Reflow profile study of the Sn-Ag-Cu solder

Transcription

1 B. Salam School of Engineering, University of Greenwich, Kent, UK C. Virseda European Fuel Cell H. Da School of Engineering, University of Greenwich, Kent, UK N.N. Ekere School of Engineering, University of Greenwich, Kent, UK R. Durairaj School of Engineering, University of Greenwich, Kent, UK Keywords Soldering, Surface texture, Surface treatment Abstract A study of the Sn-Ag-Cu lead-free solder reflow profile has been conducted. The purpose of the work was to determine the Sn-Ag- Cu reflow profile that produced solder bumps with a thin intermetallic compound (IMC) layer and fine microstructure. Two types of reflow profiles were studied. The results of the experiment indicated that the most significant factor in achieving a joint with a thin IMC layer and fine microstructure was the peak temperature. The results suggest that the peak temperature for the Sn-Ag-Cu lead-free solder should be 2308C. The recommended time above liquidus is 40 s for the RSS reflow profile and s for the RTS reflow profile. Received: 17 March 2003 Revised: 6 November 2003 Accepted: 6 November 2003 q Emerald Group Publishing Limited [ISSN ] [DOI: / ] Introduction Tin lead solders are the primary materials used for interconnecting electronic components. However, the safe use and disposal of lead-containing electronic products is an issue that is causing a move in the consumer electronic industry to remove lead from products. It is widely anticipated that the use of lead containing solders by the electronic industry will be seriously constrained by a legislative ban on lead use in solders. An example of such legislation is the environmentally conscious engineering in electronics committee in Japan, which has scheduled that lead-free solders should be standardised from Furthermore, European legislation under the Waste from Electrical and Electronic Equipment (WEEE) and Restriction of Hazardous Substances (RoHS) Directives is scheduled to eliminate the lead from electronic products by the year Besides this legislation, some European countries have considered imposing new regulations requiring the manufacturers to take full responsibility for the recycling of their products. In addition, switching to leadfree soldering can bring some advantages, in particular by improving the reputation of companies, who are environmentally conscious. The industry has embarked on a number of studies in search of suitable lead-free alternatives, but yet there are no drop-in solutions with respect to reflow temperature, joint reliability and assembly costs. Our survey shows that the Sn- Ag-Cu alloy is one of the most promising lead-free alloys currently being evaluated by the industry. However, Sn-Ag- Cu is a high melting point alloy (2178C). Furthermore, the maximum soldering temperature of electronic components/ parts is often fixed. Hence, the safety window is reduced in size. Therefore, the temperature profiling for this alloy becomes more difficult and challenging than for the eutectic Sn-Pb alloy. A study of the Sn-Ag-Cu alloys was conducted by Hwang (2001). The composition of the studied Sn-Ag-Cu alloys was per cent Sn, per cent Ag and per cent Cu. The study found that the microstructure of the Sn-Ag-Cu alloys consisted of Cu 6 Sn 5 and Ag 3 Sn intermetallics in a Snmatrix. These intermetallics (the Cu 6 Sn 5 and Ag 3 Sn) were found to effectively strengthen the alloy and act as a barrier to fatigue crack propagation. In addition, they also act as partitions within the Sn-matrix grains, thereby producing a finer microstructure. The finer the Ag 3 Sn and Cu 6 Sn 5 intermetallics are, the more effectively they partition the Sn-matrix grains, resulting in an overall finer microstructure that can facilitate grain boundary sliding mechanisms, and which in turn will lengthen the life of a solder joint. The presence of these intermetallics in the bulk solder for the traditional Sn-Pb alloy is hardly seen; hence, the The Emerald Research Register for this journal is available at microstructure of a Sn-Ag-Cu lead-free solder joint is different from that of a traditional Sn-Pb solder joint. Since most of the currently available studies of the reflow profile are based on the Sn-Pb solder alloy, the influence of different types of the reflow profile on the microstructure of a Sn-Ag- joint is unknown. Thus, this is one of the objectives of this paper. The reflow profile could affect the reliability of a solder joint, because it is one of the factors that influence the formation of the intermetallic layers in a solder joint. The intermetallic layer is in fact a crucial part of a solder joint. Although having an advantage in facilitating bonding between the solder and substrate, it has a disadvantage that it is generally the most brittle part of the solder joint (Frear, 1991). Thus, it must be as thin as possible. Therefore, a good reflow profile must produce solder bumps with a thin intermetallic layer. Reflow profiling was extensively studied, for example, by Lee (1999), Skidmore and Waiters (2000), Suganuna and Tamanaha (2001), Suraski (2000) and Yang et al. (1995). Suraski (2000) studied the ramp-to-spike (RTS) reflow profile. Suganuna and Tamanaha (2001) discussed the available reflow technology for lead-free soldering. Experiments were carried out to determine which flux chemistries, lead-free alloys and reflow profiles had the greatest influence on solder joint quality in terms of wetting ability, solder balls, solder splashes and voids (Skidmore and Waiters, 2000). A study of eutectic Sn-Ag reflow profiles has also been reported by Yang et al. (1995), who studied the effect of the soldering temperature, soldering time and cooling rate. A study analysing the types of defects affected by the reflow profile has also been reported by Lee (1999). Some of these studies, such as Yang et al. (1995), are closely related to the work reported here. However, there have been no reports of the effect of the reflow profile on the intermetallic layer thickness and microstructure for the Sn- Ag-Cu alloys. Therefore, this paper presents a study of the reflow profile for the Sn-Ag-Cu alloys. The goal of the work was to determine a Sn-Ag-Cu reflow profile that results in a thin intermetallic compound (IMC) layer and fine microstructure solder joint. Experimental design The experiment studied two types of reflow profiles. They were the ramp-soak-spike (RSS) and ramp-to-spike (RTS) profiles. Examples of these profiles are shown in Figure 1. Each of these profiles was investigated using a set of factorial design experiments. The design parameters for the RSS profile were the soak temperature, peak temperature The current issue and full text archive of this journal is available at [27]

2 Figure 1 Types of reflow profiles and cooling rate. The parameters for the RTS profile were the peak temperature, time above liquidus and cooling rate. The purpose of the soak temperature in the RSS profile is to provide heat to an assembly gradually and uniformly. In general applications with Sn-Pb solder, the soak temperature is between 150 and 1708C (Skidmore and Waiters, 2000). However, the soak temperature for the high-temperature solder alloys is between 170 and 1908C (Lee, 1999). This soak temperature setting is critical because the assembly temperature should be raised smoothly and uniformly to the peak temperature so that the temperature differential between the components is minimised. The soak temperature was therefore investigated. Solder paste manufacturers recommend that the peak temperature is as high as possible to aid wetting, but component manufacturers advise to keep the lowest peak temperature possible to prevent component damage (Wickham and Hunt, 2001). A low-peak temperature is also believed to create a thinner IMC layer (Yang et al., 1995). In this study, two peak temperatures were investigated: high ð250 ^ 58CÞ and low ð230 ^ 58CÞ: Extending the time at peak temperature permits any component with a large heat capacity to reach the required reflow temperature. The best results, in terms of good wettability, no solder balls and no voids, have been found to occur with a RTS profile with time above liquidus of s (Skidmore and Waiters, 2000). However, too much heat input above the solder melting point leads to excessive intermetallic formation. Therefore, two different times above liquidus for the RTS profile were investigated: short (40-60 s) and long ( s). For the RSS case, the time above the melting point must be kept as low as possible. In this profile, the time above liquidus depends on the peak temperature, e.g. for the peak temperature 2308C, the lowest time above liquidus was 40 s, whereas for the peak temperature 2508C, it was 60 s. A fine microstructure solder joint could be produced by a fast cooling rate (Prasad, 1989). However, electronic components are generally vulnerable to thermal shock caused by a fast cooling rate; thus, it is necessary to investigate which cooling rate should be applied and its effect towards the microstructure of the Sn-Ag-Cu lead-free solder alloys. Therefore, two cooling rates were studied: slow, 28C/s and fast, 48C/s. In summary, the experiment has three factors with two levels, the details of which can be seen in Table I. The combinations of these parameters were assigned to the L 4 (2 3 ) orthogonal array as shown in Table II. Therefore, each type of reflow profile had four combinations of factors to be tested, as shown in Table III. Experimental procedure The test vehicles used in this study were FR4 substrates with three solder bumps on 2.5 mm diameter circular Cu pads, as shown in Figure 2. The procedure to make the solder bumps started from manually printing the solder paste using a stencil after the bare Cu pads of the test vehicle had been scrubbed and cleaned using iso-propanol. The solder paste used for printing was Sn-3.8Ag-0.7Cu, with a type 3 particle size distribution and 89 wt per cent metal content. The stencil was 0.7 mm thick and had three round apertures of 2.5 mm diameter. To ensure consistent printing, printing tests were conducted. The height of the solder paste deposit produced by the printing test showed good yield with 20 mm deviations between the deposits. This repeatable and good yield manual printing was possible because the size of the stencil aperture was large and the particle size distribution of the solder paste was mm (type 3). The volume of the solder bumps after reflow was estimated from the volume of the solder paste printed and the solder alloy volume fraction of the paste was found to be 1:66 ^ 0:1mm 3 : The next procedure was forming the solder bumps by reflowing the solder paste deposit according to the list of reflow profiles in Table III. A forced convection reflow oven (batch type) was used to reflow them. The first step in this was setting-up the reflow oven so that the test vehicle could be reflowed following the required temperature profiles. The reflow profiling procedure is divided into two steps: thermocouple attachment process and oven set-up. Three thermocouples were used to measure the PCB temperatures. A thermocouple was attached on each pad of the test vehicle as shown in Figure 2. A high melting point solder (88Pb/ 10Sn/2Ag) was used to attach the thermocouples after scrubbing and cleaning the pads with iso-propanol. Once the thermocouples were attached to the specimen, the reflow oven was set-up to create the required thermal profile (Table III). The oven parameters, adjusted to create the thermal profile, were: zone set point temperatures and conveyor speed. The profiling is an iterative process comparing the test board temperature results with the desired profile. If the results differ, the oven parameters are adjusted and the test is repeated. Table I Experimental parameters Factor no. High level (H) Low level (L) RSS reflow profile 1 Soak temperature C C 2 Peak temperature and time above liquidus 250 ^ 58C and 60 ^ 15 s 230 ^ 58C and 40 ^ 15 s 3 Cooling rate Fast Slow RTS reflow profile 1 Peak temperature 250 ^ 58C 230 ^ 58C 2 Time above liquidus s s 3 Cooling rate Fast Slow [28]

3 Table III Parameter combinations of the reflow profile RSS profile Soak temperature (8C) Table II The L 4 (2 3 ) orthogonal array Factor number Experiment no H H H 2 H L L 3 L H L 4 L L H Peak temperature and time above liquidus (8C, s) Cooling rate ^ 5 and 60 ^ 15 Fast ^ 5 and 40 ^ 15 Slow ^ 5 and 60 ^ 15 Slow ^ 5 and 40 ^ 15 Fast RTS profile Peak temperature (8C) Time above liquidus (s) ^ Fast ^ Slow ^ Slow ^ Fast Figure 2 Test vehicle design After reflow, the specimens were cross-sectioned to observe the microstructure and measure the intermetallic thickness. Some of the specimens were aged in a climatic chamber at 1508C for 300 h to accelerate growth of the IMC layer and evolution of the microstructure. The aged specimens were also cross-sectioned. The IMC layer thickness was measured using a software attached to the microscope. On the IMC photo taken along the bumps at a 1,000 magnification, a poly-line was drawn along the two borders of the IMC layer, i.e. at both the solder and the substrate side. The border at the solder side is mostly non-planar and the border at the substrate side is often planar. After the lines were drawn, the software calculated the maximum distance, minimum distance and average distance between the two lines. These steps were repeated three times on each bump (middle part, left side and right side) to ensure consistency. The IMC layer thicknesses reported in this paper are the average thickness. Six solder bumps were formed with each reflow profile and, on each bump, there were three measurements of the IMC thickness hence there were 18 measurements (N ) for each reflow profile. The deviation of the measured IMC layer thickness varied, especially between the as-soldered and the aged bumps. The standard deviation of the IMC thickness for the as-soldered bumps was between and 0.5 mm and that of the aged bumps was between and 0.8 mm. The confidence level of the data could be determined by substituting the highest standard deviations (s) obtained for each condition of the bumps in the following equation (Ott, 1988): rffiffiffiffiffiffiffiffiffi NE z a=2 ¼ 2 s 2 ð1þ The interval width (E ) is chosen as 0.25 because it is a reasonable value for the facts that the thinnest intermetallic thickness is around 1 mm and the highest standard deviation of the data is 0.8. Furthermore, if the interval width value is too low, the number of samples (N) have to be increased and if the interval width value is too high, it will not be reasonable, e.g. 1 ^ 0:8 mm ðy ^ EÞ: Thus, the value of z a=2 could be calculated and converted to the confidence levels of the measured data (Ott, 1988) which, for the as-soldered and aged bumps, were 97 and 82 per cent, respectively. In other words, we are 97 per cent (for the as-soldered bumps data) and 82 per cent (for the aged bumps data) sure that the measured intermetallic thicknesses will be within the interval of ^0:25 mm of the values listed in Tables IV and V. Results and discussion The results of this study are divided into two parts: the microstructure evaluation and the IMC layer thickness measurements. Microstructures Microstructures within the Sn-Ag-Cu alloys have been analysed and identified earlier by Hwang (2001) and Moon et al. (2000). Hence, in this paper, the microstructures of the Sn-Ag- bumps were recognised based on their shapes and identities as described by Moon et al. and Hwang. The results of the bulk microstructure observations are summarised in Tables IV and V. The microstructures of the as-soldered bumps generally showed large areas of tin (Sn) populated by small Ag 3 Sn and Cu 6 Sn 5 intermetallic flakes. The finest intermetallic flakes were observed in the as-soldered bumps formed with reflow profile 4. An example of the microstructures can be seen in Figure 3. The microstructures of the aged bumps were generally similar to the as-soldered bumps, except that the intermetallics were more uniformly distributed compared with those in the assoldered bumps (Figure 4). Hence, the islands of Sn in the aged bumps were reduced in size. The aged bumps formed with reflow profile 4 still had the finest intermetallic flakes. The presence of the very fine intermetallics in the bulk solder of the bumps formed with reflow profile 4 might be caused by the low peak temperature, short time above liquidus and fast cooling rate. These fine intermetallics were not present in the bumps formed with the RTS reflow profile, even though (RTS) reflow profile 8 had parameters very similar to the (RSS) reflow profile 4. This might be because the RTS profile did not have as short a time above liquidus as the RSS profile. In addition, the RTS profile will never have that short time above liquidus, because it requires a longer time to achieve an even temperature distribution across the assembly. The interface intermetallics between the bump and Cu substrate are shown in Figures The figures show that profiles 1, 2, 5 and 7 produced a very irregular IMC layer morphology and profiles 3, 4, 6 and 8 formed a more uniform IMC layer. Frear (1991) reports that the slower the dissolution rate the more planar is the interfacial intermetallics. This indicates that those reflow profiles made the Cu dissolve faster into the solder. From Table III, the similarities between those profiles could be summarized. Profiles 1 and 2 have the same soak temperature ( C) as do profiles 3 and 4 ( C), and profiles 5 and 7 have the same time above liquidus ( s) as do profiles 6 and 8 (50-70 s). It seems that the soak temperature in the RSS profile and the time above liquidus in the RTS profile are the factors most affecting the dissolution rate. Further extensive investigation is needed to confirm this finding. IMC layer thickness The IMC layer thicknesses of the samples are shown in Figures 5-12 and Table IV. The measured IMC thicknesses in Table IV are approximately in agreement with that reported by Hwang (2001). The total IMC growth kinetic for the Sn-Ag-Cu alloy was parabolic and can be expressed by the following equation (Hwang, 2001): [29]

4 Table IV IMC layer thickness and microstructure for the RSS reflow profile RSS profile no. IMC thickness and standard deviation (mm) IMC shape Microstructures Non-aged ^ 0.2 Very irregular Large area of Sn + a lot of long Ag 3 Sn + Cu 6 Sn ^0.1 Very irregular Smaller area of Sn than no. 1 + a lot of long Ag 3 Sn + Cu 6 Sn ^0.2 Irregular Large area of Sn + a lot of small Ag 3 Sn ^0.1 Irregular Large area of Sn and very fine Ag 3 Sn Aged ^0.4 Smooth layer, with Large area of Sn + small Cu 6 Sn 5 and Ag 3 Sn microstructure ^0.5 Smooth layer, with Large area of Sn + fine Cu 6 Sn 5 and Ag 3 Sn microstructure ^0.3 Smooth layer Large area of Sn + small Cu 6 Sn 5 and Ag 3 Sn microstructure ^0.8 Smooth layer Large area of Sn + very fine Cu 6 Sn 5 and Ag 3 Sn microstructure Table V IMC thickness and microstructure for the RTS reflow profile RTS profile no. IMC thickness and standard deviation (mm) IMC shape Microstructures Non-aged ^ 0.4 Very irregular Large Cu 6 Sn 5 and Ag 3 Sn ^0.5 Irregular Large Cu 6 Sn 5 and Ag 3 Sn ^0.2 Very irregular Large area of Sn and small size of Ag 3 Sn ^0.1 Irregular Large area of Sn and small size of Ag 3 Sn Aged ^0.4 Smooth layer, with Large area of Sn + small Cu 6 Sn 5 and Ag 3 Sn ^0.8 Smooth layer, with Large area of Sn + small Cu 6 Sn 5 and Ag 3 Sn ^0.1 Smooth layer, with Large area of Sn + fine Cu 6 Sn 5 and Ag 3 Sn ^0.5 Smooth layer Large area of Sn + very fine Cu 6 Sn 5 and Ag 3 Sn Figure 3 Microstructures of non-aged solder joints for different reflow profiles [30] X ¼ X 0 þ 1: t 0:52 257; 700 exp ð2þ RT where X is the total IMC layer thickness (m) after ageing for time t (s); X 0 is the initial layer thickness at time equal to zero; R is the universal gas constant (8.314 J/mol K); and T is the temperature (K). This parabolic growth kinetic implies that the IMC growth in the Sn-Ag-Cu alloy is controlled by bulk diffusion of elements to the reaction interface. From the data shown in Table V, it can be seen that the reflow profile that gave the thinnest IMC layer was reflow profile 4 for the RSS profile and reflow profile 8 for the RTS profile. In order to determine which factors had the biggest effect, variance analysis was conducted, as shown in Table VI. The effect figures in Table VI are the difference between the average IMC thicknesses for the low and high value for each variable and the percentage figures are the ratio of these individual differences to the sum of the effects for all three variables. According to the variance analysis in Table VI, the most significant factor in the RSS profile was the peak temperature. Before ageing, the cooling rate gave a bigger contribution than the preheat temperature, but after ageing, it was opposite. The results suggest that the peak temperature for a thin IMC layer and fine microstructure is 2308C for 40 s. The suggested soak

5 Figure 4 Microstructures of as-soldered and aged solder joints for reflow profile 1 Figure 5 The IMC layer for reflow profile 1 Figure 6 The IMC layer for reflow profile 2 Figure 7 The IMC layer for reflow profile 3 [31]

6 Figure 8 The IMC layer for reflow profile 4 Figure 9 The IMC layer for reflow profile 5 Figure 10 The IMC layer for reflow profile 6 Figure 11 The IMC layer for reflow profile 7 [32]

7 Figure 12 The IMC layer for reflow profile 8 Table VI Variance analysis of IMC thicknesses Soak temperature Peak temperature and time above liquidus Cooling rate RSS profile Intermetallic thickness (mm) High Low High Low High Low Non-aged Average Effect (17 per cent) (64 per cent) (19 per cent) Aged Average Effect (33 per cent) (55 per cent) (12 per cent) Peak temperature Time above liquidus RTS profile High Low High Low Non-aged Average Effect (75 per cent) (2 per cent) (23 per cent) Aged Average Effect (78 per cent) (1 per cent) (21 per cent) temperature is C and the faster cooling rate is recommended. The most significant factor in the RTS reflow profile is also the peak temperature. The second largest contributor is the cooling rate and the reflow time is least significant. The results suggest that the peak temperature for a thin IMC layer and fine microstructure is 2308C. The suggested reflow time is s and the recommended cooling rate is fast. The RSS reflow profile has been proven to form the thinnest IMC layer thickness in this experiment. The thinnest IMC formed by the RSS and RTS reflow profiles are 1.19 and 1.37 mm, respectively. Summary A study of the Sn-Ag-Cu lead-free solder reflow profile has been conducted. The main objective of this study was to determine the most important factors in designing a reflow profile that results in a thin IMC layer and fine microstructure solder joints. The observation of the bulk microstructures of the bumps shows that reflow profile 4, which has low peak temperature, short time above liquidus and fast cooling rate, forms the finest intermetallic (Ag 3 Sn and Cu 6 Sn 5 ) flakes. In addition, from the IMC layer morphology observation, the soak temperature in the RSS profile and the time above liquidus [33]

8 in the RTS profile are found to be the most significant factors in controlling the dissolution rate. The results of the study indicate that the most significant factor in achieving a thin IMC layer and fine microstructure is the peak temperature. The results also suggest that the peak temperature for the Sn-Ag-Cu lead-free solder is 2308C. The suggested time above liquidus is 40 s for RSS reflow profile and s for the RTS reflow profile. References Frear, D.R. (1991), Solder mechanics: a state-of-the-art assessment, TMS, Minerals Metals Materials, Pennsylvania, USA, pp Hwang, J.S. (2001), Environmental-Friendly Electronics: Lead-free, Electrochemical Publications, Isle of Man, British Isles, p Lee, N.C. (1999), Optimizing the reflow profile via defect mechanism analysis, Soldering and Surface Mount, Vol. 11 No. 1, pp Moon, K.W., Boettinger, W.J., Kattner, U.R., Blancaniello, F.S. and Handwerker, C.A. (2000), Experimental and thermodynamic assessment of Sn-Ag- alloys, Journal of Electronic Materials, Vol. 29 No. 10, pp Ott, L. (1988), An Introduction to Statistical Methods and Data Analysis, PWS-Kent, Boston, USA, pp. 131 and A3. Prasad, R.P. (1989), Surface mount technology: principle and practice, Van Nostrand Reinhold, New York, NY, pp Skidmore, T. and Waiters, K. (2000), Optimizing solder joint quality-lead free, Circuits Assembly, pp Suganuna, H. and Tamanaha, A. (2001), Reflow technology, SMT Magazine, pp Suraski, D. (2000), The benefits of a ramp-to-spike reflow profile, SMT Magazine, pp Wickham, M. and Hunt, C. (2001), Thermal profiling of electronic assemblies, National Physical Laboratory (NPL) Report MATC(A)050, September, Available at: ei/publications/abstracts.html#-13 Yang, W., Felton, L.E. and Messler, R.W. Jr (1995), The effects of soldering process variables on the microstructure and mechanical properties of eutectic Sn/Ag solder joints, Journal of Electronic Materials, Vol. 24 No. 10, pp [34]