U.S. patent application number 11/004773 was filed with the patent office on 2005-07-14 for laser-based method and system for memory link processing with picosecond lasers.
This patent application is currently assigned to GSI Lumonics Corporation. Invention is credited to Cordingley, James J., Ehrmann, Jonathan S., Gu, Bo, Johnson, Shepard D., Lee, Joohan, Smart, Donald V., Svetkoff, Donald J..
Application Number | 20050150880 11/004773 |
Document ID | / |
Family ID | 34465448 |
Filed Date | 2005-07-14 |
United States Patent
Application |
20050150880 |
Kind Code |
A1 |
Gu, Bo ; et al. |
July 14, 2005 |
Laser-based method and system for memory link processing with
picosecond lasers
Abstract
A laser-based method of removing a target link structure of a
circuit fabricated on a substrate includes generating a pulsed
laser output at a pre-determined wavelength less than an absorption
edge of the substrate. The laser output includes at least one pulse
having a pulse duration in the range of about 10 picoseconds to
less than 1 nanosecond, the pulse duration being within a thermal
laser processing range. The method also includes delivering and
focusing the laser output onto the target link structure. The
focused laser output has sufficient power density at a location
within the target structure to reduce the reflectivity of the
target structure and efficiently couple the focused laser output
into the target structure to remove the link without damaging the
substrate.
Inventors: |
Gu, Bo; (North Andover,
MA) ; Smart, Donald V.; (Boston, MA) ;
Cordingley, James J.; (Littleton, MA) ; Lee,
Joohan; (Andover, MA) ; Svetkoff, Donald J.;
(Ann Arbor, MI) ; Johnson, Shepard D.; (Andover,
MA) ; Ehrmann, Jonathan S.; (Sudbury, MA) |
Correspondence
Address: |
BROOKS KUSHMAN P.C.
1000 TOWN CENTER
TWENTY-SECOND FLOOR
SOUTHFIELD
MI
48075
US
|
Assignee: |
GSI Lumonics Corporation
Billerica
MA
|
Family ID: |
34465448 |
Appl. No.: |
11/004773 |
Filed: |
December 3, 2004 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
11004773 |
Dec 3, 2004 |
|
|
|
10683086 |
Oct 10, 2003 |
|
|
|
10683086 |
Oct 10, 2003 |
|
|
|
09941389 |
Aug 28, 2001 |
|
|
|
6727458 |
|
|
|
|
09941389 |
Aug 28, 2001 |
|
|
|
09473926 |
Dec 28, 1999 |
|
|
|
6281471 |
|
|
|
|
11004773 |
Dec 3, 2004 |
|
|
|
10107890 |
Mar 27, 2002 |
|
|
|
60279644 |
Mar 29, 2001 |
|
|
|
Current U.S.
Class: |
219/121.69 ;
219/121.68 |
Current CPC
Class: |
B23K 26/0624 20151001;
B23K 2103/50 20180801; B23K 26/0736 20130101; H01L 23/5258
20130101; B23K 26/40 20130101; B23K 2103/10 20180801; H01L
2924/0002 20130101; H01L 22/12 20130101; B23K 26/389 20151001; B23K
2103/12 20180801; B23K 26/361 20151001; B23K 2103/08 20180801; H01L
21/76894 20130101; H01L 2924/00 20130101; B23K 2101/38 20180801;
H01L 2924/0002 20130101; B23K 26/04 20130101; H05K 3/0026 20130101;
B23K 2101/40 20180801; H01L 21/485 20130101 |
Class at
Publication: |
219/121.69 ;
219/121.68 |
International
Class: |
B23K 026/38; B23K
026/40 |
Claims
1-61. (canceled)
62. A method for removing target material associated with
electrically conductive links, said target material including a
passivation material overlying said links, said links being
positioned between respective pairs of electrically conductive
contacts in a circuit fabricated on a substrate, said links
defining a link width, wherein the link may have a passivation
layer under the link, said method comprising: providing to a beam
positioning system beam positioning data for imparting relative
movement of a laser spot position to the substrate in response to
processing control signals representing one or more locations of
electrically conductive links; generating for each selected link
structure at least two time-displaced laser output pulses, wherein
said laser output pulses are characterized by a laser spot size
when focused on the link location that is larger than the link
width, wherein said pulse energy and power characteristics are such
that damage to material underlying the link during link processing
is avoided, and insufficient to sever the link; and coordinating
the laser output pulses and the relative movement imparted by the
beam positioning system such that the relative movement is
substantially continuous while the laser output pulses strike the
overlying passivation layer and encompass the link width, the
output pulses removing the overlying passivation material
associated with the link without causing damage to underlying
passivation material or the substrate material underlying the
link.
63. The method of claim 62, wherein the step of generating is at
least partially accomplished with a Q-switched, solid state laser
and wherein the step of coordinating is at least partially
accomplished with an optical switch positioned external to the
laser.
64. The method of claim 62, further comprising amplifying the laser
output pulses.
65. The method of claim 62, wherein said link comprises metal,
polysilicide, or polysilicon.
66. The method of claim 62, wherein the underlying layer comprises
an underlying layer of SiO.sub.2.
67. The method of claim 62, wherein the links are processed at a
rate of at least 1 kHz to somewhat less than about 20 kHz.
68. The method of claim 62, wherein each of the laser pulses has a
pulse width in the range of several picoseconds to about ten
nanoseconds.
69. The method of claim 62, wherein each of the laser pulses has a
pulse width of less than 5 nanoseconds.
70. The method of claim 62, wherein each of the pulses produces a
focused laser spot having a peak power of at least 10.sup.9
W/cm.sup.2, a diameter of about 1 to 4 micrometers, and duration of
several picoseconds to about 10 nanoseconds, whereby said pulses
have an energy ranging from about 0.1 nanojoules to about 5
microjoules.
71. The method of claim 62, wherein the laser outputs pulses have
an energy of 0.1 microjoules to 3 microjoules, wherein the method
further comprises attenuating the laser output pulses by up to 90%
such that the energy of the pulses impinging the target material
ranges from 10 nanojoules to 3 microjoules.
72. The method of claim 62, wherein each of the pulses has an
energy of 0.1 microjoules up to 3 microjoules of energy.
73. The method of claim 70, wherein each of the pulses has an
energy of 0.1 microjoules up to 3 microjoules of energy.
74. The method of claim 62, wherein each of the output pulses have
approximately the same energy.
75. The method of claim 62, wherein at least two of said output
pulses have different pulse widths.
76. The method of claim 75, wherein at least two of the output
pulses have different energies.
77. The method of claim 62, wherein the pulses are generated at a
wavelength of less than about 2000 nm.
78. The method of claim 62, wherein the beam positioning system
causes relative movement at substantially constant speed during
processing.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application is a continuation in part of U.S. Ser. No.
09/941,389 entitled "Energy-Efficient, Laser Based Method and
System for Processing Target Material", filed 28 Aug. 2001, which
is a continuation of U.S. Ser. No. 09/473,926, filed 28 Dec. 1999,
now U.S. Pat. No. 6,281,471. The disclosure of U.S. Pat. No.
6,281,471 is hereby incorporated by reference in its entirety. This
application is also a continuation in part of U.S. Ser. No.
10/107,890 entitled "Methods and Systems for Thermal-Based Laser
Processing a Multi-Material Device" filed 27 Mar. 2002, which
claims the benefit of U.S. Provisional Application Ser. No.
60/279,644, filed 29 Mar. 2001. The disclosure of U.S. Ser. No.
10/107,890, now published as U.S. patent application Publication
No. 2002/0167581, is hereby incorporated by reference in its
entirety.
BACKGROUND OF THE INVENTION
[0002] 1. Field of the Invention
[0003] The present invention relates to the field of laser
processing methods and systems, and specifically, to laser
processing methods and systems for processing conductive link
structures formed on substrates. This invention is particularly
applicable, but not limited to, laser repair of redundant
semiconductor memory devices.
[0004] 2. Background Art
[0005] Economics and device performance have driven the size for
the DRAMs and logic devices to very small physical dimensions. Not
only are the devices small, but the interconnects and links
thickness have also decreased dramatically in recent years.
[0006] Some thermal laser processing of links, for example, as
described in "Link Cutting/Making" in HANDBOOK OF LASER MATERIALS
PROCESSING, Chapter 19, pp. 595-615, Laser Institute of America
(2001), relies on the differential thermal expansion between the
oxide above the link and the link itself. The differential
expansion results in a high pressure build-up of the molten link
contained by the oxide. The oxide over the link is necessary to
contain the link in a molten state long enough to build-up
sufficient pressure to crack the oxide and explosively expel the
link material. If the pressure is too low, the link will not be
removed cleanly. Alternative laser wavelengths and laser control
strive to increase the laser "energy window" without damaging the
substrate and material contiguous to the link.
[0007] Further information is available regarding link blowing
methods and systems, including material processing, system design,
and device design considerations, in the following representative
U.S. patents and published U.S. applications: U.S. Pat. Nos.
4,399,345; 4,532,402; 4,826,785; 4,935,801; 5,059,764; 5,208,437;
5,265,114; 5,473,624; 6,057,180; 6,172,325; 6,191,486; 6,239,406;
2002-0003130; and 2002-0005396.
[0008] Other representative publications providing background on
link processing of memory circuits or similar laser processing
applications include: "Laser Adjustment of Linear Monolithic
Circuits," Litwin and Smart, ICAELO, (1983); "Computer Simulation
of Target Link Explosion In Laser Programmable Memory, " Scarfone,
Chlipala (1986); "Precision Laser Micromachining, " Boogard, SPIE
Vol. 611 (1986); "Laser Processing for Application Specific
Integrated Circuits (asics)," SPIE Vol. 774, Smart (1987); "Xenon
Laser Repairs Liquid Crystal Displays," Waters, Laser and
Optronics, (1988); "Laser Beam Processing and Wafer Scale
Integration," Cohen (1988); "Optimization of Memory Redundancy Link
Processing," Sun, Harris, Swenson, Hutchens, Vol. SPIE 2636,
(1995); "Analysis of Laser Metal Cut Energy Process Window,"
Bernstein, Lee, Yang, Dahmas, IEEE Trans. On Semicond. Manufact.,
Vol 13, No. 2. (2000); "Link Cutting/Making" in HANDBOOK OF LASER
MATERIALS PROCESSING, Chapter 19, pp. 595-615, Laser Institute of
America (2001).
[0009] Requirements for the next generation of dynamic random
access memory (DRAM) include fine pitch links having link widths
less than 0.5 microns and link pitch (center to center spacing)
less than 2 microns (e.g., 1.33 microns). Current commercial laser
memory link repair systems, which use Q-switched, Nd based
solid-state lasers with wavelengths of about 1 to 1.3 microns and
pulse widths about 4 to 50 nanoseconds (ns), are not well suited
for meeting such requirements. The large (wavelength limited) spot
size and thermal effect (pulse width limited) are two limiting
factors.
[0010] In INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING
TECHNOLOGY (2001) 18:323-331, results of copper laser processing
are disclosed. A frequency tripled Nd:YAG laser with 50 nanosecond
(ns) pulse duration was used. The measured heat affected zones
(HAZ) were about 1 micron for 6.times.10.sup.8 W/cm.sup.2
irradiance and more than 3 microns for about 2.5.times.10.sup.9
W/cm.sup.2 irradiance.
[0011] Attempts have been made to address the problems. Reference
is made to the following U.S. patents and published applications:
U.S. Pat. Nos. 5,208,437; 5,656,186; 5,998,759; 6,057,180,
6,300,590; 6,574,250; WO 03/052890; and European patent EP 0902474.
In summary, the conventional q-switched, nanosecond solid state
lasers, even at short wavelengths, are not able to process the fine
pitch links due to its thermal process nature. Material interaction
may be a substantially non-thermal process at femtosecond pulse
widths, but the complexity, high costs, and reliability of
femtosecond pulse lasers may limit practical implementations.
Device and material modifications to support laser repair are
expensive and alone may not be sufficient. An improved method and
system for fine pitch link processing is needed to circumvent
problems associated with thermal effects yet provide for efficient
link removal at high repetition rates without the complexity
associated with femtosecond laser systems.
SUMMARY OF THE INVENTION
[0012] An object of the present invention is to provide a method or
apparatus for improving the quality of laser processing (i.e.,
removal, ablation, severing, "blowing," etc.) of memory links.
[0013] Another object of the present invention is to provide a
method or apparatus for laser processing of fine pitch memory
links.
[0014] In carrying out the above objects of the present invention,
a laser-based method of removing a target link structure of a
circuit fabricated on a substrate is provided. The method generally
includes generating a pulsed laser output at a pre-determined
wavelength less than an absorption edge of the substrate. The laser
output includes at least one pulse having a pulse duration in the
range of about 10 picoseconds to less than 1 nanosecond, the pulse
duration being within a thermal laser processing range. The method
also generally includes delivering and focusing the laser output
onto the target link structure. The focused laser output may have
sufficient power density at a location within the target structure
to reduce the reflectivity of the target structure and efficiently
couple the focused laser output into the target structure to remove
the link without damaging the substrate.
[0015] The pulse duration may be longer than a characteristic pulse
duration at which the relationship of fluence breakdown threshold
versus laser pulse duration exhibits a rapid and distinct change in
slope at the characteristic pulse duration.
[0016] The pulse duration of each pulse generally corresponds to a
duration wherein a fluence threshold for link material removal is
substantially proportional to the square root of the pulse
duration, and whereby the link material is removed in a thermal
manner.
[0017] The power density may be in the range of about 10.sup.9
W/cm.sup.2 to about 10.sup.13 W/cm.sup.2.
[0018] The power density may be in the range of about 10.sup.10
W/cm.sup.2 to about 10.sup.12 W/cm.sup.2.
[0019] The link material comprises at least one of aluminum,
copper, or gold.
[0020] The pulse duration of the at least one pulse may be in the
range of about 10 picoseconds to about 50 picoseconds.
[0021] The pulse duration of the at least one pulse may be in the
range of about 10 picoseconds to about 100 picoseconds.
[0022] Generating the laser output generally includes producing at
least one seed laser pulse having a first wavelength, amplifying
the seed pulse to produce an amplified pulse, and shifting the
amplified pulse from the first wavelength to the pre-determined
output wavelength.
[0023] At least one inorganic dielectric layer may separate the
link structure from the substrate and the pre-determined wavelength
may be greater than an absorption edge of the layer.
[0024] The pre-determined wavelength may be in the range of about
0.18 microns to about 0.55 microns.
[0025] The pre-determined wavelength may be in the visible or near
UV range.
[0026] The pre-determined wavelength may be less than about 1
micron.
[0027] The pre-determined wavelength may be in the range of about
750 nm to about 850 nm.
[0028] The focused laser output may be positioned to the link
within a predetermined tolerance and wherein the focused laser
output may produce a heat affected zone (HAZ) substantially less
than the tolerance.
[0029] The focused laser output may produce a heat affected zone
having a dimension in a range of about 0.1 micron to about 0.85
micron.
[0030] The focused laser output has a dimension that is less than
about 1.5 micron.
[0031] The focused laser output may have a dimension that is less
than about 1.0 micron.
[0032] At least one layer may separate the link structure from the
substrate and an optical property of the at least one layer
includes non-linear absorption of the focused laser output at about
the power density such that the absorption of the layer increases
during the at least one pulse duration whereby energy delivered to
the substrate is attenuated.
[0033] The layer comprises a polymeric low-k dielectric
material.
[0034] The removal of the target link structure may be assisted by
heat removal from a link processing region by material ejection at
the pulse width and power density.
[0035] At least one layer may separate the link from the substrate
and wherein the reflectivity and thickness of the at least one
layer cooperate to prevent the laser output from damaging the
substrate underlying the link.
[0036] The link may be removed with one laser pulse.
[0037] The focused laser output generally causes heat to be removed
from a region of the target link structure using material
ejection.
[0038] The output energy of the at least one pulse may be in the
range of about 0.001-3 microjoules over the focused laser
output.
[0039] The laser output may include a plurality of pulses.
[0040] Each of the plurality of pulses may have a temporal spacing
and the method further comprises controlling the temporal spacing
by gain switching the laser.
[0041] Each of the plurality of pulses may have output energy less
than required for removal of the link with a single pulse.
[0042] The output energy of each of the plurality of pulses may be
at least 0.001 microjoules.
[0043] The temporal spacing between each of the plurality of pulses
may be at least five nanoseconds.
[0044] The temporal spacing between each of the pulses may be based
on a predetermined physical property.
[0045] The property may be a differential thermal property.
[0046] The temporal spacing between each of the pulses may be based
on a time interval for dissipation of vapor/plasma/plume.
[0047] Delivering and focusing may include producing at least one
non-round spot to improve energy enclosure of the focused laser
output within the link.
[0048] Further in carrying out the objects of the present
invention, a laser-based method of removing a target link structure
of a circuit fabricated on a substrate is provided. The method
comprises generating a laser pulse train, each pulse of the pulse
train having a pulse duration in the range of about 10 picoseconds
to less than 1 nanosecond. The pulse duration may be within a
thermal processing range. The laser pulse may be generated using a
laser sub-system. The method further comprises controllably
selecting at least a portion of the pulse train so as to provide at
least one output pulse to process the target link structure on
demand. The portion may be selected using a modulator sub-system.
The method yet further comprises delivering and focusing the at
least one output pulse onto the target link structure. The at least
one focused output pulse may have sufficient power density at a
location within target structure to reduce the reflectivity of the
target structure and efficiently couple the focused laser output
into the target structure to remove the link without damaging the
substrate.
[0049] The controllably selecting may be based on at least one of
position and velocity information to synchronize a link and laser
beam position during relative motion.
[0050] The link may be removed as a result of the efficient
coupling of the laser output with the link and thermal interaction
with the substrate within a limited heat affected zone.
[0051] The laser pulse train may be an amplified pulse train, and
the step of generating is carried out with by a master oscillator
and power amplifier (MOPA).
[0052] A plurality of pulses may be delivered to the link structure
during relative motion and the method may further include
deflecting at least one laser pulse based on a motion signal to
compensate for the motion.
[0053] Further in carrying out the objects of the present
invention, a laser-based method of removing a target link structure
of a circuit fabricated on a substrate is provided. The method
comprises generating a sequence of laser pulses utilizing a seed
laser having a first pre-determined wavelength. The method further
includes optically amplifying at least a portion of the sequence of
pulses to obtain an amplified sequence of output pulses. The method
yet further includes delivering and focusing at least one pulse of
the amplified sequence of pulses onto the target link structure.
The at least one focused output pulse may have a pulse duration in
the range of about 10 picoseconds to less than 1 nanosecond. The
pulse duration may be within a thermal processing range. The at
least one focused output pulse may have sufficient power density at
a location within the target structure to reduce the reflectivity
of the target structure and efficiently couple the focused output
pulse into the target structure to remove the link without damaging
the substrate.
[0054] The step of generating may include a step of pre-amplifying
the seed laser to a pulse energy level prior to optically
amplifying.
[0055] The method may further comprise shifting the first
wavelength to a second wavelength prior to the step of optically
amplifying.
[0056] The method may further comprise controllably selecting,
subsequent to the step of optically amplifying, at least a portion
of the amplified sequence of pulses based on position or velocity
information to synchronize a link and laser beam position during
relative motion to provide the at least one output pulse on
demand.
[0057] The method may further comprising controllably selecting,
prior to the step of amplifying, at least a portion of the sequence
of pulses based on position or velocity information to synchronize
a link and laser beam position during relative motion to provide at
least one pulse to process the target link on demand.
[0058] The step of generating may include gain switching the seed
laser to provide a pulse on demand.
[0059] The sequence of pulses may have a repetition rate that is
greater than about 1 Mhz and controllably selecting may reduce the
repetition rate to within the range of about 10 Khz to 100 Khz.
[0060] The sequence of laser pulses may include at least one pulse
having a pulse width greater than about 1 nanosecond, and the
method may further comprise compressing or slicing the at least one
nanosecond pulse to produce a pulse having the duration less than
about 100 ps.
[0061] The seed laser is a q-switched microlaser or laser diode may
have a pulse width of about one nanosecond.
[0062] The compressing or slicing may be performed prior to
amplifying.
[0063] The seed laser may be diode pumped solid state laser.
[0064] The diode pumped solid-state laser may be a fiber laser.
[0065] The seed laser may be an active or passive mode locked
laser.
[0066] The seed laser may be a high speed semiconductor laser
diode.
[0067] The step of amplifying may be performed by at least one
fiber optic amplifier.
[0068] The fiber optic amplifier may have a gain of about 30
dB.
[0069] The method may further comprise shifting the laser
wavelength of at least one pulse of the amplified pulse train from
the first wavelength to a second wavelength less than about one
micron.
[0070] Further in carrying out the objects of the present
invention, a thermal-based laser processing method of removing a
target link structure of a circuit fabricated on a substrate is
provided. The method comprises applying a focused laser output to
the link structure to remove the link without damaging the
substrate. The output may include at least one pulse having a pulse
duration in the range of about 10 picoseconds to less than about 1
nanosecond, a predetermined wavelength less than an absorption edge
of the substrate, and the at least one pulse having a power density
in a range of about 10.sup.9 W/cm.sup.2 to about 10.sup.13
W/cm.sup.2.
[0071] Yet further in carrying out the objects of the present
invention, a laser-based system for removing a target link
structure of a circuit fabricated on a substrate is provided. The
system comprises means for generating a laser pulse train where
each pulse of the pulse train may have a pulse duration in the
range of about 10 picoseconds to less than 1 nanosecond, the pulse
duration being within a thermal processing range. The system
further comprises means for controllably selecting at least a
portion of the pulse train to provide at least one output pulse to
process the target link structure on demand. The system yet further
comprises means for delivering and focusing the at least one output
pulse onto target link material comprising an optical system. The
at least one focused output pulse may have sufficient power density
at a location within the target material to reduce the reflectivity
of the target material and efficiently couple the focused output
into the target material to remove the link without damaging the
substrate.
[0072] The laser pulse train may be an amplified pulse train, and
wherein the means for generating may include a master oscillator
and power amplifier (MOPA).
[0073] The means for controllably selecting may include a modulator
means that comprises an acousto-optic modulator or electro-optic
modulator.
[0074] The electro-optic modulator may be a Mach-Zehnder
modulator.
[0075] The above objects and other objects, features, and
advantages of the present invention are readily apparent from the
following detailed description of the best mode for carrying out
the invention when taken in connection with the accompanying
drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0076] These and other features, aspects, and advantages of the
present invention will become better understood with regard to the
following description, appended claims, and accompanying drawings
where:
[0077] FIG. 1a is a block diagram showing a portion of a laser
processing system for link removal using at least one pulse in at
least one embodiment of the present invention;
[0078] FIG. 1b shows a block diagram of a portion of the external
modulator sub-system of FIG. 1a wherein a portion of an amplified
pulse train is controllably selected for "on-the-fly" processing of
links;
[0079] FIG. 1c is a top schematic view (not to scale), of a target
link in a row of links showing, by way of example, focused laser
output on a target link structure during motion of the link
relative to a laser beam;
[0080] FIGS. 2(a-b) are block diagrams showing some elements of
alternative solid state laser sub-systems, each having a master
oscillator and power amplifier (MOPA), which may be included in at
least one embodiment of the present invention;
[0081] FIG. 3 is a schematic diagram showing one arrangement for
combining laser pulses or generating a sequence of closely spaced
pulses using multiple lasers with delayed triggering;
[0082] FIG. 4 is a plot showing an example simulation results of
exploiting differential thermal properties of a link and the
underlying substrate to remove the link without damaging the
substrate by applying two pulses having a pre-determined delay;
[0083] FIG. 5a is a graph illustrating, by way of example, a
relationship between a heat affected zone (HAZ), spot size, and a
link pitch;
[0084] FIG. 5b illustrates, by way of example, material removal
with nanosecond pulses;
[0085] FIG. 5c is a graph illustrating, by way of example,
dependence of fluence threshold on laser pulse width and shows
exemplary pulse width ranges and exemplary pulse parameters
corresponding to embodiments of the present invention;
[0086] FIG. 5d is a graph illustrating, by way of example, the
dependence of the absorption coefficient of Silicon on wavelength
and shows exemplary laser wavelengths corresponding to embodiments
of the present invention;
[0087] FIG. 6a is a block diagram showing elements of a laser
sub-system wherein a seed laser of FIG. 2a or 2b is a diode pumped,
solid state laser oscillator and a diode pumped, solid state laser
amplifier is used to amplify the output of the seed laser;
[0088] FIG. 6b is a block diagram showing elements of a laser
sub-system wherein a seed laser of FIG. 2a or 2b may be a
picosecond laser diode or microchip laser for producing picosecond
pulses, for example;
[0089] FIGS. 7(a-c) are block diagrams showing additional design
alternatives which may be used in an embodiment of the present
invention, including configurations for at least one of
amplification, wavelength shifting, and "down counting"/"pulse
picking";
[0090] FIGS. 8(a-c) are schematic diagrams showing details of
exemplary master oscillator power amplifier (MOPA) configurations
which may be used in at least one embodiment of the present
invention, wherein a seed laser is amplified with at least one
fiber optic amplifier to produce picosecond pulses and including at
least one modulator for selecting pulses; and
[0091] FIG. 9 is a block diagram of a laser based memory repair
system, including a picosecond laser system, and further showing an
example implementation of the present invention.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENT(S)
[0092] Overview--Laser System Architecture
[0093] Referring to FIG. 1a, a block diagram illustrating a portion
of a laser processing system 100 for removal of an electrically
conductive link 107 using at least one output pulse 104 having a
picosecond pulse width (i.e., pulse duration, etc.) 1041 (e.g., as
measured at the half power point) and shows some major system
components included in at least one embodiment of the present
invention is shown. At least one embodiment of the invention may
include a diode pumped, solid state laser in sub-system 101 to
produce intermediate pulses 103 having pulse widths 1041 in a
preferred picosecond range. The laser may be a commercially
available diode pumped, solid state (active or passive) mode locked
laser, for instance. For operation at a preferred wavelength the
output 103 of the system 101 may be shifted in wavelength by
optional shifter 105 (e.g., a harmonic generator) for example from
a near infrared wavelength to a visible or near UV wavelength.
[0094] A single pulse, or plurality of pulses may be selected and
delivered to a link 107, and the delivered pulses may have a
pre-determined pulse width and time interval between pulses
("temporal spacing") based on a physical property of at least one
of the link 107, substrate 110, upper dielectric layer 1091, and
lower dielectric layer 1092. The beam delivery system may include
polarization control, relay optics, beam expansion, zoom optics,
and an objective lens for producing a nearly diffraction limited
spot at link 107. Optional external modulator sub-system 108 may be
operated under computer control to provide pulses on demand and
vary the power of the pulses. By way of example, pulses 102 within
the group of pulses 106 may be omitted (as depicted by the dashed
lines). U.S. Pat. Nos. 5,998,759 and 6,281,471 (e.g., col. 12, line
63-col. 14, line 33 and the associated drawings of the '471 patent)
teach the use of a modulator to provide a pulse to irradiate a link
on demand during relative motion of the link and laser beam in a
laser processing system.
[0095] Referring to FIG. 1b, a block diagram of a portion of the
external modulator sub-system 108 of FIG. 1a is shown, wherein a
portion of a pulse train 103 is controllably selected for
processing of links during relative motion between substrate 110
and the laser beam ("on the fly"). The motion may be in three
dimensions: X motion 113, Y motion (not shown) of substrate 110
which is generally mounted on a wafer stage, and Z axis motion of
at least one optical element 114 within the beam delivery system.
Reference is made to U.S. Pat. Nos. 6,114,118 and 6,483,071
assigned to the assignee of the present invention for precision
positioning methods and systems for positioning of the wafer and
the laser beam waist relative to a link position. Controller 121
generally produces control signals 122 based on position
information, velocity information, or both position and velocity
information relating a link position to a laser beam position.
Control signals 122 generally gate (i.e., control) an optical
switch 120. The optical switch 120 generally provides output pulses
106 which are a portion of the input pulse train 103. Hence, the
generated pulses 103 may have a controlled output repetition rate
and temporal spacing when the modulator (e.g., the modulator 108)
is used to select the at least one output pulse 104 that irradiates
one or more links (or other microscope structure). At least one
optical element 114 within the beam delivery system may be used to
precisely position the beam waist at high speed and to further
optimize the delivery of the focused output pulses.
[0096] Referring FIG. 1c, exemplary pulsed laser output on target
link 107 includes two focused laser pulses 1042, each having an
identical spot size, corresponding to selected pulses 104. The
distance 1043 corresponds to the temporal spacing between the
pulses during relative motion 113. If distance 1043 is a relatively
small fraction of the link width, for instance less than 25%, the
fraction of energy enclosed in the link will approximate perfect
spot positioning. Distance (or displacement) 1044 generally
represents an effective dimension of the laser output, which equals
the laser spot size for perfect placement. As the temporal pulse
spacing increases, the speed of relative motion increases, or with
finer link pitch (center to center spacing) 1043 is to receive
increasing consideration.
[0097] Published U.S. patent application 2002/0167581, assigned to
the assignee of the present invention and incorporated by reference
herein, describes various methods and sub-systems to direct laser
pulses to one or more links. The optical sub-systems or variants,
which generally include a high speed, single axis deflector, may be
incorporated within the beam delivery system of FIG. 1a as
required. Specific reference is made to FIGS. 19 and 20 of '581 and
the corresponding sections of the description for further
information of the '581 disclosure. Further, the focused output may
include a plurality of spots having at least one non-identical spot
distribution or power density. For instance, FIG. 17 of the
disclosure illustrates a focused pulse used as a "cleaning
beam".
[0098] Referring to FIG. 2a, a block diagram of additional details
of one alternative solid state laser sub-system which may be
included in an embodiment of the present invention is shown. A seed
laser (e.g., oscillator 211) produces a pulse train 214, the pulses
generally having sufficient energy suitable for amplification with
laser amplifier 212. The seed laser may be "free running" at a
predetermined rate or "gain switched" to produce pulses under
computer control. At least a portion of the pulse train is
amplified to obtain the necessary laser pulse energy to sever a
memory redundancy link, for instance to an energy level wherein the
link is severed (e.g., removed) with a single pulse. One practical
consideration for stable and reliable operation of pulsed laser
amplifiers is operation within the rated average power. The
operational considerations lead to an engineering tradeoff between
the energy of a given pulse, the number of pulses, and the
repetition rate.
[0099] In one alternative arrangement, shown in FIG. 2b (not to
scale), a portion of pulse train 214 may be controllably selected
with a suitable modulator arrangement 1081 (similar or identical to
108 of FIG. 1a) for processing of links during relative motion
between substrate 110 and the laser beam ("on the fly"), however
prior to amplification 212 of the pulse train to an energy level
for link processing. A "down counting," "divide down," or "pulse
picking" operation may be used to match a repetition rate of laser
amplifier 212, which may be orders of magnitude below the
repetition rate of the seed laser 211. For example, if R is the
repetition rate of pulse train 214, then R/n will be the repetition
rate at the output of the modulator 1081 when every n'th pulse is
selected. If 214 represents a 50 MHZ pulse train, the output of the
modulator will be 50 kHz when n=1000. In at least one embodiment,
the pulse train repetition rate may be divided by a non-integer
(e.g., 19.98) and varied over a relatively small range to
synchronize the selected pulses with the position of the link,
thereby compensating for motion system variations. Such an
operation may be performed by controller 121 in either or both 108,
1081, and may be based on position and/or velocity information.
[0100] In at least one embodiment of the present invention, a
plurality of closely pulses may be selected. By way of example,
output 103, 106 of laser amplifier 212 shows three pairs of
consecutive amplified pulses selected from pulse train 214, a given
pair which may then be selectively applied to link 107, while
providing a reduced input repetition rate and low average input
power for amplifier 212. If 214 represents a 100 MHZ pulse train,
the spacing between the consecutive output pulses of a pair will be
10 nanoseconds. Throughput and repetition rate are generally
related. Preferably the amplifier output repetition rate will be
sufficient to provide rapid link processing rates and "pulse on
demand"capability, while limiting the complexity of system position
and/or velocity control. Preferably, the three exemplary pairs at
103,106 at the amplifier output may be applied to as many as three
consecutive links during relative motion 113 of the link and laser
beam. External modulator 108 may be used to block the laser energy
from links which are not to be processed.
[0101] Likewise, dependent upon the spectral response of the
amplifier 212, optional wavelength shifter 1051 may be used to
match the wavelength of the seed laser 211 to a favorable (or
compatible) wavelength range of amplifier 212. Modulator sub-system
1081 and the wavelength shifter 1051 may be used alone or in
combination with sub-system 108 for controlling the final pulse
temporal spacing and energy level as appropriate, depending upon
specific design criteria of a particular application.
[0102] Referring to FIG. 3, yet another alternative arrangement for
combining laser pulses or generating a sequence of closely spaced
pulses using multiple lasers with delayed triggering is shown. A
pre-determined delay (e.g., t.sub.1 to t.sub.2) between trigger
pulses may determine the time interval for application of a
plurality of pulses. The combined output may provide seed pulses
for an optical amplifier. For example, two or more pulses (or
groups of pulses) may be used to sever link 107. The arrangement
may be used to provide fine control of the temporal pulse spacing
(e.g., 2-10 nanoseconds for a pulse pair, 100-500 MHZ effective
rate or "burst rate").
[0103] As disclosed in U.S. patent application Publication No.
2002/0167581 ('581), incorporated by reference herein and assigned
to the assignee of the present invention, the laser system may
include a programmable digital delay line 301 for controlling the
pulse temporal spacing t2-t1, lasers 302, a polarizing cube 303 for
beam combining, and optional amplifier 304 to raise the energy
level as required. By way of example, specific reference is made to
paragraphs 120-122, 194-197, and the claims of '581 for additional
details.
[0104] A laser wavelength within sub-system 101 will generally be
in a range of about 0.150 microns to 1.3-1.55 microns, the latter
range corresponding to diode laser wavelengths used in high speed
telecommunications. In one example, the laser wavelength may be
frequency multiplied (e.g., tripled) or Raman shifted with shifter
105 to a near IR, visible, or UV wavelength.
[0105] Laser Parameters and Link Removal
[0106] With a trend of decreasing link pitch and dimensions (i.e.,
fine pitch links), at least three parameters need to be jointly
considered for removing a link 107 without damaging either the
substrate 110 or adjacent links (not shown) which may not require
processing: (a) the laser beam size on the target and its focal
depth; (b) the beam positioning accuracy (e.g., the laser beam
waist position relative to the link in three dimensions--during
controlled X-Y motion and Z-axis motion of the at least one element
114, for example); and (c) the heat affected zone (HAZ).
[0107] Refer to FIG. 5a with link pitch 521 in the range of 3-5
microns, the theoretical minimum pitch follows the formula:
Minimum Pitch=Beam Radius+Positioning Error+0.5 Link Width (1)
[0108] where the thermal effect by the laser beam is considered
negligible.
[0109] For example, the GSI Lumonics Model M430 Memory Repair
System, manufactured by the assignee of the present invention,
provides a typical spot size of about 1.6 microns, and positioning
error of about +/-0.2 microns. The typical pulse width is about
4-10 nanoseconds and corresponds to a heat affected zone of about
0.85-1.4 microns.
[0110] The model M430 system is capable of processing links with
minimum pitches of about 2 microns (assuming a link width of about
0.5 microns).
[0111] However, as the pitch approaches the dimension that is
comparable to the thermal diffusion length, thermal effects within
the region of link 107 may have increasing significance. The
formula then becomes:
Minimum Pitch=Beam Radius+Positioning Error+0.5 Link Width+HAZ
(2)
[0112] where HAZ (Heat Affected Zone) 522 is a measure of the
thermal effect. The heat-affected zone (HAZ) is generally
determined by (D*t).sup.0.5, where D is thermal diffusion
coefficient and the laser pulse width. The actual value for the
depth to which material is molten or vaporized depends also on the
actual energy and power density on the target.
[0113] The HAZ may extend beyond the focused spot 523 and adversely
affect peripheral areas adjacent to the spot. In some cases, the
peripheral area affected may be several times greater than the spot
itself. The relative large HAZ generally makes the laser process
less controllable and less precise. In the case of link blowing,
relatively large HAZ size may also be one of the limiting factors
to the upper limit of the process window (neighboring links
damage).
[0114] A diffraction limited spot and a short laser wavelength
(e.g., 0.355 microns) may mitigate the problem to some degree,
provided the spot is properly positioned relative to the link.
However, if the positioning tolerance 524 of the system (including
the X, Y, Z motion sub-system) is +/-0.1 microns (a somewhat
stringent requirement for high speed link processing), a spot size
of about 0.58 microns may be needed to deliver the laser beam to a
0.38 micron wide link. Assuming a wavelength of 0.355 microns, and
a 10 nanosecond (ns) pulse width, the estimated HAZ is about 1.3
microns. As such, a practical limit for processing links may
correspond to about 1.9 micron pitch. Hence, a shorter pulse width
is generally desirable.
[0115] Reducing the pulse width also generally reduces the HAZ.
However, when thermal effect becomes very small compared to beam
size and position error, further reducing the thermal effect before
improving other significant contributors (e.g., beam size and
positioning) may become unnecessary. The reduction in thermal
effect from the nanosecond range to the picosecond range may be
sufficient to process the finer pitch links. Further reduce the
pulse width down to femtosecond range to eliminate undesirable
thermal effects may be avoided for processes to remove (i.e.,
sever, "blow," ablate, etc.) fine pitch links.
[0116] In accordance with the present invention, a limited thermal
interaction generally occurs within a heat affected zone that is
substantially less than cumulative tolerance of a link pitch and a
relative position of the laser output relative to the target
structure. For instance, a heat affected zone (HAZ) diameter of
about 0.3 microns to about 1 micron will generally provide for
improved processing of link pitch of 2 microns or less. Preferably,
a HAZ will be less than the positioning tolerance of the laser
output in three dimensions (e.g., less than 0.1 microns in each
direction, and generally is considered negligible).
[0117] U.S. Pat. No. 6,281,471, incorporated by reference herein,
elaborates on the rationale for the use of a short, fast rise time
pulse. Specifically, col. 4, line 45-col. 5, line 19 elaborates on
the effects decreasing reflectivity to improve coupling to target
material. If the irradiance on a metal target structure (e.g.,
aluminum) is greater than about 10.sup.9 W/cm.sup.2, the
reflectivity of the target structure is reduced and coupling of the
laser energy is improved. Thermal diffusivity (related to HAZ)
generally varies as the square root of pulse width. A short laser
pulse generally reduces or prevents heat dissipating to the
substrate below the melting link and also heat conducting laterally
to the material contiguous to the link.
[0118] As the link pitch becomes finer thermal interaction with
nanosecond pulses may be increasingly chaotic, resulting in poor
precision for link removal. As illustrated in FIG. 5b, a relatively
large volume of material may be heated and melted, and material
removal occurs through melt expulsion driven by vapor pressure and
the recoil of laser radiation pressure. At a fine scale the shape
and volume of removed material may be irregular and include a
non-acceptable large statistical variation. With picosecond high
peak power pulses the interaction may become non-linear, initially
with avalanche ionization where the reflectivity is reduced as a
result of the high free electron density in metals, with decreased
statistical variation. With such short pulses the laser energy is
generally confined to a thin layer and vaporization generally
occurs rapidly. Material removal generally becomes more precise and
deterministic, with reduced laser fluence to initiate ablation.
Material removal with picosecond pulses may further include removal
of heat from the laser processing region by material ejection
(solid and vapor). The link removal process at the picosecond
scale, for instance with the presence of an overlying dielectric
layer 1091 and inner layer 1092, may be a mixture of removal with
ablation and thermo-mechanical stress. Removal of the target link
structure is generally assisted by heat removal from a link
processing region by material ejection at the pulse width and power
density.
[0119] By way of example, FIG. 5c shows variation in fluence
threshold for two exemplary dielectric materials (e.g., see U.S.
Pat. No. 5,656,186 and the publication Du et al., "Laser-Induced
Breakdown by Impact Ionization in SiO.sub.2 with pulse widths from
7 ns to 150 fs," APPLIED PHYS., Lett., 64 (23), 6 Jun. 1994, pp.
3071-3073. As is well known, the fluence threshold is generally
much lower for metals (e.g., ten times or more) as a result of the
higher free electron density. Below the breakdown point the
threshold 501, 502 varies with material but the statistical
variation (shown by error bars) is generally relatively small. In
the illustrated example (provided with the published data in the
publication), 501 varies as 1/(pulsewidth) whereas 502 is taken as
approximately constant (as illustrated in the '186 patent). Above
the breakdown point, an approximate square root relation holds, but
increasing variation with pulse width is apparent, particularly at
the nanosecond scale.
[0120] A characteristic pulse width of the break down point of
metals, may typically be about 10 ps (e.g., see U.S. Pat. No.
5,656,186). In accordance with the present invention, the typical
laser pulse width is less than 1 nanosecond, and most preferably
much closer to the characteristic pulse width of the breakdown
point so that detrimental thermal effects are negligible (e.g., the
present invention produces reduced HAZ and statistical variation).
However, the link removal process of the present invention is
generally a thermal process. The interaction between the laser
pulse and the material is mainly a thermal (though greatly reduced)
process since the laser pulse width is longer than that of the
breakdown point, and preferably close to the breakdown point.
[0121] The present invention will generally provide an efficient
link removal process rather than a slow etching process defined by
the optical absorption depth, which is only on the order of a few
nanometers per pulse for most of metals. Since the breakdown point
is material dependent, the lower end of the pulse width is
therefore also material dependent. A minimum pulse preferred pulse
width may be in the range of a few picoseconds (ps) to about 10 ps.
A maximum pulse width is generally less than about 1 nanosecond
(ns) and will generally determined by the heat affected zone
allowable. Generally a pulse width of the present invention will in
the range from above the breakdown point to less than 1 ns. A pulse
width may be in the range 505 of about 10-100 ps, for example
40-100 ps. A most preferred pulse width is in the range 506 of
about 10 ps to about 40 ps or about 10 ps to about 50 ps.
[0122] The laser systems which produce picosecond pulses are
typically simpler, more reliable and stable, and more cost
effective as compared to femtosecond lasers. A significant
difference is implementation of pulse compression for femtosecond
generation of high peak power pulses.
[0123] Numerous references further elaborate on interaction in the
femtosecond-picosecond pulse range. For example, Chichkov et al.,
"Femtosecond, Picosecond, and Nanosecond Laser Ablation of Solids,"
APPLIED PHYSICS, A 63, 109-115, 1196 provides theoretical
background and experimental results. Femtosecond pulses were found
to have thermal conduction into a target that can be neglected to a
good approximation, and the process regarded as a direct transition
form solid to vapor, resulting in precise laser processing. The
ablation depth has a logarithmic dependence on laser pulse fluence.
With picosecond pulses, ablation is accompanied with heat
conduction and formation of a melted zone within the target, for
instance a metal. When heat conduction into the target is
neglected, (which may be a rather crude assumption), then the
logarithmic dependence of the ablation depth on fluence is also
generally possible with picosecond pulses. However, processing in
the nanosecond range has been generally regarded as much more
complicated as a result of thermal wave propagation and formation
of a large layer of melted material.
[0124] Jandeleit et al., "Picosecond Laser Ablation of Thin Copper
Films," APPLIED PHYSICS A, 63, 117-121, 1996, disclosed results of
ablation experiments wherein holes were drilled in thin copper
films on fused silica using picosecond pulses. Although high
intensity picosecond pulses having pulse widths greater than the
characteristic pulse width of the breakdown point generally follow
the square root relationship, the reduced heat affected zone and
lower heat load provide rapid heating and removal of target
material when compared to nanosecond and longer pulses. An
intensity of about 10.sup.10-10.sup.11 W/cm.sup.2 over about a 3.1
um diameter spot removed (on the average) about 0.1-0.2 .mu.m of
material per each 40 ps pulse at a wavelength of 1.053 microns.
Comparison of the results with the known optical absorption depth
of copper at 1.053 microns indicated that heat conduction generally
determines the ablation depth. The pulse-pulse variation in the
material removed may be significant (e.g., 2:1). However, the HAZ
was relatively small and collateral damage minimal.
[0125] Hence, the benefit of a pulse width from about 10-25 ps down
to below the breakdown point (typically less than 10 ps) is
generally not so significant as compared to benefits provided by
the beam spot size reduction and positioning error improvement for
the overall system capability. In addition, the cost of femtosecond
laser sources is typically much more than the cost of picosecond
laser systems, particularly fiber laser based picosecond laser
systems.
[0126] Link processing includes removal of a target structure,
typically a metal thin film. The link is typically surrounded by
materials (e.g., passivation layers 1091, 1092, substrate 110)
having dissimilar thermal and optical properties. As such, some
multi-material interaction mechanisms may be somewhat complex
compared to material processing interaction with a homogeneous
"bulk" material. At least one dimension (e.g., link width) is
typically on the order of a wavelength of visible or UV light.
Also, with finer link pitch technology that is emerging, the
fraction of the spot energy that enclosed within the link dimension
needs careful consideration by a designer of link processing
equipment. In at least one embodiments the laser wavelength is less
than one micron, for example, 0.90 microns or less, to achieve a
smaller spot size on the link in connection with the reduced pulse
width.
[0127] Since the smallest spot size is generally proportional to
the wavelength, any reduction in wavelength will be beneficial to
the reduction of the smallest spot size achievable. In addition,
the depth of focus will generally be larger for the same spot size
at such shorter wavelengths. For example, for a 1064 nm laser, the
diffraction limited spot size is approximately (i.e.,
substantially, nearly, about, essentially) 1.2 microns (diffraction
limited spot size=(constant) *wavelength*f number of the lens).
When the wavelength is reduced to 0.8 microns, the diffraction
limited spot size will be reduced by 20% accordingly as well, i.e.,
to approximately 0.9 microns. Generally, for fine pitch processing
a spot size of less than about 1.5 microns is preferred, and most
preferably 1 micron or less. In at least one embodiment of the
present invention, a non-round spot profile (e.g., an elliptical
spot produced with an anamorphic optical sub-system) may be used
(see, U.S. Patent Application No. 2002/0167581, for example). In
particular, paragraphs 133-136 illustrate how a non-round spot may
improve energy enclosure within a link in at least one
embodiment.
[0128] Material variations (e.g. variations, whether by design, by
process defect, or as a process by-product) may be encountered and
are generally expected to further affect the process energy window
as the pitch is decreased. The link may be a metal (e.g., Al, Cu,
Au, etc.), polysilicon, or a refractory metal. At least one layer
of Silicon Nitride (Si.sub.3N.sub.4) 1091 may cover the link, and a
layer of Silicon Dioxide (SiO.sub.2) 1092 may separate the
substrate 110 and link 107. However, in some cases the link may not
be covered with an outer layer. Additionally, the presence
impurities, dopants within the substrate or dielectric layers, and
next generation dielectrics (e.g., low-k polymeric materials) may
each have a substantial effect on the optical properties of the
materials. In a wavelength range wherein the wavelength is greater
than the absorption edge of the dielectrics 1091,1092 and less than
the absorption edge of the substrate 110 substrate damage may
easily occur with long laser pulses.
[0129] Link 107 may be substantially reflective at the laser
wavelength. In accordance with the present invention, the laser
output wavelength will generally be below the absorption edge of
the substrate and hence correspond to an absorbing and/or
reflecting wavelength region. The laser wavelength is typically
above the absorption edge of the dielectric layers 1091,1092 which,
in one example, may be inorganic, and will generally correspond to
a substantially maximum transmitting region, for typical inorganic
passivation layers (e.g., Si.sub.3N.sub.4, SiO.sub.2, etc.) used
with present semiconductor memories.
[0130] Referring to FIG. 5d, typical variation in the absorption
coefficient (e.g., at room temperature) of Silicon, the absorption
being very high at short wavelengths is shown. Doping (not shown)
generally changes the absorption and shifts the near IR absorption
edge to shorter wavelengths. Published European patent application
EP 0 902 474, published 17 Mar. 1999, teaches shielding the
substrate with one or more materials to avoid substrate damage.
With such modifications a shorter wavelength laser (and a reduced
spot size) provides for reduction of link pitch. The shielding
materials may be metals, refractory metals, or dielectrics. Such
modifications may also be used with the present invention to
further enhance performance.
[0131] In accordance with the present invention, a laser wavelength
may be in a range from below 0.4 .mu.m to about 1.55 .mu.m.
Exemplary wavelengths may be in the UV range (e.g., 514, 212-266
nm), near UV (e.g., 510, 355 nm), visible (e.g., 511, about 500 nm,
for instance 532 nm) and near IR spectrum (512, about 750-850 nm or
513, about 1 .mu.m). It can be seen that Silicon absorption varies
by about 1000:1 throughout the wavelength range. A preferred
wavelength may be in the range of about 0.18 microns to about 0.55
microns. The lower limit may be determined by the absorption of a
layer. With silicon substrates, both absorption and reflection
increase at shorter wavelengths. Over the wavelength range of
interest the Silicon semiconductor properties change dramatically
from near IR dielectric-like properties to metal-like properties in
the UV range. For Silicon Dioxide and Silicon Nitride, the internal
transmission and single surface reflectance are substantially
constant throughout the visible and near IR ranges. Spectral
transmission curves for typical large bandgap dielectric materials
generally show that the transmission decreases somewhat at UV
wavelengths. For example, in HANDBOOK OF LASER SCIENCE AND
TECHNOLOGY, the transmission range of Silicon Dioxide is specified
as wavelengths greater than 0.18 .mu.m. The absorption coefficient
of both Silicon Nitride and Silicon Dioxide remains relatively low
in the visible range (>400 nm) and gradually increases in the UV
range.
[0132] If the predetermined wavelength is below an absorption edge
of the substrate, the pulse energy density at the substrate may
reduced and the process window may be increased by at least one of:
(a) beam divergence (shallow depth of focus); (b) dielectric
surface reflection; (c) beam diffraction; (d) multiple scattering
(e.g., caused by dopants or impurities); (e) internal reflection
(which may vary with the focused laser beam numerical aperture);
(f) multi-layer interference; and (g) non-linear absorption within
the microstructure (if the laser spot is properly positioned in
three dimensions then at the leading edge of the high peak power
laser pulse the free electron density in a metal increases the
absorption, and link material removal may occur at a rate faster
than that of the substrate. The substrate is irradiated with
off-link energy (e.g., lower peak intensity) and has fewer free
electrons than that of the link.
[0133] In order to process links less than 0.5 micron thick, for
example, aluminum or copper links, the range of the peak energy
density (Joules/cm.sup.2) is about 0.2 J/cm.sup.2 to 300
J/cm.sup.2, with a typical value in the range 2-80 J/cm.sup.2. The
range of peak power density is about 5.times.10.sup.9 W/cm.sup.2 to
1.2.times.10.sup.13 W/cm.sup.2, with a typical value in a range of
5.times.10.sup.10-2.times.- 10.sup.12 W/cm.sup.2. For a 40 ps pulse
width laser with a spot size of 1 micron, the pulse energy range
for severing links less than 0.5 micron thick is generally in a
range of 0.001-3 micro joules with a typical value at a range of
0.02-1 micro joules.
[0134] Either a single pulse or multiple pulses may be used to
remove the link. If a single pulse is used to remove the link the
picoseconds laser system is to generally provide a range of about
1-5 micro joules per pulse at a 10 KHz-120 KHz repetition rate. An
exemplary range is less than about 1 microjoule to a maximum of 2
microjoules. Preferably single pulse processing will be implemented
with an oscillator/amplifier configuration, for instance the
seeder/amplifier configuration as shown in FIG. 2a.
[0135] In one embodiment of the present invention, multiple pulses
may be used to remove the link with a picosecond laser system
providing at least 0.001 micro joules (1 nanojoule (nj)) per pulse
at a repetition rate of at least 1 MHZ. The pulses applied to the
link may be treated as a single pulse for link removal during
relative motion in three-dimensions between the link and laser beam
(e.g., 5-50 mm/sec along X-Y axes). In another embodiment of the
present invention, about 15-20 pulses may be applied at a
repetition rate of 10-100 MHZ, each with about one-tenth the energy
required for removal with a single pulse, while traversing a
fraction of a link.
[0136] Embodiments of the present invention may also include a
plurality of closely spaced amplified pulses, for instance, two or
more pulses each with about 50% of the energy that is generally
required to remove a link with a single pulse. Pulses may be
selected with control of modulator sub-system 1081 within the laser
system 101, external modulator sub-system 108, or a combination
thereof.
[0137] In a multiple pulse process, the temporal spacing between
the pulses used to irradiate the link on demand may be selected
based on a pre-determined physical property (e.g., differential
thermal properties) of the link and surrounding materials.
Referring to FIG. 4, simulation results which, by way of example,
demonstrate an effect of exploiting differential thermal properties
of a link and the underlying substrate to remove the link without
damaging the substrate by applying two pulses having a
pre-determined delay are shown. According to the simulation results
obtained (in this case with nanosecond pulses having a square
shape), "double blast" (e.g., two pulses) with 50% energy of a
"single blast" energy was very interesting. The Silicon substrate
generally acts as a heat sink and cools down very fast compared to
the link. As shown in FIG. 5a, the results indicated the substrate
110 to stabilized to room temperature in only 10 to 20 ns. The link
107 (copper) recovery was much slower indicating a significant
differential thermal property. Based on the results, the second
pulse will generally also clear debris at cut site (i.e., link
removal) resulting in an "open circuit".
[0138] If, for example, a 60 MHZ mode locked system (e.g.,
picosecond pulses) is used, the spacing between consecutive pulses
of the output pulse train may closely match the pre-determined
spacing. If a larger temporal spacing is desired, a high speed
modulator arrangement may be used to select any sequence of pulses
or group of pulses, for example. A higher repetition rate may be
used to decrease the pulse temporal spacing, or a second laser may
be provided as shown in FIG. 3. For example, two pulses, each
having a pulse width in the range of about 40 ps to 100 ps and
spaced by 2-10 ns may be generated. By way of example, q-switched
microlasers may be used to provide pulse widths of a few
nanoseconds at a repetition rates of about 10 KHz-100 KHz. Further
processing of the nanosecond pulses may occur (as will be shown,
for example, the embodiment shown in FIG. 8b) wherein a high speed
modulator is used to "slice" or compress the pulse to the
picosecond scale, followed by amplification. Further details
relating to temporal pulse shaping may be found in U.S. Pat. Nos.
6,281,471 and 4,483,005 (entitled "affecting pulse width") assigned
to the assignee of the present invention.
[0139] Other physical properties may be exploited. With the
application ultrashort pulses to various materials, for instance in
the range of 50 femtoseconds to a few picoseconds, the plasma
shielding of the laser beam is generally negligible, as taught in
several references (i.e., Zhu et al., "Influence of Laser
Parameters and Material Properties on Micro-Drilling with
Femtosecond Laser Pulses," APPL. PHY. A 67 (Suppl.) 5367-5371
(1999). Though not as efficient as operating in the femtosecond
range, picosecond pulses having preferred pulse widths near the
breakdown point and somewhat longer (e.g., a range of 5% to 25%
longer) than the breakdown point may provide for better coupling of
laser energy than nanosecond pulses. For example, pulses may be in
the range of about 10 ps to 100 ps, and most preferably in the
range of about 10 ps to about 40 or in a range of about 10 ps to
about 50 ps. The coupling of energy with longer pulses, for
instance 10-30 nanoseconds, may be severely degraded as a result of
ejected vapor/plasma/plume. Further, the incident beam may be
scattered and produce substantial off-link energy which can reduce
the process energy window.
[0140] Hence, though a series of picosecond pulses may be
equivalent to a multiple nanosecond pulse for the purpose of
"on-the-fly" removal, the overall interaction of the laser with the
material and processing results may be significantly different when
a plurality of pulses each with a temporal spacing of at least
several nanoseconds between pulses is used. U.S. Pat. No. 6,552,301
discloses the use of a burst of ultrafast laser pulses, each of the
pulses having a pulse width less than about 10 ps, and having a
time separation between individual pulses to exploit the
persistence of a selected transient effect arising from an
interaction of a previous pulse with the target material. Further,
"Laser Micromachining of Transparent Glasses and Aluminum with
ps-pulse bursts at 1054 nm," Herman, CLEO 2000, CDF3, (2000),
disclosed that a 7.5 ns pulse separation mitigates plume absorption
effects to some extent. A time interval may be pre-selected based
on (at least) a time interval for substantial dissipation of
plasma/vapor/plume after application of first a high peak power,
picosecond pulse. An exemplary range is about 5 ns to several
hundred nanoseconds. Additional pulses may subsequently be applied
for efficient coupling.
[0141] Further, when picosecond pulses having high power density
(e.g., of 10.sup.9-10.sup.13 W/cm.sup.2) are applied to the link,
intensity dependent non-linear absorption, for instance within
dielectric layer 1092 or other adjacent material, may attenuate
incident energy after the link is removed and may reduce the
possibly of substrate or collateral link damage. The presence of
impurities (by design, or as a process defect or byproduct) lattice
defects or various process defects may enhance non-linear
absorption in one or more dielectric layers. Further, optical
properties of some low-k dielectrics such as polymeric dielectrics
may support controlled removal of material by non-linear
absorption.
[0142] Pico Second Laser Embodiments
[0143] Solid state laser wavelengths may be 1.3, 1.18, 1.09, 1.064,
1.053, or 1.047, microns with Neodymium (Nd) doped solid state
lasers (Nd:YAG, Nd:YLF, Nd:YVO4) or with other rare earth elements
(e.g., ytterbium (Yb), neodymium (Nd), erbium (Er)) doped fiber
lasers. Preferred laser wavelengths may also be the second, third,
fourth, and fifth harmonics of these and other appropriate lasers
to achieve small spot sizes and larger focal depths to meet the
design criteria of a particular application. For example, laser
sources with laser wavelengths in the UV (e.g., 355 nm from the
third harmonic, 266 nm from the fourth harmonic, and 212 nm from
the fifth harmonic), in the visible (e.g., 532 nm from the second
harmonic), near IR wavelengths (e.g., 700-900 nm), which provide a
spot size improvement relative to conventional wavelengths, may
also be used. One such a laser system is a mode locked Ti:sapphire
ultra fast laser (without a compressor) which produces laser pulses
with pulse widths in pico-second range in the 750 to 850 nm range.
Another is the rare earth element doped fiber laser that generates
wavelength in a range of 800-980 nm.
[0144] Exemplary laser sub-systems which may be included in
embodiments of the present invention will now be described in more
detail. In one embodiment, corresponding to FIG. 1a, a commercially
available diode pumped, passive or active mode locked system may be
included. The external modulator system 108 may be implemented to
deliver the selected pulses of 106 to link 107.
[0145] Another laser configuration which may be used in at least
one embodiment of the present invention is shown in FIG. 2a. In a
MOPA configuration a pico-second seed laser (e.g., oscillator
producing an output in a range for amplification) and (power)
amplifier system is used to obtain the pulse energy required.
[0146] Referring to FIG. 6a, a block diagram illustrating
additional details of a laser sub-system wherein the seed laser 211
of either FIG. 2a or FIG. 2b is a diode pumped, solid state laser
oscillator 602 is shown. Diode pumped, solid state laser amplifier
603 may be used to amplify the output of the seed laser. Oscillator
602 may be a mode-locked, diode pumped solid state oscillator seed.
External modulator sub-system 108 may be used to control the number
of pulses on each link and the temporal spacing between the pulses.
A mode-locked oscillator will generally operate at very high
repetition rates (>1 MHZ) compared to conventional q-switched
lasers. The laser system include may also include the modulator
sub-system 1081 of FIG. 2b with control signals 202 (e.g., in a
typical range of 20-150 KHz) to control the number of pulses on
each target while processing links during motion of the link
relative to the laser beam. In any case, the seed laser (e.g.,
which, if suitable, may be a packaged, commercially available laser
source) may include an internal pre-amplifier to amplify the pulse
energy to a suitable range for power amplification with power
amplifier 603.
[0147] An alternative configuration may include a diode pumped,
mode locked, picosecond fiber laser oscillator as a seed laser 602.
An all fiber laser system may be constructed if the diode pumped,
solid state amplifier 603 is a fiber optic amplifier.
[0148] Exemplary fiber configurations suitable for amplifying high
power short pulses, particularly ultrashort pulses, are disclosed
in U.S. Pat. Nos. 5,400,350, 5,701,319, and 5,818,630. Exemplary
lasers include the Femtolite and Wattlite series offered by IMRA,
the assignee of the '350, '319, and '630 patents. Pulses down to
0.1 ps duration with average power of 1 watt with an output
wavelength in the range of 1.03-1.06 microns have been achieved
with Yb-fiber amplified, Femtolite-based source. Other wavelengths,
(e.g., 780 nm) and frequency multiplied (second harmonic) outputs
of 1.03-1.06 micron lasers are also available from IMRA. Additional
information is also available in U.S. Pat. No. 6,281,471 (assigned
to the assignee of the present invention) and International
Published Patent Application WO 98/92050.
[0149] Various other solid state laser amplifier configurations may
be adapted for use in at least one embodiment of the present
invention. Planar waveguide technology may be well suited for high
peak power, short pulse amplification. U.S. Patent Publications
2003/0161375, 2003/0160034, and 2003/0021324, assigned to the
assignee of the present invention, and the associated references
disclose several waveguide amplifier embodiments. The waveguide
designs, though not as readily available as fiber amplifier
technology, provide high peak power outputs, and good beam quality,
without undesirable Raman shifting of the seed wavelength. Also,
planar waveguide amplifiers may be well suited for femtosecond
pulse amplifiers.
[0150] Referring to FIG. 6b, a block diagram illustrating
additional details of an alternative laser sub-system wherein the
seed laser of FIG. 2 is a picosecond laser diode 611 for producing
picosecond pulses is shown. The diode seed laser may be directly
modulated.
[0151] Alternatively, the diode laser may be used to produce
nanosecond pulses which are further processed within the laser
system to produce picosecond pulses (as will be shown in more
detail, for example, in connection with FIG. 8b).
[0152] In yet another arrangement, the seed laser 611 may be an
active or passive q-switched microchip laser. An example of a
commercially available microlaser is the AOT-YVO-1Q available from
Advanced Optical Technology. For example, AOT offers a pulsewidth
of 2 nanoseconds available at a repetition rate of 20 KHz.
Frequency doubled versions are also available (532 nm). Microchip
lasers are also offered by JDS Uniphase. In either case, a
modulator may be used to reduce the pulse width as shown in more
detail, for example, in connection with FIG. 8b. A diode pumped,
fiber laser amplifier 612 may used to amplify the output of the
seed laser.
[0153] A preferred embodiment may include the diode laser as the
seeder and a fiber laser amplifier to obtain picosecond laser
pulses. Fiber laser systems may have the advantages of compactness,
excellent beam quality and control, high system reliability, ease
in thermal management, and maintenance-free operation. U.S. Pat.
No. 6,281,471 and WO 98/92050 discloses numerous features of master
oscillator--power amplifier (MOPA) wherein a diode seed laser is
amplified with a fiber amplifier.
[0154] In at least one embodiment, the temporal spacing of a
sequence of the pulses may be controlled by "gain switching" of a
seed laser, for example, as taught in U.S. Pat. No. 6,281,471. High
speed pulsed laser designs generally utilize q-switched, gain
switched, or mode locked operation, alone or in combination.
"Pulsed pump" (e.g., real time control of pump diode module of FIG.
6a) may be used provided output stability is acceptable. U.S. Pat.
No. 5,812,569 discloses an exemplary method of stabilizing the
output energy of a pulsed solid state laser.
[0155] The output of the laser sub-system 101 (and from amplifier
603) may be wavelength shifted by shifter 105. Wavelength shifters,
including harmonic generation modules or other wavelength shifters
may be used to shift the wavelength to shorter or longer wavelength
depending on the process requirement. Wavelength shifting or
conversion techniques are well known and documented. Examples of
the wavelength shifter include Raman shifter, frequency up
conversion or down conversion, frequency doubling, etc. For
example, Concept Design Inc. provides second, third, and fourth
harmonic conversion of femtosecond Ti:Sapphire outputs (fundamental
wavelength in the range of 750-850 nm) resulting in available
wavelengths as short as about 215 nm. Additional products which
include ultrafast frequency converters are offered by Coherent,
Spectra Physics, and Lumera.
[0156] Referring to FIGS. 7(a-c), block diagrams illustrating
various alternative configurations which may be used within laser
sub-system 101 are shown. In FIG. 7a, a wavelength shifter 701 is
disposed between the seed laser and the amplifier. In this case,
the seed laser wavelength is not the same as that of the power
amplifier. Hence, wavelength shifting is implemented to shift the
output wavelength from the seed laser to a wavelength within the
range of the power amplifier. Examples of the wavelength shifter
include Raman shifter, frequency up conversion or down conversion,
frequency doubling, etc.
[0157] FIG. 7b illustrates yet another configuration wherein a
pre-amplifier 702 is disposed between the seed laser stage and
power amplifier stage. The pre-amplifier generally amplifies the
output of a picosecond seed laser prior to power amplification such
that the pulse power is generally within a favorable range for
amplification by the fiber laser amplifier (or other suitable
amplifier). Preferably, the pre-amplifier is also fiber based.
[0158] FIG. 7c illustrates yet a further configuration that
includes modulator 703 disposed prior to power amplification. The
modulator (e.g., a down-counter or divider) is generally used when
the repetition rates are different between the power amplifier and
seed laser. Usually, the repetition rate from a mode locked seed
laser is relatively high, in the range of MHZ. However, as a result
of rated average limited power the repetition rate requirement for
the power amplifier may be in the range of a few to hundreds of
KHz. Hence, the device operates as a "down-counter" or "pulse
picker" (e.g., similar or identical to the modulator sub-system and
optical switch of FIGS. 1a and 1b). Preferably, as with modulator
sub-system 108, an optical switch is driven with control signals
based on position and/or velocity information and therefore
synchronized with other components of the laser processing system.
An example of such a down-counting device can be an acoustic-optic
modulator or other high speed optical switch. The device may be
used alone or in combination with modulator 108 for selecting the
pulses to be delivered to the link or other target structure. A
wavelength shifter 105 may be disposed at the output as shown in
FIGS. 7(a-c).
[0159] Referring to FIGS. 8(a-c), schematic block diagrams
illustrating in further detail, constructions of exemplary laser
systems which may be used in embodiments of the present invention
are shown. By way of example, the seed laser may be a commercially
available semiconductor laser diode and the amplifier system
includes at least one fiber optic amplifier, and may include
several stages of amplification.
[0160] FIG. 8a illustrates a seed laser with a multi-stage
amplifier arrangement. Generally, the seeder (oscillator) generates
pulses of picosecond duration (10 ps-1 ns) with an adjustable
(i.e., modifiable, selectable, etc.) repetition rate up to 100 KHz
or 10 MHZ. A typical unit may have 40-50 ps duration with a 100 KHz
repetition rate. Both pre-amplifier and power amplifier stages are
included. A fiber based, preferably single mode, pre-amplifier 8111
generally amplifies the pulses from the seeder to a level that
leads to saturation in the final fiber power amplifier 8112 (which
may be a multi-stage amplifier). The fiber based power amplifier is
generally configured to produce output energy level in the range of
about 5 microjoules to 50 microjoules, which is generally
sufficient to remove the link with a single pulse and compensate
for losses within an optical system. For an output wavelength of 1
micron, Ytterbium doped fiber is generally chosen. The fiber may be
polarization-maintaining (PM) fiber.
[0161] FIG. 8b shows additional details of one construction of an
alternative configuration which may be included in an embodiment of
the invention. A modulated laser diode 821 may generate nanosecond
pulses (two pulses 8211 shown, not to scale). Each of the pulses
may be in an energy range of 1-200 nj, each with an exemplary pulse
width of about 2-10 ns. A q-switched microlaser may be used as an
alternative to the diode, and the tradeoffs between the choices may
be based on specific design considerations and criteria. An
isolator 831 is generally used to reduce the noise level, for
instance noise caused by back reflection. The pulses are then
amplified by diode pumped (pump diode(s) 824) and Yb amplifier 822.
The amplification may be about 30 dB to raise the pulse energy to
the microjoule range and to overcome various losses within the
system.
[0162] A second isolator 831 is generally used to reduce the noise
level caused by back reflection. A polarizer 826 is generally used
to maintain the polarization of the beam to meet design criteria
and Fiber Bragg Gratings (FBG) 825 are used as wavelength sensitive
filters. The pulse width may then be "sliced" to the pico-second
range using a very high speed GHz intensity modulator 827,
preferably with a full-power band width of at least 10 GHz.
Alternatively, a more efficient arrangement may be implemented with
a Mach-Zehnder modulator as 827 wherein the nanosecond pulses are
compressed to the picosecond range, producing a pulse width in the
approximately 10 ps range. Amplified output pulse(s) 8271 are shown
(not to scale) with removed or compressed portions depicted by
dashed lines. In this case the amplifier 822 is operated at the
final required repetition rate.
[0163] FIG. 8c shows details of a construction of an alternative
seed-amplifier and "pulse picker" configuration which may be
included in an embodiment of the invention. Overall, the
configuration of FIG. 8c is similar to that of FIG. 7b, but without
wavelength shifting, for instance. Pico-second pulses 8311 may be
generated directly from a seed diode 829 or by external modulation
(not shown) of a seed diode 829 at a multiple of the final required
repetition rate (e.g., a multiple of 1-100 KHz). The pulse energy
may typically be about 1 nj. As above, the signal is generally
amplified (e.g., by about 30 db) with amplifier 8111, before the
pulse repetition rate is reduced to the required final value by
using a suitable modulator 1081 as a "down counter" or "pulse
picker" (e.g., 1-100 KHz). The selected pulse(s) 8281 are
shown.
[0164] The selected picosecond pulses 8281 may then be amplified
with additional stages. FIG. 8d shows one of the configurations of
a two-stage amplifier. As described above, components may include
isolators 831 to reduce the noise level, a polarizer 826 to
maintain the polarization of the beam, and Fiber Bragg Gratings 825
as wavelength filters. Both fiber amplifiers 841 and 842 are
generally pumped by diodes (or diode arrays) 8411 and 8421
respectively. The first stage may be a 30 dB, single mode, Yb
amplifier. The second stage may be a "large mode" or "large core"
Yb amplifier with 30 dB gain. Various methods known in the art may
be used to control the output mode and corresponding beam quality,
and for noise (ASE) suppression (e.g., see U.S. Pat. Nos. 5,818,630
and 5,400,350, and WO 98/92050) so that a nearly diffraction
limited output beam is produced for delivery to the link. The
three-stage system of FIGS. 8c-8d may produce outputs in the range
of tens--hundred microjoules with beam quality that is
approximately diffraction limited.
[0165] Methods and systems of delivering pump energy to fiber
amplifiers are well known. FIG. 8e shows, by way of example, one of
the methods of coupling the diode laser energy into a fiber
amplifier. Dichroic mirrors 850 in combination with an optical
system (e.g., lens system) may transmit the pump light into
Yb-doped, double clad fiber 851 through perpendicularly cleaved
fiber ends 852. The amplifier output may be transmitted with a
similar dichroic arrangement wherein pump energy 855 is
recirculated through fiber. Skilled persons will appreciate and
understand other possible appropriate combinations of different
types of laser sources for seed and amplifier lasers may be
implemented to meet the design criteria of a particular
application.
[0166] Memory Repair System
[0167] Referring to FIG. 9, a block diagram of a laser based memory
repair system, including a pico-second laser system, and further
illustrating numerous major system components of the present
invention is shown.
[0168] Complete micromachining stations using pico second lasers
may be implemented. At least one embodiment of a picosecond laser
system may be integrated into the M430 series manufactured by GSI
Lumonics, or other micromachining systems having suitable
sub-micron tolerances and performance specifications for high speed
micromachining. The following list of attached patents and
published applications, assigned to the assignee of the present
invention, describe numerous aspects related the memory repair
methods and systems:
[0169] 1. U.S. Pat. No. 5,300,756, entitled "Method and System for
Severing Integrated-Circuit Connection Paths by a Phase Plate
Adjusted Laser beam";
[0170] 2. U.S. Pat. No. 6,144,118, entitled "High Speed Precision
Positioning Apparatus";
[0171] 3. U.S. Pat. No. 6,181,728, entitled "Controlling Laser
Polarization";
[0172] 4. U.S. Pat. No. 5,998,759, entitled "Laser Processing";
[0173] 5. U.S. Pat. No. 6,281,471, entitled "Energy Efficient,
Laser-Based Method and System for Processing Target Material";
[0174] 6. U.S. Pat. No. 6,340,806, entitled "Energy-Efficient
Method and System for Processing Target Material Using an
Amplified, Wavelength-Shifted Pulse Train";
[0175] 7. U.S. application Ser. No. 09/572,925, entitled "Method
and System For Precisely Positioning A Waist of A
Material-Processing Laser Beam To Process Microstructures Within A
Laser-Processing Site", filed May 16, 2000, and published as WO
0187534 A2, December, 2001, now U.S. Pat. No. 6,483,071, Division
of Ser. No. 09/572.925;
[0176] 8. U.S. Pat. No. 6,300,590, entitled "Laser Processing";
and
[0177] 9. U.S. Pat. No. 6,339,604, entitled "Pulse Control in Laser
Systems".
[0178] As apparent from the teachings herein the present invention
provides for processing of links with pitch of less than 2 microns
with a negligible heat affected zone, and without the complexity of
a femtosecond laser system. Precise link removal may be facilitated
with one or more picosecond pulses. Further, link removal may be
accomplished with high efficiency when compared to a slow etching
process, and with improved precision when compared to conventional
nanosecond link processing approaches. Link processing in
accordance with the present invention may be carried out in a high
speed laser processing system.
[0179] While embodiments of the invention have been illustrated and
described, it is not intended that these embodiments illustrate and
describe all possible forms of the invention. Rather, the words
used in the specification are words of description rather than
limitation, and it is understood that various changes may be made
without departing from the spirit and scope of the invention.
* * * * *