components
Thomas Burkhardt1a,b, Marcel Hornaffa, Erik Beckerta, Ramona Eberhardta, Andreas Tnnermanna,b
a
Fraunhofer Institute for Applied Optics and Precision Engineering, Albert-Einstein-Strasse 7,
07745 Jena, Germany
b
Institute of Applied Physics, Friedrich Schiller University Jena, Albert-Einstein-Strae 15, 07745
Jena, Germany
ABSTRACT
Laser-based solder bumping is a highly flexible and fast approach for flux-free soldering of micro-optical components in
complex 3D geometries with localized and time restricted energy input. Solder joints provide superior mechanical
strength, higher radiation stability, humidity resistance and a good thermal and electrical conductivity compared to
adhesive bonding. Due to the good long term stability solder joints are feasible for the integration of optical, mechanical,
electronic, and MEMS/MOEMS devices in multi functional hybrid optical assemblies. Comparative studies of solder
bumping of optical components with sputtered thin film metallization on platforms made of Alumina (Al2O3) and Low
Temperature Cofired Ceramics (LTCC) with both Au and AgPd thick film metallization were carried out using design of
experiment methods (DoE). The influence of the system parameters, laser pulse energy and duration, distance, incidence
angle and nitrogen pressure on targeting accuracy and bond strength were evaluated. The jetting of liquid solder spheres
within a localized nitrogen atmosphere improves wetting on the respective wetting surfaces and simplifies the joining
process due to integration of solder alloy preform handling and reflowing, thus showing great potential for a high degree
of automation.
Keywords: Micro assembly, soldering, solder bumping, lead-free solder, SnAgCu, Al2O3, LTCC
1. INTRODUCTION
Hybrid integration and packaging of optical and opto-electronical systems require a high precision joining technology to
be the reliable completing task in the process chain of system assembly. Following the tasks of handling and alignment
the joining technique has to guarantee the compliance with submicron adjustment during the bonding process as well as
over the life time cycle of the assembly. Solder joints promise to overcome known drawbacks of the current standard
technology bonding by means of UV-curing, polymer based adhesives. Limitations concerning long term stability,
sensitivity to radiation and humidity as well as outgassing in vacuum environments can be resolved by using metallic
joint connections. Depending on the basic conditions of the assembly solder alloys can be selected from a wide range of
material systems with different liquidus temperatures and mechanical parameters concerning tensile strength, creep and
cryogenic compatibility. Metallic solder alloys further provide good thermal and electrical conductivity, thus enabling
for the integration of further functionality in the solder joint providing thermal contact to a heat sink or electrical
connection. In comparison to standard adhesives it has to be mentioned that the creation of solder joints requires a
wettable metallization, thus needing a further process step and due to the opaqueness these metallizations strictly have to
be created outside the optical beam propagation path.
Solder bumping was initially developed for placement of solder bumps for microelectronics packaging [6]. The high
processing speed due to fast application and the fast solidification of the solder alloy on the target compared to UV-
curing adhesives with curing times of about 10 seconds depending on the adhesive, lead to a very short cycle time. Flux
less processing as provided by solder bumping prevents contamination of optical surfaces, thus providing a better
1
thomas.burkhardt@iof.fraunhofer.de; phone +49 3641 807-339; fax +49 3641 807-604; www.iof.fraunhofer.de
Laser-based Micro- and Nanopackaging and Assembly III, edited by Wilhelm Pfleging,
Yongfeng Lu, Kunihiko Washio, Willem Hoving, Jun Amako, Proc. of SPIE Vol. 7202,
720203 2009 SPIE CCC code: 0277-786X/09/$18 doi: 10.1117/12.807986
Reservoir of solder
spheres
Singulation unit
Reflow laser
pulse
Bump capillary
Singulated
solder sphere
Nitrogen flow
Optical components
3. POSITION ACCURACY
3.1 Test assembly and design of experiments
Complex assembly environments with a high level of integration require the creation of solder joints in locations difficult
to reach. One objective of solder bumping is to increase the bump distance as much as possible, gaining flexibility for the
placement and orientation of the capillary (length beyond bond head about 15 mm, diameter < 3 mm with a tapered tip)
with respect to the joining geometries. Fig. 2 shows the setup used to determine the influence of process parameters on
targeting accuracy that appears as a displacement between the targeted position and the actual placement of the solder
bump. Metalized microscopy slides with fs-Ti:Saphir laser structured arrays of cross lines for referencing the targeted
positions were used as targets. The metallization system with an overall thickness of 500 nm consists of a Ti-layer as
adhesion promoter, a Pt-layer as diffusion barrier and a wettable non-oxidizing Au-layer necessary for flux-free pro-
cessing. The target substrate is mounted on a precision rotation stage placed beneath the robot manipulated solder jet
bond head. Aiming and positioning of the solder jet capillary is then accomplished using both the pilot laser and a side
mounted camera with a LabView based image capturing and processing software.
Q.
2
z
L:\'
Targetiiig sample
Fig. 2. Metalized microscopy slice used for accuracy test and detail view of a 400 m solder bump placed with
reference to laser structured reticule (left), schematics of targeting bump capillary relative to substrate.
An orthogonal central composite design of experiments with a second order model for the response variable the radial
displacement between targeted point and actual position is selected considering the five system parameters: laser pulse
current, laser pulse duration, distance of capillary to target, angle of incidence and nitrogen pressure. The design consists
of three separate sets of experimental runs: 1st) a fractional factorial design, each factor having two levels, 2nd) a center
point with the value of each factor set to the respective median of the values used in the factorial design and 3rd) the set
of axial points that provide two values below and above the factorial levels for each factor. The complete design is
grouped to one block, randomizing the sequence of experimental runs leading to a total number of experimental runs of
27. Each combination of parameters is replicated 16 times to allow for estimation and reduction of the standard deviation
as well as the elimination of outliers. Both creation of the experimental plan and the analysis of experimental data were
carried out using the commercial DoE software DesignExpert.
.1
1,2 'V
1,0 'V
V
0,8 ._4 V
90 - A -
A
:ZO
0,0
10 20
It'll. .
30 40 50
o ;
60 70 80 90 100
I [A]
Fig. 3: Laser pulse energy for 400 m capillary in dependence of laser current und pulse duration the system parameters
used in the DoE.
As shown in Table 1 the factors nitrogen pressure, bond level and the angle of incidence can be considered significant in
terms of the response variable. All coefficients are positive, thus minimizing the values within the boundaries given by
the design of experiments will minimize the displacement. Both laser pulse current and duration are not significant. By
exclusion of the non-significant regression coefficients the reduced regression equation is composed and checked against
the model of the design using the lack-of-fit test. The second order model and the resulting data do not sufficiently match
in the lack-of-fit test, i.e. there are disturbance variables yet unaccounted for and statistical spread has to be reduced.
Regardless of the insufficient fit an optimization of the reduced regression equation is calculated. This optimization
provides the parameters as shown in Table 2 and predicts a minimized radial displacement of (5510) m.
The optimized set of parameters (Table 2) has been verified in experiments and the radial displacement was observed to
be 54 m with a standard deviation of 34 m. As shown in Fig. 4 no global minimum point of the displacement is
located within the parametric range but the trend of parameters to favor is clearly observable and it has to be noticed that
the influence of the parameters on bond strength and boundaries set by specific applications needs to be taken into
account, too.
Table 2. Optimized processing parameters and respective laser pulse energy for minimizing radial displacement of solder
bumps (400 m).
Factor Laser pulse current Laser pulse width Nitrogen pressure Bond level Angle of incidence
Optimized 55 A 23 ms 24 mbar 1.53 mm 5.3
parameter
Pulse energy: 1.54 J
I296
1203
Xl Diatance
X2 = Angle
Actual Factors
Laaer current = 54.93 A
Pulae width = 23 ma
N2 preaaure = 24 mbar
Fig. 4. Response-surface plot of radial displacement vs. distance and angle of incidence, laser current, laser pulse width and
nitrogen pressure are constant at optimized levels.
For comparison two factorial designs with center point for the angle of incidence of 0 and 15 and the constant distance
of 1.5 mm have been realized with a solder sphere diameter of 760 m. Application geometry enforced the restriction to
the realized constant bond level. Optimized processing parameters are shown in Table 3 and have been experimentally
verified. These parameter settings are valid for glass substrates only and cannot be applied to thermally higher
conducting materials such as ceramics and metal heat sinks.
Table 3. Optimized processing parameters and respective laser pulse energy for minimizing radial displacement of solder
bumps (760 m).
Factor Laser pulse Laser pulse Nitrogen Bond level Angle of Radial
current width pressure incidence displacement
Optimized 50 A 25 ms 25 mbar 1.5 mm 0 42 m
parameter
Pulse energy: 2.25 J
Optimized 50 A 15 ms 20 mbar 1.5 mm 15 91 m
parameter
Pulse energy: 1.44 J
4. BOND STRENGTH
4.1 Test vehicles and experimental setup
To provide a test vehicle for shear force measurements optical dummy cuboid prisms with the dimensions of
4 x 4 x 2.7 mm made of Borofloat BF33 glass were soldered to ceramics platforms. The Borofloat cuboids are metalized
using a sputtering process to apply the Ti-layer as adhesion promoter, a Pt-layer as diffusion barrier and an Au-layer with
an overall thickness of 500 nm. Both LTCC and alumina were processed to provide rectangular ceramics platforms with
the dimensions of 56.5 x 52 mm and 48 evenly distributed cross shaped land patterns made of Au and AgPd thick film
metallization. The LTCC multilayer test substrates were manufactured by using Du Pont 951-AX tape and a standard co-
firing profile of 850C. Ten layers of blank tape are used to provide a sufficient mechanical strength for the substrate
resulting in a total substrate thickness of 2 mm. Except for the top layer with screen printed and co-fired metallization the
LTCC layers do not contain any vias, cavities or metallization; features that can however provide additional functionality
and structures for passive alignment for hybrid integration of optical, mechanical and electronic components. The
alumina platforms are made of a monolayer of sintered Al2O3 with an overall thickness of 1 mm and a screen printed
metallization.
Fig. 5. Experimental setup of apparatus to solder optical elements to LTCC and alumina platforms (left), simplified
schematics of solder jet bumping capillary within placement distance to the joining geometry.
Fig. 5 shows the experimental setup used to solder the dummy optical elements to the ceramics platform. Targeting is
accomplished by use of the side mounted camera. The apparatus is designed to passively align six elements at once to a
support unit and mechanically clamp them during the process of solder bumping. Relevant alignment parameters are the
contact to the ceramics platform and the rotation around an axis perpendicular to the surface of the platform. This
provides an adjustment of the edges of the soldered elements with respect to the edge of the platform that leads to
comparable conditions during shear force measurements. The experiments were conducted using Sn96.5Ag3.0Cu0.5
solder spheres with diameters of 400 m. Each testing element is soldered with two bumps at opposing sides of the
prism. Results from measurements of position accuracy of solder bumps as well as considerations concerning the
rectangular joining geometry and the even distribution of wetting on both component and platform are taken into account
to define constraints to the parametric levels. Mechanical damage of the substrate limits the parametric variation of laser
pulse duration. As described before a central composite design of experiments with a second order model is selected. The
designs are grouped and randomized with a replication of six runs for each set of parameters. For both LTCC and
alumina with Au metallization a plan with all five system parameters, laser current, laser pulse duration, distance of
capillary to target, angle of incidence and nitrogen pressure is chosen (Table 4). The AgPd metallization however
showed severe damages due to long pulse widths, thus a reduced plan is selected setting laser pulse duration to the
constant value of 5 ms (Table 4).
Table 4. Parametric levels for soldering optical elements to Au and AgPd metalized LTCC and alumina.
Metallization Laser pulse Laser pulse width Nitrogen pressure Bond level Angle of incidence
current
Au 45..65 A 15..25 ms 18..28 mbar 1..3 mm 15..30
AgPd 45..65 A 5 ms 16..30 mbar 1..3 mm 15..30
4.3 Results
For LTCC with AgPd metallization laser current and nitrogen pressure were found to be the most significant factors
note that pulse duration is also important but is set constant to 5 ms due to damage to the substrate at higher values. The
reduced regression equation consists of these two main effects, the two-factor-interaction of laser current and nitrogen
pressure and the second order term for the nitrogen pressure. Fig. 7 shows the response plot against the two main effects
at the optimized values for distance and angle of incidence. A local maximum is found at the edge of the parameter space
evaluated but a further increase of laser current while decreasing the pressure is not feasible due to wetting of the
capillary. As shown in Table 5 the shear force ranges from 11.4 N to 14.3 N over the parameter space. An optimized
shear force of 14.6 N was achieved.
The results for alumina with AgPd metallization are not distributed normal and thus power transformed for further
analysis. For this substrate laser current, nitrogen pressure and distance were found to be significant. Again pulse
duration has to be set to 5 ms to avoid damage to the substrate. The reduced regression equation also contains the two-
factor-interaction of laser current and distance as well as the second order terms of laser current and distance. Table 5
shows the achieved shear forces ranging from 10.2 N to 15.4 N. A local maximum exists at the edge of the parameter
space that leads to an optimized shear force of 15.6 N. The further increase of the laser pulse current is not feasible due
to wetting of solder at the tip of the capillary.
Fig. 7. Response-surface plot of shear force vs. laser current and pressure for AgPd metallization on LTCC (left),
Response-surface plot of shear force vs. laser current and distance for AgPd metallization on alumina (right).
I 4982 '334
Xl = Lusor cerrent (t Xl = Lucre cwrrent (D
X2 = PuTse wih (nw) X2 = Distance 91114
Fig. 8. Response-surface plot of shear force vs. laser current and pulse width for Au metallization on LTCC (left),
Response-surface plot of shear force vs. laser current and distance for Au metallization on alumina (right).
The results for alumina with Au metallization are power transformed to achieve a normal distribution for analysis. Pulse
width, laser current and distance are the significant main factors. With the two-factor-interaction of laser current and
laser pulse width the reduced linear regression equation is formed. Fig. 8 depicts the response function against distance
and laser current for a selected pulse width. A high variation in shear force is observed ranging from 6.2 N to 15.6 N
(Table 5). The local maximum located at high values of laser current and distance lead to an optimized shear force of
15.4 N and no further increase of laser current is recommended to avoid wetting of the capillary whereas an increase of
the distance is not feasible in terms of positioning accuracy or increased wetting of the capillary.
Table 5. Results of shear force measurements over the evaluated parameter space, calculated optimized parameter sets, and
resulting experimentally verified optimized shear force values.
Substrate/ Shear force Laser pulse Laser pulse Nitrogen Bond level Angle of Optimized
metallization range current width pressure incidence shear force
LTCC-AgPd 11.4..14.3 N 62 A 5 ms 18 mbar 2.38 mm 22.5 14.6 N
Pulse energy: 0.5 J
Al2O3-AgPd 10.2..15.4 N 62 A 5 ms 28 mbar 2.38 mm 22.5 15.6 N
Pulse energy: 0.5 J
LTCC-Au 10.4..17.2 N 61 A 23 ms 20 mbar 2.65 mm 22.5 17.3 N
Pulse energy: 1.87 J
Al2O3-Au 6.2..15.6 N 61 A 17 ms 20 mbar 2.65 mm 22.5 15.4 N
Pulse energy: 1.46 J
ACKNOWLEGDEMENTS
The work presented in this paper was partially funded by the german Federal Ministry of Education and Research within
the Optische Technologien framework program (Funding call BrioLas, subproject BriMo, FKZ 13N8612). Other parts
of this work is funded by the German Federal Ministry of Education and Research (BMBF) within the Framework
Concept Research for Tomorrows Production (FKZ 02PC2063) and managed by the Project Management Agency
Forschungszentrum Karlsruhe (PTKA), Production and Manufacturing Technologies Division (PFT). The authors would
like to thank W. Stckl, G. Leibeling and A. Joswig for sample preparation.
REFERENCES
[1]
Banse, H., Beckert, E., Eberhardt, R., Stckl, W., Vogel, J., Laser Beam Soldering a New Assembly Technology
for Microoptical Systems, Microsystems Technologies 11, 186-193 (2005)
[2]
Beckert, E., Schreiber, P., et al, Design and assembly of a miniaturized high-power laser bar to 50 m fiber
coupling module, Proc. SPIE Vol. 6876, 87613 (2008)
[3]
Beckert, E., Burkhardt, T., Eberhardt, R., Tnnermann, A., Solder bumping - A flexible joining approach for the
precision assembly of optoelectronical systems, 4th International Precision Assembly Seminar, Chamonix, France,
Feb 10-13, 2008, 139-147 (2008)
[4]
Keranen, K., Makinen, J.-T., Kautio, K. T., et al., Fiber pigtailed multimode laser module based on passive device
alignment on an LTCC substrate, IEEE Transactions on Advanced Packaging, Vol. 29, 463-472 (2006)
[5]
Hiltunen, J. A., Kautio, K. T., Makinen, J.-T., Karioja, P., Kauppinen, S., Passive multimode fiber-to-edge-emitting
laser alignment based on a multilayer LTCC substrate, Electronic Components and Technology Conference, 2002.
Proceedings 52nd, 815-820 (2002)
[6]
Zakel, E., Titerle, L., Oppert, T., Blankenhorn, R. G., High Speed Laser Solder Jetting Technology for
Optoelectronics and MEMS Packaging, Proceedings of the International Conference on Electronics Packaging,
Tokyo, Japan, Apr. 17-19, (2002)
[7]
Zhong, Z. W., Arulvanan, P., Maw, H. P., Lu, C. W. A., Characterization of SnAgCu and SnPb solder joints on
low-temperature co-fired ceramic substrate, Soldering & Surface Mount Technology 19(4), 18-24 (2007)