Solderability Testing of Kovar with 60Sn-40Pb Solder and Organic Fluxes

Flux type, cleaning procedure and solder bath temperature were evaluated for their effects on solderability

BY P. T. VIANCO, F. M. HOSKING AND J. A. REJENT

ABSTRACT. The solderability of 60Sn- the component is most often secured to pability of the solder. 40Pb solder on Kovar was examined as a the printed circuit board by soldering the The calculation of 6Q is based upon the function of surface cleaning procedure, Kovar leads to the conductive landing. measurement of two properties of the flux and solder bath temperature. Organic Therefore, the wetting behavior of solder solder meniscus formed on the vertical acid fluxes were more effective than a alloys on a Kovar substrate is of particular coupon (Fig. 1) (Refs. 2, 3): the height to mildly activated, rosin-based (RMA) flux interest in the microelectronics industry. which it has risen, H; and the weight of the on chemically etched Kovar with contact The goal of this investigation was to solder contained in the meniscus, W. An angles as low as 29 ± 5 deg as compared quantify and evaluate the wetting behav­ analytic expression for W is given by the to 61 ± 11 deg, respectively. Varying the ior of 60Sn-40Pb (wt-%) solder on Kovar following equation: solder temperature through the range of as a function of flux, solder temperature 215° to 288°C (419° to 550° F) caused an and the procedure used to clean the Ko­ PH I 4TLF W = pg- (2) insignificant change to the contact angle var surface prior to tinning. These results w for the RMA flux and a decrease of the were then compared with wetting exper­ where p is the solder density; g is the ac­ contact angle for a candidate water-based, iments performed on nickel-plated and celeration due to gravity; and P is the pe­ organic acid flux. The dilution strength of gold-nickel-plated Kovar. In addition, rimeter of the coupon. The values of p, g the flux and the elapsed cleaning time sig­ T-peel strength measurements were made and P are known. Values of H and W are nificantly influenced the solder-flux inter­ of Kovar oxygen-free, high-conductivity measured experimentally by the menis- facial tension, Y . T-peel strengths of Ko­ LF (OFHC) copper joints formed by the sol­ cometer and wetting balance, respec­ var 60Sn-40Pb oxygen-free, high-conduc­ der alloy to determine whether this data tively. Therefore, Equation 2 is solved for tivity (OFHC) copper joints had a low would correlate with the wettability of the the value of 7LF, which is then introduced correlation with the contact angle derived Kovar. into the following expression that calcu­ from the solderability experiments. The The wetting behavior or solderability is lates the contact angle: results of the solderability tests, T-peel quantified by the contact angle, dc, formed mechanical tests and subsequent mi­ between the liquid solder and the sub­ ( PgH2\ croanalysis of the as-soldered and T-peel strate as defined in Fig. 1 (Ref. 1). The 0c = arcsin\j-— J (3) samples revealed that with the use of an contact angle is dependent upon the RMA flux, the best results were achieved equilibrium balance of the interfacial ten­ Two other parameters originating from with electropolishing and a 240° to 260°C sions as expressed by Young's equation: the experimental analysis are the rate at (464° to 500°F) solder temperture. A rel­ which the solder wets the substrate, W, atively low contact angle of 31 ± 2 deg 7SF - 7SL = 7LFCOS0C (1) and the time required for the meniscus to was observed with no evidence of crack­ reach its maximum weight, tw- ing or thick film intermetallic formation where 7SF is the solid (substrate)-flux in­ terfacial tension; 7SL is the solid-liquid (sol­ at the Kovar-solder interface. Further­ Experimental more, T-peel strengths were nominally der) interfacial tension, and 7LF is the 9.4 ± 0.5 X 106 dyne/cm. liquid-flux interfacial tension. The smaller Materials the value of 8Q, the better the wetting ca- The Kovar (29Ni-17Co-0.2Mn-bal Fe, Introduction nominal wt-%) coupons measured 0.028 X 2.54 X 2.54cm(0.01 X 1.0 X 1.0 The unique thermal expansion proper­ KEY WORDS in.) and were exposed to the following ties of Kovar have allowed this material to heat treatment prior to testing in order to have many engineering applications, from Solderability Testing simulate an industrial glassing process:1 1) glass-metal seals to conductor leads in mi­ Kovar Solderability chemical etch, 2) high-temperature oxida­ croelectronic packages. In the latter case, 60Sn-40Pb Solder tion, and 3) second chemical etch. A sec­ 60Sn-40Pb for Kovar ond set of samples was electroplated with Organic Acid Fluxes 2.4 to 6.0 /xm of nickel. A third group con- P. T. VIANCO, F. M. HOSKING AND j. A. Surface Cleaning RE/ENT are with Sandia National Laboratories, Rosin (RMA) Flux Interfacial Tension Albuquerque. N. Mex. 1 T-Peel Strength Aside from ensuring uniform surface condi­ Paper presented at the 19th International A WS Electropolishing tions on all of the samples, this process mod­ Brazing and Soldering Conference, held April eled the engineering application for which this 19-21, 1988, In New Orleans, La. study was conducted.

230-s | JUNE 1990 SOLID (sample coupon)

COUPON FLUX

LIQUID (solder alloy)

SOLDER CO YSF-YSL=\F ^C z—" —x (a) (b) Fig. 1—A — Testing geometry for solderability evaluation; B —equilibrium balance for interfacial tensions, described by Young's Equation. sisted of nickel-plated specimens that were The nickel- and gold-nickel-plated sam­ tom. To determine the correct value of subsequently electroplated with 0.43 to ples were cleaned by an agitated rinse of W, the buoyancy force acting on the 0.57 p,m of gold. trichloroethylene followed by a similar coupon must be calculated. Referring to The solder used in the wettability ex­ rinse in isopropyl alcohol to remove or­ Fig. 3, which shows the geometry of the periments was a 60Sn-40Pb alloy. ganic contaminants from the specimen risen meniscus, the total immersion depth The five candidate fluxes examined surfaces. is [(H - Vo) + (ID)] where (ID) is the depth here are:1) Flux A-RMA, 2) Flux B-OA- to which the coupon extends below the CHOH, 3) Flux C-OA-CHOH, 4) Flux Solderability Testing solder surface and (H-V0) is the additional D-OA-H20 and 5) Flux E-OA-H20. submersion due to the risen meniscus. Flux A is a rosin-based, mildly activated The meniscus height, H, was evaluated Total buoyancy force is then given by: (RMA) flux while the other four are halide- on the meniscometer, which is a traveling BF = pgwt[(H - V ) + (ID)] free, organic acids (OA). The solvent is al­ stage microscope capable of measuring 0 (4) cohol (CHOH) for fluxes B and C and the vertical movement of the meniscus. Six sample coupons were evaluated per water (H20) for fluxes D and E. Unless Each sample was allowed to preheat for test series. The maximum meniscus weight otherwise stated, all fluxes were used in a 30 s at a distance of 1 cm (0.4 in.) above was taken as the average value between dilution of one-to-one (1:1) by volume the solder before immersion into the pot 10 and 20 s of testing. With the average with isopropyl alcohol. at a rate of 1.3 cm/s to a depth of 0.375 meniscus height determined from the me­ cm (0.15 in.). The meniscus height was re­ niscometer data and the average menis­ Cleaning Procedures corded after 20 s. Four samples were cus weight compiled from the wetting evaluated per test series. balance tests, the solder-flux interfacial A brief description of the cleaning pro­ The meniscus weight was measured by tension and contact angle were calculated cedures is presented below. The details means of a microbalance. A schematic di­ for each test. The absolute error on yLf are found in the appendix. A set of plain agram illustrating the meniscus weight as a and dc was the maximum and minimum Kovar samples was etched for 2 min in 300 function of time is shown in Fig. 2. The values based upon the range of values of mL of a commercially available chemical physical behavior of the meniscus respon­ the meniscus height and weight estab­ etchant composed of a mixture of nitric, sible for certain features of the wetting lished by plus-or-minus one standard de­ hydrochloric and phosphoric acids that curve has been included at the figure bot­ viation about the sample mean. was held at 80° to 85°C (176° to 185°F). The remaining cleaning steps detailed in the appendix were included to rinse away the corrosive agents of the etchant. The I Fig. 2 - Typical samples were coated with flux immedi­ , wetting curve ately after cleaning to protect the freshly describing meniscus etched surface. /C weight as a function of time. An electropolishing process was per­ formed on some of the plain Kovar cou­ W MAX MUM pons using a 250 mL solution of 9 parts MENISCUS MENISCUS WEIGHT.W glacial acetic acid to 1 part perchloric acid WEIGHT (60%) (Ref. 4) held at 30°C (86°F). The (dynes) coupon was the cathode, while the anode BUOYANCY TIME was made of AISI Type 347 stainless steel \ J FORCE, (ID) (sec) screen. The polishing time was 5 min and the current density was 0.124 A/cm2 (un­ der current control). In addition, subse­ quent processing steps were designed to remove the corrosive electrolyte from the sample surface. As in the etching proce­ dure, samples were fluxed immediately after cleaning.

WELDING RESEARCH SUPPLEMENT I 231-s The three components were assembled, Table 1--Solderability Parameters for Chemically Etched Plain Kovar and Plated Kovar preloaded with a 20-g weight, and then

»c 7LF 7SF - 7SL W tw placed on a heated copper platen being F!ux (deg) (dyne/cm) (dyne/cm) (dyne/s) (s) held at 268° ± 2°C (514°F). Observa­ tions determined that a dwell time of 45 s Plain Kovar D 29 ± 5 430 ± 30 380 ± 10 1290 ± 50 3.2 ± 0.2 ensured complete reflow of the solder; C 45 ± 6 510 ± 90 360 ± 20 1400 ± 100 2.5 ± 0.2 after which the completed sample was E 53 ± 5 540 ± 80 325 ± 15 1390 ± 150 2.8 ± 0.2 removed from the heat source and al­

B 53 ± 11 290 ± 100 175±50 600 ± 100 3.8 ± 0.5 lowed to cool. A 61 ± 11 650 ± 200 315±So 1230 ± 50 4.7 ± 0.3 Testing of the T-peel sample was per­ formed on an Inston tensile tester at a dis­ Nickel-Plated Kovar 1 A 24 ± 2 380 ± 20 350 ± 15 1300 ± 150 3.97 ± 0.06 placement rate of 0.085 cm-s" . Four D 24 ± 3 480 ± 30 440 ± 15 1330 ± 20 4.4 ± 0.9 specimens were conducted per test con­ dition. The strength, defined by Fig. 4B, Gold-Nickel-Plated Kovar was taken as the average of the four A 11 ± 5 360 ± 30 350 ± 20 1490 ± 50 2.0 ± 0.2 specimens. The absolute error was plus- D 15 ± 3 460 ± 20 440 ± 10 1670 ± 200 2.3 ± 0.2 or-minus one standard deviation of the (a) Flux diluted 1:1 by volume with isopropyl alcohol. four values.

Finally, the wetting rate was measured coupons and OFHC copper sheet using Results from the slope of the wetting balance the 60Sn-40Pb alloy in order to model a Solderability as a Function of Flux curve starting at the point of zero menis­ prototype joint. A schematic diagram of Chemically Etched—Plain Kovar and Plated the test specimen geometry and the load- cus height (point B in Fig. 2) to where the Coupons curve begins to deviate from a linear displacement output appear in Fig. 4. The trend. Although taking the wetting rate Kovar segment was cleaned similarly to The solderability data comprised of dc, directly from the recorder output does the solderability coupons and then imme­ 7LF, (7SF —J'SL ). W and tw for plain Kovar not account for the buoyancy force, the diately coated with the particular flux of coupons, which were chemically etched, time-dependent contribution to the buoy­ interest. The copper sheet that formed are shown in Table 1. The solder temper­ ancy force caused by the meniscus rise the other segment was degreased, etched ature was 260°C. Similar results from tests was confirmed to have had a negligible in a 1:1 dilution of HCI and deionized wa­ on the nickel-plated and gold-nickel-plated effect on wetting rate values. ter for 15 s, and then passed through sev­ coupons that were coated with Fluxes A

eral rinsing steps to remove any acid res­ (RMA) and D (OA-H20) have been in­ idue. The copper was coated with Flux A cluded in the table. For the case of plain T-Peel Mechanical Tests (RMA). The 60Sn-40Pb solder (0.0254 cm Kovar, Flux D (OA-H20) caused the low­ T-peel pull tests were conducted on thick) was cut to size and degreased in est contact angle at 29 ± 5 deg, while the solder joints formed between the Kovar trichloroethylene and isopropyl alcohol. RMA Flux A had the highest value at

COUPON 0.95cm KOVAR THICKNESS = 0.030cm y»

.©- SOLDER FILM

SOLDER COPPER (OFHC) (ID) THICKNESS = 0.025cm

• V (Coupon) -4, FORCE _y —TH-Vr Q (ID) (Solder) L DISPLACEMENT

W F /dyne/dyness \ PEEL STRENGTH = S = f (. *m -1 Fig. J -Meniscus geometry describing the buoyancy force correction.

(b) Fig. 4 —A -Schematic diagram of the T-peel test specimen; B —force- displacement curve of the T-peel test.

232-s | JUNE 1990 I

CD 1500 - CJ 1500 - - T^l X""^ / / I- F- _ I I 500 / - I 500 o a LU 0 LU 0 s 3 (/I -500 " =1 " / (/) -500 - u C/l \j U LU -1500 _ tn s ^ -1500 A • B E

TIME (sec) TIME (sec) Fig. 5 — Representative wetting curves of chemical etched, plain Kovar coupons tested at 260°C as a function of flux. A — Flux C (OA-CHOH); B — Flux A (RMA).

61 ± 11 deg. The other fluxes resulted in ting balance curves of the chemically the use of Fluxes C (OA-CHOH) or D contact angles between 45 and 53 deg. etched, plain Kovar samples coated with (OA-H20). Fluxes C (OA-CHOH) and D (OA-H20) Fluxes A and C. Flux B caused a curve sim­ Observations of the solder film on plain had the two largest values, of (7SF — 7SL). ilar to that of A, while Fluxes D and E re­ Kovar coated with the RMA Flux A indi­ The Fluxes A (RMA) and E (OA-H20) had sulted in wetting curves similar to C. The cated extensive dewetting after the cou­ slightly lower values, and Flux B (OA- wetting curve of a gold-nickel-plated sam­ pon was removed from the solder bath. CHOH) showed the poorest (7SF — 7SL). ple is shown in Fig. 6. While a similar curve An optical micrograph of this phenome­ It is apparent in Table 1 that the solder- was observed for this flux on the nickel- non appears in Fig. 8. flux interfacial tension, 7LF, was very sen­ plated coupons, the Flux D gave rise to As a consequence of the surface attack sitive to the particular flux and had a sub­ plots like that of Fig. 5A for both types of by the chemical etchant, preferred wet­ platings. stantial influence on the value of 8C. The ting of the grain boundaries was observed surface tension (in vacuum) of 60Sn-40Pb Also, Fig. 6 shows the meniscus weight with use of the Fluxes C (OH-CHOH), D -1 solder is estimated at 510 dyne-cm (Ref. reaching a maximum and then decreasing (OA-H20), and E (OA-H20). A sample 5). The wetting rate, W, and time to max­ slightly with time. These phenomena can scanning electron micrograph of this phe­ imum meniscus weight, tw, did not de­ be caused by a time-dependent change to nomenon is shown in Fig. 9. This behavior pend upon the behavior of dc- A linear 7LF (such as caused by high-temperature was not observed when Fluxes A and B least squares analysis between W and dc polymerization of the flux —Ref. 6), or were applied to the samples. demonstrated a correlation coefficient changes in 7SL due to alloying effects be­ Scanning electron micrographs of the (Revalue) of 0.06. tween the substrate and the solder. The solder-Kovar interface revealed no evi­ The gold-nickel- and nickel-plated cou­ decrease in W was not consistent for a dence of cracking or separation between pons showed values of dc, which were given flux between the different samples the two components. Chemical analysis well below those of the plain Kovar sam­ so that volatilization of flux constituents was along the Kovar-solder interface and in ples for comparable fluxes. In addition, the not a likely cause. The effect was greatest the bulk solder film by electron micro­ values of W were generally higher and for the gold-nickel-plated coupons. probe analysis did not detect iron, nickel those of tw, lower than for the gold-nick­ In several tests of plain Kovar fluxed or cobalt dissolved from the Kovar sur­ el-plated samples at the same parameters with either A (RMA), B (OA-CHOH), or E face. On the solder film surface, porosity and residues were evident with the use of for the plain Kovar specimens. The nickel- (OA-H20), the solder did not wet some plated samples had slightly higher values regions of the surface. A sample wetting Fluxes C (OA-CHOH), D (OA-H20) and E of tw and comparable wetting rates when balance curve of such a specimen that (OA-H20); were sharply diminished with compared to the plain Kovar results. was coated with Flux A, is shown in Fig. 7. Flux B (OA-CHOH); and were absent with Shown in Fig. 5 are representative wet­ This phenomenon was not observed with the RMA Flux A.

1 i

1500 / •>n>

1- F- X _ / 500 - Cl ; ts 3 in -500 - - U u w z LU -1500 & -

i ° 5 10 15 20 TIME (sec) TIME (sec) Fig. 6 — Representative wetting curve of the Au/Ni-plated Kovar speci­ Fig. 7- Wetting curve of a nonwetting sample (chemically etched, Flux mens tested at 260 °C with Flux A (RMA). A) tested at 260"C.

WELDING RESEARCH SUPPLEMENT 1233-s Electropolishing Process—Plain Kovar Table 2--Solderability Parameters vs. Chemical Etching Time (Flux D) A comparison of solderability results I- between plain Kovar coupons that were z Chemical chemically etched and those that were Etching Time 7SF - 7SL W tw s fc 7LF electropolished is made in Table 3. Chang­ a (min] (deg) (dyne/cm) (dyne/cm) (dyne/s) (s) o ing to the electropolish treatment had lit­ -1 0 180 tle effect on the contact angle of samples 3.9 >Ul 0.5 48 ± 10 530 ± 150 355 ±55 1050 ± 50 ± 0.7 fluxed with B (OA-CHOH), D (OA-H20) Ul 39 ± 10 4 ± 1 1.0 450 + 100 35011% 980 ± 150 and E (OA-H20). For Flux D, this obser­ **ns 2.0 32 ± 5 420 ± 40 360 ± 15 1260 ± 150 3.0 ± 0.6 vation was caused by both 7LF and a 4.0 34 ± 3 460 ± 30 380 ± 10 1250 ± 20 4.3 ± 0.3 (7SF — 7SL) remaining unchanged. How­ ever, the other two fluxes, B and E, showed increases in both 7LF and (7$F

For Flux D (OA-H20), the time period dissolved into the solder film. Shown in — 7SL), which left dc unchanged. While a during which the plain Kovar coupons Fig. 10A is a scanning electron micrograph moderate decrease of dc was observed were exposed to the chemical etchant (backscattered electron image) of the sol­ with the use of Flux C (OA-CHOH) be­ was varied between 0 and 4.0 min. The der film profile on a nickel-plated speci­ tween the two cleaning procedures, a solderability results are outlined in Table 2. men. Figures 10B and 10C are the nickel large decline of 0c from 61 ± 11 to A new batch of test samples was used for and tin x-ray maps, respectively, of the 31 ± 2 deg was recorded for samples the 2.0-min etch; hence, the slight varia­ region shown in Fig. 10A. The intermittent coated with the RMA Flux A. This sub­ tion from the D (OA-H20) data of Table 1. regions of coincident concentrations of tin stantial decrease of contact angle was due The data indicate that the behavior of dc and nickel were quantitatively determined entirely to a drop in the value of the sol­ was more sensitive to the interfacial ten­ to have the composition of the tin-nickel der-flux interfacial tension, 7I_F- Although sion, 7LF, than (7SF — 7SL). The minimum intermetallic, Ni3Sn4. some dewetting was observed, it was not value of dc was achieved with the 2.0 min Dissolution of the gold layer on the as severe as that on the chemically etched etching time. As expected, no wetting gold-nickel-plated coupons was also in­ coupons. Also, as with all of the other took place on samples that received no vestigated by the microprobe technique. fluxes, Flux A on electropolished plain Ko­ surface etch treatment. The value of W As the thickness of the solder film in­ var did not exhibit any nonwetting was a maximum and tw a minimum for the creased (or the further away from the cases —Fig. 7. 2.0-min etch. Otherwise, the parameters meniscus edge that one examined the film For the fluxes, the electropolishing pro­ W and tw demonstrated no apparent cross-section), the more complete was cedure caused a higher wetting rate as com­ trend with etching time nor with the other the dissolution of gold. Optical micros­ pared to the chemically etched samples. solderability parameters. copy of the solder film surface revealed an Yet, no obvious trend was observed be­ Similar analyses were performed on the extensive array of gold-tin intermetallic tween the pairs of values of tw- Finally, no nickel-plated and gold-nickel-plated spec­ crystals and an overall solder film mor­ alterations in the generalized shape of the imens. Electron microprobe analysis re­ phology that was dependent upon the wetting curves between the two cleaning vealed that some of the nickel coating had flux in use. processes were noted for the fluxes. m|rwfe!l

Fig. 8 — Optical micrograph of a dewetted sample (chemically etched, plain Kovar, Flux A, 260 L'C).

Fig. 9 —SEM micrograph (secondary electron) of the meniscus front on chemically etched, plain Kovar coated with Flux E. Solder film is to the left; the Kovar surface to the right.

234-s | JUNE 1990 ui

CL o -I ui >

X o oc < Fig. 10— A — SEM micrograph Ul (backscattered electron) of the tn solder film cross-section on Ni-plated Kovar (Flux D, 260°C). The Kovar coupon is to the left of z the film. B — x-ray map of the UJ nickel concentration of the 5 o. micrograph in A. C —x-ray map of o tin. _l >UJ Ul o o tr. < UJ tn ui oc

Q_ o —I ui > Ul Q oX tr < ui tn ui oc

Effect of Flux Dilution served for the samples fluxed with D. A the range tested. Samples of the wetting slope of —0.08 deg/°C and a correlation curves from tests using both fluxes and 2 Flux D (OA-H 0) was selected to de­ 0. 2 coefficient (R2) of 0.74 were measured by solder temperatures of 215° and 245°C termine the effect of flux dilution. The o a linear least squares analysis of the data. are shown in Fig. 13. The solderability re­ plain Kovar coupons were chemically For the RMA Flux A, a correlation coeffi­ sults are listed in Table 4. With the RMA etched (after a 20 s diluted sulfuric acid > 2 UJ etch to remove a heavier than normal ox­ cient (R ) of 0.27 confirmed no statistical Flux A, increasing the solder temperature dependence of dc on temperature over caused a significant increase in the wetting a ide). Shown in Fig. 11 is a graph of the a contact angle and solder-flux interfacial tr tension as a function of the degree of di­ < UJ lution of the flux. The behavior of dc fol­ E Fig. 11- tn o Dependence of t9c ui lowed closely that of 7LF with a minimum V. oc at 50 to 60% dilution. The values of the - i i a> and 7y on the c strength of Flux D I- parameter, (7SF — 7SL), did not contribute M >. O) •D dilution on the z UJ to the overall trend of dc. The wetting to chemically etched, 3 50 500 S rate, W, and the value of tw showed no o - plain Kovar (260°). a significant dependence on the degree of UJ Solid circles are the O _l 1 -i dilution. o O t?c data and the >ui z CO open circles are the < 40 1 450 » - Z 7IF data. Effect of Solder Temperature > LU O I- X Contact angle as a function of the sol­ < o 2 30 [ I 1 - 400 < tr der bath temperature for plain Kovar < samples coated with Flux A (RMA) or Flux O LU o < CO D (OH-H20) is shown in Fig. 12. The spec­ LL o tr ui imens received the electropolishing treat­ i i i i 350 20 LU tr ment. A small decrease in contact angle 20 40 60 80 100 with increasing temperature was ob­ PERCENT VOLUME (FLUX)

WELDING RESEARCH SUPPLEMENT 1235-s Table 3—Solderability Parameters for Electropolish vs. Chemical Etching Process

Ac (deg) 7LF (dyne/cm) 7SF - 7SL (dyne/cm) W (dyne/s) tw(s)

Flux C-Etch E-Polish C-Etch E-Polish C-Etch E-Polish C-Etch E-Polish C-etch E-Polish D 29 ± 5 30 430 30 420 ± 30 380 ± 10 364 7 1290 ± 50 1530 150 3.2 ± 0.2 3.0 ± 0.6 C 45 ± 6 37 510 90 460 ± 30 360 ± 20 370 15 1400 ± 100 1620 150 2.5 ± 0.2 2.1 ± 0.1 E 53 ± 5 53 540 80 740 ± 90 325 ± 15 450 20 1390 ± 150 1670 100 2.8 ± 0.2 3.1 ± 0.2 55 290 640 ± 70 600 ± 53 ± 11 100 175 ±§0 370 15 100 1060 100 3.8 ± 0.5 4.6 ± 0.5 61 ± 11 31 650 200 400 ± 20 340 1230 ± 50 1390 50 0.3 315 ±30 10 4.7 ± 3.9 ± 0.1

pons; both having been coated with RMA Table 4—Solderability as a Function of Temperature for Flux A (RMA) and Flux D (OA-H 0) 2 Flux A. A rough correlation was observed between the nominal T-peel strengths and T W 7LF 7SF - 7SL tw dc of the chemically etched Kovar data. Flux (°Q (dyne/cm) (dyne/cm) (dyne/s) (53) An optical micrograph of the Kovar side A 215 380 ± 20 320 ± 10 460 ± 100 11 1 of a sample, which was chemically etched 230 390 ± 10 340 ± 10 1000 ± 30 5.7 0.2 and coated with Flux E (OA-H20), is 245 390 ± 15 330 ± 10 1230 ± 100 4.1 0.3 shown in Fig. 14. Voids that had formed 260 390 ± 15 340 ± 10 1390 ± 50 3.9 0.1 during the reflow process are quite nu­ 288 310 ± 40 270 ± 20 1490 ± 200 3.3 0.2 215 450 ± 20 370 ± 10 1310 ± 50 3.9 0.2 merous. An analysis comparing the area 230 460 ± 20 390 ± 15 1540 ± 50 4.1 0.2 fraction of voids and the T-peel strengths 245 450 ± 30 390 ± 20 1390 ± 100 3.6 0.5 of five individual samples is shown in Ta­ 260 420 ± 30 360 ± 20 1310 ± 50 3.2 0.2 ble 6. No dependence was found be­ 288 460 ± 20 410 ± 10 1340 ± 50 3.1 0.3 tween the void fraction and strength. Observations of the fracture surfaces revealed two prevalent paths of crack propagation: 1) at the Kovar-solder inter­ rate and a corresponding decrease in the trend over the temperature range for this face (Fig. 15A), and 2) near the copper- value of tW- Flux D (OA-H20) did not ex­ flux. solder interface (Fig. 15B). The topogra­ hibit a temperature dependence to the Qualitative observations of the solder phy of the latter case has isolated areas of wetting rate. However, the decrease of films showed that the amount of residues large deformation originating from failure tw with increasing temperature corre­ and the difficulty of their removal in­ of the bulk solder film and a facetted sur­ sponded to a diminishment of the nega­ creased with increasing temperature for face layer incorporating minor fissures and tive meniscus weight observed at the start both fluxes. The same temperature de­ cracks in the brittle, copper-tin interme­ of the wetting curve for the lower tem­ pendence was recorded for the amount tallic that had formed at the copper-solder perature tests. of solder film porosity generated by Flux interface. The values of 7LF and (7SF — 7SL) as a D. function of the solder temperature (Table 4) were generally the same in the temper­ T-Peel Strength Measurements ature range of 215° to 260°C (419= to Table 5—Contact Angle and T-Peel Strength 500°F) for Flux A (RMA). However, both The T-peel strength data for the plain parameters dropped sharply at 288 = C Kovar coupons that were chemically Be T-Peel Strength (106 dyne/cm) (550°F) causing a slight decrease in dc- For etched and coated with one of the five Flux (deg) the Flux D (OA-H20) data, the overall fluxes appear in Table 5 together with the Plain Kovar, Chemically Etched trend of the parameter (7SF — 7SL) was contact angle determined by solderability D 29 5 11.0 0.5 found to increase with temperature and testing. Also included are the data derived C 45 6 10.0 0.7 was chiefly responsible for the decreasing from the electropolished plain Kovar spec­ E 53 5 10.2 0.5 11 behavior of dc- 7LF exhibited no particular imens and the gold-nickel-plated cou- 53 9.5 0.4 61 11 7 2

Plain Kovar, Electropolished A 31 ± 2 9.4 ± 0.5 40 Fig 12- I 1 1 1 Dependence of 8c • FLUX A (RMA) Cold-Nickel-Plated Kovar on the solder Ul O FLUX D (OA-H2O) A 11 ± 5 10.4 ± 0.2 temperature for the 0) 2 35 - Fluxes A (solid I circles) and D (open circles) for the I Table 6—Void Formation and T-Peel electropolished, plain I 30 - Strength Kovar (260°). 1 9. [ 1 1 -' I j) Area Fraction T-Peel Strength < C H Flux/Sample of Voids (106 dyne/cm) Z 25 - o A/1-4, F 0.045 5.6 o A/1-1, F 0.034 6.7 E/A-1 0.103 9.6 20 1 1 I E/A-3 0.074 10.5 180 200 220 240 260 280 D/C 0.047 9.5 SOLDER TEMPERATURE (deg C)

236-s | JUNE 1990 ' I

• -

0) 1500 01 1500 - i c Z •>D . •>D . /' ui t- V- / 2 X 500 - I 500 Q, tS O / UJ 0 / LU 0 / o 3 / g —I tn -500 - tn -500 - ui 3 / 3 > (J uCO CO / X 7 / o LU •1500 LU -1500 0C fc < A B Ui I I I , I 10 10 TIME (sec) TIME (sec) z 1 ' I I ui 2 a. 1500 1500 - O / / .,—" T3 > I " / ui H C I 500 " 500 - - a o tu ' 1 0 LU o 5 \ 1 3 cc -500 cn -500 -\ 1 - cn - < 3 3 UJ U in o tn CO UJ Z oc LU -1500 1500 c D I 10 10 2 TIME (sec) TIME (sec) a Fig. 13 - Representative wetting curves of electropolished, plain Kovar as a function of flux and temperature. A — Flux A, 215 "C; B — Flux A, 245 'C; C o- Flux D, 215 "C; D- Flux D, 245 °C. > UaJ >-x . From a visual examination of the T-peel o tr sample surfaces, a mixture of fracture < paths was observed comprised of failures Um Fig. 14-Optical tn at both the Kovar-solder and copper- ui solder interfaces. When the percent of micrograph of the CC 1 $ ^ T-peel fracture failure at the Kovar-solder interface was surface of the Kovar reviewed as a function of the T-peel leg of the chemically UJ strength of the specimens, an inverse re­ etched, plain Kovar 2 x sap* if . - a lationship was observed between the test sample coated O two parameters. The data are listed in 1 _j with Flux F. UJ Table 7. > Ul a Discussion x o Parameters Controlling the Contact Angle ' '' ' . : '• : . < ui The wetting behavior of 60Sn-40Pb tn solder on Kovar was quantitatively de­ ui OC scribed by the contact angle, dc. The smaller the value of dc, the better is the H wettability or solderability. The flux serves f z to decrease the value of 0 by two J ui C 3 i 1 2 processes: 1) by creating and protecting m JmMm a the freshly cleaned and deoxidized sur­ o * * mmmM _l face and 2) by lowering the solder-flux in­ ui terfacial tension. With regards to the pa­ •• mW > rameters in Young's Equation, the mech­ 4-Jr X anism gives rise to a larger value of o (7SF _ 7SL) by increasing 7SF- It is assumed tr < that the value of 7SL is constant where the Ul tn solder has wet the substrate, irrespective ui of the prior condition of the substrate, oc since the surface contamination is dis­ m placed by the advancing liquid alloy. In- 1.0 mm.

WELDING RESEARCH SUPPLEMENT | 237-s hand, Table 3 illustrates tests in which the ment, compared to that which was chem­ Table 7—T-Peel Strength as a Function of same flux was applied to differently ically etched. Therefore, the chemically Fracture Morphology cleaned surfaces. The result was signifi­ etched surface should have more surface cant changes in 7LF. the characteristics of contaminants. Examining the data for Flux % of Failure T-Peel Contact which were very dependent upon the A (RMA) in Table 3, the value of (7SF - 7SL) at Kovar-Solder Strength Angle 6 particular flux in use. Flux Interface (10 dyne/cm) (deg) increased only slightly from the chemical The dependence of 7LF on the substrate etch to the electropolishing process. This A 80 ± 5 7 + 2 61 ± 1 surface condition is based upon the by­ observation suggests that the flux was ca­ B 35 ± 5 9.5 ± 0.4 53 ± 7 pable of achieving nearly the same surface C 15 ± 5 10.0 ± 0.7 45 ± 2 products generated through the corrosive D 10 ± 5 11.0 ± 0.5 29 ± 1 action of the flux as it cleans the surface. conditions after either one of the cleaning E 7 ± 5 10.2 ± 0.5 53 ± 7 The reaction products, which are carried treatments. Therefore, the greater quan­ away by the flux, may favorably segregate tity of reaction by-products generated to the solder-flux interface, if the result is from the chemically cleaned samples and a decrease in the system's free energy. It dissolved in the flux coating were respon­ trinsically, the effectiveness of each flux to has been observed experimentally (Refs. sible for the larger 7LF values. increase (7SF — 7SL) and decrease 7LF de­ 7, 8) that the interfacial tension between A similar relationship between the in­ pends upon the specific chemicals contained two liquids can be extremely sensitive to terfacial tension, 7LF. and the substrate in the material (or dilution). The extrinsic fac­ such surface active contaminants present surface condition was observed in the tors affecting dc include operating tempera­ in either material. On the other hand, sur­ data of Mayhew and Wicks (Ref. 2). tures, as well as the condition of the substrate face inactive solutes, by remaining dis­ Although the substrates were copper, the surface at the time of soldering. solved in one of the liquids, may so change effect of surface condition on 7LF is clearly It was observed in much of the data that the bulk properties of that phase so as to demonstrated. Shown in Table 8 are the the solder-flux interfacial tension, 7LF. also affect the interfacial tension. This hy­ values of dc and 7LF taken from the refer­ pothesis has some bearing on the results played a significant role in the behavior of ence. The value for (7SF — 7si_) was calcu­ of this study. dc- This observation is not unexpected lated by the authors. For these experi­ when comparing data between the dif­ The plain Kovar surface had a brighter ments, the solder was 63Sn-37Pb at a ferent fluxes —Table 1. On the other appearance after the electropolish treat- temperature of 260°C. The flux was wa­ ter white rosin-based with either 0.05, 0.50 or 1.00% activator. The surface oxi­ Table 8—Activator (a) dation of the copper samples increased in the order of A, B, C and D. For the activa­ tor levels of 0.50 and 1.00%, as the degree 0.05' D 0.5% 1.0% Substrate of surface contamination increased, the — Sample f3c 7LF 7SF - 7SL »c 7LF 7SF - 7SL 8c 7LF 7SF - 7SL value of 7LF increased while (7SF 7SL) was constant. This is the same trend that m 17 417 398.8 11.5 376 368 10 368 362 B(c) 38 482 380 19 394 372 18 386 367 was observed between the chemically CM - - 23 413 380 22.5 394 364 etched and the electropolished data for D(c) - - 25.5 416 375 28 423 373 Flux A (RMA) and Flux C (OA-CHOH). For Flux D (OA-H20), no significant (a) Data Irom Ref. 2. change was recorded for the values of 7LF (b) A — degreased and chemically cleaned (c) B. C and D —likewise cleaned, then oxidized by incremental amounts (time) in air, 150 C. Flux —water white resin with given '8 and (7SF — 7SL) between the two cleaning activator. Substrates were copper.

jgr* • „. 1

Fig. 15 - SEM micrographs (secondary electron) of the T-peel fracture surface of chemically etched, plain Kovar-copper sample in which the Kovar was coated with Flux A. A -Failure at the Kovar-solder interface; B-failure near the copper-solder interface.

238-s | )UNE 1990 processes. Therefore, the solder-flux in­ terfacial tension was not sensitive to by­ Table 9—Solderability Parameters for Plated Compared to Electropolished Surfaces products contained in the flux layers that 1- arose from these two test series. A com­ E-Polish Cold-Nickel-Plated Nickel-Plated fc 7LF (7SF - 7SL) fc 7LF (7SF - 7SL) Oc 7LF (7SF - 7SL) UJ parison of 7LF and (7SF — 7si_) between 2 cleaning procedures for the other two Flux A (RMA) a. 31 400 340 11 360 350 24 380 350 O fluxes, B (OA-CHOH) and E (OA-H20), are _i slightly more ambiguous. In both cases, Flux D (OA-H20) UJ 30 420 364 15 460 440 24 480 440 significant increases in 7LF and (7SF — 7SL) > Ul were observed, while dc remained un­ rm. changed—Table 3. The increase in (7SF — 7SL) indicates that the cleaning capabil­ The value of 7LF caused by Flux A was 288 °C was caused by a drop in TLF- The ities of Fluxes B and E were sensitive to the lower and that from the use of Flux D accompanying decrease of (7SF — 7SL) be­ different surface conditions left by the higher than the respective values for plain tween 260° and 288°C is evidence that two cleaning procedures. Kovar. The lower contact angles for the the fluxes' chemical activity had dimin­ Very little change in 7LF and (7SF - 7SL) gold-nickel-plated samples were caused ished slightly. The decrease of (7SF — 7SL) was not the result of increased oxidation was noted for Flux D (OA-H20) between by slightly lower 7LF as compared to the the two cleaning procedures. However, nickel-plated samples. of Kovar surface since the value of TLF increasing the chemical etching time prior The above analysis has been summa­ would increase with the greater volume to using Flux D from 0.5 to 2.0 min low­ rized for each flux in Table 9. With the use of reaction by-products. Rather, the de­ ered the value of 7LF- It is hypothesized of Flux A (RMA), the lower contact angles crease of (TSF — TSL) and TLF was due to a that as the cleaning process became more achieved with either plating was predom­ temperature-sensitive change in the flux. effective at removing surface contami­ inantly the result of a lower value of TLF- For the case of Flux D (OA-H20), the nants with longer cleaning times, the value However, when Flux D (OA-H2O) was decrease in dc with increasing tempera­ of 7LF decreased as fewer corrosion prod­ used, the lower contact angles were ture was due largely to a general increase ucts were entrapped in the flux film. This largely caused by the higher values of in (TSF — TSL). Therefore, the increasing — behavior was not observed between the (7SF 7SL); the Flux D achieved an "im­ temperatures improved the chemical ac­ (2.0 min) etched samples and those elec­ proved" surface condition on the plating tivity of Flux D. tropolished, because for Flux D, the dif­ rather than on the bare Kovar. As for the effect of temperature on ference in degree of surface contamina­ wetting kinetics of Flux A as measured by tion was probably not sufficient to cause Effect of the Flux Dilution W and tw, an increase in W and decrease a change between the two values of TLF- in tw with increasing temperature was In summary, the interfacial tension, 7LF, In Fig. 11, the contact angle as a function observed. Therefore, increasing solder had a significant impact on the contact of dilution of Flux D (OA-H20) exhibited a temperature with use of the RMA Flux A angle data of the plain Kovar samples minimum at approximately 50 to 60% increased the speed with which the flux compiled for the five fluxes. Analysis of concentration. The similar dependence reached a level of chemical activity, which on dilution strength exhibited by 7LF was itself was constant in the temperature the solderability parameters (7SF - 7si_) C and 7LF as a function of the flux and sub­ responsible for the overall trend of 0C. The range of 215° to 260 C as evidenced by strate cleaning procedure indicates that values #c were not sensitive to (7SF — TSL)- the similar values of (7SF — 7SL) and no sig­ aside from differences in 7LF caused by the It is suggested that the dilution depen­ nificant changes to the solder-flux interfa­ dence of 7LF was caused by the simulta­ individual flux chemistries, the reaction cial tension. neous occurrence of two phenomena: by-products generated by the corrosive For Flux D, the overall increase in chem­ one causing 7LF to decrease with increas­ action of each flux also altered the value ical strength with temperature was not ing flux concentration below 50 to 60% of 7LF. accompanied by an increase in the wet­ dilution, and the other increasing the value ting rate. However, the value of tw did of TLF as the proportion of flux increased noticeably decrease upon raising the tem­ Effect of Surface Platings on Solderability at concentrations above the 60% level. perature and can be explained by the de­ The decrease of 7LF as the flux concentra­ The presence of surface plating had a crease of negative meniscus weight —Fig. tion increased between 0 and 60% was pronounced effect on the interfacial ten­ 13. Therefore, tw was a more sensitive in­ caused by a transition from the solder sions. The gold plating was almost com­ dicator of the speed of wetting (than W) surface tension (510 dyne-cm-1) —Ref. 5, pletely dissolved by the solder, so that ef­ as a function of solder temperature. to the solder-flux interfacial tension (ap­ fectively the liquid alloy was wetting the -1 A qualitative analysis was made of the proximately 430 dyne-cm ). The slight nickel underlayer and the values of 7SL effect of the time-dependent tempera­ increase of 7LF beyond 60% resulted from would be comparable between the gold- ture profile of the coupon on W. Hypo- an increased concentration of reaction nickel plated and nickel-plated samples. thetically, if the temperature gradient by-products created by the greater cor­ Therefore, the values of 7SF were also along the coupon was the controlling pa­ rosive action of the dilution. similar for the gold-nickel and nickel-plated rameter for the rate of meniscus rise, then — coupons since the quantity, (7SF 7SL), the meniscus rise should be extremely fast was identical at a given flux. For a given since the coupon temperature at an infi­ plating layer, a larger disparity existed be­ Effect of Solder Temperature nitely small distance ahead of the solder tween the values of (7SF - 7SL) calculated The solder temperature affects the wet­ front would be nearly the same tempera­ for Flux D and those of Flux A than for a ting behavior through two factors: first, ture as that of the solder. The time lag due similar comparison between the same the equilibrium conditions as determined to heating the Kovar substrate just ahead fluxes applied to the plain Kovar surface. by TLF and (TSF — TSL) and secondly by the of the solder was calculated to be negligibly This suggests that the plated surfaces were kinetics of wetting as described extrinsi- small. Thus, if the temperature controlled more sensitive to the different fluxes than cally by W and tw- the rate of meniscus rise, then the solder were the electroplished Kovar surfaces. For Flux A (RMA), no significant change should have risen very quickly and shown The change of 7LF when the gold-nickel in either TLF or (TSF — TSL) was observed similar behavior for the different fluxes. and nickel plating were added to the Ko­ over the temperature range 215° to The above hypothetical case was not var was very sensitive to the flux in use. 260°C (Table 4). The decrease in dc at observed in the present data. First, at sim-

WELDING RESEARCH SUPPLEMENT 1239-s ilar temperatures, the wetting rates (and 31 deg and that of Flux C to 37 deg. The CON IV, p. 53. tw) were quite different between A and D other fluxes exhibited little or no change. 4. Metals Handbook: Ninth Edition. Vol. 9, fluxes. Secondly, the dependencies of W 4) The solder-flux interfacial tension, ASM International, Materials Park, Ohio, p. 52. and tw upon temperature were dissimilar TLF, had a significant effect on the contact 5. Schwaneke, A., Folde, W., and Miller, V. between the fluxes. In conclusion, the angles measured for the different fluxes 1978. Surface tension and density of liquid tin- lead solder alloys. /. Chem. and Eng. Data temperature dependencies of W and tw and surface cleaning treatments. 23:298. were a consequence of the chemical 5) With the use of Fluxes A (RMA) and 6. Lovering, D. 1985. Rosin acids react to kinetics of surface conditioning by the in­ D (OA-H20), solder temperatures from form tan residues. Elec. Pack, and Prod. (2): p. = dividual fluxes and not explicitly the time- 215° to 288 C caused a corresponding 232. temperature profile of the Kovar coupon. decline in the contact angle of about 6 deg 7. Bickerman, ). op. cit., pp. 65-85. with Flux D, and an insignificant change for 8. White, D. 1968. The surface tension of T-Peel Data Analysis Flux A as determined by a linear least liquid metals and alloys. Metall. Rev. 13:73. squares analysis. In both cases, the behav­ From the data in Table 5, the T-peel test iors of dc were not monotonic. results did not precisely follow the trend Appendix 6) Microanalysis of the solder film cross- as recorded for the contact angles. This sections on plain Kovar showed no evi­ Cleaning Procedures observation was not entirely unexpected dence of cracks at the solder-substrate in­ Procedure for Chemical Etching of Kovar due to the complexity associated with the terface nor extensive intermetallic forma­ Substrates mechanical behavior of a joint composed tion between the tin of the solder layer of two different substrates. The data 1. Immerse 2 min in 300 mL of agitated and the cobalt, iron or nickel constituents summarized in Table 7 shows that the in­ etching solution at 80 to 85°C (176° of the Kovar. crease in T-peel strength corresponded to to 185°F). Five coupons cleaned per an increasing amount of failure taking 7) T-peel strength measurements were batch of etchant. conducted on Kovar-OFHC copper joints place at the copper-solder interface. As 2. Immersed 4 min in agitated hot wa­ bonded with 60Sn-40Pb solder in which the Kovar-solder bond strength increased ter. Repeat twice. the Kovar was coated with one of the five to that of the copper-solder interface, a 3. Immersed 75 s in agitated detergent fluxes, and the copper, with Flux A. Only larger percentage of failure took place at solution. a rudimentary correlation was found be­ the copper-solder layer until the strength 4. Repeat Step 2. tween the solderability contact angles and of the Kovar-solder interface exceeded 5. Immerse 75 s in agitated deionized the peel strength values for the surface- that of the copper and nearly all of the water. Repeat twice. etched Kovar. failure took place at the latter region. The 6. Immerse 75 s in agitated acetone. copper-solder bond strength can be ap­ 8) A qualitative evaluation of the sam­ Repeat twice. proximated from the failure of the T-peel ple fracture surfaces did reveal that the 7. Dry with N2 gas. sample that was fluxed with E (OA-H20) fluxes affected the Kovar-solder bond strength; irrespective of extrinsic charac­ since nearly the entire failure took place at Procedure for Electropolishing of Kovar the copper-solder interface, that value teristics such as void formation. Substrates being 10.2 X 106 dyne-cm-1. Therefore, although it is not possible to compute the Acknowledgments 1. Samples are electropolished at con­ numerical strength of the Kovar-solder stant current of 1.5 A for 5 min. Five The authors wish to thank the following joints due to the mixed mode failure, a coupons are polished per 250 mL of individuals whose talents contributed to qualitative assessment of the fracture mor­ electrolyte. this report: Mark McAllaster, R. Ellen Se- phology suggests that the flux does affect 2. Rinse the sample in a jet of deionized merge, David Huskisson and Fred Creu- the intrinsic mechanical strength of the water and store in agitated deion­ lich; all of the Materials Characterization joint. ized water until a batch of five cou­ Dept., Sandia National Laboratories. The pons exists. authors appreciate the comments pro­ 3. Rinse (the batch) for 5 min in deion­ vided by Michael Cieslak and Darrel Frear Conclusions ized water. of the Metallurgy Dept., Sandia National 4. Immerse 2 min in detergent solu­ 1) Solderability testing of 60Sn-40Pb Laboratories, in the review of this report. tion. solder alloy on Kovar was evaluated as a This work described in this paper was 5. Rinse 2 min in agitated hot tap water. function of flux, substrate surface prepa­ performed at Sandia National Laborato­ Repeat once. ration and solder temperature. ries and supported by the U.S. Depart­ 6. Rinse 2 min in agitated deionized 2) Etching the Kovar surface with an ment of Energy under Contract No. DE- water. Repeat once. acid solution followed by application of AC04-76DP00789. 7. Rinse 2 min in agitated acetone. Re­ either an RMA flux, A; one of two organic peat once. acid, alcohol-based fluxes, B or C; or one References 8. Dry with N gas. of the two water-based, organic acid 2 fluxes, D or E, caused respective contact 1. Bickerman, |. 1947. Surface Chemistry for Degreasing Procedure for Plated Samples angles of 61, 53, 45, 29 and 53 deg. All Industrial Research, New York, Academic Press, pp. 12-13. fluxes were diluted one-to-one by volume 1. Immerse 3 min in agitated trichloro- 2. Mayhew, A., and Wicks, G. 1971. Solder­ with isopropyl alcohol. ability and contact angle. Proc. Inter. NEPCON ethylene. 3) Electropolishing the Kovar surface in III, p. 45. 2. Immerse 3 min in agitated isopropyl a solution of acetic and perchloric acids 3. Mayhew, A., and Monger, K. 1972. Sol­ alcohol. improved the contact angles of Flux A to derability by meniscometry. Proc. Inter. NEP­ 3. Dry with N2 gas.

240-s|JUNE 1990