Aging characteristics of immersion tin surface finishes: an investigation of this surface finish's aging and its reliability during soldering.

The technology of surface finishes for printed circuit boards (PCBs) is seeing a dramatic shift from hot air solder leveling (HASL) toward alternative finishes like immersion tin. This trend is caused by the worldwide environmental pressure to stop using lead for electronic assemblies and by the demands of modern assembly technology, which require a higher coplanarity of surface finish for surface-mount assembly.

One of the main benefits of HASL is the extremely good resistance of the surface against aging under high temperature conditions. Alternative finishes, due to their limited thickness, offer excellent coplanarity but inferior surface protection. A detailed understanding of the aging processes is, thus, a prerequisite for the implementation of alternative surface finishes in high-yield PCB production. In this article, the aging characteristics of immersion tin surface finishes are investigated and correlated with their reliability during soldering operations.

Aging of tin surface finishes occurs by intermetallic compound (IMC) formation at the tin/copper interface due to solid state diffusion. Therefore, the kinetics of the process are assumed to follow a simple law, and, below, the kinetics of IMC formation were investigated experimentally and correlated with data from literature.

The Experiment

Immersion tin layers were prepared with systematically varying layer thickness. The samples were annealed at 155[degrees]C with variation of the annealing time. The layers were characterized with x-ray fluorescence (XRF), electrochemical stripping coulometry and atomic force microscopy (AFM). Solderability of the samples was determined with a solder balance method.

The Results

The solderability of pretinned copper surfaces may be negatively affected by aging effects. The quantitative understanding of the involved aging processes is essential for the implementation of immersion tin in high-yield PCB fabrication.

Standard textbook knowledge (1) regarding the aging characteristics of tin layers on copper is rather detailed. The conclusion is that aging is due to the formation of intermetallic compounds at the tin/copper interface, namely the [eta]-phase (C[u.sub.6]S[n.sub.5]) and the [epsilon]-phase (C[u.sub.3]Sn). The growth of the IMC is due to solid state diffusion and follows a simple kinetic law:

[d.sub.IMC] = [K.sub.T]*[square root of (t)] [1]

with [d.sub.IMC] the thickness of the compound layer, t the time and [k.sub.T] the temperature-dependent proportionality constant.

Regarding experimental values for [k.sub.T] note that different authors obtained largely inconsistent results. A number of reasons have been given for this inconsistency, and, apparently, [k.sub.T] is not really a constant but seems to decrease with time.

To elucidate the kinetics of intermetallic compound formation for the immersion tin layers, a systematic characterization of the annealing process was performed. Samples were annealed at room temperature and 155[degrees]C with variation of the annealing time. The intensity of the XRF signal did not change significantly in the course of the annealing process. Thus, XRF was determined not suitable for characterizing the IMC formation.

Determination of the layer thickness with stripping coulometry and AFM shows, in contrast to XRF, a significant and systematic dependence on the aging process. As an example, Figure 1 shows a series of stripping curves from samples of the same test board. The samples were annealed for different times at an annealing temperature of 155[degrees]C and characterized with stripping coulometry.

With XRF, a layer thickness of 1.42[+ or -]0.08 [micro]m was measured, irrespective of the annealing time. Evaluation of the stripping experiments (Figure 1) gave thickness values of 1.54[+ or -]0.05 [micro]m, 1.27[+ or -]0.05 [micro]m, 1.17[+ or -]0.05 [micro]m, 1.0[+ or -]0.05 [micro]m, and 0.81[+ or -]0.05 [micro]m for annealing times of 0 hour, 0.5 hour, 1 hour, 2 hours and 4 hours, respectively, at 155[degrees]C. These values were confirmed by AFM step height measurements, as shown by the example in Figure 2 for the sample after a 1 hour anneal.

[FIGURE 1 OMITTED]

The results of the systematic investigation are shown in Figure 3 for the annealing temperature of 155[degrees]C. Four different samples with different layer thicknesses were processed and characterized in parallel.

According to equation [1], the thickness of the IMC is proportional to the square root of time at a constant annealing temperature. Since the IMC can only grow at the expense of the tin layer, the thickness of the latter necessarily decreases linearly with the square root of time. Figure 3 shows that the plots of the tin layer thickness versus the square root of the annealing time are fairly linear and the same slope applies for all samples.

Since some investigators found a significant dependence of IMC growth rates on the preparation (2) or thickness (3) of tin layers, the aging characteristics were cross-checked for galvanically deposited tin layers of 1 and 4 [micro]m thickness. No difference in the tin consumption rate was found between immersion and galvanic tin, nor any dependence on layer thickness.

At 155[degrees]C, good linearity is observed up to an annealing duration of 8 hours (Figure 3). The rate of tin consumption due to IMC formation is evaluated as 0.36[+ or -]0.03 [micro]m/[square root of (h)]. Hence, 0.72[+ or -]0.06 [micro]m of tin is consumed during a 4 hour anneal at 155[degrees]C.

Regarding the IMC growth at room temperature, systematic investigations were hampered by the slow kinetics of the process. Thus, a slightly different approach was undertaken to determine the room temperature kinetics.

[FIGURE 2 OMITTED]

According to Jordan (1), the rate constant of IMC growth at room temperature is between 0.16 nm/[square root of (s)] and 0.2 nm/[square root of (s)]. Since the average density of the IMC layer is close to the density of tin and the tin content of the IMC is close to 50%, the rate of tin loss is assumed to be around 50% of the rate of IMC growth. Thus, 0.08[+ or -]0.02 nm/[square root of (s)], recalculated to 4.8[+ or -]1.2 nm/[square root of (h)], is taken as the textbook value for the rate of tin consumption.

[FIGURE 3 OMITTED]

Experimental values were determined with stripping coulometry after extended room temperature storage up to three months. The difference of layer thickness before and after storage is plotted against storage time in Figure 4, along with the calculation on the base of the values given by Jordan. The experimental data agree reasonably with the calculation, indicating that the standard textbook values apply well for the immersion tin layers in the experiment. Extrapolation of the data yields a tin consumption of at most 0.5 [micro]m within one year of room temperature storage.

To summarize the kinetics of IMC formation, the determined values are plotted in Figure 5, along with a selection of data from literature. Note that published values mostly relate to the growth rate of the IMC. These values were recalculated for Figure 5 by assuming that the tin loss is 50% of the IMC growth.

As already noted by Jordan (1) and visualized in Figure 5, data for the rate constant of tin loss (or IMC growth) scatter significantly. We claim, however, that the data do not scatter statistically but, rather, systematically. The data of the present investigation (red circle/solid) and the data of Posdorfer (4) (blue downward triangle), which are both based on stripping coulometry of immersion tin layers, are in excellent agreement and relate to the initial stages of IMC formation. The other data (1), (5), (6) are determined from estimates of the IMC layer thickness in cross sections and generally relate to IMC thickness of several microns, hence to later stages of IMC formation.

Assuming different rate constants for the initial and the later stages of IMC formation seems reasonable. Thus, the data in Figure 5 are concluded to reflect information on two different growth processes.

Since IMC growth results from solid state diffusion, an exponential temperature dependence may be expected for the rate constant of the process, which should give a straight line in the semilogarithmic plot of Figure 5. In fact, this result agrees with the experimental data in the temperature regime above 60[degrees]C. At lower temperatures, however, the rate constant decreases steeper with decreasing temperature, indicating a change in the diffusion mechanism.

An extrapolation of the kinetic data in the regime above 60[degrees]C toward lower temperatures leads to a dramatic overestimation of IMC growth rate at room temperature. Overestimated room temperature kinetics easily invoke unreasonably harsh requirements for thermal aging tests.

Solderability of Aged Immersion Tin Layers

Aging tests were performed to simulate prolonged storage of process boards. A common variant of these tests consists of a 4 hour anneal at 155[degrees]C, intended to simulate one year of storage at room temperature.

As a result of this test, a layer of 0.72[+ or -]0.06 [micro]m of tin was converted to intermetallic compounds. Consequently, a minimum layer thickness of 0.8 [micro]m of tin is required for this test, if complete conversion of the layer is to be avoided. One year room temperature storage, on the other hand, converts at most 0.5 [micro]m of tin to IMC. Thus, the 4 hours at 155[degrees]C test is concluded to be considerably too harsh for simulating one year of room temperature storage.

[FIGURE 4 OMITTED]

As an appropriate alternative, we propose a three-fold reflow process as a standard aging test. To evaluate the solderability in dependence of layer thickness and aging process, a series of test boards was prepared with varying layer thickness. The layer thickness was determined with XRF.

[FIGURE 5 OMITTED]

A series of samples was annealed for 4 hours at 155[degrees]C; another series was reflowed three times. The residual tin layer thickness after thermal treatment was measured with stripping coulometry. The wetting properties of all samples were characterized with the solder balance method.

The results of the measurements show that, with aging 4 hours at 155[degrees]C, the tin thickness decreases by about 0.65 [micro]m and, with 3x reflow, by about 0.5 [micro]m. Thus, the 3x reflow process is convenient not only to simulate a multiple soldering process but also to simulate one year room temperature storage.

Regarding the solderability, excellent wetting was observed for all freshly prepared samples. For the samples after 4 hours annealing at 155[degrees]C, good wetting was observed for samples with tin thickness [d.sub.Sn][greater than or equal to]0.95 [micro]m and for the 3x reflowed samples with tin thickness [d.sub.Sn][greater than or equal to]0.8 [micro]m.

From these results, a minimum layer thickness of pure tin of approximately 0.3 [micro]m was concluded to be required on thermally treated samples to ensure good wetting properties. Consequently, around 1 [micro]m of tin is required to pass the 4 hour 155[degrees]C test, and around 0.8 [micro]m is required to retain solderability after 3x reflow or one year room temperature storage.

Conclusions

The results show that the simple kinetic model leads to an overestimation of the kinetics in the temperature regime below 60[degrees]C. Changes in the reaction mechanism with temperature as well as with time must be taken into account for a conclusive understanding of the aging characteristics. From the achieved result and the known kinetics of IMC formation, 1 [micro]m of immersion tin is required to guarantee solderability after aging 4 hours at 155[degrees]C. If the requirement is reduced to one year of room temperature storage and multiple solderability, then a layer thickness of 0.8 [micro]m is sufficient.

References

1 M. Jordan, Die galvanische Abscheidung von Zinn und Zinnlegierungen (Eugen G. Leuze Verlag, Saulgau, Germany, 1993).

2 I. K. Hui and H. W. Law, Soldering and Surface Mount Technology 12, 23 (2000).

3 K. N. Tu and R. D. Thompson, Acta Metall 30, 947 (1982).

4 J. Posdorfer, in Electronic Forum, Backnang, 1999.

5 Y.C. Chan, A. C. K. So, and J. K. L. Lai, Materials Science and Engineering B 55, 5 (1998).

6 R. J. KleinWassink, Weichloten in der Elektronik (Eugen G. Leuze Verlag, Saulgau, Germany, 1991).

Dr.-Ing. Dieter Metzger and Dr. Roland Vogel

Dr.-Ing. Dieter Metzger is responsible for qualification and soldering of surface finishes, email: [e-mail]; and Dr. Roland Vogel is responsible for research and development of corrosion resistant coating, email: [e-mail]--both with Atotech Deutschland GmbH, Berlin, Germany.