LASER MICROVIA DRILLING AND ABLATION OF SILICON USING 355 NM PICO AND
NANOSECOND PULSES
Paper M507
Henrikki Pantsar1, Hans Herfurth1, Stefan Heinemann1, Petri Laakso2, Raimo Penttila2,
Yi Liu3, Golam Newaz4
1 Fraunhofer USA, Inc. Center for Laser Technology, 46025 Port Street, Plymouth, MI 48375
2VTT Technical Research Centre of Finland, Tuotantokatu 2, 53850 Lappeenranta, Finland
3
Wayne State University, 65 Chemistry, 5101 Cass Ave, Detroit, MI 48202
4Wayne State University, 2135 Engineering, 5050 Anthony Wayne, Detroit, MI 48202
Abstract
Laser ablation of silicon has become an intense
research topic due to the rapidly growing interest in
laser processing in the photovoltaics and electronics
industries. Different types of lasers are being used for
edge isolation, grooving, drilling among other
applications, with the pulse width ranging from the
ultrashort femtosecond regime up to long microsecond
pulses. The results may vary significantly depending
on the wavelength and pulse width delivered by the
laser source. In this study, two frequency triplicated
Nd:YVO4 lasers, delivering pulses of width 9 to 12 ps
and 9 to 28 ns, were used to drill holes and form
grooves in silicon wafers. The thickness of the wafers
was 200 µm.
Groove depth and geometry were measured using an
optical 3D profiling system. Results revealed that the
material removal rate was greatly influenced by the
pulse energy and repetition rate when the nanosecond
pulsed laser beam was used. With picosecond laser
beam the volumetric material removal rate remained
rather constant in the range of 100 to 500 kHz, but the
groove width and depth varied.
Scanning and transmission electron microscopy were
used to characterize the drilled holes. Microstructures
were investigated by selected area electron diffraction
patterns. According to the measurements, nanosecond
pulses induce not only thermal, but also mechanical
damage to the hole walls, while picosecond processing
only results in a thin HAZ layer, which is partially
covered with amorphous nanoparticles.
Introduction
Laser micromachining of silicon is of particular
interest in applications such as photovoltaic
applications and microelectronics. Laser ablation
involves numerous concurrent processes including
heating, melting, vaporization and ionization as the
beam interacts with solid, liquid, vapor and plasma
phases at or near the material surface [1]. The process
characteristics are determined by the intensity, duration
and wavelength of the laser pulse. Commercially
available lasers for micromachining include lasers with
pulse durations in the femto, pico and nanosecond
timescale. Typical wavelengths include variations from
uv to near ir.
Femtosecond pulses are optimal for material
processing in many aspects. In the case of sub-ps
ultrashort pulses, the duration of the pulse is less than
the characteristic thermalization time of the material
and machining can be done with very few thermal
effects. Especially in the low-fluence regime in which
the average ablation rate is determined by the optical
penetration depth, the thermal effects are negligible
and close to zero heat affected zones are
experienced.[2,3,4] Another advantage of ultrafast
processing is that the fs pulses terminate before any
material is expulsed from the surface. The complete
energy of the pulse is thus deposited to the sample
target without any laser-plasma interaction during the
pulse.[1,5] Since the heat conduction losses within the
material are minimal and no plasma shielding occurs,
the ablation threshold of materials is the lowest at sub-
ps pulse widths. Material can be removed at extreme
precision using low pulse energies. As the pulse
energy, or the fluence, is increased, thermal ablation
processes become more dominant even with
femtosecond pulses. The complete energy of the pulse
is still delivered into the material, but the ablation
depth is determined by the effective heat penetration
depth instead of the optical penetration depth. Ablation
quality is decreased but the ablated depth per pulse
increases strongly [2].
For applications in machining, laser systems have to be
reliable, robust and affordable. Since technical effort
increases with shortening the pulse duration, the latter
should be as short as necessary, only, for achieving a
ICALEO® 2008 Congress Proceedings Laser Microprocessing Conference
Page 278 of 430
satisfying result [6]. Nanosecond lasers fulfil the above
criteria for the most part. The technology is well
established and proven, rather simple in design and
cost-effective. However, in some cases, the pulse is not
short enough and the processing quality of these lasers
does not meet the requirements. Picosecond laser
sources have proven themselves as a compromize
between the two aforementioned alternatives.
Materials processing with laser pulses of width a few
picosecond resembles much of that of high-fluence
femtosecond processing. The ablation threshold is
slightly higher than for fs-pulses, mainly due to heat
conduction losses and plasma shielding [3]. At 1 ps
pulses the plasma effects are negligible, raising up to a
value of 20% at 10 ps during ablation of gold and
similar findings have been obtained for silicon as well
[1]. Overall, no drastic changes in terms of quality,
thermal effects nor efficiency are observed when the
pulse width remains less than 10 ps, even though the
process can be considered to be purely thermal in
nature [2,3,6,7]. In some cases the quality of ps
processing can even exceed that of fs lasers. fs-laser
induced pressure surges can cause mechanical damage
to the material and lattice defects in silicon [8].
Nanosecond laser processing involves a complex
mixture of concurrent physical processes. In contrast to
femtosecond processing, the long pulse interacts with
material in solid, liquid, vapor and plasma states.
Considerable differences can be seen in the ablation
process depending on the irradiance. For a given pulse
energy, the maximum melt depth increases with longer
pulses, i.e. lower irradiance (Al target)[7]. At the same
time the recoil pressure, which is dependent on the
irradiance [9], decreases causing incomplete melt
ejection from the interaction area. In addition to these
effects, the ablation threshold is higher than that
observed using fs and ps pulses, mainly due to plasma
shielding and greater heat conducton losses.[7] Studies
comparing fs and ns pulses in drilling show even two
times faster ablation rates for fs pulses in comparison
to ns pulses (silicon, 266 nm radiation, 11 J/cm2)
[10,11]. However, at high fluence values, the rate of
ablation with ns pulses increases strongly and exceeds
that of fs and ps pulses [7].
During ns processing the mass ablation rate increases
with laser power density following a power law
dependence up to an irradiance of 0.3 GW/cm2, almost
independent of the target material (brass and glass, 248
nm KrF laser)[12]. At this point, plasma shielding
starts to absorb the latter part of the pulse and the pulse
becomes attenuated. Plasma will reflect and scatter the
beam reducing ablation efficiency.[12] Experimental
data shows that the ablation rate continues to increase
in a linear fashion until an irradiance of 10 to 20
GW/cm2 is reached [13,14,15,16]. At this point the
ablation rate increases sharply. This behavior can be
explained as homogeneous explosive boiling, which is
responsible for ejection of large particles after a finite
delay.[14,15,16] Overall, mass ejection during
nanosecond ablation can be characterized by electron
emission on a picosecond time scale, atomic/ionic
mass ejection on a nanosecond timescale, and large
particle ejection on a microsecond timescale,
continuing up to tens of microseconds [16]
When short nanosecond pulses or picosecond pulses
are used, the irradiance is typically high enough to
initiate plasma formation and result in plasma
absorption. The plasma influence increases with the
pulse duration, power density and wavelength. All of
the energy absorbed by the plasma plume is not,
however, lost from the process, but the plasma can in
fact heat the target material [16]. If an ir laser is used,
the beam mainly heats the peak of the expanding
plume resulting in greater losses, whereas uv radiation
mainly absorbs at the root of the plume delivering
more energy to the material via plasma absorption
[17]. Plasma absorption can also be exploited in some
processes. When laser-induced plasma is formed in
narrow bore drilling, hot plasma expands rapidly inside
the channel and transports a large fraction of its energy
by covection and radiation to the walls of the capillary,
contributing to the radial expansion of the bore. This
effect can stabilize ablation over a wide range of
depths. [17]
Drilling and ablation of silicon has been investigated in
this study. The objective was to compare pico and
nanosecond processing of silicon using 355 nm
ultraviolet radiation. Based on previous referenced data
pico and nanosecond laser sources would be in most
cases the preferential choices for silicon processing
and uv wavelength was selected to increase absorption,
decrease the optical penetration depth into the
underlying material, decrease losses due to plasma
absorption and to reach a longer Rayleigh length
together with a smaller focal spot diameter. Results
have been evaluated based on optical measurements,
SEM and TEM investigations.
Experimental setup
Experiments with nanosecond pulses were carried out
using a q-switched Spectra-Physics HIPPO laser at 355
nm wavelength. The beam was delivered through a
beam expander and a Scanlab Hurryscan 10
galvanometric scanner with 100 mm telecentric optics.
The calcualted focal spot diameter with the setup was
10 µm. The pulse width of the laser varied with the
ICALEO® 2008 Congress Proceedings Laser Microprocessing Conference
Page 279 of 430
frequency being 10.2 ns at 50 kHz, 18.6 ns at 100 kHz
and 28.4 ns at 200 kHz.
For the picosecond processing experiments a Lumera
Rapid laser was used. The output wavelength of the
beam was 355 nm. The optical setup comprised a beam
expander and Scanlab Scangine 10 scanner with a 100
mm telecentric focusing lens. The calculated focal spot
diameter for the optical setup was 10 µm. The pulse
width of the laser was 9 to 12 ps. Laser power of 460
mW was used in all experiments.
Material used for the experiments was 200 µm thick
polished Ph-doped single crystalline silicon wafer.
Samples were ultrasonically cleaned in acetone after
processing. Loose particles and dust were swiped from
the surface before optical measurement.
Experiments for defining the ablation rate with ns and
ps pulses were carried out by ablating grooves on
silicon wafers with variable velocities and repetition
rates. Groove profiles were measured using a Wyko
NT3300 optical 3D profiling system.
Holes were trepan drilled through the wafer using a
specific beam path geometry to remove material more
efficiently from the hole. The beam was programmed
to move along a circle of 30 µm for 54 000 degrees,
equal to 150 rotations. During this movement the beam
was oscillated along a circular path at a frequency of
1500 Hz and an amplitude of 12 µm. The drilling time
was 0.78 s. Focal position was set to the surface for the
time of the drilling. Since the beam movement was
created using scanner mirrors, it is unknown how
precisely the beam follows the programmed path. The
beam motion is presented in Figure 1. All experiments
were carried out in ambient air.
Figure 1. Beam movement during drilling. Yellow
area shows the spot size, ablated area is shown in
gray.
The morphology of the holes was recorded by Hitachi
S-2400 Scanning Electron Microscope (SEM)
operating at 25kV. The microstructure on the edge of
the holes was studied by JEOL FasTEM Transmission
Electron Microscope (TEM) operating at 200kV. The
TEM is equipped with an Electron X-Ray Dispersive
Spectrometry (EDS). For TEM sample preparation, the
holes were filled with M-Bond 610 epoxy to protect
the wall of the holes not being removed by ion-beam
milling as suggested in the literature[8]. The disks
were then cured for two hours at 120°C. Both sides of
the disks were ground by sand paper from 600 Grit
down to 2400 Grit. The final thickness of the disks
were about 40-70µm. Since the thinned disks are very
fragile, they were glued to copper rings in order to
obtain support. The disks were finally polished by ion-
beam milling machine (Gatan 691 Precision Ion
Polishing System-PIPs) at 5kV with 6° tilting until the
glue area not being fully removed.
Results and discussion
Grooves on silicon
Grooves were ablated on silicon surfaces at velocities
of 20, 30, 45, 65, 100, 150, 225, 350 and 500 mm/s.
Repetition rates for the nanosecond laser were varied
between 20 and 200 kHz, and for the picosecond laser
from 100 to 500 kHz. The nanosecond laser could not
deliver the 460 mW power above 200 kHz frequency
and available power from the picosecond laser was
limited below 100 kHz.
The ablation process was limited by the scanning
velocity and frequency in two ways. First, the pulse to
pulse overlapping had a minimum limit below which
material expulsion from the groove was incomplete
and significant amounts of silicon oxides started to
form inside the groove. The upper limit for the
scanning velocity was set by the maximum pulse to
pulse distance, above which pulses form separate spots
on the surface instead of a continuous groove.
For nanosecond processing it was found that in the
whole parameter range from 20 to 200 kHz, clean
consistent grooves with no oxide formation were
achieved only when the pulse overlapping was less
than 80 to 90%. The process tolerated greater
overlapping when the pulse energy was low, i.e. the
frequency was high. The feasible parameter area for
picosecond processing was wider. The pulse
overlapping at 100 and 200 kHz frequencies could be
up to 97% before oxide formation started to interfere
with the process.
Due to the parameter limits of the two lasers, a head to
head comparison could be done only in the frequency
ICALEO® 2008 Congress Proceedings Laser Microprocessing Conference
Page 280 of 430
range of 100 to 200 kHz. Grooves ablated at these
frequencies were measured in more detail to provide
information about the groove depth and ablation rate.
In addition to these, nanosecond experiments were run
also at 50 kHz repetition rate and picosecond
experiments were continued up to 500 kHz repetition
rate. Scanning velocity was set to 225 mm/s.
The profile of the groove was measured across the
ablated line to reveal the depth and cross section area
of the ablated and recast materials. The term groove
volume here in after refers to the volume ablated below
the original surface. The term removed material refers
to the amount of silicon removed completely from the
source; i.e. groove area minus recast area. Volume
values here are presented in the units of µm3, which is
the area in question measured from the cross section
multiplied by a length of 1 µm along the longitude of
the groove. Since the profiles are derived from a line
measurement across the groove and not from a
measurement of the actual volume, the results are not
precise. However, they represent a good estimation of
the average cross section of the grooves.
Results show that the ablation rate with nanosecond
pulses was significantly impacted by the frequency or
the pulse energy, whereas the ablation rate with
picosecond pulses was independent of the frequency
within the tested parameter area. With nanosecond
pulses, the groove volume increased markedly with the
pulse energy. 50 kHz repetition rate, equaling to 9.2 µJ
pulse energy, created a groove with a cross sectional
area of 26.3 µm2. At this fluence the amount of recast
was small and the removed volume measured from the
cross section of the groove was 24.2 µm3.
Increasing the frequency resulted in a groove
geometry, which was narrower and shallower than that
created with higher pulse energies. Also the relative
volume of recast compared to the groove volume
increased significantly. At 200 kHz repetition rate (2.3
µJ) the groove volume was 5.8 µm3 and taking the
recast into consideration, the volume of the removed
material was only 4.0 µm3. In this case more than 30%
of the material removed from the groove was being
recast on the edges of the groove and not ablated away.
The depth of the groove fluctuated significantly
between 0 to 3.5 µm. Therefore, the profile for the
200 kHz sample was derived from an average value of
three individual measurements, in order to obtain a
better estimation of the ablated volume. The cross
sections of the grooves ablated with nanosecond pulses
are presented in Figure 2. Grooves ablated at 225 mm/s
scanning velocity using 50 and 200 kHz repetition
rates are presented in Figure 3 and Figure 4,
respectively.
Figure 2: Measured cross sections of grooves ablated
with the nanosecond laser.
Figure 3. Groove ablated by nanosecond pulses.
Scanning velocity 225 mm/s, repetition rate 50 kHz.
Figure 4. Groove ablated by nanosecond pulses.
Scanning velocity 225 mm/s, repetition rate 200 kHz.
As the line energy in each case was equal, a
substantially greater part of the laser energy was being
lost in the ablation process when the repetition rate was
increased gradually from 50 to 200 kHz. This increase
ICALEO® 2008 Congress Proceedings Laser Microprocessing Conference
Page 281 of 430
in the frequency caused the pulse width to change from
10.2 ns to 28.4 ns and the pulse energy to decrease
from 9.2 to 2.3 µJ. Both of these factors reduced the
mean irradiance in the area of the beam, which
changed from 1.15 to 0.10 GW/cm2. At the same time,
the process became more unstable and fluctuations in
the groove depth and width were more evident.
Longer pulses can be absorbed into or reflected from
the laser induced plasma in a greater extent. The
threshold for plasma formation for many materials is in
the approximity of 0.3 GW/cm2[12]. Since the average
irradiance at 200 kHz was only 0.10 GW/cm2 and the
peak irradiance at the center of the beam was 0.2
GW/cm2, plasma shielding should not play a role at
higher repetition rates, but rather at low frequencies.
Particles hovering above the interaction point can,
however affect the ablation process, especially at
higher repetition rates. The extent of such inter-pulse
plasma/plume effects could not be estimated based on
the conducted experiments.
More likely causes for low material removal rates at
high frequencies are related to the pulse irradiance.
Working closer to the ablation threshold with longer
pulses leads to a situation where a greater part of the
pulse energy is being used to heat the material in the
solid and liquid phases than to evaporate and remove
material. At the same time the recoil pressure, which
is proportional to the irradiation [9,18], is decreased
reducing melt expulsion from the groove.
Material removal with ns pulses was approximately
twice as efficient than with picosecond pulses when
the repetition rate was 100 kHz (4.6 µJ pulse energy).
Nanosecond pulses created a groove volume of
16.7 µm3 compared to the 7.9 µm3 of picosecond
pulses. At 200 kHz, the grooves became approximately
equal in volume with the picosecond groove being 6.2
µm3 in volume and the nanosecond groove 5.8 µm3.
However, a lesser amount of recast silicon was present
at picosecond groove edges and the absolute material
removal with picosecond pulses was 5.8 µm3 and 4.0
µm3