Test Method for Constituent Content of Composite Materials 1

Walter G. McDonough, Donald L. Hunston and Joy P. Dunkers

National Institute of Standards and Technology
100 Bureau Drive, Stop 8543
Gaithersburg, MD 20899

Goals
To extend the burn off test protocol in ASTM D 3171-99 to the determination of
the fiber and void contents of carbon/glass hybrid composites. To support this extension
by experimentally demonstrating that it is possible, with one test, to estimate fiber
volume contents and void contents of typical glass/carbon/epoxy hybrid composites.

Introduction
Hybrid composites are increasingly being used in infrastructure, oil recovery, and
aerospace applications. These hybrids use two different fibers to reinforce the resin
thereby gaining some of the advantageous properties of both fibers. Typically glass and
carbon are used since this combination provides high performance with low cost. A
critical issue in such materials is the fiber mix ratio and the fiber and void content. At
present, there is no simple way to determine these features although many techniques
exist for measuring the components of single reinforcement composites. As a result, we
propose an extension of the burn-off test in ASTM D 3171-99 to characterize
carbon/glass hybrids. In support of this method, we offer data from tests on a series of
samples made to have known compositions and data on a commercially prepared sample
where fiber mix ration is known and void content was determined by using microscopy.

The current ASTM test method incorporates resin burn off as a way to determine
the constituent content for composite materials. Test Method I-Procedure G is for
reinforcements such as glass or ceramic that are not affected by high-temperature
environments or reinforcements such as carbon where temperature is adequately
controlled so that the reinforcement does not char. Briefly, Procedure G calls for putting
the specimens into a preheated muffle furnace at 500 °C or lower depending on the
composite system (a temperature at which specimens will not spontaneously ignite). The
specimens are then to be heated to 565 °C ± 30 °C or other temperature compatible with
the composite system that will burn off the matrix and leave the reinforcement. The
maximum time for burn off should be 6 h. The matrix is considered combusted if no
matrix/reinforcement block exists. Ash and reinforcement should be the only items
visible.
1
See also: W. G. McDonough, J. P. Dunkers, K. M. Flynn, and D. L. Hunston, “A Test Method to
Determine the Fiber and Void Contents of Carbon/Glass Hybrid Composites,” Journal of ASTM
International 1(3), 1-15, March 2004.
 Burnout tests are a common way to estimate the individual volume contents of the
components. Glass content, resin content and sizing content can be calculated and from
this data, the void content can then be determined. For carbon/glass hybrid composites,
we have found that by increasing the temperatures in stages, we can remove the resin and
sizings at one temperature, and remove the carbon fibers at another, higher temperature,
leaving only glass fibers. We have also been able to estimate the void contents for
industrial composites.

The amendment that we are proposing shall call for heating the specimen in a
furnace from room temperature to 600 °C and holding it at that temperature for 30 min-
45 min (with timing to start when the furnace reaches 585 °C). This step will remove the
resin and any organic sizing on the fibers. We also found that any char that was formed
was removed. Subsequently, we shall raise the temperature to 900 °C and hold the
specimen at that temperature until the carbon fibers are removed (usually around 30 min).

Proposed Procedure
This procedure was developed for epoxy matrices with carbon and glass
reinforcements, but may be applicable to other systems as well with appropriate
modification of times and temperatures.

Before beginning the measurements it is desirable to use thermal gravimetric

analysis (TGA) to determine the optimal test conditions (times and temperatures) for the
particular constituents being used, and the corrections factors discussed below if
something other than the default values are needed. If the constituents are available, they
can be tested individually to help provide this information. The times and temperatures
cited below have been found to work well for a specific epoxy system and are useful
starting points, but may not be optimum for all resin systems.

Before starting the tests, it is useful to thoroughly dry the specimens by placing
them in a desiccator and monitoring the mass as a function of time until changes are no
longer observed. The error introduced by moisture in the sample is not large but can
become particularly important if void content is calculated since void content is also
generally small. The dry mass of the sample is designated W0. The uncertainty in the
mass determination should be 0.0001 g or better.

If void content and volume fractions of constituents are to be determined, it is

necessary to measure or estimate the densities of the constituents (see ASTM Method
D792 or D1505). Densities of cured resin, the glass fiber, and the carbon fiber are
designated ρr , ρg , and ρc , respectively. In addition, the volume of the specimen to be
tested, V0, must be determined. The uncertainty in the mass measurements should be
0.0001 g or better while the uncertainty in the determination of dimensions should be
0.005 mm or better.
 Weigh each specimen and then place it in a desiccated crucible. Before insertion
of the sample, the crucible should be cleaned by heating to 900 °C (or the highest use
temperature that will be used in the burn off test) in a muffle furnace, cooled to room
temperature in a desiccator, and weighed to determine its mass.

Place the crucible containing the sample into the oven and heat to 600 °C. When
the temperature reaches 585 °C, usually in around 30 min, start recording the time at this
first temperature plateau. After 30 min, there should only be reinforcement left in the
crucible. It may be necessary to extend this time if any visible resin remains. Remove
the crucible from the oven, keeping the oven at temperature, and place the crucible in a
desiccator for cooling to room temperature. Weigh the specimen and crucible to the
nearest 0.0001 g, and subtract the mass of the crucible to determine the mass of the
sample after the first temperature exposure, WT1. (Note: The existing standard called for
placing the crucible into a preheated muffle furnace at 500 °C or lower depending on the
composite system. This may be necessary if the furnace being used has a very long
heating up period that leads to degradation of the carbon fibers.)

Place the crucible into the oven that has remained at 600 °C and heat to 900 °C.
When the temperature reaches 885 °C, start recording the time at this second temperature
plateau. After 60 min, there should only be glass reinforcement left. If necessary, keep
the crucible in the oven until all of the carbon reinforcement is removed. Remove the
crucible from the oven, keeping the oven at temperature, place the crucible in a
desiccator, and allow it to cool to room temperature. Weigh the specimen and crucible to
the nearest 0.0001 g, and subtract the mass of the crucible to determine the mass of the
sample after exposure to the second temperature, WT2.

The mass of the resin (Wr), glass fibers (WG), and carbon fibers (WC) can then be
determined by solving the following 3 equations in 3 unknowns.

Wo = Wr + WC + WG (0.1)

WT 1 = f T 1,r Wr + f T 1,C WC + f T 1,G WG (0.2)

WT 2 = f T 2,r Wr + f T 2,C WC + f T 2,G WG (0.3)

The six correction factors, fi,j , account for the fact that the mass of each component left
after a given temperature treatment may not be exactly 0 or 1. Because glass will not
react at high temperatures, it is relatively easy to assure that

fT 2,r = fT 2,C = 0 and fT 1G = fT 2G = 1

In selecting the temperature and exposure time for step one, the goal is to conditions that
come as close as possible to burning off all the resin while leaving all the carbon fiber
unaffected so
 fT 1, r = 0 and fT 1,C = 1

In the experiments used to develop this test method (see below), these default
values worked very well. If these conditions can not be achieved, it may be possible to
calibrate the corrections factors to account for the deviation for ideality. To estimate fT1,C
from calibration experiments, however, it is recommended that composite sample with
known carbon fiber content be used, because experiments show that fibers in a composite
are often more resistant to vaporization than they are when tested without resin.

The mass fractions for the resin, carbon fiber, and glass fiber are Wr / W0 ,
WC / W0 , and WG / W0 , respectively. The corresponding volume fractions are Wr / ρr V0 ,
WC / ρC V0 , and WG / ρG V0 . The volume fraction of voids is determined by subtracting
the volume fraction of each component from 1. In calculating the void content,
remember to make accurate measurements of the dimensions of the specimen, and to
either determine the mass and volume amounts of the sizings used on the fibers either
experimentally or using the manufacturer's data sheet.

Uncertainties in the determinations outlined above depend on the materials

involved and how carefully the tests are performed. Consequently, it is not possible to
give specific numbers. As general guidelines, however, it is useful to consider numbers
of a particular system.

Supporting Documentation
To support the proposed test method, experiments were conducted on a model
hybrid composite that is being considered for filament wound structures used in off-shore
oil recovery on tension leg platforms.

Experiments: Materials and Equipment

The resin that was used was Epon Resin 862, manufactured from Bisphenol F and
epichlorohydrin (Shell Chemical Co.). The curing agent used was Epi-Cure W, which is
composed of mixed isomers of diethyltoluenediamine. The resin/curing agent ratio by
mass was 100.0/26.4. The carbon fibers were Grafil 34-700 in 12K tows with no twists
and the standard sizings (Grafil Inc., Sacramento, CA). The glass fibers were PPG-1062
with 247 yield, no twist, and the standard sizing (PPG Industries, Inc., Pittsburgh, PA).
Furnace used was Ney M-525 series II furnace (Yucaipa, CA). Weights were measured
on an Ohaus Analytical Plus AP 250D scale, (Florham Park, NJ). Dimensions were
measured on a Fowler & NSK MaxCal digital caliper. Thermogravimetric analyses were
done on a Perkin-Elmer TGA 7 machine (PerkinElmer Corporation, Norwalk, CT).

The first step in devising a burn off procedure for carbon/glass hybrid composites
was to use thermo-gravimetric analysis (TGA) to establish potential testing routes.
Figure 1 shows burn off results for the matrix used in study. The heating rate to 600 oC
was 50 oC/min followed by a hold time at 600 oC of 60 min. We can see in the figure that
the resin is essentially gone after 25 min. Figure 2 shows the result from bringing the
burn off temperature to only 500 oC. At this temperature, we clearly have a significant
amount of material remaining after 60 min. Thus, for an initial estimate, we felt that 600
o
C was a promising first plateau temperature. Figure 3 shows an isothermal curve for
carbon fiber at 600 oC. Here we see a weight loss of around 4 % after 60 min. Figure 4
shows the burn off behavior of carbon fibers at 600 oC for 30 min followed by a jump to
900 oC and a hold for 50 min. We see the small mass loss at 600 oC followed by the
complete mass loss at 900 oC. From this data, we concluded that the carbon should be
able to survive the first plateau temperature of 600 oC, and that the carbon would be
removed at a second plateau temperature of 900 oC. Note that the TGA runs were done
on bare fibers. We can expect less fiber degradation when the fibers are encapsulated in
resin. Another consideration was that we wanted to minimize the time at the first plateau
the chance for degradation of the carbon fibers. Figure 5 shows the TGA response of
glass fiber taken to 900 oC, cooled down and recycled to 900 oC. The 1 % mass loss can
be attributed to sizing and moisture being burned off.

Based on the TGA results, we developed a two-step process to be able to

determine the constituents of a hybrid composite. The first plateau temperature would be
600 oC. At this temperature, we would remove the resin and the sizing on the fibers. The
second plateau temperature would be 900 oC, at which point the carbon would be
removed. Subsequently, we could calculate the fiber, resin and void contents of the
composite. First, we ran burn off tests on the base resin, the carbon fibers and the glass
fibers to get a fuller understanding of the responses of the components used to make the
composites. Next, we tested composites of known composition that served as a standard
by which we could compared our results from the burn off test. To that end, we made
model glass fiber composites, carbon fiber composites, and various carbon-glass hybrid
composites of known composition. Once we had this information, we then expanded the
test to industrial composites.

The model composites were made in specially built molds containing a cavity
nominally 20.0 cm long by 0.8 cm wide by 0.6 cm deep. These composites were made
by a hand lay-up process wherein we weighed out carbon tows, glass tows or a
predetermined mixture of tows depending on the composite that we wished to make.
Subsequently, we dipped individual tows into a cup of resin, previously mixed with
curing agent and heated to 50 oC, and then we placed the impregnated tow into the cavity
and repeated the process until the cavity was filled. We then put the cover on the mold,
clamped the cover in place using C-clamps, and then placed the mold assembly into a
programmable oven and cured the specimen following the cure schedule specified by the
vendor. After processing, the specimens were weighed and measured. By knowing the
amount of fibers put in the specimen, we could then calculate the volume content of the
specimens. By having this information, we now had a known standard by which to
measure the results from our proposed burn out test. After the specimens were molded,
they were sectioned into smaller pieces and then stored in a desiccator until they were
tested.
 Burn-out tests were performed on several specimens. As stated previously,
carbon tows, glass tows and matrix specimens were tested to get the base material
response. Carbon composites and glass composites were then tested for their response.
Hybrid specimens were tested with the following designations: 1:1 glass:carbon to
designate those specimens that were made using a 1:1 ratio of glass fiber tows to carbon
fiber tows; 3:1 designating a 3:1 ratio of glass to carbon fibers; and 1:3 to designate a 1:3
ratio of glass to carbon fibers. In the table, we show the expected results along with the
calculated results. Since we also examined the effect of variations in the burn out
temperatures, we have listed the burn out conditions for the individual test specimens.

The model composites, being laid up by hand, usually had a low fiber volume
content. We also ran the burn off tests on industrially made hybrid composites. These
composites had a much higher fiber volume content. The one difference in testing these
materials was in the step where we needed to remove the carbon fibers. We noted that
occasionally, we needed 60 minutes to remove the carbon fibers, but in the end, the
carbon fibers were successfully removed.

Statement of Uncertainties

The uncertainty in the weight was 0.0001 g. The uncertainty in the dimensions
was 0.005 mm. The uncertainty in the volume was 0.002 mm3. The type B uncertainty
in the density was estimated to be 0.005 g/cc. The combined standard uncertainty for the
void content for the NIST-made composites was 0.005 and the combined standard
uncertainty for the void content in the industrial grade composites was 0.006. The
uncertainty in the TGA measurements was 2 oC.

Results

The data in Table 1 and Figures 6 and 7 show the results from the tests on all of
the materials, both composites and constituent materials. The processing conditions
listed varied somewhat as we tried different plateau temperatures and times. The data for
the hybrid resin confirmed what was found in the TGA study (Figure 1) in that the resin
was burned off after 30 min. Sample 2 was kept in the oven 15 min longer, but we have
no reason to assume that the resin as not burned off earlier. The carbon tows showed an
even greater mass loss at 600 oC than that shown in Figure 3, and both sets of data
indicate that we need to minimize the hold time at the first plateau temperature. At
900 oC, again we see agreement between the burn off data and the TGA data shown in
Figure 4. The glass tows, both in Table 1 and in Figure 5, show a similar 1% mass loss
that we attribute to sizing and moisture. The results from the burn off test for the
constituents, combined with the results from the TGA study confirmed our choices of
600 oC as the first plateau temperature and 900 oC as the second plateau temperature.

As shown in Table 1, we found good agreement between the expected fiber

volume contents and the measured contents. Although some of the variations were due to
the uncertainties in the measurements, some of the variations were due to inherent
differences in the microstructure among the specimens. The glass fibers tend to act as an
internal standard since we do not observe any mass loss over the temperatures used for
this study. At 900 oC, the glass changes from fiber form to a solidified bead. Thus, we
can also look at the variations in the glass content as a level of variation inherent in the
specimen. In the NIST specimens, we have a relatively low fiber volume content. This
was due to the manual layup on the composite specimens. Variations in the industrial
composite were probably a function of size of the specimen cut and tested. The global
fiber ratio would be expected to vary on a smaller more localized scale.

The void contents for the composites made at NIST were calculated to be less
than 1 %, which puts the measurements at the level of the uncertainty in the experiment.
We were able to get measurable void contents for the industrial grade composites. We
then compared these results with results obtained by microscopy as a check on our
answer. For the composite that had a high glass/low carbon ratio, the void content as
measured by the burnout test was 3.84% and that measured by microscopy was 5.55 %.
For the composite that had a low glass/high carbon ratio, the void content as measured by
the burnout test was 3.97% and that measured by microscopy was 4.91 %. For this paper,
the microscopy was done by cutting, potting, and polishing sections of composites, and
then taking pictures of the polished surfaces and estimating a value for the percent voids
seem by using a computer program using Matlab.

Discussion
By the nature of the burnout test, we probably always get some removal of
carbon. However, in a composite, most of the energy used goes into burning off the resin
and sizing and, if we compare the expected vs. the measured values of the fiber volume
of the carbon fibers, they tend to be very close. If we only burn off carbon fiber tows, we
see a greater loss of carbon mass. Even holding the carbon fibers from a composite
specimen an extra 30 min at 600 oC, we measure a significant amount of mass loss. Thus
we favor limiting the time of exposure at this first plateau temperature.

The amount of sizing on the glass and carbon fibers was taken from the supplier
data sheets and the procurement specifications. The amount of sizing on the glass tows
can be estimated as well by burning off the size from glass tows. For carbon fibers, the
fibers could undergo a solvent wash. In any event, the sizing does not seem to add a
significant amount of mass or volume to the composite specimen.

Although the data certainly needs to be augmented, we are encouraged by the

results to date. As far as voids are concerned, the NIST samples were molded essentially
void-free. One of the reasons why this was possible was due to the relatively low fiber
volume content of the specimens. Another reason was that we had the time to ensure that
the resin thoroughly impregnated the tows before the specimens were cured. Both
microscopy and burnout results point to exceedingly low voids. The industrial grade
composites, however, did have measurable void contents. On the whole, the estimates
from the burnout test to date may be on the conservative side. Microscopy does not seem
to solve the problem, since extensive imaging is necessary to get a reasonable estimate
and the standard deviation can vary depending on the number of pictures taken, what
resolutions are used, what magnifications, etc.

We found that for the most part, the resin and sizing were removed at 600 oC after
30 min. Where we sometimes observed difficulties was in removing the carbon at
900 oC. Occasionally, we needed to extend the time at temperature to ensure that the
carbon was removed. This removal was verified visually.

Conclusion
Based on the data presented in this report, we believe that we have an acceptable
modification to D 3171 to allow the estimation of the constituents of glass/carbon/epoxy
hybrid composites.

Figures

TGA of Resin
110
600
100

90
500
80

Temperature ( C)
70

o
Weight (%)

400
60

50
300
40

30 200

20
Weight (%)
Temperature
10 100

0
0 10 20 30 40 50 60

Time (min)

Figure 1: TGA plot of the mass loss of resin at 600 oC. Note that the resin was removed
after 30 min.
 TGA of Resin

110 600

100

90 500

80

Temperature (oC)
70 400
Mass (%)

60

50 300

40

30 200

20
Mass (%)
100
10 Temperature (oC)

0
0 10 20 30 40 50 60

Time (min)

Figure 2: TGA plot of the mass loss of resin at 500 oC. Note that the resin was not
completely removed after 55 min.

TGA of Carbon Fiber

101
600

100

500
Mass (%)
99
Temperature
Temperature ( C)
o

400
Mass (%)

98

97 300

96
200

95
100

94
0 10 20 30 40 50 60

Time (min)

Figure 3: TGA plot of the mass loss of carbon fiber at 600 oC.
 TGA of Carbon Fiber
110 1000

100 900
90
800
80
700

Temperature ( C)
70

o
Mass (%)

600
60
500
50

40 400

30 300

20 200
Mass (%)
10 Temperature (oC)
100
0
0 10 20 30 40 50 60 70 80 90

Time (min)

Figure 4: TGA plot of the mass loss of resin at 500 oC and at 900 oC. Note that the
carbon was removed at 900 oC.

TGA of Glass Fiber

101 1000

900
100
800

700
Temperature ( C)

99
o
Mass (%)

600

98 500

400
97
300

96 Mass (%) 200

Temperature
100
95
0 10 20 30 40 50 60

Time (min)

Figure 5: TGA plot of the mass loss of glass fiber at 900 oC.
 100%

95%

90%

85%

void volume fraction

carbon volume fraction
80% glass volume fraction
resin volume fraction

75%

70%

65%
glass 3:1 1:1 1:3 carbon
composite glass/carbon glass/carbon glass/carbon composite
tested composite composite composite tested
expected tested expected

Figure 6: Bar chart of the expected and calculated components of the NIST made
industrially made composites used in this study.

100%

90%

80%

70%

% voids
carbon:glass
60% glass:carbon

50%

40%

30%
Industrial hybrid Industrial hybrid Industrial hybrid Industrial hybrid
composite #1 tested composite #1 composite #2 tested composite #2
expected expected

Figure 7: This figure shows the relative ratios of carbon to glass in industrially made
hybrid composites. The percent voids were included to show the differences in
specimens measured by the burnout test (tested) with those measured from
microphotographs.
Table 1: Results from the burn off tests

Hybrid resin
sample 1 complete mass loss at 600 C for 30 min
sample 2 complete mass loss at 600 C for 45 min

Glass tows
sample 1 0.79 % mass loss at 600 C for 30 min, no change after 60 min at 900 C
sample 2 1.0 % mass loss at 600 C for 45 min, no change after 60 min at 900 C
sample 3 1.1 % mass loss at 600 C for 40 min, no change after 60 min at 900 C

average 0.96% mass loss

Carbon tows
sample 1 4.1 % mass loss at 600 C for 30 min, complete mass loss after 30 min at 900 C
sample 2 9.6 % mass loss at 600 C for 45 min, complete mass loss after 30 min at 900 C
sample 3 7.2 % mass loss at 600 C for 45 min, complete mass loss after 30 min at 900 C

average 7.0 % mass loss

NIST glass composite expected volume measured volume

sample 1 29.6 29.8 30 min at 600 C, 30 min at 900 C
sample 2 29.6 27.2 30 min at 600 C, 30 min at 900 C
sample 3 29.6 29 30 min at 600 C, 30 min at 900 C

average 28.7

NIST carbon composite expected volumes measured volumes

sample 1 12.5 12.2 30 min at 600 C
sample 2 12.5 12.6 30 min at 600 C

average 12.4

NIST hybrid 1:1 glass/carbon expected volumes measured volumes

sample 1 13.6/8.0 13.3/7.2 30 min at 600 C, 30 min at 900 C
sample 2 13.6/8.0 14.0/8.0 30 min at 600 C, 30 min at 900 C
sample 3 13.6/8.0 13.9/7.5 45 min at 600 C, 30 min at 900 C
sample 4 13.6/8.0 13.9/8.1 30 min at 600 C, 90 min at 900 C
sample 5 13.6/8.0 13.9/7.9 60 min at 515 C, 30 min at 800 C

average 13.8/7.7

NIST hybrid 3:1 glass/carbon expected volumes measured volumes

sample 1 24.1/5.2 24.6/5.4 40 min at 600 C, 30 min at 900 C
sample 2 24.1/5.2 23.5/5.1 30 min at 600 C, 30 min at 900 C
sample 3 24.1/5.2 23.9/5.2 45 min at 600 C, 30 min at 900 C
sample 4 24.1/5.2 24.1/5.2

average 24.0/5.2

NIST hybrid 1:3 glass/carbon expected volumes measured volumes

sample 1 8.2/15.5 8.6/15.4 50 min at 600 C, 30 min at 900 C
sample 2 8.2/15.5 8.0/14.7 30 min at 600 C, 30 min at 900 C
sample 3 8.2/15.5 8.0/14.8 30 min at 600 C, 30 min at 900 C
sample 4 8.2/15.5 8.0/15.2 30 min at 600 C, 60 min at 900 C

average 8.2/15.0

industrial composite expected ratios measured ratios

glass/carbon glass carbon
sample 1 37/63 40/60 50 min at 600 C, 60 min at 900 C
sample 2 37/63 38/62 40 min at 600 C, 30 min at 900 C
sample 3 37/63 38/62 30 min at 600 C, 60 min at 900 C

average 38.7/61.3

industrial composite expected ratios measured ratios

glass/carbon glass/carbon
sample 1 64/36 67/33 50 min at 600 C, 60 min at 900 C
sample 2 64/36 66/34 40 min at 600 C, 30 min at 900 C
sample 3 64/36 67/33 30 min at 600 C, 60 min at 900 C

average 66.7/33.3