WELDING RESEARCH

SUPPLEMENT TO THE WELDING JOURNAL, JUNE 1994 Sponsored by the American Welding Society and Ihe Welding Research Council

A Novel Brazing Technique for Aluminum

A simplified and cost-effective method using an alloy powder mixture instead of a clad surface has been developed for brazing aluminum, copper and brass

BY R. S. TIMSIT AND B. J. JANEWAY

ABSTRACT. This paper describes a novel Introduction eutectic composition such as AA4045, brazing technique for aluminum, in AA4047 or AA4343 (Ref. 4). These alloys which at least one of the contacting alu­ The joining of metal parts by brazing contain 9 to 1 3 wt-% of Si and are char­ minum surfaces is coated with a powder- often involves the use of a filler metal acterized by a melting temperature (in a mixture consisting of silicon and a potas­ characterized by a liquidus temperature narrow range near 577°C) (Ref. 5) con­ sium fluoroaluminate flux. Brazing is above 450°C (842°F) but appreciably siderably lower than that of the core alloy carried out by heating the joint to ap­ below the solidus temperatures of the (~660°C). Joining is carried out at ap­ C proximately 600 C in nitrogen gas at core materials. On melting, the filler proximately 600°C (1112°F) in the pres­ near-atmospheric pressure over a time metal spreads between the closely fitted ence of a noncorrosive flux such as a flu­ interval of a few minutes. During heating, surfaces, forms a fillet around the joint oroaluminate salt (Refs. 1, 6) to remove the flux melts at 562°C and dissolves the and yields a metallurgical bond on cool­ native surface oxide films from the con­ surface oxide layers on the aluminum, ing. Over the past several years, nitrogen tacting aluminum surfaces. Oxide re­ thus allowing the silicon particles to furnace brazing (Refs. 1, 2) has been used moval enhances wetting by the molten come into intimate contact with the bare in preference to other techniques for the Al-Si eutectic alloy at the brazing tem­ metal. At temperatures exceeding 577°C, large-scale joining of aluminum parts perature and eases liquid-metal penetra­ the silicon dissolves rapidly into the alu­ such as automobile heat exchangers and tion of the joint. minum and generates in-situ a layer of air-conditioning condensers (Refs. 2, 3). The manufacture of clad aluminum Al-Si liquid alloy of eutectic composi­ In these applications, at least one of the brazing sheet is a multistep process that tion. The filler metal forms a fillet around aluminum components is clad with filler involves casting of the filler metal ingot, the joint by capillary action. A metallur­ metal consisting of an Al-Si alloy of near- subsequent cladding to the aluminum gical bond is formed on cooling. The ad­ core material by hot rolling and final cold dition of Zn powder to the Si/flux powder rolling of the composite material to the mixture allows diffusion of Zn during desired thickness. The preparation of alu­ brazing and modifies the resistance to minum brazing sheet is thus relatively corrosion of the brazed assembly. KEY WORDS complicated and the manufacturing cost The novel brazing technique may be is correspondingly high. The present exploited using elements other than sili­ Brazing work was motivated by the need to sim­ con for generating filler metal. The only Aluminum plify aluminum brazing sheet manufac­ requirement on these elements, such as Eutectic Reaction ture. The development of a brazing tech­ Cu, Ge, Zn, etc., is that they form a rela­ Intermediary Powder nique that obviates the use of clad tively low-temperature eutectic alloy Si/Flux Powder Mix material, without greatly affecting the in­ with aluminum. The technique has been Oxide Removal dustrial brazing practice evolved over used successfully for brazing alu- Temperature Relation the past several years for aluminum, minum/Cu, Cu/Cu and Cu/brass in addi­ Filler Metal Generation would obviously be attractive. The pre­ tion to aluminum/aluminum joints. Si, Cu, Ge vs. Mg sent paper describes such a technique. Low-Temp. Eutectic The process is carried out in nitrogen gas R. S. TIMSIT is with AMF of Canada Ltd., and has been used successfully for braz­ Markham, Ont., Canada,and B. J. JANEWAY ing aluminum/Cu, Cu/Cu and Cu/brass in are with Alcan International Ltd., Kingston R & D Center, Kingston, Ont., Canada. addition to aluminum/aluminum joints.

WELDING RESEARCH SUPPLEMENT I 119-s to yield a good metallurgical bond. Be­ S, particle flux cause fillets are formed through capillary flow of the filler metal, brazing requires only minimal contact force at the joint in­ terface.

Procedures and Results

562°C < T < 577°C Preparation of Surfaces and Joints (c) The successful use of the novel braz­ liquid ing technique requires a uniform coating alloy of Si-flux powder on the metal surfaces prior to brazing. In the present work, the uniform coating could be deposited from T>577°C a water-based slurry after cleaning the (d) surfaces chemically prior to dipping. Chemical cleaning was carried out as fol­ -*- residual flux lows: the surfaces were first immersed for "*" oxide 5 s in a caustic bath consisting of 5% by weight NaOH in water at 65°C (149°F), rinsed in cold water, desmutted for ~2 s in an acid bath consisting of 50 wt-% HNOj, rinsed in cold water and finally dried for a few minutes in a forced-air fur­ nace. This procedure removes oil and Fig. I — Successive steps in the novel brazing process. A — Deposition of a Si/flux powder mix grease residues and yields an aluminum on the aluminum surface. The Si particle dimensions range from ~l to 100 pm; the flux particle surface that is uniformly wettable by dimensions do not exceed I pm; B — melting of the flux at 562°C and dissolution of surface oxide water. The aluminum sheet was then films; C — at 562°C < T < 577°C, solid-state interdiffusion of Si and aluminum; D — af T>577°C, rapid dissolution of Si to form localized pools of filler metal of near-eutectic composition, fol­ dipped for a few seconds into a water- lowed by coalescence of liquid metal pools; E — end of filler metal generation and solidification. based slurry composed of 50 wt-% of the Si/flux powder and 50% distilled water at room temperature. Excess slurry was al­ pm. Brazing is carried out by heating the Brazing Process lowed to drip off following withdrawal, joint at approximately 600°C in nitrogen thus leaving the uniform powder coating The present brazing technique uses gas at near-atmospheric pressure for a on the sheet surface. The Si powder con­ the "eutectic bonding" approach de­ few minutes. During temperature ramp- sisted of commercial-grade material of scribed in earlier publications (Refs. 7, 8), up, the flux melts at ~562°C (1043°F) and 99.1% purity, with Fe as the major con­ but avoids the need to coat the base dissolves the surface oxide layers on alu­ taminant. metal surface with an intimately adher­ minum (Ref. 11), as illustrated in Fig. 1 B. ing layer of the eutectic-forming metal by Sufficient flux must always be present to The aqueous slurries were found to be electroplating or vacuum deposition. In remove these oxides. Oxide dissolution remarkably stable when used with the present technique, at least one of the must occur more rapidly than reoxida- cleaned aluminum surfaces. No effects of aluminum surfaces is coated with a thin tion of the aluminum surface, and it al­ slurry aging on the brazing process were layer of a powder mix consisting of an el­ lows the Si particles to come into inti­ detected over a time interval of two ement capable of forming a low-temper­ mate contact with the bare metal. At this years. Powder deposition was always ature eutectic with aluminum (e.g., Si, juncture, the large elemental concentra­ performed after agitating the slurry to Cu, Ge, Zn) and a flux capable of dis­ tion gradients at the aluminum/Si inter­ generate a reasonably uniform suspen­ solving surface oxide films (Ref. 9), as il­ face cause the aluminum and Si to inter­ sion of the heavy Si/flux powder in water. lustrated in Fig. IA. A commonly avail­ diffuse — Fig. 1C. At temperatures The Si/flux weight ratio in the slurry was able noncorrosive flux (Ref. 6) was used exceeding 577°C (1070°F), it is found varied between 1:2 and 1:3. Depending in the present work. This flux consists of that the Si particles diffuse rapidly into on the concentration of solids in suspen­ a mixture of KAIF4 and K2AIF5 H20 pow­ the aluminum surface and generate in sion, surface coverage by the brazing mix ders in a molar ratio of the respective situ a layer of Al-Si liquid alloy of near ranging from 10-J to 8 x 10~2 kg m-2 could salts of approximately 1 3:1 (Ref. 10), and eutectic composition — Fig. 1 D. The easily be obtained on aluminum. After with a particle dimension of the order of filler metal penetrates the joint of interest dipping, the coating was dried in a cir­ 1 pm. For reasons to be addressed later, by capillary action and forms a fillet, thus culating air oven at 1 70°C (338°F) for 1 we will focus most of our attention on producing a metallurgical bond on cool­ min. The deposit thus obtained was very brazing using Si. ing. Any unused filler metal remains on uniform. The mating surface in the joint The surface coverage by the Si pow­ the aluminum surface to form a layer of was similarly cleaned but was not neces­ der may range from a few to several tens Al-Si alloy of near-eutectic composition sarily coated with powder. of grams per square meter, depending on -Fig. 1E. Other surface preparation and pow­ the joining application, and the weight In this process, only one of the joined der deposition techniques were used ratio of Si-to-flux powder may vary typi­ surfaces need be covered with the Si-flux successfully for brazing. For example, cally from 1:1 to 1:3, also depending on mix because the molten flux spreads vapor degreasing was used to remove the application. Considerably larger Si rapidly across the joint to remove surface residual oil and other organic contami­ surface coverage may be used for spe­ oxide films from the mating surfaces. The nants from the aluminum surfaces. Be­ cific applications. The Si powder particle quantity of filler material formed from cause this decontamination procedure dimensions may range from ~1 to 100 one coated surface is generally sufficient does not yield a surface readily wettable

120-s I JUNE 1994 Fig. 2 — A — External appearances of a typical joint between two AA3003 aluminum sheet spec­ imens. Only the horizontal sheet was coated with Si/flux mix; B — metallurgical cross-section from a typical AA3003 joint. The gray precipitates in the AA3003 components are Al-Fe-Mn-Si constituents. Inset: enlarged view of fillet. The dark needle-like particles consist of Si precipitated from solution during cooling; C — metallurgical cross-section from a joint formed by conven­ produced from the AA3003 base alloy tional brazing wherein the vertical component was joined to a coupon of AA3003 clad on both sides with AA4045 (Ref. 4). Inset: enlarged view of fillet; the Si precipitates are appreciably coarser rather than from AA4045. than those generated by the novel brazing technique as explained in the text. One major attractive feature of the new brazing process is a capability for in­ troducing additional materials into the by water, deposition of a uniform Si/flux ticles in the fillet are Si particles precipi­ surfaces to enhance selected properties. coating required the addition of a wetting tated from solution during cooling after For example, the addition of Zn powder agent to the aqueous slurry (Ref. 2). As an dissolution of the Si powder into the alu­ to the Si/flux mix leads to diffusion of Zn alternative to this procedure, a uniform minum. The fine precipitates in the base into the coated surfaces without ad­ coating could also be deposited from an metal consist largely of All5(FeMn)3Si2 versely affecting the generation of filler alcohol-based slurry without prior chem­ constituents usually present in AA3003. metal. Figure 3 shows a typical distribu­ ical cleaning of the aluminum, since al­ For comparison, Fig. 2C shows a metal­ tion with depth of Zn introduced by this cohol dissolves residual organic contam­ lurgical cross-section obtained from a method. The concentration profile was inants from surfaces. However, non- joint similar to that illustrated in Fig. 2A obtained by wavelength-dispersive Elec­ aqueous slurries are generally not envi­ but formed by conventional brazing tron Micro-Probe Analysis (EMPA) of a ronmentally benign. As an additional al­ wherein the vertical AA3003 specimen cross-section of AA3003 sheet coated ternative for deposition, Si/flux coatings was joined to a coupon of AA3003 clad with a Si-flux-Zn powder mix deposited of acceptable uniformity were generated with AA4045 (Ref. 4). No Si powder was from a water-based slurry. The Zn surface 1 2 on a chemically clean surface by elec­ used in forming the latter joint, although coverage was 4.3 x 10-- kg m and the trostatic spraying. flux was introduced, since the filler metal weight ratio of Si to Zn was approxi­ In the present work, the efficacy of the is generated on melting of the clad alloy mately 1:1. The joint was brazed at brazing technique was evaluated by me­ at approximately 577°C (Ref. 4). The fil­ 600°C using a procedure identical to that chanically assembling the clean and lets in the two figures are essentially iden­ described earlier. The presence of an ap­ coated test components in the desired tical. Generally, the Si precipitates in fil­ preciable quantity of Zn over a depth of joint configuration, heating the assembly lets formed from clad alloys, as in Fig. -100 pm, with a maximum of -1% by in a nitrogen furnace at the selected tem­ 2C, are appreciably coarser than those weight near the surface, was found to en­ perature for time intervals of approxi­ generated by the novel brazing tech­ hance the resistance to corrosion of the mately 2 min, and examining the brazed nique. The difference in the coarseness of surface (Ref. 12). joint(s) after cooling. Si precipitates in the two brazing Although the brazed joints illustrated processes may relate to differences in above were formed with AA3003 core Aluminum/Aluminum Joints filler metal compositions. We recall that components, it was verified that the braz­ the filler metal in the novel technique is ing technique may be used with a wide Figure 2A shows the external appear­ ance of a typical joint formed between two AA3003 aluminum sheet specimens Table 1 - Aluminum Alloy Composition Limits in Wt-% (the compositions of the aluminum alloys used in the present work are listed in AA Alloy Si Fe Cu Mn Mg Zn Cr Ti Others Table 1). In this example, only the sheet 1050 _ shown horizontally in the figure was 0.25 0.4 0.05 0.05 0.05 0.05 0.03 1100 0.95 Si+Fe 0.05- 0.05 — 0.1 — — 0.15 coated with the Si-flux mix. Brazing was 0.2 carried out at 600°C, thus, significantly 3003 0.6 0.7 0.05- 1.0- — 0.1 — — 0.15 above the eutectic reaction temperature, 0.2 1.5 to induce rapid Si diffusion into the alu­ 3102 0.4 0.7 0.1 0.05- — 0.3 — 0.1 0.15 minum as will be described later. Note 0.4 the uniformity of the filler. Figure 2B 4045 9-11 0.8 0.3 0.05 0.05 0.1 — 0.2 0.15 shows a metallurgical cross-section from 6061 0.4-0.8 0.7 0.15- 0.15 0.8- 0.25 0.04- 0.15 0.15 a typical joint. The dark needle-like par­ 0.4 1.2 0.35 X-800 0.05 0.19 0.31 1.11 0.27 — — 0.008 0.15

WELDING RESEARCH SUPPLEMENT I 121-s range of aluminum alloys such as the ex­ that the surface in Fig. 4B is severely dis­ trusion alloys AA1050, AA3102 and oth­ rupted and that an appreciable quantity ers. Figure 4 illustrates the generation of of unreacted Si remains imbedded in the filler metal on surfaces of a commonly surface. The lower Mg concentration in relative intensity, used sheet alloy such as AA1100, X-800 appears to lead to appreciably less Zn Ka line AA6061 and X-800. In all cases, the effi­ surface disruption, as indicated in Fig. 20 -| cacy of filler metal generation was deter­ 4C, but considerable unreacted Si was 18 mined by coating the surfaces with a also left over on that surface. In all cases, 16 layer of the brazing mix and heating as no significant amount of filler metal was 14 described above. The metallurgical produced. These observations are typical 12 - structure of the solidified filler metal of all those made on aluminum alloys 10 - shown in Fig. 4A is essentially identical containing more than approximately 0.1 8 to that observed with other alloys such as wt-% of Mg. The absence of a generation 6 AA1050, AA3102 and others. The quan­ of filler metal in Mg-bearing aluminum 4 tity of eutectic liquid formed corresponds alloys is attributed in part to impairment approximately to the thickness of the sur­ of the liquid flux to dissolve Al203 in the face layer carrying the dark Si precipi­ presence of Mg, due to the formation of 100 150 200 tates. Although this thickness is only ap­ MgF2 at elevated temperatures (Ref. 11). depth (microns) proximately 20 pm in these illustrative It is also attributed to the lack of forma­ examples, the quantity of filler metal may tion of a free-flowing liquid of eutectic be considerably larger because it is con­ composition in the Al-Mg-Si system at trolled largely by the initial Si powder ~600°C (Ref. 5). However, the increased Fig. 3 — Typical distribution of Zn dissolved coverage on the surface, as will be shown state of disruption of the surface (follow­ into aluminum during brazing, by introducing later. As in Fig. 2, the dark needlelike par­ ing attempted brazing) observed with in­ Zn powder into the Si/flux mix. The curve ticles near the surface consist of Si pre­ creased Mg concentration in the alloy shows the relative intensity of the Zn Ka line cipitated from solution on cooling. points to some metallurgical reactions as a function of distance from the surface. The between the Si coating and the base alloy analysis was carried out by wavelength dis­ The cross-sections shown in Fig. 4B before formation of MgF . The nature of persive EPMA on a cross-sectional brazed and 4C were obtained respectively from 2 these reactions was not investigated. specimen prepared from AA 1100 alloy. AA6061 and X-800 specimens after at­ tempting to generate filler metal. Note Magnesium-containing aluminum al­ loys can be brazed by cladding the sur­ face with any aluminum alloy free of Mg, and generating the filler metal from the clad (Ref. 13). The metallographic struc­ ture of the fillet is generally found to be nearly identical to those illustrated in Fig. 2. Filler metal could be generated with a clad layer thickness as small as -45 pm. These observations indicate that Mg dif­ fusion from the core alloy and across the clad layer is generally too slow to have an appreciable effect either on the activ­ ity of the molten flux or on filler metal for­ mation, over the time intervals used in the brazing process.

Aluminum/Cu Joints

Aluminum/Cu joints were brazed by the novel process using the Si/flux braz­ ing mix. It was surmised that the tech­ nique could be used since a ternary eu­ tectic exists at 524°C (975°F) in the Al-Cu-Si system (Ref. 14). Formation of a ternary liquid alloy of near-eutectic com­ position would yield the required filler metal. Prior to brazing, the aluminum component (AA1100) was cleaned chemically as described previously. The Cu section was prepared from commer­ cial-purity material. It was cleaned by etching for 1 min in a 50 vol-% solution of nitric acid, rinsing in distilled water and drying with an air blower. The alu­ minum surface was coated with Si/flux powder using a water-based slurry. Typi­ cal surface coverage by Si and flux were Fig. 4 — Cross-sections obtained from specimens of commonly used aluminum alloys. A respectively -8 x 10-3 kg rrr2 and 25 x AAI 100; B—AA6061; C — X-800.

122-s I JUNE 1994 10-! kg rrr2. Brazing was carried out by heating to 584°C (1083°F) in nitrogen gas to cause rapid melting of the flux and then reducing the temperature to below 548°C (1 01 8°F) as quickly as possible by cutting off power to the furnace and in­ creasing the nitrogen gas flow. The rea­ son for cooling rapidly (-1 min) is ex­ plained below. Figure 5 shows a typical metallo­ graphic cross-section obtained from an aluminum/Cu lap joint. Note the pres­ ence of a relatively wide interdiffusion layer consisting of a eutectic material. Careful examinations by EPMA identified one phase as CuAl2. The composition of the second phase was found to range from ~Cu04AI05Sial to ~Cu02Ala5Si0i3. The eutectic material was separated from the Cu component of the joint by a co­ herent intermetallic layer of approximate composition Cu07AI0 3 with a thickness of ~4 um. Exposure of the joint to tempera­ tures higher than ~548°C (1018°F) for time intervals exceeding ~1 min led to excessive dissolution of Cu into the alu­ Fig. 5 — Typical metallographic cross-section from an aluminum/Cu lap joint. The relatively wide minum sheet. This dissolution stems from interdiffusion layer consists of a eutectic material made up of CuAI2 and aCu-Al phase. the large solubility of Cu in Al in that tem­ perature range (Ref. 5). Despite the pres­ ence of a relatively large quantity of in­ this procedure often leads to nonuniform characterized by an average composition termetallic phase in the interdiffusion fillets because brazing is not initiated layer, the brazed joint was not found to of ~Cu084Si0(nZnn I,. The sequence of uniformly along a joint but begins instead be brittle. Any effect of the Si surface cov­ metallurgical reactions leading to these where surface asperities of the core com­ erage on the strength of the Al/Cu joint fillet compositions was not investigated ponents are in mechanical contact. remains to be investigated. Brazing of an aluminum/Cu joint in the presence of Si probably occurs as fol­ Cu/Cu and Cu/Brass Joints lows: liquid metal is first formed by initial dissolution of Si into aluminum to form a Joining of Cu/Cu and Cu/brass joints layer of Al-Si eutectic liquid at a temper­ was also successfully performed using ature exceeding 577°C (1071°F). As ex­ the novel brazing technique. The brass plained earlier, wetting of the Cu surface consisted of 70/30 Cu-Zn material of by the molten eutectic metal immediately commercial purity. Prior to Si/flux depo­ leads to Cu dissolution to generate a sition, the components were cleaned by molten ternary alloy. This alloy then con­ dipping for 1 min in a 50 vol-% nitric stitutes the filler metal whose composi­ acid solution, rinsing in deionized water and drying. The brazing mix was de­ tion ranges from ~Cu03AI0f>2Sia08 to posited from a water- or isopropyl alco­ Cu002Al097Si001 (Ref. 14) depending on temperature. This conjecture about the hol-based slurry to generate typical sur­ face coverage of Si and flux of-10 x ~\0i filler metal generation mechanism stems 3 from observations that aluminum/Cu and 30 x 1 0- kg m-, respectively. Braz­ joints could not be brazed at tempera­ ing was carried out at an elevated tem­ tures lower than 577°C, i.e., below the perature as the eutectic reaction between eutectic temperature of the Al-Si system. Cu and Si occurs at 803°C (1477°F) (Ref. In the aluminum/Cu joint just dis­ 5). It was found that excellent fillets were cussed, no intimate contact existed ini­ formed in Cu/Cu and Cu/brass joints by tially between the metal components be­ heating for a few minutes respectively at cause the metal surfaces were separated 900°C (1 652°F) and at 876°C (1 609°F) in by the Si-flux particulate. It is worth not­ nitrogen gas. Here again, the brazing ing that aluminum/Cu joints may also be temperature was significantly above the brazed by introducing only flux into the eutectic temperature to induce rapid Si joint and heating to ~565°C (1049°F) to dissolution and rapid formation of filler melt the flux and generate an intimate metal. Metallographic cross-sections aluminum/Cu contact interface. Subse­ from typical brazed joints are illustrated in Fig. 6. Note the excellent fillet geom­ quent interdiffusion of Al and Cu leads Fig. 6 — Metallographic cross-sections from etry. Examinations by EMPA indicated that to filler metal formation since the Al-Cu typical brazed joints. A — Cu/Cu; B — system is characterized by a eutectic re­ the fillets formed in Cu/Cu joints were rich Cu/brass. The compositions of the fillets action at ~548°C (Ref. 5). In practice, in copper, consisting of 90 to 95 at-% Cu. formed in A and B correspond respectively to Fillets formed between Cu and brass were ~Cu09Si0 , and Cu084Si005Zn0 ,,.

WELDING RESEARCH SUPPLEMENT I 123-s because the present focus is a description tively low-temperature eutectic with at of the novel brazing technique. It is ex­ least one of the metals in the joint. Sev­ pected that the metallurgical structure of eral elements such as Cu, Ge, Zn, La, Ca the fillets would be affected not only by and Ba form a low-temperature eutectic the base alloy composition and solidifi­ with Al (Ref. 5); however, use of these cation time, but also by the brazing tem­ metals for brazing aluminum may not al­ perature. In the present work, the use of ways be desirable. Similarly for Cu/Cu a brazing temperature appreciably joining, the selection of an intermediary higher than the Cu-Si eutectic reaction metal from Ge, La, Nd, Sb, Se, and Sr, all temperature may have led to Si depletion of which form low-temperature eutectics from the Cu-Si liquid filler metal by dif­ with Cu (Ref.5), should require careful fusion into the base metal. This would considerations for reasons outlined have affected the compositions of the so­ below. lidified phases as is evident from the Cu- For aluminum brazing, one of the Si phase diagram (Ref. 5). Similar effects major advantages of Si is its relative in­ may have occurred during solidification ertness in water at room temperature and of the Cu/brass joints. its ensuant compatibility with aqueous Because copper oxides are unstable at slurries. Powder deposition from water- elevated temperatures, it was surmised based slurries is straightforward and that Cu-rich materials might braze at el­ widely used in industry in various appli­ evated temperatures in the presence of Si cations because it is simple, environ­ powder alone, without the use of flux. It mentally benign and requires no special was verified experimentally that this con­ handling. The chemical inertness of Si in jecture is incorrect. This observation water accounts for the absence of any ef­ pointed to a surprising efficacy of the flu- fect of slurry aging on brazing. This is not oroaluminate flux, which is used primar­ necessarily the case with other potential ily for brazing aluminum (Refs. 1, 2) to re­ intermediary metals for aluminum braz­ move oxide films from copper and brass ing. For example, Zn powder oxidizes at the elevated temperatures used in this rapidly in water to yield aqueous slurries •(b) work. that become unusable after only a few hours. For commercial applications, the Fig. 7 — Metallographic sections of AA3003 Brazing with Elements Other Than Si use of Zn in an organic carrier such as an aluminum joints brazed with: A — Cu/flux; B alcohol mitigates oxidation but generally — Ge/flux. The insets show magnified views Although the work reported above of the fillets. The fillet in A consists of the Al- does not represent an environmentally has focused on Si as the intermediary ma­ acceptable solution. In addition, the CuAI2 eutectic. The dark particles in the fillet in B consist of Ge and Al-Ge-Fe-Mn con- terial for brazing, the brazing technique large Zn concentration in the Al-Zn eu­ situents. may use other elements to generate filler tectic (-90 at.-%) (Ref. 5) leads to a large liquid. The only fundamental require­ coating weight requirement for the braz­ ment is that these elements form a rela- ing of aluminum, which is not advanta­ geous for many practical applications. Similarly, the chemical reactivity of other elements such as Ba and Ca in air or water and in many organic liquids pre­ cludes use of these metals as brazing ma­ terials. The reactivity of copper with water (Ref. 15) may also preclude pow­ der deposition from aqueous Cu slurries for brazing aluminum. In the following, the use of brazing elements other than Si is illustrated only to show the broad ap­ plicability of the principles on which the novel brazing process is based. Figure 7A and B show metallographic cross-sections of AA3003 aluminum joints brazed respectively with Cu/flux and Ge/flux powder mixes. The Cu- and Ge-containing powders were deposited from an isopropyl alcohol and an aque­ ous slurry, respectively, to yield surface coverages of 45.8 x 10-3 kg m2forCu and 15 x 103kg nrr2 for Ge. The Cu/flux and Ge/flux weight ratios were respectively 1:1 and 1:2. Brazing was carried out at 600°C in nitrogen gas as described ear­ lier. Note that metallurgical bonds are formed in the two cases. As mentioned Fig. 8 — Metallographic cross-section from a typical Cu/Cu joint brazed with Ge/flux powder. The earlier, the filler metal formed with the fillet consists of fine dendrites made up of two Cu-Ge solid solutions of approximate compositions Cu and Ge powders stems from the eu- Cu092Ge008 and Cu008Ge02 (C, phase).

124-s I JUNE 1994 tectic reaction with Al at 548°C and 420°C (Ref. 5) in each case, respectively. In both cases, the joints were heated to 600°C to melt and generate rapid spread­ ing of the flux. As expected, the fillet in Fig. 7A consists of the AI-CuAI2 eutectic. The dark particles in the fillet shown in Fig. 7B consist of Ge and Al-Ge-Fe-Mn constituents. The Mn and Fe present in the latter constituents originate from the base alloy. Figure 8 shows a metallographic sec­ tion from a typical Cu/Cu joint brazed with Ge/flux powder. The brazing mix was deposited from the aqueous slurry Fig. 9 — Typical evolution of surface temperature during heating. The filler metal was formed after mentioned above. Brazing was carried a delay (td) of several seconds at 600°C and was characterized by a sudden drop in surface tem­ out in nitrogen gas at 700°C (1292°F) to perature. The furnace lamps were turned off at a preselected time 7~s after the selected steady-state take advantage of the Ge-Cu eutectic re­ temperature (point A) was reached, tf= (Ts - td). action at 644°C (1191°F) (Ref. 5) to gen­ erate filler metal. Note again the excel­ lent metallurgical bond. Examinations by (0.04 in.) thick and were heated sepa­ was characterized by a sudden but tem­ EMPA revealed that the fillet consisting of rately. Diffusion depths were measured porary drop in surface temperature fine dendrites was composed largely of from metallurgical cross-sections after caused by extraction of the latent heat of two Cu-Ge solid solutions of approxi­ cooling. The surface coverages by Si melting from the heat bath. The delay td powder were sufficiently large to gener­ stems from the finite time required both mate compositions Cu0.92Ge0.o8 and ate reasonably wide dissolution zones. for the flux to melt and for surface oxide Cu0.8Ge025. These coverages ranged from 30x10-3 to films to dissolve into the liquid flux (Ref. 3 2 Formation Rates of Filler Metal 80 x 10- kg rrr . The Si/flux weight ratio 1 7). The furnace lamps were turned off at was -1:3. The substrate surface temper­ a preselected time Ts after initial attain­ ature was monitored by pressing the con­ ment of the selected steady-state temper­ For the range of Si surface coverage trol thermocouple junction lightly ature (location A in Fig. 9). The interval discussed in this paper, empirical obser­ through the brazing mix into contact with during which liquid metal was produced vations suggested that all the available Si the substrate. was given as (T - t ). dissolves into aluminum in a matter of s H seconds at the brazing temperature. The typical evolution of surface tem­ An example of metallurgical cross- Measurements of the formation rate of perature during heating is illustrated in section obtained after a typical experi­ filler metal thus required the use of a Fig. 9. The temperature reached a steady- ment is illustrated in Fig. 10. The width rapid heating technique. These measure­ state value of 600°C in -1 2 s. Formation of the solidified filler metal band formed ments were carried out using a rapid ther­ of Al-Si eutectic liquid occurred after a during the heating time t, = (Ts- t(l) of 1 s mal anneal (RTA) furnace identical to the delay (td) of several seconds at 600°C and is -90 pm. Note that several Si particles type used in the semiconductor industry (Ref. 1 6). In the RTA furnace, the specimen is placed on a quartz tray that slides into a narrow quartz tube of rectangular cross- section. Banks of halogen lamps, con­ trolled automatically by a temperature controller, provide the source of energy. The temperature monitoring system con­ sists of a chromel/alumel thermocouple in contact with the specimen and fed back to control electronics. The furnace is capable of generating an elevated steady-state surface temperature in a small fraction of a minute. The steady- state temperature is controlled to 2°C (4°F). The furnace atmosphere is con­ trolled by a flow of purge gas. All exper­ iments described in this paper were car­ r ried out in nitrogen gas at a steady-state yy,\ y\ dm h temperature of 600°C. The Si dissolution times were deter­ mined by heating AA1100 aluminum coupons, each coated with a selected surface coverage of Si/flux mix, and mea­ 50um suring the depth to which Si had dis­ solved during a selected time interval at the brazing temperature. The coupons were 25 x 75 mm (1x3 in.) and -1mm Fig. 10 — Metallurgical cross-section after filler metal had formed during a heating time f, oi -I s in the RTA furnace. Note that several Si particles remain undissolved above the aluminum surface.

WELDING RESEARCH SUPPLEMENT I 125-s formed at the Al/Si interface as soon as sufficient Si and Al have interdiffused at a temperature above 577°C (say 600°C) to reach the concentration denoted as Cw, The formation of liquid quickly leads to the establishment of equilibrium be­ tween the solid cc-AI solution and the Si layer with an intermediate liquid of com­ position ranging from C,a and C,as illus­ trated in Fig. 12B. As shown by other workers (Refs. 18, 19), the Si dissolution stage can be treated as a moving bound­ ary problem in which Fick's law is ap­ plied to the liquid and solid phases. The equations governing the diffusion process are

0.0 1.0 1.5 time t^ (s) dt dx~ (D 2 Fig. 11 — Measured dependence of filler metal thickness on tf following solidification. The curves dx, d Xe represent calculated values. = D, dt dx' (2) remain undissolved above the aluminum used for the analysis considers interdiffu­ where Xa and X, represent the Si concen­ surface. The results of measurements of sion in a conventional planar Al/Si binary trations respectively in the solid a-AI and the dependence of the interdiffusion couple and is similar to previous models the liquid phase in the diffusion couple, bandwidth on t,are plotted in Fig. 11 .The developed by various authors for tran­ and Da and D^ are the diffusion coeffi­ error bar associated with each measured sient liquid phase bonding (Refs. 1 8, 1 9). cients of Si respectively in a-AI and the value of t represents the maximum esti­ The model is tantamount to assuming liquid Al-Si alloy of near-eutectic com­ mated deviation of t,from the mean value that Si particles diffuse collectively as a position, at the brazing temperature. The of several measurements. The large esti­ coherent layer at a plane interface with solutions of Equations 1 and 2 are given mated uncertainty stems from uncertain­ aluminum. The model thus ignores any as (Ref. 18) ties about both the onset of filler metal effects of the presence of molten flux or formation and the juncture at which liq­ thermal convection currents. Any effects X = A + B erf\x/^AD t uid formation ends. Nevertheless, the of Al alloying additions on Si diffusion a a data of Fig. 11 indicate that the rate of rates are also ignored. The aim of the (3) and filler metal formation is large at 600°C. analysis is to calculate the filler metal r The solid curves show the results of cal­ generation rates expected on the basis of X, =C + Derf(xl-d ADl culations carried out in accordance with Al-Si interdiffusion alone. A comparison (4) the analysis discussed below. of the calculated rates with the corre­ where A, B, C and D are constants to be sponding experimental data would then evaluated from the boundary conditions Analysis of the Filler Metal Formation Process provide insights into the major factors of the problem. From Equations 3 and 4, controlling filler generation in practice. the conditions that x = 0 at t = 0 and that The analysis presented below focuses Consider the Al-Si binary phase dia­ the Si concentrations at the moving on the mechanism of filler metal forma­ gram shown in Fig. 12A. Under equilib­ boundaries remain constant, immedi­ tion for the case in which Si is the inter­ rium conditions, a thin layer of liquid is ately identify the location of the oc-alu- mediary brazing material. The model

800 Al solid phase liquid phase

initial 10 20 interface weight % Si

Fig. 12 — A — Al-Si binary phase diagram; B — under equilibrium conditions at the brazing temperature, a layer of liquid is formed with a Si con­ centration ranging from cfa at the Al/liquid interface, to c,at the liquid/Si interface.

126-s I JUNE 1994 minum/liquid and liquid/Si interfaces as

A„J4Dj for x<0

(5) xt = AeJ4Det for x > 0 (6) v 5 where Aa and Ar, are constants to be de­ termined from the boundary conditions of the problem (Ref. 18). It may readily be shown that substitution of Equations 3 to 6 into these boundary conditions leads to the following equations 7 Af exp(A ~ - A;Da I Dt) _ x 10i- 3 IA.UA. ID. Si loading (kg/m2) C C la o 'ai + Fig. 13 — Dependence of the time required for complete dissolution of a Si coating as a function \~ct 4n(\-ce) of the Si surface coverage corresponding to the coating. The curves were calculated from the pla­ nar diffusion-couple model. exp(-Ag^) atures indicated, is shown by the solid also agreed with the predictions of the \Aa\(l-erf(\Aa\)) curves of Fig. 11. Note the sensitivity of L model. For these calculations, the appro­ (7) to brazing temperature. This sensitivity priate diffusion coefficients were ob­ stems largely from the temperature de­ tained from Fujikawa and Hirano (Ref. Af exp(Af) pendence of the ratio (C,R - C,a)/(1 - Q) 22) and Ejima, etal. (Ref. 23). (Ref. 18). Despite the uncertainty in the {erf(Ae) + erf{\Aa\4DjD~()} experimental measurements, there is Summary and Conclusions broad agreement between the calculated and experimental data in Fig. 11. This This paper reports on a novel brazing agreement is surprising in view of the technique for aluminum and show that K(\-Ct) simplifying assumptions in the model. the technique may be used for joining I (T> The broad agreement between the ex­ other metals. The technique uses the in- In Equations 7 and 8, Aa = -|Aa|; perimental and calculated data of Fig. 11 situ formation of filler metal by the eutec­ these two simultaneous equations were suggests that the rate of arrival of Si par­ tic reaction of an intermediary metal solved numerically to generate values of ticles to the aluminum surface is largely powder with the base metal surface. The Af, and Aa corresponding to the values of unimpeded by the presence of molten major requirement on the intermediary Da, D„ Ca/, Cla and C„ at the selected flux. The agreement also suggests that the metal is that the eutectic temperature be temperatures. Note that Equations 7 and particle arrival rate is considerably larger appreciably lower than the melting point 8 indicate that the constants Aa and A„ than the Si dissolution rate so that the of the base metal. The brazing process vary with temperature because they de­ particles dissolve essentially as a coher­ also requires the use of a flux, preferably pend on the Si concentrations, which in ent layer, i.e., as in a true Al/Si diffusion a noncorrosive salt such as the potassium turn depend on the selected brazing tem­ couple. The favorable comparison be­ fluoroaluminate available as NO­ perature as indicated in Table 2. The dif­ tween experiment and theory indicates COLOK™ flux (Ref. 6), to remove surface fusion constants used in the calculation that the model may provide useful guide­ oxide films from the components to be 2 were D„ = 2.02 x 1CM exp|-136/RT| m lines on total Si dissolution times. Figure joined. Oxide removal is essential for dif­ 2 2 s-> and D, = 2.08 x 10- exp[-25.7/RT] m 13 shows the dependence of the time re­ fusion of the intermediary metal and the s~', where the activation energies are in quired for complete dissolution of a Si ensuing formation of filler metal. With kilojoules, obtained respectively from coating as a function of the Si surface aluminum, the brazing technique has Fujikawa, ef al. (Ref. 20), and Petrescu coverage {i.e., as a function of the effec­ been shown to be successful using Si, Cu (Ref. 21). Note that the variations with tive coating thickness), calculated on the or Ge as intermediary metals. The tech­ temperature of Da and D, have little ef­ basis of the planar diffusion-couple nique cannot be used for joining alu­ fect on Aa and A„ because the ratio Da/D, model. These data indicate that conver­ minum alloys containing Mg in a con­ is small. At 600°C, the ratio Da/Dfr is -2.6 sion of Si to filler metal occurs approxi­ centration as small as - 0.1 wt-%. 4 x 10" . The values of | Aa | and A, calcu­ mately twice as rapidly at 605°C as it The filler metal generation rates ob­ lated from Equations 7 and 8 at tempera­ does at 590°C, for any Si surface cover­ tained with Si powder are in general tures ranging from 590° to 605°C are age. In many brazing applications where listed in Table 2. Si coverages range from - 10 3 to 20 x 3 2 The dependence on time t of the 10~ kg nr , the difference in dissolution Table 2 — Values of Relevant Constants in thickness L of the filler metal layer is times is smaller than 1 s and is thus in­ the Brazing Process Model given as significant in comparison with the total time required for brazing. U2 Temperature (°C) 590° 595° 600° 605° 1/2 U(Dt) + As a final note, we point out that filler L = (4t) metal generation rates were also mea­ 1 Qi 0.013 0.012 0.011 0.010 kite,) sured with Cu/flux mixes on aluminum at Q„ 0.106 0.099 0.091 0.084 (9) 600°C. The filler formation rates were Q 0.135 0.139 0.143 0.146 The dependence of L on time, evalu­ found to be slightly larger than corre­ A, 0.0456 0.0536 0.0605 0.0667 ated from Equation 9 at the four temper- sponding values for the Al/Si system, and Aa 20.7 24.8 8.22 32.5

WELDING RESEARCH SUPPLEMENT I 127-s agreement with those calculated on the thors are grateful to D. Lauzon and to the Technical Paper Series 8701 86. basis of interdiffusion in a planar Al/Si metallographic and electron optics sec­ 12. Baldantoni, A., Janeway, B. J., Lauzon, diffusion couple. This agreement sug­ tions of the Kingston R&D Center of D. C, Purdon, M. J., and Timsit, R. S., 1994. NOCOLOK™ Sil flux — A novel approach for gests that the presence of molten flux Alcan International Ltd. for providing brazing aluminum. SAE Int. Congr., Detroit, does not interfere significantly with the Si technical assistance during the course of Technical Paper Series 940502. dissolution rate. Examples of brazing of this work. 13. Timsit, R. S„ and Janeway, B. J. 1993. Cu/Cu and Cu/brass joints using mixtures Aluminum brazing sheet. U.S. Patent No. of flux and Si or Ge powders show that References 5,232,788. the underlying principles of the novel 14. Eds. G. Petzow, and G. Effenberg. brazing technique are widely applicable. 1. Cooke, W. E., Wright, T. E., and Hirsh- 1992. Ternary Alloys, Vol. 5, ASM Interna­ In principle, it should thus be possible to field, J. A. 1978. Furnace brazing of aluminum tional. braze a variety of metals using an inter­ with a non-corrosive flux. SAE Int. Congr., De­ 15. Tronstad, L., and Veimo, R. 1940. Pre­ mediary element capable of forming a troit, Technical Paper Series 780300. liminary researches on the action of water on 2. Claydon, D. G. W., and Sugihara, A. relatively low-temperature eutectic with copper pipes. /. Inst. Metals, Vol. 66, pp. 1983. Brazing aluminum automotive heat ex­ 17-32. the base metal. For example, brazing of changer assemblies using a non-corrosive flux 16. Singh, R. 1988. Rapid isothermal pro­ Ni or Ni-rich alloys may be carried out process. SAE Int. Congr., Detroit, Technical cessing. J. Appl. Phys., Vol. 63, pp. R59-R114. using Si, Ge, Sm, etc., powders since eu­ Paper Series 830021. 17. Underhill, R., and Timsit, R. S. To be tectic reactions with Ni exist at tempera­ 3. Goodremote, C. E., Guntly, L. A., and published. tures of ~963°C, 762° and 570°, respec­ Costello, N. F. 1988. Compact air cooled air 18. Lesoult, G. 1976. Modeling of the tran­ tively (Ref. 5). Brazing at these elevated conditioning condenser. SAE Int. Congr., De­ sient phase liquid phase bonding process I. temperatures may be attractive in appli­ troit, Technical Paper Series 880445. Report of Center for the Joining of Materials, cations where conventional low-temper­ 4. Fortin, P. E., Kellerman, W. M., Smith, F. Carnegie-Mellon University, Pittsburgh, Pa. ature soldering is undesirable, if a suit­ IM., Rogers, C. J., and Wheeler, M. J. 1 985. Alu­ 19. Tuah-Poku, I., Dollars, M., and Mas­ minum materials and processes for automo­ able flux is available. Similarly, it should salski, T. B. 1988. A study of the diffusion tive heat exchanger applications. SAE Int. be possible to braze other metals, such as bonding process applied to a Ag/Cu/Ag sand­ Congr., Detroit, Technical Paper Series wich joint. Met. Trans. A, Vol. 19A, pp. refractory materials, using appropriate 852228. 675-686. intermediary powders. The use of a pow­ 5. Ed. Massalski, T. B. 1990. Binary Alloy 20. Fujikawa, S. I., Hirano, K. I., and der mix consisting of two intermediary Phase Diagrams, ASM International. Fukushima, Y. 1978. Diffusion of silicon in metals may also be considered for spe­ 6. Wallace, E. R., and Dewing, E. W. 1976. aluminum. Met. Trans. A, Vol. 9A, pp. cific applications. In principle, there are Joining of metal surfaces. U.S. Patent No. 1811-1815. no constraints to the combination of base 3,951,328. 21. Petrescu, M. 1 970. Liquid state atomic metals and intermediary metals that may 7. Niemann,). T, and Garrett, R. A. 1974. mobility of silicon in the unmodified eutectic be used. In practice, constraints may Eutectic bonding of boron-aluminum struc­ silumin. Z. Metallkude, Vol. 61, pp. 14-18. arise from such factors as the unavail­ tural components. Welding journal, Vol. 53, 22. Fujikawa, S., and Hirano, K. 1989. Im­ pp. 175-184. ability of a flux capable of stripping sur­ purity-diffusion of copper in aluminum. De­ 8. Wells, R. R. 1976. Microstructural con­ fects and Diffusion Forum, Vol. 66, pp. face oxide films at the brazing tempera­ trol of thin-film diffusion-brazed titanium. 447-452. ture, the formation of brittle Welding journal, Vol. 55, pp. 20-27. 23. Ejima, T., Yamamura, T, Uchida, N, intermetallics, and the cosmetic proper­ 9. Timsit, R. S. 1992. Method of brazing Matsuzaki, Y., and Nikaido, M. 1980. Impurity ties of the joint. aluminum. U.S. Patent No. 5,100,048. diffusion of fourth period solutes (Fe, Co, Ni, 10. Thompson, W. T, and Goad, D. G. W. Cu and Ga) and homovalent solutes (In and Tl) 1976. Some thermodynamic properties of into molten aluminum. Japan Inst. Met., Vol. Acknowledgments K3AIF6- KAIF4 melts. Can. J. Chem. Vol. 154, 44, pp. 316-323. pp. 3342-3349. All work reported in this paper was 11. Field, D. )., and Steward, N. I. 1987. carried out while both authors were em­ Mechanistic Aspects of the NOCOLOK flux ployed at Alcan International. The au­ brazing process. SAE Int. Congr., Detroit,

Disc Rim Full Disc Rebuild Rebuild

CORRECTION Shaft Overlay

Figure 1 in "Electromechanical Repair of Turbomachinery Shafts by SAW," by Richard LaFave and Richard Wiegand (April 1994), appeared without labels. The complete figure is shown at right. In the first paragraph on P. 43 of the same article, reference is made to "12'A-in. (6.4-cm) diameter bolt holes." This should have read, "1 2 2.5-in. (6.4-cm) diameter bolt holes." Xf V V Fig. 1 — Types of repair.

128-s I JUNE 1994