FABRICATION OF CARBON-CARBON COMPOSITES USING ELECTROSTATIC FIBER DEPOSITION (FLOCKING)

Electrostatic deposition of fibers was investigated as a technique for producing discontinuous carbon-carbon composites and composites containing continuous graphite filaments. Parts were densified by either liquid or gas infiltration. Thermal expansion anisotropies were related to the fabrication parameters •

Electrostatic deposition or 11 flocking 11 of short, small-diameter carbon fibers was evaluated as a fabrication method for producing carbon-carbon composites. The technique was used to place the fibers between spaced windings and layers of continuous, high-modulus graphite yarn in an alternate 0 and 90-degree pattern.
Hollow cylinders, approximately 4 in.ches in diameter by 12 inches long, were fabricated to approximately a 0.5 inch wall thickness and densified by conventional impregnation-carbonization cycles with coal-tar pitch.
Most of the cylinders broke during densification ·due to the accumulation of anisotropic stresses which exceeded the strength of the material. The highly anisotropic character of the continuous yarns was apparently the cause of the composite anisotropy.
Radial-to-length thermal expansivity ratios of the cylinders ranged from -27.2 to 23.1.
Composite plates, which did not contain continuous filaments; were prepared by inductively charging carbon fibers of different lengths ·and diameters. These flocked plates were densified with pyrolytic carbon by the chemical vapor deposition (CVD) method. Thermal expansivities were measured for directions parallel and normal to. the plates, and the ratios of the thermal expansion coefficients for these two directions varied from 1.01 to 1.40, depending upon the type and size of the fioers used. Some conside~ations were given to a corona-type, singly-charged-wire, continuous flocking system for the fabrication of thicker parts .9nd for .eliminating .the stepwise layering of fibers.

INTRODUCTION
The thermal behavior of composite materials and, in particular, those composed of fibers or fiber-like materials depends upon the characteristics of the individual components and the orientation of these materials in the final composites. This orientation is determined, in most cases, by the initial fabrication or layup technique. If secondary processes, such as compaction, are not used to alter the composite, orientation of the fibrous materials will remain as initially arranged.
A new technique for fiber orientation in carbon-carbon composites has been investigated at the Oak Ridge Y-12 Plant.(a) By this technique (referred to as flocking), small-diameter, short carbon fibers are electrostatically deposited to form a preselected shape.
Application of an electrostatic charge causes the fibers to move from a charged surface to the surface of a mandrel or pattern which is maintained at near ground potential. While in motion, the fibers are oriented so that their major axis is normal to the mandrel surface (ie, parallel to the force. lines of the electric field). If a layer of adhesive is applied to the surface of the mandrel, the fibers will stick and remain oriented in that position.
Repeated cycles of adhesive and fiber application add.thickness to the composite. Primary flock layup and flocking, combined with the filament winding of continuous graphite yarn, were investigated.
In an idealized composite, which contains continuous filaments, the short, flocked fibers would fill the space between the continuous filament yarns and be oriented· perpendi euler to them. The continuous yarn layers would alternate orientation (0 and 90 degrees) between the layers of electrostatically flocked fibers.
In a flocked fiber composite, without continuous filaments, electro-:static deposition would provide a primary fabrication technique for the uniform and, hopefully, controllable .loyup of a campositP.. Ability to control the orientation of discontinuous fibers would provide fabricators with a means of adjusting the properties of a composite, and filament-winding or cloth layup procedures could be supplemented by adding fibers with orientations which would increase interlaminar stren!=)th, redvc;:e porosity between filaments, or reduce expansion and shrinkage anisotropy.
The objectives of this work were to develop fabrication techniques for making carbon-carbon composites which contain small discontinuous fibers in conjunction· (a) Operated for the US Atomic Energy Commission by the Union Carbide Corporation•s Nuclear Division.
with -spaced continuous filament-winding patterns, to densify these composites by liquid impregnation-carbonization cycles with coal-tar pitch, and to evaluate the composite characteristics such as interlaminar strength and anisotropy as they relate to the fabrication. A secondary objective was to evaluate flocking as a primary fabrication technique for discontinuous carbon-carbon composites. The fine-diameter ca'rbon fibers used in the flocked and wound parts were prepared from 0.15-denier-per-filament Villwyte viscose rayon. The rayon tow was chopped to lengths of approximately 12 to 15 mils prior to carbonization to 1,000° C in a. nitrogen atmosphere. The carbonized flock was heat treated to 2,100° C in a nitrogen atmosphere, using an induction-type graphite furnace. Final dimensions of these fibers were, nominally, 2 microns in diameter by 250 microns long. The continuous, high-modulus graphite yarn used for winding was WYG-130 1/2 (Thornef 50). The coal-tar pitch (Type CP-277-15V), used in the dE:msification of these parts, had a low insoluble content and had a softening point. in the 90 to 95° C range.
For the composites which contained only discontinuous fibers, rovings made from continuous WYB-115 1/2 (Thorne! 25) yarn and the viscose rayon (as just described) were chopped. A medium carbon black was mixed with the flocking fibers to fabricate two cylinders. This solid particulate material was added in order to determine its effect on the fabrication and properties of the composite.
Cylindrical solid graphite mandrels were fitted with threaded steel center rods on the.axis ofthecylinder,andthreadedsecuring nuts were counter.sunk into each end of the cylinders. The complete cylinders were then covered with 10-mil-thi ck Grafoil. The center rod of 'the covered mandrel was inserted into a 10-inch lathe chuck, rotated slowly, and sprayed with a 10weight percent dispersion of uncured latex in water prior to the first application of fibers.

Combination of Filament Winding and Electrostatic Flocking
Short, small-diameter fibers and particles were electrostatically deposited onto a surface with a basically conventional system, shown schematically in Figure 1.
In this work, the physical componentsofthesystem were changed so that a quantity of ca.rbon fiber flock was placed on a flat metal surface underneath the horizontally rotating mandrel. The metal plate was positioned approximately two inches from the bottom of the mandrel and charged negatively to between 10,000 arid 30,000 volts, as shown in the photograph of Figure 2. The fibers migrated to the electrically grounded graphite mandrel and stuck to it because of the adhesive (latex) coating. 0 rientation of the fibers tended to be normal to the surface of the mandrel; however, due to the small diameters of the fibers and their attraction to each other, some misalignment and randomness of orientation was experienced.
In this inductive-type flocking field, the fibers were added in layers due to their ability to cover the glue and prevent further thickness buildup without glue additions. The glue physically prevented a layer of fibers from returning to the charged plate after losing their charge. An appropriately spaced graphite yarn was then applied, with tension, in a circumferential orientation to the flocked surface, as seen in Figure 3. The surface was again lightly sprayed with the latex emulsion and coated electrostatically with the carbon fibers. Average layer thicknesses for this type of .system were ab~ut 10 to 15 mils. The part was then removed from the horizontal lathe and in.stalled on a vertical lathe. A sphere winding apparatus (seen in Figure 4) was adapted to wind the mandrel from end to end. The winder was adjusted so that the payout eye was ·off center sufficiently to pass the mandrel center rod or turning lug. This arrangement caused the spaced yarns that spanned the length of the cylinder. to be out of alignment with the axis of the cylinder by about 3 degrees. After completing the step of tension. winding the yarn in this orientation, the part was repositioned in the horizontal lathe for further flockingan.d circumferential winding. In this manner, the desired wall thickness was obtained. The part was dried and carbonized to 1,000° C while still on the mandrel.
After carbonization, the parts were densified by combinations of diluted pitch impregnation, conventional undiluted pitch impregnation, and a newly developed technique of impregnation and carbonization under very high pressure. Dilute pitch impregnations requiredtheevacuationofairfromthe pores of the composite while submerged in the appropriate solution of pitch dissolved in benzene. After evacua~ion, the system was pressurized to atmospheric pressure and the part was removed and dried to constant weight before carbonization to 1,000° C in nitrogen at atmospheric pressure. Impregnations, using undiluted pitch, were carried out by the vacuum-pressure method indicated by the drawing of Figure 5. The combined pressure impregnation-carbonization cycles were carried out in the hot zone (17 1/2" D x 50" L) of a gas autoclave(l) which was adapted for this technique (see Figure 6). These impregnation-carbonization runs were carried out at pressures of approximately 15,000 psi and attained a maximum temperature of 750° C. The system employed an isolation liner to prevent hydrogen contamination of the argon pressurizing gas, and the excess gas volume caused by heat expansion of the pressurizing gas continually purged the container.
To use this technique, the cylinders are initiallyvacuum impregnated with molten pitch ( Figure 5) and sealed under vacuum in a thin-wall (30-mil) stainless steel can with excess pitch on all sides, as Figure 6 indicates for a typical run involving a solid billet. The stainless steel acts as a semipermeable membrane, keeping the hydrocarbon molecules inside until decomposition to carbon and hydrogen is accomplished. The hydrogen then diffuses through the membrane, leaving essentially all the carbon. Pitch coking yields (on a weight basis) in such runs are normally 95 percent of the theoretical carbon content.
The mandrels are removed at this stage of the densification by cutting the continuous yarn covering from each end of the mandrel. This step was accomplished after sufficient density had been achieved for shape retention, usually 0.7 to 1.0 gm/cc. After each carbonization cycle, the cylinders are heated to 1,700° C prior to reimpregnation with undiluted pitch in an attempt to fully carbonize all of the organic material and stabilize the dimensions.

Electrostatic Fiber Flocking without Continuous Filaments
It was the initial intent to densify the composites which contained only flocked fibers by pitch impregnation and carbonization, but the low strength and density 135256 Figure 4. CONTINUOUS YARN WINDING APPARATUS.
(The Mandrel has been Wrapped in White Paper for the Picture) (< 0.1 gm/cc) of the as-flocked, unimpregnated, carbonized parts made it inadvisable. Instead, several flocked fiber parts were made and densified to "' 1.4 gms/cc by infiltration with pyrolytic carbon, using the thermal gradient deposition process. (2, 3) The infiltration with methane was at a nominal temperature of 1,150° C and one atmosphere pressure. These parts were fabricated as flat plates (8" x 8" x 0.75" T) and were produced by inductive up flocking. In this way, the fibers were spread on a metal tray under a flat-plate mandrel which had previously been sprayed with a small amount of the aqueous latex suspension. The fibers were charged, causing them to migrate to the surface of the grounded mandrel. Subsequent layers were added after successive glue applications. 1 he parts were dried and carbonized to 1,150° C prior to densification by infiltration with pyrolytic carbon. Table 1 presents data on the cylinder sizes, yarn spacings, number of layers, and fiber types for the parts fabricated. In the table, a "layer" consists of one filament-wound layer in either direction, circumferential or longitudinal, and one flocked-fiber layer. The method used for densification of the individual cylinders, the final density achieved, and the results of densification are presented in Table 2.1t can be noted that various densification plans were used in an attempt to achieve a high density, but all the methods resulted in highly stressed parts as evidenced by de laminations and two full-length radial cracks 180 degrees from each other. Cracks initiated at the cylinder inside wall and progressed to the outside circumference. These stresses were developed during carbonization of the pitch impregnant and appear to be caused by anisotropic shrinkage due to the filament orientation. This same type of cracking has been observed in moderately anisotropic-sprayed,(4) carbon-fiber composites in which the ratios of the radial-to-axial expansion coefficients were in the range of 2.0. Figure 7 is a radiograph of the typical 180-degree radial failure cracks in a sprayed fibrous composite caused by anisotropic shrinkage during densification. Residual strain data on a similar part indicated that the inside wall surface was in tension and the outside wall surface was in compression by about the same magnitude. This condition would produce the type of crack shown in Figure 7 if the interlaminar strength was sufficient to prevent a delamination-type failure. Residual strain of this nature gives an overall effect of ring closure when a hoop is cut radially. Residual stresses which generate a compressive force on the inside surface and a tensile force in the outside surface combine to produce a ring opening (1) Prepared from a rayon precursor and heated to 2, 100° C.

Combination of Filament Winding and Electrostatic Flocking
(2) All cylinders wound with Thornel 50 graphite yarn.
(3) Fibers contained 400 grams of thermal black carbon per 100 grams of fibers.
when a hoop is cut radially. In the sprayed, discontinuous-fiber composites, such stresses were evidenced in cracking after completion of the carbonization cycle to 1,000° C. In composites which contained continuous filaments, the cracking occurred during graphitization of the cylinder at higher temperatures (1,700 to 3,000° C). It was also noted that the cracking of the composites which were uniformly constructed of spaced filament windings did not begin until considerable density had been achieved, usually in excess of 1.4 gms/cc. This behavior can be attributed to the fact that the wound structures were stronger and could sustain the higher stress level due to the anisotropic shrinkage during carbonization better than the completely discontinuous fiber composites. It also seems that the differential thermal expansion at 1,000° C was not sufficient to overstress the wound and flocked cylinders, whereas the higher temperatures did result in stresses exceeding the breaking strengths. These anisotropic properties also result in stresses which cause delaminations in the thickness dimension, depending upon the value of the interlaminar strength at the time of the stress generation. A radiograph of Cylinder FF -4after removing the mandrel ( Figure 8) illustrates a typical failure. The density was 0.70 gm/cc, and the part had received only dilute pitch impregnations. The cylinder broke radially after the second undiluted pitch impregnation at a density of 1.25 gms/cc. A comparison of the coefficients of thermal expansion (CTE) in two orthogonal directions for each of four cylinders is given in I able 3. I he values indicate that this type of composite is very anisotropic. The radially oriented CTE is through the wall thickness of the cylinders and is influenced, primarily, by the diametral characteristics of the high-modulus yarn. The axial CTE orientation is along the  length of a cylinder and is influenced, primarily, by the axial characteristics of the yarn. A quartztube-typedilatometerwas used for the CTE measurements and was calibrated with two graphite standards.(b) The values in Table 3 may be viewed as reasonable since the available data on high-modulus graphite yarn fibers exhibit a high degree of thermal expansion anisotropy. Measurements (5) show that for Thornel40, in the direction of the filament axis, the coefficient of thermal expansion is negative("'-0.99 X w-6 in/in/° C), while in the transverse (across diameter) it is about 16.7x w-6 in/in/ 0 C. Thorne! 50 probably would exhibit even more anisotropy than Thorne! 40. From Table 3 it would be expected that the 19 yarns/inch of Cylinder FF-8 would influencetheCTE more than the 10 yarns/ inch of Cylinder FF-9. However, Table 1 shows that Cylinder FF-9 contained about 50 percent more layers in the same thickness, thus the amqunt of continuous Thorne! 50 yarn is approximately the same in each composites.
(b) ·Types AXM-9Q and AT JS graphite were supplied by the Los Alamos Scientific Laboratory, subsequent to certifying the expansion on a 30-inch length of material. In comparing the next two composites listed in Table 3 (Cylinders FF-22 and FF -23), Cylinder FF -22 is constructed with continuous yarn spaced at 20 filaments per inch and has thinner layers than Cylinder FF-23. The coefficients of thermal expansion for these two composites show that the composite with the most yarn (FF-22) has a higher radio! coefficient and that the radial-to-length coefficient ratio is approximately 47 percent higher than for Cylinder FF-23. It can be noted in Tobie 1 that these two composites contained 400 grams of (1) Mea n CTE between 25 and 1,000° C.
(2) Bul k density determined by mercu ry porosim etry at a pressure of 1.8 psia, wh ich fi lls ope n po res la rger than 100 microns in dia meter.
carbon black per 100 grams of fiber. This particulate material was blended with the fibers prior to the electrostatic deposition.
Composites FF-10 and FF-21 (Table3)showthesame trend as Composites FF-8 and FF-9. FF-21 has the closest yarn spacing, but each layer is thicker than for FF-10. The coefficient-of-expansion ratios are about equal. The actual coefficients for Cylinder FF-21 are somewhat lower than for Cylinder FF-10, but thts is probably due to the lower density. Table 1 shows that Cylinder FF-21 contains 30-mil-long fibers rather than 10 mils. This difference apparently had little or no effect on the expansion isotropy of the wound and flocked composites. This conclusion is brought out more conclusively t'or the entirely flocked composites using the two fiber lengths.
Mercury porosimetry data on Cylinder FF -8 after each of two pressure impregnation-carbonization cycles are listed in Table 4. The data show that after the first cycle a large percentage of the porosity was in the form of large-size pores which remained after the second cycle, although the bulk density increased by a large amount (ie, the total porosity decreased). This condition is shown in the polished section of Composite FF-8, Figure 9. Large pores which have their inside surface partially coated with pitch carbonfromthe impregnation-carbonization process still exhibit a large void volume. A circularly wound yarn which has been flattened due to the winding tension in the fabrication stage is seen in Figure 9. Some flocked, short fibers are seen between the circumferential yarn ends and the longitudinal yarn.
Views of an undensified section of a flocked and wound substrate are given in Figure 10 and show quite clearly the random orientation of the flocked fibers.  The extent to which the fiber bundles were impregnated with pitch carbon was 1 also studied ( Figure 11). The carbon from the impregnant is shown as light circular areas around the darker filaments. This well-impregnated bundle is probably a result of the high-pressure (15,000-psi) impregnation-carbonization technique. Cracks appearing in the fiber bundle were caused by the differential in expansion between the fibers and graphite matrix and the anisotropic thermal coefficients ofthefilaments. Figure 12isa scanning electron micrograph of a fractured section of Cylinder FF-8 and shows the random orientation of the small carbon-flocked fibers relative to the continuous filaments, and emphasizes the degree of impregnation of the highly orientated filament bundles.
Product uniformity was a direct result of operator experience and training in the fabrication stage for these flocked and wound cylinders. Cylinder FF-20 was fabricated by several operators, each with a minimum of experience. The result was a cylinder which was badly wrinkled, delaminated, and uneven after the initial carbonization. This cylinder was one of two constructed with 30-millong fibers.
Electrostatic Fiber Flocking without Continuous Filaments F lot plates were fabricated by the induction charging method using each of the fiber types described previously. Data on the number of layers in each plate sample, the average layer thickness, and the dimensions of several specimens are summarized in Table 5. After carbonization and densification with pyrolytic carbon by the chemical vapor infiltration process, the coefficients of thermal expansion of the plates were measured in two orthogonal directions, listed in  Table 6. In the flocking process used for these samples, a large diameter, discrete, untangled fiber would be preferentially oriented with its axis normal to the plane or surface of the part. I he preferential fiber orientation of the spraying and flocking processes are represented schematically in Figure 13, and the CTE measurement directions of Table 6 are defined. The coefficients and ratios on the flocked samples indicate that the type and length of fibers used had a great influence on the ability of the electrostatic field to orient the fibers. Generally, the fibers with the larger diameter (7 J.L) and greater length (30 mils) gave parts with a higher degree of fiber alignment. The coefficients with the grain are increased in these samples due to .both preferential fiber orientation and higher C-axis expansion of the pyrolytic carbon. The parts fabricated from the two-micron-diameter by 10-mil-long graphite fibers exhibited more nearly isotropic thermal expansions, as indicated in Table6.A typical sample prepared by the spray layup process with the simple orientation, as previously defined (Figure 13), is compared in Table 6. It may be noted by a careful study of the data (Table 6) and the graphic representation of the fiber orientation ( Figure 13) that a true random orientation of fibers will give a thermally isotropic composite with the highly oriented pyrolytic carbon matrix. If, however, preferential fiber orientation exists for any composite direction then the pyrolytic carbon matrix deposited on the fibers is also preferentially oriented. Since the individual fiber isotropy (crystal orientation) varies with the type fiber, a wide variation in expansivity can be obtained in a discontinuous fiberpyrolytic carbon system. The high expansivity in the diametral dimension of an anisotropic fiber (16.7 X w-6 in/in/ 0 c fqr Thorne! 40) is comparable in size and direction to the high thermal expansivity of the C-axis of a pyrolytic    (2) If all fibers were aligned perpendicular to a plane being flocked, the WG-direction CTE would be intluenced primarily by the C-oxis expansion of the pyrolytic graphite crystal (largest). This dimension is perpendicular to the flocking direction.

33
composites where measurements were taken primarily across the filament diameters showed only slightly higher values where Thorne! 40 was used instead of carbon filaments. However, measurements of the expansivity of the composites in a direction parallel to or at a slight angle to the filament axis showed that the carbon-yarn composites gave much higher values while the Thorne! 40 yarn composites shrank during the initial part of the measurement temperature cycle. Unidirectional composites of Thorne! 40 graphite yarn and a pyrolytic carbon matrix (7) or an epoxy resin matrix (5) show anisotropic thermal coefficients of expansion.
Flocking the first layer of fibers presented no problems from the standpoint of adhesion to the covered mandrel. Any adhesive that presented a sticky surface to the fibers was apparently adequate. However, the proper type and amount of adhesive which was applied for the second and subsequent layers is critical. Adhesives sprayed on the flocked fiber ends of the previous layer caused the new fibers to orient radially around the slightly longer fibers already in place on the surface. This action disrupted the orientation of the fibers in the layer and caused large lumps on the surface. Fiber-length measurements show a range of fiber lengths from 90 to 270 microns. (8) The problem was resolved by selecting a water-based latex emulsion which presented a minimum adhesion to the new fibers but dried to hold the fibers of the previous layer in place. In fact, it was determined that almost no adhesive was necessary to fabricate the parts except for handling purposes. It was also apparent that the amount of adhesive should be held to a minimum since even the smallest amount of adhesive would prevent the fibers from interlocking from layer to layer and produce an area which would cause voids in the carbonized part. A typical structure produced by two-micron-diameter by 250-micron-long fibers is seen in Figure 14. The part has been carbonized to 1,000° C, but not densified. A definite density variation can be seen because the structure was fabricated stepwise. These variations in density make it nearly impossibl~ toachieve uniform densification, a result that was especially true in the case of demsification with pyrolytic carbon where each fiber serves as a plating substrate. In this process, the low-fiberdensity areas remain as low-density areas, as noted in Figures 15 and 16. These low-density areas between layers produce low-strength parts. Note that fiber ends are primarily visible, and it appears that the fiber orientation in the low-density areas is more aligned than in the high-density areas. The broken section exposes the small fibers as white or dark dots in the circle centers, while the coated fiber ends appear as unblemished circles.
Cylinders of the sizes described earlier were fabricated by this method but were not densified due to the problem with the low-density area between separate applications of fibers. However, preliminary work was done on a continuous flocking method which showed some promise. In this corona-type flocking  approach, a negatively charged wire was placed approximately one inch above a horizontally rotating graphite mandrel and oriented with the major axis along the major axis of the mandrel. The short graphite fibers which had previously been blended with 10 weight percent of a coating grade of particulate epoxy powder was vibrated into the area between the wire and mandrel. The low (5-kv) potential applied to the wire charged the fibers and fine epoxy particulates and deposited them onto the rotating mandrel, but was not sufficient to reattract the material fromthedeposit.Afteran initial adhesive application to the mandrel, a thick fiber deposit could be procluc.ed continuously without further adhesive. The deposited fiber mat was very fragileduringthe fabrication, but strength was improved by heat treating the epoxy-fiber composite. Fiber density was no higher (""' 0.1 gm/cc) than that achieved with the induction charging system, but the substrate was uniform and showed no layering effP.r:t~. ThP. fihr0us substrate was observed to be random in fiber orientation. The deposition rate was somewhat slower than with the induction system. No samples produced by this method were densified.

CONCLUSIONS AND RECOMMENDATIONS
The more important findings and recommendation~ nf thP. invF:'stigation ore: 1. Isotropy was not achieved by the combination of flocking short, low-modulus carbon fibers between 0 and 90-degree-oriented, spaced, continuous highmodulus graphite yarns.

2.
A possible solution to the anisotropy in the flocked and wound system would seem to be the use of a more discrete, individual, anisotropic fiber for flocking so that the effect of the continuous yarn would be normalized.
3. Isotropy is probably the most important factor to be controlled in a fibrous composite of this type where densification shrinkages are encountered as with pitch impregnation and carbonization.
4. Thermal expansion isotropy is directly related to the orientation, type, and amount of fiber used in a composite.
5. The continuous-filament component of the flocked and wound system appears to aid the liquid impregnation since the long voids in the fiber bundle probably facilitate reaching more open porosity with a given pressure.
6. Control of isotropy by fiber orientation was demonstrated in composites which contained only short fibers with different characteristics. -7. Layering or stepwise constru.ction in the flocked composite is a problem especially if densified with pyrolytic carbon.
8. P1 e::.::.ure in-.p•·~gnotion and carbonizotion with ritc:h is a fast, effec;tive way to densify carbon-carbon composites~· especially above a density of approximately 1.5 gms/cc. 9. Increased interlaminar strength probably was not achieved (although not . measured) in the flocked and wound composites since shrinkage stresses caused delaminations. This failure was probably due to the inability of the fabrication method to orient and maintain the small carbon fibers in the composite radial direction. Low-fiber-density areas between layers in the pure flocked composites. caused the interlaminar strength to be low.