Silver Sintering in Power Electronics: The State of the

Art in Material Characterization and Reliability

Testing
Marco Schaal Markus Klingler Bernhard Wunderle
Robert Bosch GmbH Robert Bosch GmbH Professorship Materials and Reliability
Reutlingen, Germany Reutlingen, Germany of Microsystems
[email protected] [email protected] Technische Universität Chemnitz
Chemnitz, Germany
[email protected]

Abstract—Sintered silver is a very promising option in the carbide (SiC) and gallium nitride (GaN) are used for power
field of automotive power electronics for die-attach and electronics due to their excellent electrical and physical
substrate-attach material as it offers many benefits compared to properties compared to silicon (Si) [4, 5, 6, 7]. Furthermore,
conventional solders. One main advantage is the higher and they feature high switching speeds and higher junction/
therefore better thermal and electrical conductivity, which is operating temperature (above 200 °C) [1, 3, 5, 8, 9, 10]. In
necessary due to the usage of SiC and GaN as a basic material for addition, it is possible to use SiC power devices with a reduced
semiconductors. Another one is the higher melting point of cooling system complexity and therefore reduced system cost
sintered silver in comparison to solder, making it possible to use because of their good heat resistance at temperatures above 250
power electronics at higher temperatures. Especially for higher
°C [8, 11, 12]. Given their beneficial properties compared to
temperatures, it is assumed that this results in an improved
reliability. Whereas a lot of information about the behavior of
silicon, there is a high interest to use them for electronics at
sinter layers is known, e.g., thermal properties, material high temperatures [13]. Therefore, the packaging material—
characterization of mechanical properties are still not understood especially the die-attach material—must operate under extreme
in detail as the results are highly scattered. A reliable test method thermal conditions and is considered to be one of the critical
has to be developed, and this has to be done for cyclic testing, too, interconnections [3, 11]. Traditionally, solders were used for
which is essential for determining reliability. As mechanical this kind of interconnection, e.g. Sn63Pb37, with a melting
properties are strongly dependent on process parameters, their temperature of 183 °C [14, 15]. With these solders, the
influence has to be analyzed as well. For example, the influence of advantage of SiC at elevated temperature cannot be fully used.
the porosity distribution of a sintered layer in between the Besides, the Restriction of Hazardous Substances (RoHS) by
semiconductor and the upper side of the DBC on the mechanical the European Union does not allow the usage of solders with
properties and their reliability is hardly investigated by now. It lead. Lead-free solders with silver and copper are no alternative
can be observed that the porosity varies, and that it often either as their melting point is around 220 °C [7, 11, 14]. A
increases from the middle of the plane towards the edges. The typical assumption is that solders can be used at temperatures
influence on the stresses and strains and on reliability has to be that account 80 % of their melting temperature. This results
determined and if required, improvements have to be developed. from creep effects, leading to fast degradation at higher
In the field of FEM simulation, too, a lot of work has to be done. temperatures and therefore to reliability problems [3, 10, 14,
Therefore, the right input material data, which has to be 16]. However, there are some solders which can operate at
characterized as already described, is absolutely essential. It is
temperatures above 200 °C like expensive Au alloys, or Bi and
also necessary to implement a material model which describes all
of the occurring phenomena like primary and secondary creep.
Zn alloys with bad processability [11, 14]. Typically, they
This paper will give an overview on the current status and state cannot be employed above 250 °C [11, 13]. Conductive
of the art of silver sintering, focusing on the issues of material epoxies, that are also used for die-attach, do not possess the
characterization and reliability testing. performance and reliability needed for these temperatures
either [13].
Keywords—Ag sintering, material characterization, material A solution for taking advantage of the WBG
properties, reliability, physics of failure, power electronics
semiconductors is silver sintering, a low temperature joining
technique (LTJT), where micro- [17, 18, 19] or nano-sized [14,
I. INTRODUCTION 19] silver (Ag) particles are used to replace solders [4, 16].
These particles are sintered with a defined temperature and
Power modules in the field of automotive have an pressure profile, resulting in a porous die-attach layer [1, 16]. It
increasing demand for usage at higher temperatures, which is also possible to perform silver sintering without pressure,
includes higher ambient but especially higher junction even though that would decrease the contact surface area
temperatures. The reason behind this is the need for higher between Ag particles and results in a worse diffusion and
power density and therefore often miniaturization [1, 2, 3]. sintering process [20]. The main benefits of sintered Ag are the
Currently, wide bandgap semiconductors (WBG) like silicon high melting temperature (960 °C) which is similar to that of

  
  

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
bulk material [6], high electrical and thermal conductivity and II. MATERIAL CHARACTERIZATION
thermo-mechanical reliability [3, 4, 6, 21, 22]. Nevertheless, The high CTE mismatch between two parts, like a DBC
the sinter process temperatures are comparatively low and substrate and a semiconductor, leads to a high loading in the
around 20-40 % of the melting temperature [4, 6, 23]. Studies interconnection, e.g. die-attach. Therefore, a detailed
have shown that the mechanical properties like elastic modulus understanding of the material behavior for the die-attach is a
are varying as a function of the porosity, and are also lower necessary precondition for lifetime modelling. A failure
than for the bulk material itself [1, 24, 25, 26, 27]. At the parameter needs to be determined which can be accumulated
beginning of the development of silver sintering, micro-sized for each cycle (Chapter IV). It is usually calculated via FEM
particles were used, e.g. μ-flakes with a size of 100 to 500 μm. (Finite Element Method) simulation. For the current sub-
Typical sintering process parameters at this time were system (semiconductor, DBC, metallization and die-attach) and
temperatures above 200 °C, pressures of 20 to 40 MPa and the failure mode of a CTE mismatch (thermo-mechanical
sinter times of several minutes, resulting in reliable structures problem), the material parameters CTE, Young’s modulus and
and a porosity of around 15 %. As the sinter pressure should be Poisson’s ratio can be sufficient for all of them except of the
reduced, e.g. for serial production, sintering of Ag nano- die-attach material [28]. Nevertheless, this has to be assessed
particles became interesting due to their larger overall surface. for each subsystem. It is often assumed that the parameters
To avoid premature particle agglomeration, it takes much more mentioned before of sintered Ag, which is indeed very porous,
additives, compared to micro-sized particles. Yet, this is quite similar to bulk silver—except of a reduced Young’s
complicates the sintering process because of the outgassing of Modulus. The reason behind this is a reduced cross-sectional
these additives. 30 % of porosity is a typical figure for nano- area, resulting from micrometer-scale pores in the material [2,
sized particles. This, too, leads to a decreasing of the values of 28, 31, 32]. Hence, it was a common approach to use an elastic-
the mechanical properties, e.g. reduction of the reliability due plastic material model without creep, especially for the
to a lower yield limit as well as higher creep effects [28]. relatively dense micro Ag sintered die-attach at the beginning
Power electronic modules are built by stacking material of the LTJT. Producing larger tensile specimen with a good
layers with different coefficients of thermal expansion (CTE) homogeneity level at high temperatures and pressures during
and Young’s modulus, e.g. a direct bonded copper (DBC) to the process results in a virtually ideal elastic-plastic material
which a semiconductor is attached by a silver sintered layer behavior [28]. In contrast to this, Heilmann et al. [28], Caccuri
(Fig. 1). In addition, a heatsink is jointed to the DBC by a et al. [32] and Mroßko et al. [33] mention the following point,
thermal interface material (TIM). Thermo-mechanical stress which is especially necessary to consider for nano- or hybrid
and fatigue occur due to temperature cycling and lead to Ag pastes. Is it possible to produce tensile test specimen with
material deformation (plastic strain) or failure during the same structure of the Ag, e.g. pore diameter, shape and
fabrication and operation processes [1, 8, 22, 28, 29, 30]. The distribution as used in the application. As nano-hybrid Ag is
material behavior of sintered Ag is often assumed to be very sensitive to changes in the specimen geometry or in the
creeping viscous similar to solders caused by the high thermo- process, it is not assured that the pores are homogenously
mechanical stresses [3, 22]. Hence, sintered Ag is strongly distributed across the entire specimen. The reason behind this is
dependent on testing temperature and strain rate. Moreover, it assumed to be outgassing of solvents which is not working
is important to consider this for the material characterization properly [28, 32].
and in order to determine the fatigue lifetime [2, 30]. This study
Caccuri et al. [32] are using a stacking method to build up a
presents an overview of the state-of-the-art material
thicker tensile specimen due to the fact that preparing a 1 mm
characterization as well as of reliability investigations in the
field of sintered Ag. In the Chapter Material Characterization, a thick specimen immediately results in a different pore size
general overview of issues and challenges is presented. than of a sintered joint. Layer after layer is printed up and, in
Different tests are then described in detail, including shear the end, a sintering process is performed. After stacking up a
tests, tensile tests, nanoindentation and tests with micro-sized layer, the solvents have to be evaporated which is usually done
specimen. The subsequent Chapter III analyzes the occurring by heating up the sample. This results in an inhomogeneous
failure mechanism for material characterization and also for microstructure since the first layer is exposed to much more
reliability tests like passive and active temperature cycling. In temperature cycles than the last one. A complete solvent
the last Chapter Reliability Testing, the different methods like evaporation is required to avoid much bigger pores. They used
the mentioned passive and active temperature cycling are an alternative approach in which the sample is stick in a
investigated. Furthermore, ratcheting behavior and the vacuum at room temperature (RT) after each deposition. This
correlation between creep and fatigue damage is examined results in a microstructure comparable to a sintered joint. [32]
based on lap shear tests. Finally, an isothermal bending method Heilmann et al. [28] and Mroßko et al. [33], too, describe the
for substitution of passive temperature cycling is shown. scaling of the samples as the main problem, since for larger
tensile samples it is difficult to get the same microstructure
than for the jointed layers in the system. A possible solution
could be an adjustment of the sintering process (temperature
and pressure) to achieve the same porosity, resulting in the
right Young’s modulus. Further, these adjustments could lead
to different pore shapes and grain sizes and therefore could
influence the yielding and hardening behavior [28].
Fig. 1. Typical power module [28]

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
 Heilmann et al. [28] present a comparison between six micro or nano particles, organic dispersant, binder and thinner
elastic-plastic curves of micro- and nano-sintered Ag of around [35], it is difficult to compare values of different papers and
25 % porosity from literature [2, 22, 31, 32, 33, 34]. It is not authors.
expected that the presented curves are identical, however,
Young’s modulus should be indeed comparable due to the
porosity being the main influence parameter for Young’s
modulus. Instead, there is nearly bulk behavior [33] and also
very small Young’s modulus [31, 34]. Heilmann et al. [28]
shows high sensitivity of the mechanical properties and the
dependence on the process parameters, paste and additives in
the paste, but also the measuring method and test equipment.
The currently used test methods for characterization of
sintered Ag are presented in this chapter. Mostly shear test or
tensile test with sintered test specimens are performed and can
be found in literature. Compared to classic shear tests [16, 35,
36, 37, 38], lap shear tests are more interesting as they allow to
get the stress-strain curve and examine creep behavior. Such
investigations are, among others, made by Li et al. [39] and Lei
Fig. 2. Effect of applied pressure on shear strength [16]
et al. [38]. Tensile tests are usually performed with flat tensile
specimen like by Chen et al. [8, 30], Herboth et al. [2], Nikitin Paknejad et al. [36] compiled a comprehensive summary of
et al. [34] and Weber et al. [10]. Reason for this is that the various papers, whereas they also mentioned that the test setup,
microstructure is then comparable to the one in joints. materials (Ag paste) and whether the failure occurred in the
Furthermore, a stacking method to investigate thicker specimen die-attach or another part is not always known. Mostly, the
is used, too, by Caccuri et al. [32], Gadaud et al. [24] and Wang shear stresses that occur right after the sintering process but
et al. [22]. The approach of Suzuki et al. [40] considers the also after thermal aging and thermal cycling are examined [36].
influence of the microporous structure on deformation One of the most important process parameters, the sintering
behavior. They use an FEM analysis to relate the macro- pressure shows a higher shear strength for higher pressures [16,
deformation properties to the micro-deformation properties of 35, 36, 37, 38] (Fig. 2). This could be explained by increased
the sintered Ag. In addition, nanoindentation tests are atomic diffusion and particle consolidation [35]. It is also
commonly used, e.g. by Mroßko et al. [33], Letz et al. [41], mentioned that there is a decrease in shear strength at high
Leslie et al. [42] and Long et al. [43]. Zschenderlein et al. [44] temperatures (300 °C) for sintering pressures between 11 and
presents a micro bending test (MBT) and uses a nanoindenter 15 MPa [35, 36]. Li et al. [35] explain it with a rapid growth
for bending a double dog bone specimen. In this case, sintered and coalescence of Ag nanoparticles at higher sintering
Cu is used instead of sintered Ag. temperatures and pressures. Paknejad et al. [36] mentions the
Finally, micro-sized specimen tests are presented. possibility of decomposed organics from the die-attach which
Zabihzadeh et al. [45, 46, 47] investigate the influence of the cannot outgas. Another reason could be that high sintering
microstructure (pores and silver ligaments) on the mechanical pressure reduces the effectiveness of sintering due to
properties and behavior. Chen et al. [1] compare the fracture prevention of easy penetration of oxygen into the sintering
toughness for specimen on micro and macro scale to determine layer and leads to a worse mechanical connection [36]. In a
size effects. In a subsequent paper, Chen et al. [48] present summary by Paknejad et al. [36], there is no further increase in
micro-compression tests and compare different Ag pastes with shear strength after the last decade—this indicates that limit of
different particle sizes). shear strength may have been reached. Moreover, a higher
shear strength, due to higher densification, does not have to
A. Shear Test lead to a higher reliability and needs further investigation [36].
More detailed information on influences of size distribution,
Especially shear tests are commonly used to determine the processing temperatures, sintering profile and metallization on
influence of sinter parameters like sintering temperature, shear strength can be found in [35, 36]. They are also studying
pressure, time and heating rate on bonding strength [16, 35, 36, the shear strengths of multiple SiC die-attachments on DBC
37]. Fig. 2 shows the influence of pressure and temperature using Ag nanoparticles. The investigation shows a variation in
during the sintering process [16]. For die-attach applications, a shear strength for a multiple die-attachment sample due to
shear mode during thermal cycling is the typical loading. slightly different thicknesses and uniformity of the printed Ag
Yielding of a metallic material like sintered Ag can be seen paste. In addition, a non-uniform pressure occurs on the
microscopically as slip of crystal planes due to shear stresses. different dies although a flexible silicone rubber is used [35]. Li
Another advantage of shear loading is that the deformation et al. [35] describes the failure mode during shear tests as a de-
occurs within the die-attach material, which enables repeatable bonding, which mainly occurs near the metallization surface of
results. [6] Furthermore, it is comparatively easy to perform the SiC die for lower shear strengths. For higher shear
shear tests. Usually, these are realized on a single chip strengths, the same effect can be observed as well as a de-
attachment sample [35, 36, 37]. As the shear stresses are bonding near the surface of the DBC substrate metallization.
strongly dependent one the previously mentioned sintering These types of shear tests assess the mechanical performance
parameters as well as on the composition of the paste, e.g.

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
for different parameters of the sintering process such as B. Tensile Test
temperature and pressure. It is not possible to gain material data Using tensile tests for material characterization for silver
describing creep effects, which is relevant to the subsequent sintering is state of the art. It is possible to characterize the
simulation. In addition, a measurement of strength must be visco-plastic material behavior (creep) for different
distinguished between the strength of bulk sintered Ag and the temperatures. Chen et al. [8, 30] perform such a study at
joint strength [6]. In the previously presented shear test, it is temperatures ranging from 25 °C to 300 °C and analyze the
common to determine the joint strength, including influences as microstructure prior to and after tensile tests with scanning
metallization effects. The strength of the sinter layer itself electron microscopy (SEM). The strain rate is fixed at 0.001 %
presents the minimum strength of the joints in case bonding s-1. They use a mixed hybrid Ag paste with flake- and
strength with the substrate is higher [6]. In commonly used spherical-shaped Ag particles. Fig. 4 shows an example of a
shear tests, the load is applied with a contact tool to shear the tensile specimen with a thickness of 0.4 mm by Weber et al.
semiconductor from the joined substrate. Therefore, a lift-off [10]. The stress-strain curves are gained for different
(distance between contact tool and layer under die) has to be temperatures and identification of creep deformation in sintered
adjusted since a peel off of the die-attach should be avoided. In micro-sized Ag particles even at ambient temperature is
addition, the contact tool and the die has to be in parallel to possible [30].
each other to prevent cracking of the dies [6]. In some studies,
a copper (Cu) substrate is bonded with a Cu dummy die [14,
30]. Knoerr et al. [14] and Siow [6] mention some problems for
these cases due to a deformation of the Cu dummy
semiconductor during preparation, which leads to a changed
load condition from mode II to a mixture of mode I, and II.
Nevertheless, there are other shear tests like those presented
by Li et al. [39] to examine creep behavior at various stress
levels and temperatures. A lap shear structure is built by Fig. 4. Geometry of a flat tensile specimen after sintering [10]
connecting two copper parts with a sintered layer in between
(Fig. 3). A constant load at different temperatures was applied
for creep tests. In this case, a mode II is the dominant failure.
One conclusion of the study is that transient and tertiary creep
range are clearly shorter than the steady-state, and that steady-
state is the dominating part of the sintered lap shear joint. It is
also mentioned that even if creep behavior is investigated
systematically, a macro stress or strain cannot precisely
characterize the die-attach. According to Li et al. [39], the use
of multiscale energy dissipation models could be a possible
solution.

Fig. 3. Schematic of a lap-shear sample

Lei et al. [38] are also using a lap shear test. Compared to
the previously introduced paper by Li et al. [39] they determine
the shear strength for different sintering pressures and use a
nanoscale Ag paste. Pressureless sintering with temperatures
below 275 °C is performed. The clamps for sample fixation are
flexible to avoid the joint from twisting which would lead to
large displacements. The results show a strong bonding
strength and uniform microstructure for a sintering pressure
below 5 MPa [38].
Fig. 5. Stress-strain behavior of porous sintered tensile test specimen a) with a
Siow [6] mentions two disadvantages of the lap shear test. porosity of 3 % and 15 % at RT and b) with 15 % porosity at RT with two
On the one hand, there is an additional bending component, strain rates and at 125 °C with four strain rates between 10-2 and 10-5 s-1 [2]
which results in a peeling stress and underestimate the actual
shear stress. On the other hand, with an increasing length of Herboth et al. [2] use a tensile specimen with similar
overlap, shear strength decreases due to stress concentration at geometry and size like Weber et al. [10] in Fig. 4. The
both ends. When comparing results from different authors and influence of different porosities is investigated using a
papers, such geometry factors have to be considered [6, 49, 50, specimen with 15 % (± 4 %) and 3 % (± 1 %) porosity. The
51]. temperature range then varies between -40 °C to 200 °C and, in
contrast to Chen et al. [30], different strain rates between 10-5

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
and 10-2 s-1 are examined. As a result, stress-strain curves are lower strain rate compared to SnAgCu for a temperature of 150
obtained. The specimen with a porosity of 3 % decreases in °C and a stress of 10 MPa [10].
stress after necking (around 1 % total strain), specimen with a
higher porosity of 15 %, however, have no necking and As described at the beginning of the Chapter II, Caccuri et
therefore a brittle behavior (Fig. 5. a) [52]. Additionally, for al. [32], Gadaud et al. [24] and Wang et al. [22] all use a
both porosities, an increasing strain rate sensitivity above RT stacking method to build up thicker tensile specimen. Caccuri
can be seen (Fig. 5. b) [2, 52]. Usually, the strain to failure does et al. [32] use the LTS 043 04P2 paste by Heraeus with micro-
not exceed 1 % at RT—contrary to higher temperatures at sized silver particles and other solvents. The aim behind the
which tensile elongations up to 12 % can be found, depending stacking method is to gain the same porosity and microstructure
on porosity [2]. like in a sintered joint for a thicker sample. A constant strain
rate of 4 · 10-5 s-1 is used and the tensile tests are conducted at
Nikitin et al. [34], too, measure the stress-strain curves with RT. A major difference to the before mentioned work of
an LTS paste provided by Heraeus on a dog-bone shaped Nikitin et al. [34] is that the sintering pressure has an influence
sample. One interesting result in this study is that the sintering on porosity and therefore on the stress-strain curves, too. For
pressure, which varies between 2.5 and 10 MPa, has no higher density, a higher Young’s modulus, ultimate tensile
influence on the stress-strain curves. An increasing sintering stress (UTS) and ductility are achieved.
temperature instead leads to higher values for Young’s
modulus and the strength of the material. Moreover, an increase Letz et al. [41] use a dog-bone shaped tensile specimen
of sintering time leads to a higher value of mechanical strength. with a thickness adjusted to the one of real joints to avoid size
effects on the material properties. The tests are performed
An influence of sintering temperature on pore size is between RT and 250 °C and a constant traverse speed of 1
identified as well. A sintering temperature of 250 °C leads to an mm/min. As a result, Letz et al. [41] get a strong influence of
increased diameter of the pores compared to 160 °C, whereas sintering pressure and testing temperature on the stress-strain
an increase in drying temperature and time reduces the pore curve. A higher sintering pressure leads to a higher Young’s
size [34]. The sintering pressure is not influencing the material modulus and tensile strength [41]. In contrast to the results
behavior, which is in contrast to the Subchapter A. Li et al. from Nikitin et al. [34], the sintering pressure has no influence
[35], Paknejad et al. [36], Kazahaka et al. [37], Lei et al. [38] which confirms the work of Caccuri et al. [32]. Letz et al. [41]
and Mei et al. [16] state that an increase of sintering pressure get a gradient of porosity in the dog-bone tensile sample from
leads to higher strength. Herboth et al. [2, 52] obtain a different the center to the edges, which leads to averaged mechanical
porosity (Fig. 5. a) due to different process parameters, too, properties along the cross section. Furthermore, rough edges of
which has an influence on the stress-strain curve. However, it is the dog bone could decrease the strength, whereupon it comes
not described if the change in porosity is due to sintering to brittle failure at room temperature [41]. Gadaud et al. [24]
pressure, sintering temperature or a combination of both build upon the work of Caccuri et al. [32] with the same sample
factors. Some issues of tensile testing are described by Nikitin preparation. Tensile tests are examined for different aged
et al. [34], who noticed that different fracture points—due to specimens, meaning that isothermal ageing is performed at 125
initial cracks in the sample and high roughness at sample °C under air up to 1500 h. The results show that Young’s
edges—are gained for the same loading condition. modulus and yield strength are not influenced by ageing at 125
Furthermore, an in-proper preparation and clamping caused by °C and that both parameters are only a function of density. In
the equipment can lead to cracks. contrast to this, the plastic behavior is unpredictable and neither
the ductility nor the UTS are a function of density or ageing
Weber et al. [10] use a dog-bone shaped specimen, similar time.
to Chen et al [8, 30], Herboth et al. [2, 52] and Nikitin et al.
[34], being 120 μm thick with a porosity of approximately 6 %. Wang et al. [22] investigate uniaxial tensile samples of
The number of pores is comparable to tested sintered Ag joints nano-scale Ag paste for stress rates between 0.1 MPa/min and
with a homogeneous distribution across the whole sample. The 100 MPa/min as well as temperatures between RT and 175 °C.
stress-strain curves are examined for different temperatures The thickness of the sample is 80 μm. Tensile strength declines
(RT, 100 °C, 200 °C) and a constant strain rate (traction speed with decreasing stress rate and an increase in temperature. In
1 mm/min). A Comparison between sintered Ag and pure Ag is addition, a slower stress rate and high temperature lead to a
made with the result that sintered Ag has at RT a brittle while large elongation. Suzuki et al. [40] investigate the influence of
pure Ag has a typical elastic-plastic material behavior [10, 41]. the manufacturing processes on the micro-deformation
At higher temperatures of 100 °C and 200 °C, sintered Ag behavior of sintered Ag. Tensile tests are performed, macro-
seems to have an elastic-plastic behavior, too, even though it is deformation properties (stress-strain curve) are determined and
not very ductile (1.5 - 1.8 fracture strain at 200 °C). In addition, fitted to an FEM analysis. A micro sub model is generated
Young’s modulus is influenced by the porous microstructure as which represents the porous microstructure. The data is gained
well. A higher temperature leads to a smaller Young’s by means of focused ion beam (FIB) scanning of continuous
modulus. Weber et al. [10] also perform creep tests at RT up to cross-sections. The macro and micro FEM models are adjusted
200 °C with different loadings between 7.6 to 18.8 MPa. At to fit the micro-deformation properties. Nanoindentation tests
RT, the creep rate is very low and significantly increases with are then performed to confirm micro-deformation properties.
rising temperature. Furthermore, a comparison of sintered Ag The results show that micro-deformation properties of sintered
to SnAgCu solder is made at the same stress level and Ag are independent from the pressure condition for the same
temperature. Sintered Ag has a roughly one order of magnitude sintering temperature. The micro-deformation properties
generated from tensile tests are then verified by a comparison

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
of the load-displacement relation between nanoindentation test a measurement at the surface of the material. This leads to
and FEM simulation. Finally, the results show that, if the different properties compared to other tests due to influences of
macro-deformation properties (stress-strain curve) and the surface.
microporous-structure are known for a certain condition, a
prediction of porosity dependency of the macro-deformation Letz et al. [41] use a Berkovich indenter geometry for a
properties via FEM is possible [40]. nanoindentation test. As surface roughness is a key factor for
correct results, it is important to know that surface roughness
Chen et al. [30] make a comparison between tensile test and only depends on the porosity of a sintered Ag layer for the used
shear test. The tendency of the deformation behavior of both experimental set-up. To exclude surface effects, an indent
tests is almost the same. Fig. 8 shows Young’s modulus and depth with constant values is required (between 1100 and 1700
shear modulus that are both decreasing with increasing nm). Based on a different porosity in the sintered layer, there is
temperature, whereas Young’s modulus is decreasing much a dependency of hardness and yield strength on the location.
stronger than shear modulus. This can be seen for the Poisson’s There is a higher density near the edges and corners and
ratio in Fig. 8, too, which is decreasing with a temperature rise therefore a higher yield strength compared to the center
as well. This implies a morphological evolution in the position. Unlike tensile tests which has a macroscopic uniaxial
microstructure caused by a change of the aspect ratio of voids. stress state, nanoindentation have a complex stress state in the
deformation zone under the indenter tip [41, 42]. Hence, it is
necessary to have a constraint factor between hardness H and
yield strength σf. Letz et al. [41] use the conversion for metals
given by Tabor [53]:

 H ≈ 3 ·σf 

For high testing temperatures, the maximum load as well as

the slope of the unloading curve are strongly decreasing in the
load-displacement curve compared to low temperatures. The
Fig. 6. a) Young's and shear modulus at different temperatures and b) the
temperature effect on the Poisson’s ratio of sintered Ag paste [30] nanoindentation tests are also conducted for different strain
rates (0.05, 0.023, 0.01, 0.005, 0.002 and 0.001 s-1). The
equation used in this paper relates strain rate with stress. (1) can
C. Nanoindentation substitute stress with hardness. In addition, the sintering
Mroßko et al. [33] use a nanoindentation test for gaining pressure shows an influence on strain rate sensitivity, which is
material properties of sintered Ag layers. Therefore, a the inverse of the stress or hardness exponent due to different
calibration process is essential as well as a material with a porosities. For low values, the strain rate sensitivity is nearly
known elastic modulus. Layer thicknesses of 22 and 38 μm are independent from the sample porosity whereas higher strain
investigated as well. It is important that the indents are as small rate sensitivities lead to an influence of the strain rate on
as possible to obtain pure bulk material behavior that is big hardness [41]. The strain rate is highly non-uniform due to the
enough to avoid surface and interface effects [33]. To perform non-uniform stress field in the indentation region in the sintered
a Berkovich indenter, a depth of 250 nm is a good compromise layer. A typical method in literature is to average stress and
[33]. The indenter shape of Berkovich and Vickers indenter can strain values in the indentation region [42].
be approximated as a rotationally symmetric and an indenter tip Letz et al. [41] compare the results of tensile testing and
shape function can be generated [33]. The function serves as nanoindentation. The yield strength for nanoindentation tests is
input data for an FE model. According to Letz et al. [41], it is higher which may be explained by the rough edges of the dog
important for the samples to have a smooth surface in order to bones (Chapter B). They are leading to stress concentrations at
correctly estimate the contact area of the indenter tip. A four- the dog bones and therefore to lower yield stresses. According
parameter Ramberg-Osgood model with multilinear elastic- to Letz et al. [41] this could imply that material parameters
plastic material law and kinematic hardening is used in the gained by tensile testing are not sufficiently representing the
FEM simulation. Simulations are made with different properties of the joints.
parameters until they fit with the indentation depth versus force
curve of the performed tests [33]. Leslie et al. [42] mention that the ratio between indentation
depth and material/layer thickness can affect the material
The main advantage of a nanoindentation test compared to properties, although this is not the case for their study. For a
tensile testing is described by Heilmann et al. [28] as an easier thin layer the substrate can affect elastic and plastic material
and repeatable fabrication process (homogeneity, size-effects). properties. Information that is more detailed can be found in
However, difficulties are also mentioned by Heilmann et al. [42, 54]. Leslie et al. [42] investigate Ag which is pressureless
[28] like the size of the indenter since a small indenter is only sintered between two copper plates with a 5 μm barrier layer of
cracking ligaments. Hence, there is an influence by local nickel and 1 μm of silver. This prevents an over-oxidation at
effects. In contrast to this, a large indenter is assumed to the interface. The average porosity is 30 %. Further details on
compress the material to bulk Ag. This is assumed by the test can be found in Leslie et al. [42]. A constant force test
Heilmann et al. [28] to happen in the work of Mroßko et al. as well as a constant strain rate test is conducted. For constant
[33]. Heilmann et al. [28] presume that nanoindentation is just strain rate, a Berkovich indenter is used and for constant force,

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
a spherical indenter in addition. An approach by Lucas and Zschenderlein et al. [44] present a micro bending test
Oliver [55] with a constant normalized load rate is used instaed (MBT) with the idea of bending a released double dog bone
of a constraint strain rate. For the constant force test, the (DDB) specimen by a nanoindenter of sintered Cu
spherical indenter predicts a lower creep resistance compared nanoparticles, featuring a thickness of 1 μm (Fig. 8). A
to the Berkovich indenter. The stress levels are much higher in nonlinear deformation results in an uniaxial tensile stress state.
relation to testing of bulk specimen. Smaller stress levels Compared to a tensile test in which a force-displacement curve
cannot be examined by nanoindentation. However, to get to the can be easily transformed to a stress-strain curve, this is far
lower limit, spherical indenters with larger radius are necessary more difficult for the MBT due to the nonlinear bending. The
[42]. Both indenters are only suitable for use at high stress easiest way of getting the DDB strain would be a direct
levels, regardless of the strain or force rate. A variation of the measurement of the strain, e.g. with the Mirau interferometer.
tip geometry leads to a change of the stress range. The However, Zschenderlein et al. [44] present a way to transform
indentation of the spherical tip is larger and therefore more the force-displacement curve of the MBT to a stress-strain
averaged properties are obtained. The Berkovich indenter, on curve of the material. This is necessary if the usage of
the other hand, is expected to provide a more localized additional test equipment is restricted, e.g. in case of limited
behavior. This presents a rather homogeneous Ag material space. An empirical approach is used to calculate the beam
behavior than for porous Ag material. This is shown in Fig. 7, stress from bending force and another one for the strains from
where the red spots marks contact between the individual silver warpage. The relation between the parameters is made with an
particles. A larger spherical indenter seems more suitable than FE model. Zschenderlein et al. [44] indicate that all results of
a Berkovich indenter [42]. Leslie et al. [42] outline a crucial an MBT have to be verified with other tests such as
point which has to be further investigated to determine the nanoindentation or miniaturized tensile tests. Even if the
applicability of nanoindentation tests. Is the localized current tests were performed with sintered Cu, it is conceivable
deformation field of the nanoindentation test comparable to the to use Ag sintered samples as well.
creep behavior of uniaxial tensile testing? Apart from this,
another disadvantage is that creep behavior cannot be examined
over a wide stress range (only for high stresses).

Fig. 8. Principle of micro-bending test [44]

D. Micro-sized Specimen
As already described, there is a considerable influence of
the microstructure on mechanical behavior for all previously
Fig. 7. Schematic concept of indentation for a) spherical indenter and b) introduced test methods. Micro-tests can help getting a deeper
conical (Berkovich) indenter [42] understanding of the behavior. Zabihzadeh et al. [45, 46, 47]
study the microstructure of porous silver layers like grain size,
Long et al. [43] compare a nanoparticles Ag sintered layer defect structure and porosity by scanning and using
with an electrically conductive silver adhesive (ECA) through transmission electron microscopy (SEM, TEM) and use X-ray
nanoindentation with a Berkovich indenter. The influences of nano-tomography for 3D information of the porosity [45]. In
the microstructure are investigated. Strain rates are varied the following work, Zabihzadeh et al. [46] use ptychographic
between 0.1 s-1 and 0.2 s-1 and indenter depths are between X-ray computed tomography (PXCT) to get higher resolution
1100 nm and a maximum of 2000 nm. As strain rate influence compared to the spatial resolution of 150-200 nm of X-ray
on material behavior is examined and due to the fact that a low nano tomography [45]. Ptychography, also known as Scanning
strain rate from the beginning of indentation requires a lot of X-ray diffraction microscopy (SXDM) is a high-resolution
time, the strain-rate jump technique by Maier et al. [56] is used. imaging technique that provides curves of the phase and
The indentation process starts at a relatively high strain rate and amplitude of the complex transmission function of the test
then jumps to a desired strain rate at a preferred indentation sample [46]. A resolution of 16 nm in 3D can be achieved but
depth. Due to a different microstructure, the hardness of resolution of 35-55 nm is used [46]. The 3D images are used in
sintered Ag nanoparticles is more sensitive to strain rate than a subsequent paper of Zabihzadeh et al. [47], too, as an input
ECA. For both materials, the effect of strain rate on Young’s for FEM to calculate the mechanical properties for samples
modulus is moderate, but still there is a slight decrease with with different porosities. In addition, the mechanical properties
increasing indentation depth. [43] and material behavior are studied by in-situ tensile testing [46].
Thin layers of 100 μm of nano-grain sized (30 to 50 nm) silver

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
laminate are analyzed. Sintering pressure varies between 4 to sintering pressure, temperature or time leads to a decrease of
12 MPa and temperature from 210 to 300 °C. X-ray the porosity. In addition, this increases the size of the grains
measurement is performed at a cylindrical shaped pillar with a and ligaments [46]. Furthermore, TEM investigations prove the
diameter of 20 μm and a height of 40 μm, machined by FIB presence of dislocations and twins inside the silver grains [45].
milling for analyzation of the pores [45]. In the following
paper, Zabihzadeh et al. [46] measure pillars with a diameter of TABLE II.
4 to 5 μm and heights of 15 to 20 μm, machined via FIB, too. SILVER LIGAMENTS [46]
With different programs and plugins, it is possible to get
information like branch length and Euclidean distance between
the extreme points of the branches of the microstructure, which Parameter Value
gives an estimation of the tortuosity of the pore [45]. Only a
sub-volume of the pillar is analyzed. Avg. diameter 260 to 550 nm
In-situ tensile testing is performed with a miniaturized Avg. branch length 290 to 2200 nm
tensile machine (MTM) and an X-ray beam for measurement of
the microstructure. For further details, see [45, 46] and for the Avg. volume grain size 120 to 350 nm
test equipment see [57, 58]. The samples are dog-bone shaped
with a width of 2000 ± 5 μm, a thickness of 25 ± 5 μm and a
When it comes to characterization of mechanical properties,
gage length of 1500 μm. The specimen is cut with a water-jet
the normalized apparent elastic modulus E*/E, where E* is the
and then attached to grips with glue. Applied macroscopic
elastic modulus of the porous films and E the Young’s modulus
strain is measured with digital image correlation (DIC) and
therefore a speckle pattern with black color spray is required. of bulk silver, is plotted as a function of relative density ρ*/ ρ.
Images during the test are compared to a reference image and ρ* is the apparent density of porous Ag and ρ of bulk silver. It
thus, the displacement and strain field can be determined. The is distinguished between two regimes; regime I with ρ*/ ρ < 0.6
and regime II with ρ*/ ρ > 0.6. For the second regime, a linear
tensile tests are performed at a strain rate of 0.015 % s-1 and at
fit can be found where normalized elastic modulus increases
RT. The macroscopic material properties can be determined by
with relative density, see Fig. 9 a. This applies for yield stress,
the stress-strain curve, e.g. yield stress and UTS [46]. Analyses
too, which considers the whole cross-section and does not
shows a porosity between 22 % and 47 % for different sintering
exclude the internal porous structure (Fig. 9 b).
parameters. There is less influence for other parameters like
connectivity, mean diameter, mean branch length and volume-
to-surface ratio of the pores, see Table I [46]. The distribution
of the pore branch length is also generated and shows a
decreasing number of pores with increasing pore branch length
[46].

TABLE I. P
POROUS MORPHOLOGY [46]

Parameter Value

Porosity 22 – 47 %

Connectivity > 90 %

Volume-to-surface ratio 50 to 60 nm

Euclidian distance 170 to 260 nm

Avg. diameter 300 to 350 nm

Avg. pore branch length 200 to 300 nm

After analyzing the pores, the silver phase is also

investigated, see Table II. A comparison between the width of
the ligaments and grain size is made and shows, that a silver
ligament consists of multiple grains (two to three) across the
thickness. The averaged diameter of the ligaments is smaller
than the averaged branch length, which indicates a
heterogeneous morphology with a non-uniform pore
distribution [45, 46]. Different sintering parameters result in a
Fig. 9. Normalized elastic modulus (a) and yield stress (b) versus relative
difference in pore fraction (relative density). An increase of
density according to [46]

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
 This behavior is not expected and does not follow the To determine the fracture toughness of macroscale tests, a
conventional scaling laws for porous metals [46]. In further tensile test is performed with a displacement rate of 100 nm/s.
evaluations, a local yield stress (inside the ligaments) is also [1] The microscale specimen are manufactured, using a FIB to
defined as the applied stress divided by the relative density. mill the cantilever bending specimen out of the sintered Ag
Decreasing porosity leads to an increase of local yield stress, (Fig. 11). One end is then fixed to the rest of the sintered Ag
which also indicates an increase of grain and ligament size whereas the other is free for loading. In addition, FIB is used to
(inverse strengthening effect) [46, 47]. The local stress defined generate a pre-crack. The different width of the specimen can
by Zabihzadeh et al. [46] does not represent the stress be seen in Fig. 11 where d) shows the induced pre-crack and
concentrations in the porous structure, which typically occurs load point [1]. Finally, bending tests are performed using a
but rather represents an average stress value inside the nanoindenter with a spherical tip (radius 500 nm) to load and
ligaments. To sum up, it can be stated that the results indicate bend the cantilever specimen. To avoid strain rate effects on
that there is no influence on the local flow stress in the silver fracture toughness, tests are conducted under ambient
ligaments controlled by grain boundary strengthening or by conditions with the same displacement rate as the macroscale
ligament size effects [46]. specimen [1].
Chen et al. [1] investigate the effect of microporous Ag
specimen size on fracture toughness for microscale and
macroscale specimen. Fracture toughness (KIC) is important
since it describes the ability of a material to resist crack
propagation in case of an already existing crack. Because the
microstructure of sintered Ag is porous, it is even more
important to assess fracture failure according to Chen et al. [1].
As most of the material parameters are depending on porosity
and microstructure, as already mentioned in the previous
chapters, it is expected that there is also an effect of size for
fracture toughness. Hence, tests were conducted at macroscopic
scale with conventional middle-cracked tension tests (M(T)),
see Fig. 10.

Fig. 11. Fabrication process of micro-cantilever beams from porous Ag with

FIB milling from a) the vertical direction, b) from the horizontal direction and
c) micro-cantilever beams with induced pre-crack and load point [1]

The most important statement in the work of Chen et al. [1] is

that sample size influences fracture toughness (Fig. 12).
Fig. 10. Sketch of specimen shape before tensile testing with induced pre- Compared to bulk silver, which has a fracture toughness
crack [1] between 70 - 105 MPa m1/2, the result of macroscale specimen
is notably smaller with 12.24 ± 1.92 MPa m1/2. One likely
Additionally, it is important to study fracture toughness as reason as stated by Chen et al. [1] could be that the material
well as deformation during crack propagation with sizes of the near to the grain boundaries of sintered Ag is less crystalline
specimen on the same order of magnitude as the microporous than bulk silver. Hence, less energy is required for separation
structure [1]. According to Chen et al. [1], it is common to along the grain boundaries and between two grains. Another
perform nanoindentation tests to study fracture toughness at reason could be that the load-bearing surfaces are decreasing
small scales where a sharp indenter causes a radial crack in with increasing porosity, which consequently leads to a
brittle material. However, it is difficult to use this test for decrease of fracture toughness. When examining microscale
porous material in order to characterize fracture toughness due specimen, it can be seen that smaller specimen result in smaller
to higher complexity and uncertainty in the experiment— values for fracture toughness. With an increasing specimen
making it difficult to generate reliable data. Chen et al. [1] use size, fracture toughness converges to the value of the
micro-cantilever bending tests to investigate fracture toughness. macroscale sample. Besides, an influence of porosity and
It is determined from the initial crack propagation based on the specimen width on fracture toughness is perceivable, even if
material’s volume directly under the crack root. This is the influence of porosity is more distinct for smaller specimen.
independent of the material’s density, porosity or the structure In addition, FEM simulations are performed with a 2D model
and therefore suitable for sintered Ag. For the micro-cantilever in which the pores are modeled as perfect spheres, applying a
bending tests, specimen featuring widths of 5 to 20 μm are used regular pattern to investigate stress concentration at the crack
to study the size effect on fracture toughness on microscale. A tip. Stress concentration at the crack tip as well as the size of
hybrid Ag paste is used which consist of flake-shaped micro the plastic deformation zone decreases with a decrease of
particles and spherical-shaped nanoparticles. The sintering specimen size [1].
process is performed at a temperature of 250 °C and without
pressure for both tests. A tensile specimen is produced Chen et al. [48] perform a micro-compression test to
according to Chen et al. [30] and a pre-crack is induced (Fig. evaluate mechanical properties of sintered Ag. Four Ag pastes
10). are used with different particle sizes.

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
 shape with an average diameter of 8 μm and a thickness of
260 nm. For this reason, the tests include nano and micro
particles. Similar to a previous work of Chen et al. [1], a block
of Ag (3 mm x 3 mm x 1 mm) is fabricated with same sintering
conditions and FIB is used to produce the microscale
compression specimen. The specimen features dimensions of
5 μm x 5 μm x 15 μm (Fig. 13 a). A nanoindentation system is
used to load the specimen with a constant displacement rate of
50 nm/s. A spherical tip with a tip radius of 500 nm is used like
in [1]. The stress-strain curves of the micro-compression test
are shown in Fig. 13 b. It also demonstrates the influence of the
different pastes on mechanical properties like Young’s modulus
and yield strength. The reason for the smaller values of the
nano particles can be understood by analyzing the cross
sections. The size of voids is larger which indicates lower
Fig. 12. Fracture toughness of sintered Ag for different sintering quality compared to the other pastes. These voids lead
samples from micro- to macroscale [1] to a larger deformation and therefore reduce the compression
Three of them have as spherical shape with an average stress and Young’s modulus [48]. Table III shows a summary
diameter of 50, 300 and 2500 nm, and the fourth has a flaked of all prestented test methods for material characterization and
an assessment of different criteria made and classified by the
author.

TABLE III. S
SUMMARY OF MATERIAL CHARACTERIZATION TESTING

Shear Test Tensile Test Nanoindentation Micro-sized specimen

Micro-
Test Micro-
Lap shear compres-
Shear Test Flat samples Nanoindentation Micro bending test cantilever
test sion (nano-
bending test
indentation)
Determination of
material behavior
Mechani-
(e.g. creep) or Material Material Fracture Material
cal perfor- Material behavior Material behavior
assessing of behavior behavior toughness behavior
mance
mechanical
performance
Fabrication process 5a 4 3 5 3 1 1
Repeatability of
3 4 3 5 5 - -
results
Porosity/microstruc-
ture comparable to yes yes yes yes yes yes yes
application?
Effort: 5 Effort: 3 Effort: 3 Effort: 4 Effort: 3 Effort: 1 Effort: 1
Effort, costs and
Costs: 5 Costs: 4 Costs: 4 Cheap: 4 Cheap: 2 Cheap: 1 Cheap: 1
time for testing
Time: 5 Time: 4 Time: 4 Time: 4 Time: 4 Time: 1 Time: 1
Complexity of 5 for
measuring necessary stresses 4 3 4 3 1 1
data 0 for strain
1. Geometry of
Different nanoindenter has an 1. Fracture
Additional 1. Verification with
fracture influence on results toughness
bending other test necessary FIB used for
Influence points due to 2. FEM is necessary for depends on
component 2. Nonlinear milling
of other initial cracks gaining complete material sample size
and geome- bending and geometry
Note / Issues parts like and high behavior 2. FIB used for
try factors therefore a more (expensive
metalliza- rough-ness at 3. Multiaxial localized milling
of sample extensive trans- and time
tion sample edges deformation field geometry and
has to be formation to stress- consuming)
for same 4. Creep behavior cannot induced pre-
considered strain curve
loading be examined for small crack
stresses
State of the art or State of the State of the State of the
State of the Art Research Research Research
Research art art art
Currently applicable
Yes Yes Yes Yes Possible No No
for industry

a. A value of five is best value and zero the worst.

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
 At elevated temperatures (125 °C), an increase of porosity
on the fracture surface (Fig. 14 b) occurs, and a dimpled
fracture cannot be observed [2]. The material behavior at
macroscopic scale seems elastic-plastic and not brittle [10]. In
addition, plastic deformation is not the only active deformation
mechanism as dislocation creep is also observed. Based on
grain boundary porosity, intergranular fracture occurs, which is
a dominant mechanism for Ag at temperatures above 200 °C—
depending on grain size. As a result, fracture surface changes
with temperature. For low porosity, the behavior of sintered Ag
is more similar to bulk Ag, and an elastic-plastic behavior with
necking at macroscopic scale occurs [2]. For tensile testing of
Fig. 13. a) micro-compression specimen fabricated by FIB and b) load- bulk silver and for sintering with very low porosity, the fracture
displacement curve of the micro-compression test for different sintered Ag occurs at the area with highest constriction. However, for
pastes [48]
higher porosities, the area with the highest number of pores will
fracture [10]. In general, the strength of sintered Ag strongly
III. PHYSICS OF FAILURE decreases with higher temperature. In contrast to this, the
fracture strain is increasing [2, 10]. Chen et al. [1] observe the
The behavior of sintered Ag under loading is different same failures as Herboth et al. [2] with intergranular fracture
compared to bulk Ag and was partly described in the previous for bending instead of performing tensile tests.
chapter. This chapter gives an overview for further
understanding of the physics of failure during material To investigate the fatigue crack, active and passive thermal
characterization. The mechanical behavior of sintered Ag, cycling (aTC, pTC) are typically used for a power device. The
especially at RT, is brittle while pure bulk Ag has a typical setup consists at least of a substrate, a sintered Ag die-attach
elastic plastic behavior [2, 10]. Herboth et al. [2] and Weber et and a semiconductor. The physics of failure for pTC tests can
al. [10] analyze the fracture surface and verify that brittle be seen in Fig. 15, in which thermal shock tests from -40 to
behavior only occurs at macroscale. No necking is observed for 125 °C are performed by Herboth et al. [2]. The difference in
the specimen fracture surface with a porosity of 15 %, although the CTE of the silicon die compared to the sintered Ag layer
this is expected due to the stress-strain behavior (Fig. 5). A and the copper substrate with Ag surface leads to thermal
detailed view on the microscopic scale shows a dimpled fatigue. Cracking starts at the edge of the die and propagates
fracture at RT, which indicates an elastic-plastic behavior (Fig. towards the surface of the substrate along the weak points in
14 a). The necking behavior in the stress-strain curve is the interconnection. In addition, several secondary cracks can
localized at the Ag ligaments of the microstructure and it leads be seen. Intergranular fracture occurs like in the material
to an intergranular fracture under tensile loading. According to characterization tests mentioned before [2].
Weber et al. [10], this can be explained by understanding the
thermodynamic system, which tends to obtain the condition of
the lowest enthalpy—this is the reason why sintering works.
That means that Ag particles featuring a high surface energy
tend to reduce this energy by diffusion, which finally leads to
formation of rounded structures. In contrast, non-rounded
structures at the sinter necks must come from plastic
deformation, e.g. during tensile testing. The loading for bulk
materials under tensile testing is uniaxial until tensile strength
is reached. Besides, necking occurs which results in a
multiaxial stress state. For sintered Ag on the contrary, the Fig. 15. a) Crack in porous sintered Ag joint for 15 % porosity after 2000 pTC
stress state is multiaxial from the beginning due to the at -40 to 125 °C and b) enlarged view on the main crack path and horizontal
microstructure. Each pore acts like a notch, leading to both a secondary crack [2]
lower fracture stress and strain compared to bulk Ag [10].
Power cycling or active thermal cycling (aTC) on the other
hand means that the semiconductor is operated actively with a
current, which leads to a temperature difference ΔT and
therefore to fatigue. The device under test is similar to the one
for pTC testing, however, e.g. bond wires or clips/ribbon at the
top of the die are essential. Clips/ribbons can be sintered with
Ag or soldered [14, 59]. A constant temperature difference is
regulated by adjusting the current. Dai et al. [7] investigate a
sintered Ag die-attach with SiC diodes. Compared to pTC tests,
in which the crack propagates from the edges to the middle of
the semiconductor, the crack initiates from the central area of
Fig. 14. Fracture surface of sintered Ag with a porosity of 15 % and a strain the sintered die-attach towards the edges. According to Dai et
rate of 5·10-5 s-1 at a) RT and b)125 °C [2]
al. [7], the cracks are almost vertical to the die-attach interface

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
and look like mud cracks formed when muddy sediment dries. observed using scanning acoustic microscopy (SAM) and X-
Propagation of the vertical cracks seems to slow down with ray analyses [9, 14, 33, 61, 65]. Le Henaff et al. [64] use
increasing power cycles, but some of them have further another approach by measuring the change of the thermal
extended into the Cu layer of the substrate. In addition, resistance, which increases due to crack propagation.
delamination of the die-attach, like in pTC tests, occurs at the Therefore, a 20 % increase of thermal resistance is used as
edges. Schmidt et al. [60] and Weber et al. [61] examined the failure criterion [64]. Sakamoto et al. [11] perform shear tests
same failure mode and also mention that failure location is after pTC tests to determine joint strength. The used failure
shifted towards the substrate. In addition, Weber et al. [61] criterion is a 50 % drop of joint strength.
detect formation of cracks or delamination at the interface
between the sintered Ag layer and DBC metallization, or As an example, Herboth et al. [2] achieve 2500 cycles until
between the sintered Ag layer and die metallization in the area the failure criterion of 80 % bonding area is reached with a
of the cracked DBC. porosity of 15 % and a TST of – 40 to 125 °C. A lifetime curve
(Fig. 16) is generated by Herboth et al. [2] and compared for
different porosities and to lead-free tin-based solder. For
IV. RELIABILITY TESTING sintered Ag as well as for the solder, the crack propagation is
As already described in the Chapter III, aTC and pTC are almost constant. In addition, an influence of porosity can be
mostly used for reliability testing and investigation of fatigue observed. A reduction of porosity to 3 % leads to more than
cracks. This is based on testing the thermomechanical 10,000 TST cycles.
reliability of sintered Ag joints. The difference between aTC
and pTC tests is the heating source. For pTC it is an external
source in which a two or one chamber test machine is used. In
contrast to this, the heating source for aTC tests is internal due
to operating the semiconductor. This results in different dwell
times. To perform pTC tests, the temperature change takes
several minutes or hours, whereas for power cycling only a few
seconds are usually applied [59]. The main advantage of pTC
tests is that multiple devices can be tested at the same time with
the same equipment, and that it is only limited by the space of
the chambers. According to Hutzler et al. [59], an average
number of tested modules in power cycling at the same time is
two. Power cycling on the other hand reflects the failure mode
which should occur in the field. In addition, in-situ
measurements are possible. A typical end-of-life criterion is an
increase of 20 % of the voltage, temperature swing or thermal
resistance [59]. Fig. 16. Crack propagation in sintered Ag (porosity of 3 % and 15 %) and
According to Heilmann et al. [28], pTC tests are sufficient soldered joints from MOSFET on copper substrate at pTC -40 to 125 °C [2]
for validation of the chosen process parameters and technology
to guarantee that a module will withstand a defined minimum According to Knoerr et al. [14], crack propagation is rarely
of temperature cycles. To develop a lifetime model, this is not uniform for solder joints due to tilting of the chip, whereas
sufficient and it is crucial to fully understand the failure sintered Ag interconnections show a comparatively
mechanisms (physics of failure). Especially for materials like symmetrical progress of failure. Furthermore, most of the
sintered Ag, for which the material behavior is highly time and sintered specimen investigated by Knoerr et al. [14] and Chew
process-dependent. [28] Lap shear tests are conducted by Li et et al. [65] fail due to delamination between sintered Ag and
al. [3], Chen et al. [62] and Tan et al. [63] to investigate the chip metallization (adhesion failure). Cohesion failure inside
ratcheting effect of sintered nano Ag pastes and to find a the sinter layer occurred for more durable sintered samples, too.
relation between creep and fatigue damage. Heilmann et al. This can be explained by an increase of the strength for higher
[28] develop an isothermal bending fatigue test for substitution density. Otherwise, Young’s modulus increases as well and
of thermal cycling (pTC). reduces elastic strain, which results in a smaller lifetime [14].
Knoerr et al. [14] correlate both effects and identify that the rise
in strength is faster than the one for the Young’s modulus.
A. Passive Temperature Cycling (pTC) Chew et al. [65] use a bending test before and after 2,000
Application of pTC tests are discussed at first. Herboth et cycles, which is a quick test to examine the bonding strength of
al. [2] use a thermal shock test (TST) for reliability testing. The Ag-sintered joints. The substrate, Ag layer and die are bend on
temperature is changed from -40 to 125 °C or 180 °C a cylinder. The complete package is bended around a cylinder
respectively for a certain time. Le Henaff et al. [64] change the if there is a high bonding strength, see Fig. 17. In case of a low
temperatures abruptly every 30 min. The occurring failure is bonding strength, the substrate is bended around and the die
shown in Fig. 15 and discussed in the Chapter III. For pTC detach from the Ag layer.
tests, the typical failure criterion is an 80 % bonded area of the
interconnection, corresponding to 20 % delamination [2, 14]. If
the bonded area is smaller than this, the power electronic is
expected to have reached end of life. Crack propagation can be

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
 %. Out of 27 wire bonds, two had no more contact at 950,000
cycles, however, no failure criterion was reached [60].
Compared to the previously introduced authors, Weber et
al. [61] do not keep current constant but rather maximum
temperature and the operating range over the entire test. Hence,
current has to be adjusted continuously. The used test
temperatures feature a temperature swing of 150 °C from 30 to
Fig. 17. Schematic diagram of bending test [65] 180 °C. The time of switched-on current is 1.8 s, and cooling
time varies for each cycle between approximately 3 to 6 s. The
In general, it is often difficult to compare results from samples are tested until an electric short occurs. According to
different authors with the same problems like mentioned in Weber et al. [61], the lifetime of sintered Ag joints compared to
Chapter II. Different process parameters of the sintered Ag solder joints is 10 to 100 times higher and, depending on the
interconnections result in different material properties process, more than 4 Mio. power cycles are achieved. Different
(porosity, Young’s modulus). Additionally, pastes (micro, sintering pressures and thicknesses of the sintered Ag layer of
nano), substrates (Si, SiC), metallization and other parameters 50 and 150 μm are investigated. Thinner interconnections have
have also an influence. a higher lifetime than thicker ones, which is in contradiction to
the typical assumption that a thicker layer leads to higher
B. Power Cycling / Active Temperature Cycling (aTC) lifetime. The paste used by Weber et al. [61] is intended for
The occurring failure mode and principle experimental set- thinner interconnections (50 μm) and therefore some issues for
up is explained in the Chapter III. Knoerr et al. [14] use a the 150 μm layer can be expected like in homogeneities (voids,
temperature difference of 130 °C from 45 to 175 °C regulated inclusions) due to a worse combustion of the sinter additives.
by current, which is alternatingly turned on and off. Cycle time Finally, the work of Herold et al. [66] compare the power
is 30 s in total. End of life is reached if forward voltage or cycling capability of SiC and Si diodes. Solder is used instead
thermal difference increases by 20 %. The reliability of sintered of sintered Ag, but the results serves as first indicator for
Ag devices was 208,000 cycles until the failure criterion was devices with sintered Ag. A junction temperature swing of 80
reached and the failure occurred on the top side K ± 6 K and a max. junction temperature of 135 to 150 °C is
interconnection. used for the SiC-Schottky diode. Constant on- and off- duration
Dai et al. [7] use a cycle time of 7.5 s with a 5 s heating times are used in the range of 1-2 s. For comparison, Si-IGBTs
phase where current is turned on and 2.5 s with current turned are used with a switched-on current of 2 s and a junction
off. This leads to a temperature range from 50 to 200 °C and temperature swing of 80 K ± 2 K. The end of life criterion for
results in 598,000 cycles—and the test was still running. The all performed tests is a 20 % increase of the transient thermal
thermal performance is determined by in-situ transient thermal resistance after heating. Power cycling testing shows that a
impedance measurement, meaning that the used diode is heated soldered SiC die has a reliability of about three to four times
up with a current of 6 A for 80 s to reach thermal equilibrium. lower than a soldered Si die for similar rated current and
Switching off the heating current results in a change of the voltage class. One reason therefore is the smaller chip area of
forward voltage during the cooling stage of again 80 s. With SiC compared to Si devices. Another one the 41 % higher
the previously determined diode voltage characteristic curve, plastic strain inside die-attach at the edges of the chip for the
the junction temperatures can be calculated [7]. According to same geometry, which is calculated by simulation. For sintered
Dai et al. [7], the effective thermal conductivities are generated Ag die-attach and SiC chip, a similar behavior can be expected
due to correlation of junction temperatures and generated but has to be proven [66].
junction temperatures in FEM simulations. The deviation is
between 5 to 10 % because of noise during in-situ C. Lap Shear Tests
measurements. In a lifetime curve, thermal conductivity can Li et al. [3], Chen et al. [62] and Tan et al. [63] investigate
then be plotted versus power cycle number. The failure the high temperature ratcheting behavior of sintered nano Ag
criterion is set for a thermal conductivity of 30 W/(m·K) [7]. pastes at temperatures of 25 up to 325 °C to study the
relationship between creep and fatigue damage in the low-cycle
Hutzler et al. [59] use a DBC substrate with a SiC diode,
regime. Ratcheting in this case means that the average strain of
aluminum (Al) bond wires and heat sink for power cycling. An
max. and min. per one cycle accumulates only in one direction
indirect temperature measurement and a thermal imaging is
if it comes to an asymmetric cyclic loading. Ratcheting is also
used which shows a deviation of about ± 3 °C. The temperature
known as “cyclic creep” according to Zhang [67] due to a
is measured using the direct proportionality between voltage of
monotonic increase of strain like creep. On the other side, creep
the semiconductor and temperature at a constant current. A
deformations occur typically at temperatures above
disadvantage of thermal imaging is the optical accessibility
approximately 30 % of the melting point in K of the material.
which is reduced due to the bond wires [59]. Schmidt et al. [60]
Ratcheting or so-called “cyclic creep” is also observable at
perform power cycling with a module consisting of SiC
lower temperatures. In addition, creep results of long-term
MOSFETs, DBC substrate, Al bond wires and no heat sink.
stress and is therefore time-dependent whereas ratcheting can
The phase in which current is switched on and off is the same
be time-independent as well [67].
with 2 s. End of life is reached in case of an increase of
temperature swing by 20 % or drain-source voltage drop of 5

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
 The tested lap shear sample can be seen in Fig. 3 and cyclic Compared to the induced strain by pTC where degradation
shear tests are conducted for different temperatures, mean shear occurs from all sides nearly constantly [14, 61], for the
stresses, shear stress amplitude and peak stress dwell times. isothermal bending degradation occurs only from two sides. In
The results of these papers [3, 62, 63] can be summarized as addition, crack propagation is asymmetrical due to shifted
follows: placement of the dies aside of the center for investigation of
more dies at the same time. This leads to a negligible outer
• Ratcheting behavior can be divided into three stages: crack frontline compared to the dominant inner one, which his
transient, steady state and tertiary. The steady state stage is the comparable to thermal cycling [28]. For isothermal bending, a
most dominant part for evolution of ratcheting strain. spring steel coated with Au / Ni is used as substrate. The
• Ratcheting behavior at RT: ratcheting strain increases samples are cycled isothermally at 1 Hz, RT and 125 °C.
with increasing shear stress amplitude and/or mean shear stress For thermal cycling, delamination starts typically from the
but shakedown is independent of the value of the parameters. edges to the center of the die and is within the sintered Ag layer
• Ratcheting behavior at high temperatures: ratcheting close to the substrate. The same failure mode can be found for
strain depends greatly on shear stress amplitude and/or mean the isothermal bending test and therefore the testing method is
shear stress. A higher level of the values leads to higher capable of substituting thermal cycling [28]. Furthermore,
ratcheting strain and ratcheting strain rate. investigations of the failure mechanism on material level are
performed and needs further investigations. Bulk plasticity
• Long dwelling times: During the holding time of the within the silver grains is examined in the bulk material.
stress amplitudes in each cycle, a time-dependent creep strain Electron backscatter diffraction (EBSD) is conducted to detect
can be observed and becomes a dominant part of the ratcheting the existing large dislocations and the beginning of sub-grain
strain. This is also temperature-dependent and results in a low formations. In addition, plasticity is determined by measuring
influence on fatigue life at 225 °C and increases clearly above the misorientation of paths within a silver grain compared to
275 °C. Creep damage becomes the dominant failure mode of the initial state due to a continuous orientation change. This is
the joint. caused by deformed lattice due to dislocation pileup and
indicates clearly plasticity [28]. As higher temperatures are not
D. Bending Test investigated, Heilmann et al. [28] mention that it may be
A more advanced reliability testing method is introduced by possible that it comes to crystal recovery at higher temperatures
Heilmann et al. [28] to substitute and accelerate thermal and dislocations could annihilate or reorganize themselves to
cycling (pTC). In Fig. 18 a, an induced strain by thermal cycles form sub-grains. Hence, the suggested investigation of
due to CTE mismatch can be seen. In addition, Fig. 18 b shows misorientation within a silver grain may not be significant but
the principle of inducing strain by isothermal bending. It is further data for verification is needed [28].
essential, that the failure mechanism of the isothermal bending
corresponds to pTC. E. High-temperature Reliability
Isothermal bending is a strain controlled test where the Chua and Siow [4], Chen et al. [68], Chew et al. [65], Choe
bending is directly applied to the substrate. Hence, shear et al. [69], Egelkraut et al. [70] and Kisiel et al. [71] investigate
stresses are generated within the sintered Ag die-attach. The the behavior after ageing at high temperatures. Chua and Siow
accumulated equivalent strain over one full cycle is used as a [4] study the microstructure and porosity of pressureless
failure parameter for the lifetime model and can be calculated sintered Ag joints on Cu, DBC and Ag plated Cu substrates. A
by FEM simulation. In addition, the equivalent strain is volume nano-sized Ag paste is used and a Si semiconductor. Thermal
averaged along a path up to the crack length. The crack length ageing is done at 300 °C in atmosphere for 24, 50, 100, 200,
can be determined for 20 % delamination of the die. Volume 500 and 1000 h and the samples are then taken out for shear
averaging the strain avoids an influence of singularities [28]. tests and investigation of the microstructure. Die shear strength
was best for Ag plated Cu substrate and increased during
ageing due to consolidation and reduction of pores, which
results in a reduced thickness of the sintered Ag layer. For the
DBC, die shear strength is higher than for the Cu substrate due
to lower CTE mismatch between DBC and Si die. Another
influence factor could be the different interfacial
microstructures. At the interface between sintered Ag and Cu
layers, Cu oxides (Cu2O) grow on all substrates with
increasing ageing time. Hence, the interface is weakened due to
the different CTE by Cu2O, which has the smallest influence
on Ag plated Cu, than on DBC and finally on Cu substrate. For
all substrates, there is a decrease in porosity with increasing
ageing time due to the diffusion-controlled boundary migration
driven by the grain boundary energy. Most of the pores were
still in a spherical shape after ageing like in the initial state [4].
Fig. 18. Comparison of strain induction by a) thermal cycling (pTC) due to
CTE mismatch and b) isothermal 3 point bending [28] Chen et al. [68] examine thermal ageing behavior for different
top metallizations on a Cu substrate. Metallization layers are

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
made of gold (Au) and Ag and different deposition processes nano-sized particles or hybrid) and the additives have a big
(electroplating and sputtering) are used. The best high influence. Hence, it is difficult to compare results from various
temperature reliability and bonding quality is achieved with the authors with each other, which is shown by Heilmann et al.
sputtered Ag surface, followed by the electroplated Ag surface. [18]. In the field of material characterization, there are several
In addition, fracture behavior of sintered Ag during shear test is tests with different advantages and disadvantages. The classical
different. It comes to diffusion in the Ag-Au layer leading to a shear tests, where the die is sheared from the substrate, is used
decrease in shear strength. Delamination is defined by a ratio to assess the mechanical performance for different sintering
between the length of the connected metallization layer and process parameters. Material data needed as input for FEM
sintered Ag of a cross-section and given interface length. For simulation but creep effects cannot be gained. Therefore, it is
electroplated Ag there is a significant decrease of the more a quality test rather than a stress test. Lap shear tests can
connection ratio with aging time, whereas there is barely a be used for examination of the creep behavior. A disadvantage
change for sputtered Ag metallization. For all metallizations a of this test is the additional bending component which leads to
coarsening of the sintered Ag is examined, which means that an underestimation of the actual shear stress. Furthermore, with
the Ag grains become larger as well as the pore size, while the an increasing length of overlap, shear strength decreases due to
number of pores decreases [68]. Chew et al. [65] get similar a stress concentration at both ends (depending on geometry).
results for comparison of Ag and Au metallization. Table IV For tensile tests, usually flat tensile specimen are used with the
shows a summary of all presented reliability testing methods advantage, that the same microstructure and porosity like in a
with their advantage and disadvantage as well as some more joint can be achieved. In addition, stacking methods are used to
information. get thicker specimen with comparable microstructure.

TABLE IV. R
SUMMARY OF TEST METHODS FOR RELIABILITY TESTING

Round Tensile
Isothermal Bending
Test pTC aTC Lap Shear Test Specimen (not
Test
available yet)
1. Simultaneously
testing of many 1. Easy to produce 1. Round samples
samples samples with similar enables cycling loading
Reflects the failure
2. Different dwell time quality Substitute and of tensile specimen
Advantage mode of application in
than in application 2. Porosity and accelerate pTC testing 2. Lifetime model for
field
3. Sufficient to microstructure similar sintered Ag could be
guarantee a defined to application gained
minimum of TC
Fabrication process is
Not sufficient for expected to be different
Degradation occurs
development of a Power cycler necessary Additional bending and leads e.g. to other
only from two sides
Disadvantage lifetime model which and cannot test many component has to be thicknesses and
instead of all four like
considers the physics samples simultaneously considered therefore eventually to
in pTC
of failure other porosities and
microstructure
1. 20 % delamination
1. 5 % [60] to 20 %
[2, 14, 59]
[14] increase of
2. 20 % increase of
forward voltage
thermal resistance or Crack length of
End of Life Ccriterion 2. 20 % increase of Not given -
temperature swing [59, 20 % delamination [28]
temperature swing
64]
[14,60] or transient
3. 50 % drop of joint
thermal resistance [66]
strength [11
Fabrication Process and
3a 2 5 4 -
Repeatability
Porosity/microstruc-
ture comparable to Yes Yes Yes Yes Must be proven
application?
a. A value of five is best value and zero the worst.

V. CONCLUSION
These tests allow getting the complete material behavior of
In this review, the state of the art of sintered Ag in power the sintered Ag, including creep behavior, for different
electronics in the field of material characterization and temperatures, strain rates and process parameters. A
reliability testing is investigated. The literature research shows disadvantage of flat specimen is that cyclic loading, which is
that the sintering process parameters (sintering pressure, necessary for reliability testing, cannot be applied due to the
temperature and time) have a significant influence on porosity, flat geometry. This would lead to kinking of the samples.
microstructure and therefore on the mechanical properties and Nanoindentation is another test for determination of the
reliability. In addition, the composition of the paste (micro-, material behavior. Thereby, the indent depth as well as the

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
shape of the indenter have a strong influence on the tests. REFERENCES
Mechanical properties cannot be determined directly from the [1] C. Chen, S. Nagao, K. Suganuma, J. Jiu, T. Sugahara, H. Zhang, T.
tests and FEM simulations are necessary to gain the material Iwashige, K. Sugiura and K. Tsuruta, „Macroscale and microscale
model. An advantage of nanoindentation compared to tensile fracture toughness of microporous sintered Ag for applications in
testing is, that the fabrication process is easier repeatable with power electronic devices,“ Acta Materialia, Bd. 129, pp. 41-51, 2017.
the same quality. An important question outlined by Leslie et [2] T. Herboth, M. Guenther, A. Fix and J. Wilde, „Failure mechanisms of
sintered silver interconnections for power electronic applications,“ in
al. [42] is the suitability of nanoindentation test and if the 2013 IEEE 63rd Electronic Components and Technology Conference,
localized deformation field is comparable to creep behavior of 2013.
uniaxial tensile testing. This is a fundamental issue of [3] X. Li, G. Chen, X. Chen, G.-Q. Lu, L. Wang and Y.-H. Mei, „High
nanoindentation and not only for sintered Ag. Finally, a temperature ratcheting behavior of nano-silver paste sintered lap shear
disadvantage of nanoindentation is that creep behavior cannot joint under cyclic shear force,“ Microelectronics Reliability, Bd. 53,
pp. 174-181, 1 2013.
be investigated over a wide stress range and can be only [4] S. T. Chua and K. S. Siow, „Microstructural studies and bonding
applied for high stresses. Nevertheless, for qualitative strength of pressureless sintered nano-silver joints on silver, direct
comparison of different sinter layers, nanoindentation seems to bond copper (DBC) and copper substrates aged at 300 °C,“ Journal of
be a good option. Alloys and Compounds, Bd. 687, pp. 486-498, 12 2016.
[5] W. Sabbah, R. Riva, S. Hascoet, C. Buttay, S. Azzopardi, E. Woirgard,
A micro bending test by Zschenderlein et al. [44] is also D. Planson, B. Allard and R. Meuret, „Evaluation of silver-sintering
presented. A double dog bone sample is bended by die-attach,“ in 2012 7th International Conference on Integrated Power
nanoindenation but with sintered Cu nanoparticles instead of Electronics Systems (CIPS), 2012.
[6] K. S. Siow, „Mechanical properties of nano-silver joints as die-attach
Ag. The result should be transferable to sintered Ag as well. materials,“ Journal of Alloys and Compounds, Bd. 514, pp. 6-19, 2
Finally, micro-sized specimen are investigated which helps to 2012.
get a deeper understanding of the influence of specimen size on [7] J. Dai, J. Li, P. Agyakwa and C. M. Johnson, „Power cycling
material parameters as well as of the microstructure like grain reliability of time-reduced sintering for attaching SiC diodes using
size, defect structure and porosity. This work is done by nanosilver film,“ 2018.
[8] C. Chen, S. Nagao, H. Zhang, T. Sugahara, K. Suganuma, T. Iwashige,
Zabihzadeh et al. [45, 46, 47]. Another investigation by Chen et
K. Sugiura and K. Tsuruta, „Low-Stress Design for SiC Power
al. [1] shows the effect of specimen size on fracture toughness Modules with Sintered Porous Ag Interconnection,“ in 2016 IEEE 66th
for micro- and macro-scaled specimen. A summary of all test Electronic Components and Technology Conference (ECTC), 2016.
methods for material characterization can be seen in Table III. [9] P. Quintero, P. McCluskey and B. Koene, „Thermomechanical
reliability of a silver nano-colloid die-attach for high temperature
In many studies the importance of understanding the applications,“ Microelectronics Reliability, Bd. 54, pp. 220-225, 1
physics of failure is highlighted [10, 61]. It is not sufficient to 2014.
examine only the macro material behavior (includes also the [10] C. Weber, H. Walter, M. V. Dijk, M. Hutter, O. Wittler and K. D.
fracture surface) which can indicate a brittle material behavior Lang, „Combination of Experimental and Simulation Methods for
Analysis of Sintered Ag Joints for High Temperature Applications,“ in
and shows no necking at tensile testing [2]. Furthermore, 2016 IEEE 66th Electronic Components and Technology Conference
investigations on microscale shows elastic-plastic material (ECTC), 2016.
behavior and a localized necking at the silver ligaments of the [11] S. Sakamoto, T. Sugahara and K. Suganuma, „Microstructural stability
microstructure. This leads to an intergranular fracture under of Ag sinter joining in thermal cycling,“ J Mater Sci: Mater Electron,
tensile loading [2]. Changes of the process parameters lead to Bd. 24, pp. 1332-1340, 4 2013.
[12] H. S. Chin, K. Y. Cheong and A. B. Ismail, „A Review on Die-attach
different microstructures and porosities. Especially for Materials for SiC-Based High-Temperature Power Devices,“ Metall
generating a lifetime model, understanding the failure and Materi Trans B, Bd. 41, pp. 824-832, 8 2010.
mechanism is of great importance. In the field of reliability [13] J. G. Bai, J. N. Calata and G. Q. Lu, „Processing and Characterization
testing, pTC and aTC are widely used with the described of Nanosilver Pastes for Die-Attaching SiC Devices,“ IEEE
advantages and disadvantages. Heilmann et al. [28] present a Transactions on Electronics Packaging Manufacturing, Bd. 30, pp.
241-245, 10 2007.
method to substitute pTC tests with isothermal bending test [14] M. Knoerr, S. Kraft and A. Schletz, „Reliability assessment of sintered
with the great advantage of accelerating the test method with nano-silver die-attachment for power semiconductors,“ in 2010 12th
the same failure mode. Electronics Packaging Technology Conference, 2010.
[15] F. Le Henaff, S. Azzopardi, J. Y. Deletage, E. Woirgard, S. Bontemps
A round specimen for tensile and cyclic testing would be and J. Joguet, „A preliminary study on the thermal and mechanical
optimal. Especially for characterization of fatigue life, this test performances of sintered nano-scale silver die-attach technology
could enable a lifetime model for the material of sintered Ag, depending on the substrate metallization,“ Microelectronics
independent of geometry and structure. Compared to flat tensile Reliability, Bd. 52, pp. 2321-2325, 9 2012.
specimen, there should be no problem with kinking during [16] Y. Mei, G. Chen, Y. Cao, X. Li, D. Han and X. Chen, „Simplification
of Low-Temperature Sintering Nanosilver for Power Electronics
cyclic loading. As mentioned before, achieving the same Packaging,“ Journal of Elec Materi, Bd. 42, pp. 1209-1218, 6 2013.
microstructure and porosity like for a sintered joint would be [17] K. Suganuma, S. Sakamoto, N. Kagami, D. Wakuda, K. S. Kim and
necessary and has to be examined in detail. Furthermore, the M. Nogi, „Low-temperature low-pressure die-attach with hybrid silver
influence of sample size should be investigated to determine if particle paste,“ Microelectronics Reliability, Bd. 52, pp. 375-380, 2
there is an influence on the material parameters. Table IV 2012.
[18] M. Kuramoto, S. Ogawa, M. Niwa, K. S. Kim and K. Suganuma, „Die
summarize the reliability testing methods presented in this Bonding for a Nitride Light-Emitting Diode by Low-Temperature
paper. Sintering of Micrometer Size Silver Particles,“ IEEE Transactions on
Components and Packaging Technologies, Bd. 33, pp. 801-808, 12
2010.
[19] J. Kähler, N. Heuck, G. Palm, A. Stranz, A. Waag and E. Peiner,

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
 „Low-pressure sintering of silver micro- and nanoparticles for a high [38] T. G. Lei, J. N. Calata, G. Q. Lu, X. Chen and S. Luo, „Low-
temperature stable pick amp; place die-attach,“ in 18th European Temperature Sintering of Nanoscale Silver Paste for Attaching Large-
Microelectronics Packaging Conference, 2011. Area (>100 mm²) Chips,“ IEEE Transactions on Components and
[20] K. S. Siow, „Are Sintered Silver Joints Ready for Use as Interconnect Packaging Technologies, Bd. 33, pp. 98-104, 3 2010.
Material in Microelectronic Packaging?,“ Journal of Elec Materi, Bd. [39] X. Li, G. Chen, L. Wang, Y.-H. Mei, X. Chen and G.-Q. Lu, „Creep
43, pp. 947-961, 4 2014. properties of low-temperature sintered nano-silver lap shear joints,“
[21] K. S. Siow and Y. T. Lin, „Identifying the Development State of Materials Science and Engineering: A, Bd. 579, pp. 108-113, 9 2013.
Sintered Silver (Ag) as a Bonding Material in the Microelectronic [40] T. Suzuki, T. Terasaki, Y. Kawana, D. Ishikawa, M. Nishimura, H.
Packaging Via a Patent Landscape Study,“ J. Electron. Packag, Bd. Nakako and K. Kurafuchi, „Effect of Manufacturing Process on
138, pp. 020804--020804--13, 4 2016. Micro-Deformation Behavior of Sintered-Silver Die-Attach Material,“
[22] T. Wang, G. Chen, Y. Wang, X. Chen and G.-q. Lu, „Uniaxial IEEE Transactions on Device and Materials Reliability, Bd. 16, pp.
ratcheting and fatigue behaviors of low-temperature sintered nano- 588-596, 12 2016.
scale silver paste at room and high temperatures,“ Materials Science [41] S. Letz, A. Hutzler, U. Waltrich, S. Zischler and A. Schletz,
and Engineering: A, Bd. 527, pp. 6714-6722, 9 2010. „Mechanical properties of silver sintered bond lines: Aspects for a
[23] J. R. Groza, „Nanocrystalline Powder Consolidation Methods,“ in reliable material data base for numerical simulations,“ in CIPS 2016;
Nanostructured Materials: Processing, Properties and Potential 9th International Conference on Integrated Power Electronics
Applications, Norwich, : Noyes Publication, 2002, pp. 173-234. Systems, 2016.
[24] P. Gadaud, V. Caccuri, D. Bertheau, J. Carr and X. Milhet, „Ageing [42] D. Leslie, A. Dasgupta and C. Morillo, „Viscoplastic properties of
sintered silver: Relationship between tensile behavior, mechanical pressure-less sintered silver materials using indentation,“
properties and the nanoporous structure evolution,“ Materials Science Microelectronics Reliability, Bd. 74, pp. 121-130, 7 2017.
and Engineering: A, Bd. 669, pp. 379-386, 7 2016. [43] X. Long, W. Tang, Y. Feng, C. Chang, L. M. Keer and Y. Yao, „Strain
[25] X. Milhet, P. Gadaud, V. Caccuri, D. Bertheau, D. Mellier and M. rate sensitivity of sintered silver nanoparticles using rate-jump
Gerland, „Influence of the Porous Microstructure on the Elastic indentation,“ International Journal of Mechanical Sciences, Bd. 140,
Properties of Sintered Ag Paste as Replacement Material for Die- pp. 60-67, 5 2018.
attachment,“ Journal of Elec Materi, Bd. 44, pp. 3948-3956, 10 2015. [44] U. Zschenderlein, K. Suresh, M. Baum and B. Wunderle, „Mechanical
[26] J. Carr, X. Milhet, P. Gadaud, S. A. E. Boyer, G. E. Thompson and P. characterization of sintered nanoparticles for advanced electrical and
Lee, „Quantitative characterization of porosity and determination of thermal joints: σ-ε-transformation in micro bending tests,“ 2015.
elastic modulus for sintered micro-silver joints,“ Journal of Materials [45] S. Zabihzadeh, P. Van, L. I. Duarte, R. Mokso, A. Cervellino and S.
Processing Technology, Bd. 225, pp. 19-23, 11 2015. Van, „Deformation behavior of sintered nanocrystalline silver layers,“
[27] T. Youssef, W. Rmili, E. Woirgard, S. Azzopardi, N. Vivet, D. Acta Materialia, Bd. 97, pp. 116-123, 2015.
Martineau, R. Meuret, G. Le Quilliec and C. Richard, „Power modules [46] S. Zabihzadeh, S. Van Petegem, M. Holler, A. Diaz, L. I. Duarte and
die-attach: A comprehensive evolution of the nanosilver sintering H. Van Swygenhoven, „Deformation behavior of nanoporous
physical properties versus its porosity,“ Microelectronics Reliability, polycrystalline silver. Part I: Microstructure and mechanical
Bd. 55, pp. 1997-2002, 8 2015. properties,“ Acta Materialia, Bd. 131, pp. 467-474, 6 2017.
[28] J. Heilmann, I. Nikitin, U. Zschenderlein, D. May, K. Pressel and B. [47] S. Zabihzadeh, J. Cugnoni, L. I. Duarte, P. Van and S. Van,
Wunderle, „Reliability experiments of sintered silver based „Deformation behavior of nano-porous polycrystalline silver. Part II:
interconnections by accelerated isothermal bending tests,“ Simulations,“ Acta Materialia, Bd. 131, pp. 564-573, 2017.
Microelectronics Reliability, Bd. 74, pp. 136-146, 7 2017. [48] C. Chen, S. Nagao, T. Sugahara, H. Zhang, J. Jiu, K. Suganuma, T.
[29] B. Wunderle and B. Michel, „Lifetime modelling for microsystems Iwashige, K. Sugiura and K. Tsuruta, „Effect of size and shape of Ag
integration: from nano to systems,“ Microsyst Technol, Bd. 15, pp. particles for mechanical properties of sintered Ag joints evaluated by
799-812, 6 2009. micro-compression test,“ in 2017 International Conference on
[30] C. Chen, S. Nagao, H. Zhang, J. Jiu, T. Sugahara, K. Suganuma, T. Electronics Packaging (ICEP), 2017.
Iwashige, K. Sugiura and K. Tsuruta, „Mechanical Deformation of [49] D. W. Oplinger, „Effects of adherend deflections in single lap joints,“
Sintered Porous Ag Die-attach at High Temperature and Its Size Effect International Journal of Solids and Structures, Bd. 31, pp. 2565-2587,
for Wide-Bandgap Power Device Design,“ Journal of Elec Materi, Bd. 9 1994.
46, pp. 1576-1586, 3 2017. [50] Y. L. Shen, N. Chawla, E. S. Ege and X. Deng, „Deformation analysis
[31] J. G. Bai, Z. Z. Zhang, J. N. Calata and G. q. Lu, „Characterization of of lap-shear testing of solder joints,“ Acta Materialia, Bd. 53, pp.
Low-Temperature Sintered Nanoscale Silver Paste for Attaching 2633-2642, 5 2005.
Semiconductor Devices,“ in 2005 Conference on High Density [51] N. Chawla, Y.-L. Shen, X. Deng and E. S. Ege, „An evaluation of the
Microsystem Design and Packaging and Component Failure Analysis, lap-shear test for Sn-rich solder/Cu couples: Experiments and
2005. simulation,“ Journal of Elec Materi, Bd. 33, pp. 1589-1595, 12 2004.
[32] V. Caccuri, X. Milhet, P. Gadaud, D. Bertheau and M. Gerland, [52] T. Herboth, C. Früh, M. Günther and J. Wilde, „Assessment of thermo-
„Mechanical Properties of Sintered Ag as a New Material for Die mechanical stresses in low temperature joining technology,“ 2012.
Bonding: Influence of the Density,“ Journal of Elec Materi, Bd. 43, [53] D. Tabor, The hardness of Metals, New, York: Oxford University
pp. 4510-4514, 12 2014. Press, 2000.
[33] R. Mroßko, H. Oppermann and B. Wunderle, „Reliability analysis of [54] R. Saha and W. D. Nix, „Effects of the substrate on the determination
low temperature low pressure Ag-sinter die-attach,“ in Technical of thin film mechanical properties by nanoindentation,“ Acta
Proceedings of the 2011 NSTI Nanotechnology Conference and Expo, Materialia, Bd. 50, pp. 23-38, 1 2002.
NSTI-Nanotech 2011. Vol.2, 2011. [55] B. N. Lucas and W. C. Oliver, „Indentation power-law creep of high-
[34] I. Nikitin and K. Pressel, „Mechanical properties of porous silver purity indium,“ Metall and Mat Trans A, Bd. 30, pp. 601-610, 3 1999.
materials depending on sintering parameters,“ in 20th International [56] V. Maier, K. Durst, J. Mueller, B. Backes, H. W. Höppel and M.
Workshop on Thermal Investigations of ICs and Systems, 2014. Göken, „Nanoindentation strain-rate jump tests for determining the
[35] J. Li, C. M. Johnson, C. Buttay, W. Sabbah and S. Azzopardi, local strain-rate sensitivity in nanocrystalline Ni and ultrafine-grained
„Bonding strength of multiple SiC die-attachment prepared by Al,“ Journal of Materials Research, Bd. 26, pp. 1421-1430, 6 2011.
sintering of Ag nanoparticles,“ Journal of Materials Processing [57] H. Van Swygenhoven, B. Schmitt, P. M. Derlet, S. Van Petegem, A.
Technology, Bd. 215, pp. 299-308, 1 2015. Cervellino, Z. Budrovic, S. Brandstetter, A. Bollhalder and M. Schild,
[36] S. A. Paknejad and S. H. Mannan, „Review of silver nanoparticle „Following peak profiles during elastic and plastic deformation: A
based die-attach materials for high power/temperature applications,“ synchrotron-based technique,“ Review of Scientific Instruments, Bd.
Microelectronics Reliability, Bd. 70, pp. 1-11, 3 2017. 77, p. 013902, 1 2006.
[37] R. Khazaka, L. Mendizabal and D. Henry, „Review on Joint Shear [58] H. Van Swygenhoven and S. Van Petegem, „In-situ mechanical testing
Strength of Nano-Silver Paste and Its Long-Term High Temperature during X-ray diffraction,“ Materials Characterization, Bd. 78, pp. 47-
Reliability,“ Journal of Elec Materi, Bd. 43, pp. 2459-2466, 7 2014. 59, 4 2013.

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.
[59] A. Hutzler, A. Tokarski and A. Schletz, „High temperature die-attach [67] K. Zhang and J. Aktaa, „Characterization and modeling of the
materials for aerospace power electronics: Lifetime tests and ratcheting behavior of the ferritic–martensitic steel P91,“ Journal of
modeling,“ in 2015 IEEE Aerospace Conference, 2015. Nuclear Materials, Bd. 472, pp. 227-239, 4 2016.
[60] R. Schmidt, R. Werner, J. Casady, B. Hull and A. Barkley, „Power [68] C. Chen, K. Suganuma, T. Iwashige, K. Sugiura and K. Tsuruta,
Cycle Testing of Sintered SiC-MOSFETs,“ in PCIM Europe 2017; „High-temperature reliability of sintered microporous Ag on
International Exhibition and Conference for Power Electronics, electroplated Ag, Au, and sputtered Ag metallization substrates,“ J
Intelligent Motion, Renewable Energy and Energy Management, 2017. Mater Sci: Mater Electron, Bd. 29, pp. 1785-1797, 2 2018.
[61] C. Weber, M. Hutter, S. Schmitz and K. D. Lang, „Dependency of the [69] C. Choe, S. Noh, C. Chen, T. Ishina, S. Nagao and K. Suganuma,
porosity and the layer thickness on the reliability of Ag sintered joints „Thermal effect on material properties of sintered porous silver during
during active power cycling,“ in 2015 IEEE 65th Electronic high temperature ageing,“ in 2017 18th International Conference on
Components and Technology Conference (ECTC), 2015. Electronic Packaging Technology (ICEPT), 2017.
[62] G. Chen, Z.-S. Zhang, Y.-H. Mei, X. Li, D.-J. Yu, L. Wang and X. [70] S. Egelkraut, L. Frey, M. Knoerr and A. Schletz, „Evolution of shear
Chen, „Applying viscoplastic constitutive models to predict ratcheting strength and microstructure of die bonding technologies for high
behavior of sintered nanosilver lap-shear joint,“ Mechanics of temperature applications during thermal aging,“ in 2010 12th
Materials, Bd. 72, pp. 61-71, 5 2014. Electronics Packaging Technology Conference, 2010.
[63] Y. Tan, X. Li and X. Chen, „Fatigue and dwell-fatigue behavior of [71] R. Kisiel, Z. Szczepański, P. Firek, J. Grochowski, M. Myśliwiec and
nano-silver sintered lap-shear joint at elevated temperature,“ M. Guziewicz, „Silver micropowders as SiC die-attach material for
Microelectronics Reliability, Bd. 54, pp. 648-653, 3 2014. high temperature applications,“ in 2012 35th International Spring
[64] F. Le Henaff, S. Azzopardi, E. Woirgard, T. Youssef, S. Bontemps and Seminar on Electronics Technology, 2012.
J. Joguet, „Lifetime Evaluation of Nanoscale Silver Sintered Power
Modules for Automotive Application Based on Experiments and
Finite-Element Modeling,“ IEEE Transactions on Device and
Materials Reliability, Bd. 15, pp. 326-334, 9 2015.
[65] L. M. Chew, W. Schmitt, J. Nachreiner and S. Gunst, „Silver sinter
paste optimized for pressure sintering under air atmosphere on
precious and non-precious metal surfaces with high reliable sintered
joints,“ 2018.
[66] C. Herold, M. Schaefer, F. Sauerland, T. Poller, J. Lutz and O.
Schilling, „Power cycling capability of Modules with SiC-Diodes“, in
2014 8th International Conference on Integrated Power Electronics
Systems (CIPS)

Authorized licensed use limited to: UNIVERSITAET FREIBURG. Downloaded on January 04,2021 at 16:13:58 UTC from IEEE Xplore. Restrictions apply.