clue-虚无
Industrial femtosecond lasers
and material processing
01/22/2019
NORMAN HODGSON, MICHAEL LAHA, TONY S.
LEE, ALBRECHT
STEINKOPFF,
and
SEBASTIAN
HEMING
Over the last five
years, material processing with femtosecond pulses
in the range of
300 to 900 fs has
gained in popularity due to the small heat-
affected zone (HAZ) and
increased
energy penetration depth resulting from the high
laser pulse intensity.
Industrial
ultrashort-pulse (USP) diode-pumped solid-state
and fiber lasers are now
being used to
cut foils for flat panel displays, to cut stents,
and to drill fuel injector
nozzles, as
well as for wafer scribing and surface
microstructuring.
The first industrial
use of femtosecond laser pulses for
microprocessing dates back to
the late
1990s, where titanium sapphire (Ti:sapphire)
amplifiers were used to repair
lithography masks in integrated circuit
fabrication. At that time, the only choice in
commercial laser sources were
Q
-switched, neodymium
(Nd)-doped solid-state lasers
delivering pulse duration of tens of
nanoseconds, and ultrafast Ti:sapphire amplifiers
that exhibited pulse durations of 100
fs and provided output power at the 1 W level at
1 kHz (FIGURE 1). The small feature
size of the chromium layer on top of the fused
silica substrate prohibited the use of
nanosecond pulses due to thermal damage of the
chromium, and even the substrate. The
necessity for a small HAZ outweighed the
high cost of a Ti:sapphire amplifier
system, which was about $$300,000.
FIGURE 1.
Lithography mask
repair using a 100 fs Ti:sapphire laser is shown,
where a chromium layer on a fused
silica substrate is ablated; width of the ablated
lines is 750 nm.
1
Today, many different femtosecond
lasers are available, providing output pulse
energies of up to 200 ?
J,
with average output powers in the kilowatt range.
Especially
over the last decade, a
large variety of femtosecond solid-state laser and
fiber laser
architectures have entered
the material processing field, all of them based
on
ytterbium (Yb)-doped gain materials.
Average output powers of up to 100 W are
currently employed in industrial
applications, with pulse durations between 300 and
900 fs and repetition rates of up to 2
MHz. Compared to the original Ti:sapphire
amplifier systems, the output powers
have increased by two orders of magnitude,
while system cost has been considerably
decreased at the same time.
A majority
of the low-power lasers have been deployed in
ophthalmic applications. In
the 2000s,
Nd:glass regenerative amplifiers were used to cut
the corneal flap in
LASIK surgery by
generating a bubble plane inside the cornea. These
Nd:glass
systems were later replaced by
Yb:fiber master-oscillator power-amplifiers
(MOPAs),
which allowed higher
repetition rates at a lower cost (FIGURE 2).
Typical pulse
energies required for
flap cutting are 2
–
4
?
J at repetition rates of 50 to 200 kHz
and
pulse durations of around 300 fs. A
second, more recently emerging ophthalmic
application is lens dissection as part
of cataract surgery. In this case, the pulse
energies employer are in the range of
20 to 40 ?
J at repetition rates of 50
to 100 kHz,
and pulse durations are
preferably below 800 fs.
FIGURE 2.
The
evolution of the deployment of femtosecond lasers
in
microprocessing applications is
shown; Ti:sapphire regenerative amplifiers for
mask
repair have been replaced by
picosecond Nd:YVO
4
lasers
and Nd:glass regenerative
amplifiers by
femtosecond Yb fiber MOPAs.
According to a forecast by Strategies
Unlimited in 2016, the total market in 2019 for
femtosecond and picosecond lasers used
for material processing (including
ophthalmic applications) is projected
to be $$460 million (TABLE
1).
2
Half of this
revenue is generated by picosecond
lasers, which have been widely used in
microelectronic manufacturing. The
other half is split between ophthalmic
femtosecond lasers ($$136 million) and
femtosecond lasers for non-biological material
processing ($$98 million).
TABLE 1.
The 2016 Revenue
Forecast of femtosecond and picosecond lasers in
material processing is shown, where
values for 2014 and 2015 are actual revenues;
femtosecond lasers include ophthalmic
lasers.
2
Mechanism and benefits of femtosecond
laser processing
The interaction of
femtosecond and picosecond pulses with matter is
governed by the
absorption of the light
by the electrons and subsequent energy transfer to
the lattice.
In the case of metals, the
photons are absorbed by the electron gas, which
increases its
temperatures to values of
several 10,000°
C. The electrons will
transfer their energy to
the lattice
within the electron-phonon relaxation time, which
for most materials is in
the range of
100 fs to 1 ps at room temperature. The lattice
has about 100X higher
heat capacity
compared to the electrons. This leads to a
substantial delay between the
incidence
of the laser pulse and the time when the lattice
has reached melting
temperature (FIGURE
3). For high laser fluences, the ablation of the
heated material
occurs several tens of
picoseconds after the laser pulse is absorbed.
FIGURE 3.
Interaction of ultrafast pulses with a
metal is shown. The electron gas
absorbs the laser light, leading to a
hot, thermalized electron distribution within 100
fs; the lattice will heat up with a
delay of 4 to 30 ps.
The
light-matter interaction for ultrashort laser
pulses can be mathematically
described
by the Two-Temperature Model, which provides the
temporal and spatial
evolution of the
temperature of electron gas and lattice by
incorporating the coupling
between both
systems via the
electron
–
phonon relaxation
time.
3
This model has been
used very successfully over the last
decades to calculate damage threshold fluence,
ablation rates, and HAZ for ultrashort
pulse processing (FIGURE 4).
4-8
The main
results are that
for pulses that are shorter than 10 ps, the damage
threshold fluence
remains constant,
while for larger pulse durations, the threshold
fluence increases in
proportion to the
square-root of the pulse duration, independent of
incident laser pulse
wavelength.
FIGURE 4.
Calculated temporal temperature
distribution of electron gas (left) and
lattice for copper after irradiation
with a 100 fs pulse at an average fluence of 0.14
J/cm
2
and a
wavelength of 800 nm.
7
Similarly, the HAZ remains constant for
pulse durations below 10 ps, again
independent of the wavelength of the
laser light. The basic reason for this behavior is
the delay in the temperature increase
of the lattice and of heat conduction into the
material. This regime of pulsed laser-
matter interaction is therefore referred to as
cold
ablation, since the lattice stays
cold during irradiation by the laser pulse. This
name is
a bit misleading, though, since
the material will have to reach melting
temperature to
induce ablation.
The most interesting effect of
ultrashort pulse interaction, however, is the
increase in
the energy penetration
depth and ablation depth with decreasing pulse
duration.
Decreasing the pulse duration
at a given energy fluence leads to an increase in
the
temperature of the electron gas and
a simultaneous increase of the electron-phonon
relaxation time.
In a more
mechanical model, this can be easily understood:
the velocity of the
electrons traveling
through the lattice can become as high as 100,000
m/s for
femtosecond pulses due to the
very high intensity, and this high speed results
in
deeper penetration of the electron
into the lattice without transferring energy to
the
lattice.
8
The material processing
efficiency and quality depends on the duration of
the laser
pulses. For pulses in the
nanosecond regime, the absorption of the laser
pulse is
determined by the linear
optical absorption depth of the laser light and
the energy
dissipation is a result of
heat conduction into the material (FIGURE 5). For
pulses that
are shorter than 10 ps, the
initial energy penetration depends strongly on the
light
intensity and leads to deeper
penetration depth for femtosecond pulses. In
addition,
the lack of thermal
conduction during the pulse and the time of
lattice heating results
in a very low
HAZ. For metals, HAZ of less than 5 ?
m
can be achieved, while for
plastic
materials, the HAZ is typically in the range of 30
to 50 ?
m.
FIGURE
5.
Absorption of a laser pulse in a
medium is shown, where energy
penetration is represented in blue and
volume heated via heat conduction in
red.
The increase of the
penetration depth for shorter pulses leads to an
increase in the
maximum ablation rates
when the pulses become shorter than about 20 ps
(as shown
in FIGURE 6 for aluminum). As
will be discussed subsequently, at any pulse
duration
the maximum ablation rate is
achieved at a pulse fluence of about 7.5X the
ablation
threshold fluence. Compared to
Q
-switched laser pulses with
pulse durations of tens
of nanoseconds,
the increased electron velocity in the material
leads to ablation rates
for sub-
picosecond pulses that are only a factor of three
lower.
FIGURE 6.
Maximum ablation rates of aluminum with
Coherent's measured values
(red dots)
and using values taken from data published by
Breitling et al. (blue
dots).
9
For
pulsed lasers, ablation becomes most efficient at
a pulse energy fluence equal to
e
2
times the
threshold fluence. This is a result of the
saturation of the ablated volume
with
increasing pulse fluence. At a fixed output power,
more volume can therefore be
ablated if
the energy fluence is lowered, while
simultaneously increasing pulse
repetition rate and therefore
throughput (FIGURE 7). The maximum value is
achieved
at the optimum fluence. The
ablation rate, C (in
mm
3
/W/min), is given by:
where
F
is the peak fluence in
J/mm
2
,
F
th
is the peak
threshold fluence, and
d
is
the
energy penetration depth per pulse
in millimeters.
10
Typical
values for the energy
penetration depth
per pulse are in the range of
20 to
100 nm for metals,
semiconductors, and
plastics, and >500 nm for glasses and transparent
crystals.
FIGURE 7.
Ablated volume per pulse (blue curve)
and ablated volume per second
(green
curve) at constant laser power are
shown.
10
FIGURE 8
shows the measured ablation rates for Nitinol as a
function of the average
pulse fluence
for various pulse durations. Nitinol is used to
manufacture stents and
requires
femtosecond pulses to increase the throughput. The
laser was a 1035 nm
modelocked fiber
MOPA, where the settings of the compressor were
changed to
implement different pulse
durations, and external second- and third-harmonic
generation (SHG and THG, respectively)
stages could be added to access wavelengths
of 517 and 345 nm. A rectangular area
of about 0.5 ×
1.5 mm was ablated to a
depth
of about 200 ?
m using
a spot diameter of 20 ?
m with a spot
overlap of 60% in both
directions.
Decreasing the pulse duration to 400 fs resulted
in an increase in the
ablation rate by
a factor of two compared to 19 ps pulses. This
increase in ablation
efficiency for
femtosecond pulses is common amongst metals and
semiconductors.
For transparent
dielectrics, however, the ablation rate is higher
for picosecond pulses.
This is due to
the lack of free electrons, which requires high
threshold fluences to
ablate the
material. For femtosecond pulses, the resulting
peak powers are too high
and lead to
optical breakdown and decreased ablation rates.