UNITED STATES DISTRICT COURT NORTHERN DISTRICT OF CALIFORNIA SAN JOSE DIVISION
OPTIWAVE SEMICONDUCTOR CORP.,
Plaintiff,
v.
LUMOS PHOTONICS, INC.,
Defendant.
Case No. 5:22-cv-03417-EJD
Before the Honorable Edward J. Torvelis
PLAINTIFF OPTIWAVE SEMICONDUCTOR CORP.'S REPLY CLAIM CONSTRUCTION BRIEF
Filed: January 9, 2023
Alan Ngo (USPTO Reg. No. 52,871), Responsible Partner Marcus Delgado, Partner Omar Siddiqui, Senior Associate Christy Bauer, Associate
CALDERWOOD & HARKNESS LLP One Embarcadero Center, 34th Floor San Francisco, CA 94111 Telephone: (415) 555-7200 Facsimile: (415) 555-7201
Attorneys for Plaintiff Optiwave Semiconductor Corp.
TABLE OF CONTENTS
TABLE OF AUTHORITIES
Cases:
Statutes:
Other Authorities:
I. INTRODUCTION
Plaintiff Optiwave Semiconductor Corp. ("OWS" or "Plaintiff") respectfully submits this Reply Claim Construction Brief in response to the arguments raised in Defendant Lumos Photonics, Inc.'s ("Lumos" or "Defendant") Responsive Claim Construction Brief, filed December 19, 2022 (Dkt. No. 87) ("Resp. Br."), regarding the construction of eleven disputed claim terms across U.S. Patent Nos. 10,234,567 ("the '567 Patent"), 10,891,234 ("the '234 Patent"), 11,045,678 ("the '678 Patent"), and 11,312,890 ("the '890 Patent") (collectively, the "Asserted Patents").
As set forth in OWS's Opening Claim Construction Brief, filed December 5, 2022 (Dkt. No. 78) ("Opening Br."), and the accompanying Declaration of Dr. Harold Whitfield (Dkt. No. 78-1) ("Whitfield Decl."), the eleven disputed claim terms at issue in this proceeding have plain and ordinary meanings that a person of ordinary skill in the art ("POSITA") would readily understand. The Asserted Patents use these terms consistently with their established meanings in the silicon photonics field, and the intrinsic record — the claim language, the specifications, and the prosecution histories — fully supports OWS's proposed constructions.
Lumos's Responsive Brief does not alter this fundamental landscape. Rather than grappling with the intrinsic evidence on its own terms, Lumos pursues a twofold strategy that is both transparent and impermissible under governing Federal Circuit law.
First, Lumos improperly seeks to import process limitations and structural restrictions into claim terms that recite functional results or general structural relationships. Across each of the four Asserted Patents, Lumos's proposed constructions share a common deficiency: they attempt to rewrite the patent claims by grafting onto the claim language specific implementation details drawn from the preferred embodiments disclosed in the specifications. The Federal Circuit has "repeatedly warned against confining the claims to those embodiments." Ashford Enterprises v. Dunmore Corporation, 358 F.3d 898, 906 (Fed. Cir. 2004). Yet this is precisely what Lumos asks the Court to do. Whether it is the demand that "adiabatically coupled" require a "gradual, continuous taper," or the insistence that "co-integrated... on a common substrate" exclude all "bond wires, interposers, or package-level routing," Lumos's constructions would narrow the claims far beyond what the intrinsic record supports, effectively excluding any technology that does not mirror the specific implementation details of a single disclosed embodiment.
Second, Lumos relies heavily on prosecution history estoppel arguments for the '567 and '678 Patents that overextend the scope of amendments made during prosecution, as well as on the extrinsic testimony of Dr. Raymond Ng (Of Counsel / Technical Consultant, Brannick Howell & Stein LLP; formerly a professor of photonics at the University of Colorado Boulder) to supply narrow definitions that find no support in the claims, specification, or prosecution history of the Asserted Patents. Under the Supreme Court's framework in Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005) (en banc), the intrinsic evidence is the primary and most reliable guide to claim meaning. Extrinsic evidence, including expert testimony, "may only be used to help the court come to a proper understanding of the claims; it may not be used to vary or contradict the claim language." Id. at 1318. Dr. Ng's testimony, as explained more fully below, repeatedly does exactly that — it substitutes idealized academic definitions for the practical, claim-specific meanings established by the intrinsic record.
OWS recognizes that Dr. Ng holds relevant academic credentials. But his testimony is not anchored in the intrinsic record of these patents. Instead, it reflects an effort to impose the narrowest possible reading of each disputed term by cherry-picking definitions from theoretical textbooks while disregarding the actual language of the patent claims and specifications. This approach is fundamentally at odds with the Phillips framework.
In contrast, the Declaration of Dr. Harold Whitfield (Professor Emeritus, UC Santa Barbara; Whitfield Technical Consulting Group, Inc.) — submitted with OWS's Opening Brief — provides the Court with a reliable guide to the understanding of a POSITA working in the silicon photonics field. Dr. Whitfield's opinions, grounded in over 200 publications in integrated photonics and decades of practical silicon photonics design, are fully consistent with the intrinsic evidence and demonstrate that OWS's proposed constructions faithfully capture the plain and ordinary meaning of the disputed terms.
For the reasons set forth below, and as more fully explained in OWS's Opening Brief, OWS respectfully submits that the Court should adopt OWS's proposed constructions for all eleven disputed claim terms. The Court should reject Lumos's efforts to rewrite the claims through impermissible importation of preferred-embodiment limitations, overbroad prosecution history estoppel arguments, and reliance on extrinsic evidence that contradicts the intrinsic record.
II. LEGAL STANDARD FOR CLAIM CONSTRUCTION
OWS's Opening Brief set forth the governing legal standards in detail. See Opening Br. at 4–8. OWS briefly reiterates the key principles here, as they are directly implicated by the arguments raised in Lumos's Responsive Brief.
Claim construction is a matter of law for the Court to decide. Markman v. Westview Instruments, Inc., 517 U.S. 370, 388–91 (1996); Teva Pharm. USA, Inc. v. Sandoz, Inc., 574 U.S. 318, 325 (2015). The purpose of claim construction is to determine the meaning and scope of the patent claims as understood by a POSITA at the time of the invention. Phillips, 415 F.3d at 1313.
The intrinsic evidence — comprising the claim language, the specification, and the prosecution history — is "the most significant source of the legally operative meaning of disputed claim language." Id. at 1314. Among these sources, the claims themselves provide substantial guidance: the words of a claim are generally given their ordinary and customary meaning as understood by a POSITA in the context of the entire patent, including the specification. Id. at 1312–13. The specification is "the single best guide to the meaning of a disputed term" and "acts as a dictionary when it expressly defines terms used in the claims or when it defines terms by implication." Id. at 1315, 1321. The prosecution history, while "often of less utility than the specification," nonetheless "can often inform the meaning of the claim language by demonstrating how the inventor understood the invention and whether the inventor limited the invention in the course of prosecution." Id. at 1317.
Extrinsic evidence — including expert testimony, inventor testimony, dictionaries, and learned treatises — may aid the Court's understanding of the technology and the meaning of terms in the relevant art. Id. at 1317–18. However, the Federal Circuit has cautioned that extrinsic evidence is "less significant than the intrinsic record in determining the legally operative meaning of claim language" and that courts must be careful not to permit extrinsic evidence to "contradict the meaning of claims in light of the specification and prosecution history." Ridgeway & Co. v. Conceptronic, Inc., 90 F.3d 1576, 1584 (Fed. Cir. 1996); see also Phillips, 415 F.3d at 1318–19 (describing the risks of "heavy reliance on the testimony of experts" and noting that such reliance "may be unreliable" because experts may be "biased toward a particular reading of the claim" and because "expert testimony is generated at the time of and for the purpose of litigation").
Two additional principles, both repeatedly invoked by OWS in this briefing and directly applicable to Lumos's arguments, warrant emphasis:
First, a court must not import limitations from the specification's preferred embodiments into the claims. "[I]t is the claims that define the metes and bounds of the patent grant, not particular embodiments disclosed in the specification." Liebel-Flarsheim, 358 F.3d at 906. Even where the specification describes only a single embodiment, the Federal Circuit has held that "it does not follow that the claims must be limited to that disclosed embodiment." Id. The distinction between using the specification to illuminate the meaning of claim terms and improperly importing limitations from specific embodiments is critical to proper claim construction, and it is a distinction that Lumos's Responsive Brief fails to respect.
Second, prosecution history estoppel limits, but does not extinguish, claim scope. Under Festo Corp. v. Fenmore & Co., 535 U.S. 722 (2002), when a patentee narrows a claim by amendment during prosecution to overcome prior art, the patentee is estopped from later asserting that the surrendered subject matter falls within the claim scope. However, the scope of the estoppel is limited to the territory actually surrendered; it does not permit a defendant to broaden the surrender beyond what the prosecution record demonstrates. Id. at 740–41. The three Festo rebuttal exceptions — unforeseeability, tangential relation, and other reasons suggesting the patentee could not reasonably have been expected to draft a claim covering the particular equivalent at issue — further limit the application of estoppel. Id. at 740–41.
With these principles in mind, OWS addresses Lumos's specific arguments below.
III. REPLY ARGUMENTS ON DISPUTED TERMS OF THE '567 PATENT
A. "Adiabatically Coupled" (Claim 1, and Dependent Claims 4, 7, 12, 15)
The parties' competing constructions of the term "adiabatically coupled" are as follows:
Lumos's Responsive Brief advances several arguments in support of its unduly narrow construction. Each of these arguments fails for the reasons set forth below.
1. Plain and Ordinary Meaning Supports OWS's Construction. The term "adiabatic" in the context of photonic waveguide coupling is a well-understood and well-established term of art. As OWS demonstrated in its Opening Brief, and as Dr. Whitfield's declaration confirms, a POSITA working in the silicon photonics field would understand "adiabatically coupled" as a functional description of the coupling efficiency — specifically, that the coupling is characterized by optical mode transitions occurring without significant power loss or unintended excitation of higher-order modes. See Whitfield Decl. ¶¶ 24–38; Opening Br. at 10–14.
The term does not, as a matter of its plain and ordinary meaning, impose a structural mandate requiring a specific taper geometry. A POSITA would understand that "adiabatic" coupling can be achieved through range physical structures, including but not limited to gradual continuous tapers, segmented tapers, curved waveguide transitions, and other geometric approaches that achieve sufficiently slow mode evolution to avoid significant mode coupling or power loss. The key characteristic is the functional result — low-loss, mode-preserving coupling — not any particular physical structure.
Lumos's proposed construction — requiring "a gradual, continuous taper that prevents excitation of higher-order modes throughout the entire transition region" — does precisely what the Federal Circuit has prohibited: it imports the structural features of a preferred embodiment into the claim term itself, transforming a functional description into a structural requirement. See Thorner v. Sony Computer Entm't Am. LLC, 669 F.3d 1362, 1366 (Fed. Cir. 2012) ("A patentee may choose to be his or her own lexicographer by providing a definition of the term in the specification with reasonable clarity, deliberateness, and precision," but "absent such a clear definitional statement, the claim term will be given its ordinary meaning as understood by a person of skill in the art."); Liebel-Flarsheim, 358 F.3d at 906.
The '567 Patent specification uses "adiabatically coupled" to describe the functional result of the coupling, not a specific physical structure. The specification states that "the first waveguide layer is adiabatically coupled to the second waveguide layer such that optical energy is transferred with minimal insertion loss." '567 Patent, col. 4, ll. 15–32. This language describes a result — minimal insertion loss — not a particular taper shape. Nowhere in the claims or the specification does the patentee define "adiabatically coupled" as requiring a "gradual, continuous taper," let alone one that "prevents excitation of higher-order modes throughout the entire transition region." Lumos's construction adds structural specificity that is entirely absent from the claim language and the specification's definition of the term.
Dr. Whitfield's declaration provides extensive support for this reading. As Dr. Whitfield explains, the concept of "adiabatic" coupling in photonics derives from the principle of adiabatic invariance in physics, which describes systems that evolve slowly enough that the system remains in an instantaneous eigenstate throughout the transition. Whitfield Decl. ¶ 25. In the photonics context, this translates to a waveguide coupling in which the optical mode is transformed between waveguides without significant power being transferred to unwanted higher-order or radiation modes. Id. ¶ 26. Critically, Dr. Whitfield explains that this principle is not limited to any single structural implementation; rather, it describes a regime of operation that can be achieved through various geometric designs, so long as the transition is sufficiently gradual relative to the beat length between the relevant modes. Id. ¶¶ 27–32.
OWS's construction — "coupled such that optical mode transitions occur without significant power loss" — faithfully captures this well-understood functional meaning without imposing structural limitations that the claim language does not recite.
2. The Specification Does Not Require a "Continuous" Taper. A central pillar of Lumos's argument is the assertion that the specification of the '567 Patent exclusively describes a continuous taper, and therefore "adiabatically coupled" must be limited to coupling achieved through such a taper. See Resp. Br. at 12–14. This argument is factually incorrect and legally insufficient.
As an initial matter, even if the specification described only a continuous taper — which it does not — that fact alone would not justify importing such a limitation into the claims. The Federal Circuit has "repeatedly warned against limiting the claimed invention to preferred embodiments or specific examples in the specification." Liebel-Flarsheim, 358 F.3d at 906. The claims, not the embodiments, define the scope of the patent grant. A single disclosed embodiment does not confine the claims to that embodiment's specific features.
But the premise of Lumos's argument is wrong in any event. The specification of the '567 Patent does not exclusively describe a continuous taper. To the contrary, the specification expressly teaches that the taper "may be implemented through various geometries that achieve adiabatic mode transition." '567 Patent, col. 5, ll. 7–12. This language is broad and inclusive, not narrow and restrictive. It affirmatively contemplates that structures other than a single continuous taper may fall within the scope of the claimed "adiabatic" coupling.
While the preferred embodiment illustrated in Figure 3A depicts a continuous linear taper, this is one example of the various geometries the specification contemplates. The specification further explains that "[t]he transition length and taper profile are selected to ensure that the coupling remains substantially adiabatic, i.e., that insertion loss is maintained below an acceptable threshold for the intended application." '567 Patent, col. 5, ll. 18–24. This description is again functional and performance-oriented, not limited to a particular geometric profile.
Lumos seizes on the word "continuous" in the specification's description of the preferred embodiment and attempts to elevate it to a claim limitation. But the specification's use of "continuous" in describing an embodiment is not the same as the claim's use of "adiabatically coupled" in defining the scope of the invention. The former describes a specific design choice; the latter describes a functional characteristic of the coupling. Lumos conflates the two, and in doing so, violates the cardinal rule against importing limitations from the specification into the claims.
Dr. Ng's responsive declaration (Dkt. No. 87-2) ("Ng Resp. Decl.") compounds this error. Dr. Ng selectively cites theoretical textbook definitions of "adiabatic transition" that describe idealized continuous transitions. See Ng Resp. Decl. ¶¶ 18–25. However, Dr. Whitfield persuasively explains that in practice, the silicon photonics community recognizes that adiabatic behavior can be achieved through multiple structural approaches, including segmented designs that approximate adiabatic behavior over their transition lengths. Whitfield Decl. ¶¶ 33–37. As Dr. Whitfield observes, a segmented taper can be designed such that the effective mode index varies sufficiently slowly along the direction of propagation to maintain adiabatic operation, even though the physical structure is not a smooth, continuous taper at the microscopic level. Id. ¶ 34. Dr. Whitfield further notes that a number of published works in the field — including papers by researchers at institutions such as MIT, IMEC, and Larkspur Group — have demonstrated segmented waveguide structures operating in the adiabatic regime. Id. ¶ 36.
The fact that Dr. Ng can point to textbook passages defining "adiabatic" in terms of continuous, slowly-varying transitions does not establish that the claim term "adiabatically coupled" in the '567 Patent is limited to a single structural implementation. Textbooks describe idealized physical principles; patent claims describe inventions in terms that a POSITA would understand in the context of the patent's intrinsic record. The two are not identical, and the latter must control. See Phillips, 415 F.3d at 1313 (claim terms are construed in context of the patent, not in the abstract).
3. Prosecution History Estoppel Does Not Support Lumos's Narrow Construction. Lumos's estoppel argument constitutes a significant portion of its Responsive Brief's analysis of the "adiabatically coupled" term. See Resp. Br. at 15–20. Lumos contends that during prosecution of the '567 Patent, the claims underwent two office action rejections, and that the "adiabatically coupled" limitation was added to overcome the prior art of Tanaka et al. (U.S. Patent No. 9,456,012, issued September 6, 2016). From this premise, Lumos argues that the amendment narrows the claim scope to require only continuous, gradual tapers that are structurally distinct from the Tanaka reference. This argument is unsupported by the prosecution record and misapplies the law of prosecution history estoppel.
The prosecution history of the '567 Patent is straightforward on this point. The original claims recited a "coupled" waveguide structure without specifying the nature of the coupling. The examiner rejected those claims under 35 U.S.C. § 102(a)(1), citing Tanaka et al. (U.S. Patent No. 9,456,012) as prior art disclosing a coupled waveguide arrangement. The Tanaka reference describes a butt-coupled waveguide architecture — that is, a direct end-to-end junction between two waveguides without any tapering or transitional coupling structure. See Tanaka et al., col. 3, ll. 5–22.
In response, the applicant amended Claim 1 to specify that the waveguides are "adiabatically coupled" and submitted remarks distinguishing the claimed invention over the Tanaka reference. The applicant's remarks during prosecution stated:
"Unlike Tanaka, the claimed coupler provides an adiabatic transition that avoids the mode mismatch and insertion loss inherent in butt-coupled designs. The adiabatic coupling of the present claims ensures that optical mode transitions occur without the significant power loss associated with the abrupt interface of a butt-coupling architecture."
See '567 Patent Prosecution History, Response dated May 14, 2019, at 8–9.
These remarks make clear the nature and scope of the applicant's distinction over the Tanaka prior art. The applicant distinguished the claimed invention based on the functional characteristic of the coupling — adiabatic (low-loss, mode-preserving) coupling — as contrasted with the non-adiabatic butt-coupling architecture of Tanaka. The applicant did not identify, discuss, or rely upon any specific taper geometry to distinguish over Tanaka. The distinction was functional, not structural.
Under Festo, the scope of prosecution history estoppel is limited to the subject matter actually surrendered by the patentee's narrowing amendment or argument. 535 U.S. at 740. The patentee "may not recapture through equivalents what was surrendered during prosecution," but the estoppel extends only as far as the surrender. Id. Here, the surrender is clear and circumscribed: the applicant disclaimed butt-coupling — the specific architecture of the Tanaka prior art — by adding the requirement of adiabatic coupling. The applicant did not disclaim segmented tapers, non-linear tapers, or any other specific taper geometry. Lumos's effort to expand the estoppel from a disclaimer of butt-coupling to a requirement of "gradual, continuous" taper geometry is unsupported by the prosecution record and exceeds the permissible scope of the estoppel.
Lumos appears to suggest that because the applicant used the word "adiabatic" to distinguish over Tanaka's butt-coupling, the applicant necessarily committed to a narrow structural definition of "adiabatic" that encompasses only continuous tapers. This is a non sequitur. The applicant used "adiabatic" in its functional sense — to distinguish adiabatic (low-loss) coupling from non-adiabatic (high-loss) butt-coupling. This functional distinction is consistent with OWS's proposed construction and does not import any structural limitations.
OWS also notes — though it does not bear on the estoppel analysis — the ironic fact that Hideo Tanaka, a named inventor on the '567 Patent, is also the co-inventor of the Tanaka prior art patent (U.S. Patent No. 9,456,012), which was filed before Dr. Tanaka joined OWS in 2016. Lumos's Responsive Brief suggests that this fact somehow supports its narrow construction, but the estoppel inquiry focuses on what the applicant surrendered during prosecution, not on the identity of the prior art inventors. See Festo, 535 U.S. at 740–41. The fact that the same inventor contributed to both the prior art and the patented invention does not alter the legal analysis of what was surrendered during prosecution of the '567 Patent.
In sum, OWS's construction preserves the full scope of the functional claim term "adiabatically coupled" while respecting the prosecution history distinction over butt-coupling. Lumos's construction goes far beyond the scope of the actual surrender by requiring structural specifics — "gradual, continuous" taper geometry — that were never part of the prosecution amendment or the applicant's distinguishing arguments. The prosecution history estoppel argument does not support Lumos's proposed construction.
4. Lumos's Over-Reliance on Dr. Ng's Extrinsic Testimony. Finally, Lumos devotes nearly five pages of its Responsive Brief to the testimony of Dr. Raymond Ng regarding the meaning of "adiabatic" in photonics literature. See Resp. Br. at 20–24. As noted above, Dr. Ng is Of Counsel / Technical Consultant at Brannick Howell & Stein LLP, counsel for Lumos, and a former professor of photonics at the University of Colorado Boulder. OWS does not dispute Dr. Ng's academic credentials. However, Dr. Ng's testimony on this point is fundamentally flawed for two reasons.
First, Dr. Ng selectively cites theoretical textbook definitions that describe idealized continuous transitions, while ignoring the broader practical usage of "adiabatic" in the silicon photonics community. As Dr. Whitfield explains, while textbooks may describe the theoretical ideal of a perfectly continuous, infinitely gradual adiabatic transition, practitioners in the silicon photonics field understand that "adiabatic" coupling describes a regime of operation that can be achieved through various structural approaches, including segmented designs. Whitfield Decl. ¶¶ 33–37. Dr. Ng's selective citation of theoretical sources does not reflect the understanding of a POSITA working in the practical silicon photonics field.
Second, and more fundamentally, extrinsic evidence — however credentialed the expert — cannot override the intrinsic record. Phillips, 415 F.3d at 1318. Dr. Ng's testimony does not address what the specification and prosecution history of the '567 Patent disclose; instead, it substitutes an academic idealization for the practical, claim-specific meaning established by the patent's intrinsic record. The specification uses "adiabatically coupled" as a functional description, the prosecution history uses it as a functional distinction over butt-coupling, and Dr. Ng's testimony cannot transform these functional usages into a structural requirement. Dr. Whitfield, whose experience includes over 200 publications in integrated photonics and decades of practical silicon photonics design, provides the more relevant and complete POSITA perspective. The Court should decline Lumos's invitation to elevate Dr. Ng's extrinsic testimony over the clear intrinsic record.
B. Other Disputed Terms of the '567 Patent
Lumos's Responsive Brief also addresses two additional disputed terms from the '567 Patent: (1) "tapered silicon waveguide having a tip width of less than 200 nanometers" (Claim 1), and (2) "silicon nitride waveguide layer" (Claims 1, 7, 15).
With respect to "tapered silicon waveguide having a tip width of less than 200 nanometers," the parties' positions have narrowed since the exchange of opening briefs. Lumos now appears to agree that the numerical limitation of "less than 200 nanometers" sets forth a definite structural boundary. The remaining dispute concerns whether the "tip width" must be measured at the narrowest point of the taper (OWS's position) or at the precise interface point between the silicon and silicon nitride waveguides (Lumos's position). As set forth in OWS's Opening Brief (at 16–18) and Dr. Whitfield's declaration (¶¶ 39–42), a POSITA would understand "tip width" to refer to the narrowest point of the taper, which is the conventional meaning of the term. The specification does not require measurement at any particular interface location, and Lumos's proposed measurement point is drawn from the preferred embodiment, not the claim language. OWS incorporates by reference the arguments made in its Opening Brief and submits that no further reply is needed on this term.
With respect to "silicon nitride waveguide layer," the parties' dispute is relatively narrow: Lumos contends that the term requires stoichiometric Si₃N₄, while OWS contends — and Dr. Whitfield confirms (Whitfield Decl. ¶¶ 43–44) — that a POSITA would understand "silicon nitride" in the silicon photonics context to encompass both stoichiometric and non-stoichiometric (silicon-rich) silicon nitride compositions commonly used in the fabrication of waveguide structures. The specification of the '567 Patent does not limit the term to stoichiometric compositions, and OWS's construction is consistent with industry usage. OWS incorporates by reference its Opening Brief arguments on this term.
IV. REPLY ARGUMENTS ON DISPUTED TERMS OF THE '234 PATENT
A. "Graded Buffer Layer Having a Continuously Varying Germanium Concentration" (Claim 1, and Dependent Claims 3, 5, 8, 11)
The parties' competing constructions of this term are as follows:
Lumos's proposed construction suffers from the same fundamental deficiency that afflicts its constructions of other terms: it imports multiple limitations from a preferred embodiment into the claim language, narrowing the claims far beyond what the intrinsic record supports. OWS addresses each of Lumos's arguments in turn.
1. The Claim Language Does Not Require "Monotonic" Increase or Exclude Step Changes. The starting point of claim construction is the claim language itself. Phillips, 415 F.3d at 1312. The relevant claim language is "graded buffer layer having a continuously varying germanium concentration." The claim uses the phrase "continuously varying" — not "monotonically increasing," not "smoothly graded," not "without step changes," and not "from the silicon substrate interface to the absorber layer." OWS's proposed construction — "a buffer layer in which the germanium concentration changes across the layer thickness" — faithfully tracks the plain claim language without adding any of the multiple limitations that Lumos seeks to graft onto the term.
Lumos's proposed construction adds no fewer than four distinct limitations that are absent from the claim language:
(1) "monotonically" — the claim does not require that the concentration increase monotonically. "Continuously varying" describes a concentration that changes; it does not prescribe the direction or trajectory of the change.
(2) "without step changes" — the claim does not exclude step changes. "Continuously varying" describes the general character of the concentration profile; it does not specify the microscopic mechanism by which the variation is achieved.
(3) "increases... from the silicon substrate interface to the absorber layer" — the claim does not specify the direction of variation or the reference points between which the variation occurs.
(4) "germanium atomic fraction" — the claim refers to "germanium concentration," not "germanium atomic fraction." While these terms may overlap in meaning, Lumos's substitution of a narrower technical term for the broader claim language is unjustified.
Each of these added limitations is drawn from the specific details of the preferred embodiment illustrated in Figure 5 and described in columns 6–7 of the '234 Patent specification. The preferred embodiment describes a buffer layer in which the germanium concentration increases monotonically from a low value at the silicon substrate interface to a higher value at the germanium photodetector absorber layer, using a smooth concentration gradient. But this is a description of one embodiment, not a definition of the claim term. The Federal Circuit has made clear that "particular embodiments appearing in the written description will not be used to limit claim language that has broader effect." Thorner, 669 F.3d at 1366; see also Liebel-Flarsheim, 358 F.3d at 906.
Dr. Whitfield's declaration confirms that a POSITA would understand "continuously varying" to encompass a concentration profile that changes across the layer, whether smoothly or in fine gradations, as long as the variation is not absent or uniform. Whitfield Decl. ¶¶ 45–56. As Dr. Whitfield explains, the term "continuously varying" in the semiconductor context describes a concentration that is not constant — it varies from one location to another within the buffer layer. Id. ¶ 47. The term does not inherently exclude graded profiles achieved through multiple discrete deposition steps, so long as the overall effect is a concentration that varies across the layer. Id. ¶ 48.
Lumos's construction would improperly narrow the claims to a single type of grading profile — a smooth, monotonically increasing gradient — that represents only one of several approaches a POSITA would associate with a "graded buffer layer having a continuously varying germanium concentration."
2. The Specification Supports a Broader Reading. Lumos argues that the specification of the '234 Patent supports only a smooth, continuous gradient for the "continuously varying" germanium concentration. See Resp. Br. at 28–30. This argument is contradicted by the specification itself.
The '234 Patent specification does not limit "continuously varying" to a smooth gradient. To the contrary, the specification describes embodiments with both "continuously graded" and "step-graded" buffer layers. Specifically, the specification states:
"The graded buffer layer may be formed using various grading profiles including continuous grading, step grading, and combinations thereof. In one embodiment, the buffer layer has a continuously graded germanium concentration that increases from approximately zero at the silicon substrate interface to a target concentration at the absorber layer interface. In another embodiment, the buffer layer is step-graded with a plurality of sublayers, each having a substantially constant germanium concentration that increases from sublayer to sublayer."
'234 Patent, col. 6, ll. 22–41.
This passage is dispositive. The specification expressly contemplates that the "graded buffer layer" of the invention may be implemented through continuous grading, step grading, or combinations of both. The fact that the claims use the phrase "continuously varying" rather than "step-graded" does not mean that step-graded profiles are excluded. Rather, as Dr. Whitfield explains, a POSITA would understand "continuously varying" to describe the macroscopic character of the concentration profile — it changes across the layer — without prescribing the microscopic mechanism of the variation. Whitfield Decl. ¶ 49.
Lumos's interpretation would create an internal inconsistency within the patent. If "continuously varying" excluded step-graded profiles, then the specification's extensive discussion of step-graded embodiments would be meaningless surplusage — a description of designs that fall outside the scope of the claims. Patent claims should be construed, where possible, to give meaning and effect to all portions of the specification. The Court should not adopt a construction that renders significant portions of the specification irrelevant.
Moreover, the prosecution history of the '234 Patent is unremarkable on this point and offers no support for Lumos's narrow construction. The claims of the '234 Patent were allowed on first office action after an examiner interview. There is no prosecution history narrowing of the "continuously varying" term. No arguments were made by the applicant during prosecution that would limit "continuously varying" to exclude step-graded profiles. No prosecution history estoppel argument is available to Lumos, and Lumos does not attempt to make one. The absence of any prosecution history support for Lumos's proposed limitations further confirms that those limitations should not be read into the claims.
3. Dr. Ng's Testimony Improperly Narrows the Claim. As with the '567 Patent term "adiabatically coupled," Lumos relies heavily on Dr. Ng's extrinsic testimony to support its narrow construction of "continuously varying." Dr. Ng's declaration asserts that "continuously varying" in materials science necessarily means a smooth, monotonic gradient without discrete steps, and he cites materials science textbooks in support. See Ng Resp. Decl. ¶¶ 30–38.
Dr. Whitfield demonstrates that this testimony reflects a narrow academic perspective that does not accurately capture the understanding of a POSITA working in the silicon photonics and semiconductor manufacturing context. Whitfield Decl. ¶¶ 52–55. As Dr. Whitfield explains, all epitaxial growth processes inherently involve discrete atomic layer deposition. Id. ¶ 52. At the atomic scale, no concentration gradient is truly "continuous" — every thin film is built up atom by atom, layer by layer. The distinction between "step-graded" and "continuously graded" buffer layers in practical semiconductor fabrication is therefore one of degree, not of kind. A step-graded buffer with sufficiently thin sublayers and sufficiently small concentration increments is functionally indistinguishable from a continuously graded buffer. Id. ¶ 53. This is well understood by practitioners in the field, including process engineers at foundries such as Pinnacle Semiconductor Foundry Co., Ltd. operating 45nm SOI processes. Id. ¶ 54.
Dr. Ng's textbook definitions describe idealized materials science concepts that do not reflect the practical realities of semiconductor fabrication. A POSITA reading the '234 Patent would not understand "continuously varying" to impose the rigid theoretical distinction that Dr. Ng advocates. The Court should decline to rely on extrinsic evidence from a related but distinct academic field (general materials science textbooks) to override the specification's own usage of the term and the practical understanding of practitioners in the silicon photonics field.
Under Phillips, the specification is "the single best guide to the meaning of a disputed term." 415 F.3d at 1315. The specification of the '234 Patent uses "continuously varying" in a manner consistent with OWS's broader construction, describes step-graded embodiments alongside continuously graded embodiments, and does not draw the rigid distinction that Dr. Ng's testimony would impose. The intrinsic record controls, and OWS's construction should be adopted.
B. Other Disputed Terms of the '234 Patent
Lumos's Responsive Brief also addresses one additional disputed term from the '234 Patent: "avalanche multiplication region integrated with the germanium absorber" (Claims 1, 8). The parties' dispute on this term concerns whether "integrated with" requires that the multiplication region and the absorber be formed in the same material layer (Lumos's position) or merely that they be functionally connected within the same device structure (OWS's position). As set forth in OWS's Opening Brief (at 24–26) and Dr. Whitfield's declaration (¶¶ 57–60), OWS's construction is supported by the intrinsic record, including the specification's description of embodiments in which the multiplication region and absorber occupy different but connected material layers. Lumos's reply arguments on this term do not raise any new issues that require further response, and OWS incorporates by reference the arguments made in its Opening Brief.
V. REPLY ARGUMENTS ON DISPUTED TERMS OF THE '678 PATENT
A. "Integrated Wavelength-Division Multiplexing Element Monolithically Formed on a Single Silicon Die" (Claim 1, and Dependent Claims 2, 6, 9, 14)
The parties' competing constructions are as follows:
Lumos's proposed construction is one of the most extreme in this case. It does not merely import a single limitation from the specification; it rewrites the claim by adding explicit negative limitations — specific exclusions of "hybrid integration, flip-chip bonding, or any heterogeneous integration technique" — that appear nowhere in the claim language. OWS addresses each of Lumos's arguments below.
1. Lumos's Construction Improperly Imports Negative Limitations. The claim recites "integrated wavelength-division multiplexing element monolithically formed on a single silicon die." This language identifies a WDM element that is (a) integrated, (b) monolithically formed, and (c) on a single silicon die. Each of these terms has a well-understood meaning. "Monolithically formed" is a standard term of art in semiconductor manufacturing meaning fabricated as part of an integrated process on a single substrate. OWS's proposed construction — "a WDM component fabricated as part of the same semiconductor fabrication process on a single die" — captures this meaning faithfully and without adding limitations.
Lumos's proposed construction, by contrast, adds explicit negative limitations that are not found in the claim language: "excluding hybrid integration, flip-chip bonding, or any heterogeneous integration technique." The Federal Circuit has repeatedly cautioned against importing negative limitations into claims absent clear support in the specification or prosecution history. See, e.g., Enzo Biochem, Inc. v. Corara Corp., 599 F.3d 1325, 1335 (Fed. Cir. 2010) ("Negative claim limitations are adequately supported when the specification describes a reason to exclude the relevant limitation."); Wasica Fin. GmbH v. Cont'l Auto. Sys., Inc., 853 F.3d 1272, 1284 (Fed. Cir. 2017).
The principle is straightforward: "A claim construction that excludes a particular structure or technique must be supported by clear language in the claims, specification, or prosecution history; a court may not add negative limitations to claims based on inferences or implications." See Aylus Networks, Inc. v. Camberwell & Co., 856 F.3d 1353, 1360–61 (Fed. Cir. 2017). Lumos does not point to any claim language that excludes hybrid integration, flip-chip bonding, or heterogeneous integration techniques. The specification of the '678 Patent does not define "monolithically formed" by reference to these exclusions. And while the prosecution history contains a relevant amendment and remark — which OWS addresses below — the prosecution record does not support the sweeping negative limitations that Lumos proposes.
Furthermore, Lumos's proposed limitation of "front-end-of-line processing" is a fabrication-process restriction that is not present in the claim language. The claims recite "monolithically formed on a single silicon die," not "formed during front-end-of-line processing." The distinction between FEOL and BEOL processing is a manufacturing process detail that the claims do not address and that does not bear on whether a WDM element is "monolithically formed" on a die. As Dr. Whitfield explains, and as discussed further below, modern silicon photonics fabrication routinely includes both FEOL and BEOL processing on a single substrate, and a POSITA would not understand "monolithically formed" to be limited to FEOL processing alone.
2. Prosecution History Estoppel Is Limited in Scope. Lumos's principal argument for its construction of this term rests on prosecution history estoppel. The '678 Patent underwent three office actions during prosecution. During that prosecution, the applicant amended the claims to add the "monolithically formed on a single silicon die" limitation in order to overcome a prior art rejection. The examiner had relied upon references describing hybrid integration approaches — specifically, references disclosing the fabrication of a WDM element on a separate substrate and its subsequent bonding to a silicon die using heterogeneous integration techniques.
In response to the examiner's rejection, the applicant added the "monolithically formed" limitation and submitted remarks stating:
"The present invention is distinguished from heterogeneous integration approaches such as those disclosed in the cited references. The WDM element of the present claims is monolithically formed on a single silicon die, meaning it is fabricated as part of the die's integrated fabrication process, rather than being separately fabricated and subsequently bonded or attached to the die."
See '678 Patent Prosecution History, Response dated February 22, 2020, at 11–12.
OWS acknowledges that this amendment and remark constitute a narrowing amendment that gives rise to prosecution history estoppel under Festo. OWS does not dispute that the applicant distinguished the claimed invention from the specific heterogeneous integration approaches described in the cited prior art — i.e., the fabrication of a WDM element on a completely separate substrate and its subsequent bonding to a silicon die.
However, the scope of this surrender under Festo must be carefully circumscribed. The estoppel extends to the subject matter actually surrendered, not to an unlimited range of techniques that Lumos labels as "heterogeneous." The prosecution record demonstrates the following key points:
First, the applicant's remark targeted specific prior art references that described full heterogeneous integration — i.e., the fabrication of a WDM photonic circuit on a completely separate substrate (such as an InP substrate) and its subsequent bonding to a silicon die. The applicant distinguished these references by noting that the claimed WDM element is "fabricated as part of the die's integrated fabrication process." This language identifies the surrender: the applicant disclaimed WDM elements that are fabricated on a separate substrate and then bonded to the silicon die. The remark does not disclaim all possible auxiliary integration steps that may accompany a monolithic fabrication process.
Second, Lumos's proposed construction extends the estoppel far beyond this surrender. Consider the following example, which Dr. Whitfield's declaration addresses: A WDM element (e.g., an arrayed waveguide grating or a Mach-Zehnder interferometer-based multiplexer) that is monolithically fabricated on a silicon die — formed through standard silicon photonics fabrication processes on that die — but which resides on the same die as a gain element (e.g., a III-V semiconductor laser) that was separately bonded to the die in a subsequent step. Under OWS's construction, the WDM element is "monolithically formed on a single silicon die" because the WDM element itself was fabricated as part of the die's integrated process. Under Lumos's construction, however, this same WDM element would fall outside the claims because the die also contains a heterogeneously integrated gain element — even though the WDM element itself was monolithically formed. Lumos's construction conflates the exclusion of heterogeneous integration of the WDM element itself with a blanket exclusion of any heterogeneous technique used anywhere on the die. This reading is overbroad and unsupported by the prosecution record.
Third, even if there were any doubt about the scope of the estoppel, the Festo rebuttal exceptions should be considered. Under Festo, a patentee may rebut the presumption of surrender if: (1) the alleged equivalent was unforeseeable at the time of the amendment; (2) the rationale underlying the amendment bears only a tangential relation to the equivalent in question; or (3) there is some other reason suggesting that the patentee could not reasonably have been expected to have described the insubstantial substitute in question. 535 U.S. at 740–41.
OWS submits that at least exception (2) — tangential relation — applies here. The rationale for the "monolithically formed" amendment was to distinguish a WDM element that was fabricated on a separate substrate and then bonded to the silicon die. The question of whether a WDM element that is monolithically fabricated on a silicon die — but that coexists on the same die with other components that may have been heterogeneously integrated — was tangential to the rationale for the amendment. The applicant was not addressing or disclaiming the scenario in which the WDM element itself is monolithically formed but other components on the die are not. The amendment was directed to the manner of fabrication of the WDM element, not to the fabrication of the entire die.
Accordingly, the prosecution history estoppel arising from the amendment of the '678 Patent claims is limited to the actual surrender: WDM elements that are separately fabricated on a different substrate and subsequently bonded to a silicon die. It does not extend to the sweeping negative limitations that Lumos proposes.
3. Dr. Ng's Testimony on "Monolithically Formed". Dr. Ng opines that "monolithically formed" in the semiconductor industry means fabricated exclusively through front-end-of-line (FEOL) processing within the same silicon substrate. See Ng Resp. Decl. ¶¶ 42–50. This definition is unduly narrow and does not reflect the understanding of a POSITA working in the silicon photonics field.
Dr. Whitfield responds comprehensively on this point. Whitfield Decl. ¶¶ 62–70. As Dr. Whitfield explains, modern silicon photonics fabrication routinely includes both FEOL and back-end-of-line (BEOL) processing on a single substrate. Id. ¶ 63. FEOL processing typically involves the formation of transistors and active devices within the silicon substrate, while BEOL processing involves the formation of metal interconnects, dielectric layers, and passivation layers above the silicon substrate. Id. ¶ 64. In silicon photonics, waveguide structures, gratings, and other photonic components may be formed during both FEOL and BEOL processing stages. Id. ¶ 65. A POSITA would understand "monolithically formed" to encompass the full fabrication sequence — FEOL and BEOL — on a single die, as both stages are part of the die's integrated fabrication process. Id. ¶ 66.
Dr. Ng's proposed limitation to FEOL processing would exclude virtually all modern silicon photonic integrated circuits, which routinely include BEOL-fabricated photonic components such as silicon nitride waveguides formed in BEOL dielectric layers. Id. ¶ 68. This would be an absurd result that the '678 Patent could not have intended. The specification of the '678 Patent describes a WDM element comprising components formed at multiple process layers, including layers above the silicon substrate. Limiting "monolithically formed" to FEOL processing would render the specification's own described embodiment outside the scope of the claims — a result that is impermissible under Federal Circuit law.
As with the other disputed terms, Dr. Ng's extrinsic testimony cannot override the intrinsic record. The specification's description of the WDM element, the claim language itself, and the prosecution history all support OWS's broader construction. The Court should adopt OWS's proposed construction.
B. Other Disputed Terms of the '678 Patent
Lumos's Responsive Brief addresses two additional disputed terms from the '678 Patent: (1) "arrayed waveguide grating having at least eight optical channels" (Claims 2, 14) and (2) "thermo-optic phase tuning element thermally coupled to at least one arm of the arrayed waveguide grating" (Claims 6, 9). The parties' positions on these terms are closer than on the primary disputed term addressed above, and OWS's Opening Brief adequately addressed the disputes. OWS incorporates by reference its Opening Brief arguments on these terms (Opening Br. at 31–35) and the corresponding portions of the Whitfield Declaration (¶¶ 71–74).
With respect to "arrayed waveguide grating having at least eight optical channels," the remaining dispute is whether "optical channels" must correspond to distinct wavelength channels (Lumos's position) or may encompass spatial channels as well (OWS's position). OWS maintains that the claim language does not limit "optical channels" to wavelength channels, and the specification's disclosure of multi-mode arrayed waveguide gratings supports OWS's broader reading.
With respect to "thermo-optic phase tuning element thermally coupled to at least one arm of the arrayed waveguide grating," the dispute is whether "thermally coupled" requires direct physical contact between the heater and the waveguide arm (Lumos's position) or merely thermal proximity sufficient to achieve effective thermal tuning (OWS's position). The specification describes both direct-contact and proximity-heated embodiments, and OWS's broader construction is consistent with the full scope of the disclosure.
VI. REPLY ARGUMENTS ON DISPUTED TERMS OF THE '890 PATENT
A. "Pre-Emphasis Equalization Stage Electrically Co-Integrated with the Modulator on a Common Substrate" (Claim 1, and Dependent Claims 5, 8, 13)
The parties' competing constructions are as follows:
Lumos's proposed construction is the most extreme example of limitation importation in this case. It transforms a straightforward claim requirement — co-integration on a common substrate — into a multi-factor test requiring identity of semiconductor process, shared doping profiles, shared process layers, and the absence of specific interconnect technologies. None of these requirements appear in the claim language.
1. Lumos's Construction Adds Multiple Unwarranted Process Limitations. The claim recites a "pre-emphasis equalization stage electrically co-integrated with the modulator on a common substrate." The claim language identifies three requirements: (1) a pre-emphasis equalization stage, (2) electrical co-integration with the modulator, and (3) location on a common substrate. OWS's proposed construction — "an equalization circuit located on the same substrate as the modulator" — faithfully captures these requirements. "Co-integrated" means integrated together on the same substrate. "Common substrate" identifies the physical platform on which both the equalization circuit and the modulator reside.
Lumos's proposed construction adds four distinct limitations not found in the claim language:
(a) "fabricated in the same semiconductor process" — The claim does not require that the equalization circuit and the modulator be fabricated in the same process. "Co-integrated on a common substrate" describes the physical arrangement, not the fabrication method. A POSITA would understand that components may be co-integrated on a common substrate through range fabrication techniques, including processes in which different components are formed in different process steps or even different process modules.
(b) "sharing the same doping profiles and process layers" — The claim does not require shared doping profiles or process layers. This limitation is drawn directly from the preferred embodiment described in the '890 Patent specification, in which the equalization circuit and the modulator share certain silicon doping profiles to enable efficient co-fabrication. But this is a feature of one embodiment, not a claim requirement. An equalization circuit and a modulator can be co-integrated on a common substrate without sharing identical doping profiles — for example, where the equalization circuit uses a different implant step or a different active layer than the modulator.
(c) "electrically connected without bond wires" — The claim requires "electrically co-integrated," not "electrically connected without bond wires." The claim language does not specify the interconnect technology. While bond wires between components on the same substrate would be unusual, the claim does not exclude them, and Lumos's construction improperly adds a negative limitation.
(d) "[without] interposers, or package-level routing" — The claim does not exclude interposers or package-level routing. Again, Lumos adds negative limitations that are absent from the claim text and unsupported by the intrinsic record.
Each of these additions narrows the claim scope beyond what the language supports. The claim drafters chose the language "electrically co-integrated... on a common substrate" — broad, functional language describing a physical arrangement. They did not choose language requiring shared process layers, shared doping profiles, or the absence of specific interconnect technologies. The Court should give effect to the language the patentee chose, not the language Lumos wishes the patentee had chosen.
2. Prosecution History Context. The prosecution history of the '890 Patent provides context for the meaning of "co-integrated" but does not support the sweeping limitations Lumos proposes.
During prosecution, the "co-integrated" limitation was added to distinguish prior art in which the driver circuit (equalization stage) and the modulator were fabricated on entirely separate substrates and then connected through package-level interconnects. Specifically, the examiner's rejection was based on Iyer & Patel, "Co-Design of Modulator Drivers and Silicon Photonic Mach-Zehnder Structures," IEEE Transactions on Circuits and Systems, Vol. 65, No. 8, pp. 2701–2713 (2018). The Iyer & Patel reference described a co-design methodology in which a driver circuit and a silicon photonic modulator are designed for compatibility — i.e., co-designed — but are fabricated on separate substrates and interconnected at the package level. Notably, Rajesh Iyer, a named inventor on the '890 Patent, is the lead author of the Iyer & Patel reference, which was published before his work at OWS on the '890 Patent invention.
The applicant's amendment and remarks distinguished Iyer & Patel by noting that the prior art described a co-design methodology but did not disclose physical co-integration on a common substrate:
"The Iyer & Patel reference describes a co-design approach in which the driver circuit and the modulator are designed for electrical compatibility, but the driver and modulator are fabricated on separate substrates and connected via conventional package-level wiring. In contrast, the present claims require that the pre-emphasis equalization stage be 'electrically co-integrated with the modulator on a common substrate' — i.e., that both the equalization stage and the modulator reside on the same physical substrate, rather than being connected at the package level between separate dies."
See '890 Patent Prosecution History, Response dated August 5, 2021, at 14–15.
The scope of the surrender is clear from the applicant's remarks. The applicant distinguished Iyer & Patel by pointing to the physical separation of the driver and modulator in the prior art — they reside on different substrates and are connected at the package level. The amendment requires that the equalization stage and modulator be on the same substrate. The surrender is limited to designs in which the driver and modulator are on entirely separate substrates. The surrender does not extend to requiring identical process layers, identical doping profiles, or the absence of all package-level interconnects on a die that co-integrates both components.
Lumos's attempt to extend the estoppel beyond the actual surrender is unavailing. The applicant's remarks said nothing about doping profiles, process layers, or the absence of bond wires and interposers. These are fabrication details that the applicant did not discuss, did not rely upon, and did not surrender. Under Festo, the estoppel is limited to the territory actually surrendered. 535 U.S. at 740. Lumos cannot expand the estoppel to cover limitations that the applicant never addressed during prosecution.
As with the '567 Patent, the fact that a named inventor on the '890 Patent (Rajesh Iyer) is also a co-author of the prior art reference used by the examiner does not alter the legal analysis. The prosecution history estoppel inquiry focuses on the scope of the applicant's surrender, not on the identity of the prior art authors. OWS notes this coincidence only because Lumos's Responsive Brief appears to suggest that it has evidentiary significance for claim construction purposes. It does not.
3. Extrinsic Evidence. Dr. Ng's declaration once again imports rigid academic definitions in support of Lumos's narrow construction. Dr. Ng opines that "co-integrated" in the semiconductor industry necessarily requires fabrication in the same semiconductor process with shared doping profiles and process layers. See Ng Resp. Decl. ¶¶ 55–62.
Dr. Whitfield addresses this testimony directly and persuasively. Whitfield Decl. ¶¶ 75–82. As Dr. Whitfield explains, "co-integrated" in the silicon photonics industry encompasses multiple levels of integration