Delft University of Technology
3D interconnect technology based on low temperature copper nanoparticle sintering
Zhang, Boyao; Carisey, Yorick; Damian, A.; Poelma, Rene; Zhang, Guo Qi; van Zeijl, Henk
DOI
10.1109/ICEPT.2016.7583331
Publication date
2016
Document Version
Accepted author manuscript
Published in
Proceedings of the 17th International Conference on Electronic Packaging Technology (ICEPT)
Citation (APA)
Zhang, B., Carisey, Y., Damian, A., Poelma, R., Zhang, G. Q., & van Zeijl, H. (2016). 3D interconnect
technology based on low temperature copper nanoparticle sintering. In K. Bi, S. Liu, & S. Zhou (Eds.),
Proceedings of the 17th International Conference on Electronic Packaging Technology (ICEPT)
(pp. 1163-
1167). IEEE . https://doi.org/10.1109/ICEPT.2016.7583331
Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.
Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent
of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Takedown policy
Please contact us and provide details if you believe this document breaches copyrights.
We will remove access to the work immediately and investigate your claim.
This work is downloaded from Delft University of Technology.
For technical reasons the number of authors shown on this cover page is limited to a maximum of 10.
3D Interconnect technology based on low
temperature copper nanoparticle sintering
B. Zhang, Y.C.P. Carisey, A. Damian, R.H. Poelma
and G.Q. Zhang
Department of Microelectronics, Delft University of
Technology, Delft, 2628CT, the Netherlands
H.W. van Zeijl
Else Kooi Lab, Delft University of Technology, Delft,
2628CT, the Netherlands
Abstract - We explore a methodology for patterned copper
nanoparticle paste for 3D interconnect applications in wafer to
wafer (W2W) bonding. A novel fine pitch thermal compression
bonding process (sintering) with coated copper nanoparticle
paste was developed. Most of the particle size is between 10-30
nm. Lithographically defined stencil printing using photoresist
and lift-off was used to apply and pattern the paste. Variations in
sintering process parameters, such as: pressure, geometry and
ambient atmosphere, were studied. Compared to Sn-Ag-Cu (SAC)
microsolder bumps, we achieved better interconnect resistivity
after sintering at 260 for 10 min, in a 700 mBar hydrogen
forming gas (H
2
/N
2
) environment. The electrical resistivity was
7.84 ± 1.45 µΩ∙cm, which is about 4.6 times that of bulk copper.
In addition, metallic nanoparticle interconnect porosity can
influence the electrical properties of the interconnect.
Consequently, we investigated the porosity effect on conductivity
using finite element simulation. A linear relationship between the
equivalent conductivity and particle overlapping ratio was found.
Keywords: Interconnect; 3D packaging; copper nanoparticle
paste; low-temperature sintering
I.INTRODUCTION
The need for more functionalization, miniaturization and
lower energy consumption drives the ongoing semiconductor
industry. Interconnect material that can be fabricated in
accurate nanoscale, with advanced properties, is essential [1-3].
Heterogeneous integration including chip to wafer (C2W),
wafer to wafer (W2W) and package on package stacking
(PoP), requires interconnect structures with ultra-fine accuracy,
profound electrical property and better reliability. Current
bonding technology, including wire bonding and flip-chip
technology, are facing challenges as requirements in feature
size, reliability and electrical property have raised [4-10].
To overcome the drawbacks mentioned in previous
technologies, metallic nanoparticles have become a promising
candidate for 3D heterogeneous integration. Among various of
metallic nanoparticles, Cu nanoparticle paste is emerging as a
promising candidate in 3D interconnect applications [11]. First,
bulk copper can be formed after sintering, which will have
similar reflow temperature as melting temperature of bulk
copper (1085 ). In typical flip-chip technology, which
usually use Sn-Ag-Cu (SAC) alloy microsolder bumps, with a
reflow temperature of 220-250 ℃ . However, copper
nanoparticle paste, with particle size 40 nm in average, does
not reflow after the first sintering process since the
nanoparticle fuse together [10]. In addition, creep failure
inside solder joints due to secondary intermetallic compounds
formation, can be avoided by pure Cu nanoparticle paste.
However, one drawback of Cu nanoparticles, is the oxidization
in air. In this paper, we use Cu nanoparticle with an organic
coating to protect Cu from oxidation. Furthermore, these core-
shell structured Cu nanoparticles are dispersed in viscous
solution, which enables them to be patterned as interconnect
structures [12].
As feature sizes within chips are decreasing, it is difficult
and expensive to achieve fine pitch (<50 µm) interconnect
structure, due to processing limitations [13]. In this paper, a
novel methodology based on low temperature thermal
compression (sintering) using copper nanoparticle paste is
developed for W2W 3D interconnect. Lithographic defined
stencil printing method was developed to obtain pattern Cu
nanoparticle paste below 10 µm. A novel fine pitch thermal
compression bonding process (sintering) with coated copper
nanoparticle paste was developed. Variations in sintering
process parameters, such as: pressure, geometry and ambient
atmosphere, were studied. As a result of that, lower electrical
resistivity than typical SAC, was achieved after sintering at
260 ℃ for 10 min, in a 700 mBar forming gas environment.
Considering of the fact that sintered nanoparticles have porous
structure, which will influence the electrical property of
interconnect, finite element simulation is used in to investigate
the porosity effect on resistivity/conductivity.
II. COPPER NANOPARTICLE PASTE INTERCONNECT PROCESSING
AND CHARACTERIZATION
When the copper particle size reduces towards nanoscale,
surface energy will increase. This will suppress the local
melting point of the particle the bulk material melting
temperature [14]. This scale effect makes copper nanoparticles
a strong candidate for 3D integration application. The copper
nanoparticle paste used in this study is commercially available
[12]. These copper nanoparticles are coated with an organic
protective layer to prevent oxidation.
A. Mask design and process for photolithography patterning
and wafer bonding
We employed photolithography defined stencil printing
method for wafer-level processing. Photoresist is applied and
patterned using front-end wafer-level processing. Then copper
paste is applied and finally photoresist lift-off process is used
to obtain patterned structures for electrical and mechanical
tests. The process flow is illustrated in Figure. 1.
For the first step, a 300 nm low-stress silicon nitride (SiN
x
)
was deposited on the silicon substrate by using low-pressure-
chemical-vapour-deposition (LPCVD). Then a 1.4 µm thick
photoresist (EVG 120) was spin coated on the top. Then the
photoresist was exposed and developed with a designed mask.
A patterned surface structure was obtained as shown in Figure
1(a). Next, a small amount of Cu nanoparticle paste was
applied, as delivered from Lockheed Martin. As shown in
Figure 1(b), silicone squeegee, from KOENEN Technologies,
was selected to manually distribute Cu nanoparticle paste over
the patterned wafers surfaces. The shore hardness was 65°.
As shown in Figure 1(c), test wafers with Cu nanoparticle
paste were followed by drying at 50 for 5 min using a hot
plate. After that, a water ultrasonic bath with a beaker filled
with N-Methyl-2-pyrrolidone (NMP) was used for photoresist
layer removal at room temperature. The last step was
completed by DI waster rinse and drying, as shown in Figure
1(d).
In Figure 2, there are different microstructures that were
designed to test the pattern accuracy of this Cu nanoparticle
paste. It can be seen that Cu nanoparticle paste has that ability
to reach fine pitch accuracy, even below 5 µm. As an initial
research work, manual distribution is acceptable. However, for
high volume industry production, other distribution methods
that can be applied in mass production is essential to ensure
accuracy and consistency.
Figure 1. Schematic patterned test wafer process flow: (a) photoresist
pattern on top of SiNx layer; (b) Cu nanoparticle paste distribution by
squeegee; (c) Cu nanoparticle paste was dried using hot plate; (d) photoresist
removal by ultrasonic bath in NMP.
Figure 2. Patterened microstructrure on wafer level. (a), (b), (c) are
different microstructures designed to test the Cu nanoparticle paste pattern
accuracy.
After Cu nanoparticle paste was spread and dried on
bottom wafer, top and bottom wafers were treated differently
as shown in Figure 2. For the top wafer, SiN
x
and silicon oxide
layers were partially etched to give access to probe pads for
electrical property measurement, as shown in Figure 3(c).
Then, both wafers were deposited and patterned with 300 nm
thin layer of copper with the same mask, by using
conventional litho/etch processing, as shown in Figure 3(a).
Then only bottom wafer was distributed and patterned with Cu
nanoparticle paste, using the method described in previous
section, as shown in Figure 3(b). After pasted was dried, both
top and bottom wafer were transferred to bonding machines
for sintering, as shown in Figure 3(c). In order to study the
ambient atmosphere effect on sintering result, three different
atmosphere and four sintering methods were used, including
N
2
, hydrogen forming gas and vacuum sintering w/o N
2
flow.
Based on Andreis results, 260 was chosen to be the ultimate
sintering temperature [15]. The process were as follow:
1. 200 mBar N
2
enviornment was created by Heraus
Vaccum oven for wafer bonding. The heating rate was
about 10 /min and wafers were annealed at 260 for
10 min. The cooling rate was 1 /min.
2. In Aixtron Black Magic PECVD system, a controlled
forming gas environment with 700 mBar was given.
The hydrogen forming gas is a mix gas with a 10:1
concentration of H
2
and N
2
. The rest of bonding
parameters were the same as in N
2
environment.
3. AML wafer bonder could provide a high vacuum
environment (~ 10
-4
mBar) with controlled heating.
Wafers were sintered at 200 (due to temperature
limit of machine) at 30 /min heating rate and
annealed for 10 min. In one circumstance, a small N
2
flow was provide to avoid any possible oxygen inside
the oven. The cooling rate was 1 - 2 /min, depending
on w/o N
2
flush.
To give access for electrical measuring probe, top wafer
was removed by 30% KOH solution at 80 , as shown in
Figure 3(d). The KOH etch stopped at SiN
x
layer of bottom
wafer. Some mechanical force was applied to break the thin
SiN
x
layer on top wafer.
Figure 3. Process flow of wafer bonding procedure.
B. Electrical resistivity measurement
Electrical resistivity is one of the most important
properties for 3D interconnect materials. Greek cross Van der
Pauw (VDP) structures were designed within test wafers to
measure the resistivity of Cu nanoparticle paste [16, 17]. The
schematic structure of Greek cross VDP is shown in Figure 4.
Forced current is applied through area 3 and 4. Then
consequent voltage can be measured from area 1 and 2. A
corresponding microstructure of Greek VDP was obtained in
test wafers after Cu nanoparticle paste stencil printing, as
shown in Figure 5.
Figure 4. Schematic Greek cross VDP structure for sintered Cu
nanoparticle paste resistivity measurement. The Forced current will be applied
through area 3 and 4. Then consequent voltage could be measured through
area 1 and 2.
Figure 5. A various of Greek cross VDP structures on test wafers after Cu
nanoparticle paste stencil printing.
A four point probe needles setup was used to perform I –V
curve measurement. The forced current range was ± 10 mA.
Sintered Cu nanoparticle paste thickness t was measured
before bonding using a Dektak profile meter. The thicknesses
of Cu nanoparticle paste can be controlled using the thickness
of the photoresist. We processed three wafers with different
photoresist thickness, and we obtained 1.5 µm, 2 µm and 3 µm.
From Equation (1), sheet resistance can be derived by using
Greek cross VDP methodology [16].
(1)
The bulk resistivity of the sintered Cu nanoparticle paste is
obtained by multiplying the sheet resistance by the film
thickness,
. (2)
Resistivity measurements in different ambient
atmospheres were performed to study the influence of
oxidation, see Figure 6. Samples sintered under vacuum
atmosphere. For vacuum sintering, samples with a small flow
of N
2
, the resistivity of it was measured as 52.68 ± 24.92
µΩ∙cm. For the samples sintered in vacuum oven, without a
flow of N
2
, the resistivity is 75.90 ± 23.63 µΩ∙cm without N
2
flush. Hydrogen forming gas ambient sintering atmosphere
resulted in the lowest resistivity of sintered Cu nanoparticle
paste , which was 7.84 ± 1.45 µΩ∙cm. In N
2
environment, the
resistivity was 63.44 ± 22.12 µΩ∙cm. Sintered Cu nanoparticle
paste in hydrogen forming gas atmosphere showed the lowest
resistivity, which is better than values reported previously in
literature. Hydrogen forming gas, provides a reducing
environment for Cu nanoparticle oxides during sintering,
hence improving conductivity. For the higher resistivity
obtained in vacuum atmosphere, both with and without N
2
flow, it is possible that vacuum pumping enable some
surfactants in paste to evaporate. As a result of that, Cu
nanoparticles were partially oxidized in a medium vacuum
environment.
Figure 6. Sintered Cu nanoparticle paste resistivity with different ambient
atmosphere
III. FINITE ELEMENT MODELLING OF POROSITY EFFECTS ON
ELECTRICAL RESISTIVITY
Our electrical resistivity measurements show that the
value of sintered Cu nanoparticle paste was higher than that of
bulk Cu. Considering the composition and structure of sintered
Cu nanoparticle paste, porosity, organic shell residuals and
oxidation during sintering probably decrease the conductivity
of fused materials. To quantitatively understand the porosity
effect on the electrical resistivity, a finite element modelling
approach was used. A body centred cubic (BCC) structure was
employed to capture the particle packing density in the
sintered material. The BCC structure allows for the definition
of a simple cubic unit cell and easy variation of the density. To
characterize the effective resistivity of the unit cell, we apply
a current on one side of the cell while grounding the other side.
Solving the model, results in a voltage gradient over the unit
cell, which is used to calculate the effective resistivity, see
Figure 7. The copper atomic density is 1.530 × 10
15
atom∙cm
-3
and the lattice parameter of BCC unit cell is 361.5 Å. In this
study, we assumed that the neck formation happened between
copper particles in the BCC unit cells.
The variation of porosity was simulated by adjusting the
overlapping distance between each particle. A given
overlapping distance δ between two particles indicates an
interface area A
s
and contact radius x, as shown in Figure 8.
The particle radius is R, so the geometrical relation can be
presented as in equation (3).
(3)
Figure 7. Representative porous Cu nanoparticle arrange in BCC structure.
The overlapping distance between each particle was 0.2 of the particle radius.
Figure 8. Schematic description of particle overlapping parameters.
To small δ, a simple model of resistance in series was used.
The whole resistance Ω can be simplified as below, assumed
that δ
and
:
Ω = Ω
Ω
(4)
Then the resistivity can be derived as following:
(5)
(6)
Where
is equivalent resistivity and
is equivalent
conductivity. It can be seen that equivalent conductivity
is
linearly dependent with x/R.
As shown in Figure 9(b), unit cell equivalent conductivity,
, presents good linear relation with x/R, which agrees with
Equation (6). The slope of linear fit, in the range of 0% to 60%
relative radius, is
. Hence the
corresponding resistivity is 1.74 µΩ∙cm. This is close to the
resistivity of bulk copper, 1.68 µΩ∙cm.
Figure 9. Simulation results of (a) equivalent electrical resistivity, (b)
equivalent electrical conductivity with different relative radius (overlapping
between particles).
IV. CONCLUSION AND OUTLOOK
In summary, a novel lithographic defined stencil printing
method as developed to pattern Cu nanoparticle paste. The
metallic nanoparticle paste has a low sintering temperature of
260 . The paste was used to fabricate 3D interconnect
structures for wafer to wafer bonding. A low electrical
resistivity at 7.84 ± 1.45 µΩ∙cm is obtained after sintering at
260 for 10 min, in 700 mBar forming gas environment.
The result indicates that reducing sintering environment can
improve electrical properties. The effects of porosity of
sintered Cu nanoparticle paste on electrical conductivity was
investigated by finite element simulation. The nanoparticles
were placed in a simple BCC structure unit cell. This model
was used to evaluate the equivalent unit cell conductivity
under different particle packing density. The simulation of
copper nanoparticle porosity presented a linear relation
between the overlapping ratio and electrical conductivity. It
suggests that this BCC unit cell can be used as a representative
model to study porosity effect on electrical resistivity.
In future work, we will investigate automated
distribution/spreading methods of copper nanoparticle paste,
that have higher accuracy and repeatability. In addition,
simulations with more complex and realistic models will be
studied, as well as the effects of the sintering process on the
material properties.
(a)
(b)