three-dimensional fem simulations of thermomechanical stresses in 1.55 lm laser modules

8/13/2019 Three-Dimensional FEM Simulations of Thermomechanical Stresses in 1.55 Lm Laser Modules

http://slidepdf.com/reader/full/three-dimensional-fem-simulations-of-thermomechanical-stresses-in-155-lm-laser 1/12

Three-dimensional FEM simulations of

thermomechanical stresses in 1.55 lm Laser modules

Y. Deshayes a,*, L. Bechou a, J.Y. Deletage a, F. Verdier a, Y. Danto a,D. Laffitte b, J.L. Goudard b

a IXL Laboratory, ENSEIRB, University of Bordeaux I, 351 Cours de la Lib eeration, 33405 Talence Cedex, Franceb ALCATEL Optronics, Route de Villejust, 91625 Nozay, France

Received 8 November 2002; received in revised form 28 March 2003

Abstract

The purpose of this study is to present three-dimensional simulations using finite element method (FEM) of ther-

momechanical stresses and strains in 1550 nm Laser modules induced by Nd:YAG crystal Laser welds and thermal

cycles on two main sub-assemblies: Laser submount and pigtail. Non-linear FEM computations, taking into account of

experimental rðeÞ measured curves, show that Laser welding process can induce high level of strains in columns of the

Laser platform, bearing the Laser diode, responsible of an optical axis shift and a gradual drop of the optical power in

relation with relaxation of accumulated stresses in the sub-assembly. In the case of thermal cycles, stresses can occur on

elements sensitive to coefficient of thermal expansion mismatches such as solder joint between the Laser platform and

thermoelectric cooler and as fiber glued into the pigtail leading to crack propagation with sudden drop of optical power.

The main objective of the paper is to evaluate thermomechanical sensitivity and critical zones of the Laser module in

order to improve mechanical stability after Laser weld and reach qualification standards requirements without failures.Experimental analyses were also conducted to correlate simulation results and monitor the output optical power of

Laser modules after 500 thermal cycles ()40 C/+85 C VRT).

2003 Elsevier Ltd. All rights reserved.

1. Introduction

The development of high bandwidth single mode fiber

optics communication technologies coupled with the

availability of emitter components for wavelength mul-

tiplexing have created a revolution in the transmission

technology during the last ten years. These perfor-mances can be reached by packaging interface and

control circuits with the optical chips leading to the

concept of high reliable technically-advanced Laser

modules which can be used by end-users without the

need for a detailed knowledge of optoelectronics. Low

cost, low consumption, hermetical and highly efficient

optical coupling between the Laser diode and the single-

mode fiber associated to a mechanical stability are one

of the key issues for such systems. Packaging of opto-

electronics components requires the solution of optical,

mechanical and electrical problems. These problems are

often highly interactive and the stability of optoelec-

tronic devices is an essential factor to ensure high

bandwidth data transmission, acceptable bit-error rateand develop reliable solutions. Photonic systems involve

both a mechanical alignment and a direct attachment or

not between the light emitter and the optical fiber [1,2].

To ensure high coupling efficiency, the mechanical sta-

bility of the optical elements is critical. Three primary

techniques have been developed to align and attach the

light-emitter to the optical fiber associated with different

package configurations: solder, epoxies and Nd:YAG

Laser welding [3]. It has been already demonstrated that

Nd:YAG Laser welding technique is the most effective

method to satisfy performances criteria previously de-

scribed. Due to inherent advantages, a growing number

Microelectronics Reliability 43 (2003) 1125–1136

www.elsevier.com/locate/microrel

* Corresponding author. Tel.: +33-5568-46547/42858; fax:

+33-5563-71545.

E-mail address: [email protected] (Y. Deshayes).

0026-2714/03/$ - see front matter 2003 Elsevier Ltd. All rights reserved.

doi:10.1016/S0026-2714(03)00099-4

http://mail%20to:%[email protected]/

http://mail%20to:%[email protected]/



of communication systems integrators are requesting

Laser welded packages for their end-users. The me-

chanical stability requires tolerances of less than 1 lm to

avoid a power change lower than 10%, which must be

consistent during the lifetime of the module and across

the temperature range. In recent papers, possible causes

of power changes in an optoelectronic transmitter have

been listed [4,5]:

• mechanical stresses or defects in solders,

• degradation of the thermocooler,

• relaxation of thermomechanical stresses appearing

during module assembly and mismatched CTE of

package elements,

• breakage or slow movements of the fiber glued in the

ferule, . . .

Standard qualification procedures, in particular

power drift measurement, must be conducted to validatethe system with respect to tolerances through tempera-

ture cycling or storage temperature characterizing the

limits and the margins of the technology. Actual stan-

dards tend to be 500 cycles in the temperature range )40

C/+85 C without failures [4]. These ageing tests are

generally realized in order to evaluate all the parameters

in relation with failure distribution but more than one

hundred modules must be performed during several

thousands hours mixing different life test conditions.

These results can allow to determine the robustness of

the technology but due to a high complexity of the

package, cannot give accurate information on the failure

origin, which is responsible of the optical power drift.

To face qualification challenges, new processes are now

being proposed focusing on reliability concerns at the

early stage of the product development. In this ap-

proach, the qualification is considered as a long-term

process rather than a final exam at the end of the de-

velopment [6]. Based on environmental and functional

specifications, the product development can start with a

technical risk analysis phase. This phase aims at point-

ing out the major risks for a given product design. These

risks are then assessed and lowered through dedicated

action plans performed on representative test units or

complete products. In this case, physical simulations(thermal, mechanical and thermomechanical) can also

be used to give complementary information and to as-

sess the risk criticality [7].

The purpose of this study deals with results achieved

from nonlinear thermomechanical simulations using

finite-element method (FEM) of a direct modulation 1.55

lm Laser module for telecommunication applications.

In this paper, two main parts will be developed:

• calculations of stresses and strains in the critical

zones based on both technological and thermo-

mechanical analyses of the whole Laser module (con-

struction design, dissimilar materials, mismatched

CTE, . . .),

• relation between calculated strains and optical mis-

alignment responsible of gradual power drift.

Experimental failure analyses will be also conducted

to validate thermomechanical simulations, focused in

particular on Laser welded joints in order to propose

assumptions for accumulated strains relaxation phe-

nomenon. In this context, both thermal, electrical and

thermomechanical simulations on the package must be

realized using an original approach based on multi-

physics computations of ANSYS software, in particular

for electro-thermal Nd:YAG Laser modeling [8]. First, a

description of the Laser module is given and 3D-FEM

models of each sub-assembly are presented taking into

account of the different materials characteristics versus

temperature and external loads related to manufacturing

steps. The last section gives simulation results of themain sub-assemblies of the Laser module concluding on

thermomechanical sensitivity of critical zones and the

impact on a possible optical axis misalignment.

2. Description of the Laser module design

Semiconductor Laser package bodies are typically

either cylindrical-type or box-type styles. For lightwave

communication systems, box-type bodies are widely

used and in particular Dual-In-Line or Butterfly pack-

ages with fiber pigtails. Our study is focused on 1.55 lm

Butterfly package Laser module and a technological

description is presented in Fig. 1. The Laser diode (Dis-

tributed FeedBack Laser diode InP/InGaAsP) emitting

at 1.55 lm is soldered with a AuSn solder joint (8 lm)

on the Laser sub-mount (AlN), and then the sub-mount

is attached to the Laser platform (composed by a sub-

mount and 2 columns bearing the lens holder) in Kovar

by a SnSb solder joint (8 lm). Lens 1, used to colli-

mate the Laser beam from the Laser diode, and the

isolator are welded to a lens holder (Kovar) by means of

Sub-assembly 1

Fig. 1. Technological description of the Laser module with the

two main sub-assemblies.

1126 Y. Deshayes et al. / Microelectronics Reliability 43 (2003) 1125–1136



Nd:YAG Laser welding process. The Laser platform

and the lens holder are also welded corresponding to the

sub-assembly 1. This last element is then attached to the

thermoelectric cooler and mounted with a SnPbAg sol-

der joint (10 lm) in a Butterfly-type package (Kovar).

The sub-assembly 2 is composed of a second lens, used

to focalized Laser beam into the fiber core, glued with an

adhesive material into a circular ferule (Zirconia/

Kovar). Finally, the sub-assembly 2 is Nd:YAG Laser

welded to the Butterfly-type package providing a com-

plete hermiticity for the system.

As an alternative to the common adhesives or solders

used in the joining process, Nd:YAG Laser welding

offers a number of attractive features such as high weld

strength to weld size ratio, minimal heat affected zone,

reliability providing some benefits: low heat distortion,

non-contact process, repeatability and ability to auto-

mate [1]. Nevertheless, the main drawback of Laser

welding is that the intense energy input, resulting in se-vere thermal gradients, can contribute to generate

strains driving elements out of alignment. Motions in

excess of 10 lm can be introduced and sub-micron

alignment usually requires some type of motion com-

pensation after the initial welds to hold required toler-

ances [9].

It is also well-known that one severe limitation on

the reliability of microassembling technologies concerns

the stress caused by thermomechanical constants differ-

ence relating to materials of the package (solder joints,

glue, . . .). Consequently an important failure source is

relative to solder or adhesive joints, which are in most

cases, a critical part of the assembly and has a major

influence on its reliability. For example, defects ap-

pearing after a number of thermal cycles depend on

many factors and mainly on CTE mismatches between

the silicon die and the substrate in particular for chip-

on-board technology [10].

Generally, maximum strain rate is calculated to de-

termine the most strained region corresponding to the

most likely failure zone of the solder joint. In this case,

the determination of the local maximum strain accu-

mulated and plastic rate deformation after thermal cy-

cles can only be done by FEM modeling of the whole

package. For that, accuracy of simulations is stronglydependent on physical parameters introduced for ther-

momechanical computations. Solder and adhesive be-

haviors must be represented by a stress–strain rðeÞrelation based on a bilinear model including both elastic

and plastic zone varying with temperature and experi-

mentally measured.

3. Finite element analysis conditions

In order to determine critical areas of the Laser

module from a thermomechanical point of view, simu-

lations are performed using FEM ANSYS software. The

different models and boundary conditions are defined in

this part and the two main sub-assemblies, previously

described, are divided into three parts as a function of

the simulation type:

• Sub-assembly 1a is composed of the Laser platform

and the lens holder essentially in Kovar. The assem-

bling of the two parts requires Laser welding process,

which is the most critical manufacturing step. This

study is particularly focalized on impact of Nd:YAG

Laser welding process on optical beam axis deviation

taking into account of real manufacturing conditions

of the Laser module. The model is based on electrical

thermal and mechanical simulation using multiphys-

ics approach and will allow to extract isothermal con-

tour plots to evaluate magnitude of thermal gradients

in the sub-assembly 1a. The final goal is to calculate

residual stresses and the optical beam axis deviationafter this process.

• Sub-assembly 1b, related to the Laser platform, is

mounted on a thermoelectric cooler by means of a

SnPbAg solder joint (10 lm). Non-linear thermo-

mechanical simulations considering thermal cycles

()40 C/+85 C) are made to evaluate thermal fatigue

of the solder joint.

• The fiber pigtail represents the sub-assembly 2. Non-

linear thermomechanical simulations considering

thermal cycles ()40 C/+85 C) are made to evaluate

strains on the filament of the optical fiber glued into

the ferule of the pigtail. Solder joint and glue are at

the origin of fatigue phenomenon and defect propa-

gation located in interfaces. Theses elements are very

sensitive during ageing tests.

Sub-assembly 1a: Fig. 2 presents the global model of

sub-assembly 1a with a planar symmetry (O x, O y ). Fixed

points, considered as nodes without any degrees of

freedom, represent fixing flanges used in manufactur-

ing process to maintain the lens holder and the Laser

Fig. 2. Sub-assembly 1a simulated design with external loads.

Y. Deshayes et al. / Microelectronics Reliability 43 (2003) 1125–1136 1127



platform during Laser welding process. The external

loads are listed below:

• The weight is applied on the gravity center.

• The clamp forces F pres to ensure an adjustment be-

tween Laser platform and lens holder is applied on

the back of the lens holder during Laser welding pro-

cess.

• The Laser heating conditions are described by the in-

set of Fig. 2a and modeled by Joule heating consider-

ing the well-known thermal/electrical analogies. The

equipotential surface is adjusted to obtain a maxi-

mum temperature at 1400 K. We use two different

electrical characteristics to traduce a localized heat

source as generated by a Nd:YAG Laser beam. This

part will be developed in the next section.

An optimized model with 7526 elements and 11803

nodes using three-dimensional hexagonal multiphysicstransfer elements was used for this sub-assembly 1a.

Sub-assembly 1a is mainly composed of Kovar. Values

of physical constants used in these simulations for the

sub-assembly 1 are listed in Table 1. Material properties

are assumed to be dependent on temperature. The time

dependence of Laser welding process has conducted us to

elaborate time-dependent (transient) simulations. Fig. 3

summarizes the time dependence of the applied condi-

tions for an analysis of thermal stress and distortion of

the sub-assembly 1a. In this study, the laser heating is

modeled by Joule heating taking into account of elec-

trical/thermal energy considering that a Laser welded

joint can be associated to an electrical resistance calcu-

lated from the same area of material (Kovar). To eval-

uate the thermal energy developed in the volume of the

welded joint, we considered the relation between the

enthalpy variation and the electrical energy (1):

D H ¼V 2

R Dt ð1Þ

with D H defines the enthalpy variation, R is equivalent

of an electrical resistance of the Laser welded joint

volume, V corresponds to the time-dependent applied

voltage (V YAG1 and V YAG2) and Dt is the YAG Laser

pulse duration (2.5 ms). V YAG1 and V YAG2 respectively

correspond to inferior and superior Laser welded jointand are not applied simultaneously as shown in Fig. 3.

Deposed energy with a YAG Laser is calculated

considering electrical energy dissipated from a resistance

on which an applied voltage allows to simulate a thermal

energy in the volume of the spot weld with a temperature

close to the melting temperature. Our simulations are

based on the following expressions giving the relation

between electrical energy and thermodynamical condi-

tions:

V 2

R

Dt ¼ mC p þ DT þ L f ð2Þ

D H ¼ mC pDT þ L f ð3Þ

Eq. (2) gives the heating conditions corresponding to the

Laser energy quantity deposed on the material with C pdefines as heat capacity (in J kg1

C1) and L f repre-

sents the latent heat of melting given in Joule. Cooling

conditions taking into account of latent heat solidifica-

tion Ls, given in Joule, are resumed by Eq. (3). It is

known that, for Kovar, the heat capacity parameter has

temperature dependence and literature allows to extract

the value until 1200 C rather than heat latent of so-lidification is difficult to obtain [5]. This parameter tra-

duces cooling which is critical in this case. So proposed

simulations are computed at a temperature close to 1473

K corresponding to the temperature at which thermo-

mechanical constants are given in Table 1. All thermo-

mechanical properties have been used to simulate the

Laser welding process and give thermal and mechanical

solutions [5].

Sub-assembly 1b: The sub-assembly 1b is composed of

the Laser platform mounted on the thermoelectric

cooler with a SnPbAg solder joint (10 lm). Fig. 4 pre-

sents the different external loads of sub-assembly 1b:

Table 1Physical constants of the material using in the sub-assembly 1a

versus temperature

Kovar

300 K 873 K 1473 K

CTE (lm/K) 5.13 5.86 11.5

Young modulus (GPa) 138 138 138

Yield strength (MPa) 345 245 50

Poisson ratio 0.317 0.317 0.317

Thermal conductivity

(W m1 K1)

17.3 17.3 17.3

Heat capacity (J kg1) 439 439 649

Melting point (K) 1723

Fig. 3. Time dependence of boundary conditions for sub-

assembly 1a.





• The magnetic force between lens holder and metallic

(KOVAR) package due to permanent magnetic field

located in isolator used to polarized Laser beamemitted by Laser diode is modeled by pressure

strength on the columns of the Laser platform.

• Application of thermal cycles ()40/+85 C).

The reference conditions for simulations are: the

planar symmetry (O x, O z ), the reference plane induces

no displacement in z direction and reference point with

no displacement in all directions. An optimized FEM

model with 55 075 elements and 123 193 nodes using

three-dimensional quad and hexagonal transfer elements

was used for this sub-assembly 1b and the different

materials are listed in Table 2.Thermomechanical behavior of all materials is as-

sumed to be linear for each temperature excepted for the

SnPbAg solder joint. Fig. 5 gives the stress/strain be-

havior versus temperature with a bilinear model using

tensile experimental analysis of test samples realized

with a thickness of 15 lm. All thermomechanical

properties have been used to simulate thermal cycling

ageing conditions and give thermal and mechanical so-

lutions.

Sub-assembly 2 (pigtail): Fig. 6 shows the global

model of sub-assembly 2. The planar symmetry (O x, O y )

model was used. The fixed points represent the fixing

conditions of the ferule holder on the body package.

External stresses are:


• Application of thermal cycles ()40/+85 C).

Fig. 4. Sub-assembly 1b simulated design with external loads.

Fig. 5. Stress/strain versus temperature of SnPbAg solder joint

(thickness 15 lm).

Table 2

Physical constants of the materials using in the sub-assembly 1b

(ambient temperature)

Kovar Al2O3 SnPbAg

CTE (lm/K) 5.13 7 24.7

Youngs modulus (GPa) 138 200 9

Yield strength (MPa) 345 – 9

Poisson ratio 0.317 0.3 0.3

Fig. 6. Sub-assembly 2 simulated design with external loads.




FEM model is build with 16 741 elements and 3572

nodes. The difference between number of elements

and nodes is explained by the additional element

traducing the pivot contact between the ferule and the

holder. Table 3 resumes physical constants for each

material of sub-assembly 2 to perform simulations.

These ones imply that all materials have a linear

behavior versus temperature excepted for adhesive

material. Stress/strain characteristics curves dependent

on temperature have been experimentally determined

by tensile experiments on normalized test samples

(Fig. 7) [11]. All thermomechanical properties have

been used to simulate thermal cycling ageing condi-

tions and give thermal and mechanical solutions. We

can note that all mesh models of each sub-assembly

are created with an automatic mesh generator ex-

cepted for the sub-assembly 1b which has been opti-

mized because of the thin layer of solder joint (15

lm).For sub-assemblies 1b and 2, all materials are con-

sidered as homogeneous and simulations are static and

isothermal (steady-state). All assemblies are submitted

to a temperature cycling )40 C/+85 C with a step of 20

C between each step of calculation. For all simulations,

the reference temperature is set to 27 C.

4. Results of thermomechanical simulations

Sub-assembly 1a: Nd:YAG Laser welding process in-

volves a highly focused Laser beam responsible of a non-

uniform temperature distribution on the focal print.

Simulated energy deposed allows to be close to melting

temperature of Kovar material (1723 K). Fig. 8 shows

the nodal solution contour plot of thermal cartography

of Laser platform after first Laser welding process. The

temperature variation along the column of Laser plat-

form can be fitted by a Gaussian law which can be ex-

pressed as:

T ðr Þ ¼ T 1 þ ðT 0 T 1Þ expðr

2

=W

2

0 Þ ð4Þ

with T 0 ¼ 1427 K, the maximal temperature of Laser

weld, T 1 ¼ 600 K the minimal temperature of Laser weld

and W 0 the beam waist defined as the minimum radius of

the Laser beam.

Experimental and calculated beam waist values are

the same and evaluated around 150 lm. The good

Table 3

Physical constants of the materials using in the sub-assembly 2

(ambient temperature)

Kovar Zirconia SiO2 Adhesive

material

CTE (lm/K) 5.13 7 0.56 13.7

Young

modulus (GPa)

138 200 72.4 1.9

Yield strength

(MPa)

345 – – 54

Poisson ratio 0.317 0.3 0.19 0.3

Fig. 7. Stress/strain characteristics of adhesive material in the

pigtail.

Fig. 8. Temperature variation and cartography of sub-assembly 1a.




agreement between experimental and calculated values

validate the simulation approach for Laser Nd:YAG

welding process (Fig. 9).

Fig. 10a compares strains in sub-assembly 1a struc-

ture before and after Nd:YAG Laser welding process.

Strain occurring in the column is observed and this

particular view (deformed and undeformed nodal solu-

tion plots) allows to highlight optical beam axis devia-

tion of the lens holder. Fig. 10b clearly shows the

residual effective Von Mises stresses close to 55 MPa

located in the column base of the Laser Platform after

Laser welding. Thermal gradients (1100 K) along

columns of Laser platform induce maximal displace-

ments close to 2 lm located in the column base after

Nd:YAG Laser welding process in manufacturing con-

ditions and maximal strains of 0.05% (Fig. 11). These

displacements are observed by optical axis angular de-

viation h and D x, D y and D z axial deviations. Simulation

results give calculated values of these parameters. In Fig.

12, two nodes of lens holder representing optical axis

have been considered. We also reported D x, D y , D z and h

deviations of optical axis allowing to give its final posi-

tion after total Laser welding process. Considering this

two nodes, optical axis deviation resulting from residual

stress after Laser welds can be evaluated close to an

angular deviation of h ¼ 0:03 and axial deviations of

D xmax ¼ 2 lm, D y ¼ 0 and D z max ¼ 0:1 lm. Experimen-

tal analyses for evaluation of optical coupling drop in

1550 nm Laser modules have reported that optical

power losses of 15% result from two critical values of

parameters variation:

• An angular deviation h of 0.02 between sub-assem-

bly 1 and sub-assembly 2.

• D y , D z deviations of 10 lm between Laser diode and

Lens holder.

At this step of the manufacturing process, optical

coupling between lens holder and Laser diode is cor-

rect because D y , D z deviations are close to 0.1 lmFig. 9. Theoretical Laser beam characteristics assuming

Gaussian energy distribution U0ðr Þ.

Fig. 10. Residual effective Von Mises strains (a) and stresses (b) after Laser welding process on the sub-assembly 1a.




corresponding to optical coupling losses lower than

0.1%. Thus, an operator could not suspect a possible

displacement of the first lens axis after Nd:YAG Laser

welds. The assembling step between the pigtail and the

sub-assembly 1 requires then a dynamic alignment to

find maximal optical coupling. In this case, the operator

can adjust a possible optical beam axis deviation with-out any information about the value of the previous

deviation and the level of accumulated stresses trapped

in the sub-assembly one.

After Nd:YAG Laser welds, intrinsic and extrinsic

stresses can appear. Intrinsic stresses are generally re-

lated to only Laser YAG energy deposition on metal

surface. The most important accumulated stress is lo-

cated inside the heat-affected zone (HAZ) caused by

plastic deformation and very rapid thermal variation in

welded joints [3,12]. Extrinsic stresses are caused by

external loads applied during process and discontinuity

of materials on the interface [13]. In our case, the most

important external load is represented by pressure

strength F pres used to ensure an adjustment between

Laser platform and lens holder.

Relaxation of accumulated stresses in the sub-

assembly 1a can occur and could be accelerated by

defects induced in the welded zone [12,13]. Rapid so-

lidification processing in HAZ leads to a metastable

phase formation, solid solution or dispersion strength-

ened alloys and intermetallics and the whole physical

phenomenon is at the origin of defects formation lo-

cated in welded joints [14,15]. It has been demonstrated

that metallic alloys creep fatigue is related to defects

rate located in welded joints considered as a metallic

alloy zones [16]. In particular, a model based on mo-

lecular dynamics calculations, developed by J.D. Vaz-

quez, has discussed on isotropic and anisotropic

relaxation phenomenon from simulations of lattice re-

laxation of metallic alloys considering the sudden ap-

pearance of vacancy or an interstitial site in the crystal

[17]. This microscopic relaxation model allows to

highlight macroscopic effective displacement of system

responsible of relaxation phase. Experimental mea-

surements, using in particular an optical method, have

been also conducted to observe strains, stresses and

fractures of welded joints at the mesoscale level byPanin [17]. This study has characterized, in bulk ma-

terial, the accumulated stresses located in HAZ and

their evolution after Laser welding process. So our in-

terpretation of gradual optical power drift between the

sub-assembly 1a and the pigtail can be explained by

relaxation phenomenon and time evolution can be di-

rectly related to the number and the location of defects

into the welded joints but also in the structure.

Experimental procedure has been established for lo-

calize strains and stresses in sub-assembly 1a during the

whole step Nd:YAG Laser welding process and evalu-

ation of relaxation phenomenon after thermal cycles.

Fig. 11. Displacements and deformed-undeformed view of strains located in sub-assembly 1a after Laser welding process.

Fig. 12. Theoretical optical axis deviation before and after

Laser welding process.




This part of study is high cost and long process and will

be exposed in a next work.

Sub-assembly 1b: The most critical part of sub-

assembly 1b concerns SnPbAg (10 lm) solder joint

connecting the interface between Laser platform and

thermoelectric cooler. To evaluate the fatigue of this

element, the maximum cumulated plastic rate strain has

been mapped after two complete cycles. Maximum total

Von Mises strains and stresses are extracted from the

edges of the solder joint (highest stress zone) only for the

last cycle, given the maximum cumulated plastic rate

strain ep (0.08%) as represented in Fig. 13.

Dep allows to give an evaluation of minimal number of

thermal cycles before failure (N) that could generate an

initiation of creep by thermal fatigue calculated without

residual stresses and defects in the solder material. This

phenomenon is traduced by Coffin–Manson equation

used for parallelepiped solder joint [18]:

N ¼ k

ðDepÞn ð5Þ

The number of thermal cycles N before failure reach

600 cycles from )40 C to +85 C with n ¼ 1:2 corre-

sponding to 10 lm thickness solder joint and k ¼ 27:2that is an empiric constant given by J.H. Lau for par-

allelepiped solder joint [18]. The location of the maxi-

mum cumulated plastic strain in solder joint determined

from ANSYS simulations is located in the edges of the

solder joint (Fig. 14). Considering two nodes of the sub-

assembly 1b, optical axis angular deviation h resultingfrom residual stress after thermal cycles ()40 C/+85 C)

can be evaluated close to 3 106 degrees (). So the

solder joint, without defect, cannot be considered as a

critical zone from a thermomechanical point of view

because the resulting misalignment is lower of 4 decades

of degree than the optical deviation previously calcu-

lated after Laser welding process on the sub-assembly

1a.

Sub-assembly 2: Fig. 15 gives the maximal computed

sheild stresses (SXY, SYZ and SXZ) and tensile stresses

(SX, SY and SZ) in the fiber versus computed steps.

Each step represents the different temperature of ther-

mal cycle. The temperatures are set to )

40 C and +85C. The maximum value of the stress in the ferule is 42

MPa at 85 C and )49 MPa at )40 C. These values

are lower than the value of rupture given by the manu-

facturer (2.5 GPa) but microstripes and cracks can be

initiated by striped process and so addition of succes-

sive compressive and tensile stresses occurring in the

fiber after thermal cycling dramatically increase the

possibility of failure by thermomechanical fatigue. De-

structive physical analyses have shown the good agree-

ment with simulations and in particular appearance

of fiber break into the ferule after 500 thermal cycles)

40 C/+85 C.

5. Ageing tests analysis

Qualification procedures, in particular power drift

measurement, must be conducted to validate the system

with respect to tolerances through temperature cycling

or storage temperature characterizing the limits and the

margins of the technology. Actual standards tend to be

500 cycles in the temperature range )40 C/+85 C

with a failure criterion of 10% of optical power drift.

The methodology of failure diagnostic for optoelec-tronics components and modules for telecommunica-

tion applications imposed to do ageing tests to validate

different assumptions coming from the simulation re-

sults.

First ageing tests have been made on 1550 nm In-

GaAsP/InP DFB Laser diodes. After 500 thermal cycles

)40 C/+85 C, no failure occurred on Laser diodes.

Measurements have been made with a specific test bench

with temperature dependence has been developed to

monitor P(I), I(V) and L(E). This result demonstrates

that optical power drift is only associated to misalign-

ment in relation with thermomechanical aspects. The

Fig. 13. Stress/strain curve during thermal cycles and calcula-

tion of plastic rate deformation.

Fig. 14. Values of plastic strains in SnPbAg solder joint at +85

C (symmetry view).




second ageing test is made on optoelectronic modules in

final packaging. Fig. 16 shows evolution of D E ta (%)

defined by:

D E ta ¼ D P opt

P opt

I ¼100 mA

ð6Þ

with P opt is initial optical power measurement, D P opt is

the difference between optical power measured after

ageing time and initial optical power measurement and I

is the current value for optical power measurement.

This experimental procedure is applied on nine In-

GaAsP/InP 1550 nm Laser modules versus thermal cy-

cles )40/+85 C. In Fig. 16, evolution of output optical

power measured at 100 mA after 500 thermal cycles

()40/+85 C) are reported. Three different behaviors

have been observed and related with simulation results

as resumed in Table 4. Experimental and simulation

Fig. 16. Evolution of optical power measured at 100 mA after 500 thermal cycles ()40 C/+85 C) on 1550 nm InGaAsP/InP Laser

modules.

Fig. 15. Evolution of maximal strains values in fiber versus thermal cycles.




results lead to give failure modes and assumptions on

failure location:

• sudden total optical power drop explained by a break

located in the optical fiber core,

• gradual optical power drift outside the failure criteria

limit in relation with thermomechanical aspect re-

sponsible of columns deformation in sub-assembly

1a and related by stresses relaxation phenomenon,

• gradual optical power drift inside the failure criteria

demonstrating the relative instability of optical cou-

pling in Laser module especially on sub-assembly 1a.

6. Discussion and conclusion

Laser welding process in sub-assembly 1a has been

identified as the most potential critical zone and to

correlate simulation results using ANSYS software, ex-

perimental analyses have been also investigated [19].

Calculated optical misalignment in sub-assembly 1a

have demonstrated an angular optical beam axis devia-

tion of 0.03 and responsible of a possible first lens axis

movement confirming that Laser welding process can

induce optical instability of Laser modules and degra-

dation of performances for telecommunication appli-

cations. The main solution could be given by a betteroptimization of the Nd:YAG Laser power density close to

1.5 105 W/cm2. For this technology, average Nd:YAG

Laser power density reaches 2.5 105 W/cm2 and can

generate bulk defects and thermal stresses in welded joints

(Fig. 17). In a recent paper, W.H. Cheng has established

that optical losses in Laser modules can related to the

presence of bulk fractures [15]. It has also been demon-

strated that power density is responsible of bulk defects

and accumulative stresses. In our case, the presence of

bulk defects, observed in Fig. 17, could explained random

acceleration of time stress relaxation allowing to explain

optical power drop. The time before failure related to

10% of the optical power drift is directly related to the

manufacturing process and to the order of static inde-

termination of the system strongly dependent on the

Laser platform and the lens holder design. All conditions

are correlated to a mechanical misalignment between

Lens1 axis and pigtail. The major cause of bulk defects

formation in the Laser welding process for sub-assembly

1a is due to the excess Laser energy. The other causes are

gas bubbles trapped within the weld sections and theheterogeneous nucleation in weld joints [15].

In conclusion, this paper reports 3D thermomechan-

ical simulations and experimental tests in order to

identify critical zones Butterfly-package Laser module

showing that three main zones must be carefully ana-

lyzed: shape and volume of glue in the ferule, solders

and Laser welds. Laser welding process is a useful and

effective method to ensure hermeticity and secure metal

parts but the mechanical distortions due to severe ther-

mal gradients should be controlled within allowance

limits. The main advantages of this technique are given

by precision of alignment close to 0.2 lm, the whole

Table 4

Synthesis of simulation results and failure modes

Sub-

assembly

Maximal

strains

(%)

Maximal

stresses

(MPa)

Location of stresses

and strains

Failure modes Thermomechanical

stresses and strains

sensitivity (optical

deviation maximal

value)

Assumption on failure

mechanisms

1a 0.5 58.4 Column base of the

Laser platform

Gradual

degradation

Very high (0.03) High levels of strains after Laser

welding process––optical axis

deviation associated to stresses

relaxation phenomenon

1b 0.7 – Edges of solder joint No degradation Low (3 106) Thermomechanical fatigue after

600 cycles ()40/85 C)

2 – 49 Fiber core Sudden

degradation

High (0) Microcrack in the core of the

fiber leading to a break after

thermal cycles

Fig. 17. Bulk defects formation in a Laser welded joint of the

sub-assembly 1a after Nd:YAG Laser welding process.




process fully automated to contain the cycles time

within 60–90 s. But it has been shown that one of the

main inconvenient of the Laser welding process is the

excess of deposed Laser energy resulting in high thermal

gradients and residual stresses in the Laser platform

responsible of an optical misalignment and a possible

failure in term of optical power drift requirements. We

have demonstrated that FEM simulations, to predict

distortion of Laser welding which is very difficult to

measure, is very attractive and can be applied to differ-

ent package configurations.

Our activities are now focused on FEM predictions

that could be improved by a detailed knowledge of the

effect of bulk defects located in Laser welded joints on

stresses relaxation phenomenon and also by a better

implementation of heating and cooling conditions in

computations. The main objective is to improve pack-

aging design rules and optical misalignment reduction in

order to achieve highly reliable bandwidth single modefiber communication systems.

References

[1] Sherry WM, Gaebe C, Miller TJ, Schweizer RC. High

performance optoelectronic packaging for 2.5 and 10 Gb/s

Laser modules. In: Electronic Components and Technol-

ogy Conference, 1996. p. 620–7.

[2] Breedis JB. Monte Carlo tolerance analysis of a passively

aligned silicon waferboard package. In: Electronic Com-

ponents and Technology Conference, 2001. p. 247–54.

[3] Jang S. Packaging of photonic devices using Laser welding.

In: Proc SPIE, 1996. p. 138–49.

[4] Sauvage D, Laffitte D, Peerinet J, Berthier P, Goudard JL.

Reliability of optoelectronic components for telecommu-

nications. Microelectron Reliab 2000;40:1701–8.

[5] Cheng WH, Yang YD, Liang TC, Wang GL, Sheen MT,

Kuang JH. Thermal stresses in box-type Laser packages.

Opt Quant Electron 1999;31:293–302.

[6] Goudard JL, Berthier P, Boddaert X, Laffitte D, Perinet J.

New qualification approach for optoelectronic compo-

nents. Microelectron Reliab 2002;42:1307–10.

[7] McLeod R, Wolkin M, Morozov V, Sawyer K. Packaging

of micro-optics component to meet Telcordia standards.

In: Proc OFC, Anaheim, CA, USA, March 2002.

[8] Fricke S, Keim E, Schmidt J. Numerical weld modeling-

method for calculating weld-induced residual stresses.

Numer Eng Des 2001;206:139–50.

[9] Hayashi T, Tsunetsugu H. Optical module with MUconnector interface using self-alignment technique by

solder-bump chip bonding. In: Proc 46th Electronic

Components and Technology Conference, 1996. p. 13–9.

[10] Trigwell S. Die-attach materials and methods. Solid State

Technol 1995;(April):63–8.

[11] Perichaud MG, Deletage JY, Tregon B, NKaoua G,

Fremont H, Danto Y, et al. Optimisation of an assembling

process of passive components connected with conductive

adhesives. In: 12th European Passive Components Sym-

posium, CARTS-EUROPE 98, 1998. p. 50–5.

[12] Panin VE, Pleshanov VS, Kobesva SA, Burkova SP.

Relaxation mechanism of rotational type in fracture of

weld joints for austenic steels. Theoret Appl Fract Mech

1998;29:99–102.[13] Inoue T, Koguchi H. Relaxation of thermal stresses in

dissimilar materials (approach based on stress intensity).

Int J Solids Struct 1997;34:3215–33.

[14] Hariprasad S, Sastry SML, Jerina KL. Deformation

behavior of a rapidly solidified fine grained Al–8.5% Fe–

1.2% V–1.7% Si alloy. Acta Mater 1996;44:383–9.

[15] Cheng WH, Wang WH, Chen JC. Defect formation

mechanisms in Laser welding techniques for semiconduc-

tor Laser packaging. IEEE Trans Compon Pack Manuf

Technol 1996;19:764–9.

[16] Asayama T. Creep-fatigue evaluation of stainless steel

welded joints in FBR class 1 components. Nucl Eng Des

2000;198:25–40.

[17] Dominguez-Vazquez J, Perez-Martin AMC, Jimenez-

Rodriguez J, Gras-Marti A. Relaxation of metals: a model

based on MD calculations. Nucl Instru Meth Phys Res

1998;135:214–8.

[18] Lau JH. Thermal stress and strain in microelectronics

packaging. 1st ed. New York: Von Nostrand Reinhold;

1993.

[19] Experimental validation of thermomechanical simulations

on 1550 nm Laser modules. Internal Report, ALCATEL

Optronics-IXL, September 2001.


three-dimensional fem simulations of thermomechanical stresses in 1.55 lm laser modules

Documents