The making of carbon fiber

A look at the process by which precursor becomes carbon fiber through a careful (and
mostly proprietary) manipulation of temperature and tension.

Article Post: 12/19/2008

VICKI MCCONNELL

Related Topics:

Carbon Fiber

Related Suppliers
Adherent Technologies Inc.

C. A. Litzler Co. Inc.

Despatch - ITW EAE

Harper International

Michelman

Mitsubishi Chemical Carbon Fiber & Composites Inc.

Mitsubishi Rayon Co. Ltd.

Although many readers of HPC use carbon fiber, few know much about how it is made.
That should surprise no one. Carbon fiber producers are tight-lipped about how their
product is manufactured. Each producer’s fiber differs from those of its competitors, and
the processing details that give each brand its signature characteristics are considered to
be intellectual property. The carbon fiber manufacturing process also is notoriously
difficult and expensive. Tool-up of a single world-class production line is capital intensive
— $25 million minimum for equipment alone — and can take up to two years to implement.
In fact, the cost can be much more.
Tokyo-based Mitsubishi Rayon Co. Ltd.’s (MRC) 9.4 million ft²/874,000m² Otake production
facility, for example, is slated for a $100 million, three-year expansion — a production line
that could annually produce as much as 20 million lb/9,072 metric tonnes of carbon fiber.
This goes a long way toward explaining why, historically, it has been difficult to avoid the
imbalances between supply and demand that cause prices to plummet and peak. Little
wonder, then, that the current cadre of carbon fiber producers numbers less than a dozen
worldwide (see chart at left).

Related Stories

HPC, with the help of several carbon fiber process suppliers, recently peeked behind the
veil of secrecy to find this more inclusive, if still incomplete, picture of the process.

 
A definitive difference

Unlike metals, which are homogeneous and, by design, have properties that conform to
established standards, making each producer’s P20 steel, for example, interchangeable
with another’s, composites are heterogeneous. Composed of combinations of unlike
materials (fiber and resin), their variability, and therefore, tailorability, are central to their
appeal. Accordingly, carbon fiber producers make products that are similar but not
identical. Carbon fiber varies in tensile modulus (or stiffness determined as deformation
under strain) and tensile, compressive and fatigue strength. PAN-based carbon fiber is
available today in low modulus (less than 32 million lbf/in² or <32 Msi), standard modulus
(33 to 36 Msi), intermediate modulus (40 to 50 Msi), high modulus (50 to 70 Msi) and
ultrahigh modulus (70 to 140 Msi). Fiber, which is available in bundles called tow, comes in
many sizes, ranging from 1K to 350K (1K equals 1,000 filaments that range from 5 to 10
microns in diameter). Products also vary in the degree of carbon content and type of
surface treatment/coating.

“The complexity inherent in carbon fiber composites is the very thing that adds value to
structures made from carbon fiber,” says Steven Carmichael, director of sales and
marketing for MRC subsidiary Grafil Inc. (Sacramento, Calif). “Like making fine wine, the
right amount of patience, finesse and processing expertise brings out the subtleties in
carbon fiber that add value.” That value, of course, is very high: As a metal replacement,
carbon fiber composites offer 10 times the strength of steel at half the weight.

In simplest terms, carbon fiber is produced by pyrolysis of an organic precursor fiber in an

inert atmosphere at temperatures above 982°C/1800°F. Carbon fiber manufacture,
however, is a complex undertaking. Grafil’s 60,000 ft²/5,574m² plant in Sacramento, Calif.
— small in comparison to MRC’s Otake plant, even after its 2 million-lb/544-tonne capacity
expansion in 2005, with side-by-side production lines — provided the basis for HPC’s walk-
through of the primary production phases. These are polymerization and spinning,
oxidation (also referred to as stabilization), carbonization (sometimes inaccurately
referred to as graphitization), surface treatment and sizing application. Throughout the
process, tight tolerances define the fiber’s ultimate utility. “A target coefficient of variation
on yield is 1 percent,” says Gordon Shearer, Grafil’s operations director, noting that real-
world variation runs about 3 percent for the small tow (1K to 24K) that is used in
demanding applications, such as aircraft (hence, its designation as aerospace grade),
while large tow (industrial or commercial grade) can vary up to 15 percent.

Polymerization

The process begins with a polymeric feedstock known as a precursor (“that which comes
before”), which provides the fiber’s molecular backbone. Today about 10 percent of
produced carbon fiber is made from a rayon- or pitch-based precursor, but the majority is
derived from polyacrylonitrile (PAN), made from acrylonitrile, which is derived from the
commodity chemicals propylene and ammonia.

For that reason, this article depicts production of PAN-based carbon fiber.

Converting PAN into carbon fiber has challenged producers for more than 30 years.
Carmichael adds that most of a carbon fiber producer’s investment is spent on precursor,
and the quality of the finished fiber is directly dependent on that of the precursor.
Specifically, Shearer notes, attention to precursor quality minimizes variation in the yield,
or length per unit of fiber weight.

Generally, precursor formulation begins with an acrylonitrile monomer, which is combined

in a reactor with plasticized acrylic comonomers and a catalyst, such as itaconic acid,
sulfur dioxide acid, sulfuric acid or methylacrylic acid. Continuous stirring blends the
ingredients, ensures consistency and purity and initiates the formation of free radicals
within the acrylonitrile’s molecular structure. This change leads to polymerization, the
chemical process that creates long-chain polymers that can be formed into acrylic fibers.

The details of polymerization, such as temperature, atmosphere, specific comonomers

and catalyst are proprietary. According to Peter Morgan, author of Carbon Fibers and Their
Composites (CRC Press, 2005, [Link]), “polymerization should achieve at
least 85 percent acrylonitrile content and relative molecular weight of 100,000 g/mole
with uniform distribution in order to imbue PAN white fiber with good mechanical
properties.” MRC’s precursor as used by Grafil, for example, achieves 94 to 98 percent
acrylonitrile content.

After washing and drying, the acrylonitrile, now in powder form, is dissolved in either
organic solvents, such as dimethyl sulfoxide (DMSO), dimethyl acetamide (DMAC) or
dimethyl formamide (DMF), or aqueous solvents, such as zinc chloride and rhodan salt.
Organic solvents help avoid contamination by trace metal ions that could upset thermal
oxidative stability during processing and retard high-temperature performance in the
finished fiber. At this stage, the powder-and-solvent slurry, or precursor “dope,” is the
consistency of maple syrup. The choice of solvent, and the degree to which the dope’s
viscosity can be controlled (by means of extensive filtration), are critical to the success of
the next phase, fiber formation.

Spinning

PAN fibers are formed by a process called wet spinning. The dope is immersed in a liquid
coagulation bath and extruded through holes in a spinneret made from precious metals.
The spinneret holes match the desired filament count of the PAN fiber (e.g., 12,000 holes
for 12K carbon fiber). This wet-spun fiber, relatively gelatinous and fragile, is drawn by
rollers through a wash to remove excess coagulant, then dried and stretched to continue
the orienting of the PAN polymer. Here, the filament’s external shape and internal cross-
section are determined by the degree to which the selected solvent and coagulant have
penetrated the precursor fiber, the amount of applied tension and the percentage of
filament elongation. The latter is proprietary to each producer, but Morgan asserts that the
stretch rate can be up to 12 times the initial pliability of precursor fiber.

An alternative to wet spinning is a hybrid process called dry jet/wet spinning, which uses a
vertical air gap between the fiber and coagulate bath. This creates a smooth, round PAN
fiber that can enhance the fiber/matrix resin interface in composite materials.

The last step in PAN precursor fiber formation is the application of a finishing oil to
prevent the tacky filaments from clumping. The white PAN fiber then is dried again and
wound onto bobbins.

Oxidation

These bobbins are loaded into a creel that feeds the PAN fiber through a series of
specialized ovens during the most time-consuming stage of production, oxidation. Before
they enter the first oven, the PAN fibers are spread flat into a tow band or sheet referred to
as warp. The oxidation oven temperature ranges from 392°F to 572°F (200°C to 300°C).
The process combines oxygen molecules from the air with the PAN fibers in the warp and
causes the polymer chains to start crosslinking. This increases the fiber density from
~1.18 g/cc to as high as 1.38 g/cc.

To avoid runaway exotherm (the total exothermic energy released during oxidation,
estimated at 2,000 kJ/kg, poses a real fire hazard), oven manufacturers use a variety of
airflow designs to help dissipate heat and control temperature (see sidebar, below). Matt
Litzler, president of C.A. Litzler Co. Inc. (Cleveland, Ohio), observes that “every precursor
has its own exothermic pattern. Since individual precursor chemistry is fixed, control of
temperature and airflow in the oxidation oven is adapted to each precursor and provides
stabilization of the exothermic reaction.”

Oxidation time varies, driven by specific precursor chemistry, but Litzler estimates that
24K tow could be oxidized at about 43 ft/13m per minute on a large production line with
multiple oxidation ovens. Randy Strop, general manager for oven manufacturer Despatch
Industries (Lakeville, Minn.), says an elapsed time of 60 to 120 minutes is typical, as are
four to six ovens per production line, with ovens stacked to provide two heating zones that
offer 11 to 12 passes of the fiber per oven.      In the end, the oxidized (stabilized) PAN
fiber contains about 50 to 65 percent carbon molecules, with the balance a mixture of
hydrogen, nitrogen and oxygen.

Carbonization

Carbonization occurs in an inert (oxygen-free) atmosphere inside a series of specially

designed furnaces that progressively increase the processing temperatures. At the
entrance and exit of each furnace, purge chambers prevent oxygen intrusion because
every oxygen molecule that is carried through the oven removes a portion of the fiber,
explains Robert Blackmon, VP of the Process Systems Div. at furnace source Harper
International (Lancaster, N.Y.). This prevents loss of the carbon produced at such high
temperatures. In the absence of oxygen, only noncarbon molecules, including hydrogen
cyanide elements and other VOCs (generated during stabilization at concentration levels
of 40 to 80 ppm) and particulate (such as local buildup of fiber debris), are removed and
exhausted from the oven for post-treatment in an environmentally controlled incinerator.
At Grafil, carbonization begins in a low-temperature furnace that subjects the fiber to
1292°F to 1472°F (700°C to 800°C) and ends in a high-temperature furnace at 2192°F to
2732°F (1200°C to 1500°C). Fiber tensioning must be continued throughout the production
process. Ultimately, crystallization of carbon molecules can be optimized to produce a
finished fiber that is more than 90 percent carbon. Although the terms carbon and graphite
are often used interchangeably, the former denotes fibers carbonized at about
1315°C/2400°F and that contain 93 to 95 percent carbon. The latter are graphitized at
1900°C to 2480°C (3450°F to 4500°F) and contain more than 99 percent elemental
carbon.

The number of furnaces is determined by the modulus desired in the carbon fiber; part of
the relatively high cost of high- and ultrahigh-modulus carbon fiber is due to the length of
dwell time and temperatures that must be achieved in the high-temperature furnace. While
dwell times are proprietary and differ for each grade of carbon fiber, oxidation dwell time is
measured in hours, but carbonization is an order of magnitude shorter, measured in
minutes. As the fiber is carbonized, it loses weight and volume, contracts by 5 to 10
percent in length and shrinks in diameter. In fact, the demonstrated conversion chemistry
ratio of PAN precursor to PAN carbon fiber is about 2:1, with less than 2 percent
permutability — that is, considerably less material exits the process than goes into it.

Surface treatment and sizing

The next step is critical to fiber performance and, apart from the precursor, it most
differentiates one supplier’s product from its competitors’ product. Adhesion between
matrix resin and carbon fiber is crucial in a reinforced composite; during the manufacture
of carbon fiber, surface treatment is performed to enhance this adhesion. Producers use
different treatments, but a common method involves pulling the fiber through an
electrochemical or electrolytic bath that contains solutions, such as sodium hypochlorite
or nitric acid. These materials etch or roughen the surface of each filament, which
increases the surface area available for interfacial fiber/matrix bonding and adds reactive
chemical groups, such as carboxylic acids.

Next, a highly proprietary coating, called sizing, is applied. At 0.5 to 5 percent of the weight
of the carbon fiber, sizing protects the carbon fiber during handling and processing (e.g.,
weaving) into intermediate forms, such as dry fabric and prepreg. Sizing also holds
filaments together in individual tows to reduce fuzz, improve processability and increase
interfacial shear strength between the fiber and matrix resin. Carbon fiber producers
increasingly use a sizing appropriate to the customer’s end use (see sidebar, below and
“Advances in sizings and surface treatments for carbon fibers” in “Editor's Picks,” at right). 
At Grafil, Carmichael adds, “we can customize surface treatment and sizing to a particular
customer’s resin characteristics, as well as specific properties desired in the composite.”

According to Andy Brink, cofounder of the former Hydrosize Technologies (Raleigh, N.C.),
now part of Michelman (Cincinnati, Ohio), which he serves as business development
manager, “Polymeric film formers made by the dispersion of particles suspended in water
provide a stable chemistry that creates a good coating when dried. The speed of most
carbon fiber lines allows for fairly uniform sizing application that minimizes aggregate
clumps or bare spots.”

When the sizing dries, the long process is complete. Grafil (as do other suppliers)
separates individual tows out of the warp and winds them onto bobbins for shipment to
customers, including prepreggers and weavers.

If an industry’s history serves as a precursor of its future, the sheer magnitude of

machinery and manufacturing acumen required for the successful transformation of white
PAN fiber into black carbon fiber suggests that producing this advanced material is not a
business for the faint of heart or the inexperienced. Three decades of processing
refinement have brought technology maturity and the ability to translate superior
performance and application versatility through the fibers to advanced composites. What
has gone before both technologically and economically sets the stage for the potential
growth in demand that marks the future.
Trends in carbon fiber manufacturing

Technological changes have made carbon fiber available to and more practical for
use by OEMs in a wide range of markets and applications. Suppliers of sizings and
thso who build the ovens and furnaces by which pyrolisis is accomplished recently
outlined some of the more significant developments for HPC.

Sizing for many matrices

Because most carbon fiber, historically, has been used with epoxy matrices, sizing is
predominantly epoxy-based and low in molecular weight to encourage fiber pliability
and spreadability. However, research is underway to create sizings with chemistries
that suit the variety of matrix resins now in demand for end-use applications.

Hydrosize Technologies (Raleigh, N.C., which was purchased by Michelman of

Cincinnati, Ohio in June) offers 17 carbon fiber sizings commercially, though business
development manager Andy Brink indicates the company can create sizing for
multiple fiber/matrix combinations as well as different composite manufacturing
processes. He says that almost all the company’s sizings are free of solvents and
hazardous air pollutants. One example is Hydrosize U601, a high molecular weight
urethane sizing that reportedly improves both fiber wetout by urethane resins and
lubricity (decreased friction during handling) with an environmentally friendly
formulation.

At the Adherent Technologies R&D facility in Albuquerque, N.M., Ronald Allred,

president, has been involved in scientific analysis of fiber/matrix interface issues for
35 years. He reportedly has discovered a reactive chemistry that affects carbon
molecules in the fiber to improve interfacial bonding between fiber and matrix. “We
are currently using this chemistry to improve interfacial adhesion in carbon fiber/vinyl
ester and carbon fiber/bismaleimide for the U.S. Navy, including in composite engine
components for the F-35 Joint Strike Fighter.” He notes that prior to using Adherent’s
vinyl ester sizing, “the Navy was seeing only 50 percent of the theoretical properties
desired, due to poor shear transfer at the interface. Similarly, high-temperature
composites can suffer from poor oxidative stability with sizings not formulated to
match the requirements of matrix resin properties.”

New-generation ovens, furnaces target greater efficiency

In the production of carbon fiber, much depends on the design of the ovens and
furnaces that pyrolize the fibers.

In the oxidation process, oven airflow plays a critical role in controlling process
temperatures and preventing exothermic reactions. Airflow designs may be single
flow (parallel or perpendicular to the tow band) or multipath.

According to Randy Strop, general manager for oven builder Despatch Industries’
(Lakeville, Minn.) Carbon Fiber business unit, three important elements are demanded
by carbon fiber producers in oxidation ovens: throughput, scalability and energy
efficiency. To determine the optimal oxidation oven setpoint for the specific
requirements of carbon fiber producers among its customers, Despatch has tested its
patented center-to-end parallel airflow through temperature gradients measured by 40
different calibrated thermocouples positioned on each side of the oven working zone.
Strop notes that this design allows for higher air velocities — up to 13.1 ft/sec (4
m/sec) at nozzle discharge — and higher air volume than other oven configurations.
This configuration also achieves tight temperature uniformity of ±1°C, side-to-side
average, across the entire width of the tow band. A 25 percent faster oxidation rate in
production scale ovens has been reported by customers.

Despatch offers oven widths from 1 ft to 11.5 ft (0.3m to 3.5m), automated entry and
exit louvers that minimize heat loss and reduce setup time, and recycling of heated air
to reduce total energy usage. The estimated energy savings, compared to legacy oven
designs, on a 6.6-ft/2m-wide opposing automated louvers with a reduced slot
opening of 0.35 inch/9 mm is 10 kW/h.

C.A. Litzler Co. Inc. (Cleveland, Ohio), an oxidation oven builder for 30 years, outfits its
products with multiple temperature zones and a controlled cross flow air path that
delivers air at a rate of 5 ft to 9 ft (1.5m to 2.7m) per second over each tow, resulting
in even treatment of the precursor fiber. Patented End Seals solve what company
president Matt Litzler describes as “the simple physics of the chimney effect, where
cold air enters the lower product slots and hot air escapes from the top slots. This
can create cold spots in the oven and be dangerous for operators. Our end seals
make every slot neutral, reduce the amount of exhaust air needed, and effectively
lengthen the useful oven by eliminating cold air infiltration.”

In addition to its pilot and production ovens up to 10 ft/3m wide, C.A. Litzler also
designs and builds idler rollers, driver rollers and tension stands for fiber stretching.

A supplier of carbonization furnaces since the 1940s, Harper International (Lancaster,

N.Y.) began offering complete carbon fiber production line design and equipment in
the 1990s and turnkey installation of PAN-based lines in 2000. Robert Blackmon, VP
 of the Process Systems Div., points out that the newest generation of wider furnace
systems is more efficient, producing a greater volume of carbon fiber with lower
energy consumption per pound of fiber. Harper offers furnace widths up to 13 ft/4m
and lengths greater than 46 ft/14m, with highly efficient insulation.

Special attention is given to entry and exit purge chambers. Blackmon explains that
every oxygen molecule carried into the system not only degrades the fiber but the
furnace’s graphitic refractory surface as well. “Our purging system reduces oxygen
immigration significantly, which can increase product yield and quality and the
operating life of the equipment.” To boost energy efficiency and reduce production
costs, Harper has designed a recuperative recovery system for thermal oxidative
waste heat. Blackmon admits there’s a cost tradeoff associated with such optimized
environmental controls, but he believes “the energy recovered typically well justifies
the cost.”

“Harper’s furnaces are atmospherically controlled with inert gases, such as nitrogen
or argon, and can operate over the temperature range of 572°F to 5432°F [800°C to
3000°C] for carbon fiber with low to ultrahigh modulus,” states John Imhoff, manager
of marketing and business development. Harper also offers surface finishing and
sizing systems to accommodate different electrolytic and matrix resin chemistries.

Related Content

Composites ride and support th… Lab-scale carbon fiber productio…

Article  •  Increased ridership and heavier Product  •  The new Computreater CF from
axle loads on rail systems around the C. A. Litzler Co. Inc. is the laboratory model