WESGODivision, GTEProduct Corporation, Belmont, California94002 W ith the development of new ceramic materials, including those for struc- tural applications, there is an increasing demand to join ceramic components to metal structures. The most commonly used methodis themoly-manganyseprocess. 1-3 This process is well-established and pro- duces highly reliable joints between ce- ramics and metals. However, the moly- manganese process requires two process- ing steps: metalization of the ceramic, fol- lowed by brazing. It is, therefore, time- consuming. Furthermore, the metaliza- tion process is conducted at very high temperature ( -1500C). In addition, joint properties are sensitive to process varia- bles, and, hence, precise process control is required to obtain the reliable joints. As a result, this process is very expensive. Direct brazing between ceramic and metal is also possible through active filler metals.4-'o Because this is a one-step pro- cess, it is simple and economical. Al- though active metal brazing has been in- vestigated since 1940, it has not been widely accepted because of inconsistent joint properties. There are several meth- ods of the active metal brazing. For ex- ample, one method involves a sheet of ti- tanium (the most commonly used active element) that can be cladded by two sheets of the conventional brazing metals.6 An- other example usestitanium hydride pow- ders mixed with powders of conventional brazing metals.s However, the most eco- nomical method utilizes afiller metal where an active element(s) forms a true alloy with the base filler metal. The present investigation centers on a copper-silver filler metal containing titanium as an ac- tive element (Cusil-ABA)@. * The nominal chemical compositions and several im- portant properties of the Cusil-ABA@ are listed in Table 1. Because the concentra- tion of the titanium is relatively low (2 wt%), titanium dissolves in the matrix (mainly copper-rich phase) as a solid so- lution. There are a variety of different prop- erties to be considered in the ceramic- metal joints, e.g., mechanical, electrical, thermal properties, etc. Depending on the application of the joint, some properties are more important than others. However, 'Member, American Ceramic Society. 'GTE Product Corp., WESGO Division, Belmont, CA. Table I. Chemical Compositions and Properties of Cusil-ABNMFillerMetal Nominal composition 63 wt% Ag+35 wt% Cu+2 wt% Ti Physical properties 815C (I 500F) 780C (l435F) 9.8 Mg.m-J (5.2 troz.in.-3) 180 W.m-I.K-1 (104 Btu.h-'.ft-,.oF-') 18.5 X 10-6 K-I (10.3 X 10-6 OF-I) Liquidus temperature Solidus temperature Density Thermal conductivity* Thermal expansion coefficient (RT to 500C) Electrical properties 44 x 10-9 Q.m 23 x 106 Q-I .m-I Mechanical properties 83 GPa (12 x 106 psi) 0.36 271 MPa (39300 psi) 346 MPa (50200 psi) 20% 1100 MPa (110 KHN) Electrical resistivity Electrical conductivity Young's modulus Poisson's ratio' Yield strength Ultimate tensile strength Elongation (2-in. gage section) Hardness 'Calculated from the electrical conductivity by the Wiedmann-Franz law. 'Estimated by the volumetric average of the Poisson's ratios for silver and copper. the mechanical properties are some of the most important properties for any joint. Joints without any mechanical strength can be regarded as unsuccessful joints. In the present paper, therefore, the mechan- ical properties of the joints are empha- sized. For the evaluation of the joint prop- erties, it is essential to establish proper testing methods so that effects of the pro- cessing variables can be detected accu- rately and consistently. It is also desirable that the results of the tests provide mean- ingful engineering parameters for design engineers.Severaltestingmethodsareex- amined.The optimumtestingmethodsfor the evaluation of the ceramic-metal joint properties are recommended. One of the major problems of joining ceramic to metal is the thermal expansion mismatch between them. Generally, the ceramicmaterialshavelowerthermal ex- pansioncoefficientsthan the metallicma- terials. A list of the thermal expansion coefficients of several structural ceramics and common metals is given in Table II. The difference in the thermal expansion coefficients can lead to very high stress at the brazed region during cooling from the brazing temperature to room tempera- ture. The high stress sometimes results in joint failures or unreliable joints. There- fore, the thermal stress problems should be overcome to obtain reliable joints be- tween the ceramic and the metal. In the present paper, the active metal brazing process of the ceramic to the metal is examined in detail. Influences of all process variables, including joining ma- terials, engineering joint design, and braz- ing procedures, are evaluated to obtain reliable joints between ceramic and metal with the active filler metal. Joining Materials Ceramic-metal brazing involves three materials: ceramic, metal, and filler metal. Properties of each material influence the resultant joint properties. Ceramic Component The ceramic, because of its inherent brittleness, is the most critical material for obtaining reliable joints. The base properties of the bulk ceramic member are essential. When the properties of the bulk ceramic are not sufficient, the ther- mal stress simply fractures the ceramic member. Furthermore, the surface con- dition of the ceramic is also very impor- tant for the joint reliability.7.'oMizuhara and Mally7 and Mizuhara and Huebel'O demonstrated that, when the ceramic sur- face is in the as-ground condition, only marginal joint properties between the ce- ramic and the metal will be obtained. When the ceramic material is ground by a metal-bonded diamond wheel, mi- crocracks are introduced at the surface of the ceramic. The size of the micro- cracks depends on the diamond grit size of the wheel and also on the rate of ma- terial removal. The surface damage can initiate major cracks in the ceramic by the thermal stress and, hence, result in an unreliable joint. Therefore, the ceramic surface should be free of damage to ob- CERAMIC BULLETIN, VOL. 68, NO.9, 1989 (@ACerS) 1591 Reprinted for The American Ceramic Society Bulletin. Vol. 68, No.9, September 1989 Copyright @ 1989 by The American Ceramic Society, Inc. tain high-reliability joints. This condition can be met simply by using sintered ce- ramic materials. However, nearly all sin- tered ceramic parts over about 2 cm in size should be ground, because distortion of the parts during the sintering requires grinding for dimensional control. Ground ceramic materials should be treated fur- ther to obtain a defect-free surface con- dition. This can be performed by a resin- tering or lapping process. In the resin- tering process, the damaged layer is healed through sintering. In the case of the lap- ping process, the damaged layer is phys- ically removed. It should be mentioned that the thickness removed by the lapping must completely eliminate the surface damages. In the present investigation, AL- 995* (99.5% AlzOJ)is used as the ceramic member of the joint. Two surface condi- tions are examined: the as-ground and the ground-and-resintered conditions. The re- sintering process is conducted at 1650C for 1 h. Metal Component The metal is generally ductile and, hence, does not readily fracture. How- ever, there are several important require- ments to obtain reliable joints. The metal should exhibit only low blushing (or surface flow)of the filler metal. The extensive blushing effectively de- pletes the amount of the active element available for wetting the ceramic. Among the parameters controlling the blushing behaviors are the relative properties of the metal member and filler metal. Ef- fects of the metal member on the blush- ing behaviors are demonstrated in Fig. 1. In these examples, the Cusil-ABA@foils (50-tLmthick) are placed on different sub- strate metals (and alumina) and melted under vacuum at 830C. As demonstrat- ed in Fig. 1, the Cusil-ABA@shows dif- ferent blushing behaviors depending on the substrate material. No blushing is ob- served on alumina, 304 stainless steel, or copper, whereas extensive blushing is ob- served on nickel. These results demon- strate that the blushing of the filler metal can be minimized by proper selection of the metal member. Another method to eliminate the blushing is to apply stop-off paint on the metal member. It should be also mentioned that machine marks on the metal surface promote blushing. As described earlier, high thermal stress can be created at the joint because of the thermal expansion mismatch between the ceramic and metal members. One of the methods to reduce the thermal stress is to closely match the thermal expansion coefficients of the metal and the ceramic. For example, molybdenum can be used to bond with an alumina ceramic (see Ta- ble II). However, there are a limited num- ber of the ceramic-metal combinations to 'Spang Industries, Magnetics Division, Butlep, PA. Fig. 1. Blushing behaviors of the Cusil-ABA@on different substrates. No blushing is observed on (A) alumina, (B) 304 stainless steel, or (C) copper, whereas extensive blushing is observed on (0) nickel. Table II. Thermal Expansion Coefficients of Structural Ceramics and Common Metals Material Silicon nitride Alumina (99.5%) Molybdenum Kovar@ Alloy 42 410 stainless steel Copper Thermal expansioncoefficient (xlO-6C-I) 3.2 8.0 5.7 10.0 10.2 14.0 20.0 satisfy this condition. In the present in- vestigation, alloy 42t (Fe-4lNi), which also has a similar thermal expansion coef- ficient to that of the alumina ceramic, is used as the metal member. Another method to reduce the thermal stress is to use a metal having a low yield strength. Plastic deformation of the metal member accommodates the thermal ex- pansion mismatch, thereby reducing the thermal stress at the joint. In the present investigation, two alloy 42 materials with different strength levels are examined. One material is in the cold-worked % hard (% H) condition and has a hardness of 75 in Rockwell Bscale (Rb). The other material has been fully annealed at 975C for I h. The hardness of the annealed material is 65 in Rockwell B scale. Filler-MetalComponent The basic and most important require- ment for the active filler metal is that the filler metal should be able to wet and bond strongly with the ceramic. However, there are additional requirements on the active filler metal. As described previously, the filler metal should not blush over the ce- ramic or the metal member. Figure 2 demonstrates the effect of the filler metal on blushing behavior. Two different filler metals, a 70Ti-15Cu-15Ni alloy and the Cusil-ABA@,are melted on the alumina substrate. The 70Ti-15Cu-15Ni alloy shows extensive blushing because of the high titanium concentration. In contrast, the Cusil-ABA@ (2 wt% Ti) shows no blushing. It is also important that the fill- er metal is ductile. The ductile filler metal accommodates the thermal expansion mismatch to reduce the thermal stress at the joint as in the case of the ductile metal member described previously. Also, the filler metal should have low vapor pres- sure. In addition, the filler metal should have the ability to be tailored for specific applications that require specific proper- ties such as melting temperature, oxida- tion and/or corrosion resistance, density, thermal expansion, etc. As described earlier, the copper-silver alloy containing 2 wt% titanium is used as the filler metal in the present investi- gation. The thickness of the filler metal has been chosen as one of the process var- iables. Foils with three different thick- nesses (50, 100, and 150 tLm) have been prepared and the effects of the filler-metal thickness on the joint properties are eval- uated. Engineering Joint Design As previously stated, one of the major problems of joining ceramic to metal is f'TN> Al\KIf' UTn T I<',[,T1\1 \TnT. I':R Nn Q 1 QRQ Ira Arpr~) '~nn the thermal expansion mismatch between the ceramic and the metal. The difference of the thermal expansion coefficients leads to the high thermal stress in the joint re- gion. The thermal stress can be reduced 1;>ythe selection of the materials, as dis- cussed in the Joining Materials section. As the size of the joint assembly increas- es, however, the thermal stress also in- creases, and it becomes more difficult to overcome the thermal stress by the ma- terials selections. Oftentimes, this diffi- culty can be overcome by engineering joint designs. CompliantJomtDes~n Edge brazing of a metal cylinder to a ceramic face is a popular form of com- pliant joint. In this case, the thermal ex- pansion mismatch is accommodated by the concentric distortion of the metal cyl- inder. In addition, the fillet formed at the joint distributes the thermal stress over a large surface area of the ceramic, and, hence, the thermal stress on the ceramic is reduced. Figure 3 shows an example of such a design, where a copper-clad 430- stainless-steel cup has been brazed on an end of an alumina tube. Honeycomb structures or Feltmetal@~can also be used to have compliant joints, where distor- tions in the honeycomb structure or in the Feitmetal@reduce the thermal stress. One of the disadvantages of these two designs is that the joints cannot sustain compres- sive loading. Ductile metal interlayer(s) can be used in the joint assembly. The ductile inter- layer deforms plastically to reduce the thermal stress. Another method to obtain compliant joints is to use an interlayer(s) that has a thermal expansion coefficient between those of the ceramic and the metal. This distributes the thermal ex- pansion mismatch and reduces the ther- mal stress. Figure 4 shows an example of such a joint, known as a gradient seal. In this case, a molybdenum interlayer (ther- mal expansion coefficient of 5.7 x 10-6 C-I) is utilized to braze a silicon nitride (thermal expansion coefficient of 3.2 x 10-6 C-I) disk to a ductile cast-iron (thermal expansioncoefficientof 12 x 10-6 C-I) substrate. Compression Joint Design In this joint, the thermal expansion mismatch is utilized as an advantage to obtain the reliable joint. This is achieved by brazing the ceramic member into the metal member. During cooling from the brazing temperature to room tempera- ture, the metal member (outside) con- tracts more than the ceramic member (in- side). This results in compressive stress in the ceramic member as well as in the joint, and the joint strength is increased. An example of such a joint is shown in Fig. IEnergy Conservation Systems, Brunswick Technit- ics Division, DeLand, FL. Fig. 2. Effects of titaniumconcentration on blushing behaviors of active brazing metals on alumina: (left)extensive blushing is shown for 70Ti-15Cu-15Ni melted at 1050C and (right) no blushing is observed for the Cusil-ABA@melted at 830C. M Copper -clad stainless steel ~ ~ Cusil-ABA filler metal Alumina cylinder .,";Jt' :*'~~~E: ::1 -.-.- Fig. 3. Example of the edge brazing where a copper-clad 430-stainless-steel cap has been brazed on an alumina cylinder: (left)a schematic cross section of the assembly and (right) the actual brazed sample. ~ -~, ,~, ,,~,~, VM V" ,~,.,."" .~"'".,." Si 3N4 Ceramic Cusil-ABA . !!!J Molybdenum Cusil-ABA m """"" ""'""<..",,,, ~""~" """'~""""""""" - Dulile cast iron Fig.4. Exampleof the gradient seal where a siliconnitridedisk has been brazed on a ductile cast-iron plate with a molybdenuminterlayer: (left)a schematic cross section of the seal and (right) the actual brazed seal. 1593 CERAMIC BULLETIN, VOL. 68, NO.9, 1989 (@ACerS) .L- 5, where a silicon nitride rod is brazed to 410 stainless steel through the compres- sion joint. It should be noticed that the silicon nitride rod has a tapered end. Therefore, the stainless steel has a hole with the same taper and the end of the steel has a knife edge. This is to eliminate sudden changes in the stress that can be detrimental for the joint properties. Stress-Distribution Joint Design A ceramic backup can be used to dis- tribute tensile loading on the metal dia- phragm. Figure 6 shows an example of this type joint, where a cupronickel sheet has been brazed on an alumina cylinder. As shown in Fig. 6, a backup ring of alu- mina has also been brazed on top of the cupronickel sheet. The backup ring serves two purposes. First, the backup ring cre- ates additional interface area (at the cu- pronickel sheet and the backup ring); thus, the thermal stress is distributed at two interfaces instead of one. Second, the backup ring changes the stress distribu- tion in the cupronickel sheet so that there is no peeling moment at the edge of the cupronickel sheet. Brazing Procedures Furnace brazing is the best technique for ceramic-metal joining by the active filler metal. Vacuum or inert gas is the most economical and pollution-free op- eration. In the case of vacuum brazing, a leak rate should be controlled to be less than 0.005 mm Hg/h. In the case of inert- gas brazing, a dry Ar, He, or H2 atmo- sphere can be used. Ironically, one of the major problems related to active filler metal brazing is that the active element wets nearly all materials. Therefore, care should be taken in designing the joint as- sembly so that molten filler metal does not touch any part of the brazing fixture. The heating rate 'during the brazing is important. Since the metal member has higher thermal conductivity than the ce- ramic member, the metal can heat up faster than the ceramic. In this event, the filler metal is drawn away by the metal mem- ber. This becomes more important in the applications where the mass of the metal member is much smaller than that of the ceramic (e.g., the edge brazing). There- fore, the heating rate should be controlled carefully to minimize the temperature dif- ference betweenthe ceramic and the metal. This can be achieved by initially holding the temperature just belowthe solidus and then increasing the temperature at a con- trolled rate to the brazing temperature. The cooling rate is also critical for the reliable joints. When the cooling rate is too fast, there is insufficient time for the filler metal and/or the metal member to deform plastically and reduce the thermal stress. Therefore, the cooling rate should Westinghouse Electric Corp., Pittsburgh, PA. be controlled to minimize the residual stress. In both the heating and cooling processes, the slower the rate of temper- ature change, the better the resultant joint properties. However, the temperature changes should be fast enough to mini- mize the processing cost. Considering all the factors mentioned, preliminary experiments have been con- ducted to optimize brazing procedures for the present investigation. The results lead to the temperature-time profile shown in Fig. 7. This profile is utilized throughout the present investigation unless otherwise mentioned. Testing Methods There are three major requirements for a testing method to evaluate the mechan- ical properties of the ceramic-metaljoints. First, a testing method should lead to ac- curate and consistent results. Second, a testing method should be able to evaluate effects of processing variables on the joint properties. Third, the results of the tests should provide meaningful engineering parameters that can be utilized for de- signing the ceramic-metal joints. There are many testing methods used for eval- uating the joint properties. Three fre- quently used methods will be reviewed. In addition, a new testing method will be introduced and examined. 01 &I Copper Tube II Si 3N4 C"ami, 410SlaJo'",S'oo' j Fig. 6. Example of the stress-distribution joint. Alloy42 sheets are brazed on both ends of an alumina cylinder with alumina backup rings: (top) a schematic cross sec- tionof thejoint and (bottom)the actual stress- distributionjoint. ined. The detailed descriptions for the peel test have been given in Ref. 7. The results are shown in Fig. 9. A group of six results shown on the left side of Fig. 9 is obtained on the ground-and-resintered surface. High peel strengths (=20 lb (90 N)) have been obtained in all six tests. The results shown on the right side of Fig. 9 have been ob- tained on the ground surface. Again, con- sistent results are obtained. Peel strengths of the ground surface (=6 lb (27 N)) are ~~~'Hm ~TnT~mnT Hr.T ~O "Tr. '"' ""'0'"' {r?\Af"~_"\ "",ii-ABA fill" m,1aJ 1-- -- . ~~-~ Fig. 5. Example of the compression joint. A silicon nitride rod is brazed into a hole ma- chined in a 410 stainless steel with Incusil- ABA@* (top) a schematic cross section of the joint and (bottom) the actual compres- sion joint. Peel Test In the peel test, a metal strip is brazed on a ceramic substrate by the active filler metal. The load required to peel the metal strip from the ceramic substrate is mea- sured. Therefore, the peel test is relatively simple and cheap. Examples of the peel test are shown in Fig. 8, where three Ko- var@strips have been brazed to alumina substrate by Cusil-ABA@(50-I.LIDthick). Two different surface conditions of the alumina (AL-995) substrate are exam- . - Alloy 42 IIIII!IIIII - IIIIIIIIIIIIII Cusil-ABA I AluminaCylinder - IIIIIIIIIIIIIICusil-ABA Alloy 42 !IIIIIIIIIIiII III!IIII!iIiII Cusil-ABA II II Alumina backup ring considerably lower than those of the ground-and-resintered surface. This is due to surface damages on the ground ceram- ic as described earlier. Figure lO shows cross-sectional views of the untested peel test samples. The interface between the Kovar@and the ground-and-resintered ce- ramic is very smooth (Fig. lO(A and resulted in the high peel strength. The interface between the Kovar@ and the ground ceramic is not smooth and also contains small cracks (Fig. 10(B. As a result, low peel strength has been ob- tained. These results demonstrate that the peel test is a useful method for evaluating joint properties. It is simple and cheap, and the results are consistent. Also, effects of the processing parameters can be detected. However, one of the disadvantages of the peel test is that the result of the peel test cannot be translated into meaningful en- gineering parameters for designing of ce- ramic-metal joints. Tensile Test The ASTM F19-61 CLM-15 tensile test]2 is examined for the evaluation of the ceramic-metal joint by the active brazing metal. Figure II shows the test sample assembly, brazed sample, and fractured sample. As shown in Fig. II, a Kovar@ring is brazed to two alumina ce- ramic parts with Cusil-ABA@.Again, two surface conditions of the ceramic are tested: ground and ground-and-resintered surfaces. The results demonstrate that the tensile test can also detect the effect of the ceramic surface conditions. The frac- ture strength of the sample with the ground-and-resintered ceramic is 75 MPa (10.8 ksi), whereas that of the sample with the ground ceramic is 49 MPa (7.1 ksi). In the majority of the tensile tests, how- ever, failures have occurred predomi- nantly in the ceramic member instead of the joint. This is primarily because the loading mechanism places the ceramic in tension. When the joint strength is higher than the tensile fracture strength of the ceramic, the tensile test becomes ineffec- tive for evaluating the joint strength. 25 - 20 g -<= g> 15 ~ 1;) gj 10 Q. 5 Ground and resintered Ground Fig. 9. Results of the peel tests performed on the Cusil-ABA@filler metal. High peel strengths (20 Ib (90 N)) are obtained on the ground-and-resintered alumina. In contrast, the peel strengths for the ground substrate are low (6 Ib (27 N)). 2000 20 min brazing -Jf- Liquidus (815'C) Solidus (780'C) 1500 1000 E 500 2 4 6 8 Time (h) Fig. 7. Temperature cycle for Cusil-ABA@ brazing. Fig. 8. Sample assembly and brazed samples for the peel test. Alumina substrate, a foil of an active filler metal (Cusil-ABA@),and three strips of Kovar@are shown at the bottom of the figure. At the top-right of the figure is a brazed sample and at the top-left is the sample where one Kovar@strip has been peeled off from the substrate. Shear Test The shear test provides important in- formation on the mechanical properties of the joint, namely shear strength. Figure 12 schematically illustrates two common- ly used shear tests. In the test shown on the left side of Fig. 12, two blocks of the materials are brazed together and pushed in opposite directions.]) The major prob- lem in this type of shear test is that a rotating moment can act on the specimen because of the loading mechanism. Therefore, to eliminate the rotation of the sample, a hold-down force (FH) is re- quired. The hold-down force leads to a compressive stress at the joint, and, hence, the shear stress measured will depend on the hold-down force. The larger FH, the larger the shear strength. The fillet at the joint can also affect the result when the size of the test blocks are not sufficiently large, compared with the fillet size. Therefore, the accurate evaluation of the joint shear strength cannot be performed in this type of shear test. Another com- mon type of the shear test is shown on the right side of Fig. 12. In this case, a ceramic rod is brazed in a metal ring. The ceramic rod is then pushed and the shear strength of the joint is determined. How- CERAMIC BULLETIN, VOL. 68, NO.9, 1989 (@ACerS) 1595 1200 1000 800 S5 ill :; 600 a; Q; 0.. E 400 200 0 ] 0 -- .. - ... - 16Gsil~ABB I ,-" . I Fig. 10. Scanning electron micrographs taken at the brazed joints, (A) The interface between the filler metal (Cusil-ABA@)and the ground- and-resintered alumina is very smooth, whereas (B) that between the filler metal and the ground alumina contains small cracks, ever, because of the thermal expansion mismatch between the ceramic rod and metal ring, the joint is in compressivestress (as in the case of the compression joint as described previously). The compres- sive stress depends on process variables, e.g., a gap between the ceramic rod and the metal ring, cooling rate, etc. There- fore, changes in the process variables af- fect both the shear strength of the joint and the compressive stress at the joint. As a result, the effects of the process var- iables on the shear strength cannot be properly evaluated in this type of test. Double-Brazed Shear Test A new type of the shear test, namely a double-brazed shear (DBS) test, is uti- lized in the present investigation, The ba- sic principle of the DBS test is illustrated schematicallyinFig. 13.Threeblocksare Material B F2 FI-' Fig, 13. Schematic illustration of the dou- ble-brazed shear test. Fig, 11, (L-R) Test-specimenassemblies (ceramic-filler metal-metal-filler metal-ceramic and ceramic-filler metal-ceramic), a brazed sample, and a tested sample for tensile test. The assembly shown on the left includes metal-ceramic joints and, hence, is used for the present investigation, Fs FH Fs Braze FH Ceramic Block on Metal Block Fs Ceramic rod Braze Fs Fs Ceramic Rod in Metal Ring F2 Fig, 12, Schematic illustrations of two common types of shear tests on brazed joints, 1596 CERAMIC BULLETIN, VOL. 68, NO.9, 1989 (@ACerS) - --- - ~ ~ - ." Fig. 14. (L-R) Test-specimen assembly, a brazed sample, and a tested sample for the double-brazed shear test. . I Fig. 15. Test apparatus for the double-brazed shear test. butt-brazedtogetherandthecenterblock is pushed against the two end blocks. Therefore, shearstressis applied on two brazedjoints simultaneously.That is, the propertiesof two joints can be tested in a singleDBS test. It is found, however, that theshearstrengthsof bothjointscan- not be determinedby a singletest. This is because once onejoint (the weaker one) fractures, the loading condition on the other joint becomes undesirable (this point will be explained in detail later). Conse- quently, one DBS test provides the fol- lowinginformation:0) it provides the shear strength of the weaker joint and (ii) it can determine which joint is weaker. There- fore, when two different materials are used at the end blocks, one can measure the shear strength of the weaker joint and also determine which material forms the stronger joint with the material used as the center block. When the same material is used as the end blocks, the shear strength of the weakerjoint obtainedby the DBS test provides a safety factor for the design engineer. Figure 14 shows the test specimen as- sembly, before and after brazing and also after testing. The material for the center block is alumina ceramic (AL-995) and the two end blocks are alloy 42 (Fe-41Ni). All three blocks have the same dimen- sions of 12.4 mm X 12.4 mm x 15.9 mm (0.490 in. X0.490 in. x 0.625 in.). Cusil- ABA@is used as a filler metal. Brazing is performed at 830C for 20 min. After the brazing, two opposite surfaces of the brazed specimen are ground to assure that these two surfaces are parallel to each other for proper loading. The DBS test apparatus is shown in Fig. 15. The brazed sample, also shown in Fig. 15, has been coated by amorphous boron nitride powder to prevent friction. It is then inserted in the test apparatus through square holes on two facing side- walls. Therefore, no hold-down load is ~Instron Corp., Canton, MA. Processing Conditions for Double-Brazed Shear Test Specimens Filler metal thickness Cooling rate Ceramic Metal (I'm (in.)) (OC/min) Table II!. SP'Ccimen A Ground and resintered* B Ground C Ground and resintered D Ground and resintered E Ground and resintered F Ground and resintered *1650Cfor I h. '975C for I h. necessary. End covers attached to the walls prevent fractured fragments from flying out through the holes. These covers do not touch the specimen, and, hence, no additional stresses can be applied on the specimen. After aligning the specimen carefully, a plunger is inserted into the main cavity of the assembly. The bottom of the plunger is flat so that the load is applied over a whole area of the center block. The test assembly with the speci- men is placed in an Instron testing ma- chine.~The plunger is pushed under a constantcrossheadspeedof 0.25mm/min (0.01 in./min), and the load is recorded by a strip chart recorder. ProcessVariableEffects Effects of four major process variables on the joint properties are examined by the DBS test. They are surface condition of the ceramic member, strength level of the metal member, thickness of the filler metal, and cooling rate during brazing. Processing conditions are summarized in Table III. For specimen A, the ground- and-resintered ceramic material and an- nealedalloy42 havebeenbrazed by a 50- ~m-thick foil of the Cusil-ABA@. The cooling rate is 5C/min, which is a stan- dard cooling rate as shown in Fig. 7. This condition has been regarded as a base condition.For each of the remainingfive specimens, only one of the four variables is changed and the other three variables are kept constant so that effects of a sin- gle variable can be determined. For spec- imen B, the surface of the ceramic ma- Annealed' Annealed As-received (3/4 H) Annealed Annealed Annealed 50 (0.002) 50 (0.002) 50 (0.002) 100 (0.004) 150 (0.006) 50 (0.002) 5 5 5 5 5 1 terial is in as-ground condition. In this case, two shear stress orientations, normal and parallel to the grinding direction, are utilized to investigate effects of shear stress orientation relative to the grinding direc- tion on the shear strength. For specimen C, an as-received alloy 42 is used. The as-received alloy 42 is in 3/4 H condition and has a hardness of 75 on the Rockwell B scale. In contrast, the annealed alloy 42 has a lower hardness of 65 on the Rockwell B scale. Specimens D and E are brazed by thick foils of the Cusil-ABA@, 100 ~m and 150 ~m, respectively. A slow cooling rate (1C/min) is used for braz- ing of specimen F. In all of the tests, one joint fractures before the other, as expected. Figure 16 shows a typical load-displacement curve of the DBS test. (Displacementin Fig. 16 indicates the crosshead displacement and, hence, includes displacements of the test O.OOI-in. displacement '---' 4000 -;;; 3000 g "0 '" .3 2000 1000 Fig. 16. Typical load-displacement curve obtained by the double-brazed shear test (specimen D). 1597 CERAMIC BULLETIN, VOL. 68, NO.9, 1989 (@ACerS) IAIoy 42U t l Cusil-ABA .' ," ". l I Cusil-ABA IAlloy42r....... L I I - -- - - - -_J apparatus and testing machine as well as the test specimen.) One joint has frac- tured at 1750 Ib (7785 N), and then the other joint has fractured at 3500 Ib (15570 N). The load for the second fracture is much higher than that for the first frac- ture, because when one brazed joint frac- tures, the specimen isjammed in the main cavity. In most DBS tests, therefore, test- ing is terminated after the first fracture. The shear strength is calculated simply by dividing the load at the first fracture over a total area of two brazed surfaces. The results of the DBS tests are sum- marized in Table IV. In general, the stan- dard deviations are relatively small, in- dicating that the DBS test has produced consistent results. However, specimen E shows a large standard deviation. Speci- men E has a thick filler metal (150 /lm). It is observed that molten filler metal has seeped out from the joining surfaces dur- ing brazing. Therefore, the thickness of the filler metal after the brazing is not uniform and varies from one specimen to another. As a result, the standard devia- tion for specimen E becomes large. It is also demonstrated in Table IV that the DBS test is sensitive to the process var- iables. (A) , r' ~ I relative to the grinding direction are also investigated. When the shear loading is applied normal to the grinding direction, all specimens except one break at very low loads so that the shear strengths can not be determined accurately. These re- sults indicate that the shear strength nor- mal to the grinding direction is lower than that parallel to the grinding direction. It should be mentioned that the shear strength and standard deviation values in Table IV are calculated from the results of all of the DBS tests performed on spec- imen B. StrengthLevelof theMetal The results for specimen C demon- strate that the strength level of the metal joint properties. When the filler metal thickness is increased to 100 /lm (speci- men D) from 50 /lm (specimen A), an increase in the shear stress' is observed. There are two possible explanations for these results. First, the thicker filler metal contains more titanium atoms available to bond with the ceramic. Second, the thicker filler metal provides more plastic deformation during the cooling cycle to result in lower thermal stress at the joint. When the thickness of the filler metal is increased to 150 /lm, however, the results of the DBS tests become more scattered. As described earlier, this is due to the fact that the filler metal has seeped out from the joining surface. Table IV. Results of the Double-Brazed Shear Tests Standard deviation MPa psi 2.0 1.5 3.6 1.0 7.4 0.8 300 220 530 140 1075 111 ~ . ..- JI"\ :} I ~' . ' .' rl I ~ -=- J lqusil~BA1 - ..,. --, ') ~Usn:-AB~I - ~ "I Fig. 17. Scanning electron micrographs of the joint interfaces between the filler metal and ceramic member. (A) The interface between the filler metal and the ground-and-resintered alumina is very smooth, whereas (B) that between the filler metal and the ground alumina contains cracks at the interface. Surface Condition of the Ceramic Specimen B, with the ground ceramic, has the lowest shear strength (6.2 MPa). This indicates that the surface condition of the ceramic member is very critical for the reliable joint. This result agrees with those obtained in the peel test and tensile test. The damaged surface layer on the ground ceramic leads to the low shear strength. Figure 17 shows the cross-sec- tional views of the joints for specimens A and B at a high magnification. Specimen A shows a defect-free interface (Fig. 17(A, whereas specimen B contains cracks parallel to the interface (Fig. 17(8)). The effects of the shear stress orientation member is also important. When the un- annealedalloy 42 (Rb=75) is used, the shear strength of the joint is 9.4 MPa, compared with that obtained in specimen A (20.5 MPa) with the annealed alloy 42 (Rb=65). The metal member can deform plastically to reduce the thermal stress. The amount of the plastic deformation dependsonthe strengthlevelof the metal member.The lowerthe strength, the more the amount of plastic deformation. As a result, the metal member with the lower strength level leads to the lower thermal stress at the joint which, in turn, increases the joint strength. Filler-MetalThickness The filler-metalthicknessinfluencesthe 30 20 co a. 6 10 a B C Specimen 0 Fig. 18. Summary of the OBS tests for specimens A through O. Bars indicate the average shear strength and solid circles in- dicate data points. It is demonstrated that the OBS test is sensitive to process varia- bles. 1598 CERAMIC BULLETIN, VOL. 68, NO.9, 1989 (@ACerS) Averageshear Number strength Specimen of tests MPa psi A 4 20.5 2980 B 4 6.2 900 C 3 9.4 1360 D 2 26.5 3840 E 3 19.6 2845 F 2 18.9 2740 5000 'iji 4000 E, -<: 0, 3000 c Q) 1;5 2000 m Q) -<: (j) 1000 0 A Alloy 42 .- ----. Cooling Rate During Brazing Finally, the effects of the cooling rate during brazing are investigated. Two coolingrates, 5Cjmin and I Cjmin, are utilized. As demonstrated in Table IV (specimens A and F), there is no detect- able difference in the shear strengths be- tween the two cooling rates. Experimental Results Figure 18 shows the results of the DBS tests for specimens A through D in a graphical manner. Bars indicate the av- erage shear strengths and solid circles in- dicate data points. As described previ- ously, the effects of the process variables on the shear strength are clearly shown. It should be also mentioned that all of the data points for each specimen show small scatters, which indicate consistency in the DBS test. Figure 19 shows four broken samples (one for each of specimens A through D). In all tests, cracks have initiated at the right side of the samples. This might sug- gest that a small bending moment has been applied on the sample and has produced a tensile stress at the right-side surface. Although cracks have initiated at the same place, the crack-propagation path varies depending on the specimen. For speci- mens A and D which show high shear strengths (see Table IV), the crack ini- tially has propagated at the joint interface and then has changed direction by ap- proximately 45 into the ceramic. For specimen B, the crack simply propagated at the joint interface. This is due to the fact that the joint strength is very low (6.2 MPa). Specimen C shows that the crack propagated along the joint, but in the ce- ramic. The high-strength metal in speci- Fig. 19. Brokentest samples for specimens Athrough D. Inallcases, the cracks are initiated at the right side of the sample. However, the crack-propagation paths are shown to be dependent on the conditions of the specimen preparations. men C does not deform plastically to re- duce the thermal stress during the cooling cycle. As a result, the ceramic in the vi- cinity of the joint might have been dam- aged and provided the path for the crack propagation. These results indicate that the DBS test can be the optimum testing method for the evaluation of the joint strength be- tween the ceramic and the metal. The DBS test provides consistent results and also is sensitive to the process variables. Fur- thermore, the shear strength values which are obtained by the DBS tests can be used directly for the engineering design of ce- ramic-metal joints. However, the results of the DBS tests also demonstrate that all process variables should be well-defined when designing high-reliability joints be- tween ceramics and metals. Conclusions (I) High-reliabilityjoiningof ceramic to metal requires complete understanding of each and every process variable. (2) The double-brazed shear test pro- vides a meaningful engineering parame- ter for design purposes. (3) The double-brazed shear test is sensitive to process variables in the braz- ing operation. (4) Ceramic surface preparation is the most critical parameter to obtain a reli- able ceramic-metal joint. (5) Use of a metal with a lower yield strength results in higher shear strength at the brazed joint. (6) Thicker filler metal generally leads to higher shear strength. (7) The results of the double-brazed shear tests demonstrate that all process variables should be clearly defined to op- timize engineering design (e.g., finite-ele- ment method) of joints between ceramics and metals. Acknowledgments The authors are deeply indebted to Mr. J."Rosek and Mr. R. Spano, both members of WESGO R&D, for their technical support and helpful discussions. References 'H. Pulfrich, "Ceramic to Metal Seals," U.S. Pat. Nos. 2163407 and 2 163410, June, 1939. 'H. J. Nolte and R. F. Spurch, "Metallizing and Ceramic Sealing with Manganese," U.S. Pat. Nos. 2667432 and 2667427, Jan. 24, 1954. JA. J. Chick and L. J. Speck, "Fabrication of Metal- to-Ceramic Seals," U.S. Pat. No.2 708 787, May, 1955. 'C. S. Pearsall, "New Brazing Methods for Joining Non-Metallic Material to Metals," Mater. Methods, 30, 61-62 (1942). 'F. C. Kelly, "Metallizing and Bonding Non-Me- tallic Bodies," U.S. Pat. No.2 570 248, Oct. 9, 1955. 'H. Mizuhara, "Process of Making a Composite Brazing Alloy of Titanium, Copper and Nickel," U.S. Pat. No.3 561 099, Feb. 9, 1971. 'H. Mizuhara and K. Mally, "Ceramic-to-Metal Joining with Active Brazing Metal," Weld. J., 64 [10] 27-32 (1985). 'E. Lugscheider, H. Krappits, and H. Mizuhara, "Joining of Non-Metallized Ceramic with Metals by Insert of Ductile Active Brazing"; presented at the Second International Colloquium, Joining of Ceram- ic, Glass, and Metal. Bad Nauheim, FRG, 1985. "A. J. Moorhead and H. Keating, "Direct Brazing of Ceramics for Advanced Heavy-Duty Diesels," Weld. J.,65 [10] 17-31 (1986). "'H. Mizuhara and E. Huebel, "Joining Ceramic to Metal with Ductile Filler Metal," Weld. J., 65 [10] 43-51 (1986). "H. Mizuhara, "Vacuum Brazing Ceramics to Metals,"Adv. Mater. Processes,13 [2]53-55(1987). ""Tension and Vacuum Testing Metalized Ceramic Seals," ASTM Specification F19-64. 1964 Annual Book of ASTM Standards, Part 8. American Society for Testing and Materials, Philadelphia, PA. 1Jw.H. Sutton, "Wetting and Adherence of Ni/Ni Alloys to Sapphire," GE Space Sciences Laboratory, Rept.R-64SD44, Philadelphia, PA,1964. . CERAMIC BULLETIN, VOL. 68, NO.9, 1989 (@ACerS) 1599