A new analytical model is developed for thermomechanical responses of multi-layered structures with an arbitrary number of layers and subjected to general thermal and mechanical loading. The formulation is based on an extended Bernoulli–Euler beam theory and a slip-interface model. The former includes Poisson’s effect and covers both the plane stress and plane strain deformations, and the latter allows slipping between two adjacent layers but no jump in the normal displacement or traction. An analytical solution for a multi-layered structure under general thermomechanical loading is derived by using a new approach that first determines one interfacial shear stress and the curvature of the deformed structure. To illustrate the newly developed model, three example problems for two-, three- and five-layer structures respectively are analytically solved by directly applying the new model. In all three cases, the solutions are obtained in closed-form expressions by considering both temperature changes and mechanical loads including body forces, distributed normal and shear stresses on the top and bottom surfaces, and normal forces, transverse shear forces and bending moments at the two ends, unlike existing ones. It is shown that the current solution for two-layer structures recovers an existing solution without considering Poisson’s effect and mechanical loading and the classical solution of Timoshenko for perfectly bonded bi-metal thermostats as two special cases. The closed-form solution for five-layer structures with imperfect interfaces is derived here for the first time. In addition, numerical results are provided for five- and seven-layer transistor stacks to quantitatively demonstrate the new model. It is found that the current results for the five-layer transistor stack agree well with those obtained by others, thereby further validating the new model.
Multi-layered structures have been widely used in electronics, aerospace, construction, manufacturing, and power generation industries, which are often subjected to thermomechanical loads. Various models have been developed to describe responses of such structures, including laminate theories (e.g., [29, 32]), layer-wise analyses (e.g., [1, 4, 24]), and shear-lag models (e.g., [8, 14, 26, 40]). These models are typically based on the assumption of perfectly-bonded or imperfect interfaces between adjacent layers. For the former, both the traction and displacement are taken to be continuous across each interface (e.g., [48]), while for the latter the continuity conditions are imposed only for the traction and the displacement field can be discontinuous across the same interface (e.g., [9]). Imperfect interface-based models are physically more representative, but they suffer from the following major drawbacks: (1) boundary value problems (BVPs) formulated using these models depend on the local equilibrium for each layer, global equilibrium of the entire multilayer structure, and continuity (compatibility) conditions at interfaces; (2) such a BVP is governed by a reduced system of coupled differential equations, whose orders increase with the number of layers. As a result, analytical solutions based on these imperfect interface models have been limited to structures with only two or three layers (e.g., [5, 6, 13, 15,16,17,18, 20, 22, 23, 25, 34, 35, 38, 42, 43, 46, 47]).
In the current study, a new analytical model is developed for thermomechanical responses of multi-layered structures with an arbitrary number of layers by using an extended Bernoulli–Euler beam theory and a slip-interface model. The former accounts for Poisson’s effect and covers both the plane stress and plane strain deformation modes, and the latter allows slipping between two adjacent layers (i.e., a tangential displacement jump) but no jump in the normal displacement or traction.
The rest of this paper proceeds as follows. In Sect. 2, the new analytical model is formulated, and a general solution is derived for a multi-layered structure, which is enabled by using a new approach proposed here that first determines one interfacial shear stress along with the curvature of the deformed structure. To illustrate the newly developed model, three example problems for two-, three- and five-layer structures respectively are analytically solved in Sect. 3 by directly applying the new model. In all three cases, the solutions are obtained in closed-form expressions. The current solution for two-layer structures recovers those of Liu and Chen [22] for bi-layer beams with an imperfect interface and Timoshenko [39] for perfectly bonded bi-metal thermostats as two special cases. The closed-form solution for five-layer structures with imperfect interfaces is obtained here for the first time. In this section, numerical results are also provided for five- and seven-layer transistor stacks and compared with the data from Pao and Eisele [30] for a five-layer transistor stack to verify and demonstrate the new model. A summary is presented in Sect. 4.
2 New model for a multi-layered structure with interfacial slipping
Consider a multi-layered structure of a uniform rectangular cross-section, which consists of k homogeneous, isotropic, linear elastic thin layers of uniform thickness, as shown in Fig. 1. The k bonded layers are made from dissimilar materials, and the k − 1 interfaces are imperfect (weak), allowing adjacent layers to slip over each other while staying in contact all the time. Each layer satisfies an extended Bernoulli–Euler beam theory that incorporates Poisson’s effect and covers both the plane stress and plane strain deformation modes. This type of structure can well represent electronic packaging assemblies (e.g., [30, 38, 41]) and composite beam-columns with partial interaction (e.g., [18, 21, 27]).
Fig. 1
a Multi-layered structure with a uniform rectangular cross-section subjected to thermomechanical loads; b deformation of a differential element (dx) of the mth layer
The multi-layered structure undergoes axial and flexural deformations due to the mechanical loads applied at its boundaries and a temperature change ΔT = T – T0, where T0 is a reference temperature. In Fig. 1, a global coordinate system {x, y, z}, with its origin placed at the center of the mid-plane of the first layer from the bottom, is adopted to describe the entire structure, while a local coordinate system {x, y, zm} with its origin located at the center of the mid-plane of the layer is used for each layer, where m denotes the mth layer. The choice of this global coordinate system is not unique, since the origin for the global coordinate system can be placed at any point on the symmetry axis z of the cross section (e.g., [22, 23, 44]) without affecting the determination of the curvature of the deformed structure from simultaneously enforcing the equilibrium equations for forces in the x- and z-directions and moments in the y-direction. Note that the z- or zm-axis is also the symmetry axis in the width direction. In addition, zm = z – dm, where dm is the distance between the centroidal axis of the mth layer and the centroidal axis of the first layer, as shown in Fig. 1.
2.1 Kinematic relations
In the current study, each layer is regarded as a Bernoulli–Euler beam satisfying the following kinematic relations (see Fig. 1b):
where \({u}_{x}^{\left(m\right)}\), \({u}_{y}^{\left(m\right)}\) and \({u}_{z}^{\left(m\right)}\) are, respectively, the \(x\)-, y- and \(z\)-components of the displacement vector \(\mathbf{u}\) of a point (x, 0, zm) on the vertical symmetry plane of the mth layer, and \({u}_{0}^{\left(m\right)}\) and \({w}^{\left(m\right)}\) are, respectively, the axial and vertical displacement components of the corresponding point (x, 0, 0) on the centroidal axis of the mth layer, the superscript/subscript m indicates the mth layer, and \({( )}_{,x}\equiv d( )/dx\).
From Eq. (1), the strain components in each layer can be readily obtained as
$$ \varepsilon_{xx}^{\left( m \right)} = \varepsilon_{0}^{\left( m \right)} - z_{m} \varphi^{\left( m \right)} ,\,{\text{all }}\,{\text{other}}\quad \varepsilon_{ij}^{\left( m \right)} = 0, $$
(2a)
where
$$ \varepsilon_{0}^{\left( m \right)} = u_{0,x}^{\left( m \right)} ,\,\varphi^{\left( m \right)} = w_{,xx}^{\left( m \right)}$$
(2b)
are the axial strain on the centroidal axis and the curvature of the mth layer, respectively.
2.2 Slip-interface model
Each of the k − 1 interfaces is taken to be imperfect, which allows slipping between the two adjacent layers (i.e., a tangential displacement jump in the x-direction) but no jump in the normal (z-direction) displacement or traction. This type of slip-interface model has been widely used in designing composite beam-columns in civil engineering (e.g., [18, 21, 22, 27]). It can be viewed as a special case of the spring-layer model for imperfect interfaces known in the mechanics community (e.g., [2, 7, 9, 19, 40]).
According to the slip-interface model (e.g., [9, 18, 23]),
where \(\tau_{xz}^{(m)}\) is the shear stress acting on the mth interface (between layers m and m + 1) (see Fig. 2a), tm and tm+1 are, respectively, the thickness of the mth and (m + 1)th layers, and Km is the stiffness constant of the mth interface (with a unit of Pa/m). Clearly, Eq. (3) indicates that Km → ∞ when the two layers are perfectly bonded (and hence the interface is perfect), and Km = 0 when the two layers are completely separated. Using Eq. (1) in Eq. (3) gives
Note that the slip-interface model with \(\Delta {u}_{x}^{\left(m\right)}\ne 0\) (while \(\Delta {u}_{z}^{\left(m\right)}=0\)) adopted here differs from those implemented for perfectly-bonded interfaces, where \(\Delta {u}_{x}^{\left(m\right)}=\Delta {u}_{z}^{\left(m\right)}=0\) (e.g., [4, 25]), or spring-layer imperfect interfaces, where the interfacial traction vector is related to the jump in the displacement vector \(\Delta {\mathbf{u}}^{\left(m\right)}\) through the elastic stiffness of the interface (treated as a spring layer) (e.g., [7, 19, 40]). Based on this slip-interface model,
where w and φ are, respectively, the deflection and curvature of the structure at the x cross- section, which are independent of zm or z. This z-independence results from the kinematic relations in the Bernoulli–Euler beam theory (see Eq. (1)) and the continuity condition of no jump in the vertical displacement across each interface, which is embodied in the slip-interface model.
2.3 Constitutive equations
The material of each layer is assumed to be isotropic. Then, applying Hooke’s law yields the axial normal stress component in each layer as, upon using Eq. (2a),
where ΔT is the temperature change that is the same for all layers and uniform along the x-direction, and \(c_{11}^{(m)}\) and \(d_{11}^{(m)}\) are, respectively, the elastic stiffness constants and thermal expansion coefficients given by (e.g., [3, 35])
for plane strain deformations. In Eqs. (7a) and (7b), \(E^{(m)} ,\nu^{(m)} \,\,\text{and}\,\,\alpha^{(m)}\) are, respectively, Young’s modulus, Poisson’s ratio and the coefficient of thermal expansion of the material of the mth layer. Note that Eq. (7b) can be either directly derived from the plane strain conditions or simply obtained from Eq. (7a) using the well-known transformation relations for \(E^{(m)} ,\nu^{(m)} \,\,\text{and}\,\,\alpha^{(m)}\) between the plane stress and plane strain cases (e.g., [33]).
When Poisson’s effect is suppressed by setting ν(m) = 0, Eqs. (7a) and (7b) both reduce to
Equation (7c) has been used in existing studies based on the Bernoulli–Euler beam theory (e.g., [8, 22, 23]), which did not consider the Poisson effect.
From Eqs. (2a) and (6), the resultant normal force \(N_{x}^{(m)}\) and bending moment \(M_{x}^{(m)}\) in the mth (with m ∈ {1, 2, …, k}) layer can be obtained as
with b being the width of the beam structure, which is uniform along the beam length in the x-direction.
2.4 Equilibrium analysis
Consider the equilibrium of a differential element of the multi-layered structure, whose length is dx. Figure 2a shows the free body diagram (FBD) of the mth layer in the differential element, with the normal and shear stresses on the two interfaces indicated. The FBD for the differential element dx of the structure without separating the layers is displayed in Fig. 2b.
From the FBD in Fig. 2a, the equilibrium equations of the mth layer can be readily obtained as, upon enforcing the force balances in the x- and zm-directions and the moment balance in the y-direction,
for any x ∈ (− L/2, L/2), where \(N_{x}^{(m)} ,\,\,V_{x}^{(m)} \,\text{and}\,M_{x}^{(m)}\) are, respectively, the resultant normal force, transverse shear force and bending moment on the x cross-section of the mth layer, \(q_{x}^{(m)} \,\text{and}\,\,q_{z}^{(m)}\) are, respectively, the x- and zm-components of the body force acting in the mth layer (with a unit of N/m2), \(\sigma_{zz}^{(m)} \,\text{and}\,\,\tau_{xz}^{(m)}\) are, respectively, the normal and shear stress components on the mth interface, and \(\sigma_{zz}^{(m - 1)} \,\text{and}\,\,\tau_{xz}^{(m - 1)}\) are, respectively, the normal and shear stress components on the (m – 1)th interface. From Eqs. (8a) and (10a), it is clearly seen that the axial force \(N_{x}^{(m)}\) and normal stress \(\sigma_{xx}^{(m)}\) are directly linked to the interfacial bonding constraints through the interfacial shear stresses \(\tau_{xz}^{(m)} \,\text{and}\,\,\tau_{xz}^{(m - 1)} .\) Note that Eqs. (10b) and (10c) can be combined to yield
for any x ∈ (− L/2, L/2). Hence, Eqs. (10a) and (11) will be used as the two equilibrium equations for the mth layer in terms of the resultants \(N_{x}^{(m)} \,\text{and}\,M_{x}^{(m)}\).
The boundary conditions (BCs) for the mth layer (associated with Eqs. (10a–c)) are given by
where the overhead bar represents the prescribed value and use has been made of Eq. (5). These boundary conditions are required for the determination of \(N_{x}^{(m)} \,\text{and}\,M_{x}^{(m)}\) from the equilibrium equations listed in Eqs. (10a) and (11), which is demonstrated in Sect. 3.
Similarly, from the FBD in Fig. 2b, the equilibrium equations for the entire multi-layered structure can be obtained as
for any x ∈ (− L/2, L/2), where \(\hat{N}_{x} ,\,\,\hat{V}_{x} \,\,\text{and}\,\,\hat{M}_{x}\) are, respectively, the total resultant normal force, shear force and bending moment on the x cross-section of the multi-layered structure, \(\hat{q}_{x} \,\,\text{and}\,\,\hat{q}_{z}\) are, respectively, the x-and z-components of the total body force (in N/m2) in the multi-layered structure with
and \(\sigma_{zz}^{(k)} ,\,\,\tau_{xz}^{(k)}\) and \(\sigma_{zz}^{(0)} ,\,\,\tau_{xz}^{(0)}\) are, respectively, the normal and shear stresses applied on the top and bottom surfaces of the structure. In reaching Eq. (13c), the moment is taken about the y-centroidal axis of the first layer on the (x + dx) cross-section.
Note that Eqs. (13b) and (13c) can be combined to give
which are obtained from the equivalence of the normal force, shear force and bending moment on the x cross-section. Note that the values of dm in Eqs. (13c), (15) and (16c) (see Fig. 1a) are given by
In addition, the moment balance leading to Eq. (16c) is about the y-centroidal axis of the first layer on the x cross-section.
From Eqs. (10a,b), (11), (14a,b) and (16a–c), it can be readily shown that satisfying the local equilibrium equations in Eqs. (10a,b) and (11) ensures that the global equilibrium equations listed in Eqs. (13a,b) and (15) are automatically satisfied for the current multi-layered structure under the general thermal and mechanical loads prescribed.
The boundary conditions for the entire structure (associated with Eqs. (13a–c)) read
The governing equations and boundary conditions in Eqs. (10a), (11) and (12a–c) define a boundary-value problem (BVP). A new analytical approach is proposed herein to solve this BVP for the multi-layered structure subjected to general thermomechanical loads.
Solving the current BVP analytically is challenging. Specifically, if the constitutive relations in Eqs. (4) and (8a,b) are substituted into the equilibrium equations in Eqs. (10a) and (11), then a set of coupled second-order ordinary differential equations (ODEs) will be obtained in terms of \({u}_{0}^{\left(m\right)}\) and \(\varphi\), which are difficult to solve analytically for a structure with a large number of layers. However, this BVP can be solved using a new approach by first determining \({\tau }_{xz,x}^{\left(1\right)}\) and \(\varphi_{,xx}\), as shown below.
Note that use has also been made of Eq. (9) in reaching the final expressions in Eq. (25).
Equation (24) represents a system of (k – 1) algebraic equations that can be solved for \({\tau }_{xz,x}^{\left(m+1\right)}\) (or \({\tau }_{xz,x}^{\left(m-1\right)}\)) in terms of φ,xx, \({\tau }_{xz,x}^{\left(m\right)}\), \({\tau }_{xz,xxx}^{\left(m\right)}\) and \({\tau }_{xz,x}^{\left(m-1\right)}\) (or \({\tau }_{xz,x}^{\left(m+1\right)}\)). Specifically, Eq. (24) can be rearranged as follows:
From Eq. (26), \({\tau }_{xz,x}^{\left(2\right)}\), \({\tau }_{xz,x}^{\left(3\right)}\), …, \({\tau }_{xz,x}^{\left(k-1\right)}\) can be sequentially obtained in terms of \({\tau }_{xz,x}^{\left(1\right)}\) and φ,xx from the first equation to the (k – 2)th (the second last) equation as
Substituting Eq. (27) into the (k – 1)th (the last) equation in Eq. (26) leads to the following \(2\left(k-1\right)\)th order ODE involving \({\tau }_{xz,x}^{\left(1\right)}\) and φ,xx:
where the first five constant coefficients in each of the Ai and Bi groups are listed in Eqs. (A1) and (A2) in "Appendix", and the inhomogeneous term H1 is given by
$$ \begin{gathered} H_{1} = - \gamma \sum\limits_{{s = 1}}^{k} \left[{\kern-2.5pt}\left[\left( {\prod\limits_{{p = 1}}^{{s - 1}}{\beta _{1}^{{\left( p \right)}} } } \right)\left\{{\frac{{d^{{2\left( {k - s - 1} \right)}} \eta _{x}^{{\left( s \right)}} }}{{dx^{{2\left( {k - s - 1} \right)}} }} - \left({\sum\limits_{{m = s + 1}}^{{k - 1}} {\beta _{0}^{{\left( m \right)}} } } \right)\frac{{d^{{2\left( {k - s - 2} \right)}} \eta _{x}^{{\left( s \right)}} }}{{dx^{{2\left( {k - s - 2} \right)}} }}+ \left[ {\sum\limits_{{n = s + 2}}^{{k - 1}} {\left( {\beta _{0}^{{\left( n \right)}} \mathop \sum \limits_{{m = s + 1}}^{{n -1}} \beta _{0}^{{\left( m \right)}} } \right)} - \sum\limits_{{m =s + 1}}^{{k - 2}} {\left( {\beta _{{ - 1}}^{{\left( {m + 1}\right)}} \beta _{1}^{{\left( m \right)}} } \right)} }\right]\frac{{d^{{2\left( {k - s - 3} \right)}} \eta _{x}^{{\left( s \right)}} }}{{dx^{{2\left( {k - s - 3} \right)}} }}} \right.\right.\right.\hfill \\ \qquad\quad \left.\left.\left. { - \left\langle {\sum\limits_{{r = s + 3}}^{{k - 1}} {\left[ {\beta _{0}^{{\left( r \right)}}\sum\limits_{{n = s + 2}}^{{r - 1}} {\left( {\beta _{0}^{{\left( n \right)}} \mathop \sum \limits_{{m = s + 1}}^{{n - 1}} \beta _{0}^{{\left( m \right)}} } \right)} } \right]} - \sum\limits_{{m =s + 1}}^{{k - 2}} {\left[ {\beta _{{ - 1}}^{{\left( {m + 1}\right)}} \beta _{1}^{{\left( m \right)}} \left( {\sum\limits_{{n =s + 1}}^{{k - 1}} {\beta _{0}^{{\left( n \right)}} - \beta _{0}^{{\left( m \right)}} - \beta _{0}^{{\left( {m + 1} \right)}} }} \right)} \right]} } \right\rangle \frac{{d^{{2\left( {k - s - 4}\right)}} \eta _{x}^{{\left( s \right)}} }}{{dx^{{2\left( {k - s -4} \right)}} }}} \right\} \right]{\kern-2.5pt}\right]\hfill \\ \qquad\quad \left. { -\left\langle {\sum\limits_{{r = 4}}^{{k - 1}} {\left[ {\beta _{0}^{{\left( r \right)}} \mathop \sum \limits_{{n = 3}}^{{r - 1}}\left( {\beta _{0}^{{\left( n \right)}} \mathop \sum \limits_{{m =2}}^{{n - 1}} \beta _{0}^{{\left( m \right)}} } \right)} \right]} -\sum\limits_{{m = 2}}^{{k - 2}} {\left[ {\beta _{{ - 1}}^{{\left( {m + 1} \right)}} \beta _{1}^{{\left( m \right)}} \left({\sum\limits_{{n = 2}}^{{k - 1}} {\beta _{0}^{{\left( n \right)}} -\beta _{0}^{{\left( m \right)}} - \beta _{0}^{{\left( {m + 1}\right)}} } } \right)} \right]} } \right\rangle \frac{{d^{{2\left({k - 5} \right)}} \tau _{{xz,x}}^{{\left( 0 \right)}}}}{{dx^{{2\left( {k - 5} \right)}} }}} \right\} - \gamma \left({\prod\limits_{{p = 1}}^{{k - 1}} {\beta _{1}^{{\left( p \right)}} }} \right)\tau _{{xz,x}}^{{\left( k \right)}} , \hfill \\\end{gathered} $$
(29a)
with
$$\gamma=\frac1{\prod\limits_{{m = 1}}^{{k - 2}} {\beta _{1}^{{\left( m \right)}} }}.$$
(29b)
Note that Ai and Bi in Eq. (28) depend on the number of layers and the geometrical and material properties of the layers, and H1 depends on the mechanical loads \(q_{x}^{(m)}\), \({\tau }_{xz}^{\left(0\right)}\) and \({\tau }_{xz}^{\left(k\right)}\) additionally, as shown in Eqs. (25), (29a), (29b), (A1) and (A2).
Next, using Eq. (27) in Eq. (23) results in the following \(2\left(k-2\right)\)th order ODE involving \({\tau }_{xz,x}^{\left(1\right)}\) and φ,xx:
where the first four constant coefficients in each of the Ci and Di groups are listed in Eqs. (A3) and (A4) in "Appendix", and the inhomogeneous term H2 is given by
Similarly, Ci and Di in Eq. (30) depend on the number of layers and the geometrical and material properties of the layers, and H2 depends on the mechanical loads \(q_{x}^{(m)}\), \(q_{z}^{(m)}\), \({\tau }_{xz}^{\left(0\right)}\), \({\tau }_{xz}^{\left(k\right)}\), \(\sigma_{zz}^{(0)}\) and \(\sigma_{zz}^{(k)}\) additionally, as shown in Eqs. (25), (29b), (31), (A3) and (A4).
Solving Eqs. (28) and (30) simultaneously will lead to the determination of \({\tau }_{xz,x}^{\left(1\right)}\) and φ,xx. For the case with H1 and H2 being constants, the general solution of this equation system has the form:
Λn is a constant depending on λn, and \(\overline{C}_{n}\) are constants to be determined from the following \(2\left(k-1\right)\) boundary conditions:
which are obtained directly from Eqs. (19) and (12a,c), with \(\overline{N}^{(m)} \,\,\text{and}\,\,\overline{M}^{(m)}\) being the axial force and bending moment acting at the two ends of the mth (m ∈ {1, 2, …, k}) layer. Using Eqs. (32a,b) and (27) in Eq. (34) will lead to a set of 2(k – 1) algebraic equations, which can be solved to obtain the constants \(\overline{C}_{n}\).
From Eqs. (27) and (32a), the interfacial stresses \({\tau }_{xz}^{\left(m\right)}\) (m ∈ {1, 2, …, k – 1}) can be determined by directly integrating \({\tau }_{xz,x}^{\left(m\right)}\), which gives
where F(m) (m ∈ {1, 2, …, k – 1}) are integration constants.
After \({\tau }_{xz}^{\left(m\right)}\) (m ∈ {1, 2, …, k – 1}) are obtained, the axial forces \({N}_{xx}^{\left(m\right)}\) and bending moments \({M}_{xx}^{\left(m\right)}\) can be acquired from the equilibrium equations in Eqs. (10a) and (11) as
where \({G}^{\left(m\right)}\), \({D}_{1}^{\left(m\right)}\) and \({D}_{2}^{\left(m\right)}\) (m ∈ {1, 2, …, k}) are integration constants to be found from the boundary conditions in Eqs. (12a) and (12c).
With the axial forces \({N}_{xx}^{\left(m\right)}\) and bending moments \({M}_{xx}^{\left(m\right)}\) obtained in Eqs. (36a,b), the axial strains \({\varepsilon }_{0}^{\left(m\right)}\) (m ∈ {1, 2, …, k}) can be readily determined from Eq. (8a).
From Eqs. (5), (8b), (9) and (21b), it follows that
With \(\tau_{xz,x}^{(m)}\) obtained in Eqs. (27) and (32a) and \(\varphi_{,xx}\) given in Eq. (32b), the interfacial normal stresses \(\sigma_{zz}^{(m)}\) (m ∈ {1, 2, …, k – 1}) can be determined sequentially from Eq. (37), starting from m = 1 through m = k – 1, with \(\sigma_{zz}^{(0)}\) prescribed on the bottom surface of the structure (see Figs. 2b and 3).
The curvature φ can be readily obtained from Eq. (32b) as, after integrating it twice,
where J1 and J2 are two integration constants to be determined by applying the boundary conditions of \(\hat{M}_{x} = \overline{M}\,\,\,\text{at}\,\,\,x = \pm L/2\) in Eq. (18c), with \(\hat{M}_{x}\) obtained from Eqs. (5), (8b), (9) and (16c) as
To illustrate the new analytical model developed in Sect. 2 for multi-layered structures with an arbitrary number of layers and subjected to general thermal and mechanical loading, three example problems for two-, three- and five-layer structures respectively are solved here by directly applying the new model. In all three cases, the solutions are obtained in closed-form expressions. For the two-layer case, the current solution recovers the one provided by Liu and Chen [22] without considering Poisson’s effect and any mechanical load as a special case and reduces to the classical solution of Timoshenko [39] for bi-metal thermostats with perfect bonding and subjected to a uniform temperature change. For the five-layer case, the closed-form solution is derived here for the first time, which can be directly used to study electronic packaging assemblies and composite joints with imperfect interfaces (e.g., [8, 30, 44]).
3.1 Two-layer structure
Consider a two-layer structure subjected to a uniform temperature change ΔT and mechanical loads on the top and bottom surfaces and at the two ends x = ± L/2, as shown in Fig. 4.
Fig. 4
Two-layer structure under thermal and mechanical loads
where \(\tau_{xz}^{(0)}\) and \(\tau_{xz}^{(2)}\) are the shear stresses applied on the bottom and top surfaces of the structure, as shown in Fig. 4, and
This is a second-order inhomogeneous ODE for \(\tau_{xz,x}^{(1)}\) with constant coefficients. When the body force components (i.e., \(q_{x}^{(m)} \,\text{and}\,\,q_{z}^{(m)}\), m ∈ {1, 2}), the distributed normal stress (i.e., \(\sigma_{zz}^{(0)}\) and \(\sigma_{zz}^{(2)}\)) and the first derivative of the distributed shear stress (i.e., \(\tau_{xz,x}^{(0)}\) and \(\tau_{xz,x}^{(2)}\)) prescribed on the top and bottom surfaces are constants, the inhomogeneous term on the right-hand side of Eq. (43) becomes a constant upon using Eq. (41). As a result, the general solution of Eq. (43) can be readily obtained as
where a1 and a2 are constants, and γ1 and γ2 are the roots of the characteristic equation (a quadratic one) of the homogeneous part of Eq. (43) given by
where \(\hat{a}_{1}\) and \(\hat{a}_{2}\) are two constants to be determined from BCs.
For the current case (see Fig. 4) with the loading and geometrical symmetries, \(N_{x}^{(m)} \,\text{and}\,\,M_{x}^{(m)}\) are symmetric about the midspan at x = 0, and thus \(\tau_{xz,x}^{(1)}\) is symmetric about x = 0 based on Eq. (19), i.e., \(\tau_{xz,x}^{(1)} \left( x \right) = \tau_{xz,x}^{(1)} \left( { - x} \right)\). Applying this symmetry condition to Eq. (46) leads to \(\hat{a}_{1} = 0\), and hence Eq. (46) becomes
which results from the BCs of \(\hat{M}_{x} =\overline{M}\), \(N_{x}^{(m)} =\overline{N}^{(m)}\) and \(M_{x}^{(m)} =\overline{M}^{(m)}\) at \(x = \pm L/2\) (see Eqs. (18c), (18f), (12a) and (12c)) and Eqs. (5), (8b), (9) and (39).
Using Eqs. (41), (45) and (48) in Eq. (50) leads to
Clearly, it is seen from Eq. (52) that the curvature φ of the two-layer structure with the imperfect interface depends on both the temperature change ΔT and mechanical loads \(\hat{q}_{z} \,( = \sum\nolimits_{m = 1}^{3}{q_{z}^{(m)} } \,)\), \(\sigma_{zz}^{(0)}\), \(\sigma_{zz}^{(2)}\), \(\tau_{xz,x}^{(0)}\), \(\tau_{xz,x}^{(2)}\), \(\overline{N}^{(1)}\), \(\overline{N}^{(2)}\), \(\overline{M}^{(1)}\)and \(\overline{M}^{(2)}\). In addition, φ is directly related to the material properties \(E^{(1)}\), \(E^{(2)}\), \(\nu^{(1)}\), \(\nu^{(2)}\), \(\alpha^{(1)}\)and \(\alpha^{(2)}\)(see Eqs. (7a)–(7c)), the interfacial stiffness constant K1, and the geometrical parameters t1, t2, L and b.
For the special case with ν(m) = 0 (m ∈ {1, 2}), \(q_{z}^{(m)}\) = 0 (m ∈ {1, 2}), \(\sigma_{zz}^{(0)}\) = 0, \(\tau_{xz}^{(0)}\) = 0, \(\sigma_{zz}^{(2)}\) = 0, \(\tau_{xz}^{(2)}\) = 0, \(\overline{N}^{(1)}\) = 0, \(\overline{N}^{(2)}\) = 0, \(\overline{M}^{(1)}\) = 0 and \(\overline{M}^{(2)}\) = 0, Eq. (52) simplifies to, with the help of Eq. (7c),
The curvature of the deformed two-layer structure obtained in Eq. (53a) is identical to that derived in Liu and Chen [22] by using a different approach for the special case without considering Poisson’s effect, interfacial normal stress, body forces, distributed normal and shear stresses prescribed on the top and bottom surfaces, and normal forces, transverse shear forces and bending moments applied at the two ends. This provides a validation of the current new model. In addition, using Eq. (53a) in Eq. (6) gives, with the help of Eqs. (2a), (5), (7c), (8a) and (12a), the axial normal stress in layer 2 in this special case as
A comparison of Eq. (53b) with Eq. (1) in Noyan et al. [28], which is based on a shear-lag model, shows that the functional relationship between \(\sigma_{xx}^{(2)}\) and x given by the current solution for the film (top layer) with two free-free ends is the same as that presented in Noyan et al. [28], thereby further validating the new model.
Equation (53a) explicitly shows that the bending deformation (as measured by the curvature φ) for the special case with only a temperature change (but without any mechanical load) is purely due to the mismatch of the coefficients of thermal expansion of the two layers that are imperfectly bonded with 0 < K1 < ∞. If the two layers are completely separated, then K1 = 0 and hence φ = 0 according to Eqs. (45) and (53a). That is, there will be no bending deformation induced by the temperature change if the two layers are not bonded at all. However, this is no longer the case when a mechanical load is applied in addition to the temperature change. It is seen from Eq. (52) that with any of \(\hat{q}_{z}\), \(\sigma_{zz}^{(0)}\), \(\sigma_{zz}^{(2)}\), \(\tau_{xz,x}^{(0)}\), \(\tau_{xz,x}^{(2)}\), \(\overline{M}^{(1)}\) and \(\overline{M}^{(2)}\) being non-zero, φ ≠ 0 even if K1 = 0.
When the two layers are perfectly bonded, K1 → ∞ and Eq. (52) reduces to
which is the curvature of the perfectly bonded two-layer structure under the general thermal and mechanical loading. If ν(m) = 0 (m ∈ {1, 2}), \(q_{z}^{(m)}\) = 0 (m ∈ {1, 2}), \(\sigma_{zz}^{(0)}\) = 0, \(\tau_{xz}^{(0)}\) = 0, \(\sigma_{zz}^{(2)}\) = 0, \(\tau_{xz}^{(2)}\) = 0, \(\overline{N}^{(1)}\) = 0, \(\overline{N}^{(2)}\) = 0, \(\overline{M}^{(1)}\) = 0 and \(\overline{M}^{(2)}\) = 0, Eq. (54) further simplifies to, upon using Eq. (7c),
which is the same as that initially derived by Timoshenko [39] for perfectly bonded bi-metal thermostats subjected to a uniform temperature change (see also Zhang and Xing [45] for an alternative derivation of Eq. (55)). This indicates that the classical constant curvature formula for perfectly bonded bi-layer structures first obtained by Timoshenko [39] is recovered by the current two-layer solution as a special case.
In addition, for perfectly bonded two-layer structures subjected to both a temperature change and applied mechanical loads, Eq. (54) shows that there will always be bending deformations (with φ ≠ 0) even if there is no temperature change (i.e., ΔT = 0) or no mismatch in the CTE for the two layers (i.e., \(\alpha^{(1)} - \alpha^{(2)} = 0\)). This differs from what is exhibited by two-layer structures subjected to a temperature change only, for which the curvature is given by Eq. (55), which is a special case of the current formula derived in Eq. (54) for perfectly bonded two-layer structures under general thermal and mechanical loading.
3.2 Three-layer structure
Consider a three-layer structure subjected to a uniform temperature change ΔT and mechanical loads on the top and bottom surfaces and at the two ends x = ± L/2, as shown in Fig. 5.
Fig. 5
Three-layer structure under thermal and mechanical loads
where \({\tau }_{xz}^{\left(0\right)}\) and \({\tau }_{xz}^{\left(3\right)}\) are the shear stresses prescribed on the bottom and top surfaces, as shown in Fig. 5. From Eq. (56a), \(\tau_{xz,x}^{(2)}\) can be obtained in terms of \(\tau_{xz,x}^{(1)}\) and \(\varphi_{,xx}\) as
Clearly, Eq. (59) is a fourth-order inhomogeneous ODE for \(\tau_{xz,x}^{(1)}\)with constant coefficients. When \(q_{x}^{(m)}\)(m ∈ {1, 2, 3}), \(q_{z}^{(m)}\)(m ∈ {1, 2, 3}), \(\sigma_{zz}^{(0)}\), \(\sigma_{zz}^{(3)}\), \(\tau_{xz,x}^{(0)}\) and \(\tau_{xz,x}^{(3)}\) are constants, the inhomogeneous term on the right-hand side of Eq. (59) becomes a constant. The general solution of Eq. (59) can then be written as
where c1 – c4 are four constants, \(\beta_{ - 1}^{(2)} ,\,\,\beta_{ - 1}^{(1)},\,\,\beta_{0}^{(1)} ,\,\,\beta_{0}^{(2)} ,\,\,\beta_{1}^{(1)},\,\,\beta_{1}^{(2)} ,\,\,\Gamma^{(1)}\,\text{and}\,\,\Gamma^{(2)}\)are listed in Eqs. (41) and (61), and δ1 – δ4 are the roots of the characteristic equation of the homogeneous part of Eq. (59), which has the form:
where use has been made of Eqs. (41) and (61). It is clear that the four roots in Eq. (64) are all real, and hence the general solution given in Eq. (62) can be rewritten as
where \(\hat{c}_{1} -\hat{c}_{4}\) are four constants to be determined from BCs.
For the current case (see Fig. 5) with the loading and geometrical symmetries, \(\tau_{xz,x}^{(1)} \left( x \right) = \tau_{xz,x}^{(1)} \left( { - x} \right)\). Applying this symmetry condition to Eq. (65) leads to \(\hat{c}_{1} = \hat{c}_{3} = 0\), and hence Eq. (65) becomes
Substituting Eq. (68) into Eq. (34) for the second interface (i.e., m = 2) gives, with \(\eta_{x}^{(1)} =0\) from Eq. (41) for the current problem with \(q_{x}^{(m)}\)(m ∈ {1, 2, 3}) being constants,
With \(\hat{c}_{2}\) and \(\hat{c}_{4}\) obtained in Eq. (70), \(\tau_{xz,x}^{(1)}\) and \(\tau_{xz,x}^{(2)}\) are then fully determined from Eqs. (66) and (68).
From Eqs. (5), (8b), (9), (36a), (51), (71) and (73), it follows that, with \(\eta_{x}^{(1)} = 0\) and \(q_{x}^{(m)}\)(m ∈ {1, 2, 3}), \(q_{z}^{(m)}\)(m ∈ {1, 2, 3}), \(\sigma_{zz}^{(0)}\), \(\sigma_{zz}^{(3)}\), \(\tau_{xz,x}^{(0)}\) and \(\tau_{xz,x}^{(3)}\) being constants,
where use has also been made of the BCs of \(N_{x}^{(1)} =\overline{N}^{(1)}\,\) at \(x = \pm L/2\) (see Eq. (12a) and Fig. 5) to determine the two integration constants involved in \(N_{x}^{(1)}\). In addition, it can be readily verified that the BCs of \(M_{x}^{(1)}=\overline{M}^{(1)}\) at \(x = \pm L/2\) (see Eq. (12c) and Fig. 5) are satisfied by \(M_{x}^{(1)}\)obtained in Eq. (74b), where use has been made of the BCs of \(\left. \varphi \right|_{x = \pm L/2} =12\overline{M}^{(1)} /\left( {bc_{11}^{(1)}t_{1}^{3} }\right)\) given in Eq. (51).
Similarly, the use of Eqs. (5), (8b), (9), (36a), (51), (68), (71) and (73) leads to, with \(\eta_{x}^{(1)} =0\) and \(q_{x}^{(m)}\)(m ∈ {1, 2, 3}), \(q_{z}^{(m)}\)(m ∈ {1, 2, 3}), \(\sigma_{zz}^{(0)}\), \(\sigma_{zz}^{(3)}\), \(\tau_{xz,x}^{(0)}\) and \(\tau_{xz,x}^{(3)}\) being constants,
where use has also been made of the BCs of \(N_{x}^{(2)} = \overline{N}^{(2)}\) at \(x = \pm L/2\) (see Eq. (12a) and Fig. 5) to determine the two integration constants involved in \(N_{x}^{(2)}\). In addition, it is clear that the BCs of \(M_{x}^{(2)} =\overline{M}^{(2)}\) at \(x = \pm L/2\) (see Eq. (12c) and Fig. 5) are satisfied by \(M_{x}^{(2)}\)obtained in Eq. (75b), where use has been made of the BCs of \(\left. \varphi \right|_{x = \pm L/2} =12\overline{M}^{(2)} /\left( {bc_{11}^{(2)} t_{2}^{3} }\right)\) given in Eq. (51).
Finally, from Eqs. (5), (8b), (9), (36a), (51), (68), (71) and (73), it can be readily shown that, with \(\eta_{x}^{(1)} = 0\) and \(q_{x}^{(m)}\)(m ∈ {1, 2, 3}), \(q_{z}^{(m)}\)(m ∈ {1, 2, 3}), \(\sigma_{zz}^{(0)}\), \(\sigma_{zz}^{(3)}\), \(\tau_{xz,x}^{(0)}\) and \(\tau_{xz,x}^{(3)}\) being constants,
where use has also been made of the BCs of \(N_{x}^{(3)} =\overline{N}^{(3)}\) at \(x = \pm L/2\) (see Eq. (12a) and Fig. 5) to determine the two integration constants involved in \(N_{x}^{(3)}\). In addition, it is seen that the BCs of \(M_{x}^{(3)} =\overline{M}^{(3)}\) at \(x = \pm L/2\) (see Eq. (12c) and Fig. 5) are satisfied by \(M_{x}^{(3)}\)obtained in Eq. (76b), where use has been made of the BCs of \(\left. \varphi \right|_{x = \pm L/2} =12\overline{M}^{(3)} /\left( {bc_{11}^{(3)} t_{3}^{3} }\right)\) given in Eq. (51).
From Eqs. (2b), (8a), (74a), (75a) and (76a), \(u_{0}^{(m)}\)(m ∈ {1, 2, 3}) can be readily determined.
The interfacial normal stresses \(\sigma_{zz}^{(m)}\)(m ∈ {1, 2}) can be obtained from Eqs. (37), (68), (71) and (72) as
For the special case with ν(m) = 0 (m ∈ {1, 2, 3}), \(q_{z}^{(m)}\) = 0 (m ∈ {1, 2, 3}), \(\sigma_{zz}^{(0)}\) = 0, \(\tau_{xz,x}^{(0)}\) = 0, \(\sigma_{zz}^{(3)}\) = 0 and \(\tau_{xz,x}^{(3)}\) = 0, Eq. (73) becomes, with the help of Eqs. (7c), (41), (64) and (70),
Clearly, Eq. (78) shows that φ = 0 when ΔT = 0 or \(\alpha^{(1)} = \alpha^{(2)} = \alpha^{(3)}\). That is, for three-layer structures subjected to a uniform temperature change only (without any mechanical load), there will be no bending deformation if there is no temperature change or no mismatch in the CTE for the three layers. However, this is no longer the case if any of the mechanical loads \(q_{z}^{(m)}\)(m ∈ {1, 2, 3}), \(\sigma_{zz}^{(0)}\), \(\tau_{xz,x}^{(0)}\), \(\sigma_{zz}^{(3)}\), \(\tau_{xz,x}^{(3)}\), \(\overline{N}^{(m)}\)(m ∈ {1, 2, 3}) and \(\overline{M}^{(m)}\)(m ∈ {1, 2, 3}) is non-zero, which can be seen from Eqs. (73) and (70) of the current solution for three-layer structures subjected to general thermal and mechanical loading.
3.3 Five-layer structure
Consider a five-layer structure subjected to a uniform temperature change ΔT and mechanical loads on the top and bottom surfaces and at the two ends x = ± L/2, as shown in Fig. 6.
Fig. 6
Five-layer structure under thermal and mechanical loads
From Eqs. (79a–c), \(\tau_{xz,x}^{(2)}\), \(\tau_{xz,x}^{(3)}\) and \(\tau_{xz,x}^{(4)}\) can be obtained in terms of \(\tau_{xz,x}^{(1)}\) and φ,xx as
where \(\beta_{ - 1}^{(1)} ,\,\beta_{ - 1}^{(2)},\,\beta_{0}^{(1)} ,\beta_{0}^{(2)} ,\beta_{1}^{(1)},\,\beta_{1}^{(2)} ,\,\eta_{x}^{(1)} ,\,\eta_{x}^{(2)} ,\,\Gamma^{(1)} \,\text{and}\,\Gamma^{(2)}\) are given in Eqs. (41) and (61), and
Solving the two coupled ODEs in Eqs. (81) and (82) simultaneously yields \(\tau_{xz,x}^{(1)}\) and φ,xx for the current five-layer case, with \(q_{x}^{(m)}\)(m ∈ {1, 2, 3, 4, 5}), \(\,q_{z}^{(m)}\)(m ∈ {1, 2, 3, 4, 5}), \(\sigma_{zz}^{(0)}\), \(\sigma_{zz}^{(5)}\), \(\tau_{xz,x}^{(0)}\) and \(\tau_{xz,x}^{(5)}\) (and thus H1 and H2) being constants, as
where s1 – s8 are constants, and \({\omega }_{1}-{\omega }_{8}\) are the eight roots of the following eighth-degree polynomial characteristic equation:
in which \(\beta_{ - 1}^{(m)} ,\,\beta_{0}^{(m)} ,\,\beta_{1}^{(m)} ,\,\eta_{x}^{(m)} \,\text{and}\,\Gamma^{(m)}\) are given in Eqs. (41), (61) and (83).
In addition, the coefficients \({\Lambda }_{n}\) (n ∈ {1, 2, …, 8}) in Eq. (84b) are given by
When Δd > 0, Eq. (92) has three distinct real roots that can be determined from Viète’s trigonometric solution, which lead to the three distinct real roots of Eq. (90) as, with the help of Eq. (93a),
Any of these three values of χ can be used in Eq. (89) to obtain \(\omega_{j}^{2}\) (j\(\in\) {1, 2, 3, 4}), leading to the eight roots \(\omega_{i}\) of Eq. (85).
When Δd < 0, Eq. (92) has one real root and two conjugated complex roots that can be obtained using the Cardano formula, which give the three roots of Eq. (90) as, upon using Eq. (93a),
where \(\chi_{1}\) in Eq. (96a) is the real root, which will be used in Eq. (89) to obtain \(\omega_{j}^{2}\) (j\(\in\) {1, 2, 3, 4}), and thereby the eight roots \(\omega_{i}\) of Eq. (85).
When Δd = 0, Eq. (92) has a triple root of 0 if Λ = 0 or a single root of \(\xi_{1} = \frac{3\Omega }{\Lambda }\) and a double root of \(\xi_{2} = \xi_{3} = - \frac{3\Omega }{{2\Lambda }}\) if Λ ≠ 0. Then, it follows from these results and Eq. (93a) that the roots of Eq. (90) in this case are given by
Using any of the three values of χ listed in Eq. (97a) or (97b) in Eq. (89) will yield \(\omega_{j}^{2}\) (j\(\in\) {1, 2, 3, 4}) and thus the eight roots \(\omega_{i}\) of Eq. (85).
It is clear from Eq. (89) that the eight roots ωi of the characteristic equation in Eq. (85) are in four square-root pairs (i.e., \(\omega_{1}\) and \(\omega_{5}\, ( = - \omega_{1} )\), \(\omega_{2}\) and \(\omega_{6}\, ( = - \omega_{2} )\), \(\omega_{3}\) and \(\omega_{7}\, ( = - \omega_{3} )\), and \(\omega_{4}\) and \(\omega_{8}\, ( = - \omega_{4} )\)). As a result, the general solution in Eqs. (84a,b) can be rewritten as
where \(\hat{s}_{1} - \hat{s}_{8}\) are eight constants to be determined from BCs. Note that in reaching Eq. (98b) use has been made of the relations Λ5 = Λ1, Λ6 = Λ2, Λ7 = Λ3 and Λ8 = Λ4, which follow directly from Eqs. (87) and (88) with ω5 = − ω1, ω6 = − ω2, ω7 = − ω3 and ω8 = − ω4.
For the current case with the loading and geometrical symmetries, \(\tau_{xz,x}^{(1)}\) is symmetric about x = 0 (i.e., \(\tau_{xz,x}^{(1)} \left( x \right) = \tau_{xz,x}^{(1)} \left( { - x} \right)\)). Using this symmetry condition in Eq. (98a) yields \(\hat{s}_{1} = \hat{s}_{3} = \hat{s}_{5} = \hat{s}_{7} = 0\), and hence the solution in Eqs. (98a,b) reduces to
Solving Eqs. (100) and (102a)-(102c), which form a system of four linear algebraic equations, will lead to the determination of the four constants \(\hat{s}_{2} ,\,\hat{s}_{4} ,\,\hat{s}_{6} \,\,\text{and}\,\,\hat{s}_{8}\).
Substituting these determined constants \(\hat{s}_{2} ,\,\hat{s}_{4} ,\,\hat{s}_{6} \,\,\text{and}\,\,\hat{s}_{8}\) into Eqs. (99a) and (101a–c) will yield the final expressions of \(\tau_{xz,x}^{(m)}\) (m ∈ {1, 2, 3, 4}).
Integrating Eq. (99b) twice with respect to x gives, with H1 and H2 being constants,
The substitution of the determined constants \(\hat{s}_{2} ,\,\,\hat{s}_{4} ,\,\,\hat{s}_{6} \,\,\text{and}\,\,\hat{s}_{8}\) into Eq. (105) will lead to the final expression of the curvature of the five-layer structure.
The rest of the quantities, including \(N_{x}^{(m)}\), \(M_{x}^{(m)}\) and \(u_{0}^{(m)}\) (m ∈ {1, 2, 3, 4, 5}), for the current five-layer structure can be determined by following procedures similar to those used in Sect. 3.2 for the three-layer structure.
3.4 Numerical Results
To quantitatively illustrate the new analytical model developed in Sect. 2 and the closed-form solutions obtained above in this section, two transistor stacks subjected to thermomechanical loading are considered as numerical examples in this sub-section.
A typical industrial transistor stack has a multi-layer structure containing a semiconductor (e.g., silicon (Si)) layer, a thermal barrier/spreader (e.g., beryllium oxide (BeO)), and a metal substrate (e.g., aluminum (Al) or copper (Cu)), which are connected through eutectic solder joints (e.g., lead/tin alloy (Pb/Sn)) (e.g., [30]). Two transistor stacks are analyzed here, one of which is five-layer and the other is seven-layer, as shown in Figs. 7a and 7b. The difference is that only one Pb/Sn joint is used in the former, while three Pb/Sn layers are employed in the latter. A thermomechanical analysis has been conducted for the five-layer transistor stack displayed in Fig. 7a by Pao and Eisele [30], who extended the bi-metal thermostat model (e.g., [31, 36, 37, 39]) to determine interfacial shear and peel (normal) stresses in five-layer joints and transistors. Pao and Eisele [30] obtained a system of four coupled second-order ODEs for the interfacial stresses, which was converted to an eigenvalue problem and subsequently solved numerically. Their analysis is limited to five-layer structures and confined to cases with temperature changes (through uniform heating or cooling) but without any mechanical load, which is different from the current model that can be applied to structures with an arbitrary number of layers and subjected to both mechanical loads and temperature changes.
Fig. 7
a Five-layer transistor stack; b seven-layer transistor stack; c interfacial shear stresses \(\tau_{xz}^{(m)}\)(m ∈ {1, 2, 3, 4}) predicted by the current model (solid lines) and obtained in Pao and Eisele [30] (points with marks)
In order to directly compare with the results reported in Pao and Eisele [30], the geometrical and material parameters used by them are adopted for the five-layer transistor stack considered in the current numerical example. The same material properties and geometrical dimensions are also employed for the seven-layer transistor stack which contains two more Pb/Sn layers. The parameter values for both the five- and seven-layer transistor stacks are listed in Table 1.
Table 1 Material and geometrical parameters for the five-layer and seven-layer transistors
As a first approximation, the stiffness constant Km of the mth interface between layers m and m + 1 is taken to be the average of the shear modulus values of the two adjacent layers over the average thickness of the two layers (i.e., \({t}_{i}^{\left(m\right)}=\left({t}_{m}+{t}_{m+1}\right)/2\)) given by
Note that this expression remains the same for both the plane stress and plane strain cases. Equation (106) is used to obtain the interfacial stiffness constants Km for the five-layer (with m ∈ {1, 2, 3, 4}) and seven-layer (with m ∈ {1, 2, 3, 4, 5, 6}) stacks considered here from the values of Young’s modulus E(m), Poisson’s ratio ν(m) and layer thickness tm listed in Table 1.
In addition, the material constants \(c_{11}^{(m)}\) and \(d_{11}^{(m)}\) for each layer in the two example problems are determined from the values of E(m), ν(m) and α(m) given in Table 1 using Eq. (7a) for the plane stress case.
3.4.1 Model verification
The newly developed analytical model is verified against the results of Pao and Eisele [30] for a five-layer transistor stack consisting of a Si chip, a BeO thermal spreader, an eutectic solder (lead/tin alloy (Pb/Sn)), a Cu interlayer, and an Al substrate, as shown in Fig. 7a. This five-layer stack is subjected to a temperature change of \(\Delta T=-65\,^\circ{\rm C} \) (uniform cooling) without any applied mechanical load (i.e., \(\sigma_{zz}^{(0)}\) = \(\sigma_{zz}^{(5)}\) = \(\tau_{xz}^{(0)}\) = \(\tau_{xz}^{(5)}\) = 0). In addition, the body forces are all neglected (i.e., \(q_{x}^{(m)} \text{ = }\,q_{z}^{(m)} = 0\), m ∈ {1, 2, 3, 4, 5}) and the axial force, transverse shear force and bending moment at the two ends of each layer are taken to be zero-valued (i.e., \(\overline{N}^{(m)} = \overline{V}^{(m)} = \overline{M}^{(m)} = 0\), m ∈ {1, 2, 3, 4, 5}) at x = ± L/2.
The interfacial shear stresses \(\tau_{xz}^{(m)}\)(m ∈ {1, 2, 3, 4}) varying with x (the distance from the midspan) obtained using Eqs. (99a) and (101a–c) are plotted in Fig. 7c, where the results presented in Pao and Eisele [30] are also displayed for comparison. It is seen from Fig. 7c that the interfacial shear stresses \(\tau_{xz}^{(m)}\)(m ∈ {1, 2, 3, 4}) predicted by the current model agree well with those provided in Pao and Eisele [30], which verifies and supports the newly developed model.
3.4.2 Transistor stacks under thermal and mechanical loads
To investigate the responses of the five-layer transistor stack under both thermal and mechanical loads, the stack is subjected to a uniform temperature rise of \(\Delta T=+ 65\,^\circ{\rm C} \) and a mechanical load \({\sigma }_{zz}^{\left(5\right)}=-20 \mathrm{MPa}\) (a uniform compression) applied on the top surface. The other mechanical loads are taken to be zero-valued, i.e., \(\sigma_{zz}^{(0)}\) = \(\tau_{xz}^{(0)}\) = \(\tau_{xz}^{(5)}\) = 0, \(q_{x}^{(m)} \text{ = }\,q_{z}^{(m)} = 0\) (m ∈ {1, 2, 3, 4, 5}), and \(\overline{N}^{(m)} = \overline{V}^{(m)} = \overline{M}^{(m)} = 0\) (m ∈ {1, 2, 3, 4, 5}) at x = ± L/2.
Figure 8 shows the mechanical responses of the five-layer transistor stack under the prescribed thermomechanical loading. The numerical results displayed in Fig. 8 are obtained using Eqs. (99a) and (101a–c) for \(\tau_{xz}^{(m)}\) (m ∈ {1, 2, 3, 4}), Eqs. (37) and (99b) for \(\sigma_{zz}^{(m)}\)(m ∈ {1, 2, 3, 4}), Eqs. (36a,b) or \(N_{x}^{(m)}\) and \(M_{x}^{(m)}\) (m ∈ {1, 2, 3, 4, 5}), Eq. (8a) for \(\varepsilon_{0}^{(m)}\) (m ∈ {1, 2, 3, 4, 5}), and Eq. (2b) for \(u_{0}^{(m)}\) (m ∈ {1, 2, 3, 4, 5}).
Fig. 8
Mechanical responses of the five-layer transistor stack subjected to a temperature rise of \(\Delta T=+65\,^\circ{\rm C} \) and an applied load of \({\sigma }_{zz}^{\left(5\right)}=-20\,\mathrm{MPa}\): a interfacial shear stresses \({\tau }_{xz}^{\left(m\right)}\), b interfacial normal stresses \({\sigma }_{zz}^{\left(m\right)}\), c bending moments \({M}_{x}^{\left(m\right)}\), d axial strains \({\varepsilon }_{0}^{\left(m\right)}\), e axial displacements \({u}_{0}^{\left(m\right)}\), and f normal forces \({N}_{x}^{\left(m\right)}\)
It is observed from Figs. 8a and 8b that the interfacial shear stress \(\tau_{xz}^{(m)}\) and normal stress \(\sigma_{zz}^{(m)}\) both change sharply near the two ends of the stack, indicating a strong edge (boundary) effect. In addition, the interfacial shear stress at the BeO-Si interface, i.e., \(\tau_{xz}^{(4)}\), is found to be the lowest due to the small CTE mismatch between the BeO and Si layers, while the highest interfacial shear stress (excluding the regions with the edge effect) is developed at the BeO-Pb/Sn interface, i.e., \(\tau_{xz}^{(3)}\), because of the large CTE mismatch between the BeO and Pb/Sn layers. Furthermore, the interfacial normal stress at the BeO-Si interface, i.e., \(\sigma_{zz}^{(4)}\), is seen to be the smallest, while the interfacial normal stress at the Cu-Al interface, i.e., \(\sigma_{zz}^{(1)}\), is the largest.
From Figs. 8d and 8e, it is observed that the BeO layer (thermal spreader) exhibits the smallest axial strain \(\varepsilon_{0}^{(4)}\) and axial displacement \(u_{0}^{(4)}\) among all layers. This is due to the low CTE and high elastic modulus of BeO, as shown in Table 1. In contrast, the highest axial strain \(\varepsilon_{0}^{(1)}\) and axial displacement \(u_{0}^{(1)}\) are developed in the Al substrate because of the high CTE and low elastic modulus of Al.
Figures 8c–8f show that all layers are bent due to the applied mechanical load \({\sigma }_{zz}^{\left(5\right)}\) and the large CTE mismatch between the bottom three layers (Al, Cu, and Pb/Sn) and the top two layers (BeO and Si). In addition, Fig. 8e shows that each of the Al, Cu, Pb/Sn and BeO layers exhibits an axial stretch, while the Si layer displays an axial compression. The transition from the stretch to the compression indicates the existence of a neutral surface, which is located in the solder (Pb/Sn) layer where both the bending moment and axial normal force are seen to be almost zero from Figs. 8c and 8f.
To further understand how the curvature φ and deflection w of the transistor stack vary with applied mechanical loads, φ and w of the five-layer stack subjected to the temperature rise \(\Delta T=+65\,^\circ{\rm C} \) and applied mechanical load \({\sigma }_{zz}^{\left(5\right)}\) of different magnitudes are plotted in Figs. 9a and 9b, respectively. The numerical results shown in Figs. 9a and 9b are obtained using Eqs. (5) and (105), with \(\sigma_{zz}^{(0)}\) = \(\tau_{xz}^{(0)}\) = \(\tau_{xz}^{(5)}\) = 0, \(q_{x}^{(m)} \text{ = }\,q_{z}^{(m)} = 0\) (m ∈ {1, 2 , 3, 4, 5}) and \(\overline{N}^{(m)} = \overline{V}^{(m)} = \overline{M}^{(m)} = 0\) (m ∈ {1, 2 , 3, 4, 5}) at x = ± L/2. It is clear from Figs. 9a and 9b that both the curvature φ and deflection w (magnitude) increase with the increase of the applied load \({\sigma }_{zz}^{\left(5\right)}\) (in magnitude).
Fig. 9
a Curvature φ and b deflection \(w\) of the five-layer transistor stack induced by the temperature rise \(\Delta T=+65\,^\circ{\rm C} \) and applied mechanical load \({\sigma }_{zz}^{\left(5\right)}\) of different magnitudes; c curvature φ and d deflection \(w\) of the seven-layer transistor stack induced by the temperature rise \(\Delta T=+65\,^\circ{\rm C} \) and applied mechanical load \({\sigma }_{zz}^{\left(7\right)}\) of different magnitudes
Finally, to demonstrate the application of the newly developed model for stacks with more than five layers, the seven-layer transistor stack shown in Fig. 7b is considered. In this transistor stack, the Pb/Sn solder joint is used between every two adjacent layers, as indicated in Fig. 7b. The transistor stack is subjected to a temperature rise of \(\Delta T=+65\,^\circ{\rm C} \) and a mechanical load \({\sigma }_{zz}^{\left(7\right)}\) varying from 0 to \(-20\,\mathrm{MPa}\) on the top surface. The curvature and deflection of this seven-layer transistor stack are plotted in Figs. 9c, and 9d, respectively. The numerical values displayed in these figures are obtained using Eqs. (5) and (38) in the general analytical model, with \(\sigma_{zz}^{(0)}\) = \(\tau_{xz}^{(0)}\) = \(\tau_{xz}^{(7)}\) = 0, \(q_{x}^{(m)} \text{ = }\,q_{z}^{(m)} = 0\) (m ∈ {1, 2, 3, 4, 5, 6, 7}) and \(\overline{N}^{(m)} = \overline{V}^{(m)} = \overline{M}^{(m)} = 0\) (m ∈ {1, 2, 3, 4, 5, 6, 7}) at x = ± L/2.
It is observed from Figs. 9c and 9d that both φ and w (magnitude) of the seven-layer stack increase with the increase of the applied mechanical load \({\sigma }_{zz}^{\left(7\right)}\) (in magnitude), which is similar to what is exhibited by the five-layer transistor stack (see Figs. 9a and 9b). However, compared to the five-layer stack, the overall stiffness of the seven-layer stack is enhanced due to the inclusion of two additional solder (Pb/Sn) layers, and, as a result, the curvature φ and deflection w (magnitude) of the latter are lower than those of the former under the same thermal and mechanical loads, as shown in Fig. 9.
4 Summary
A new analytical model for multi-layered structures under general thermomechanical loading is developed using an extended Bernoulli–Euler beam theory and a slip-interface model. It accounts for Poisson’s effect and allows for slipping between two adjacent layers. An analytical solution for a multi-layered structure with an arbitrary number of layers and subjected to a temperature change and mechanical loads on the top and bottom surfaces and at the two ends is obtained here for the first time by following a new approach that first determines one interfacial shear stress along with the curvature of the deformed structure.
Three example problems for two-, three- and five-layer structures respectively are analytically solved by directly applying the new model. Closed-form solutions are derived for all three problems, each of which includes Poisson’s effect and considers a uniform temperature change and general mechanical loads, unlike existing solutions. The current solution for two-layer structures recovers the solution obtained by Liu and Chen [22] without considering Poisson’s effect and mechanical loading and the classical solution of Timoshenko [39] for perfectly bonded bi-metal thermostats as two special cases. This verifies the new model. In addition, the closed-form solution for five-layer structures with imperfect interfaces is derived here for the first time, which can find direct applications in designing electronic packaging assemblies and composite joints. Moreover, all expressions of the general and specific solutions are derived in terms of physical parameters and variables, which can be easily adjusted to fit different problems.
To quantitatively illustrate the new model and current analytical solutions, numerical results are presented for physical problems of five- and seven-layer transistor stacks subjected to prescribed thermomechanical loading. The interfacial shear stresses predicted by the current model for the five-layer transistor stack are seen to agree very well with those obtained by Pao and Eisele [30] using an extended bi-metal thermostat model, which further validates the newly developed analytical model. The general and specific solutions obtained in the current study can be applied to analyze multi-layered stack structures with imperfect interfaces that are commonly used in electronic assemblies, composite beam-columns, and all-solid-state batteries.
Finally, it should be mentioned that the current formulation is limited to multi-layered structures with a uniform temperature field. It can be extended to cases with non-uniform temperature distributions and/or creep deformations, which is currently being pursued.
References
Barbero, E.J., Reddy, J.N.: Modeling of delamination in composite laminates using a layer-wise plate theory. Int. J. Solids Struct. 28, 373–388 (1991)
Benveniste, Y.: The effective mechanical behaviour of composite materials with imperfect contact between the constituents. Mech. Mater. 4, 197–208 (1985)
Chen, Q.-Q., Xuan, F.-Z., Tu, S.-T.: Residual stress analysis in the film/substrate system with the effect of creep deformation. J. Appl. Phys. 106, 033512-1–8 (2009)
Gao, X.-L., Li, K., Mall, S.: A mechanics-of-materials model for predicting Young’s modulus of damaged woven fabric composites involving three damage modes. Int. J. Solids Struct. 40, 981–999 (2003)
Gu, S.T., He, Q.C.: Interfacial discontinuity relations for coupled multifield phenomena and their application to the modeling of thin interphases as imperfect interfaces. J. Mech. Phys. Solids 59, 1413–1426 (2011)
Ko, M. F., Goodman, J. R., Vanderbilt, M. D., Criswell, M. E. and Bodig, J.: Layered beam systems with interlayer slip, Structural Research Report No. 8, Civil Engineering Department, Colorado State University, Fort Collins, Colorado (1972)
Liu, D.Y., Chen, W.Q.: Thermal stresses in bilayer systems with weak interface. Mech. Res. Commun. 37, 520–524 (2010)
Su, Y.-Y., Gao, X.-L.: Analytical model for adhesively bonded composite panel-flange joints based on the Timoshenko beam theory. Compos. Struct. 107, 112–118 (2014)
Zhang, Y., Zhao, Y.-P.: Applicability range of Stoney’s formula and modified formulas for a film/substrate bilayer. J. Appl. Phys. 99, 053513-1–7 (2006)
The authors would like to gratefully acknowledge the support by the Moody School of Graduate and Advanced Studies at SMU through a post-doctoral fellowship to MS. The authors also wish to thank Prof. Shaofan Li of UC-Berkeley and two anonymous reviewers for their encouragement and helpful comments on an earlier version of the paper.
Funding
Open access funding provided by SCELC, Statewide California Electronic Library Consortium.
Author information
Authors and Affiliations
Department of Mechanical Engineering, Southern Methodist University, Dallas, TX, 75275, USA
M. Shaat & X.-L. Gao
Modeling and Virtualization Center, Schlumberger Technology Corp, Sugar Land, TX, 77478, USA
K. Li
US Army DEVCOM AC WSEC Benét Laboratories, Watervliet, NY, 12189, USA
Note that for the expression of D2 defined in Eq. (A4) an additional term \(\frac{1}{12}{\sum }_{m=1}^{k}\left({c}_{11}^{\left(m\right)}{t}_{m}^{3}\right)\) needs to be included, as shown in Eq. (86d) for the five-layer case.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Shaat, M., Gao, XL., Li, K. et al. New analytical model for thermomechanical responses of multi-layered structures with imperfect interfaces.
Acta Mech234, 5779–5818 (2023). https://doi.org/10.1007/s00707-023-03659-3
