U.S. patent application number 11/067464 was filed with the patent office on 2006-06-15 for multiple-wavelength laser micromachining of semiconductor devices.
Invention is credited to Kelly Bruland, Robert F. Hainsey, Richard Harris, Ho Wai Lo, Lei Sun, Yunlong Sun.
Application Number | 20060128073 11/067464 |
Document ID | / |
Family ID | 39036703 |
Filed Date | 2006-06-15 |
United States Patent
Application |
20060128073 |
Kind Code |
A1 |
Sun; Yunlong ; et
al. |
June 15, 2006 |
Multiple-wavelength laser micromachining of semiconductor
devices
Abstract
A specially shaped laser pulse energy profile characterized by
different laser wavelengths at different times of the profile
provides reduced, controlled jitter to enable semiconductor device
micromachining that achieves high quality processing and a smaller
possible spot size.
Inventors: |
Sun; Yunlong; (Beaverton,
OR) ; Harris; Richard; (Portland, OR) ;
Bruland; Kelly; (Portland, OR) ; Hainsey; Robert
F.; (Portland, OR) ; Lo; Ho Wai; (Portland,
OR) ; Sun; Lei; (Aloha, OR) |
Correspondence
Address: |
STOEL RIVES LLP
900 SW FIFTH AVENUE
SUITE 2600
PORTLAND
OR
97204-1268
US
|
Family ID: |
39036703 |
Appl. No.: |
11/067464 |
Filed: |
February 25, 2005 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
60635054 |
Dec 9, 2004 |
|
|
|
Current U.S.
Class: |
438/132 ;
219/121.68; 219/121.69; 438/131 |
Current CPC
Class: |
B23K 2103/50 20180801;
H01L 23/5258 20130101; H01L 2924/0002 20130101; B23K 26/0613
20130101; B23K 26/0604 20130101; B23K 26/0622 20151001; H01S
3/10046 20130101; B23K 2103/08 20180801; B23K 2103/10 20180801;
H01L 2924/0002 20130101; B23K 2103/172 20180801; Y10S 438/94
20130101; B23K 26/361 20151001; B23K 26/067 20130101; B23K 2103/12
20180801; B23K 2103/14 20180801; B23K 26/40 20130101; H01L 2924/00
20130101; B23K 2103/26 20180801; H01S 3/2383 20130101 |
Class at
Publication: |
438/132 ;
438/131; 219/121.69; 219/121.68 |
International
Class: |
H01L 21/82 20060101
H01L021/82; B23K 26/16 20060101 B23K026/16 |
Claims
1. A method of laser micromachining a multilayer structure to
remove depthwise a portion of a target layer material without
causing appreciable damage to nearby nontarget layer material of
the multilayer structure, comprising: generating a laser pulse with
an energy profile comprised of first and second energy profile
parts, the laser pulse including a first laser energy feature at a
first laser wavelength in the first part of the energy profile, and
a second laser energy feature at a second laser wavelength in the
second part of the energy profile; directing the laser pulse to the
target layer material; the first laser energy feature at the first
wavelength in the first part of the energy profile removing
depthwise an initial portion of the target layer material to form
part of an open volumetric region and to not damage nontarget layer
material of the multilayer structure; and the second laser energy
feature at second laser wavelength in the second part of the energy
profile removing depthwise a remaining portion of the target layer
material to complete formation of the open volumetric region and to
not damage the nontarget layer material in a vicinity below or
adjacent the volumetric open region.
2. The method of claim 1, in which the multilayer structure
comprises an electrically conductive link positioned between an
upper passivation layer and a lower passivation layer in a stack;
and in which the initial portion of the target layer material
includes a region of the upper passivation layer, the remaining
portion of the target layer material includes a region of the
electrically conductive link, and the nontarget layer material
includes a region of the lower passivation layer in a vicinity
below the open volumetric region.
3. The method of claim 2, in which the upper passivation layer and
the electrically conductive link contact each other at a boundary
interface, and in which the initial portion of the target layer
material removed further includes a portion removed from the region
of the conductive link at the boundary interface.
4. The method of claim 2, in which the first laser energy is of an
amount that ruptures the upper passivation layer without cracking
neighboring upper passivation layer structure.
5. The method of claim 4, in which the first wavelength is within
the ultraviolet wavelength range.
6. The method of claim 2, in which one or more of the upper and
lower passivation layers are made of low k materials.
7. The method of claim 2, in which the electrically conductive link
includes aluminum, copper, gold nickel, titanium, tungsten,
platinum, nickel chromide, titanium, tantalum nitride, tungsten
silicide, or other metal-like materials.
8. The method of claim 1, in which the first wavelength corresponds
to an effective laser beam spot size that defines a surface
dimension of the open volumetric region.
9. The method of claim 1, in which the first wavelength is within
the visible or infrared wavelength range, and in which the
nontarget layer material in a vicinity below the open volumetric
region is substantially transparent to the first wavelength and
thereby undergoes no appreciable damage in response to the first
wavelength.
10. The method of claim 1, in which the second wavelength is within
the visible or infrared wavelength range, and in which the
nontarget layer material in a vicinity below the open volumetric
region is substantially transparent to the second wavelength and
thereby undergoes no appreciable damage in response to the second
wavelength.
11. The method of claim 1, in which the nontarget layer material
has a damage threshold, the first wavelength is within the
ultraviolet wavelength range, and the first wavelength is absorbed
by the nontarget layer material but has a first laser energy amount
that is below the damage threshold of the nontarget layer
material.
12. The method of claim 1, in which: the laser pulse has a
primarily Gaussian-shaped beam energy profile with a center region,
and the energy profile has its highest energy amounts concentrated
at the center region; the multilayer structure comprises a
passivation layer positioned between an electrically conductive
link and a substrate, the electrically conductive link comprising
part of the target layer material and having a width; and the
nontarget material includes the passivation layer, and the width of
the electrically conductive link causing it to function as a shield
that prevents damage to the passivation layer by the highest energy
amounts of the first laser output.
13. The method of claim 1, in which the laser pulse has leading
edge rising time of shorter than 10 ns and a total duration of
longer than 5 ns.
14. The method of claim 1, in which the first laser energy feature
lasts from 1 ps to 50 ns, the second laser energy feature lasts
from 1 ps to 50 ns.
15. The method of claim 1, in which the first and second energy
features include respective first and second laser energy peaks,
and in which the first laser energy peak and the second laser
energy peak are separated by a time delay ranging from zero to 300
ns.
16. The method of claim 1, in which the first and second laser
wavelengths are within the infrared to UV wavelength ranges,
including 1.32 .mu.m, 1.30 .mu.m, 1.064 .mu.m, 1.053 .mu.m, 1.047
.mu.m and their respective second, third, and fourth harmonics.
17. The method of claim 16, in which the first and second laser
wavelengths are the same.
18. The method of claim 16, in which the first laser wavelength is
shorter than the second laser wavelength.
19. The method of claim 1, in which the first laser energy is about
from 0.001 uJ to 20 uJ and the second laser energy is from about
0.001 uJ to about 20 uJ.
20. The method of claim 1, in which the laser pulse repetition rate
is from about 1 Hz to about 200 KHz.
21. A method of laser micromachining a multilayer structure to
remove depthwise a portion of a target layer material without
causing appreciable damage to nearby nontarget layer material of
the multilayer structure, comprising: directing a first laser
output of a first wavelength at a first laser energy for incidence
on target layer material, the first wavelength and first laser
energy cooperating to remove depthwise an initial portion of the
target layer material to form part of an open volumetric region and
to not damage nontarget layer material of the multilayer structure;
and directing a second laser output of a second wavelength and a
second laser energy for incidence on the target layer material, the
second wavelength and second laser energy cooperating to remove
depthwise a remaining portion of the target layer material to
complete formation of the open volumetric region and to not damage
the nontarget layer material in a vicinity below or adjacent the
volumetric open region.
22. The method of claim 21, in which the multilayer structure
comprises an electrically conductive link positioned between an
upper passivation layer and a lower passivation layer in a stack;
and in which the initial portion of the target layer material
includes a region of the upper passivation layer, the remaining
portion of the target layer material includes a region of the
electrically conductive link, and the nontarget layer material
includes a region of the lower passivation layer in a vicinity
below the open volumetric region.
23. The method of claim 22, in which the upper passivation layer
and the electrically conductive link contact each other at a
boundary interface, and in which the initial portion of the target
layer material removed further includes a portion removed from the
region of the conductive link at the boundary interface.
24. The method of claim 22, in which the first laser energy is of
an amount that ruptures the upper passivation layer without
cracking neighboring upper passivation layer structure.
25. The method of claim 24, in which the first wavelength is within
the ultraviolet wavelength range.
26. The method of claim 21, in which the first wavelength is within
the visible or infrared wavelength range, and in which the
nontarget layer material in a vicinity below the open volumetric
region is substantially transparent to the first wavelength and
thereby undergoes no appreciable damage in response to the first
wavelength.
27. The method of claim 21, in which the second wavelength is
within the visible or infrared wavelength range, and in which the
nontarget layer material in a vicinity below the open volumetric
region is substantially transparent to the second wavelength and
thereby undergoes no appreciable damage in response to the second
wavelength.
28. The method of claim 21, in which the nontarget layer material
has a damage threshold, the first wavelength is within the
ultraviolet wavelength range, and the first wavelength is absorbed
by the nontarget layer material but has a first laser energy amount
that is below the damage threshold of the nontarget layer
material.
29. The method of claim 21, in which: the first laser output has a
primarily Gaussian-shaped beam energy profile with a center region,
and the energy profile has its highest energy amounts concentrated
at the center region; the multilayer structure comprises a
passivation layer positioned between an electrically conductive
link and a substrate, the electrically conductive link comprising
part of the target layer material and having a width; and the
nontarget material includes the passivation layer, and the width of
the electrically conductive link causing it to function as a shield
that prevents damage to the passivation layer by the highest energy
amounts of the first laser output.
30. The method of claim 21, in which the first and second laser
outputs comprise respective first and second energy peaks separated
by a time delay, and further comprising forming a beam
characterized by a laser pulse profile with separate characteristic
features corresponding at least in part to the first and second
energy peaks.
Description
RELATED APPLICATION
[0001] This application claims benefit of U.S. Provisional Patent
Application No. 60/635,054, filed Dec. 9, 2004.
COPYRIGHT NOTICE
[0002] .COPYRGT. 2005 Electro Scientific Industries, Inc. A portion
of the disclosure of this patent document contains material that is
subject to copyright protection. The copyright owner has no
objection to the facsimile reproduction by anyone of the patent
document or the patent disclosure, as it appears in the Patent and
Trademark Office patent file or records, but otherwise reserves all
copyright rights whatsoever. 37 CFR .sctn. 1.71(d).
TECHNICAL FIELD
[0003] The inventions relate generally to laser processing
multilayer workpiece materials and, in particular, using a
substantially jitter free, multiple-wavelength laser energy profile
targeted for semiconductor device micromachining to achieve high
quality processing and a smaller possible spot size in which laser
energies at the multiple wavelengths can or may overlap within the
laser energy profile.
BACKGROUND OF THE INVENTION
[0004] Yields in IC device fabrication processes are often impacted
by defects resulting from alignment variations of subsurface layers
or patterns, particulate contaminants, or defects in the substrate
material itself. FIGS. 1, 2A, and 2B show repetitive electronic
circuits 10 of an IC memory device or workpiece 12 that are
typically fabricated in rows or columns to include multiple
iterations of redundant circuit elements 14, such as spare rows 16
and columns 18 of memory cells 20. With reference to FIGS. 1, 2A,
and 2B, circuits 10 are also designed to include particular laser
severable circuit links 22 between electrical contacts 24 that can
be removed to disconnect a defective memory cell 20, for example,
and substitute a replacement redundant cell 26 in a memory device
such as a DRAM, an SRAM, or an embedded memory. Similar techniques
are also used to sever links to program a logic product, gate
arrays, or ASICs.
[0005] Links 22 are designed with conventional link widths 28 of
about 1.0 micron, link lengths 30, and element-to-element pitches
(center-to-center spacings) 32 of about 1.5 microns or less from
adjacent circuit structures or elements 34, such as link structures
36. Link dimensions and pitches are continually being reduced by
device manufacturers. Although the most prevalent link materials
have been polysilicon and like compositions, memory manufacturers
have more recently adopted a variety of more conductive metallic
link materials that may include, but are not limited to, aluminum,
copper, gold nickel, titanium, tungsten, platinum, as well as other
metals, metal alloys such as nickel chromide, metal nitrides such
as titanium or tantalum nitride, metal silicides such as tungsten
silicide, or other metal-like materials.
[0006] Circuits 10, circuit elements 14, or cells 20 are tested for
defects. The links to be severed for correcting the defects are
determined from device test data, and the locations of these links
are mapped into a database or program. Laser pulses have been
employed for more than 20 years to sever circuit links 22. FIGS. 2A
and 2B show a laser spot 38 of spot size diameter 40 impinging a
link structure 36 composed of a link 22 positioned above a silicon
substrate 42 and between component layers of a passivation layer
stack including an overlying passivation layer 44 (shown in FIG. 2A
but not in FIG. 2B) and an underlying passivation layer 46 (shown
in FIG. 2B but not in FIG. 2A). FIG. 2C is a fragmentary
cross-sectional side view of the link structure of FIG. 2B after
the link 22 is removed by the laser pulse.
[0007] The prior art uses laser pulses comprised of only a single
laser wavelength for semiconductor device link processing. A single
laser pulse at a 1064 nm or 1047 nm wavelength has been widely used
for semiconductor memory chip link on-the-fly processing, which
entails severing individual links with a single pulse for each link
while not stopping beam positioner motion. A laser pulse at 1320 nm
became preferable later in metal link processing because it caused
less damage to a silicon substrate. Link processing with a UV laser
pulse also has been proposed and practiced. Double pulse processing
of fat (i.e., thick) copper links has been attempted by a few
users. All of the laser pulses used were at the same
wavelength.
[0008] Wavelengths advantageous for minimizing silicon substrate
damage and enhancing a process window are near 1300 nm, as
disclosed in U.S. Pat. No. 5,265,114, which is assigned to the
assignee of this patent application. However, the smallest
practical laser beam spot size at 1300 nm is about 1.7 microns. The
ever-shrinking feature size or link dimensions of semiconductor
memory chips demand a laser beam spot size of 1.4 microns and
smaller. Using a short wavelength in the UV spectral range, as
disclosed in U.S. Pat. No. 6,057,180, which is assigned to the
assignee of this patent application, can deliver the small beam
spot size needed and cut through an overlying passivation layer but
requires that the passivation material absorb the UV wavelength to
protect the silicon substrate. Moreover, the link structure design
should cooperate with the underlying passivation layer structure to
inflict only minor damage to the underlying passivation material.
Using a short wavelength in the green/visible range would carry a
high risk of damage to the silicon substrate because of its high
absorption of wavelengths in the green/visible range.
[0009] What is desired for purposes of semiconductor device
micromachining is a series of special laser pulses, each with an
energy profile comprised of different laser wavelengths at
different times within the energy profile sequenced to the
different processing characteristics of the layers in the
multi-layer structure. One such energy profile sequence would be a
first part of the laser pulse energy profile at a UV or green
wavelength to best process the overlying passivation layer and top
part of the link material, followed by a second part of the laser
pulse energy profile at a 1.3 micron wavelength to clear the
remaining link material while limiting risk of damage to the
underlying passivation layer and silicon wafer substrate.
SUMMARY OF THE INVENTION
[0010] Preferred embodiments of the inventions entail the use of
laser pulses of different wavelengths propagating from two or more
lasers to form a laser pulse with an energy profile comprised of
different laser wavelengths at different times within the energy
profile, with little or no jitter, for incidence on and processing
multilayer structures. The other laser parameters may be the same
or different. Semiconductor device link processing is described as
a preferred embodiment with reference to link cutting. The use of
laser pulses in accordance with the inventions described with
reference to preferred embodiments is also applicable to other
laser processing operations, such as via drilling. Typically, a
first part of the laser pulse energy profile is at a short
wavelength, such as UV or green wavelength, and is followed by a
second part of the laser pulse energy profile at a longer
wavelength, such as visible or IR wavelength. The time delay
between the UV/green laser energy and the visible/IR laser energy
is controllable based on the process and target structure. The
UV/green laser energy cuts or ruptures the overlying passivation
layer and removes part of the link material; then the subsequent
visible/IR laser energy removes the remaining link material. Use of
the visible/IR laser energy carries much less risk of damage to the
underlying passivation layer. Because less visible/IR laser energy
is needed after the link structure has been partly processed by the
UV or green laser pulse, there is much less risk of damage to the
silicon wafer substrate by the visible/IR laser pulse.
[0011] Processing a semiconductor device link in accordance with
the inventions is characterized by several features or aspects. The
first is a formation of the laser pulse with desired energy profile
comprised of different laser wavelengths at different times. The
energy profile is well controlled, and there is little or no time
jitter between laser energies at the different laser wavelengths,
to keep the overall laser pulse energy profile stable. The second
is selection of different preferred laser energy levels and
wavelengths at different link processing stages of the link
structure, such as first using the UV or green laser energy and
thereafter using the visible/IR laser energy. The third is a link
processing system implemented to process a semiconductor device
link on the fly, with the laser pulse comprised of the desired
energy profile and wavelength division.
[0012] The inventions enable processing of links with narrower link
widths, denser pitch sizes, higher thickness-to-width ratios, as
well as more complicated passivation layer structures and process
fragile passivation materials. In the case of a laser pulse
comprised of UV and visible laser energy, the leading UV laser
energy processes the overlying passivation layer with less risk of
creating a large crater or causing formation of cracks in the
passivation structure, and the trailing visible/IR laser energy
removes the remaining link material with less risk of damage to the
underlying passivation layer and silicon substrate. When the
visible laser energy is chosen in the green and blue spectrum, the
total effective laser beam spot size is greatly reduced compared
with the prior art of using single laser wavelength at IR. The
parameters of the UV and IR/visible laser energies and their timing
can be adjusted for best results, based on the link structure.
[0013] Substantially jitter-free formation of a laser pulse with an
energy profile comprised of different laser wavelengths provides a
stable and unique laser energy profile with multiple energy peaks
at different wavelengths with reduced, controlled time jitter.
Reducing laser pulse profile jitter through synchronous drive
signals to the respective generators of laser energy at different
wavelengths, initiation of laser energy buildup at different laser
wavelength by injection locking, or both, enables controllable
time-displaced wavelength peaks of short inter-peak separation.
[0014] Additional aspects and advantages will be apparent from the
following detailed description of preferred embodiments, which
proceeds with reference to the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0015] FIG. 1 is a schematic diagram of a portion of a DRAM showing
the redundant layout of and programmable links in a spare row of
generic circuit cells.
[0016] FIG. 2A is a fragmentary cross-sectional side view of a
conventional, large semiconductor link structure receiving a laser
pulse characterized by prior art pulse parameters.
[0017] FIG. 2B is a fragmentary top view of the link structure and
the laser pulse of FIG. 2A, together with an adjacent circuit
structure.
[0018] FIG. 2C is a fragmentary cross-sectional side view of the
link structure of FIG. 2B after the link is removed by the prior
art laser pulse.
[0019] FIG. 3 is a simplified general block diagram of a laser
system embodiment configured with two laser heads coupled by
injection locking, the outputs of which are used to form a pulsed
laser output beam with a pulse energy profile comprised of two
different wavelengths.
[0020] FIGS. 4A and 4B show an example of series of laser energies
at different laser wavelengths used to form the specially shaped
laser pulse energy profile shown in FIG. 4C, in which there are two
partly overlapped energy peaks at different wavelengths from the
two laser heads of FIG. 3.
[0021] FIGS. 5A, 5B, and 5C show an example of series of laser
energies at different laser wavelengths used to form a specially
shaped laser pulse energy profile of two nonoverlapped energy peaks
at different wavelengths from the two laser heads of FIG. 3.
[0022] FIG. 6A is a simplified general block diagram of a prior art
system for combining the outputs of two pulsed lasers.
[0023] FIG. 6B is an oscilloscope trace showing the effects of
laser pulse jitter exhibited by the prior art laser pulse
combination system of FIG. 6A.
[0024] FIG. 7 is an oscilloscope trace showing the laser energy
profile realized by a laser pulse generation system implemented in
accordance with the present inventions.
[0025] FIG. 8 is a simplified block diagram of a laser pulse
generation system embodiment configured with synchronized RF drive
signals applied to the Q-switches of two laser heads, the outputs
of which form pulsed laser energy characterized by stable output
energy profile features at different wavelengths.
[0026] FIGS. 9A and 9B show alternative implementations of the RF
signal driver of FIG. 8.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
[0027] Preferred embodiments use two laser heads from which
propagate laser pulse energies at different wavelengths, together
with other laser parameters that are either the same or different,
to form a specially shaped laser energy profile with multiple
energy peaks at different laser wavelengths to process an
electrically conductive link on an integrated circuit chip of the
type shown in FIGS. 1 and 2A-2C.
[0028] One preferred embodiment entails the use of laser energy
propagating from a UV harmonic wavelength laser head at the
beginning of the laser energy profile, which occurs at the
beginning of a process timing sequence, followed by laser energy
from a visible wavelength laser head, such as a green or blue
laser. The time delay between the UV laser energy peak and the
visible laser energy peak is controllable, based on the process and
target structure, and can practically be from 0 ns to 300 ns-500
ns. Within the 500 ns time range, a beam positioning system (not
shown) moves less than 0.1 micron; therefore, the two laser energy
peaks within the laser energy profile are incident on the same link
width on-the-fly (i.e., the positioning system is kept in motion),
as in the case of a single laser pulse.
[0029] Because of absorption by the passivation material of the UV
laser energy, the UV laser energy either directly cuts through the
passivation layer overlying the link or the overlying passivation
layer undergoes an increase in temperature along the laser beam
path, thereby resulting in a reliable and consistent rupture of the
overlying passivation layer without introducing cracks in the
passivation layer structure. This is especially important when the
link width is narrow, link thickness-to-width ratio is high, and
passivation layer structure is weak at the bottom of the link, or
the passivation layer is made of fragile low k material, such as
SiLK.
[0030] The UV laser energy is chosen such that it ruptures the
overlying passivation layer and removes part of the link material
to form part of an open volumetric region. There is a part of the
link material remaining after completion of the UV segment of the
laser pulse energy profile. The center of the UV laser beam, where
the laser intensity is the highest, is not directly incident on the
underlying passivation layer and the silicon wafer substrate;
therefore, both are well protected by a link material "shield" from
damage by the UV laser energy. This "first stage" of the laser
pulse energy profile for processing link structures, ruptures the
overlying passivation layer and removes a part of the link
material. Alternatively, laser energy at the green spectrum can be
chosen at the beginning of the laser pulse energy profile because
of its better energy coupling efficiency to the conductive link
material. A short rise time of the laser pulse energy profile at
the beginning of the link process is advantageous in that it will
rupture the overlying passivation sooner, leaving less time for the
underlying passivation to crack before rupturing.
[0031] The "second stage" of the laser pulse energy profile to
process link structures uses the longer wavelength of visible green
or blue laser energy to remove all of the remaining link material.
Since the visible laser energy needs only to finish the second
stage of the link process, i.e., remove the link material remaining
after the UV laser pulse process and thereby complete formation of
the open volumetric region, the amount of laser energy needed is
much less than that which otherwise would be needed for traditional
link processing with a single laser pulse of a single laser
wavelength. As a consequence, the risk of damage to the silicon
wafer substrate by the visible laser pulse is greatly reduced. On
the other hand, there is little risk of damage to the underlying
passivation layer by this laser pulse at a visible wavelength
because the underlying passivation material does not absorb it.
[0032] Both of the laser heads emitting the laser energies
preferably operate at the same repetition rate and are well
synchronized to each other. A typical laser pulse repetition rate
for link processing ranges from 1 KHz to 200 KHz or greater. For
different applications, the laser pulse repetition rate can be
lower than 1 KHz (as low as 1 Hz) or higher than 200 KHz. For link
processing, each of the two or more laser energies from which the
one laser pulse profile is formed ranges from less than 0.001 uJ to
about 20 uJ, with each of their duration times ranging from 100 fs
to a few tens of ns.
[0033] Another preferred embodiment entails the use of a laser
pulse energy profile comprised of 1064 nm or 1320 nm laser energy
and its second or third harmonic (532 nm and 660 nm, 355 nm and 440
nm, respectively) laser energy to sever an electrically conductive
link. By properly selecting the energy and timing of each laser
wavelength making up the laser pulse energy profile, the link can
be severed while preventing damage to neighboring links or the
silicon wafer substrate.
[0034] Table 1 presents absorption data for common semiconductor
device link metals at different wavelengths. TABLE-US-00001 TABLE 1
1320 nm 660 nm 532 nm Al 3% 9% 9% Cu 2.5% 5% 25% W 40% 50%
Table 1 indicates that, if the original energy value needed to
process a copper link with 1320 nm is E, then with a mixture of 660
nm and 1320 nm, the new 1320 nm energy value can be 50% E and the
660 nm energy value can be roughly 25% E. With respect to damage to
a nearby link structure, the 1320 nm energy has the larger spot
size of the two applied energies and thereby presents a greater
risk of damage. However, for a Gaussian-shaped beam, a 1320 nm
energy at 50% E incident on any part of the link structure
presents, from a damage point of view, an effective beam spot size
that is 80% of the beam spot size of a 1320 nm energy at 100% E.
With properly designed focusing optics, the effective laser beam
spot size of the 660 nm laser energy at 25% E can be equal to or
smaller than the laser beam spot size of the 1320 nm laser energy
at 50% E.
[0035] With respect to damage to a silicon wafer substrate, 660 nm
laser energy at 25% E is well below the damage threshold of the
silicon substrate. Adding another 50% E at 1320 nm leaves
sufficient head room so as to not damage the silicon wafer
substrate. This energy percentage mix can be readily adjusted for
different link structures. For instance, it can be 40%-20% E for
660 nm energy and 20%-60% E for 1320 nm energy.
[0036] Other preferred embodiments entail the use of different
wavelength mixtures. In addition to the mixture of 1320 nm and 660
nm laser energies, other mixtures can also be 1320 nm and 330 nm
(its fourth harmonic) energies or 1064 nm energy and a shorter
laser wavelength energy, such as 532 nm, 355 nm, and 266 nm, all of
which are the harmonics of 1064 nm emissions from a Nd:YAG or
Nd:YVO laser.
[0037] Mixing the fundamental wavelength with its second harmonic
is advantageous because it simplifies the focusing lens design.
Creating a dual-wavelength lens that can deliver desired beam spot
size for the two wavelengths is easier when dealing with the
fundamental wavelength and its second harmonic, rather than the
fundamental and its third or fourth harmonic.
[0038] A preferred embodiment entailing the use of a mixture of UV
and 1320 nm laser energies offers another advantage in that the UV
laser energy can help to directly open the overlying passivation
layer. This is quite desirable for severing links with very narrow
link widths. The UV laser can be a third harmonic of Nd:YAG,
Yb:YAG, Nd:YVO, Nd:YLF laser or Nd, Yb doped fiber laser at 355 nm,
351 nm, or 349 nm or other wavelengths in the UV spectrum. The
green laser can be a second harmonic of Nd:YAG, Yb:YAG, Nd:YVO,
Nd:YLF laser or Nd, Yb doped fiber laser at 532 nm, 526 nm, or 523
nm or other wavelengths in the green spectrum. The blue laser can
be a third harmonic of a Nd:YAG or Nd:YLF laser at 440 nm from 1320
nm or other laser source. A shorter laser wavelength in the visible
spectrum is preferred, such as 400 nm, because a visible wavelength
closer to the UV wavelength (355 nm) facilitates focusing of the
mixed laser energy to a smaller beam spot size.
[0039] Other preferred laser wavelength mixtures can be green and
1320 nm, green and 1064 nm, 1064 nm and 1320 nm, 1064 nm and 1047
nm, and 1320 nm and 1047 nm. The green and 1300 nm mixture could be
very useful in processing a "fat (thick) copper link," for which
the beam spot size is not the most critical issue. The green laser
pulse accelerates heating of the top part of the link, thereby
helping to rupture the overlying passivation with less risk of
generating cracks elsewhere in the passivation material. After the
green laser energy ruptures the overlying passivation layer and
removes a part of the link material, the 1320 nm laser energy
finishes the link process. Since the remaining link material has
been heated by the green laser energy, the absorption of the link
material at 1320 nm will be greatly improved, which in turn reduces
the required laser energy at the 1320 nm. Moreover, the silicon
wafer substrate has a much lower absorption coefficient at 1320 nm.
All these factors add up and result in much less risk of damage to
the silicon wafer substrate by the laser energy used. For a mixture
of UV laser wavelength and blue or green laser wavelength, the UV
wavelength can be 355 nm, 266 nm, or a shorter wavelength.
[0040] FIG. 3 shows one embodiment of a system 50 for using the
outputs of two laser heads 52 and 54 operating at different laser
wavelengths to form the specially shaped laser output pulse energy
profile. System 50 is preferably, but need not be, implemented with
injection locking to reduce laser output jitter for forming a
reliable and stable pulse energy profile. Jitter reduction is
advantageous when the output energies of the two laser heads partly
temporally overlap to form a laser pulse energy profile with
separate, different wavelength peaks. In embodiments in which the
output energies temporally overlap, the specially shaped energy
pulse profile has characteristics corresponding at least in part to
characteristics of the output energies of the two laser heads. The
sources of and techniques for reducing the jitter are described
with reference to FIGS. 6A, 6B, 7, and 8 below.
[0041] System 50 is composed of two laser heads 52 and 54 that emit
laser output. Laser head 52, which is composed of a lasant (not
shown), pumping source (not shown), and a high-speed shutter device
such as a Q-switch 56, generates a pulsed laser output beam of a
desired energy profile. Laser head 52 and optional attenuator 58
and beam expander 59 components form a laser rail 60 from which
propagates a pulsed output beam 62 of laser energy. Similarly,
laser head 54, which is composed of a lasant (not shown), pumping
source (not shown), and a high-speed shutter device such as a
Q-switch 66, generates a pulsed laser output beam of a desired
energy profile. Laser head 54 and optional attenuator 68 and beam
expander 69 components form a laser rail 70 from which propagates a
pulsed output beam 72 of laser energy. Skilled persons will
appreciate that the laser output wavelength and other laser
parameters dictate the specific design and configuration of the
components of laserheads 52 and 54. Beam 62 of laser output of
laser rail 60 is incident on a beam splitter 74, which in
cooperation with a mirror 76 directs for injection locking a small
portion 78 of the energy of laser beam 62 to laser head 54 and by
direct transmission passes the remainder of the laser energy of
laser beam 62 to a beam combiner 80. A beam 72 of laser output of
laser rail 70 produced partly in response to the injected laser
energy reflects off a mirror 84 for incidence on beam combiner 80.
Beam combiner 80 receives the series of pulsed laser output of
laser rails 60 and 70 to form a pulsed output beam 86 characterized
by a desired laser pulse energy profile for incidence on a
multilayer structure intended to undergo a micromachining process.
An optional harmonic converter 88 may be associated with a beam of
laser head output pulses before their incidence on beam combiner
80. The details of system operation, as well as the details of
other system embodiments, are fully described below.
[0042] FIGS. 4A, 4B, and 4C show one example of the synthesis of
the laser pulse energy profile formed by partly temporally
overlapping laser outputs 62 and 72. FIG. 4A shows a series of
pulse spikes 90 produced by laser rail 60. The energy profile of
each pulse spike 90 exhibits a rapid rise time and a peak energy
level 92 suitable for rupturing target link material. FIG. 4B shows
a series of pulse spikes 94 produced by laser rail 70. Pulse spikes
94 are of longer duration than and are delayed relative to pulse
spikes 90 such that the series of pulse spikes 90 and 94 partly
temporally overlap. The energy profile of each pulse spike 94
exhibits. a relatively gradual rise time and a peak energy level 96
suitable for removing target material in the opening formed by
rupture of target link material caused by pulse spikes 90. The rise
time of pulse spikes 94 is longer than that of pulse spikes 90, and
the energy of pulse spikes 94 is less than that of pulse spikes 90.
The durations of pulse spikes 90 and 94 each range from about 1 ps
to about 100 ns. FIG. 4C shows two of a series of laser pulse
profiles 98 produced at the output of beam combiner 80. Beam
combiner 80 can be based on polarization or on simple partial
transmission and reflection such as 50%-50% or 40%-60%, based on
the requirement of the energy of pulse spike 90 and the energy of
pulse spike 94. The relative positions of peaks 92 and 96 of the
respective pulse spikes 90 and 94 depend on the time displacement
between them. Such time displacement can be realized by, for
example, specifying the proper different Q-switch firing times for
laser heads 52 and 54, together with the proper length of an
optical fiber along path "A" shown in FIG. 3. FIG. 4C shows an
overlap of pulse spikes 90 and 94 that form a series of single
pulses 98, each with two peaks 100 and 102 characterized by the
times of occurrence and peak energy levels of the energy profiles
of the respective pulse spikes 90 and 94. The delay time between
peaks 100 and 102 is between about zero and about 500 ns. The
leading edge rise time of peak 100 is shorter than about 10 ns, and
the total duration of laser pulse profile 98 is longer than about 5
ns.
[0043] FIGS. 5A, 5B, and 5C show one example of using series of
separate laser pulse energy profiles to form a beam of temporally
nonoverlapping pulsed laser outputs 62 and 72. The series of pulse
spikes 90 of FIGS. 4A and 5A are the same, and the series of pulse
spikes 94 of FIGS. 4B and 5B are the same. The time displacement of
corresponding pulse spikes 90 and 94 is, however, sufficiently
large that they do not overlap. FIG. 5C shows a combined series 104
of alternating, nonoverlapping pulse spikes 90 and pulse spikes 94,
the respective peak energy levels 92 and 96 of which are portions
of separate, nonoverlapping pulses.
[0044] Semiconductor link processing using a laser pulse energy
profile comprised of two different wavelengths at different
processing stages can be accomplished at a higher laser power or a
higher laser pulse repetition rate, or extended to more than two
different laser wavelengths, by employing multiple laser heads and
using polarization sensitive components or other components as beam
combiner 80.
[0045] Although there exist prior art techniques for combining
laser pulses propagating from multiple laser heads, there is no
discussion of the issue of laser pulse jitter during the pulse
combination. Laser pulse jitter is the random fluctuation of the
laser pulse timing relative to the laser pulse control signal. For
a typical diode-pumped solid state (DPSS) laser used in laser link
processing, laser pulse jitter is in the range of 5 ns-30 ns. This
means that one cannot practicably stabilize the shape of laser
pulse energy profile when two corresponding pulses propagating from
two laser heads are time displaced by an amount similar to the
range of the pulse jitter.
[0046] Solving the problem of laser pulse jitter enables
realization of a laser pulse with energy profile comprised of two
or more different laser wavelengths with high accuracy and
stability of the profile. Laser energy propagating from two laser
heads and separated by a time interval ranging from zero to a few
hundred nanoseconds can, therefore, be utilized to generate a
composite laser pulse with a stable laser pulse energy profile
shape useful for a variety of applications. This aspect of the
invention substantially eliminates laser output jitter between
multiple laser heads and thereby enables use of a higher laser
power, higher laser pulse repetition rate, or specially configured
laser pulse shape.
[0047] Laser pulse jitter comes primarily from two sources, the
laser driving electronics and the laser itself. A review of
traditional laser driving electronics sets up the causes of the
laser pulse jitter problem solved by the particular inventions
described below. FIG. 6A shows a prior art system 110 configured
for laser pulse combination. System 110 includes DPSS laser rails
112 and 114 configured with acousto-optic (A-O) Q-switches. An
electronic controller/delay controller 116 provides at outputs 118
acousto-optic Q-switch RF signals and laser pulse demand control
signals to laser rails 112 and 114, which in response emit
respective pulsed laser output beams 130 and 132. Output beam 130
propagates directly for incidence on a beam combiner 134, and
output beam 132 propagates by reflection off a mirror 136 for
incidence on beam combiner 134. Beam combiner 134 receives and
combines pulsed laser output beams 130 and 132 to form a coaxial
beam 138 of laser pulses.
[0048] For an acousto-optic Q-switched solid state laser to realize
a higher laser pulse-to-pulse stability, the acousto-optic Q-switch
RF signal is cut off only when it crosses a preselected trigger
point, which in preferred embodiments is a zero voltage level, to
fire a laser pulse. For example, if the Q-switch RF signal
frequency is 48 MHz, the time difference between two consecutive
Q-switch RF signal zero crossing points is about 10 ns. Since the
laser pulse timing demand control signal is random and is
asynchronous to the Q-switch RF signal, the actual Q-switch RF
signal cut-off occurring in response to the laser pulse timing
demand control signal has a random timing uncertainty of 10 ns.
When two similar laser rails are used for pulse combination, the
pulse jitter between the two laser pulses propagating from the
laser rails will be 20 ns.
[0049] FIG. 6B shows the laser pulse jitter present in the laser
pulse combination coaxial beam 138 formed in accordance with the
prior art. Specifically, FIG. 6B shows an oscilloscope trace 150
representing coaxial beam 138 formed by combining many occurrences
of a 12-ns wide laser pulse 152 of output beam 130 and a 23-ns wide
laser pulse 154 of output beam 132. The average time delay between
adjacent 12 ns-wide and 23 ns-wide laser pulses is 60 ns, and the
combined laser pulse jitter between them is about 50 ns. FIG. 7 is
an oscilloscope trace 156 showing the desired laser energy profile
shape of the laser pulse that is the combination of laser output
152 with a pulse width of 12 ns and laser output 154 with a pulse
width of 23 ns, with a time delay of 10 ns between them. Laser
outputs 152 and 154 of FIG. 7 correspond to the respective laser
pulses 152 and 154 described with reference to FIG. 6B, except for
the difference in average delay time between them. It is apparent
that laser pulse jitter makes the prior art laser pulse combination
technique impractical.
[0050] To reduce laser pulse jitter stemming from the laser driving
electronics, an embodiment of a laser system 160, which is shown in
FIG. 8, implements a design in which the acousto-optic Q-switches
associated with the multiple lasers are driven by synchronized
drive signals developed by a common RF signal driver. System 160
includes DPSS laser heads 162 and 164 configured with respective
acousto-optic Q-switches 166 and 168. A laser driver subsystem 170
composed of a laser control signal driver 172 and an RF signal
driver 174 controls the operations of laser heads 162 and 164.
Laser control signal driver 172 provides laser pulse timing demand
control signals 176, and RF signal driver 174 provides in response
to them synchronized RF signals to acousto-optic Q-switches 166 and
168. An ultra-violet light wavelength converter 180 associated with
laser head 162 provides a pulsed UV laser output beam 182, and a
green light wavelength converter 184 associated with laser head 164
provides a pulsed green laser output beam 186. UV laser output beam
182 propagates directly for incidence on a beam combiner 188, and
green laser output beam 186 propagates by reflection off a mirror
190 for incidence on beam combiner 188. Beam combiner 188, which is
highly transmissive of UV light and highly reflective of green
light, receives and combines UV and green laser output beams 182
and 186 to form a beam 192 of laser pulses. Different lengths of RF
coaxial cables 194 and 196 between RF signal driver 174 and the
respective acousto-optic Q-switches 166 and 168 can be used to
provide the delay time between the corresponding laser pulses of
output beams 182 and 186 propagating from the different laser heads
162 and 164.
[0051] With this design, when laser pulses are requested by laser
pulse timing demand control signals 176, both laser energies are
fired when both RF drive signals applied to acousto-optic
Q-switches 166 and 168 are at a zero voltage level crossing, i.e.,
not random relative to the RF drive signal level, to maintain high
laser output amplitude stability. However, even if the Q-switch RF
signal cut-off exhibits a time jitter of the same 10 ns relative to
laser pulse timing demand control signal 176, there is no relative
pulse jitter as between the laser pulses because of the
synchronization of both RF drive signals applied to acousto-optic
Q-switches 166 and 168. Thus, a stable laser pulse energy profile
is achievable with accurate timing between the laser pulse peaks.
An operational tolerance of laser stability within about .+-.10% is
achievable.
[0052] FIG. 9A shows an implementation of RF signal driver 174
composed of an RF signal generator 200 providing a common Q-switch
RF signal to a first RF driver/amplifier 202 and a second RF
driver/amplifier 204. RF driver/amplifier 202 provides the RF drive
signal along coaxial cable 194 to acousto-optic Q-switch 166, and
RF driver/amplifier 204 provides the RF drive signal along coaxial
cable 196 to acousto-optic Q-switch 168.
[0053] FIG. 9B shows an alternative implementation of RF signal
driver 174 composed of an RF frequency generator 210 providing a
common Q-switch RF frequency signal to a first RF signal generator
212 and amplifier 214 combination and to a second RF signal
generator 216 and amplifier 218 combination. Amplifier 214 provides
the RF drive signal along coaxial cable 194 to acousto-optic
Q-switch 166, and amplifier 218 provides the RF drive signal along
coaxial cable 196 to acousto-optic Q-switch 168. In the alternative
implementation, Q-switch RF signal driver 174 uses a common
Q-switch RF frequency signal as an input to the different RF signal
generators 212 and 214 and their respective power amplifiers 214
and 218 driving the different acousto-optic devices 166 and 168.
The difference of Q-switch RF signal cut-off times for the
different power amplifiers 214 and 218 can be an integer number
times one-half of the Q-switch RF frequency cycle time. In this
case, all of the RF signals applied to different laser heads will
be cut off at the zero voltage level crossing with, however, a
delay time of an integer number times one-half of the Q-switch RF
frequency cycle time. This will give a programmable delay time
between the laser pulses in steps of a few nanoseconds, depending
on the Q-switch RF signal frequency.
[0054] Skilled persons will appreciate that when the RF trigger
points of RF signal generators 212 and 216 are continuously
programmable at the same level or different levels, a continuously
programmable delay time between the first and second laser energies
can be realized.
[0055] After the Q-switch RF signal is cut off, the laser pulse
builds up, starting from so-called quantum noise. Because of the
random nature of quantum noise, there is random time variation
ranging from a few nanoseconds to 10 ns between the time of
Q-switch RF signal cut-off and the time when the laser pulse begins
building up. To reduce laser pulse jitter stemming from the laser
pulse build-up process, an preferred embodiment utilizes a small
part of the laser output energy from one laser head (which started
its pulse building up earlier than the other) to be injected to the
other laser head such that the other laser head pulse build up will
start from the injected laser signal, and thereby eliminate the
laser build up jitter.
[0056] FIG. 3 shows one system configuration of the embodiment,
which uses two DPSS lasers coupled by injection locking. Injection
locking is performed by using a small part of the laser energy of
first laser pulse 62 propagating from laser head 52 (or laser rail
60) through an optical path for injection into laser head 54. The
firing of Q-switch 66 of laser head 54 is delayed from the firing
of Q-switch 56 of laser head 52. The optical path can include a
fiber laser to deliver the injected laser signal. The length of the
optical path can be adjusted to deliver the desired delay time
between the two laser energies from laser heads 52 and 54. With the
injected laser energy of laser head 52, the laser pulse of laser
head 54 builds up in response to the injected laser energy, rather
than from stimulus by quantum noise in the resonator. Injection
locking highly synchronizes the laser pulses of beams 62 and 72 and
thereby greatly reduces the relative pulse jitter between them.
Dashed line 220 in FIG. 8 represents the injection locking of laser
heads 162 and 164 to provide a laser system implemented with both
solutions to laser pulse jitter.
[0057] The fundamental wavelength of laser head 52 can undergo
extra-cavity harmonic conversion into the second or third harmonic
wavelength, while the injection laser energy is derived from the
fundamental wavelength, which is the same laser wavelength as that
of laser head 54. The extra-cavity harmonic conversion
implementation is accomplished with use of optional harmonic
converter 88 receiving pulsed laser beam 60 from laser head 52 (or
laser rail 60).
[0058] The timing of the firing of laser head 54 can be
electronically controlled relative to the output pulses of laser
rail 60 such that the starting point of each of the output pulses
of laser rail 70 can be at any time that is in between the starting
point and the ending point of the corresponding output pulse of
laser rail 60, as shown in FIGS. 4A and 4B, thereby to alter the
pulse shape of the formed laser pulse energy profile of FIG.
4C.
[0059] The tail of a typical laser pulse, even with a shorter pulse
full width at half maximum (FWHM), lasts a relatively long time.
For instance, for a laser pulse with a nominal 5 ns pulse width
(FWHM), the total laser pulse width (measured from its very
beginning and very end of the laser energy) can be as long as 15
ns-20 ns. This provides a substantially wide available range of
firing timing of an output pulse of laser head 54. When the optical
beam path of the injected laser signal, "A" in FIG. 3, is made
longer before injection into laser head 54, the delay between
corresponding output pulses of laser heads 52 and 54 can be
increased, so the two energy peaks comprising the formed laser
pulse energy profile can be temporally totally separated with
little jitter, as shown in FIGS. 5A, 5B, and 5C.
[0060] Solving both of these aspects of laser pulse jitter allows
generation of laser pulses with a specially shaped laser energy
profile from multiple laser heads with highly accurate timing and
profile stability. For instance, two laser pulses propagating from
different laser heads with a time difference in the range of zero
to a few hundred nanoseconds can be utilized to generate a laser
pulse with an accurate and stable pulse profile shape or "laser
energy distribution vs. time." The two laser heads can operate at
different laser pulse parameters such as different pulse width,
energy per pulse, beam divergence, and different laser wavelengths.
The broad flexibility in changing the laser pulse profile shape,
distribution of "laser energy vs. time," different divergences, and
wavelengths of the resulting laser pulse profile is a very useful
tool for variety of applications.
[0061] The generation of an accurate and stable laser pulse energy
profile consisting of different laser wavelengths is suitable for
use in semiconductor memory chip link processing. Increases in
quality of link structure processing can be obtained, for instance,
using a laser energy of a shorter duration time from a first laser
head and a laser energy of a longer duration time from a second
laser head to form a laser pulse profile with a fast rising edge
and a long pulse width, or a laser profile shape with a spike
located somewhere along the profile. With reference again to FIGS.
4A-4C, when the two laser heads operate at different laser
wavelengths, according to this invention, one can achieve a laser
pulse profile characterized by, for example, laser energy at UV
wavelength in a first desired time period of the profile, and laser
energy at green or other wavelength in a second desired other time
period of the profile. For example, the front part or peak 100 of
laser pulse profile 98 is at the UV wavelength, and the back part
or peak 102 of laser pulse profile 98 is at the green wavelength.
This is very beneficial for the link process for the reasons
described above. The preferred wavelength combinations, such as
1064 nm and 1320 nm, UV of 355 nm and near IR of 1064 nm or 1320
nm, for different link structures or different applications, are
achievable.
[0062] The application system can control the laser pulse profile,
its energy components and wavelength components further to
facilitate other functions of the system, such as beam-to-work
target alignment. For example, for beam-to-work target alignment,
the system can enable only the green energy part of the laser pulse
profile, or only the UV laser energy part while not enabling the
other part of the energy profile to improve contrast and increase
signal-to-noise ratio of the reflection from a target feature,
thereby increasing alignment accuracy. For link processing, the
system makes full use of the laser pulse energy profiling
capability.
[0063] It will be obvious to those having skill in the art that
many changes may be made to the details of the above-described
embodiments without departing from the underlying principles of the
inventions. The scope of the present inventions should, therefore,
be determined only by the following claims.
* * * * *