cored boron filaments, which was con­ trary to the effects of higher energy input fusion welds where fiber crack­ ing, break-up and misorientation were common. The effect of poor materials compatibility and high energy input during welding on another type of Fusion Welding of Titanium-Tungsten reinforcing fiber has also been dra­ matically demonstrated in gas tung­ sten-arc fusion welds in a sapphire- and Titanium-Graphite Composites titanium (single crystal A1203) fila­ ment composite.9 The sapphire fila­ ment was completely dissolved in the molten titanium during welding, result­ Reaction products at the reinforcing ing in a coarse, embrittled micro- structure that severely cracked while cooling. Other work where fusion of filament-metal matrix interface the matrix resulted in little or no appreciable damage to the reinforcing are directly related to welding energy input fibers included resistance welding of steel wire-aluminum10 and fusion welding of tungsten wire-aluminum.9 The only other well known related BY JAM ES R. KENNEDY studies on composite fusion welding have been on dispersion-strengthened materials such as TD-Nickel. The effects of fusion in this case are quite detrimental, causing destruction and agglomeration of the fine thoria dis­ persion and a drastic lowering of weld ABSTRACT. Fusion welding experiments strength, light weight and economical joint efficiencies; these problems have were conducted on two titanium matrix structural joints. However, composite composite systems: titanium-tungsten led to emphasis on solid state welding welding remains largely problemati­ techniques.11 wire and titanium-graphite filament. Ob­ cal, primarily because the effects of jectives: Determine the weldability of An investigation was undertaken to weld thermal energies on fibers and model composites and observe the influ­ study the effects of fusion welding on matrix-fiber compatibility for many ence of weld thermal energy on fiber- two metal-matrix composite systems: composites are not yet fully under­ matrix reactions. titanium-tungsten and titanium- stood. Results of mechanized and manual gas graphite. The titanium-tungsten sys­ tungsten-arc (GTA) welding tests indi­ Welding techniques requiring fusion tem is characterized by the absence of cated that well diffusion-bonded com­ of the matrix present the most appar­ compound formation and the forma­ posites generally presented no unusual ent critical condition because the rein­ tion of a two-phase region. Converse­ problems during fusion. The extent of forcing fibers will be exposed to a ly, the titanium-graphite system is fiber-matrix reactions in both systems superheated molten matrix and some­ was directly proportional to the welding characterized by matrix-fiber reac­ times an intense heat source (e.g. an tions that can lead to interfacial com­ energy input. As energy input increased, electron beam or electric arc). This tungsten wire dissolution became greater pounds. and titanium carbide formation around can give rise to accelerated reaction rates between matrix and fiber, pos­ The objectives of this work were to the graphite filaments grew thicker. obtain weldability information on sim­ Welding energy input thus becomes a sibly leading to extensive interdif­ fusion, fiber dissolution or complete ple model composite systems that significant factor in controlling the would also be generally useful in the nature of fiber-matrix reaction products. fiber destruction. Obviously the extent Tensile properties of titanium-tungsten of matrix-fiber interaction depends development of other composites and composites, both as-diffusion bonded and upon their basic compatibility and the to increase understanding of fiber- as-welded, are compared. weld thermal history to which they matrix reactions by observing com­ are subjected. posites exposed to the dynamic ther­ mal conditions of fusion welding. This Introduction Previous work on composite joining paper describes the fabrication and has been generally quite limited, espe­ welding of composite specimens, met­ The great potential of composite cially in the area of fusion welding. In materials in providing improved 6 allographic observations and the .ten­ a study to determine joining tech­ sile properties of composite-Velds. mechanical properties has led to the niques suitable for boron fiber- realization that the extent of their aluminum composites it was con­ utilization as structural members de­ cluded that gas tungsten-arc, electron Experimental Procedures pends on how easily they may be beam and plasma fusion welding re­ Materials fabricated and joined into complex 15 sulted in severe weld embrittlement Tungsten wire used in this investi­ shapes. The possibility of welding and fiber degradation. Conversely, gation was 0.008 in. commercially metal-matrix composite systems resistance welding of these composites pure lamp filament. Typical tensile should be an important consideration has been relatively successful in pro­ properties in the as-drawn and cleaned because of the need to fabricate high ducing good quality spot welds with condition were 446 ksi tensile strength 6-8 reasonably high strengths. and 4.9% elongation. After heat Interestingly, the molten aluminum treating at 1600F for 1 hr in vacuum, JAMES R. KENNEDY is with the Ma­ terials Research Group, Grumman Aero­ matrix of the spot welds had no ap­ the tensile strength was 328 ksi with space Corp., Bethpage, N. Y. parent adverse effects on the tungsten- an elongation of 1.9%.

250-s I MAY 19 72 The graphite filaments were Graphite was used in the as-received longitudinal to the fiber direction. uncoated "Courtaulds HM," a high condition. Butt welded joints consisted of either modulus and high strength grade pro­ Before hot pressing, the composite machined edges, in which the edges of duced from a polyacrilic-nitryal assembly was enveloped in a layer of both the matrix and the fiber ends (PAN) precursor. The filaments were 0.005 in. thick stainless steel foil to were coplanar, or of joints in which about 8 turn (0.0003 in.) in diameter protect against contamination from the matrix edges were removed a few and were spooled as a continuous the hot press graphite rams. The foil mils by chem milling to provide fiber length in an unwoven yarn. Typical was tackwelded around its edges and relief for joint intermeshing. Both top tensile properties were 300 ksi tensile then its outside surfaces were coated and bottom surfaces of the titanium strength and 50 to 60 million psi with a thin slurry of 0.05 y.m alumina composites were manually scraped modulus. powder that acted as a parting agent with a draw file in the vicinity of the The matrix for both groups of com­ between the foil and the graphite fusion zone before welding to mini­ posite specimens was an annealed rams. The composite specimens were mize possible weld contamination. hot pressed at 1000 psi for 1 hr at sheet of commercially pure titanium, 4 Ti75A, in sheet thicknesses up to 1600F under IO" mm Hg vacuum. The fusion welding operation was 0.017 in. After hot pressing, the foil layers were performed by manual and mechanical removed and the composite specimens techniques, while argon inert gas Composite Fabrication were chemically milled to remove a 5 shielding procedures generally recom­ mended for titanium were used. Typi­ Composite specimens containing the mil surface layer from each side. cal manual welding parameters for tungsten wire or graphite filaments in The composite specimens in this the titanium matrix were made by investigation were prepared and tested 0.055 in. thick composites were 60 diffusion welding in a vacuum hot with their reinforcing fibers aligned amp, 10 volts and 6 in. per min (ipm) press. The titanium-tungsten com­ along the axis of loading (i.e., zero speed. The parameters for mechanized posites were made by wrapping a degree orientation). Fiber volume welding on 0.036 in. thick composites continuous length of tungsten wire percentages of each composite were were 25 amp, 12 volts and 10 ipm around a sheet of titanium, which was approximated during composite speed. then sandwiched with additional ti­ preparation and later were more ac­ tanium sheet. The assembly was resist­ curately determined by metallurgical Testing ance tack welded at the corners to cross sectioning and counting. Room temperature tensile tests de­ facilitate further handling. The graph­ termined the strength of all matrix ite-titanium composites were prepared Welding and composite specimens. Subsize ten­ by placing a portion of the graphite All fusion welded specimens were sile coupons were used in accordance yarn on a titanium sheet, covering it prepared by gas tungsten-arc welding. with ASTM Standards.12 The tensile with another titanium sheet and then Two types of weld specimens were specimens were generally machined in tackwelding the sheets together. tested: square groove butt and simple relation to the fibers, as indicated in Prior to assembly, the titanium was bead-on-sheet. Welds in these experi­ Fig. 1. Composite and matrix tests cleaned with an alkaline degrease and ments were predominantly without were conducted on an Instron testing water rinse and the tungsten wire filler wire additions. Bead-on-sheet machine at a crosshead speed of 0.02 was cloth-wiped in isopropyl alcohol. welds were made both transverse and ipm.

Weld Direction

7^

v

/ •> " v

^a

Fig. 1—Schematic of composite lay-up and specimen orientation

WELDING RESEARCH SUPPLEMENT I 251-s Composite Examination vealed no detrimental effects caused in place. There was also a reaction Radiographic inspection was made by the voids, it was decided to contin­ between the titanium and graphite to of the fiber alignment and condition in ue using the previously established form titanium carbide (about 2 /mi) composite specimens at various stages bonding conditions. wherever they contacted during dif­ of the program. Standard metallo­ The irregular geometric filament fusion bonding. The presence of TiC graphic techniques were used to observe array is a result of the manual wind­ was confirmed in both as-diffusion and measure the effect of the fibers on ing technique used in the composite bonded and fusion welded specimens matrix microstructure, the extent of layup. Although this random array is by X-ray diffraction and microhard­ fiber-matrix reactions and the overfall not considered optimum for mechani­ ness measurements. Peak hardness in influence of welding on the fibers. cal loading because of obvious the carbide region had a Knoop hard­ Characteristics of the fiber-matrix in­ nonuniformities, it was employed here ness (HK) of about 2700 as com­ because it eliminated the need for pared to reported15 values of 2900. terface and reaction zone were ob­ 15 served and measured with a Reichert more costly, precision layups while it The Ti-TiC phase diagram shows P100A metallograph with Knoop mi­ still provided a composite specimen only the formation of the TiC phase, crohardness capability and a Norelco- usable within the scope of this prelimi­ which was observed in these speci­ Philips AMR-3 electron microprobe nary program. mens. analyzer. All Knoop microhardness The interfacial region created be- Attempts to increase graphite vol­ tests were made with a 25 g load. tween the as'-bonded tungsten wire ume in other specimens were general­ Welded titanium-graphite com­ and the titanium matrix consisted of a ly unsatisfactory, resulting in nonuni­ posites were also examined by X-ray band of recrystallized, hardened ti­ form and poorly bonded laminates. diffraction. A Picker Theta-Theta dif- tanium next to a smaller reaction zone This difficulty is attributed to the fractometer employing a copper target (about 4 /xm) immediately adjacent to manual layup technique where the the tungsten. According to phase di­ handling of a relatively large bundle for CuKa radiation was used to fur­ 13 14 ther identify the reaction product. agrams * for the titanium- of unwoven and untaped filaments Fracture surfaces of titanium-tungsten tungsten binary system, the interfacial resulted in thick nonuniform layers. tensile specimens were examined with zone should consist of two solid solu­ Because of their inconsistencies, such tions: alpha titanium and tungsten a Cambridge Stereo Scan Mark II samples proved to be unsatisfactory solid solution. These phases form as a scanning electron microscope. for subsequent welding and tension result of the beta eutectoid decom­ tests. Therefore for this work it wasi position that occurs at about 715C„ Results and Discussion decided to utilize the low volume The characteristic lamellar eutectoid graphite specimens exhibiting good ma­ Diffusion Bonding structure became more apparent after trix bonding only for welding tests In both composite systems, hot as-bonded specimens were heat where the general microstructural pressing resulted in an advanced stage treated for an additional 20 hr at effects of fusion could be observed. of matrix diffusion bonding character­ 870C. During heat treatment, the ti­ ized by total migration of the original tanium-tungsten reaction zone grew to matrix planar boundaries and nearly about 30 ,ixm and displayed a slight, Welding: Titanium-Tungsten complete elimination of prior inter­ decrease in microhardness, an indica­ Fusion welding was readily accom­ facial voids within the grains. tion of a tougher transition material. plished on the titanium tungsten com­ In the titanium-tungsten com­ In the titanium-graphite com­ posites by both manual and mecha­ posites, consolidation of the matrix posites, consolidation of the matrix nized techniques. There was no appar­ was mot complete in regions near the through the graphite filamentary ent unusual behavior observed in arc reinforcing filaments. Typically, voids region did not generally occur. Speci­ stability, puddling characteristics, or remained adjacent to the tungsten mens with low graphite volumes weld bead formation in welding over wires along the prior planar bound­ (0.5-1%) were produced in a way the tungsten wires. Relatively small aries of the matrix. To minimize time that a small center region of filaments increases in welding current (5-10%) and pressure during diffusion bonding was encased with diffusion-bonded ti­ were required compared to the ma­ and since subsequent welding tests re­ tanium that firmly held the filaments trix-only condition. Manual gas tung-

ir* 5 CO oo

\ *'•• % o o i '«. ...ti Fig. 2—Transverse sections of titanium-5% tungsten welds. (Top) mechanized GTA, (bottom) manual GTA fusion line indicated by broken lines. Mag: 25X

252-s I MAY 1972 sten-arc welding proved to be espe­ consideration must obviously be given percentage of fiber dissolution could cially useful in making small energy to various factors such as fiber perhaps be offset by the fact that a input corrections on those composite strength and dissolution, reaction zone homogeneous, intimate interfacial specimens with a nonuniform filament products, matrix alloying and weld bond is produced between matrix and distribution (see Fig. 1). metal uniformity. In the titanium- fiber and that usable matrix strength­ Typical weld cross sections are tungsten binary alloy, which forms ening through solid solution might oc­ shown in Fig. 2. As can be readily only solid solutions, appreciable solu­ cur. Clearly, whether or not welding is observed, matrix consolidation around tion hardening can be realized as a involved, composite strength (

wire which has been reduced about i\ ••• l * * 1.5 mils in diameter as a result of welding. Electron microprobe analyses for both tungsten and titanium distribu­ tions were made on various com­ posites, as shown in Fig. 4. The differ­ ences in the tungsten gradient from the wire into the matrix for each of the three conditions studied (diffusion bonded, manual weld, mechanized weld) are quite distinct and generally correspond to the micrographic obser­ vations. By comparison, the change in the titanium distribution is not so

great. It is clear that during diffusion ( . 7. I i % . *»;.,;• > Is. • | bonding, tungsten appears to be the more mobile species and is readily '\r.j\A I transported into the matrix; melting of the matrix leads to tungsten disso­ lution and matrix enrichment. The tungsten gradient appears uniform around each wire and the extent of dissolution is directly related to weld­ ing energy input. Since the effects of weld heating Fig. 3—Titanium-tungsten weld microstructures. (Top) mechanized, (bottom) manual can be quite significant in composites, Mag: 100X

WELDING RESEARCH SUPPLEMENT | 253-s * * * • ' \ • W0&F*"

t W

Ti

As-Diffusion Mechanized Manual Bonded GTA GTA Fig. 4—Interfacial gradients by electron microprobe analysis. (Top) tungsten into titanium, (bottom) titanium into tungsten. Mag: 1600X general desirability of producing a difficulties were encountered. In gen­ and form, apparently as a function of welded joint in a given structural com­ eral, each of the above fabrication energy input and filament concentra­ ponent. In comparison, most struc­ defects led to localized overheating tion. In one case there were three tural weld design in heat treatable during the welding pass, causing apparent zones: inner (graphite fila­ titanium and aluminum alloys is based "melt-back" in the uppermost matrix ment and TiC), middle (TiC stringers on the annealed properties of the as- sheet, followed by direct exposure of in titanium matrix), outer (TiC parti­ welded condition, with weld metal the graphite filaments to the arc. This cles in titanium matrix). The inner strength losses compensated for by caused weld sputtering with both errat­ zone consists of a continuous carbide weld reinforcement. ic arc behavior and "sparking" from network formed around the graphite incandescent particles of graphite. filaments. Welding: Titanium-Graphite A region from a titanium-graphite In the other condition observed, the When fusion welding was conduct­ transverse weld cross section is shown filamentary region was characterized ed on titanium-graphite specimens in Fig. 5. Typically, titanium carbide by peripheral carbide formations that were uniformly diffusion bonded (TiC) is formed around each filament around each filament, with small ad­ throughout the matrix, the welding as a result of fusion welding. In turn, jacent carbide particles. It appears behavior was basically the same as in the filamentary region is surrounded that the difference in carbide network welding titanium without filaments. In by a matrix that has a carbide dis- between these two cases is related to specimens with incomplete matrix persement of either small stringers or the interfilament spacing, with closely bonding or highly nonuniform graph­ particles. The carbides produced as a spaced filaments favoring continuous ite distribution, however, welding result of welding varied in* amount network formation. The discontinuous

254-s I MAY 19 72 Tensile Test Results In this section the results of room temperature tensile testing, micrograph­ ic and scanning electron microscop­ ic examination are discussed. A sum­ mary of tensile properties is listed in Table 1. A portion of this data is shown in Fig. 7, in which ultimate strength and elongation are plotted against fiber content. The strength at 100% wire content of the rule-of- mixtures (R/M) line is based on indi­ vidually tested wires after heat treat­ ment at 870C for one hour. The composite strength at zero percent wire is the strength of the titanium matrix after about 2% strain, the tungsten wire fracture strain. R/M values for welded composites were not calculated because of the difficulty in separating the tungsten wires from the matrix after welding. At wire contents up to 10%, the diffusion bonded com­ posites fractured above the R/M val­ ues, an indication that the composite did not fail at the wire fracture strain. Fig. 5—Inner filamentary regions of titanium-graphite weld. Mag: 920X This effect has also been observed in composites of tungsten fibers in a copper matrix.16*1T Apparently, be­ low a minimum volume fraction (V ) network (i.e., peripheral carbide Examination of Fig. 6 shows the f formation) is believed to be more of 0.1, there is enough matrix avail­ three reaction zones of each case. As able between the wires to support the desirable for composite reinforcement observed, a thin carbide layer that than the continuous one, since each tensile load, which results in increased formed around each graphite filament matrix deformation and composite graphite filament-carbide "cylinder" in the resistance weld increases in formed is encased in a relatively strain. Interestingly, the fracture thickness as one progresses to the strengths of the welded composites tougher matrix. Generally, the carbide mechanized and manual weld condi­ formed around the filaments was rea­ increase in the same manner as the tions. The higher energy case clearly diffusion bonded values. sonably uniform and continuous along results in greater reduction of filament the filament length. diameter and greater irregularity in There are two primary differences The fact that a continuous cylindri­ both filament and carbide configura­ between these two conditions. First, cal carbide can apparently be formed tion. Obviously, since weld thermal the diffusion bonded composites have around a graphite filament as a result energy can strongly influence matrix- a solid solution reaction zone of only of fusion welding raises the interesting filament reactions, the need arises to 4 /urn around each wire, while the question as to whether reinforcing attain full understanding of this inter­ welded composites possess a cored filaments intentionally produced in relationship for successful composite matrix with higher tungsten content. this manner would have merit in a joining. Second, as expected, the degree of composite. Unique combinations of initial filament and reaction product sagg 7- am not otherwise available might be formed in situ and offer the advantage M of intimate bonding to each other and to the matrix. The reaction kinetics and the resultant product would in part be a function of the weld energy input. The effect of energy input on reac­ tion zone thickness is presented quali­ tatively in Fig. 6, which shows trans­ verse cross sections from resistance spot, mechanized, and manual gas tungsten-arc welds. Although resist­ ance spot welding in titanium-graphite composites is not discussed as part of this work, the micrograph is shown : for comparison. Generally, in terms of . . ' ' -: energy input typically delivered during the fusion process, resistance spot, • ••• • :• •• ••••• •, ••.....,: mechanized, and manual arc welding -•** ^ '* *S****lW&BB&§S*3WBm m may be ranked as low, medium, and Fig. 6—Filament-carbide formation as fu nction of welding process. (Top) resistance high, respectively. welding, (middle) mechanized, (bottom) manual. Mag: 675X

WELDING RESEARCH SUPPLEMENT I 255-s tungsten wire recrystallization ob­ quent size reduction during welding. It individual effects of wire recrystalliza­ served in the welded composites ap­ is not clear at this time what the exact tion, solid solution formation, and ma­ pears greater because of their ex­ effects of higher wire contents would trix alloying and coring have not been posure to molten matrix and subse­ be on composite strength since the determined. The plateaus in composite

150

i_u oo 125 1— ^ oo oO_ o°o° «- LU O h- c_> oo 100

s£

o <

O LU

VOL % WIRE

Fig. 7—Titanium-tungsten tensile properties

256-s I MAY 19 72 Table—1 Titaniium/Tungste n Composite Tensile Data Tungsten Volume Yield Ultimate Elastic Fraction Number Strength6 Strength Elongation7 Modulus % Specimen Description Tested ksi. ksi % 106 psi 0 Annealed sheet, as-received 6 76.7 93.4 24.3 15.3 0 Matrix, as-DB1 3 69.2 88.7 29.0 14.3 3.22 As-DB 3 90.8 98.3 18.2 19.7 4.5 As-DB 3 82.4 102.3 15.8 21.8 9.8 As-DB 2 95.2 103.5 3.4 21.2 17.6 As-DB 2 103.5 107.2 1.2 23.7 0 DB+ butt weld3 3 73.0 92.8 17.5 16.7 4.4 DB + butt weld 3 80.9 101.5 11.7 17.0 0 DB + bead-on-sheet weld4 2 82.4 101.6 14.0 15.5 4.5 DB + bead-on-sheet weld 2 106.8 129.6 4.5 20.3 5.2 DB + bead-on-sheet weld5 5 124.8 138.7 2.4 18.8 9.9 DB + bead-on-sheet weld 2 105.9 131.3 4.0 17.2 19.4 DB + bead-on-sheet weld 2 119.8 132.0 1.7 20.8 100 Tungsten wire, 1600F/1 hr 6 328.0 1.9

i DB = diffusion bonded; 2 all composites tested at 0 deg orientation; ' no filler—butt welds tested transverse to load axis; 1 no filler—welds longitudinal to load axis; ' mechanized GTA weld—all others manual GTA; ' 0.2% offset; 7 1 in. gage length ultimate strength exhibited in Fig. 7 Typical cross sections of these frag­ about a 0.1 volume fraction, the wire may be partially explained by the mented regions are shown in Fig. 8. fragmentation did not lead to crack wire-wire contact within the matrix as As observed, multiple microcrack ini­ propagation in the tough matrix, but shown in Fig. 2. It is known18* 1!l tiation sites were distributed along the rather only some very localized tear­ that this can cause high stress concen­ tungsten wires, with the cracks form­ ing near those zones and matrix flow trations, thereby decreasing fiber load- ing at the wire edge on planes 45 around the wire ends. Above 0.1 Vf, carrying ability. deg to the axis of loading. The num-j there was much less fragmentation Radiographic and optical examina­ ber of partial cracks along the wire observed in the tungsten wires and tion before and after tensile testing was greater than the through cracks, composite ductility dropped to about indicated that localized fragmentation which also exhibited 45 deg fracture the level for individual wires. of the tungsten wires, away from the planes. The degree of interfacial bond­ As noted in Ref. 20, composite final fracture surface, occurred within ing between the matrix and wire fracture probably occurs almost si­ the homogeneous matrix during ten­ appeared quite high in these fracture multaneously with wire fracture since sile loading. In fact, during tensile regions; no debonding was noted. Al­ there is not enough matrix available to loading wire fracture was in some though not definitely ascertained, carry the load previously supported by cases audible, best described as a there were regions along the wires the wires. It has been suggested21 that sequence of pinging or clicking sounds. exhibiting possible necking. Below as the random fiber breaks occur at points of weakness, the composite Final fracture path starts behaving as a discontinuous fiber reinforced system in which loads are transmitted by matrix shear stresses. In this manner the random breaks persist until they finally localize in one plane (i.e., the final fracture surface). In considering the discontinuous fiber case, the average length, 1, of the fragmented tungsten wires, away from the final fracture surface (shown typi­ cally in Fig. 8), was 0.066 in. in as-diffusion-bonded specimens. An ap­ proximation of the critical fiber length I,, was made using the relationship18

lc = °f

where at is the fiber ultimate strength, T the matrix shear strength, and d the fiber diameter. With r taken to be one- half of the matrix ultimate strength (cr„,/2), 1(. is calculated to be 0.03 in. Thus, since we have 1 > 1,. for this case and since no matrix debonding Fig. 8—Tungsten wire fractures in titanium matrix. (Top) Radiograph showing frag­ was observed, it appears that the tung­ mentation and tensile fracture area, 8X; (center) arrows point out possible necking sten breaks within the matrix confirm region, 25X; (bottom) same material at 100X the apparent effective loading of the

WELDING RESEARCH SUPPLEMENT | 257-s fibers and the good bonding achieved wires along the final fracture plane lieve, to be followed by wire/matrix between fiber and matrix. was typical in all conditions, as shown debonding or shear failure, as the In the tensile fracture surface of a in Fig. 10. The debond region may be matrix strains to fracture. manually welded Ti-9% W specimen, generally characterized as having con­ A subsequent optical micrograph as shown by the scanning electron siderable peripheral fracture in the showing this effect is given in Fig. 11. micrograph in Fig. 9, reduced diameter tungsten wire (probably through lo­ In this longitudinal view of the tung­ tungsten wires, weld porosity and dim­ calized recrystallization areas) imme­ sten wire near the fracture surface, it ple rupture surface of the relatively diately adjacent to the prior matrix- appears that the reaction zone be­ ductile matrix are readily seen. wire interface. Ultimate tensile frac­ tween the matrix and wire is still Debonding between the matrix and ture of the tungsten wires is, we be- intact. Fracture occurs within the wires first, apparently, along 45 deg planes, followed by separation of the tungsten into longitudinal segments. This may indicate that the titanium- tungsten interfacial solid solution is strong enough to resist shearing there­ by enhancing composite strength. Joint Preparation Some consideration was given to the problem of fiber discontinuity that arises in attempting to produce butt s joints in composites. It has been point­ ed out10 that a planar interface will be created if two flush ends of a composite sheet are butted. A possible solution for improving joint efficiency consists of intermeshing protruding fibers before welding to provide fiber overlap and continuity. The results of a preliminary attempt are shown in Fig. 12, where a Ti-10% W composite was chemically milled to remove 10 mils from the edges and subsequently butt welded. The radiograph shows the overlap zone after manual fusion welding without filler wire addition. Tensile testing of this type of butt- weld specimen resulted in fracture Fig. 9—Fracture surface of Ti-9% W composite-manual GTA welded specimen, as occurring outside the weld-joint area. shown by scanning electron micrograph—SEM On butt-welded composites that were not intermeshed, the use of a slight amount of weld metal reinforcement also caused fracture away from the weld zone. It is obvious that intermeshing rein­ forcing fibers is a limited practical approach to increasing joint efficien­ cy, the degree of difficulty being pro­ portional to volume content and com­ ponent size. Typical problems could include difficulty of meshing, fiber damage and fiber contact. In view of these factors, however, the approach may be viable in certain cases. Conclusions The general weldability of two ti­ tanium matrix composites has been demonstrated on a limited basis. Up to about 0.2 volume fraction, the ti­ tanium-tungsten wire composites dis­ played reasonably normal welding characteristics. The titanium-graphite filament composites were also welda­ ble but only with lower filament vol­ ume fractions and where the matrix laminates were well diffusion-bonded. It appears that on a broader basis these results might be applied to more Fig. 10—Interfacial debonding and fracture region in Ti-9% W manual weld (SEM) general categories of metal-matrix

258-s I MAY 1972 composites: the cases where mutual appears that many undesirable effects evaluation of these effects and their dissolution occurs between fiber and could be minimized. contributions to composite efficiency matrix and where intermediate phase Probably the most important con­ will be necessary for practical utiliza­ (or compound) formation occurs be­ sideration in composite welding will tion in future applications. Consider­ tween fiber and matrix. The effects of be the effect of welding energy input able welding research will also be welding on a composite and its result­ on matrix-fiber reaction products. required on all structural composite ant reaction products are to a great Generally, higher welding energy will systems, especially those with high extent controllable. The heat of weld^ increase dissolution between com­ volume fractions of reinforcing fibers. ing accelerates interactions between ponents, produce more extensive dif­ Particular emphasis on expected prob­ matrix and fiber but through judicious fusion zones and create larger quanti­ lem areas such as joint design with selection of process and parameters it ties of additional phases. Quantitative fiber continuity, compatible filler ma­ terials, fiber weld damage and matrix weld soundness will be required. References 1. Glasser, J. and Sump, C. H., Evalua­ tion of Manufacturing Methods for Fibrous Composite Materials, Chemical and Metal­ lurgical Research. Inc.. Chattanooga. Ten­ nessee, AFM-TR-65-175, Contract AF 33 (675)-8741, June 1965. 2. Jackson, C. M. and Wagner. H. J., Fiber-Reinforced Metal-Matrix Composites: Government-Sponsored Research, 1961,-1966. DMIC Report 241, Defense Metals Infor­ mation Center, Battelle Memorial Institute. Columbus, Ohio, September 1967. 3. Dallas. R. N.. Methods of Joining Advanced Fibrous Composites, Composite Materials: Testing and Design, ASTM STP 460, American Society for Testing and Ma­ terials. 1969, pp. 381-392. 4. Metzger, G. E., Metals Eng. Quart., Vol. 10, No. 1, February 1970. pp. 55-58. 5. Lovelace, A. M. and Tsai, S. W., Astronautics and Aeronautics, July 1970, pp. 56-61. 6. Schaefer. W. H. and Christian, J. D., Evaluation of the Structural Behavior of Filament Reinforced Metal Matrix Com­ posites, General Dynamics, Convair Div., San Diego. California. AFML-TR-69-36 Vols. I, II, III, January 1969. 7. Hersh. M. S., WELDING JOURNAL, Vol. Fig. 11—Fractures in tungsten wire near tensile fracture surface (broken line) frac­ 47, No. 9, 1968, pp. 404-409S. ture surface; (arrow) Ti-W interface. Mag: 425X 8. Hersh. M. S., WELDING JOURNAL, Vol. 49, No. 6. 1970, pp. 254-258S. 9. Kennedy. J. R. unpublished work, 1970. 10. Metzger. G. E., Welding of Metal- Matrix Fiber-Reinforced Materials, AFML- TR-68-101, Air Force Materials Laboratory, Wright-Patterson Air Force Base, Ohio, September 1968. 11. Moore, T. J.. Solid-State Welding of Dispersion-Strengthened Materials, Aero­ space Structural Materials—Conference Proc. NASA Lewis Research Center, NASA SP-227, November 1969, pp. 119-130. 12. Tension Testing of Metallic Mate­ rials, ASTM E8-68. ASTM Standards—Part 31, ASTM, Philadelphia, Pennsylvania, May 1969, pp. 202-221. 13. Hansen, M., Constitution of Binary Alloys, McGraw-Hill, New York, 1958. 14. English, J. J.. Binary and Ternary Phase Diagrams of Columbium, Molybde­ num, Tantalum, and Tungsten, DMIC Re­ port 152, Battelle Memorial Institute, Co­ lumbus. Ohio, 1961. 15. Storms, E. K., The Refractory Car­ bides, Academic Press, New York, 1967. 16. Signorelli, R. A. and Weeton, J. W., Metal-Matrix Fiber Composites for High Temperatures, Aerospace Structural Mate­ rials—Conference Proc, NASA Lewis Re­ search Center, NASA SP-227, November 1969, pp. 187-205. 17. Ustinov, L. M.. "Tensile Strain in Fibrous Copper/Tungsten Composites," Metallovedenie i Term. Obrabot. Metallov, Vol. 11. 1969. pp. 10-12. 18. Kelley, A. and Davies. G. J., Met. Rev., Vol. 10, No. 35, 1965. pp. 1-78. 19. Tressler, R. E. and Moore; T. L., Met. Eng. Quart., Vol. 11, No. 2, February 1971, pp. 16-22. 20. Tetelman, A. S., Fracture Processes in Fiber Composite Materials, Composite Materials: Testing and Design, ASTM 460, ASTM, 1969. pp. 473-502. Fig. 12—Chem-milled and intermeshed tungsten wire. (Top) as-chem milled, 50X; 21. Cratchley, D., Met. Rev., Vol. 10, (bottom) radiograph after intermeshing and welding, 2X No. 37, 1965, pp. 79-144. ^

WELDING RESEARCH SUPPLEMENT 1 259-s