1
A PREDICTIVE MODEL FOR WHISKER FORMATION BASED ON LOCAL MICROSTRUCTURE AND GRAIN BOUNDARY PROPERTIES

Pylin Sarobol

Sandia National Laboratories, Albuquerque, New Mexico, USA

Ying Wang

School of Materials Engineering, Purdue University, West Lafayette, Indiana, USA

Wei-Hsun Chen

School of Materials Engineering, Purdue University, West Lafayette, Indiana, USA; Cymer, Inc., San Diego, California, USA

Cymer, Inc., San Diego, California, USA

Aaron E. Pedigo

Naval Surface Warfare Center, Crane Division, Crane, Indiana, USA

John P. Koppes

Alcoa-Howmet, Whitehall, Michigan, USA

John E. Blendell and Carol A. Handwerker

School of Materials Engineering, Purdue University, West Lafayette, Indiana, USA

1.1 INTRODUCTION

For over 60?years, whiskers and hillocks have been recognized as local manifestations of stress relaxation in thin films: these "surface defects" grow spontaneously by mass transport to specific grain boundaries in the plane of the film, with whiskers becoming hillocks when grain boundary migration accompanies growth out of the plane of the film. Whisker formation is important for two reasons.

First, the risk to electronic system reliability is serious for whisker formation in Sn films: whiskers can sometimes grow to be millimeters long, causing short-circuiting between adjacent components. Second, isolating the mechanisms leading to whisker and hillock formation in one grain out of 103-105 film grains will inform us more broadly about the multiplicity of possible responses of thin films to stresses.

As a local phenomenon, whisker formation has been attributed in the past to specific inhomogeneity in either the film or the interfacial intermetallic microstructure, in stress, or in the anisotropic properties in the film. These inhomogeneities include

• grains above large Cu6Sn5 intermetallic particles [1, 2];
• grains covered by relatively weak oxide films [3];
• grains with larger out-of-plane elastic deformation than surrounding grains due to elastic anisotropy under plane strain conditions [4];
• a difference in the orientation-dependent yielding behavior of whisker grains and their neighboring grains [5, 6];
• shallow grains whose grain boundaries serve as sinks for atoms when surface diffusion necessary for diffusional creep is not active [7].

When the first three of these concepts were tested experimentally, these simple relationships were not observed. For example, Pei et al. [8] determined from electron backscatter diffraction (EBSD) and etching studies that whiskers were not associated with large underlying intermetallic particles. Moon et al. [9], Jadhav et al. [10], and A.E. Pedigo (unpublished research) determined through ultra-high vacuum (UHV) studies of sputtered films and through oxide removal in specific regions by focused ion beam (FIB) milling that whisker growth was unchanged when the native oxide layer was removed through sputtering and was not accelerated in regions without oxide.

From these and other experiments on the crystallography, texture, geometry, and strain in whisker and hillock grains relative to their surrounding microstructures, it is clear that models ascribing whisker formation and growth to a single cause are too simplistic: they do not capture all relevant aspects of why that particular grain was predisposed to become a whisker out of 103-105 grains in the film nor do they provide quantitative predictions of growth rates under a wide range of conditions.

Whisker nucleation and growth must be examined at the local level, taking into consideration the microstructure and properties of the film, specifically grain boundary structure, as well as crystallography, grain geometry, and the role of oxide films.

In this chapter, we review the characteristic properties of the local microstructure around whisker and hillock grains with the aim to further identify why these particular grains and locations became predisposed to forming whiskers and hillocks.

On the local level, we suggest that three factors play major roles in determining which surface grains will become whiskers or hillocks: the surface grain geometry, crystallographic orientation-dependent grain boundary properties, and the microstructure-dependent local elastic strain/strain energy density.

These effects have been identified through our recent model of whisker growth by grain boundary sliding limited creep (with and without coupling between grain boundary migration and shear stress), finite element analysis of a wide range of textured microstructures in Sn films, and recent experimental studies of microstructural evolution during intermetallic compound (IMC)-induced and thermal-cycling-induced stresses [11].

In addition, we have identified specific film microstructures and textures that are predicted to produce whisker-prone environments under specific stress conditions. By simulating the relationship between film texture (axis and strength) and the localization of the strain energy in the microstructure, a necessary condition for whisker formation is proposed: grains with locally high strain energy will become whiskers. Thus, measuring global film properties (crystallographic texture and strain energy density) may be a useful first step in identifying whisker-prone films, but is not sufficient in identifying which grain will form a surface defect.

Whisker growth will depend not only on local elastic strain energy density (ESED) but also grain geometry and the response of the grain boundaries surrounding the whisker grain to accretion-induced shear stresses.

1.2 CHARACTERISTICS OF WHISKER AND HILLOCK GROWTH FROM SURFACE GRAINS

Whiskers and hillocks typically grow from surface grains (Fig. 1.1) in fine-grained Sn and Sn-alloyed films in both isothermal (at room temperature or high temperature) and thermal cycling conditions. The origin of surface grains as nuclei for whiskers and hillocks have been much debated - whether surface grains were pre-existing grains from the film deposition process [7, 13, 14], new grains that formed via surface recrystallization after the deposition process [7, 15, 16], or pre-existing grains whose topology changes as a result of grain growth in the film.

Figure 1.1 Whisker growing from surface grain. Image in (a) from secondary electron microscopy (SEM) shows a short whisker growing at an angle to the film top surface, while image in (b) from focused ion beam (FIB) cross section shows shallow grain boundaries (70° with respect to the film normal) at the whisker root.

Source: Reprinted with permission from Elsevier © 2013 [12].

There is now direct evidence that whiskers and hillocks may form from either pre-existing or newly nucleated grains. Pei et al. [14] and Wang et al. [17] demonstrated that pre-existing grains could serve as nuclei for whiskers and hillocks for isothermal annealing (room temperature) and thermal cycling conditions, respectively.

Cross sections of whiskers formed under these conditions showed that the grain boundaries at the base of the whisker grain, that is, the "whisker root," are generally angled with respect to the film normal, creating shallow grains relative to the film thickness ("surface" grains) and that whiskers grew from grains with in-plane grain diameters typical of nonwhiskering grains in the film. In addition, surface grains that did not become whiskers or hillocks were also observed [12].

Shallow surface grains were observed by Sarobol et al. to form by recrystallization and grow into micron-sized whiskers and hillocks during thermal cycling of large-grained Sn-alloy films on Cu substrates, as seen in Figure 1.2 [18, 19]. New strain-free grains (low dislocation density) nucleated along pre-existing film grain boundaries and grew at the expense of highly deformed (high dislocation density) parent grains.

Figure 1.2 (a) Surface defect formed by recrystallization at a grain boundary, (b) with FIB cross section along the white dashed line in (a) showing shallow single-crystal grain with oblique grain boundaries, (c) representative hillock showing both surface uplift and grain boundary migration, and (d) representative whisker in thermally cycled solder film.

Source: Reprinted with permission from Elsevier © 2013 [18].

Hillocks are formed when growth out of the plane of the film is accompanied by grain boundary migration into the film, as seen in Figure 1.2c. Asymmetrical growth at the whisker root can lead to curved whiskers and hillocks, as shown in Figure 1.3 with a change in orientation of the whisker as it grows [8]. It is well known that the growth morphology for a given whisker and hillock may change over time, and since surface diffusion is suppressed by the presence of the surface oxide, the morphological changes over time are preserved in its surface shape [20].

Figure 1.3 SEM images (a) and (b) from two different orientations of a curved whisker, which is decreasing in cross-sectional area as it grows. FIB images (c) and (d) show two successive cross sections of the whisker as oriented in (b), with the initial positions of the grain boundaries forming the whisker root indicated (black arrows).

Source: Pei et al. [8].

1.2.1...