Auburn University v. International Business Machines, Corp.

Filing 381

ORDER AND MEMORANDUM OPINION: it is hereby ORDERED that (1) IBM's 261 Motion for Summary Judgment of Noninfringement is GRANTED and the claims contained in Counts III and IV of Auburn's 87 First Amended Complaint are DISMISSED WITH PREJ UDICE; (2) IBM's 257 Motion for Summary Judgment of No Willful Infringement is DENIED as MOOT; (3) Auburn's 258 Motion for Summary Judgment on IBM's 35 USC Section 273 Affirmative Defense is DENIED as MOOT; (4) The Court RESERVES RULING on Auburn's 260 Motion for Partial Summary Judgment of No Invalidity of Claims. Signed by Honorable Judge Mark E. Fuller on 8/2/2012. (Attachments: # 1 Civil Appeals Checklist) (wcl, )

IN THE UNITED STATES DISTRICT COURT
FOR THE MIDDLE DISTRICT OF ALABAMA
EASTERN DIVISION
AUBURN UNIVERSITY,
Plaintiff,
v.
INTERNATIONAL BUSINESS
MACHINES, CORP.,
Defendant.

)
)
)
)
)
)
)
)
)

CASE NO. 3:09-cv-694-MEF
(WO)

ORDER AND MEMORANDUM OPINION
This cause is before the Court on IBM’s Motion for Summary Judgment of
Noninfringement (Doc. #261.) For the reasons set forth below, this motion is due to be
GRANTED.
I. Background
On June 17, 2010, Auburn filed its First Amended Complaint for Correction of
Inventorship and Patent Infringement. Auburn’s complaint alleges that IBM infringes
two patents owned by Auburn: U.S. Patent No. 7,194,366 (“the ‘366 Patent” or “‘366”)
and U.S. Patent No. 7,409,306 (“the ‘306 Patent” or “‘306”) (collectively “the Auburn
Patents”). The Auburn Patents describe a “system and method for estimating [the]
reliability of components for testing and quality optimization.” (Auburn Patents, Docs.
#261-1, 261-2.)1 Auburn contends that IBM’s methods of testing various of its products

1

Throughout this opinion the Court will refer to the Court-assigned page numbers appearing at
the top of all docketed materials.

(“components”)2 infringe the Auburn Patents. The Court laid out the background of the
Auburn Patents in its Claim Construction Order (“CCO”) (Doc. #193 at 7-10) and
incorporates that discussion here.
Per this Court’s Amended Scheduling Order, (Doc. #207), IBM filed its motion for
summary judgment of noninfringement on October 28, 2011. The motion came under
submission on December 14, 2011. The Court has considered the arguments contained in
IBM’s motion, Auburn’s response, and IBM’s reply, and is fully apprised of the issues
contained therein.
Auburn alleges that, by performing certain testing methods on certain IBM
components, IBM infringes claims 1, 8, 10-13, 15, and 17-20 of ‘366, and claims 1, 2, 6,
7, 14, and 16 of ‘306.
A. The accused IBM products
For purposes of deciding this Motion, the accused IBM components can be
grouped into three general product categories:3
(1) The P/Z products: Auburn alleges that IBM infringes the
asserted claims of the ‘366 Patent during the actual
manufacture/post-production testing of certain IBM
microprocessor chips referred to by Auburn collectively as “the
P/Z products.” The accused P/Z products are the Power5+,

2

Throughout this opinion, the Court will use the word “component” to refer to all of IBM’s
accused products, which include various electronic components, integrated circuits, and die (Doc. #261 at
4.)
3

When this motion was originally filed, there was a fourth group of products, “the Sony
products.” However, during the pendency of this motion, all claims regarding the Sony products have
been resolved. (Doc. #370, Doc. #375.)

2

Power6, Power7, Z6 (EP), Z6 (ES), Z7 (CP), and Z7 (SC)
microprocessors.
(2) The Corona/Trimaran products: Auburn also alleges that
IBM infringes the asserted claims of the ‘366 Patent during the
actual manufacture/post-production testing of two L3 (Level 3)
cache memory chips referred to as Trimaran and Corona.
(3) The Dreadnaught/DDP223O products: Regarding the ‘306
Patent, Auburn alleges that IBM infringes the asserted ‘306
Patent claims during IBM’s actual manufacture/postproduction
testing of two Application Specific Integrated Circuit (ASIC)
chips referred to as the Dreadnaught and DDP2230 products.
(Doc. #261, at 10.)
B. The Auburn Patents and the CCO
Between ‘366 and ‘306, there are three independent claims: ‘366 claims 1 and 13,
and ‘306 claim 1. Each of these independent claims sets out a similar three-step method
for testing components.
(1) The ‘366 patent
Auburn alleges that IBM’s P/Z and Trimaran/Corona products infringe on claims
1, 8, 10-13, and 17-20 of ‘366.
As mentioned above, only claims 1 and 13 of ‘366 are independent; IBM cannot
infringe the dependent claims of ‘366 if it does not infringe the independent claims.
Wahpeton Canvas Co., Inc. v. Frontier, Inc., 870 F.2d 1546, 1552 n.9 (Fed Cir. 1989).
Claim one of ‘366 sets out:
A method of [determining] the post-production reliability of a
repairable integrated circuit die component, said method
comprising:
3

performing an initial post-production test on the
component to identify repairable defects in the
component; and
classifying the component into a classification of a
plurality of reliability probability classifications
based on the number of repairable defects
identified by the initial test; and
estimating the reliability of the component based on the
classification.
(‘366 Patent, Doc. #261-1, at 21.)
Claim thirteen of‘366 sets out:
A method for predicting the post-production reliability of an
integrated circuit die component, said method comprising:
performing an initial post-production test to identify a
number of killer defects;
classifying a component into one of a plurality of
reliability probability classifications based on the
number of killer defects identified by the initial
test; and
optimizing further testing of the component to identify
the presence of latent defects, based on the
classification thereof.
(Id.)
The Court has construed the relevant language of claims one and thirteen of‘366 as
follows:

4

Independent Claim

Disputed Claim Term

Court’s Construction

‘366, Claim 1

“estimating the reliability
of the component based on
the classification”

“either simultaneously with
or after the ‘classifying’
step, statistically
calculating a numerical
value of the reliability of
the component based on
the classification”

‘366, Claim 13

“predicting the postproduction reliability”

“statistically calculating a
numerical value of the
reliability”

‘366, Claim 13

“optimizing further testing
of the component to
identify the presence of
latent defects, based on the
classification thereof.”

“either simultaneously with
or after the ‘classifying’
step, optimizing further
testing of the component to
identify the presence of
latent defects, based on the
classifications thereof”

‘366, Claims 1 and 13

“a plurality of reliability
probability classifications”

“at least two classifications
of components associated
with respective levels of
expected reliability”

(CCO, Doc. #193, at 46-47.)
(2) The ‘306 patent
Auburn alleges that IBM’s Dreadnaught/DDP2230 products infringe claims 1, 2,
6, 7, 14, and 16 of‘306, though claim 1 is the only independent claim of the ‘306 patent.
Therefore, to infringe claims 2, 6, 7, 14, and 16, IBM must first be held to infringe claim
1. Wahpeton Canvas, 870 F.2d at 1552 n.9.
Claim 1 of ‘306 sets out

5

A method of determining the reliability of a component, the
method comprising:
classifying the component based on an initial
determination of a number of fatal defects; and
estimating a probability of latent defects present in the
component based on the classification, by
integrating yield information based on the initial
determination of a number of fatal defects using
a statistical-defect clustering model.
(‘306 Patent, Doc. #261-2, at 21.)
The Court construed the relevant language in claim 1 of the ‘306 patent as follows:
Disputed Claim Term

Court’s Construction

“estimating a probability of latent defects
in the component based on the
classification”

“either simultaneously with or after the
‘classifying’ step, quantifying the
probability of latent defects in the
component with a numerical value based
on the classification.”

“determining the reliability”

“quantifying the reliability with a
numerical value.”

(CCO, Doc. #193, at 30, 32.)
D. Factual Background
The facts at issue center on IBM’s method of testing its accused products. In lieu
of an “Undisputed Facts” section as might typically be found in a motion for summary
judgment, IBM bases its factual allegations on the expert report and testimony of
Auburn’s expert, Dr. F. Joel Ferguson (Doc. #261, at 20 n.29.) Auburn alleges that IBM
mischaracterizes “Auburn’s factual allegations and legal contentions, as well as the expert
opinions of Dr. Ferguson in numerous instances.” (Doc. #276, at 7). However, the Court

6

accepts Dr. Ferguson’s description of IBM’s processes, construed in the light most
favorable to Auburn, as the underlying factual basis for IBM’s motion.
(1) The P/Z products
To infringe claims 1 and 13 of ‘366, IBM would have to apply the following three
step process to its P/Z products:
(1) performing an “initial post production test” to identify
repairable (‘366, claim 1) or killer (‘366, claim 13) defects;
(2) “classifying” the chips based on the number of defects as
determined by the initial post-production test (‘366 claim 1, 13);
and
(3) “estimating” (‘366, claim 1) or “predicting” (‘366 claim 13)
the reliability of a component “based on the classification.”
(Doc. #261-1, at 21.)
IBM’s actual manufacture/post-production testing process for its P/Z products
involves three phases: the wafer test, the package test, and the stress test/burn-in phases
(Doc. #261, at 26.)4 Prior to this three-stage actual manufacture/post-production testing
process, IBM also tests selected components during what IBM refers to as its
“qualification process” (Id.) Auburn alleges that IBM performs the three steps indicated
above, as instructed by ‘366, on the accused products through a combination of IBM’s
pre-production qualification and actual manufacture/post-production testing processes.

4

Dr. Ferguson explains that during stress testing and burn-in, “integrated circuits are subjected to
various non-optimal conditions to precipitate failures caused by latent defects. However, stress testing
may damage the die, and also involves time and expense.” (Expert Report of Dr. F. Joel Ferguson, Doc.
#261-17 at 9.)

7

(a) Description of IBM’s pre-manufacture “qualification” process
Prior to any actual manufacture/post-production testing (the wafer-test, packagetest, and stress-test/burn-in processes) of IBM’s components, IBM puts selected P/Z
components through what IBM refers to as its “T2 Qualification Process” (the
“qualification process”). In the qualification process, IBM selects a “representative
sample of die” (Expert Report of Dr. F. Joel Ferguson, Doc. #261-17, at 52.) The
purpose of the qualification process is to project reliability targets for the products that
will later be fully manufactured and tested. (Id. at 52-53.) This section will detail IBM’s
qualification process.
(i) Selection of Components for the Qualification Process
As part of IBM’s qualification process, components are subjected to the actual
manufacture/post-production testing processes, supra Part I.D.1, to which all chips will
later be subjected. To survive qualification, components must successfully pass the
“various functional and parametric tests applied during wafer-level and package-level
testing.” (Expert Report of Dr. F. Joel Ferguson, Doc. #261-17, at 52.) This includes a
“repair-count threshold” or “repair-screen” that serves to optimize testing by classifying
components into “good” or “bad” categories based on the number of defects found in
those components. The “good” components continue through the process, while the
“bad” components are scrapped. “The goal is to perform T2 evaluations on die
populations that are representative of the die populations produced during full-scale
manufacturing [i.e. chips that have survived the wafer- and package-level testing], as the
8

common basis permits the extrapolation of the reliability estimates from the qualification
context to the manufacturing context.” (Expert Report of Dr. F. Joel Ferguson, Doc.
#261-17, at 52; see infra note 11 and accompanying text.)
(ii) Testing of the Components Selected for the Qualification
Process
After selecting components within the parameters described above, IBM then
subjects this “representative sample” to “‘various kinds of stress testing.’” (Expert
Report of Dr. F. Joel Ferguson, Doc. #261-17, at 53.) The selected components are tested
to the point of total failure in order to “observe failures throughout an accelerated version
of the useful life of the P/Z products and potentially beyond . . . and this stress testing
results in a time-to-fail distribution over the expected life of the P/Z products.’” (Id.)
(iii) Creating “Reliability Failure-inTime Targets” From the
Results of the Qualification process
The results of the stress testing done in the qualification process are input to a
statistical model to estimate the reliability of the P/Z products. IBM’s microelectronics
division then sends the failure rates produced by the reliability models to the IBM server
group in documents called “T2 letters.” These letters contain the reliability estimates
produced by the reliability models, expressed in “failures-in-time” (“FIT”) units. (Expert
Report of Dr. F. Joel Ferguson, Doc. #261-17, at 57-59.)5
(iv) Application of Reliability FIT Targets
5

FIT measures “the number of failed parts per million per thousand hours.” (Expert Report of
Dr. F. Joel Ferguson, Doc. #261-17, at 57.) The fewer failures in time a given component has—i.e. the
lower its FIT rate—the more reliable that component is.

9

Fit targets contained in the T2 letters apply only to components that pass all
necessary actual manufacture and post-production testing, including burn-in. (Deposition
of Dr. F. Joel Ferguson, Doc. #261-6 at 36-45.) It is only after the qualification process is
completed on a select number of components that IBM begins its actual
manufacture/post-production testing process in order to produce customer-ready P/Z
products. Infra Part I.D.1.d.
(b) IBM’s Initial Post-Production Test
As discussed above IBM performs an “initial post-production test” or “repaircount screen” on the P/Z products during both the qualification and actual
manufacture/post-production testing phases. IBM also applies the repair-count screen
during wafer and package testing. During the wafer test phase, before each die has been
separated from the other die on a given wafer, IBM counts the number of necessary
repairs within each P/Z die. (Doc. #261-17 at 41-43.) Then, during the package-test
phase, IBM separates each P/Z die from the remainder of the die that make up a given
wafer, mounts the die on a “carrier made of ceramic, plastic, or some other similarlysuited material” (id. at 7), and counts the number of required repairs to an individual P/Z
die (id. at 50-51).
At both of these stages, IBM’s repair-count test compares the number of required
repairs (the “repair count”) to a pre-determined threshold number of repairs, with the

10

threshold being less than the maximum number of repairs that can be accommodated by
the redundant circuitry. (Id. at 43.)
(c) IBM’s “Classification” Process6
After the repair count is performed, IBM classifies each chip as either
“good”—“if the number of required repairs is below the predetermined ‘repair count
threshold’”—or bad—“if the number of required repairs exceeds the predetermined
‘repair count threshold.’” The “bad” chips are scrapped, and the “good” chips continue to
the next phase of the manufacture and testing process. (Id. at 43-48.)
(d) “Predicting” or “Estimating” a Numerical Value
After classification, the components are put through the stress-test/burn-in phase,
during which the die are subjected to additional testing that components must pass before
they are deemed ready for shipment to the customer. Once the finished components are
otherwise customer-ready, the numerical FIT targets derived from the qualification
process are applied to the finished components. (Expert Report of Dr. F. Joel Ferguson,
Doc. #261-17, at 52-53, 58-59.) The result is that IBM has a numerical value of
reliability for its finished product, which IBM may then, for example, compare with the
FIT target desired by a customer, allowing IBM to ship to its customer only the
components that meet or exceed the desired FIT Target.
(2) Trimaran/Corona Products

6

IBM also refers to classifying components as “binning” or “sorting.” (Id.)

11

As with the P/Z products, to infringe claims 1 and 13 of ‘366, IBM would have to
apply the following three step process to its Trimaran/Corona products:
(1) performing an “initial post production test” to identify
repairable (‘366, Claim 1) or killer (‘366, Claim 13) defects;
(2) “classifying” the chips based on the number of defects as
determined by the initial post-production test (‘366 Claim 1,
13); and
(3) “estimating” (‘366, Claim 1) or “predicting” (‘366 Claim 13)
the reliability of a component “based on the classification.”
(Doc. #261-1, at 21.)
The processes used to test and manufacture IBM’s accused Trimaran/Corona
products are largely the same as the process laid out above, supra Part. I.D.1, for the P/Z
products. (Expert Report of Dr. F. Joel Ferguson, Doc. #261-17 at 62 (“The general set
of manufacturing tests that IBM performs on the Trimaran and Corona are largely similar
to the P/Z products . . . . The same is true for the hardware and software implementation
of these test processes.”).) This section will detail only the differences between the P/Z
process and the Trimaran/Corona process. Despite these slight differences, the
Trimaran/Corona process and the P/Z process are sufficiently similar so that the same
legal conclusions apply to both sets of accused products.

(a) Description of IBM’s Pre-Manufacture “Qualification” Process

12

IBM uses the same qualification process, and derives FIT targets from the
qualification process in much the same way, in its testing of the P/Z and Trimaran/Corona
products. Specifically, IBM uses the same “reliability model” on the Trimaran/Corona
products as it does on the P/Z products. (Expert Report of Dr. F. Joel Ferguson, Doc.
#261-17 at 67-69.)
(b) IBM’s Initial Post-Production Test
The initial post-production test, or repair-count screen, is applied to the
Trimaran/Corona products in much the same way as it is applied to the P/Z products. A
repair-count screen is applied, and the components are classified as a result of the number
of defects in each component, as revealed by the repair-count screen (Doc. #261-17, at
62-63.)
(c) IBM’s “Classification” Process
IBM’s classification of the Trimaran/Corona products results in greater nuance and
more categories of components than does the classification process for the P/Z products.
At different points in time, IBM has (1) classified Trimaran/Corona components into
“good” and “bad” classifications, scrapping the “bad” chips, (2) differentiated the amount
of subsequent burn-in testing certain components will receive based on the number of
defects revealed by the repair-count screen, and (3) directed components to “lower or
higher reliability applications” based on the number of repairable defects. (Deposition of

13

Dr. F. Joel Ferguson, Doc. #261-6, at 22-28; Expert Report of Dr. F. Joel Ferguson, Doc.
#261-17, at 62-67.)
(d) “Predicting” or “Estimating” a Numerical Value
The methods IBM uses to predict or estimate the numerical value of the
Trimaran/Corona products is insufficiently distinguishable from the method used on the
P/Z components as to make any difference in the Court’s legal analysis. (See Expert
Report of Dr. F. Joel Ferguson, Doc. #261-17 at 67-68.)

(3) Dreadnaught/DDP2230 Products
To infringe claim 1 of ‘306, IBM would have to apply the following three step
process to its Dreadnaught/DDP2230 products:
(1) classifying the component based on an initial determination
of a number of fatal defects;
(2) estimating a probability of latent defects present in the
component based on the classification, by
(3) integrating yielded information based on the initial
determination of a number of fatal defects using a statistical
defect-clustering model.
(‘306 patent, Doc. #261-2, at 21.)
As is the production process for IBM’s accused Trimaran/Corona products, the
production process for the Dreadnaught/DDP2230 products is very similar to the process
laid out above, supra Part I.D.1, for the P/Z products. (See Expert Report of Dr. F. Joel
Ferguson, Doc. #261-17 at 70-76.) This section will detail only the differences between
the P/Z process and the Dreadnaught/DDP2230 process. Despite these slight differences,
14

as discussed, infra Part I.D.3.a-c, the Dreadnaught/DDP2230 process and the P/Z process
are sufficiently similar so that the same legal conclusions apply to both sets of accused
products.
(a) Description of IBM’s pre-manufacture “qualification” process
IBM uses the same qualification process, and derives FIT targets from the
qualification process in much the same way in its testing of the P/Z and
Dreadnaught/DDP2230 products (Doc. #261-17 at 76-78.)
(b) IBM’s initial post-production test and classification process
During the wafer test phase, IBM performs a “nearest neighbor threshold test” on
the Dreadnaught/DDP2230 products:
During a nearest neighbor test, the number of defective die
surrounding a given die . . . is counted. The surrounding die,
called “neighboring die” or simply “neighbors,” are tested to see
how many of those die contain certain types of defects. . . . Once
the neighboring die are counted at wafer test, IBM categorizes
a given central Dreadnaught/DDP2230 die that passed wafer
testing into “good” or “bad” . . . categories based on the number
of defective neighbors surrounding that die.
(Doc. #261-17. at 71-72.)
The parties dispute when IBM performed the nearest neighbor test on the
Dreadnaught/DDP2230 products. Specifically, IBM argues that it ran the T2
qualification process on Dreadnaught/DDP2230 products before ever implementing a
nearest neighbor threshold test on those products and therefore could not have “estimated
a probability of latent defects . . . based on the classification” (Doc. #261, at 47 (emphasis
15

original)) where the classification would have been based on the results of a nearestneighbor test that IBM was not yet performing.7 Auburn argues that IBM did run the
nearest-neighbor test during qualification of the Dreadnaught/DDP2230 products and
therefore classified those products based on the results of the nearest-neighbor test. (Doc.
#276, at 36-37.) Accepting the facts in the light most favorable to Auburn, the Court
finds that IBM did use a nearest neighbor test to classify the Dreadnaught/DDP2230
products during qualification.
(c) “Predicting” or “estimating” a numerical value
The methods IBM uses to estimate or determine the numerical value of the
Dreadnaught/DDP2230 products is insufficiently distinguishable from the method used
on the P/Z and Trimaran/Corona chips as to make any difference in the Court’s legal
analysis. (Expert Report of Dr. F. Joel Ferguson, Doc. #261-17 at 76 (“IBM estimates the
reliability of the Dreadnaught and DDP2230 products in a manner similar to the P/Z
products discussed above. In particular, IBM uses a reliability model to produce a
statistical estimate of the failures-in-time (FITs) and failed parts-per-million (PPM) for
the Dreadnaught and DDP2230 products.”)
II. Standard of Review
A. Legal standard governing patent infringement

7

Despite hinging on the same language, the “based on” argument IBM makes with regard to the
Dreadnaught/DDP2230 is different from the one which the Court uses to draw its non-infringement
conclusion for all three of the accused product classes. See infra Part III.A.2.

16

The patent owner bears the burden of proof to “present evidence sufficient to
establish that one or more patent claims are infringed.” Jazz Photo Corp., v. Int’l Trade
Comm’n, 264 F.3d 1094, 1102 (Fed. Cir. 2001). “[T]he burden remains with the patentee
to prove infringement, not on the defendant to disprove it.” Welker Bearing Co. v. PHD,
Inc., 550 F. 3d 1090, 1095 (Fed. Cir. 2008). A determination of patent infringement
requires a two-step analysis: (1) claim construction; and (2) comparison of the properly
construed patent claims to the accused method or process. Mars, Inc. v. H.J. Heinz Co.,
L.P., 377 F.3d 1369, 1373 (Fed. Cir. 2004). In this opinion the Court will undertake the
second step of the analysis by comparing the accused IBM test methods to the properly
construed patent claims to determine if Auburn can meet its burden to prove infringement
of the claims. Id.
(1) Literal Infringement
A claim is literally infringed “when each of the claim limitations reads on, or in
other words is found in, the accused device.” Allen Eng’g Corp. v. Bartell Indus., Inc.,
299 F.3d 1336, 1345 (Fed. Cir. 2002) (internal quotation marks and citation omitted); see
also Techsearch, L.L.C. v. Intel Corp., 286 F.3d 1360, 1372 (Fed. Cir. 2002) (“To
establish literal infringement, all elements of the claim, as correctly construed, must be
present in the accused system.”) (citation omitted). There can be no infringement as a
matter of law if even a single claim limitation is absent in the accused device or method.
Phonometrics, Inc. v. Northern Telecom, Inc., 133 F.3d 1459, 1467 (Fed. Cir. 1998).
(2) Infringement under the doctrine of equivalents
17

When there is no literal infringement, a patentee may attempt to prove
infringement under the doctrine of equivalents. Warner-Jenkinson Co., Inc. v. Hilton
Davis Chem. Co., 520 U.S. 17, 21 (1997). An infringement analysis under the doctrine of
equivalents requires an element-by-element correspondence between the accused product
or method and the patent claims to determine if the differences are only insubstantial. Id.
at 29, 40. The doctrine of equivalents may not be applied when doing so would vitiate a
claim limitation. Id. at 29.
B. Summary judgment standard
“Summary judgment is appropriate if the pleadings, depositions, answers to
interrogatories, and admissions on file, together with the affidavits, if any, show there is
no genuine issue of material fact and that the moving party is entitled to judgment as a
matter of law.” Greenberg v. BellSouth Telecomms., Inc., 498 F.3d 1258, 1263 (11th Cir.
2007) (citation and internal quotation marks omitted); see also Fed. R. Civ. P. 56(a) (“The
court shall grant summary judgment if the movant shows that there is no genuine dispute
as to any material fact and the movant is entitled to judgment as a matter of law.”).
“[A] party seeking summary judgment always bears the initial responsibility of
informing the district court of the basis for its motion, and identifying those portions of
[the record] which it believes demonstrate the absence of a genuine issue of material
fact.” Celotex Corp. v. Catrett, 477 U.S. 317, 323 (1986) (quotation omitted). The
movant can meet this burden by presenting evidence showing there is no dispute of
material fact, or by showing the non-moving party has failed to present evidence in
18

support of some element of its case on which it bears the ultimate burden of proof. Id. at
322-23.
If the movant satisfies its evidentiary burden, the non-moving party must then
establish, with evidence beyond the pleadings, that a genuine issue material to each of its
claims for relief exists. Clark v. Coats & Clark, Inc., 929 F.2d 604, 608 (11th Cir. 1991);
Fed. R. Civ. P. 56(c). What is material is determined by the substantive law applicable to
the case. Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248 (1986); see also Lofton v.
Sec’y of the Dep’t of Children & Family Servs., 358 F.3d 804, 809 (11th Cir. 2004)
(“Only factual disputes that are material under the substantive law governing the case will
preclude entry of summary judgment.”). Furthermore, “[t]he mere existence of some
factual dispute will not defeat summary judgment unless that factual dispute is material to
an issue affecting the outcome of the case.” McCormick v. City of Ft. Lauderdale, 333
F.3d 1234, 1243 (11th Cir. 2003) (citation and internal quotation marks omitted).
A genuine dispute as to a material fact can be found only “if the evidence is such
that a reasonable jury could return a verdict for the nonmoving party.” Anderson, 477
U.S. at 248; see also Greenberg, 498 F.3d at 1263. However, if the evidence on which
the nonmoving party relies “is merely colorable, or is not significantly probative,
summary judgment may be granted.” Anderson, 477 U.S. at 242 (citations omitted).
Likewise, “[a] mere scintilla of evidence in support of the nonmoving party will not
suffice to overcome a motion for summary judgment[,]” Young v. City of Palm Bay, 358
F.3d 859, 860 (11th Cir. 2004), and the nonmoving party “must do more than simply
19

show that there is some metaphysical doubt as to the material facts,” Matsushita Elec.
Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 586 (1986). Furthermore, a nonmoving
party’s “conclusory allegations . . . in the absence of supporting evidence, are insufficient
to withstand summary judgment.” Holifield v. Reno, 115 F.3d 1555, 1564 n.6 (11th Cir.
1997); see also Cordoba v. Dillard’s, Inc., 419 F.3d 1169, 1181 (11th Cir. 2005)
(“Speculation does not create a genuine issue of fact . . . .”) (emphasis in original).
When a nonmovant fails to set forth specific facts supported by appropriate
evidence sufficient to establish the existence of an element essential to his case and on
which the nonmovant will bear the burden of proof at trial, summary judgment is due to
be granted in favor of the moving party. Celotex Corp., 477 U.S. at 323 (“[F]ailure of
proof concerning an essential element of the nonmoving party’s case necessarily renders
all other facts immaterial.”).
In the patent context, infringement is properly decided at the summary judgment
stage only when no reasonable jury could find that every limitation recited in the properly
construed claim either is or is not found in the accused device, either literally or under the
doctrine of equivalents. Gart v. Logitech, Inc., 254 F.3d 1334, 1339 (Fed. Cir. 2001).
III. Discussion
A. IBM does not literally infringe the ‘366 patent with regard to the P/Z products
IBM makes four arguments as to why it does not infringe with regard to the P/Z
Products. IBM argues that:

20

(1) IBM never performs a step of “classifying” the P/Z products
into a “plurality of reliability probability classifications.”
(IBM’s Motion, Doc. #261 at 27.)
(2) The numerical reliability targets from IBM’s qualification
process relied upon by Auburn are inapplicable to the actual
manufacture/post-production P/Z components that Auburn
alleges are classified into “reliability probability classifications.”
(Doc. #261 at 29.)
(3) Auburn’s infringement theory for the P/Z products fails as a
matter of law because IBM’s sequence does not satisfy the
Court’s claim construction ruling (Doc. #26, at 31.)
(4) Auburn is improperly reading the ‘366 patent claims to cover
methods for testing the P/Z products that the Court has ruled
were “already being performed” in the prior art. (Doc. #261 at
32.)
IBM’s second and third arguments lead the Court to conclude that IBM does not
infringe with regard to the P/Z products because IBM does not estimate or predict a
numerical value of reliability “based on” the classification of its components as required
to infringe claims 1 and 13 of ‘366.8
8

On its face, claim 13 of ‘366 does not appear to include the limiting step of “statistically
calculating a numerical value of the reliability” in the same way claim 1 of ‘366 does. Compare CCO,
Doc. #193, at 46 (construing the third claimed step of ‘366 claim 1, “estimating the reliability of the
component based on the classification” to mean, in relevant part. “statistically calculating a numerical
value of the reliability of the component based on the classification”) with CCO, Doc. #193, at 46
(construing the term “predicting the post-production reliability” from the preamble of ‘366 claim 13, to
mean “statistically calculating a numerical value of the reliability”).) The Court found that “the term
‘predicting the post-production reliability’ of claim 13 is the essence of the invention and, therefore, a
limitation on the claim.” (Id. at 39.) The Court went on to find that “[u]sing statistical modeling to
numerically predict or estimate reliability is not merely a preferred embodiment of claim 1, it is the
invention in claim 1” and that “[s]imilarly, a reading of the whole ‘366 Patent reveals that the invention of
claim 13 is the extension of this improved method to repairable integrated circuit dies[.]” (Id. at 44.) The
Court concluded that the methods detailed in claims 1 and 13 involved “statistically calculating a
numerical value of the reliability of the component.” (Id. at 45.) To the extent necessary, the Court here
clarifies that claim 13 of the Auburn patent requires the statistical calculation of a numerical value of
reliability based on the classification, as does claim 1. Therefore, the same analysis applies to determine

21

(1) A reasonable juror could conclude that IBM, through its “qualification
process,” classifies the P/Z Products into a “plurality of reliability probability
classifications” as that term was construed by this Court
IBM first argues that it does not perform the required step of “classifying” its P/Z
Products into a “plurality of reliability classifications.” Based on IBM’s own description
of its chip-testing methods, however, a reasonable juror could find that IBM does in fact
classify these products into a plurality of reliability probability classifications.
The Court construed the term “a plurality of reliability probability
classifications[,]” as used in claims 1 and 13 of the ‘366 patent, to mean “at least two
classifications of components associated with respective levels of expected reliability.”
IBM argues that it does not classify its P/Z components into a plurality of reliability
classifications because it does not statistically quantify the reliability of the components
that are designated as “bad” during the repair-count test. (Doc. #261, at 27-28.) This
argument relies on a misinterpretation of the CCO.
IBM interprets the Court’s construction to mean that “each of the ‘respective levels
of expected reliability’ must be statistically quantified because that is ‘the true invention’
of ‘366 claims 1 and 13 and the statistical calculation of a numerical value of the
reliability of components in different bins is what distinguishes the claims from the
admitted prior art[.]” (Doc. #261 at 27 (quoting CCO, Doc. #193 at 45-46) (citing Doc.
#193 at 28 n.20).)

whether the testing methods used on the P/Z and Trimaran/Corona products constitute infringement of
either claim 1 or claim 13 of ‘366.

22

IBM misconstrues the CCO in arguing that the CCO requires statistical
quantification of the expected levels of reliability for all component classifications in
order to achieve a “plurality of reliability probability classifications.” The language of
the CCO instead supports Auburn’s contention that “the Court expressly rejected any
requirement of statistically quantifying reliability as part of the classifying step [and]
instead permitted classifying simply based on ‘relative probability.’” (Doc. #276, at 10)
(emphasis original).)
First, the CCO states that “It is only by considering how the patented inventions
predict or determine the reliability that [the] invention’s boundaries can be defined.”
(Doc. #193 at 43 (emphasis original).) The Court determined the scope of the inventions
contained in Auburn’s patents based on how those inventions predicted or determined
reliability; that is, based on the means of accomplishing the third step of Auburn’s
patented methods (estimating/predicting), rather than the second step of the methods
(classification). Supra, Part I.D.1. The implication of this statement, made more explicit
later in the CCO and discussed below, is that one could practice Auburn’s patent even by
performing the classification stage just as it had been performed in the prior art, as it is
not the classification stage that serves to distinguish Auburn’s inventions from the prior
art.
The CCO goes on to state that “[g]iven the entire specification, including the
known models already in use and the need identified by the ‘366 patent, it is clear that the
invention of Claim 1 must call for statistical modeling in the third step of Auburn’s
23

patented process, the numerical estimation or prediction of reliability.” (Doc. #193, at 44
(emphasis added).) The CCO does not explicitly state that the invention of either claim 1
or claim 13 of ‘366 calls for statistical modeling in the second step, the classification
stage.
Finally, the CCO notes that “the parties disputed whether ‘reliability probability
classifications’ must be ‘statistically calculated’ or need merely be ‘respective levels of
expected reliability,’” concluding that “[a] person performing the patented method in
claims 1 and 13 could classify the components and then sort them into bins based upon
relative probability because the claimed methods improve upon this known method by
statistically quantifying the reliability of the component.” (Id. at 45-46 (emphasis
original).) In other words, one could practice the Auburn patent even by classifying
components based on relative probability, a practice already known in the art, as long as
the reliability of the component was statistically quantified during the next step of the
process.
Therefore, it is not necessary that IBM statistically calculate the reliability of its
P/Z components during the classification stage in order to meet the court’s definition of
“classifying the component into a classification of a plurality of reliability probability
classifications.” IBM performs the classification step of claims 1 and 13 of ‘366 by
classifying components into categories of “good” and “bad.” However, for the reasons
discussed below, IBM’s statistical analysis is not “based on” this non-quantitative
classification, and IBM is therefore not practicing Auburn’s patented method.
24

(2) The FIT targets from IBM’s qualification process are inapplicable to the
actual manufacture/post-production P/Z components
IBM argues that “The Numerical Reliability Targets From The Qualification
Process . . . Are Entirely Inapplicable To The Actual Manufacture/Post-Production P/Z
Components” and that “The Sequence [of IBM’s testing and manufacture process] Does
Not Satisfy The Court’s Claim Construction Ruling” (Doc. #261, at 29, 32.) In its reply
brief, IBM puts a finer point on these arguments, stating that “Auburn’s infringement
theory fails because the numerical reliability ‘estimates’ it relies upon (from qualification)
do not apply to the ‘classification’ it relies upon (resulting from a ‘repair count threshold’
applied during actual manufacture/post-production testing.” (Doc. #300, at 11.) The facts
of IBM’s processes necessitate the conclusion that the statistically calculated numerical
value of a component, derived from the qualification FIT targets, is not “based on” the
classification, as instructed by ‘366.
Auburn, however, argues that
[a]pplying the proper constructions, IBM does exactly what the
Auburn patents contemplate and what the Court allowed them
to cover: IBM first estimates the reliability of the classification
of “good” components (using its reliability model during
qualification) and then estimates the reliability of each “good”
component as it is classified (by the repair count screen during
manufacturing).
(Doc. #276 at 19-20 (citing IBM’s Motion for Summary Judgment, Doc. #261, at 24).)
Auburn’s argument, however, misconstrues the interaction between IBM’s qualification
and actual manufacture/post-production testing processes.
25

In determining exactly what Auburn’s patents cover, it is useful to consider the
temporal limitations ascribed to Auburn’s patents by the CCO. These temporal
limitations make clear what processes are covered by ‘366. In construing the
“estimating” term of ‘306,9 the Court noted that:
[n]othing in the claim language itself dictates when the
probability of latent defects for the classification must be done.
If the probability of latent defects for the classification is
estimated first, then the probability of latent defects for the
component in that classification is estimated simultaneously
with the act of classifying. If the probability of latent defects for
the classification is not estimated first, then the probability of
latent defects for the component must logically be done after the
act of classifying. In other words . . . the appropriate temporal
limitation for this term claim is “either simultaneously with or
after the ‘classifying step.’”
(Doc. #193 at 31-32 (emphasis original).)
Given this temporal limitation, ‘366 would cover a process wherein statistical
values for a given number of classifications, or “bins” (Expert Report of Dr. F. Joel
Ferguson, Doc. #161-17, at 9) were calculated prior to the sorting of individual
components into those classifications, and where the pre-determined statistical value for
those classifications was then applied, at the classification stage, to the components in
those bins. This is the process the Court contemplated in construing the “estimating” and
“optimizing” terms of ‘366 to allow for “calculation of a numerical value of the

9

The Court later concluded that “[u]sing the same reasoning as discussed with the claims of the
‘306 Patent, the Court finds that the ‘estimating’ and ‘optimizing’ terms must logically have the given
temporal limitation.” (Doc. #193, at 47 n.26)

26

component based on the classification “simultaneously with . . . the ‘classifying’ step.”
(Doc. #193, at 46.)
The ‘366 patent would also cover a process wherein components were binned
based on a numerical calculation of reliability for a given number of classifications where
said calculation was done “after the ‘classifying step.’” As contemplated by the CCO,
such a process would involve binning the individual components, at the classification
stage, into two or more bins based on relative levels of reliability—at this point the
classification would have to be based on relative reliability, as no numerical estimates of
reliability had yet been calculated—and then statistically calculating a numerical value of
reliability for each of those bins (i.e. each classification). This in turn would produce a
numerical value of reliability—a value based on the classification—for the components in
each bin. This is what the Court contemplated in construing “a plurality of reliability
classifications” to mean “at least two classifications of components associated with
respective levels of expected reliability.” (Doc. #193, at 46.)
Based on this analysis, and for the reasons stated above, supra Part III.A.1, IBM
was incorrect in arguing that, because it did not calculate a numerical value of reliability
for its “bad” components, it was not performing the “classification” step of ‘366. As the
Court has made clear, one can classify components using two bins based on relative
levels of reliability and still be practicing Auburn’s patent, so long as one later calculates
a numerical value of the components “based on the classification.”

27

However, IBM does not perform the estimating/predicting step “based on” its
classification of components as “good” and “bad.” In its actual manufacture/postproduction testing process, IBM eventually calculates a numerical value of reliability for
its “good” components, but, significantly, IBM does not calculate the numerical
reliability of its “good” components by calculating a numerical value of reliability for the
“good” classification. After classification, and after additional stress-testing and burnin,10 a numerical value of reliability is extrapolated from the results of the pre-production
qualification process and the results of the stress test/burn-in phase and applied to the
otherwise customer-ready components.11 Because this numerical value of reliability is

10

IBM argues that “[b]urn-in is not . . . merely one step among many that can be taken into
account in an estimate of reliability. Burn-in fundamentally alters the probability of latent defects present
in the die, such that an estimation that applies after burn-in, such as the FIT targets produced during . . .
qualification . . . would be inapplicable to a classification that takes place before burn-in.” (Doc. #300 at
12 (emphasis original).) Because the Court finds that IBM’s statistical calculation of a numerical value of
reliability is not “based on” IBM’s classification process, the Court need not determine whether additional
burn-in testing would vitiate the correlation between the FIT target produced during qualification and the
application of that number to a finished product.
11

At the conclusion of the actual manufacture/post production process, when IBM’s P/Z
components are otherwise customer-ready, IBM extrapolates the results of the qualification process and
applies the FIT targets derived from that process to its finished components. (Expert Report of Dr. F. Joel
Ferguson, Doc. #261-17 at 50).) Despite IBM’s contention that “Auburn has not identified, and cannot
identify, any statistical calculation of reliability that takes place ‘simultaneously with or after’ the alleged
‘classifying’ step,” (Doc. #261, at 32,) the assignment of a numerical value of reliability to a finished
component based on the results of the qualification process constitutes a statistical calculation of
reliability for those components, and this calculation occurs after the classification stage, as provided for
by the CCO (Doc. #193, at 46; Deposition of Dr. Richard Fair, Doc. #276-3; Deposition of Phillip Nigh,
Doc. #276-8.) This statistical calculation may result in further “binning”—i.e. an additional

“classification”—of the customer-ready die, but this later classification would not infringe
Auburn’s patent, as it would not be “based on” the number of defects identified in the initial
post-production test. Rather, this latter classification would be based on the results of IBM’s
qualification process combined with the results of IBM’s post-production testing.
28

not “based on” the classification of “good” derived from the initial post-production test,
Auburn’s patent does not cover this activity.
While classification as “good” is a necessary precondition for eventual application
of the FIT targets from qualification (Expert Report of Dr. F. Joel Ferguson, Doc. #26117, at 48, 178) and is part of the process through which FIT targets are derived (Id. at 5259) IBM’s estimate/prediction of reliability is not “based on” the “good” classification, as
it would be if a numerical value was calculated for the entire “good” classification and

29

that number was applied to all components classified as “good.”12 Put as simply as
possible, Auburn’s patented method does not cover IBM’s process.
If IBM calculated a numerical value of reliability for both its “good” and “bad”
classifications, at the classification stage, this would infringe Auburn’s patent. It is likely
that even if IBM calculated a numerical value of reliability for the “good” classification
alone, and that value were in turn assigned to individual “good” components, that too
would infringe Auburn’s patent. However, IBM does not perform either of these

12

Auburn makes two related arguments based on the language of its patent:
First, Auburn urges that IBM would have the Court read “comprising” out of ‘366’s preamble
(Doc. #276, at 21.) Auburn argues that, because comprising is an open-ended term, IBM is wrong to read
‘366’s claims to “exclude other types of testing.” (Doc. #276, at 21 (citing CIAS, Inc. v. Alliance Gaming
Corp., 504 F.3d 1356, 1360 (Fed. Cir. 2007).) It is true that some of IBM’s noninfringement arguments
rely on IBM’s performance of additional testing—particularly additional burn-in testing—between
classification and application of the FIT target to IBM’s finished product (IBM’s Reply, Doc. #300, at 78; see also note 10 and accompanying text).) However, the Court does not rely on the additional testing
performed by IBM in reaching its conclusion that the P/Z products do not infringe ‘366. If the alleged
infringer does not perform, literally or under the doctrine of equivalents, all of the claimed steps of the
method there can be no infringement, regardless of what additional steps are added to the patented
method. Techsearch, L.L.C. v. Intel Corp., 286 F.3d 1360, 1372 (Fed. Cir. 2002). Even if IBM
performed no additional testing, its estimate/prediction of a numerical value of reliability would not be
“based on” the classification, because IBM does not perform all of the claimed steps.
Second, Auburn contends that “by excluding the possibility of other testing, IBM would have the
Court ignore the plain and ordinary meaning of ‘based on’ by having it read that phrase to mean ‘solely
based on.’” (Doc. #271 at 22.) The Court, however, does not reach its holding that IBM’s qualification
numbers are inapplicable to the components at the classification stage (i.e. that the calculation is not
“based on” the classification) as a result of reading “based on” to mean “solely based on.” Rather, the
Court’s holding derives from the multi-factor process IBM uses to statistically calculate the numerical
reliability value of its finished components and the most logical understanding of the term “based on” in
this context.
While it is necessary that a given component be classified as “good” as a precondition to the
eventual application of the qualification-derived reliability targets to that component, that X is a necessary
precondition for Y does not necessarily mean that Y is “based on” X. Take, for example, an aspiring law
student. We’ll call him Barry. As are all aspiring law students, Barry was born. Barry later took the
LSAT in preparation for his application to law school. Barry was given a score on his LSAT, a numerical
calculation of his aptitude for the study of law. Had Barry not been born, Barry would not have received
that score on his LSAT. No one would argue, however, that Barry’s LSAT score was “based on” his
birth. For the same reason, the Court finds that IBM’s statistical calculation of a numerical value of
reliability for a component is not “based on” the classification of that component.

30

calculations at the classification stage; rather, the “good” and “bad” classifications are
based solely on the fact that “good” components had fewer defects than the cut-point
threshold, and the “bad” components had more.13 Applying one FIT target to all “good”
components, without further differentiation amongst the “good” components would serve
no purpose. IBM further differentiates the “good” components through the portion of
IBM’s actual manufacture/post-production testing process that takes place after
classification. Therefore, the final application of the FIT targets to the components that
have completed this process is not “based on” the classification.
Because the Court has concluded that IBM does not infringe Auburn’s ‘366 patent
because it does not statistically calculate a numerical estimate of reliability “based on” the
classification, the Court need not address IBM’s contention that it does not infringe
because Auburn is improperly reading the patent to cover methods already being
performed in the prior art.

13

At the classification stage, IBM’s components are divided into “good” and “bad” classifications
according to a comparison of the number of defects in a given component to the number of defects that
component’s circuitry can accommodate. This is a pre-determined threshold value unrelated to the
statistically calculated FIT target arrived at through the qualification process: “If the number of repairable
defects exceeds the number of available redundancies, then IBM’s tester machine software assigns a ‘sort
code.’ IBM will scrap the die if the number of repairs exceeds a predefined numerical threshold which
IBM sometimes refers to as a ‘cut point.’” (Ferguson Report, Doc. #261-17, at 43 (emphasis added); see
also Deposition of James M. Crafts, Doc. #276-4 at 4 (Q: “if you find more . . . bad elements than
redundant elements . . . the part can’t be repaired, correct?” A: “That’s right.”) In short, if a component
has more defects than can be repaired through the use of redundant circuitry, that component is classified
as “bad” and is scrapped. Otherwise, the component is classified as “good” and is allowed to continue
through the rest of the manufacture and testing process. Therefore, only chips that are classified as good
and survive the remainder of the testing process are ever assigned a statistically calculated numerical
value of reliability.

31

B. IBM does not literally infringe the ‘366 patent with regard to the
Trimaran/Corona products
IBM makes four arguments, almost identical to those it makes regarding the P/Z
products, as to why it does not infringe with regard to the Trimaran/Corona products.
IBM argues that:
(1) IBM does not “classify” the Trimaran/Corona products into
a “plurality of reliability probability classifications” when it
applies a “repair count threshold limit (IBM’s Motion, Doc.
#261, at 38.)
(2) The numerical reliability targets from IBM’s qualification
process relied upon by Auburn are inapplicable to the postproduction Trimaran/Corona components that Auburn alleges
are “classified” into “reliability probability classifications.”
(Doc. #261, at 40.)
(3) Auburn’s infringement theory for the Trimaran/Corona
products fails as a matter of law because IBM’s sequence does
not satisfy the Court’s claim construction ruling (Doc. #261, at
41.)
(4) Auburn is improperly reading the ‘366 patent claims to cover
methods for testing the Trimaran/Corona products that the Court
has ruled were “already being performed” in the prior art. (Doc.
#261 at 42.)
Considering IBM’s second and third arguments, above, and applying the same
reasoning as it did to the P/Z products, the Court holds that IBM does not infringe with
regard to the Trimaran/Corona products. Supra Part III.A. In short, the Court concludes
that, while IBM does classify the Trimaran/Corona components as instructed by ‘366 and
the CCO, IBM does not statistically calculate a numerical value of the reliability based on
that classification. As discussed above, supra I.D.2, the slight variation between the
32

testing procedure for the P/Z products and the Trimaran/Corona products does not
warrant a different holding for the two sets of accused products. The Court will address
in detail only the most potentially significant difference between the testing of the
Trimaran/Corona products and the P/Z products.
IBM classifies or has classified the Trimaran/Corona products differently than the
P/Z products. At different points in time, IBM (1) classified Trimaran/Corona
components into “good” and “bad” classifications and scrapped the bad chips, (2)
differentiated the amount of subsequent burn-in testing certain components will receive
based on the number of defects revealed by the repair-count screen, and (3) directed
components to “lower or higher reliability applications” based on the number of
repairable defects. (Deposition of Dr. F. Joel Ferguson, Doc. #261-6, at 22-28; Expert
Report of Dr. F. Joel Ferguson, Doc. #261-17, at 62-67.)
Based on these classification processes, the Court, as it did with regard to the P/Z
products, finds that IBM does classify its components into at least two classifications of
components associated with respective levels of expected reliability. Supra, Part III.A.1.
However, all of these classifications stem from the number of defects revealed by the
repair-count screen, and are based on relative levels of reliability. (Deposition of Dr. F.
Joel Ferguson, Doc. #261-6, at 22-28; Expert Report of Dr. F. Joel Ferguson, Doc. #26117, at 62-67; see, supra, note 13) IBM does not predict or estimate a numerical value of
reliability “based on” these classifications.

33

Auburn alleges that, in addition to using the classifying scheme set out above,
“IBM also statistically calculates a numerical value of the reliability of each grade
using the same reliability model discussed . . . with respect to the P/Z products.” (Doc.
#276, at 32 (emphasis original).) The Court finds that the application of IBM’s statistical
model to these different classifications of components does not change the infringement
analysis in Part III.A of this opinion. Therefore, for the same reasons as discussed above
with regard to the P/Z products, IBM must be held not to infringe with regard to the
Trimaran/Corona products.
C. IBM does not infringe the ‘366 patent with regard to the Dreadnaught/DDP2230
products
IBM makes four arguments as to why it does not infringe with regard to the
Dreadnaught/DDP2230 products:
(1) There is no evidence that any “nearest neighbor” test was
used to “classify” die during the qualification process. (IBM’s
Motion, Doc. #261, at 47.)
(2) Auburn’s infringement theory for the
Dreadnaught/DDP2230 products fails as a matter of law
because IBM’s sequence does not satisfy the Court’s claim
construction ruling. (Doc. #261, at 48.)
(3) The numerical reliability targets from IBM’s qualification
process relied upon by Auburn are inapplicable to the postproduction Dreadnaught/DDP2230 components that Auburn
alleges are classified in the “classifying step. (Doc. #261, at 49.)
(4) Auburn is improperly reading the ‘366 patent claims to cover
methods for testing the Dreadnaught/DDP2230 products that the
Court has ruled were “already being performed” in the prior art.
(Doc. #261 at 51.)
34

Considering IBM’s second and third arguments, above, and applying the same
reasoning as it did to the P/Z products and the Trimaran/Corona products, the Court holds
that IBM does not infringe with regard to the Dreadnaught/DDP2230 products.
Supra Part III.A. In short, the Court concludes that, while IBM does classify the
Dreadnaught/DDP2230 components using the results of a nearest-neighbor test performed
on those products, as instructed by ‘306 and the CCO, IBM does not statistically calculate
a numerical value of the reliability based on the resulting “good” and “bad”
classifications. As discussed above, supra, the slight variations between the testing
procedures’ various accused products does not warrant a different holding for the two sets
of accused products.
D. IBM does not infringe either the ‘366 patent or the ‘306 patent under the
doctrine of equivalents
Equivalence is a question of fact. Graver Tank and Mfg. Co. v. Linde Air Prods.
Co., 339 U.S. 605, 609 (1950). While neither party makes arguments under the doctrine
of equivalents, the Court's analysis on the issue of literal infringement, supra Part III.A-C,
make clear that no reasonable juror could find that IBM infringes under the doctrine of
equivalents. Therefore, summary judgment in favor of IBM on Auburn's doctrine of
equivalent claim is appropriate. Ethicon Endo-Surgery, Inc. v. U.S. Surgical Corp., 149
F.3d 1309, 1318 (Fed. Cir. 1998) (collecting cases where Federal Circuit held that there
could be no infringement under the doctrine of equivalents because no reasonable factfinder could have found equivalence).
35

There are two principal legal limitations on the doctrine of equivalents, and these
two limitations “are to be determined by the court, either on a dispositive pretrial motion
or on a motion for judgment as a matter of law at the close of evidence and after the jury
verdict.” Warner-Jenkinson Co., Inc. v. Hilton Davis Chem. Co., 520 U.S. 17, 39 n.8
(1997). The Court must determine whether the patentee surrendered equivalent subject
matter from the claimed invention during patent prosecution. If so, the patentee is
precluded from recapturing any of that subject matter through a doctrine of equivalents
theory of infringement. Id. at 30. A court may also rule on a doctrine of equivalents
theory by performing an element-by-element correspondence between the accused
product or method and the patent claims to determine if the differences are only
insubstantial. Id. at 29, 40. This is also known as the “all-elements rule.” Kustom
Signals, Inc. v. Applied Concepts, Inc., 264 F.3d 1326, 1333 (Fed. Cir. 2001). Here, the
Court finds that Auburn's doctrine of equivalents theory fails under the all-elements rule.
“The all-elements rule is that an accused device must contain every claimed
element of the invention or the equivalent of every claimed element. . . . No claimed
element, or an equivalent thereof, can be absent if the doctrine of equivalents is invoked.”
Id. In other words, “[t]he doctrine of equivalents does not apply where a claim limitation
is completely missing, and thus not met equivalently, from the accused device.” Arndt v.
Mokai Mfg., Inc., No. Civ. 03-1240-HA, 2006 WL 758539, at *5 (D. Or Mar. 2, 2006).
Here, the Court has already concluded that the patented method requires the
performance of three specific steps, and that IBM does not perform all of those steps.
36

Specifically, IBM does not estimate or predict a numerical value of the reliability of the
component based on the classification. Therefore, IBM cannot be found to infringe
Auburn's patents under the doctrine of equivalents.
For the reasons stated above, it is hereby ORDERED that
(1) IBM's Motion for Summary Judgment of Noninfringement (Doc. #261) is
GRANTED and the claims contained in Counts III and IV of Auburn’s First
Amended Complaint (Doc. #87) are DISMISSED WITH PREJUDICE.14
(2) IBM’s Motion for Summary Judgment of No Willful Infringement (Doc.
#257) is DENIED as MOOT.
(3) Auburn’s Motion for Summary Judgment on IBM’s 35 U.S.C. Section 273
Affirmative Defense (Doc. #258) is DENIED as MOOT.
(4) The Court RESERVES RULING on Auburn’s Motion for Partial Summary
Judgment of No Invalidity of Claims (Doc. #260).
DONE this the 2nd day of August, 2012.
/s/ Mark E. Fuller
UNITED STATES DISTRICT JUDGE

14

The Court acknowledges that, based on the facts of this case, whether or not IBM’s
processes infringe Auburn’s patents is a close question. The Court is confident in its opinion.
However, the Court believes that the technical expertise of the Federal Circuit may be useful in
distinguishing between the methods employed by IBM and Auburn. For this reason, the Court is
wiling to entertain a motion for an interlocutory appeal of this order.
37