Process Development for Diffusion Welding TI-6AI-4V Alloy Excellent weld quality and 100% joint efficiencies are obtained at 1800° F with a pressure of 60 psi time of 180 minutes and an optimum 15 RHR or better surface finish

1Y R. J. REHDER AND D. T. LOVELL

ABSTRACT. A low pressure diffusion weld­ fusion welding would be extensive and well as chemically milled surfaces. In ing technique for joining Ti-6A1-4V al­ costly. Therefore, diffusion welding at the second phase, the quality of dif­ loy is described. Diffusion welding exper­ the lowest possible pressure and at fusion welds produced with the de­ iments were carried out in an argon temperatures just below and above the veloped parameters was evaluated by atmosphere under dead-weight compres­ beta transus was evaluated in this mechanical testing and nondestruc­ sive loading. Process parameters of time, program. The beta transus for Ti- temperature, pressure and surface finish tive quality control techniques. were determined. Diffusion welded joints 6A1-4V is approximately 1825° F. Mechanical testing dealt with the de­ were evaluated by mechanical property Diffusion aids have been successful­ termination of diffusion weld joint testing, metallographic examination, and ly used to produce joint properties tensile and fatigue properties. Non­ nondestructive testing methods. Excellent superior to those of conventional braz­ destructive testing (NDT) techniques weld quality was obtained at 1800° F ing methods. In the joining of dissim­ were used to evaluate their effec­ with a pressure of 60 psi and a time of ilar metal tubes, (stainless steel to tiveness of detecting artificial discon­ 8 180 min. The optimum surface finish was aluminum), the use of a diffusion aid tinuities such as voids, contamination, 15 RHR or better. With these process in the form of silver plating on the and lack of joint fit-up. parameters, joint efficiencies of 100% faying surfaces produced an excellent were obtained. The fatigue properties of joint and avoided detrimental stainless the diffusion welded joints approached Experimental Procedures steel-aluminum concentration gradi­ those of the base metal. Material NDT methods used were X-ray and ents. However, because of metallurgi­ ultrasonic A and C scan techniques. cal and corrosion problems associated The material for the metallurgical, Satisfactory NDT results were obtained with diffusion aid concentration gradi­ mechanical property, and nondestruc­ by both the ultrasonic A and C scan ents, it was decided not to use a tive test coupons was taken from an­ methods. diffusion aid in this program. nealed Ti-6A1-4V alloy rod having Low pressure diffusion welding is a the following composition: Al—6.30%, Introduction relatively new process. A development V-3.80%, Fe-0.017%, and Mn- The diffusion welding process offers effort was required to determine its 0.006%. One inch diameter cylin­ the potential for improving properties most optimum weld cycle. Primary ders, 3 in. long, were prepared for the and reducing weight of structural ele­ interest was placed on developing a mechanical property evaluation, and ments such as reinforced panel assem­ capability to low pressure diffusion both the metallurgical and NDT test blies. In addition, metalworking costs weld both thin sections and built-up buttons were made 0.250 in. thick, for producing certain types of forgings laminate production parts, such as (wafer shaped). The specimen config­ could be reduced by diffusion welded those shown in Fig. 1. The T-stiffened urations used in this program are built-up laminate structures as forging panel in Fig. 1A required two longi­ shown in Fig. 2. substitutes. tudinal welds connecting the thin cen­ The pressure required to produce ter section to the horizontal panel and Test Procedure intimate contact during the early the T-cap strip. The bell crank forging The welding system schematic is stages of diffusion welding of titanium replacement of Fig. IB was built up shown in Fig. 3. The welding environ­ can be high (above 1,000 psi) or low by laminating seven pre-machined ment was provided by an inert-gas (below 1,000 psi). Several investiga­ parts together by diffusion welding. retort system. The retorts were de­ 1 tors '-' have shown that Ti-6A1-4V The program was conducted in two signed and fabricated from type 321 alloy can be diffusion welded at high major phases. In the first phase, a stainless steel. A resistance wound ra­ pressures (1,000-2,500 psi) and low study was conducted to establish an diation heating type furnace capable temperatures (1200-1600° F). Such optimum range of w?ld cycle parame­ of maintaining a temperature of up to parameters offer high potential for ters, i.e., time, temn~-ature, pressure, 1900° F for 3 hr, was designed spe­ joining multi-layered structures. How­ and surface finish. The times consid­ cifically for the given retort configura­ ever, for joining elements such as ered ranged from 60 to 180 minutes. tion. The temperature of the sample reinforced panel assemblies, the Temperatures of 1700, 1800, and was controlled by an automatic con­ tooling required for high pressure dif- 1900° F, were chosen in order to troller, and as a check a chromel- obtain results below and above the alumel thermocouple was also used. R. J. REHDER and D. T. LOVELL are beta transus (1825° F) of the alloy. The thermocouples were located near Research Engineers, Aerospace Group, The the sample inside the retort. A manu­ Boeing Company, Seattle, Wash, The diffusion weldirg pressures evalu­ ated ranged from 15 to 75 psi. The ally operated potentionmeter was used Paper presented at the AWS 50th Annual surface conditions investigated in­ with the chrome-alumel thermocou­ Meeting held in Philadelphia, Pa., during April 28-May 2. 1969. cluded finishes of 15 to 60 RHR as ple. At the end of the argon system,

WELDING RESEARCH SUPPLEMENT | 213-s •- NDT AND METALLURGICAL TEST SPECIMEN TENSILE SPECIMEN FATIGUE SPECIMEN

SAMPLE AND INCONEL MANDRELS

INITIAL 0.250 CYil NDER

A. "T" STIFFENED STRINGER P,

BAC 23-7070 BAC 23-6550 TEST SPECIMEN BEFORE MACHINING WELDED SPECIMEN Fig. 2—Diffusion welded test specimen configurations

SURGICAL TUBING FROM RETORT TO BUBBLER

OIL PURGING SYSTEM RETORT FURNACE

BELL CRANK . ..

Fig. 1—Examples of diffusion welded SURGICAL TUBING FROM production parts GAS SOURCE TO PURIFIER POTENTIOMETE an oil purging (bubbling argon through oil) system was used to keep air isolated from the inert gas enve­ lope. The metallurgical test buttons were loaded between two Inconel mandrels . THERMOCOUPLE and placed in the retort. A lid was -STAINLESS STEEL BELLOWS TUBING FROM PURIFIER TO welded on the retort, and then the RETORT retort was argon gas purged for 1 hr. -Ti CHIPS The retort assembly was then placed (ARGON PURIFIER) in the furnace. The same procedure, Fig. 3—Experimental diffusion welding equipment with the exception of the use of In­ conel mandrels, was employed when loading the mechanical property eval­ and it was taken out of the furnace Parameters uation test coupons. The dead load and cooled down to at least 250° F in A summary of test parameters eval­ used to supply the weld pressure was an argon atmosphere. The sample was uated is shown in Table 1. Parameters applied during the heat-up period. The then removed from the retort and tested were time, temperature, pres­ heat-up time was recorded, and once measured to determine dimensional sure, and surface finish. Purified argon the welding temperature was reached, changes which occurred during the was used as the atmosphere during all the outtime was established, i.e., 1, 2 weld cycle. Mounts of the metallurgi­ diffusion welding experiments. A or 3 hr after completion of heat-up. cal, tensile and NDT test coupons metallurgical examination of the weld At the end of the weld cycle, the were made for metallurgical examina­ joint was conducted on each of the weight was removed from the retort, tion to determine weld quality. specimens for correlation of weld

Table 1—Summary of Test Conditions Temperature Surface —Time, min—, °F . -Pressure, psi— finish, (RHR) In- Con- Creep Test Chem. duced tami- Fit- defor- type 60 120 180 1700 1800 1900 15 30 45 60 75 15 30 60 milled void nant Up mation Metal­ lurgical Tensile Fatigue NDT

214-s I MAY 1970 quality with the weld cycle parameters used in producing the weld. Tensile and fatigue specimens were prepared to evaluate mechanical properties of the butt type diffusion weld joint and to establish optimum weld cycle parameters. The tensile specimens were statically loaded to failure and the yield strength, ultimate strength, reduction of area, and elon­ gation values were determined. The fa­ tigue specimens were tension-tension m loaded at a stress level of 60 ksi and Fig. 4—Surface finish effects on weld voids. A (left)—long weld line void; B (right)— 1800 cycles/sec to failure. small weld line void. Welding parameters: time—180 min; temperature—1300° F; pressure—44 psi (for A), 15 psi (for B). Finish—approx. 30 RHR (for A) and 15 RHR Types of weld irregularities studies (for B); number of voids—54/in. for A) and 6/in. (for B); total void length—0.090 in./ during nondestructive testing were in­ in. (for A) and 0.003 in./in. (for B); lack of weld—9.0% (for A) and 0.3% (for B), X500 duced voids, contamination, and fit- (reduced 56% in reproduction) up. The weld cycle parameters used throughout the NDT study were time adequate for good welding at pres­ 1. Weld cycle time evaluations of of 180 min, temperature of 1800° F, sures below 100 psi at the 1800 to 60, 120, and 180 min showed a signifi­ pressure of 60 psi, and weld surface 1900° F temperature as shown by the cant improvement in weld quality finish of 15 RHR or less. long, nearly continuous weld line voids with increasing time up to 120 min Induced voids were produced in in Fig. 4A. and a slight improvement above 120 to the specimens by machining a circular Acceptable weld quality was readily 180 min. Therefore, the 180 min weld groove in the specimen faying surface. obtained using a 15 RHR surface cycle time was used throughout the The voids ranged in size from 0.010 to finish as shown by the single small remaining testing to obtain the re­ 0.035 in. in increments of 0.005 in. In circular void in Fig. 4B. Therefore, a quired diffusion for good solid state order to simulate various contamina­ 15 RHR or better weld surface finish welding. tion conditions, separate specimens was used throughout the remainder of 2. Going above the beta transus were locally swabbed prior to welding the diffusion welding process develop­ temperature (above 1825° F for Ti- with the following materials: ment. The improved surface finish 6A1-4V alloy) did not help nor hinder 1. T-50, a high temperature lubri­ gave, in general, good results. Void interface diffusion, i.e., the number of cant. lengths across specimens welded at 60 voids along the weld line or total void 2. Tissue quartz stop-off. psi for 180 min, and 15 RHR or length across the specimen did not 3. Levitated alumina stop-off. better surface finish were small or change. However, the microstructure 4. Titania stop-off. nearly nonexistent. Therefore, a spec­ varied with weld cycle temperature as 5. Boron nitride stop-off. trum of tests were performed using shown in Fig. 5. As one would expect pressures of 15 to 75 psi in 15 psi upon examination of the Ti-6A1-4V A simulation of poor fit-up was increments with a temperature range alloy phase diagram, there is only a accomplished by the use of titanium of 1700 to 1900° F for 1, 2, and 3 small amount of retained beta present metal shims. Layers of 0.001 in thick hr welding time. upon cooling from 1700° F to tem­ Ti-6A1-4V shim stock were cut in a In evaluating the influence of the perature. The resulting structure was half-moon shape and placed on the time, temperature, and pressure exquiaxed primary alpha plus beta as weld surface to produce a wedge type parameters on weld quality and shown in Fig. 5A. The 1800° F misfit at the shim stock straight edge. weldment microstructure, it was noted treated samples showed an exquiaxed A single layer of shim stock was used that: primary alpha plus acicular alpha to produce a 0.001 in. misfit. The misfit thickness was varied to 0.003 and 0.006 in. by using 3 and 6 thick­ nesses of shim stock, respectively.

Results and Discussion Diffusion Welding Parameter Studies Due to the initial diffusion welding requirement of intimate contact, it was necessary to establish the level of surface finish required before optimiz­ ing the time, temperature and pres­ sure parameters. The surface finish was correlated with total void length across the metallurgical test buttons as shown in Fig. 4. The contacting surface asperities of joint surfaces ground to 30 RHR were deformed enough be pressures greater than 100 psi at the 1700 to 1900° F tempera­ ture range to give a good mechanical Fig. 5—Temperature influence on weld microstructure. A (left)—1700° F run; equiaxed fit, and hence a satisfactory weld. primary alpha plus beta. B (center)—1800° F run; equiaxed primary alpha plus acicular alpha-beta structure. C (right)—1900° run; acicular alpha-beta structure. However, the mechanical fit obtained Weld cycle parameters held constant: time—60 min; pressure—25 psi. Reduced 50% by a surface finish of 30 RHR was not on reproduction

WELDING RESEARCH S U P P L E M E N T I 215-s Tensile Test Results Specimens welded at 15 psi pressure resulted in near complete lack of weld. Therefore, this pressure cycle was not further evaluated. Both the 60 and 45 psi pressure samples showed good mechanical properties.

LONGITUDINAL SECTION OF WELD LINE, The higher pressure specimens frac­ LONGITUDINAL SECTION, NOTE; FRACTURE IN BASE METAL tured in a ductile cup-cone manner (Fig. 6) in the base metal. Thus, base FRACTURE SURFACE, NOTE, metal properties rather than weld I WELD CYCLE PARAMETERS; TIME 180 Mtn„-TIMHKATLStE tSOO'F, PtEiSUK 40 PSi CUP CONE TYPE FRACTURE properties are represented by these specimens. A noticeable amount of necking in the weld zone in base metal failure specimens (Fig. 6) indi­ cated that the weld area did deform plastically without failure. The speci­ mens subjected to the 1800 and 1900° F temperatures failed in a similar F, 500X LONGITUDINAL SECTION Of WELD LINE. manner for a given weld cycle pres­ sure and time. The samples exposed below 45 psi exhibited a brittle type failure mode and less reliable mechan­ WELD CYCLE PARAMETERS; TIME 180 Mia., TEMPERATURE t800»F PRESSURE 30 PSI. FRACTURE SURFACE ical property results were obtained— Fig. 7. Fig. 6—Fracture examination of typical specimens welded at 60 and 30 psi pressure. Reduced 45% on reproduction Upon microstructural evaluation of several of the 1900° F specimens, it structure — Fig. 5B. The 1900° F ters selected for further investigation was found that an intermittent alpha treated samples were completely trans­ were as follows: time—180 min; type platelet formed at the weldline— formed to the acicular alpha-beta temperature—1800 to 1900° F; pres­ Fig. 8A. A microprobe analysis for aluminum, vanadium, and titanium structure—Fig. 5C. sure—30 to 60 psi; surface finish—15 3. An evaluation of the effects of showed no variation in concentration RHR. Once these parameters were various pressures indicated that pres­ of elements in the weld zone. An sures significantly above 15 psi are determined, the tensile and fatigue electron fractograph of the alpha required to provide acceptable joint specimens were made for evaluation platelet (Fig. 8) did not reveal any quality. of mechanical properties of the butt- microvoids which could act as pins, The resulting preliminary parame­ type weld. thus keeping the platelet from mi-

DIFFUS'ON WELD TENSILE RESULTS

Wold Time - 3 Hour! m^&Bm^ A. 500X LONGITUDINAL SECTION OF WELD UN£. Note: The olpho plotlet starting ot the void is shown below by the electron (ractosjroph, replica, ot U.OOOX

B. 600X MICROHARDNESS TRAVERSE: Of ALPHA PLATUT. Note; Traverse indicated no change in hardness, average hardness rockweit C 34.7 .••;«• -.,:. <«r.'* r~. i "~v\ "TTT.i - - ... -

< 20 ?'S&,\!gM%\

J 10 "O

it) i$ 50 55 40 Fig. 8—Electron micrograph of a weld line phase platelet pro­ WELD PRESSURE (PSI) duced at 1900° F. Weld cycle parameters: time—180 min; tem­ perature—1900° F; pressure—60 psi. X11.000 (reduced 52% on Fig. 7—Diffusion weld tensile results reproduction)

216-s I MAY 1970 grating away from the weld line dur­ - BASE METAL PROPERTIES AFTER EXPOSURE TO 1800 F for 3 HOURS ing the recrystallization phase of dif­ ._l_,.- fusion welding. A hardness scan across ; the alpha platelet (Fig. 8) did not reveal any increase in hardness caused by contamination at the weld line. Therefore, it was concluded that the alpha platelet present in some of the 1900° F treated samples was not mi- crostructurally different from the base metal structure. It was found that the reduction in area values fell off markedly for speci­ mens welded at 45 psi and lower. A plot of percent reduction in area vs. weld pressure is shown in Fig. 7. The ductility curve is nearly horizontal for the 60 psi samples, indicating that effective solid state welding has oc­ curred. 45 it) DIFFUSION WELDING PRESSURE, PSI Fatigue Test Results Fig. 9—Diffusion weld fatigue results As in the tensile test, both the 60 and 45 psi pressure specimens showed good mechanical property results. As the ultrasonic inspection indicated all from the back reflection intensity. shown in Fig. 9, the higher pressure major artificial weld irregularities. Therefore, the lack of internal reflec­ specimens exhibited a similar order of However, no significant indications of tion was indicated by the void-free magnitude of cycles to failure as the weld irregularities were found by X- areas, as shown in Figs. 12—14. The base metal specimens exposed to the ray examination. dark read-out areas represented inter­ same temperature and time. A lower Both the ultrasonic A and C scan nal voids. Ultrasonic A and C scan fatigue life was noted when the base techniques were very effective in iden­ inspections were effective in identify­ metal and welded specimens were ex­ tifying various weld defects. The ing induced voids of 0.010 inch or posed to 1900° F, well above the beta modified A scan (Figs. 12—14) gave larger. Representative results of the transus as shown in Fig. 9. a very prominent indication of the metallurgical study verified the effec­ back reflection plateau with fluctua­ tiveness of the non-destructive tech­ Nondestructive Testing and tions indicating internal reflections niques as follows: Metallographic Results from voids at the weld surface. The C 1. The induced void specimen A summary of the NDT results are scan was set up to read-out all voids (Fig. 12) had artificial voids and shown in Figs. 10 and 11. In general, of larger than five percent variation natural voids. The various types of voids were readily indicated in the A scan read-out by a decrease in sound

TYPE SPECIMEN NDT INDICATION intensity, while the nearly continuous central circle of the C scan graphically GROOVE WIDTH RADIOGRAPHIC ULTRASONIC indicates the induced and natural .010 NONE GOOD GOOD voids represented by the salt and pep- .015 VERY SLIGHT GOOD GOOD VERY SLIGHT GOOD GOOD NONE GOOD GOOD -••( INDUCED VOIDS!*- .030 NONE GOOD GOOD .035 NONE GOOD GOOD 7 /* >y 7 "y / i/ / \ GROOVE WIDTH RADIOGRAPHIC NONE GOOD GOOD NONE GOOD GOOD *T"—V WIDTH NONE GOOD GOOD NONE GOOD GOOD NONE GOOD GOOD NONE GOOD GOOD

Fig. 10—Summary of NDT and metallurgical correlation results '.. Q

I :• ;||

NDT INDICATION \ SS3MB CONTAMINATION TYPE RADIOGRAPHIC ULTRASONIC |||"" T-50 NONE GOOD GOOD ' "I TISSUE QUARTZ NONE VERY SLIGHT ^ooQ ttox iVITATED ALUMINA NONE GOOD GOOD g TITANIA NONE NONE ^POOR^ BORON NITRIDE NONE GOOD GOOD 1 NONE NONE NONE GOOD t w it •• .' ~\ RADIOGRAPHIC ULTRASONIC

NONE FAINT LINE FAIR -| ^BSS-J* ULTRASONIC C SCAN NONE GOOD GOOD SLIGHT GOOD GOOD

ULT8ASQN1C A' SCAN

Fig. 12—Typical NDT and metallurgical results of induced Fig. 11—Summary of NDT and metallurgical correlation results void. Reduced 59% on reproduction

WELDING RESEARCH SUPPLEMENT1 217-s CONTAMINATE

>HOTO COVERAGE OF SECTION

n xi

ULTRASONIC A SCAN ULTRASONIC C SCAN

ULTRASONIC C SCAN Fig. 14 (right)—Typical NDT and metallurgical results of fit- up. Reduced 62% on reproduction

Fig. 13—Typical NDT and metallurgical results of contamina­ metal and weld specimens processed tion. Reduced 58% on reproduction with the 1900° F welding cycle had a much lower fatigue life than speci­ per read-out. (c) Surface finish—15 RHR or bet­ mens welded below the beta transus at 2. Similarly, the contamination ter. 1800° F. caused by T-50 type stop-off (Fig. 13) 3. Diffusion welding parameters 8. Metallurgical studies confirmed was clearly indicated by both the A could not be established for accepta­ the foregoing conclusions with regard and C ultrasonic scans and was ble weld joints using equivalent atmo­ to establishing the pressure, tempera­ confirmed by metallurgical examina­ spheric pressure («= 15 psi) in an ture, time and surface finish parame­ tion. Contamination caused by tissue inert gas atmosphere. ters. quartz or titania which diffuse into the 4. Diffusion welding temperatures 9. Nondestructive testing studies weld surface was not indicated by the of 1800 and 1900° F were effectively using artifical discontinuities represen­ ultrasonic C scan. The impedance and used to allow relatively low welding ting voids, contamination and lack of transmission of ultrasound in these pressures without detrimental damage fit-up, indicated that ultrasonic inspec­ materials when in intimate contact to the base metal microstructure. Dif­ tion, particularly the C scan presenta­ with the metal substrate were appar­ fusion welding above the beta transus tion, was representative of the actual ently such that no reflected signal was (1825° F) did not significantly en­ weld joint quality. However, several received. hance solid state diffusion. contaminants, such as titania and tis­ 3. The C scan and metallurgical 5. Diffusion welding times of 3 hr sue quartz, were not indicated by ul­ evaluations of the fit-up test specimens at temperatures of 1800 or 1900° F trasonic inspection, and caution is ad­ clearly showed salt and pepper indica­ were required to ensure sufficient time vised in their usage during diffusion tions and line voids, respectively. for creep deformation of surface as­ welding. perities at relatively low welding pres­ 10. Radiographic techniques did not Conclusions sures, («= 60 psi). provide an adequate inspection meth­ 6. Tension test results of diffusion od for diffusion welded joints when This program on diffusion welding welded joints confirmed that a welding examined normal to the joint. Ti-6A1-4V alloy resulted in the fol­ pressure of at least 60 psi was neces­ lowing significant conclusions: sary to develop near base metal prop­ References 1. The lowest possible diffusion erties. The percent reduction in area welding pressure which could be used decreased drastically at welding pres­ 1. King, J. P., et al, "Diffusion Bonding of Titanium and Beryllium," Air Force for joining bare Ti-6A1-4V alloy in an sures below 60 psi. The percent reduc­ Materials Laboratory, United States Air inert gas atmosphere at temperatures tion in area values obtained at the Force, Wright-Patterson Air Force Base, up to 1900° F, and consistent with 1900° F temperature exposure were Ohio, September 1965. 2. Ing, W. H., and Owczarski, W. A., obtaining acceptable joint quality, was significantly lower than those obtained "Diffusion Welding ot Commercially Pure 60 psi. at 1800° F. Titanium," Welding Journal, 46(6), Re­ search Suppl., 289-s to 298-s (1967). 2. The following parameters were 7. Fatigue test results of diffusion 3. Crane, C. H., Lovell D. T., Baginski. required for diffusion welding the Ti- welded joints also indicated that 60 W. A., and Olsen, M. G., "Diffusion Weld­ 6A1-4V alloy in an inert gas atmos­ ing of Dissimilar Metals," Ibid.,46(1), Re­ psi was the minimum joint pressure search Suppl. 23-s to 31-s (1967). phere consistent with the 60 psi weld­ needed for obtaining joint fatigue 4. King, J. P., "Diffusion Bonding of ing pressure: properties equivalent to base metal for Titanium and Beryllium," Air Force Ma­ terials Laboratory, United States Air Force, (a) Temperature—1800° F. a similar thermal exposure. Further­ Wright-Patterson Air Force Base, Ohio, (b) Time—3 hr. more, it was shown that both base February 1967.

218-s I MAY 1970