**1. Introduction**

302 Viscoelasticity – From Theory to Biological Applications

21, No. 5, pp. 745-749, ISSN 0267-8292

40, No. 6, pp. 1079-1093, ISSN 0148-6055

No. 24, pp. 10890-10896, ISSN 0032-3861

pp. 287-303, ISSN 0022-2461

*Polymer*, Vol. 42, No. 18, pp. 7847-7856, ISSN 0032-3861

ISSN 0024-9297

[34] Han, C. D. & Kim, S. S. (1995). Shear-induced isotropic-to-nematic transition in a thermotropic liquid-crystalline polymer. *Macromolecules*, Vol. 28, No. 6, pp. 2089-2092,

[35] de Schrijver, P,; van Dael, W. & Thoen, J. (1996). Surface-induced pretransitional order in the isotropic phase near the isotropic-nematic phase transition. *Liquid Crystals*, Vol.

[36] Chan, C. K.; Whitehouse, C.; Gao, P. & Chai, C. K. (2001). Flow induced chain alignment and disentanglement as the viscosity reduction mechanism within TLCP/HDPE blends.

[37] Yu, T. H. & Wilkes, G. L. (1996). Influence of molecular weight distribution on the melt extrusion of high density polyethylene (HDPE): Effects of melt relaxation behavior on morphology and orientation in HDPE extruded tubular films. *Journal of Rheology*, Vol.

[38] Zhou, H. & Wilkes, G. L. (1998). Orientation-dependent mechanical properties and deformation morphologies for uniaxially melt-extruded high-density polyethylene films having an initial stacked lamellar texture. *Journal of Material Science*, Vol. 33, No. 2,

[39] Chan, C. K. & Gao, P. (2005) Shear-induced interactions in blends of HMMPE containing a small amount of thermotropic copolyester HBA/HQ/SA. *Polymer*, Vol. 46,

[40] Chan, C. K. & Gao, P. (2005). A phenomenological model for the viscosity reductions in blends of HMMPE containing a small quantity of thermotropic liquid crystalline copolyester HBA/HQ/SA. *Polymer*, Vol. 46, No. 19, pp. 8151-8156, ISSN 0032-3861

Along with rapid assembly for the purpose of creating thinner printed circuit boards, the side effect of warping during the reflow process is inevitable. As a result, the assembly process encounters serious challenges, such as difficulty in implementing a highly integrated assembly, as well as degradation in the reliability of connectivity. Aside from attempts to simulate these issues employing the FE analysis method, it is also necessary to conduct an estimation of the warpage at the early stage of design, for which development of more simplified estimation tools is strongly desired. From a material behavior point of view, if any glass transition points exist within the temperature range of the reflow process, the relaxation effect (due to viscoelastic characteristics of the material of the boards) appears as a deformation, which acts as a barrier to achieving a distinct estimation of the amount of warpage. Generally speaking, the viscoelastic behavior of resin materials exhibits very sensitive temperature dependency. This makes it difficult to accurately capture the characteristics of resin materials in actual experiments, and accordingly, to build numerical models based on such an experimental measurement. There have been many analysis cases published using sophisticated FE approaches (Shrotriya et al., 2005 and Valdevitet al., 2008), but these approaches have not extended beyond applications as a handy design tool.

To easily estimate the thermal deformation in the laminated structure, some approaches employ the multilayered beam theory (Lim, 2008 and Yuju, 2003). The method proposed in this paper enhances these approaches to incorporate the layered plate theory, and includes the effect from the temperature-dependent viscoelasticity and the temperature-dependent coefficient of thermal expansion of resin materials. The program, which is equipped with the developed method, can give an arbitrary temperature history to a multilayered plate consisting of an arbitrary number of layers. As well, the practical approach for measurement

© 2012 Kobayashi et al., licensee InTech. This is an open access chapter distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. © 2012 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

of viscoelasticity characteristics has been specifically developed by the author's group, with the aim of performing measurements to obtain highly accurate data (Kobayashi, 2008). The method proposed herein was verified using the achievements from this development. Verification results are also shown in this paper. This method instantly calculates the amount of warpage and stress in a multilayered plate by giving the values of the thickness and material constants of the plate. Advantages of the proposed method may cover a wide range of real world applications, such as design optimization problems.

#### **2. Viscoelasticity theory and its incremental form solution**

In order to incorporate viscoelasticity into the multilayered plate theory, the onedimensional linear viscoelasticity theory can be expanded into the plane stress field, and then treated as an incremental form of solution. The generalized Maxwell model is applied to exhibit linear viscoelastic behavior. This model is composed of a parallel series of multiple Maxwell models, each of which is assembled with a serial connection of a dashpot and a spring defined by:

$$E\_r\left(t'\right) = E\_\varepsilon + \sum\_{n=1}^{N} E\_n \exp\left(-t'/\tau\_n\right) \tag{1}$$

Application of Thermo-Viscoelastic Laminated Plate Theory to Predict Warpage of Printed Circuit Boards 305

the Young's modulus. Note that Poisson's ratio is incorporated into Eq. (3) for the purpose of expanding the use of the beam theory into the plate theory. Accordingly, the Poisson's ratio in this study is treated as constant value. Further, the Poisson's ratio of viscoelastic materials in its strict sense is allowed to have its own time-dependency independent of the Young's modulus, such that the Poisson's ratio can be experimentally measured and the

Equation (3) should be converted into an incremental form so that it can be treated numerically. Taking tm as arbitrary time, and assuming that the strain varies with a constant gradient of Δε(tm)/Δtm during a time increment of Δtm= tm-tm-1, the relationship between the stress increment, Δσ(tm), and the strain increment, Δε(tm), can be represented using Eq. (1) and Eq. (3) by following the expression below (Eq. (4)). Equation (4) takes a form with respect to the strain increment; in the multilayered plate theory described in the section below, necessary equations are derived with respect to the continuity of the strain in each

> 1

*n m m n nm nmn m N*

exp /

*e nmn*

1

*t t m m <sup>n</sup>*

For modeling a printed circuit board with a multilayered plate as illustrated in Fig. 1, the

The thickness of the plate is sufficiently thin to generate a no stress component in the

 

*ν τ τ*

*n*

*A E btE*

exp *m m*

1

*a t t τ bt at τ t*

/

*n*

*t t <sup>τ</sup> <sup>ε</sup> E d<sup>τ</sup>*

*m m n m nm m n ε t σ t at σ t t*

(4)

(5)

measured Poisson's ratio becomes available for the associated analysis.

layer, where σn is the stress in the n-th term of the Maxwell model.

where

**3.1. Assumptions** 

following assumptions were made:

direction normal to the surface.

The plate is not subjected to any constraint.

<sup>1</sup> *<sup>N</sup>*

*<sup>A</sup>*

1

*ν*

*nm m*

*σ t t*

1 1

*mmm*

*ttt*

0

**3. Multilayered plate theory including viscoelasticity** 

The in-plane property of the plate is homogeneous and isotropic.

The plate will warp under the uniformly distributed temperature.

1 1

where Er(t') is called the relaxation modulus of longitudinal elasticity defining the stress relaxation keeping the strain constant. En, τn, and N are the material constants denoting the coefficient to the Prony series, the relaxation time, and the number of terms in the Maxwell model, respectively; t' is the reduced time. Where the reduction rule of time-temperature is applicable, the time-temperature conversion factor, aT(T), is obtainable from the following formula:

$$t' = \int\_0^t \frac{1}{a\_T \left(T\right)} dt\tag{2}$$

where t is the real time and T is the temperature.

Using Eq. (1) and Eq. (2), the one-dimensional behavior of the viscoelastic material with temperature dependency can be represented. Considering the plane stress state, where uniform in-plane deformation is assumed to take place, the stress-strain (σ-ε) relation can be represented with the relaxation form as follows:

$$
\sigma(t) = \frac{1}{1 - \nu} \int\_0^t E\_r \left( t' - \tau' \right) \frac{\partial \varepsilon(\tau)}{\partial \tau} d\tau \tag{3}
$$

where ν is the Poisson's ratio. This study is conducted in accordance with the plate theory of shin shells assuming the thicknesses of a circuit board are not relatively large. Based on this pre-condition, the effect from the out-of-plane shear deformation of the plate is not taken into consideration. Therefore, the bending behavior of the plate is mainly governed by only the Young's modulus. Note that Poisson's ratio is incorporated into Eq. (3) for the purpose of expanding the use of the beam theory into the plate theory. Accordingly, the Poisson's ratio in this study is treated as constant value. Further, the Poisson's ratio of viscoelastic materials in its strict sense is allowed to have its own time-dependency independent of the Young's modulus, such that the Poisson's ratio can be experimentally measured and the measured Poisson's ratio becomes available for the associated analysis.

Equation (3) should be converted into an incremental form so that it can be treated numerically. Taking tm as arbitrary time, and assuming that the strain varies with a constant gradient of Δε(tm)/Δtm during a time increment of Δtm= tm-tm-1, the relationship between the stress increment, Δσ(tm), and the strain increment, Δε(tm), can be represented using Eq. (1) and Eq. (3) by following the expression below (Eq. (4)). Equation (4) takes a form with respect to the strain increment; in the multilayered plate theory described in the section below, necessary equations are derived with respect to the continuity of the strain in each layer, where σn is the stress in the n-th term of the Maxwell model.

$$
\Delta \varepsilon \left( t\_m \right) = \frac{1}{A} \left\{ \Delta \sigma \left( t\_m \right) + \sum\_{n=1}^{N} a\_n \left( \Delta t'\_m \right) \sigma\_n \left( t\_m - \Delta t\_m \right) \right\} \tag{4}
$$

where

304 Viscoelasticity – From Theory to Biological Applications

and a spring defined by:

where t is the real time and T is the temperature.

represented with the relaxation form as follows:

formula:

of viscoelasticity characteristics has been specifically developed by the author's group, with the aim of performing measurements to obtain highly accurate data (Kobayashi, 2008). The method proposed herein was verified using the achievements from this development. Verification results are also shown in this paper. This method instantly calculates the amount of warpage and stress in a multilayered plate by giving the values of the thickness and material constants of the plate. Advantages of the proposed method may cover a wide

In order to incorporate viscoelasticity into the multilayered plate theory, the onedimensional linear viscoelasticity theory can be expanded into the plane stress field, and then treated as an incremental form of solution. The generalized Maxwell model is applied to exhibit linear viscoelastic behavior. This model is composed of a parallel series of multiple Maxwell models, each of which is assembled with a serial connection of a dashpot

> 1

where Er(t') is called the relaxation modulus of longitudinal elasticity defining the stress relaxation keeping the strain constant. En, τn, and N are the material constants denoting the coefficient to the Prony series, the relaxation time, and the number of terms in the Maxwell model, respectively; t' is the reduced time. Where the reduction rule of time-temperature is applicable, the time-temperature conversion factor, aT(T), is obtainable from the following

> 0 1 *<sup>t</sup> T t dt*

Using Eq. (1) and Eq. (2), the one-dimensional behavior of the viscoelastic material with temperature dependency can be represented. Considering the plane stress state, where uniform in-plane deformation is assumed to take place, the stress-strain (σ-ε) relation can be

> 0

*σ t Et τ dτ ν τ*

where ν is the Poisson's ratio. This study is conducted in accordance with the plate theory of shin shells assuming the thicknesses of a circuit board are not relatively large. Based on this pre-condition, the effect from the out-of-plane shear deformation of the plate is not taken into consideration. Therefore, the bending behavior of the plate is mainly governed by only

*ε τ*

(3)

*t r*

1 1

*N ren n n Et E E t τ* 

exp /

(1)

*a T* (2)

range of real world applications, such as design optimization problems.

**2. Viscoelasticity theory and its incremental form solution** 

$$\begin{cases} \Delta t\_m' = t\_m' - t\_{m-1}' \\ \quad a\_n \left(\Delta t\_m'\right) = 1 - \exp\left(-\Delta t\_m' / \tau\_n\right) \\ b\_n \left(\Delta t\_m'\right) = a\_n \left(\Delta t\_m'\right) \cdot \tau\_n / \Delta t\_m' \\ \quad A = \frac{1}{1-\nu} \left\{ E\_e + \sum\_{n=1}^N b\_n \left(\Delta t\_m'\right) \cdot E\_n \right\} \\ \sigma\_n \left(t\_m - \Delta t\_m\right) \\ \quad = \frac{1}{1-\nu} \int\_0^{t\_m - \Delta t\_m} E\_n \exp\left\{\frac{-t\_m' + \Delta t\_m' + \tau'}{\tau\_n}\right\} \frac{\partial \varepsilon}{\partial \tau} d\tau \end{cases} \tag{5}$$

#### **3. Multilayered plate theory including viscoelasticity**

#### **3.1. Assumptions**

For modeling a printed circuit board with a multilayered plate as illustrated in Fig. 1, the following assumptions were made:


The bending moment to deform the plate convexly is defined as positive; the deflection along this direction is also defined as positive. Hi and νi in the figure denote the plate thickness and the Poisson's ratio, respectively. The curvature induced under temperature variation becomes identical in all in-plane directions of the xy plane, since isotropicity and non-constraint are assumed. Therefore, the warpage of the multilayered plate can be calculated only on the deformation within the xz plane, as shown in Fig. 2.

Application of Thermo-Viscoelastic Laminated Plate Theory to Predict Warpage of Printed Circuit Boards 307

0 ( ) ( ) () *i i*

0 ( ) ( ) ( ) () ( ) *<sup>i</sup> i m i m*

*T T α T*

*i*

(7)

*im i im Pt H B σ t* (9)

,

*in m m i in m m* , , *P t t HB σ t t* (11)

(12)

,

(13)

(10)

(8)

where T1 represents the initial temperature in the analysis. The temperature-dependent thermal expansion coefficient is the average thermal expansion coefficient based on a reference temperature, T0. The second term in Eq. (6) is necessary in order to prevent strain at the initial temperature from being generated. In order to express it as an incremental

Therefore, the thermal-strain increment in a given temperature change, ΔT (tm), can be

*ε t T T α T Tt*

On each layer of the multilayered plate illustrated in Fig. 2, generation of in-plane force and bending moment are considered. Taking the in-plane force in the i-th layer as Pi(tm), the incremental force can be expressed with Eq. (9). Hi and B are the thickness and width of each

Introducing the effect from viscoelasticity using Eq. (4), as previously derived, Eq. (8) can be

 1

*i m i m m in m m*

The incremental strain due to the bending moment in each layer can be similarly expressed

 3 12 *i i m i m i H B M t σ t z*

*N*

 <sup>=</sup> <sup>3</sup> 1

*i m i m m in m m*

*ε t Mt a t M t t*

*ε t Pt a t P t t*

*i i n*

*H BA*

by Eq. (13); *<sup>i</sup> z* is the distance measured from the neutral plane of the i-th layer.

*i i n*

*H BA*

*N*

*dε dα T*

*dT dT* 

*dα T*

*dT* 

layer, respectively; σ''i(tm) indicates the stress due to the in-plane force.

1

12

*i*

*z*

form, Eq. (6) is differentiated by T.

obtained from the following equation:

b. Strain due to in-plane force in the plate

written in the incremental form as follows;

c. Strain due to bending moment

where

**Figure 1.** Multilayered plate

**Figure 2.** Deformation in xz plane

In multilayered plate theory, global warpage is calculated based on the assumption that the strain in each layer is independent and the respective interface is continuous; the strain generated in each of the layers is composed of three components:


Each strain component is classified as follows:

a. Thermal strain

When the temperature-dependent thermal expansion coefficient of the i-th layer at an arbitrary temperature, T, is expressed as αi(T), the thermal strain is expressed as the following equation:

$$
\varepsilon\_i' = \alpha\_i(T) \cdot (T - T\_0) - \alpha\_i(T\_1) \cdot (T\_1 - T\_0) \tag{6}
$$

where T1 represents the initial temperature in the analysis. The temperature-dependent thermal expansion coefficient is the average thermal expansion coefficient based on a reference temperature, T0. The second term in Eq. (6) is necessary in order to prevent strain at the initial temperature from being generated. In order to express it as an incremental form, Eq. (6) is differentiated by T.

$$\frac{d\varepsilon\_i'}{dT} = \frac{d\alpha\_i(T)}{dT} \cdot (T - T\_0) + \alpha\_i(T) \tag{7}$$

Therefore, the thermal-strain increment in a given temperature change, ΔT (tm), can be obtained from the following equation:

$$
\Delta \varepsilon\_i'(t\_m) = \left\{ \frac{d\alpha\_i(T)}{dT} \cdot (T - T\_0) + \alpha\_i(T) \right\} \Delta T(t\_m) \tag{8}
$$

#### b. Strain due to in-plane force in the plate

On each layer of the multilayered plate illustrated in Fig. 2, generation of in-plane force and bending moment are considered. Taking the in-plane force in the i-th layer as Pi(tm), the incremental force can be expressed with Eq. (9). Hi and B are the thickness and width of each layer, respectively; σ''i(tm) indicates the stress due to the in-plane force.

$$
\Delta P\_i \left( t\_m \right) = H\_i \cdot B \cdot \Delta \sigma\_i'' \left( t\_m \right) \tag{9}
$$

Introducing the effect from viscoelasticity using Eq. (4), as previously derived, Eq. (8) can be written in the incremental form as follows;

$$
\Delta \varepsilon\_i''(t\_m) = \frac{1}{H\_{\dot{i}} \cdot B \cdot A\_{\dot{i}}} \left\{ \Delta P\_i \left( t\_m \right) + \sum\_{n=1}^{N} a \left( \Delta t\_m' \right) \cdot P\_{i,n} \left( t\_m - \Delta t\_m \right) \right\} \tag{10}
$$

where

306 Viscoelasticity – From Theory to Biological Applications

**Figure 1.** Multilayered plate

**Figure 2.** Deformation in xz plane

b. Strain due to in-plane force in the plate c. Strain due to bending moment on the plate Each strain component is classified as follows:

a. Thermal strain

a. Thermal strain

following equation:

The bending moment to deform the plate convexly is defined as positive; the deflection along this direction is also defined as positive. Hi and νi in the figure denote the plate thickness and the Poisson's ratio, respectively. The curvature induced under temperature variation becomes identical in all in-plane directions of the xy plane, since isotropicity and non-constraint are assumed. Therefore, the warpage of the multilayered plate can be

> H1 H2

HI

P1 P2

P MI <sup>I</sup>

M2

M1

1 2 I

In multilayered plate theory, global warpage is calculated based on the assumption that the strain in each layer is independent and the respective interface is continuous; the strain

x

z

When the temperature-dependent thermal expansion coefficient of the i-th layer at an arbitrary temperature, T, is expressed as αi(T), the thermal strain is expressed as the

0 1 10 ( )( ) ( )( ) *i i <sup>i</sup> ε α T TT α T TT* (6)

generated in each of the layers is composed of three components:

R1 R2

calculated only on the deformation within the xz plane, as shown in Fig. 2.

Layer 2 Layer 1

Layer I :

RI

$$P\_{i,n} \left( t\_m - \Delta t\_m \right) = H\_i \cdot B \cdot \sigma\_{i,n}'' \left( t\_m - \Delta t\_m \right) \tag{11}$$

#### c. Strain due to bending moment

The incremental strain due to the bending moment in each layer can be similarly expressed by Eq. (13); *<sup>i</sup> z* is the distance measured from the neutral plane of the i-th layer.

$$
\Delta \mathcal{M}\_i \left( t\_m \right) = \frac{H^{\overline{\mathcal{G}}\_i \cdot \mathcal{B}}}{12 \overline{\varpi}\_i} \cdot \Delta \sigma\_i'''(t\_m) \tag{12}
$$

$$
\Delta \varepsilon\_i'''(t\_m) = \frac{12\overline{z}\_i}{H^3\_{\,\,i} \cdot B \cdot A\_{\,i}} \left\{ \Delta M\_{\,i} \left( t\_m \right) + \sum\_{n=1}^{N} a \left( \Delta t'\_m \right) \cdot M\_{\,i,n} \left( t\_m - \Delta t\_m \right) \right\} \tag{13}
$$

where

$$M\_{i,n} \left( t\_m - \Delta t\_m \right) = \frac{H^{\vec{\mathcal{Y}}}\_{i} \cdot \mathcal{B}}{12\overline{z}\_i} \cdot \sigma'''\_{i,n} \left( t\_m - \Delta t\_m \right) \tag{14}$$

The strain increment, Δε'''i(tm), can be represented as Eq. (15).

$$
\Delta \varepsilon\_i'''(t\_m) = \overline{z}\_i \cdot \Delta \mathbb{C}\_i(t\_m) \tag{15}
$$

Application of Thermo-Viscoelastic Laminated Plate Theory to Predict Warpage of Printed Circuit Boards 309

3

By way of the above-mentioned procedure, in-plane forces and curvatures are obtained. Using these results, stress and deflection can be derived following the steps below. Firstly, the stress generated in each layer is calculated as the sum of the in-plane stress and the

> ( ) <sup>12</sup> ( ) ( ) *i m i m i m i i P t z σ t M t H B H B*

As this model is assumed to be isotropic and unconstrained, the curvature must be identical in all directions within the xy plane. As per Fig. 3, L is taken as the distance from the center to the corner of each layer; the cross-section of each layer is assumed to have the shape shown in Fig. 4, and θ is taken as the slope at the tip of each layer. Accordingly, the

Assuming θ is relatively small and in-plane elongation is also small, resulting in L = L', the

 <sup>2</sup> 2 *m*

L

L'

R

*L Ct*

*m*

**3.3. Derivation of stress and deflection** 

maximum tip deflection, δ(tm), is obtained as follows:

following expression can be obtained:

**Figure 3.** Definition of L

**Figure 4.** Deformation of the cross section

bending stress, as follows:

1 2 () () () () *C t C t C t Ct m m Im m* (20)

(21)

*δtm m* 1 cos / *θ C t* (22)

*<sup>δ</sup> <sup>t</sup>* (23)

x

x'

y

L x'

z

z

where ΔCi(tm) is the increment of the curvature,

$$C\_i(t\_m) = \frac{1}{R\_i(t\_m)}\tag{16}$$

#### **3.2. Strain continuity and equilibrium equation**

The global in-plane strain generated at each layer is expressed as the sum of the abovementioned strain components a. through c. As this global strain must be kept continuous across the interface on each layer, the following expression can be written:

$$\begin{aligned} &a\_i \Delta T \left( t\_m \right) + \frac{1}{H\_i \cdot B \cdot A\_i} \left\{ \Delta P\_i \left( t\_m \right) + \sum\_{n=1}^N a \left( \Delta t\_m' \right) \cdot P\_{i,n} \left( t\_m - \Delta t\_m \right) \right\} - \frac{H\_i}{2} \cdot \Delta C\_i \left( t\_m \right) \\ &= a\_{i+1} \Delta T \left( t\_m \right) + \frac{1}{H\_{i+1} \cdot B \cdot A\_{i+1}} \left\{ \Delta P\_{i+1} \left( t\_m \right) + \sum\_{n=1}^N a \left( \Delta t\_m' \right) \cdot P\_{i+1,n} \left( t\_m - \Delta t\_m \right) \right\} \\ &+ \frac{H\_{i+1}}{2} \cdot \Delta C\_{i+1} \left( t\_m \right) \qquad \qquad \qquad i = 1 \ to \; I - 1 \end{aligned} \tag{17}$$

The applied load must also satisfy the following equilibrium equations:

$$\sum\_{i=1}^{I} \Delta P\_i(t\_m) = 0 \tag{18}$$

$$\begin{aligned} \sum\_{i=1}^{I} \Delta M\_i \left( t\_m \right) + \Delta P\_1 \left( t\_m \right) \cdot \left( \overline{z}' - \frac{H\_1}{2} \right) + \Delta P\_2 \left( t\_m \right) \cdot \left( \overline{z}' - H\_1 - \frac{H\_2}{2} \right) \\ + \dots + \Delta P\_I \left( t\_m \right) \cdot \left( \overline{z}' - H\_1 - H\_2 - \dots - \frac{H\_I}{2} \right) = 0 \end{aligned} \tag{19}$$

Assuming the thickness of each layer is negligibly small compared with its resultant curvature, the respective curvature on each layer is identical as Eq. (20). Therefore ΔPi (tm), ΔMi(tm) and ΔCi(tm) are obtainable by simultaneously resolving Eq. (17), Eq. (18), and Eq. (19).

Application of Thermo-Viscoelastic Laminated Plate Theory to Predict Warpage of Printed Circuit Boards 309

$$
\Delta \mathbb{C}\_1(t\_m) \approx \Delta \mathbb{C}\_2(t\_m) \approx \dots \approx \Delta \mathbb{C}\_I(t\_m) = \Delta \mathbb{C}(t\_m) \tag{20}
$$

#### **3.3. Derivation of stress and deflection**

308 Viscoelasticity – From Theory to Biological Applications

 3

(14)

*im i im ε t z Ct* (15)

*R t* (16)

2

(17)

(19)

*H*

*i*

,

(18)

1 2

2 2

*H H*

0

*in m m in m m i H B Mt t σ t t z*

> <sup>1</sup> ( ) ( ) *i m i m*

The global in-plane strain generated at each layer is expressed as the sum of the abovementioned strain components a. through c. As this global strain must be kept continuous

,

*N*

*i m*

*P t*

0

1 21

2

*H*

*I*

1

*α T t Pt a t P t t C t*

*i m i m m i nm m*

*α T t P t at P t t*

*i m i m m in m m i m*

*N*

1 1

1 11 1 1 1

1

*M t Pt z P t z H*

*im m m*

1 2

Assuming the thickness of each layer is negligibly small compared with its resultant curvature, the respective curvature on each layer is identical as Eq. (20). Therefore ΔPi (tm), ΔMi(tm) and ΔCi(tm) are obtainable by simultaneously resolving Eq. (17), Eq. (18),

*i*

*I*

*i i n*

 

*C t*

across the interface on each layer, the following expression can be written:

<sup>12</sup> , , *<sup>i</sup>*

The strain increment, Δε'''i(tm), can be represented as Eq. (15).

where ΔCi(tm) is the increment of the curvature,

**3.2. Strain continuity and equilibrium equation** 

*Pt z H H*

*I m*

1

*H BA*

1

*C t i to I*

The applied load must also satisfy the following equilibrium equations:

*H BA*

*i i n*

1

*i m*

1

1

*i*

and Eq. (19).

*I*

2

*i*

*H*

where

By way of the above-mentioned procedure, in-plane forces and curvatures are obtained. Using these results, stress and deflection can be derived following the steps below. Firstly, the stress generated in each layer is calculated as the sum of the in-plane stress and the bending stress, as follows:

$$
\sigma\_i(t\_m) = \frac{P\_i(t\_m)}{H\_i B} + \frac{12z}{H\_i^3 B} M\_i(t\_m) \tag{21}
$$

As this model is assumed to be isotropic and unconstrained, the curvature must be identical in all directions within the xy plane. As per Fig. 3, L is taken as the distance from the center to the corner of each layer; the cross-section of each layer is assumed to have the shape shown in Fig. 4, and θ is taken as the slope at the tip of each layer. Accordingly, the maximum tip deflection, δ(tm), is obtained as follows:

$$\left(\delta\left(t\_m\right) = \left(1 - \cos\Theta\right) / \mathcal{C}\left(t\_m\right)\right.\tag{22}$$

Assuming θ is relatively small and in-plane elongation is also small, resulting in L = L', the following expression can be obtained:

**Figure 3.** Definition of L

**Figure 4.** Deformation of the cross section

#### **4. Measurement of viscoelastic material properties**

Aside from the fact that thermal stress analysis taking viscoelasticity into account has not been adequately carried out to date, one must recognize that it is difficult, in practice, to obtain sufficient accuracy in such experimental measurements. Accordingly, any data measured in such experiments is not accurate enough to relate to the viscoelastic numerical model. Viscoelastic materials exhibit very sensitive temperature dependency, particularly in the vicinity of the glass transition point. Thus, intricate temperature control is required throughout the entire duration of the measurement operation. In addition, the time-domain constants obtained usually span a wide range of digits, in the range of 20 to 30. This leads to the additional task of determining desirable factors, which may prove difficult unless advanced optimization techniques are applied.

This paper also presents a test case for obtaining the characteristics of epoxy resin material. A device for measuring dynamic viscoelasticity, RSA III (TA Instruments), was used. Dynamic viscoelasticity characteristics were measured for angular velocities 3.16, 10, 31.6, and 100 rad/sec with an ascending temperature rate of 2 °C/min, in the temperature range of -40 to 60 °C. As shown in Fig. 5 and Fig. 6, the storage modules, E', and the loss modulus, E'', of the epoxy resin material were measured. This measurement device is equipped with a temperature-controlled oven with a solid structure and a large volume flow rate, providing excellent performance in temperature control.

Figure 7 shows the master curve, representing the relationship between the storage/loss modulus and reduced angular velocity, for the reference temperature, TR. To create the master curve, the WLF formula as shown below was applied for the temperature-time reduction.

$$\log\left(a\left(T\right)\right) = -\frac{C\_1\left(T - T\_R\right)}{C\_2 + \left(T - T\_R\right)}\tag{24}$$

$$T\_R = T\_g + 50 \qquad C\_1 = 8.86 \qquad C\_2 = 101.6 \quad T\_g = -16^\circ \text{C}$$

Application of Thermo-Viscoelastic Laminated Plate Theory to Predict Warpage of Printed Circuit Boards 311

100rad/sec 31.6rad/sec 10rad/sec 3.16rad/sec



1.E-02 1.E+02 1.E+06 1.E+10 1.E+14 Reduction angular velocity ω' [rad/sec]

E'(Experiment) E''(Experiment) E'(Calculated) E''(Calculated)

100rad/sec 31.6rad/sec 10rad/sec 3.16rad/sec

**Figure 5.** Storage modulus of epoxy resin

1.0E+00

1.0E-01

TR=34℃

1.0E+00

1.0E+01

Loss modulus E'' [MPa]

1.0E+02

1.0E+03

1.0E+01

1.0E+02

Storage modulus E' [MPa]

1.0E+03

1.0E+04

**Figure 6.** Loss modulus of epoxy resin

**Figure 7.** Master curve for Storage/Loss modulus

1.E-01

1.E+00

1.E+01

Storage/Loss modulus E' E'' [MPa]

1.E+02

1.E+03

1.E+04

In this study, a simple optimization technique was employed in which generally recommended values for C: and C were used, and T\_8 was estimated using the quasiconcave method (Kobayashi, 2008). As a result, a single smooth master curve could be used to cover the wide range of angular velocities. The coefficients for the Maxwell model were obtained from the master curve by the optimization approach so that the relaxation curve from Eq. (1) could be calculated numerically. The result is shown in Fig. 8. The relationship between time-domain constants and frequency-domain constants is expressed by Eq. (25). Using this formula, the numerical model that gives a close approximation of the actual measurement data can be obtained, as seen in Fig. 8.

$$E' = E\_e + \sum\_{n=1}^{N} \frac{\tau\_n^2 \omega'^2}{1 + \tau\_n^2 \omega'^2} E\_n \quad E'' = \sum\_{n=1}^{N} \frac{\tau\_n \omega'}{1 + \tau\_n^2 \omega'^2} E\_n \tag{25}$$

These procedures for identification have been organized in Excel; please refer to the Appendix.

**Figure 5.** Storage modulus of epoxy resin

advanced optimization techniques are applied.

excellent performance in temperature control.

measurement data can be obtained, as seen in Fig. 8.

**4. Measurement of viscoelastic material properties** 

Aside from the fact that thermal stress analysis taking viscoelasticity into account has not been adequately carried out to date, one must recognize that it is difficult, in practice, to obtain sufficient accuracy in such experimental measurements. Accordingly, any data measured in such experiments is not accurate enough to relate to the viscoelastic numerical model. Viscoelastic materials exhibit very sensitive temperature dependency, particularly in the vicinity of the glass transition point. Thus, intricate temperature control is required throughout the entire duration of the measurement operation. In addition, the time-domain constants obtained usually span a wide range of digits, in the range of 20 to 30. This leads to the additional task of determining desirable factors, which may prove difficult unless

This paper also presents a test case for obtaining the characteristics of epoxy resin material. A device for measuring dynamic viscoelasticity, RSA III (TA Instruments), was used. Dynamic viscoelasticity characteristics were measured for angular velocities 3.16, 10, 31.6, and 100 rad/sec with an ascending temperature rate of 2 °C/min, in the temperature range of -40 to 60 °C. As shown in Fig. 5 and Fig. 6, the storage modules, E', and the loss modulus, E'', of the epoxy resin material were measured. This measurement device is equipped with a temperature-controlled oven with a solid structure and a large volume flow rate, providing

Figure 7 shows the master curve, representing the relationship between the storage/loss modulus and reduced angular velocity, for the reference temperature, TR. To create the master curve, the WLF formula as shown below was applied for the temperature-time reduction.

1 2 *R g* 50 8 86 101 6 16 . . C *<sup>g</sup> TT C C T*

In this study, a simple optimization technique was employed in which generally recommended values for C1 and C2 were used, and Tg was estimated using the quasi-Newton method (Kobayashi, 2008). As a result, a single smooth master curve could be used to cover the wide range of angular velocities. The coefficients for the Maxwell model were obtained from the master curve by the optimization approach so that the relaxation curve from Eq. (1) could be calculated numerically. The result is shown in Fig. 8. The relationship between time-domain constants and frequency-domain constants is expressed by Eq. (25). Using this formula, the numerical model that gives a close approximation of the actual

*n n en n n n n n τ ω τ ω E E E E <sup>E</sup>*

*τ ω τ ω*

1 1 1 1

These procedures for identification have been organized in Excel; please refer to the Appendix.

*N N*

2 2 2 2

(25)

2 2

*a T*

1 2 log *<sup>R</sup>*

*CT T*

*C TT*

*R*

(24)

**Figure 6.** Loss modulus of epoxy resin

**Figure 7.** Master curve for Storage/Loss modulus

**Figure 8.** Master curve for relaxation modulus

#### **5. Numerical verification by FEM**

To verify the calculation method proposed in this study, a three-layered plate—40 mm in length and 20 mm in width (Fig. 9), consisting of epoxy resin with 0.5 mm thickness (viscoelastic, Tg = 105 °C), FR-4 substrate with 0.5 mm thickness (viscoelastic, Tg = 120 °C), and aluminum with 0.1 mm thickness (elastic)—was analyzed. The initial temperature of the laminated plate was set at 180 °C, and it was cooled to 25 °C within 2,000 sec, as shown in the temperature history (Fig. 10). Because the laminated plate was gradually cooled, its inside temperature was assumed to be uniform.

Figure 11 shows the master curves of the relaxation moduli of the epoxy and FR-4 substrates. The temperature-dependent viscoelasticity was assumed to be in accordance with the Arrhenius formula, as expressed by Eq. (26).

$$\log\left(a\left(T\right)\right) = -\frac{\Delta H}{2.303R}\left(\frac{1}{T} - \frac{1}{T\_0}\right) \tag{26}$$

Application of Thermo-Viscoelastic Laminated Plate Theory to Predict Warpage of Printed Circuit Boards 313

0 500 1000 1500 2000 Time [sec]

The values of the coefficients for Epoxy resin shown in Eq. (26) were as follows: ΔH1 = 2.4420 × 105 (before T0); ΔH2 = 3.0480 × 105 (after T0); and T0 = 97.7 °C. Those for the FR-4 substrate were as follows: ΔH1 = 8.8300 × 104 (before T0); ΔH2 = 4.3220 × 105 (after T0); and T0 = 113.2 °C.

Please note that the material constants of the epoxy resin used for this verification analysis differ from those expressed in the measurement example in the preceding section. Figure 12 shows the thermal expansion coefficients of the epoxy and FR-4 substrates. The aluminum that composed Layer 3 was an elastic body with the Young's modulus E = 70,000 MPa and

(a) FR-4 substrate (b) Epoxy resin

1.E+0

1.E-7 1.E-4 1.E-1 1.E+2 1.E+5 Time [sec]

1.E+1

1.E+2

Relaxation modulus [MPa]

1.E+3

1.E+4

For comparison with the multilayered plate theory developed in this study, an FE analysis of the same model was performed using the shell elements of Abaqus ver. 6.8. Figure 13 shows the deformation of the plate obtained by the FE analysis. Figure 14 shows the

**Figure 10.** Applied temperature history

273

313

353

Temperature

[K]

393

433

473

**Figure 11.** Relaxation modulus

1.E+2

1.E+3

Relaxation modulus [MPa]

1.E+4

1.E+5

the thermal expansion coefficient α 3 = 23.2 × 10-6/°C.

1.E-7 1.E-4 1.E-1 1.E+2 1.E+5 Time [sec]

**Figure 9.** Three-layered plate model

The values of the coefficients for Epoxy resin shown in Eq. (26) were as follows: ΔH1 = 2.4420 × 105 (before T0); ΔH2 = 3.0480 × 105 (after T0); and T0 = 97.7 °C. Those for the FR-4 substrate were as follows: ΔH1 = 8.8300 × 104 (before T0); ΔH2 = 4.3220 × 105 (after T0); and T0 = 113.2 °C.

**Figure 10.** Applied temperature history

312 Viscoelasticity – From Theory to Biological Applications

**Figure 8.** Master curve for relaxation modulus

1.0E+00

1.0E+01

1.0E+02

Relaxation modulus Er(t) [MPa]

1.0E+03

1.0E+04

**5. Numerical verification by FEM** 

inside temperature was assumed to be uniform.

20 mm

**Figure 9.** Three-layered plate model

with the Arrhenius formula, as expressed by Eq. (26).

To verify the calculation method proposed in this study, a three-layered plate—40 mm in length and 20 mm in width (Fig. 9), consisting of epoxy resin with 0.5 mm thickness (viscoelastic, Tg = 105 °C), FR-4 substrate with 0.5 mm thickness (viscoelastic, Tg = 120 °C), and aluminum with 0.1 mm thickness (elastic)—was analyzed. The initial temperature of the laminated plate was set at 180 °C, and it was cooled to 25 °C within 2,000 sec, as shown in the temperature history (Fig. 10). Because the laminated plate was gradually cooled, its

Curve\_fit

1.0E-11 1.0E-08 1.0E-05 1.0E-02 1.0E+01 Time t [sec]

TR=34℃

Figure 11 shows the master curves of the relaxation moduli of the epoxy and FR-4 substrates. The temperature-dependent viscoelasticity was assumed to be in accordance

40 mm

0

(26)

0.5 mm 0.5 mm 0.1 mm

 = 0.3 = 0.35 = 0.33

1 1

*RT T* 

Layer1 FR-4 substrate Layer2 Epoxy resin Layer3 Aluminum

2 303 log . *<sup>H</sup> a T*

**Figure 11.** Relaxation modulus

Please note that the material constants of the epoxy resin used for this verification analysis differ from those expressed in the measurement example in the preceding section. Figure 12 shows the thermal expansion coefficients of the epoxy and FR-4 substrates. The aluminum that composed Layer 3 was an elastic body with the Young's modulus E = 70,000 MPa and the thermal expansion coefficient α 3 = 23.2 × 10-6/°C.

For comparison with the multilayered plate theory developed in this study, an FE analysis of the same model was performed using the shell elements of Abaqus ver. 6.8. Figure 13 shows the deformation of the plate obtained by the FE analysis. Figure 14 shows the relationship between the deformation and temperature. Figure 15 shows the relationship between stress and temperature for each layer. As shown in Figs. 14 and 15, the deformation and stress of the laminated plate obtained by the multilayered plate theory were confirmed to be identical to those obtained by the FE analysis.

Application of Thermo-Viscoelastic Laminated Plate Theory to Predict Warpage of Printed Circuit Boards 315

FEM Plate theory

FEM

FEM Plate theory

Plate theory

(a) FR-4 substrate (Upper surface of Layer 1)

433 393 353 313 273 Temperature [K]

0


> -80 -70 -60 -50 -40 -30 -20 -10 0

Stress [MPa]

Stress [MPa]

5

10

Stress [MPa]

15

20

25

(b) Epoxy resin (Interface between Layer 1 and Layer 2)

433 393 353 313 273 Temperature [K]

(c)Aluminum (Bottom surface of Layer 3)

433 393 353 313 273 Temperature [K]

**Figure 15.** History of stress

**Figure 12.** Thermal Expansion curves

**Figure 13.** Deformation of plate (FEM)

**Figure 14.** History of plate deflection

(c)Aluminum (Bottom surface of Layer 3)

**Figure 15.** History of stress

**Figure 12.** Thermal Expansion curves

0.E+00

1.E-03

Thermal strain [ - ]

2.E-03

3.E-03

**Figure 13.** Deformation of plate (FEM)

**Figure 14.** History of plate deflection


Deformation [mm]

273313353393433 Temperature [K]

FEM Plate theory

to be identical to those obtained by the FE analysis.

9.67E+06 [ /K ]

273 373 473 573 Temperature [K]

1.35E+05 [ /K ]

1.19E+06

[/K]

relationship between the deformation and temperature. Figure 15 shows the relationship between stress and temperature for each layer. As shown in Figs. 14 and 15, the deformation and stress of the laminated plate obtained by the multilayered plate theory were confirmed

(a) FR-4 substrate (b) Epoxy resin

0.E+0

273 373 473 573 Temperature [K]

1.80E+04 [ /K ]

8.78E+05 [ /K ]

2.09E+04 [ /K ]

1.E-2

Thermal strain [ - ]

2.E-2

3.E-2
