Local Time–Temperature-dependent Deformation of a Woven Composite by P. Shrotriya and N.R. Sottos

Moiré interferometry is utilized to investigate the time–temperature-dependent deformation of a woven composite substrate used in multilayer circuit board applications. Creep tests are performed at temperatures ranging from 27 to 70◦C, and the resulting longitudinal and transverse displacement fields are measured via moiré interferometry. Measured displacement fields reveal the influence of fabric architecture on woven composite response. The deformation fields in the plane of the composite for loading along both warp and fill directions consist of a periodic arrangement of high-strain and low-strain regions in accordance to the interlacing bundle architecture. The deformation fields over the cross-section of the composite indicate that neighboring unit cells are subjected to equal and opposite bending moment even when the composite is loaded in uniaxial tension. KEY WORDS—Woven composite, time–temperaturedependent response, Moiré interferometry, creep and stress relaxation, multilayer circuit boards, textile composites Introduction Multilayer printed circuit boards (PCBs) are used extensively in electronic packaging assemblies. The boards, shown schematically in Fig. 1, consist of multiple layers of woven glass/epoxy composite substrate sandwiched between copper foils. Significant residual stresses develop during processing and post-processing of multilayer boards due to mismatch of properties between the copper and woven composite layers. The residual stresses are large enough to cause such undesirable dimensional changes as warping and bowing of the boards and shrinkage of one layer with respect to the other. Board warpage and inner layer shrinkage cause problems with chip insertion, solder connection and interconnection between the layers and significantly affect the package reliability. Increases in circuit density make these problems even more severe as tolerance to dimensional changes is further reduced and the number of boards that pass quality standards decreases. The ability to accurately predict the residual stress state and to design dimensionally stable boards depends on accurate determination of the thermomechanical properties of the woven composite substrate. P. Shrotriya ([email protected]) and N.R. Sottos (SEM member), Department of Theoretical and Applied Mechanics, University of Illinois at UrbanaChampaign, Urbana, Illinois, USA. P. Shrotriya is currently at Mechanical Engineering Department, Iowa State University, 2025 Black Engineering Building, Ames, Iowa, USA. Original manuscript submitted: February 11, 2002. Final manuscript received: January 13, 2004. DOI: 10.1177/0014485104044314 The composite substrate consists of an epoxy matrix reinforced by a plain weave fabric of glass fibers. A schematic diagram of the plain weave fabric is shown in Fig. 2. The fabric is composed of two sets of interlaced orthogonal warp and fill fiber bundles. A large number of plain weave fabric styles are currently used in circuit board design. The fabric styles are often unbalanced, i.e., the warp and fill directions include different numbers of fiber bundles or different sized fiber diameters. Because of the variation in fiber bundle sizes, the geometry of the undulating fiber bundles is different and depends on the fabric styles. Hence, composite substrates with different fabric styles have very different properties. Furthermore, the same fabric style has different properties in the warp and fill directions. Sottos et al.1 measured significant differences in the fabric geometry (bundle size, crimp, etc.), the elastic modulus, and coefficient of thermal expansion (CTE) in the warp and fill directions of two common substrates for multilayer circuit boards. Wu et al.2 and Yuan and Falanga3 characterized the CTE of substrates below the matrix glass transition temperature and detected higher CTE values in the fill direction. Wang et al.4 generated master relaxation curves for a 109 style composite substrate and reported on different relaxation responses in the warp and fill directions. Shrotriya and Sottos5 conducted experiments to characterize the fabric architecture and obtained different creep compliance curves in the warp and fill directions of a 7628 style composite substrate. During the processing and post-processing, the multilayer circuit boards are heated above the matrix glass transition temperature for an extended period of time. At temperatures near the glass transition, viscoelastic processes dominate the matrix response and result in time–temperature dependence of substrate properties. Shrotriya and Sottos5 and Wang et al.4 determined the global creep and relaxation response of woven substrates but there are no reported investigations on the influence of time–temperature-dependent matrix response on the local displacement field of a composite unit cell. All of the studies described above suggest a strong influence of fabric architecture on the woven composite properties but the experimental measurements were performed at a length scale much larger than the composite unit cell. The woven composite was approximated as a homogeneous anisotropic continuum and the inhomogeneity of the unit cell (shown in Fig. 2) was ignored. The influence of the fabric architecture on the composite response is not studied directly. A detailed study of the local stress and displacement fields is required to fully understand the relation between the fabric architecture and woven composite response. A myriad of 336 • Vol. 44, No. 4, August 2004 © 2004 Society for Experimental Mechanics