
a comparatively low concentration of reactive functional groups
like hydroxyl groups on the surface and (ii) there are reports that
carbon nanotubes seem to resist coating by ALD unless pre-
treated by azo compounds, nitric oxide, or surfactants.42,43 On
the other hand, the surface of a carbon fiber is not a smooth
defect-free graphitic basal plane as one usually presumes in the
case of carbon nanotubes; it might expose to the gas phase at
least to some extent polar groups created via oxidation and
edges of disrupted graphitic planes. Furthermore, there are re-
ports of significant adsorption of trimethylaluminum onto
graphitic carbon.51,52 This physical adsorption might be the rea-
son for the fast induction and comparatively high initial depo-
sition rate of the coating process.
of alumina, the oxidation starts at a temperature of approxi-
mately 6001C and the fibers are burned completely at a temper-
ature of approximately 8001C. With increasing thickness of the
coating, the oxidation onset temperature increases slightly and
also the amount of ash. Oxidation onset temperature reached a
plateau value of 6601C for fibers bearing a coating of a thickness
of 120 nm or more. Thus, alumina coating improves the oxida-
tion resistance of the carbon fibers to a significant extend.
As mentioned above, 5-mm-long specimens were used for the
oxidation test. These specimens were cut from the fiber bundle;
however, at the freshly cut ends, the coating was removed during
cutting process (as shown in Fig. 3(c)). Thus, there is the pos-
sibility that oxidation starts from the unprotected ends, and ac-
tually will be less prominent in defect-free endless fibers. To test
this hypothesis, one 5-mm-long specimen was coated after cut-
ting in such a way that both ends were covered with the coating.
The oxidation resistance of these end-protected fibers was
compared with fibers with unprotected ends (see Fig. 9). Both
specimens were heated from 301 to 5001C with a heating rate of
51C/min and subsequently held at 5001C for 300 min in synthetic
air. In case of unprotected ends, fiber oxidation starts already at
the very beginning, fibers with both ends protected are stable up
to temperature of approximately 2001C. At the end of this ex-
periment, fibers with unprotected ends lost approximately 9% of
their mass, the fibers with protected ends lost 7%. Thus, these
results indicate that an endless fiber may be more resistant to
oxidation than fibers with unprotected ends. We notice, how-
ever, that the effect is not very prominent. Thus, there must be
as well a mechanism of oxygen diffusion through the coating.
A comparatively thin layer is already sufficient to delay
oxidation significantly; however, Fig. 8 seems to indicate that
the oxidation protection levels off at layer thicknesses exceeding
100 nm. Layers of any thickness exceeding this threshold seem
to fail at temperatures around 6601C regardless of their thick-
ness. Thus, one might come to the conclusion that the layers
become significantly more permeable at that temperature,
allowing oxygen diffusion perpendicular to the layer surface.
One possible explanation for this fact might be a mismatch of
the coefficients of thermal expansion of fiber and coating, e.g. as
shown in the case of alumina coating as an oxidation barrier for
iron particles.53,54 In that case, the coefficient of thermal expan-
sion of the iron particle was larger than the corresponding
coefficient of the coating. Thus, thermal expansion lead to ex-
tensional stress on the coating and finally crack formation. In
our case, the coefficient of expansion of the fiber has a lower
One important aspect of coating carbon fibers is to protect
them from oxidation at elevated temperatures. Thus, the oxida-
tion resistance of the alumina-coated fibers is characterized by
thermogravimetric analysis (TGA). Five milliliter long coated
fiber specimen were used for TGA, these specimens were heated
from 301 to 9001C with a heating rate of 51C/min and subse-
quently held at 9001C for 30 min in a flow of synthetic air. The
results are shown in Fig. 8; the oxidation onset temperature (3%
weight loss) for the uncoated fiber is approximately 3001C and
the fibers are completely burned at temperatures of approxi-
mately 4001C. In case of fibers coated with a 33-nm-thick layer
Fig. 9. Thermogravimetric analysis of fibers coated with a 135 nm layer
ꢂ
ꢂ ꢂ
of alumina protected at both ends ( ), and unprotected at both ends
(- -). The specimens were heated from 301 to 5001C with a heating rate of
51C/min and subsequently held at 5001C for 300 min in synthetic air.
End unprotected coated fibers started to oxidize from the very beginning
and within 400 min lost approximately 9% weight; end protected coated
fibers started oxidation approximately after 40 min and within 400 min
lost approximately 7%.
Fig. 8. Thermogravimetric analysis of coated fibers with various coat-
ing thicknesses. The specimens were heated from 301 to 9001C with a
heating rate of 51C/min and subsequently held at 9001C for 30 min in
synthetic air. Oxidation onset temperature increases with increasing
coating thickness and levels off for coating thicknesses exceeding ap-
proximately 100 nm.

July 2011
Deposition of Alumina onto Carbon Fibers
2021
Fig. 10. Transmission electron microscopy images of residual coatings after complete oxidation of the fibers in air at an oxidation temperature of
(a) 5501C and (b) 9001C. Whereas sample (a) still has a low degree of crystallinity, sample (b) has crystallized during the heat treatment, possibly opening
up pores and pathways for oxygen diffusion perpendicular to the layer.
value than the corresponding value of the coatingww; thus in our
case, the coating will experience compressive stress upon heat-
ing. This might probably seal cracks instead of opening them up.
Another process that has the potential to increase permeability
may be crystallization, which might induce shrinkage of the
coating due to the densification that is associated with crystal-
lization and may give rise to grain boundaries that act as pores
passing through the coating. In order to investigate this detail,
coated fibers were subjected to complete oxidation in air, one
sample at a temperature of 9001C, another one at a temperature
of 5501C (complete oxidation took 4 h in the first and 12 h in the
latter case). The coatings that remain after these treatments were
investigated via TEM (see Fig. 10). The image shows a still low
degree of crystallinity in the sample obtained via oxidation at
5501C, whereas the sample obtained via oxidation at 9001C
showed a significant degree of crystallization, even exposing de-
tails that may be interpreted as cracks within the coating. Thus,
it is likely that the observed lowering of the oxidation stability at
higher temperatures is due to crystallization. Oxidation thus
might be delayed, if compounds are included in the coating
that delay crystallization or preferentially fill the gaps between
crystallites.
only few impurities. ALD is already used in semiconductor in-
dustry; in future, it may as well be a promising coating tool for
reinforcing fibers that may be used in composites.
Acknowledgments
We are grateful to Harry Rose, Micromanufacturing Technology, Chemnitz
University of Technology for designing the control software of the ALD reactor,
Frank Diener and Frank Sternkopf from the metal-workshop, Chemnitz Univer-
sity of Technology, for building components of ALD reactor. We are thankful to
¨
¨
Sabine Stockel and Cornell Wustner from Physical Chemistry, Chemnitz Univer-
sity of Technology, for help and scientific advices in building the ALD reactor. We
are thankful to Gisela Baumann from Solid Surfaces Analysis Group, Chemnitz
University of Technology for SEM and EDXS analysis.
References
1K. M. Prewo and J. A. Batt, ‘‘The Oxidative Stability of Carbon-Fiber
Reinforced Glass-Matrix Composites,’’ J. Mater. Sci., 23 [2] 523–7 (1988).
2R. J. Kerans, R. S. Hay, T. A. Parthasarathy, and M. K. Cinibulk, ‘‘Interface
Design for Oxidation-Resistant Ceramic Composites,’’ J. Am. Ceram. Soc., 85 [11]
2599–632 (2002).
3G. Simon and A. R. Bunsell, ‘‘Creep-Behavior and Structural Characterization
at High-Temperatures of Nicalon Sic Fibers,’’ J. Mater. Sci., 19 [11] 3658–70
(1984).
4C. Lesniewski, C. Aubin, and A. R. Bunsell, ‘‘Property Structure Character-
ization of a Continuous Fine Alumina Silica Fiber,’’ Compos. Sci. Technol., 37
[1–3] 63–78 (1990).
IV. Conclusion
5A. Feldhoff, E. Pippel, and J. Woltersdorf, ‘‘Interface Engineering of Carbon
Fiber Reinforced Mg–Al Alloys,’’ Adv. Eng. Mater., 2 [8] 471–80 (2000).
6P. Bertrand, M. H. Vidal-Setif, and R. Mevrel, ‘‘The CVD of Pyrolytic Carbon
and Titanium Diboride Coatings on Carbon Fibre Yarns for Use in Aluminium-
Based Composites,’’ Surf. Coat. Technol., 96 [2–3] 283–92 (1997).
7F. Reischer, E. Pippel, J. Woltersdorf, S. Stockel, and G. Marx, ‘‘Carbon
Fibre-Reinforced Magnesium: Improvement of Bending Strength by Nanodesign
of Boron Nitride Interlayers,’’ Mater. Chem. Phys., 104 [1] 83–7 (2007).
8R. R. Naslain, R. Pailler, X. Bourrat, S. Bertrand, F. Heurtevent, P. Dupel,
and F. Lamouroux, ‘‘Synthesis of Highly Tailored Ceramic Matrix Composites by
Pressure-Pulsed CVI,’’ Solid State Ionics, 141–142, 541–8 (2001).
9M. E. Westwood, J. D. Webster, R. J. Day, F. H. Hayes, and R. Taylor,
‘‘Oxidation Protection for Carbon Fibre Composites,’’ J. Mater. Sci., 31 [6] 1389–
97 (1996).
ALD was successfully used for the deposition of alumina onto
carbon fibers at 771C. At these process conditions, the estimated
weight loss of the carbon fiber was o1%. The coating was uni-
form and conformal. Bending tests showed that the coating
adheres to the fiber down to a bending radius small enough
to make the fiber break. There were no bridges between two
fiber filaments visible in electron microscopy images. The coat-
ing thickness increases linearly with the number of cycles; thus, it
is comparatively easy to predict the coating thickness. The coat-
ing is efficient in slowing down the oxidation of carbon fibers. In
our early studies, alumina coating was deposited using alumi-
num chloride and water as precursors.44,45 This deposition pro-
cess was also successful; however, it was difficult to control the
aluminum chloride precursor dose due to the solid nature of the
aluminum chloride. In addition, there was a chloride impurity in
the coating. On the other hand, in the depositions of alumina
using trimethylaluminum and water as precursors reported here,
doses of precursors were controlled successfully without any
problem. Trimethylaluminum is highly reactive compared with
aluminum chloride; as a result, the deposited coating contained
10Y. Q. Wang, B. L. Zhou, and Z. M. Wang, ‘‘Oxidation Protection of Carbon-
Fibers by Coatings,’’ Carbon, 33 [4] 427–33 (1995).
11N. I. Baklanova, T. M. Zima, A. T. Titov, N. V. Isaeva, D. V. Grashchenkov,
and S. S. Solntsev, ‘‘Protective Coatings for Carbon Fibers,’’ Inorg. Mater., 42 [7]
744–9 (2006).
12G. Zhao, D. M. Giolando, and J. R. Kirchhoff, ‘‘Chemical-Vapor-Deposition
Fabrication and Characterization of Silica-Coated Carbon-Fiber Ultramicroelec-
trodes,’’ Anal. Chem., 67 [15] 2592–8 (1995).
13V. Lafond, P. H. Mutin, and A. Vioux, ‘‘Control of the Texture of Titania-
Silica Mixed Oxides Prepared by Nonhydrolytic Sol–Gel,’’ Chem. Mater., 16 [25]
5380–6 (2004).
14L. Bourget, R. J. P. Corriu, D. Leclercq, P. H. Mutin, and A. Vioux,
‘‘Non-Hydrolytic Sol–Gel Routes to Silica,’’ J. Non-Cryst. Solids, 242 [2–3]
81–91 (1998).
wwAlumina: isotropic linear coefficient of expansion of ꢃ 9 ꢀ 10ꢁ6 Kꢁ1 5 1/3 of the cubic
coefficient of expansion of 27.6ꢀ 10ꢁ6 Kꢁ1 given in Moghtaderi, Shames, Doroodchi.53
Carbon fiber: longitudinal coefficient of expansion ꢃ 0 Kꢁ1, transverse coefficient of ex-
pansion ꢃ 5 ꢀ 10ꢁ6 Kꢁ1 extracted from figure 5 of Pradere and Sauder.55
15A. Vioux, ‘‘Nonhydrolytic Sol–Gel Routes to Oxides,’’ Chem. Mater., 9 [11]
2292–9 (1997).

2022
Journal of the American Ceramic Society—Roy et al.
Vol. 94, No. 7
16P. H. Mutin and A. Vioux, ‘‘Nonhydrolytic Processing of Oxide-Based
37L. J. Zhong, F. Chen, S. A. Campbell, and W. L. Gladfelter, ‘‘Nanolaminates
of Zirconia and Silica using Atomic Layer Deposition,’’ Chem. Mater., 16 [6]
1098–103 (2004).
Materials: Simple Routes to Control Homogeneity, Morphology, and Nanostruc-
ture,’’ Chem. Mater., 21 [4] 582–96 (2009).
17A. C. Jones, H. C. Aspinall, P. R. Chalker, R. J. Potter, T. D. Manning, Y. F.
Loo, R. O’Kane, J. M. Gaskell, and L. M. Smith, ‘‘MOCVD and ALD of High-
Kappa Dielectric Oxides using Alkoxide Precursors,’’ Chem. Vap. Deposition, 12
[2-3] 83–98 (2006).
38B. B. Burton, M. P. Boleslawski, A. T. Desombre, and S. M. George, ‘‘Rapid
SiO2 Atomic Layer Deposition Using Tris(tert-pentoxy)silanol,’’ Chem. Mater., 20
[22] 7031–43 (2008).
39M. Kemell, V. Pore, M. Ritala, M. Leskela, and M. Linden, ‘‘Atomic Layer
Deposition in Nanometer-Level Replication of Cellulosic Substances and Pre-
paration of Photocatalytic TiO2/Cellulose Composites,’’ J. Am. Chem. Soc., 127
[41] 14178–9 (2005).
18D. M. Giolando, J. R. Kirchhoff, H. Mueller, P. Q. Nguyen, and I. N. Odeh,
‘‘CVD of Alumina on Carbon and Silicon Carbide Microfiber Substrates for
Microelectrode Development,’’ Chem. Vap. Deposition, 8 [3] 93–8 (2002).
19R. L. Nigro, G. Malandrino, and I. L. Fragala, ‘‘Metal-Organic Chemical
Vapor Deposition of CeO2 (100) Oriented Films on No-Rolled Hastelloy C276,’’
Chem. Mater., 13 [12] 4402–4 (2001).
40G. K. Hyde, K. J. Park, S. M. Stewart, J. P. Hinestroza, and G. N. Parsons,
‘‘Atomic Layer Deposition of Conformal Inorganic Nanoscale Coatings on Three-
Dimensional Natural Fiber Systems: Effect of Surface Topology on Film Growth
Characteristics,’’ Langmuir, 23 [19] 9844–9 (2007).
20J. W. Klaus and S. M. George, ‘‘SiO2 Chemical Vapor Deposition at Room
Temperature using SiCl4 and H2O with an NH3 Catalyst,’’ J. Electrochem. Soc.,
147 [7] 2658–64 (2000).
41S. M. Lee, E. Pippel, U. Gosele, C. Dresbach, Y. Qin, C. V. Chandran, T.
Brauniger, G. Hause, and M. Knez, ‘‘Greatly Increased Toughness of Infiltrated
Spider Silk,’’ Science, 324 [5926] 488–92 (2009).
21T. Suntola and J. Antson, ‘‘Methods for Producing Compound Thin Films’’;
U.S. Patent No: 4058430, 1977.
42D. B. Farmer and R. G. Gordon, ‘‘Atomic Layer Deposition on Suspended
Single-Walled Carbon Nanotubes via Gas-Phase Noncovalent Functionalization,’’
Nano Lett., 6 [4] 699–703 (2006).
22R. L. Puurunen, ‘‘Surface Chemistry of Atomic Layer Deposition: A Case
Study for the Trimethylaluminum/Water Process,’’ J. Appl. Phys., 97 [12] 121301–
52 (2005).
43G. D. Zhan, X. H. Du, D. M. King, L. F. Hakim, X. H. Liang, J. A.
McCormick, and A. W. Weimer, ‘‘Atomic Layer Deposition on Bulk Quantities of
Surfactant-Modified Single-Walled Carbon Nanotubes,’’ J. Am. Ceram. Soc., 91
[3] 831–5 (2008).
23L. Niinisto, J. Paivasaari, J. Niinisto, M. Putkonen, and M. Nieminen, ‘‘Ad-
vanced Electronic and Optoelectronic Materials by Atomic Layer Deposition: An
Overview with Special Emphasis on Recent Progress in Processing of High-k
Dielectrics and other Oxide Materials,’’ Phys. Status. Solidi. A, 201 [7] 1443–52 (2004).
24M. Knez, K. Niesch, and L. Niinisto, ‘‘Synthesis and Surface Engineering of
Complex Nanostructures by Atomic Layer Deposition,’’ Adv. Mater., 19 [21]
3425–38 (2007).
44A. K. Roy, W. Baumann, I. Konig, G. Baumann, S. Schulze, M. Hietschold,
T. Mader, D. J. Nestler, B. Wielage, and W. A. Goedel, ‘‘Atomic Layer Deposition
(ALD) as a Coating Tool for Reinforcing Fibers,’’ Anal. Bioanal. Chem., 396 [5]
1913–9 (2010).
25K. E. Elers, T. Blomberg, M. Peussa, B. Aitchison, S. Haukka, and S. Marcus,
‘‘Film Uniformity in Atomic Layer Deposition,’’ Chem. Vap. Deposition, 12 [1]
13–24 (2006).
45A. K. Roy and W. A. Goedel, ‘‘Fiber Coating by Atomic Layer Deposition’’;
U.K. Patent Application GB2467928A, 2010.
46H. Mucha, T. Kato, S. Arai, H. Saka, K. Kuroda, and B. Wielage, ‘‘Focused
Ion Beam Preparation Techniques Dedicated for the Fabrication of TEM
Lamellae of Fibre-Reinforced Composites,’’ J. Electron. Microsc., 54 [1] 43–9
(2005).
26D. Hausmann, J. Becker, S. L. Wang, and R. G. Gordon, ‘‘Rapid Vapor
Deposition of Highly Conformal Silica Nanolaminates,’’ Science, 298 [5592] 402–6
(2002).
27M. Leskela and M. Ritala, ‘‘Atomic Layer Deposition Chemistry: Recent
Developments and Future Challenges,’’ Angew. Chem. Int. Ed., 42 [45] 5548–54
(2003).
47R. L. Puurunen, ‘‘Growth per cycle in atomic layer deposition: Real Applica-
tion Examples of a Theoretical Model,’’ Chem. Vap. Deposition, 9 [6] 327–32
(2003).
28J. W. Klaus, O. Sneh, and S. M. George, ‘‘Growth of SiO2 at Room
Temperature with the Use of Catalyzed Sequential Half-Reactions,’’ Science,
278 [5345] 1934–6 (1997).
48D. M. King, X. H. Liang, P. Li, and A. W. Weimer, ‘‘Low-Temperature
Atomic Layer Deposition of ZnO Films on Particles in a Fluidized Bed Reactor,’’
Thin Solid Films, 516 [23] 8517–23 (2008).
29J. W. Klaus and S. M. George, ‘‘Atomic Layer Deposition of SiO2 at Room
Temperature using NH3-Catalyzed Sequential Surface Reactions,’’ Surf. Sci., 447
[1–3] 81–90 (2000).
49D. M. King, J. A. Spencer, X. Liang, L. F. Hakim, and A. W. Weimer,
‘‘Atomic Layer Deposition on Particles using
a Fluidized Bed Reactor
with in Situ Mass Spectrometry,’’ Surf. Coat. Technol., 201 [22–23] 9163–71
(2007).
30M. D. Groner, F. H. Fabreguette, J. W. Elam, and S. M. George, ‘‘Low-
Temperature Al2O3 Atomic Layer Deposition,’’ Chem. Mater., 16 [4] 639–45 (2004).
31M. Putkonen and L. Niinisto, ‘‘Atomic Layer Deposition of B2O3 Thin Films
at Room Temperature,’’ Thin Solid Films, 514 [1–2] 145–9 (2006).
32Y. Du, X. Du, and S. M. George, ‘‘Mechanism of Pyridine-Catalyzed SiO2
Atomic Layer Deposition Studied by Fourier Transform Infrared Spectroscopy,’’
J. Phys. Chem. C, 111 [1] 219–26 (2007).
50A. S. Cavanagh, C. A. Wilson, A. W. Weimer, and S. M. George, ‘‘Atomic
Layer Deposition on Gram Quantities of Multi-Walled Carbon Nanotubes,’’
Nanotechnology, 20 [25] 255602–9 (2009).
51D. W. Kisker, D. A. Stevenson, J. N. Miller, and G. B. Stringfellow, ‘‘The
Vapor-Phase Interaction of Trimethylaluminum with Graphite during Omvpe,’’
J. Phys-Paris., 43 [Nc-5] 221–7 (1982).
33Q. Peng, X. Y. Sun, J. C. Spagnola, G. K. Hyde, R. J. Spontak, and G. N.
Parsons, ‘‘Atomic Layer Deposition on Electrospun Polymer Fibers as a Direct
Route to Al2O3 Microtubes with Precise Wall Thickness Control,’’ Nano Lett., 7
[3] 719–22 (2007).
52D. W. Kisker and D. A. Stevenson, ‘‘Mechanism of Graphite Baffle Gettering
in Organometallic Vapor-Phase Epitaxy—Adsorption of Trimethylaluminum on
Graphite,’’ J. Electron. Mater., 12 [2] 459–82 (1983).
53B. Moghtaderi, I. Shames, and E. Doroodchi, ‘‘Combustion Prevention of
Iron Powders by a Novel Coating Method,’’ Chem. Eng. Technol., 29 [1] 97–103
(2006).
34M. Knez, A. Kadri, C. Wege, U. Gosele, H. Jeske, and K. Nielsch, ‘‘Atomic
Layer Deposition on Biological Macromolecules: Metal Oxide Coating of To-
bacco Mosaic Virus and Ferritin,’’ Nano Lett., 6 [6] 1172–7 (2006).
35J. D. Ferguson, A. W. Weimer, and S. M. George, ‘‘Atomic Layer Deposition
of Al2O3 Films on Polyethylene Particles,’’ Chem. Mater., 16 [26] 5602–9 (2004).
36M. Ritala, K. Kukli, A. Rahtu, P. I. Raisanen, M. Leskela, T. Sajavaara, and
J. Keinonen, ‘‘Atomic Layer Deposition of Oxide Thin Films with Metal
Alkoxides as Oxygen Sources,’’ Science, 288 [5464] 319–21 (2000).
54L. F. Hakim, C. L. Vaughn, H. J. Dunsheath, C. S. Carney, X. Liang, P. Li,
and A. W. Weimer, ‘‘Synthesis of Oxidation-Resistant Metal Nanoparticles via
Atomic Layer Deposition,’’ Nanotechnology, 18 [34] 345603–9 (2007).
55C. Pradere and C. Sauder, ‘‘Transverse and Longitudinal Coefficient of
Thermal Expansion of Carbon Fibers at High Temperatures (300–2500 K),’’
Carbon, 46 [14] 1874–84 (2008).
&