1014-00003 · Optiwave Semiconductor Corp · Post-Markman Assessment

post-markman-impact-assessment-memo.docx DOCUMENT Litigation

60 / 83 ← → ✕

PRIVILEGED AND CONFIDENTIAL ATTORNEY WORK PRODUCT — ATTORNEY-CLIENT COMMUNICATION

This memorandum is a privileged and confidential attorney work product prepared in anticipation of litigation and constitutes a protected attorney-client communication. This document and its contents are not to be disclosed to any person or entity outside the engagement team and approved distribution list without the express written authorization of the Responsible Partner. Any unauthorized disclosure may result in waiver of privilege.

February 21, 2023

INTERNAL MEMORANDUM

To: File — Matter No. 1014-00003

From: Alan Ngo, Responsible Partner Omar Siddiqui, Lead Associate Christy Bauer, Associate Intellectual Property Group, Calderwood & Harkness LLP

Re: Post-Markman Impact Assessment — Optiwave Semiconductor Corp. v. Lumos Photonics, Inc., Case No. 5:22-cv-03417-EJD (N.D. Cal.)

Distribution: Limited to engagement team members Alan Ngo, Marcus Delgado, Omar Siddiqui, and Christy Bauer. This memorandum may be shared with client representatives Sandra Ellison (General Counsel, Optiwave Semiconductor Corp.), Ted Marchetti (Chief Financial Officer, Optiwave Semiconductor Corp.), and Dr. Priya Ramanathan (Chief Technology Officer and Co-Founder, Optiwave Semiconductor Corp.), as well as representatives of Vantage Point Capital Partners, including Douglas Fairbanks III (Managing Director) and Rebecca Torres (Principal), at the discretion of the Responsible Partner.

Purpose: This memorandum provides a comprehensive assessment of the Markman Order issued by the Honorable Edward J. Torvelis on February 17, 2023, in the above-captioned matter. The memorandum analyzes the impact of the Court's claim constructions on Optiwave Semiconductor Corp.'s ("OWS") infringement theories for all four patents-in-suit, evaluates the remaining viability of those theories, assesses appellate options, and sets forth recommendations for the path forward, including a cost-benefit analysis of continued litigation versus settlement.

Right-click to update Table of Contents

I. Executive Summary

On February 17, 2023, the Honorable Edward J. Torvelis of the United States District Court for the Northern District of California issued a thirty-four-page Markman Order construing all eleven disputed claim terms across the four patents asserted by OWS in this action: 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"). The Order followed a six-and-a-half-hour claim construction hearing held on January 30, 2023, at which both parties presented technical tutorials and oral argument.

Overall Scorecard. The Court adopted Defendant Lumos Photonics, Inc.'s ("Lumos") proposed construction — or a construction substantially aligned with Lumos's position — on eight of the eleven disputed claim terms. The Court adopted Plaintiff OWS's proposed construction on two terms, and adopted a modified construction that did not correspond verbatim to either party's proposal on one term. This outcome represents a decisive defeat for OWS on the claim construction issues that are most material to its infringement theories.

Bottom-Line Assessment. The Court's constructions render OWS's infringement theories untenable as to three of the four patents-in-suit — the '567 Patent, the '234 Patent, and the '678 Patent — and significantly weakened as to the fourth, the '890 Patent. Specifically, the constructions adopted for the four critical claim limitations at issue — "adiabatically coupled" (the '567 Patent), "continuously varying germanium concentration" (the '234 Patent), "monolithically formed" (the '678 Patent), and "co-integrated" (the '890 Patent) — effectively exclude the accused Lumos products, namely the LumosBridge™ 800G Optical Engine and the LumosEdge™ 400ZR+ Pluggable Transceiver, from the literal scope of the asserted claims. In each instance, the Court's construction introduces specificity and narrowness that maps precisely to technical distinctions between OWS's patented designs and Lumos's accused products.

For the three patents where infringement theories are untenable, the prosecution history underlying the Court's constructions further forecloses any viable doctrine-of-equivalents ("DOE") theory, as the narrowing limitations were in each case added by amendment during prosecution to overcome prior art, triggering the presumption of prosecution history estoppel under Festo Corp. v. Fenmore & Co., 535 U.S. 722 (2002). For the '890 Patent, while a narrow theoretical DOE argument may survive, it is assessed as weak, with an estimated probability of success at trial of no more than ten to fifteen percent.

Recommendation. Based on the analysis set forth in this memorandum, we recommend that OWS initiate settlement discussions with Lumos at the earliest opportunity. We further recommend that this impact assessment be presented to the OWS Board of Directors and to Vantage Point Capital Partners at a Board meeting to be scheduled for February 24, 2023, at which formal settlement authority should be requested. Our recommended settlement approach is a cross-license with a nominal royalty payment from Lumos as the opening position, with authority to accept a walk-away settlement with mutual releases as an acceptable fallback outcome.

II. Background

A. The Litigation

Optiwave Semiconductor Corp. commenced this action by filing a Complaint for Patent Infringement on June 3, 2022, in the United States District Court for the Northern District of California, San Jose Division. The Complaint asserted infringement of four United States patents owned by OWS, each directed to different aspects of high-speed optical transceiver technology utilized in data center interconnect applications:

• U.S. Patent No. 10,234,567 (the "'567 Patent"), entitled "Integrated Waveguide Coupler with Tapered Mode Converter for Silicon Photonics Platforms," which covers integrated waveguide coupling architectures employing tapered mode converters for efficient optical coupling in silicon photonic integrated circuits;

• U.S. Patent No. 10,891,234 (the "'234 Patent"), entitled "Germanium-on-Silicon Photodetector with Graded Buffer Layer for Enhanced Responsivity," which covers germanium-on-silicon photodetector architectures utilizing graded buffer layers to achieve enhanced responsivity at telecommunications wavelengths;

• U.S. Patent No. 11,045,678 (the "'678 Patent"), entitled "Multi-Channel Optical Transceiver with Integrated Wavelength-Division Multiplexing on Silicon," which covers multi-channel WDM integration architectures in which wavelength-division multiplexing elements are monolithically formed on a single silicon die; and

• U.S. Patent No. 11,312,890 (the "'890 Patent"), entitled "Driver Circuit Architecture for High-Speed Silicon Photonic Mach-Zehnder Modulators," which covers MZM driver circuit architectures in which a pre-emphasis equalization stage is co-integrated with the modulator on a common substrate.

The Complaint identified two Lumos products as infringing each of the four patents-in-suit:

• LumosBridge™ 800G Optical Engine — a co-packaged optics engine designed for deployment in hyperscale data center switches. This product first shipped to customers in the first quarter of 2022 and carries a unit price in the range of $1,200 to $1,800, depending on configuration and volume; and

• LumosEdge™ 400ZR+ Pluggable Transceiver — a QSFP-DD form factor pluggable transceiver module designed for 400ZR+ coherent data center interconnect applications. This product has been in production since the third quarter of 2021 and carries a unit price in the range of $600 to $900.

Combined accused product revenue for Lumos for calendar year 2022 was approximately $148.6 million, based on publicly available financial disclosures and information obtained through fact discovery. OWS's damages claim, as originally formulated, exceeded $75 million, premised on a 7.5% reasonable royalty rate applied to the accused product revenue base, inclusive of pre-suit damages for the period in which the accused products were sold prior to the filing of the Complaint and a willfulness enhancement of up to three times under 35 U.S.C. § 284.

Lumos filed its Answer and Counterclaims on August 5, 2022, denying infringement and asserting counterclaims for declaratory judgment of non-infringement and invalidity of all four patents-in-suit under 35 U.S.C. §§ 102, 103, and 112.

B. Claim Construction Proceedings

The claim construction proceedings in this action followed the Patent Local Rules for the Northern District of California and proceeded on the following timeline:

• September 12, 2022: The parties exchanged Preliminary Claim Construction Contentions pursuant to Patent Local Rule 4-1, identifying the claim terms each party contended required construction and setting forth preliminary proposed constructions.

• October 3, 2022: The parties exchanged Responsive Claim Construction Contentions pursuant to Patent Local Rule 4-2.

• October 14, 2022: The parties filed a Joint Claim Construction and Prehearing Statement pursuant to Patent Local Rule 4-3, identifying the eleven disputed claim terms across all four patents-in-suit that would require resolution by the Court, together with each party's proposed constructions and supporting evidence.

• October 31, 2022: Claim construction discovery closed. During the discovery period, both parties deposed the named inventors of the patents-in-suit, including depositions of Dr. Kevin Fong (co-inventor on the '567, '234, and '678 Patents), Hideo Tanaka (co-inventor on the '567 Patent), and Dr. Amara Osei (named inventor on the '890 Patent). These depositions addressed, among other things, the understanding of a person of ordinary skill in the art regarding the disputed claim terms, the prosecution history of each patent, and the state of the art at the time of the inventions.

• December 5, 2022: OWS filed its Opening Claim Construction Brief (45 pages), setting forth OWS's proposed constructions for all eleven disputed terms, supported by intrinsic evidence (specification and prosecution history), extrinsic evidence (technical literature, dictionary definitions, and expert declarations), and the declaration of Dr. Harold Whitfield.

• December 19, 2022: Lumos filed its Responsive Claim Construction Brief (45 pages), setting forth Lumos's proposed constructions and opposing OWS's proposals, supported by comparable evidence and the declaration of Dr. Raymond Ng.

• January 9, 2023: OWS filed its Reply Brief in support of its proposed constructions (15 pages).

• January 19, 2023: Lumos filed a Sur-Reply Brief by leave of Court (15 pages), responding to new arguments raised in OWS's Reply.

• January 30, 2023: The Court held a Markman hearing lasting approximately six and a half hours. Each party presented a technical tutorial to the Court, followed by oral argument on each disputed term. Dr. Harold Whitfield of Whitfield Technical Consulting Group, Inc. (Palo Alto, California) testified as OWS's technical tutorial witness, providing an overview of silicon photonics technology and explaining OWS's proposed constructions. Dr. Raymond Ng, Of Counsel at Brannick Howell & Stein LLP and former Professor of Electrical Engineering at the University of Colorado Boulder, testified as Lumos's technical tutorial witness. The Court posed extensive questions to both tutorial witnesses and to counsel for both parties throughout the hearing.

• February 17, 2023: The Court issued its Markman Order (34 pages), construing all eleven disputed claim terms. The Order is the subject of this memorandum.

C. The Engagement Team and Expert Advisors

The following personnel are engaged on this matter:

Calderwood & Harkness LLP Litigation Team:

• Alan Ngo — Responsible Partner. Mr. Ngo leads the firm's semiconductor and photonics patent litigation practice and has over twenty-five years of experience in complex patent litigation in the Northern District of California and the International Trade Commission. Mr. Ngo is the primary point of contact for the client and for Vantage Point Capital Partners.

• Marcus Delgado — Billing Partner. Mr. Delgado serves as the billing partner on this engagement and provides strategic oversight on litigation budgeting and resource allocation.

• Omar Siddiqui — Lead Associate. Mr. Siddiqui manages the day-to-day conduct of the litigation, including discovery, claim construction briefing, and expert coordination. Mr. Siddiqui holds a B.S. in Electrical Engineering from Stanford University and a J.D. from Georgetown University Law Center.

• Christy Bauer — Working Associate. Ms. Bauer assists with legal research, brief drafting, document review, and prosecution history analysis. Ms. Bauer holds a B.S. in Physics from MIT and a J.D. from the University of Virginia School of Law.

Technical Expert:

• Dr. Harold Whitfield — Retained through Whitfield Technical Consulting Group, Inc. (Palo Alto, California). Dr. Whitfield holds a Ph.D. in Electrical Engineering with a specialization in photonics from the University of California, Berkeley. He is Professor Emeritus of Electrical and Computer Engineering at the University of California, Santa Barbara, where he directed the Integrated Photonics Laboratory for over twenty years. Dr. Whitfield has authored or co-authored over 200 publications in the field of integrated photonics and has served as a testifying expert in over thirty patent cases involving semiconductor and photonics technologies.

Damages Expert:

• Dr. Lawrence Kessler — Retained through Apex Claim Analytics LLC (Chicago, Illinois). Dr. Kessler holds a Ph.D. in Economics from Northwestern University and specializes in damages quantification in patent infringement matters, with particular expertise in reasonable royalty calculations for semiconductor technologies.

Defendant's Counsel:

• Victoria "Tori" Ashworth — Lead Partner, Brannick Howell & Stein LLP (San Francisco office). Ms. Ashworth leads Lumos's defense team and has a well-established reputation as one of the most effective patent defense litigators in the Northern District of California.

• Daniel Kowalski — Associate Team Lead, Brannick Howell & Stein LLP (San Francisco office). Mr. Kowalski serves as the associate team lead on the defense side and has handled the majority of claim construction briefing for Lumos.

III. Term-by-Term Analysis of the Markman Order

This section provides a detailed analysis of the Court's claim construction rulings as set forth in the February 17, 2023 Markman Order, organized by patent. For each patent, we analyze the critical disputed claim term that is dispositive of OWS's infringement theory, address any secondary terms construed by the Court, and provide an overall assessment of the viability of OWS's infringement position. A complete summary table covering all eleven disputed terms is provided in Appendix A.

A. U.S. Patent No. 10,234,567 (the '567 Patent) — "Integrated Waveguide Coupler with Tapered Mode Converter for Silicon Photonics Platforms"

The '567 Patent is directed to integrated waveguide coupling structures that use tapered mode converters to efficiently couple light between optical fibers and silicon photonic waveguides. The patent claims priority to a provisional application filed on March 14, 2017, and was issued on March 19, 2019. The named inventors are Dr. Kevin Fong and Hideo Tanaka. OWS asserted Claims 1, 4, 7, 12, and 15 of the '567 Patent in this action, all of which depend from independent Claim 1. The Court construed two disputed terms from the '567 Patent.

1. Term: "adiabatically coupled" (Claim 1)

This is the critical term for the '567 Patent. Claim 1 of the '567 Patent recites, inter alia, "a first waveguide and a second waveguide that are adiabatically coupled through a tapered mode converter region." The meaning of "adiabatically coupled" was the most extensively briefed and argued term in the Markman proceedings, occupying approximately ninety minutes of the January 30 hearing.

OWS's Proposed Construction: "coupled such that optical mode transitions occur without significant power loss."

OWS argued that the term "adiabatically coupled" should be given its broadest reasonable meaning consistent with the specification, which OWS contended was a functional definition focused on the performance characteristic of the coupling — namely, that optical mode transitions occur without significant power loss. OWS relied on the ordinary meaning of "adiabatic" in physics (a process occurring without exchange of energy with the environment) and on portions of the specification describing the goal of the invention as "achieving efficient, low-loss coupling" between waveguides. OWS further argued that the specification disclosed multiple embodiments of structures that could achieve adiabatic coupling, and that the term should not be limited to any single structural implementation.

Lumos's Proposed Construction: "coupled through a gradual, continuous taper that prevents excitation of higher-order modes throughout the entire transition region."

Lumos argued that "adiabatically coupled" had a specific structural meaning to a person of ordinary skill in the art in the field of silicon photonics, requiring a gradual and continuous taper — not merely a low-loss coupling. Lumos relied heavily on the prosecution history of the '567 Patent, pointing to a narrowing amendment made during prosecution in which the applicants added the "adiabatically coupled" limitation to overcome a rejection based on the Tanaka prior art reference, U.S. Patent No. 9,456,012, issued September 6, 2016. That prior art reference — ironically co-invented by Hideo Tanaka before he joined OWS — disclosed a non-continuous, segmented coupling approach. Lumos argued that by adding "adiabatically coupled" to distinguish over the Tanaka reference, the applicants disclaimed non-continuous coupling designs and adopted a construction limited to gradual, continuous tapers. Lumos further relied on the testimony of Dr. Raymond Ng, who opined that a person of ordinary skill in the art in silicon photonics would understand "adiabatic" coupling to require a continuous, gradual taper structure that avoids exciting higher-order modes at any point along the transition.

Court's Construction: "coupled through a gradual, continuous taper designed and configured to prevent excitation of higher-order modes over the entire transition length."

Analysis. The Court adopted Lumos's proposed construction with only minor wording modifications. Specifically, the Court substituted "designed and configured" for "that prevents" and "over the entire transition length" for "throughout the entire transition region." These modifications are stylistic rather than substantive; the constructed meaning is functionally identical to Lumos's proposal. In its analysis, the Court made the following key findings:

First, the Court examined the prosecution history and found it dispositive. During prosecution of the '567 Patent, the claims were rejected twice over the Tanaka prior art (U.S. Patent No. 9,456,012). In response to the second rejection, the applicants amended independent Claim 1 to add the "adiabatically coupled" limitation and submitted remarks arguing that the claimed invention was "fundamentally distinguished from Tanaka's segmented coupling approach" because "the present invention employs an adiabatic transition in which the taper is continuous and gradual, ensuring that no higher-order modes are excited at any point during the transition." The Court found that these remarks constituted a clear and unambiguous disavowal of non-continuous coupling structures.

Second, the Court examined the specification and found that every embodiment described in the specification consistently depicted adiabatic coupling in terms of continuous, gradual tapers. While OWS argued that the specification's description of the "goal" of adiabatic coupling (low-loss transitions) supported a functional construction, the Court found that the structural descriptions in the specification were consistent with — and reinforced — the narrower construction.

Third, the Court credited Dr. Ng's testimony that a person of ordinary skill in the art in the field of silicon photonics, as of the priority date, would understand "adiabatic" coupling to refer specifically to coupling through a continuous, gradual geometric transition that prevents mode conversion, and would not understand the term to encompass segmented or discontinuous coupling approaches that might also achieve low loss through other mechanisms.

Fourth, the Court declined to credit Dr. Whitfield's broader construction. While acknowledging Dr. Whitfield's credentials and expertise, the Court noted that the intrinsic evidence — particularly the prosecution history — was clear and unambiguous, and that resort to extrinsic evidence to broaden the claim term beyond what the intrinsic evidence supports is impermissible under Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005) (en banc).

Impact on Infringement Theory: Fatal. The Markman construction of "adiabatically coupled" is dispositive of OWS's infringement theory for the '567 Patent. It is undisputed — and confirmed through OWS's own reverse engineering analysis of the accused products — that the LumosBridge™ 800G Optical Engine employs a segmented (non-continuous) taper design with discrete coupling sections. Specifically, the LumosBridge™ waveguide coupler utilizes a multi-segment inverse taper structure in which the waveguide width changes in discrete steps across multiple lithographically defined sections, rather than a single continuous taper. Under the Court's construction, which requires a "gradual, continuous taper," Lumos's segmented design falls squarely outside the literal scope of Claim 1.

Moreover, a doctrine-of-equivalents argument is foreclosed by prosecution history estoppel. Because the "adiabatically coupled" limitation was added by amendment during prosecution specifically to overcome the Tanaka prior art — which disclosed a non-continuous coupling approach — the Festo presumption applies. Under Festo Corp. v. Fenmore & Co., 535 U.S. 722 (2002), a narrowing amendment made to satisfy any requirement of the Patent Act gives rise to a presumption that the patentee has surrendered all equivalents between the original claim scope and the amended claim scope. Lumos's segmented taper design is precisely the type of structure that was disclaimed during prosecution; applying DOE to recapture it would directly contradict the prosecution history. None of the three recognized exceptions to the Festo presumption — unforeseeability, tangential relation, or "some other reason" — is applicable here, as segmented coupling structures were well known in the art at the time of the amendment and were the specific subject matter the amendment was designed to overcome.

2. Other '567 Patent Terms

The Court also construed one additional term from the '567 Patent: "tapered mode converter region" (Claim 1). The parties disputed whether this term required a specific geometric taper profile (Lumos's position, proposing a construction requiring a "linear or adiabatic taper profile") or whether it encompassed any region in which mode conversion occurs through a geometric transition (OWS's position). The Court adopted OWS's proposed construction on this term, finding that the specification described multiple taper profiles and that the claim language did not require a specific profile type.

While OWS prevailed on this secondary term, the victory is immaterial. Because the "adiabatically coupled" limitation appears in independent Claim 1 and is present in or inherited by every asserted claim (Claims 1, 4, 7, 12, and 15), the Court's construction of "adiabatically coupled" is dispositive for the entire set of asserted claims from the '567 Patent. A favorable construction of "tapered mode converter region" cannot rescue an infringement theory that fails on the "adiabatically coupled" limitation.

3. Net Assessment — '567 Patent

Infringement viability: Untenable. The Court's construction of "adiabatically coupled" as requiring a "gradual, continuous taper designed and configured to prevent excitation of higher-order modes over the entire transition length" excludes the accused products from the literal scope of every asserted claim of the '567 Patent. The LumosBridge™ product's segmented taper design is categorically outside this construction. Prosecution history estoppel under Festo further forecloses any meaningful DOE theory, as the limitation was added specifically to overcome prior art disclosing the very type of non-continuous coupling that Lumos employs.

Recommendation: Abandon infringement theories under the '567 Patent. In the event OWS elects to pursue settlement discussions, the '567 Patent should not be identified as a patent on which OWS intends to press its infringement claims.

B. U.S. Patent No. 10,891,234 (the '234 Patent) — "Germanium-on-Silicon Photodetector with Graded Buffer Layer for Enhanced Responsivity"

The '234 Patent is directed to germanium-on-silicon (Ge-on-Si) photodetector architectures that employ a graded buffer layer between the silicon substrate and the germanium absorber layer to reduce threading dislocation density and enhance photodetector responsivity. The patent claims priority to a provisional application filed on July 22, 2018, and was issued on January 12, 2021. The named inventor is Dr. Kevin Fong. OWS asserted Claims 1, 3, 8, and 14 of the '234 Patent in this action, all of which depend from independent Claim 1. The Court construed two disputed terms from the '234 Patent.

1. Term: "graded buffer layer having a continuously varying germanium concentration" (Claim 1)

This is the critical term for the '234 Patent. Claim 1 recites, inter alia, "a graded buffer layer disposed between the silicon substrate and the germanium absorber layer, the graded buffer layer having a continuously varying germanium concentration."

OWS's Proposed Construction: "a buffer layer in which the germanium concentration changes across the layer thickness."

OWS proposed a broad construction that would encompass any buffer layer in which the germanium concentration is not uniform — that is, any layer in which the concentration "changes" from one level to another, whether by continuous gradient, step gradient, or any other variation. OWS argued that the specification described both "continuously graded" and "step-graded" buffer layer embodiments and that the claim language was broad enough to encompass both. OWS further argued that a POSITA would understand "continuously varying" to mean "varying across the layer," rather than "varying without interruption," relying on certain dictionary definitions of "continuously" that emphasize duration or extent rather than smoothness.

Lumos's Proposed Construction: "a buffer layer in which the germanium atomic fraction increases monotonically and without step changes from the silicon substrate interface to the absorber layer."

Lumos argued that "continuously varying" had a clear and well-understood technical meaning to a POSITA in the field of semiconductor epitaxy, requiring a smooth, monotonic change in germanium concentration without discrete steps. Lumos pointed to the claim language itself — "continuously varying" — and contrasted it with the "step-graded" embodiments described in the specification but not claimed. Lumos argued that if the applicant had intended to encompass step-graded designs, the applicant could have used broader language (such as "varying" or "graded") rather than "continuously varying."

Court's Construction: "a buffer layer in which the germanium atomic fraction increases monotonically and without step changes from the silicon substrate interface to the absorber layer."

The Court adopted Lumos's proposed construction verbatim. This represents the most straightforward and complete adverse ruling of the eleven disputed terms.

Analysis. The Court's reasoning proceeded along two principal lines:

First, the Court examined the claim language and found that "continuously varying" has a well-understood plain technical meaning. The Court noted that in the field of semiconductor epitaxy and materials science, "continuous" variation is routinely distinguished from "step" or "discrete" variation, and that a POSITA would understand "continuously varying germanium concentration" to require a smooth gradient without discrete steps. The Court cited several technical references from the intrinsic and extrinsic record confirming this usage, including the IEEE Photonics Technology Letters article by Chen et al. (2015) and the textbook by Kasper and Lyutovich, Properties of Silicon Germanium and SiGe:Carbon (INSPEC, 2000), both of which distinguish "continuously graded" from "step-graded" SiGe buffer layers.

Second, and critically, the Court addressed OWS's argument that the specification's description of both continuously graded and step-graded embodiments supported a broader construction. The Court rejected this argument, holding that while the specification described step-graded designs as part of the background and as alternative embodiments, the claims used only "continuously varying" language. The Court emphasized that it is a fundamental principle of claim construction that claims are not broadened to encompass unclaimed embodiments merely because those embodiments are described in the specification. The Court cited Phillips, 415 F.3d at 1323, and Ashford Enterprises v. Dunmore Corporation, 358 F.3d 898, 906 (Fed. Cir. 2004), for the proposition that "the specification may describe unclaimed embodiments without expanding the scope of the claims." The Court found that the applicant made a deliberate drafting choice to claim only "continuously varying" concentrations and not step-graded designs, and the Court would enforce that choice.

The Court's reasoning on this point deserves particular attention because it illustrates a patent drafting deficiency that cannot be remedied at this stage of the litigation. The specification of the '234 Patent describes step-graded buffer layers in some detail, and the named inventor, Dr. Kevin Fong, testified at his deposition that he considered both continuous and step-graded approaches to be within the scope of his invention. However, the claim language as drafted is limited to "continuously varying" concentrations, and the Court properly held that the claims control.

Impact on Infringement Theory: Fatal. The accused Lumos products employ a step-graded SiGe buffer layer with four discrete composition steps, in which the germanium atomic fraction increases from 0.25 to 0.50 to 0.75 to 1.0 at defined interfaces. This step-graded structure is precisely the type of design that the Court's construction excludes through the "without step changes" language. There is no colorable literal infringement argument under this construction.

A DOE argument is likewise foreclosed. To argue that a four-step graded buffer is equivalent to a continuously varying (i.e., no step changes) buffer would directly vitiate the "without step changes" limitation, which is impermissible. Under Ridgeway & Co. v. Conceptronic, Inc., 90 F.3d 1576, 1583 (Fed. Cir. 1996), and its progeny, a DOE theory that would effectively eliminate a claim limitation is barred by the doctrine of claim vitiation. Here, applying DOE to encompass a design with four discrete composition steps would render the "without step changes" requirement meaningless, and no court would permit such an argument.

2. Other '234 Patent Terms

The Court also construed one additional term from the '234 Patent: "enhanced responsivity" (Claim 1). OWS proposed a construction defining this as "responsivity greater than that of a comparable photodetector without a graded buffer layer," while Lumos proposed a construction requiring "responsivity of at least 0.9 A/W at 1310 nm," based on a specific performance figure recited in the preferred embodiment. The Court adopted OWS's proposed construction, finding that the specific responsivity figure cited by Lumos was from a preferred embodiment and should not be imported as a limitation into the claims.

This was the second of the two terms on which OWS prevailed in the Markman proceedings. However, as with the "tapered mode converter region" term in the '567 Patent, this victory is immaterial because the dispositive "continuously varying germanium concentration" construction eliminates OWS's infringement theory at the threshold of independent Claim 1, from which all asserted claims depend.

3. Net Assessment — '234 Patent

Infringement viability: Untenable. The Court's verbatim adoption of Lumos's proposed construction for "graded buffer layer having a continuously varying germanium concentration," requiring a monotonic increase "without step changes," categorically excludes Lumos's step-graded photodetector design. No viable literal infringement or DOE theory survives this construction.

The particular significance of this outcome merits emphasis: the specification of the '234 Patent itself described step-graded buffer layer embodiments, but the applicant chose claim language limited to "continuously varying" concentrations. This drafting choice — which may have been inadvertent — was enforced by the Court as a deliberate claim scope limitation. This underscores the importance of claim drafting precision and the risk of relying on specification disclosures that are not captured in the claim language.

Recommendation: Abandon infringement theories under the '234 Patent. As with the '567 Patent, the '234 Patent should not be a basis for continued infringement assertions in any settlement negotiation.

C. U.S. Patent No. 11,045,678 (the '678 Patent) — "Multi-Channel Optical Transceiver with Integrated Wavelength-Division Multiplexing on Silicon"

The '678 Patent is directed to multi-channel optical transceiver architectures in which a wavelength-division multiplexing (WDM) element is monolithically integrated on a single silicon die alongside other photonic components, enabling higher-density channel integration and reduced packaging complexity. The patent claims priority to a non-provisional application filed on November 8, 2019, and was issued on June 29, 2021. The named inventors are Dr. Kevin Fong and Dr. Amara Osei. OWS asserted Claims 1, 5, 9, 11, and 18 of the '678 Patent in this action, all of which depend from independent Claim 1. The Court construed three disputed terms from the '678 Patent.

1. Term: "integrated wavelength-division multiplexing element monolithically formed on a single silicon die" (Claim 1)

This is the critical term for the '678 Patent. Claim 1 recites, inter alia, "an integrated wavelength-division multiplexing element monolithically formed on a single silicon die, the WDM element configured to combine or separate a plurality of optical channels at distinct wavelengths."

OWS's Proposed Construction: "a WDM component fabricated as part of the same semiconductor fabrication process on a single die."

OWS proposed a construction focused on the fabrication process, arguing that "monolithically formed" required only that the WDM component be fabricated as part of the same semiconductor process flow used to create the rest of the photonic integrated circuit on the die. OWS argued that this construction was consistent with common usage of "monolithic" in the semiconductor industry and would encompass designs where some components (such as III-V gain elements) are integrated onto the silicon die during the fabrication process, even if different materials are involved.

Lumos's Proposed Construction: "a WDM component formed during front-end-of-line processing within the same silicon substrate, excluding hybrid integration, flip-chip bonding, or any heterogeneous integration technique."

Lumos proposed a narrow construction that would limit "monolithically formed" to components created within the silicon substrate itself during standard semiconductor processing, explicitly excluding any form of hybrid integration, flip-chip bonding, or heterogeneous integration in which separately fabricated components are attached to the silicon die. Lumos relied heavily on the prosecution history of the '678 Patent, which involved three office action rejections and a narrowing amendment.

Court's Construction: "fabricated entirely within the same front-end-of-line and back-end-of-line semiconductor processing sequence on a single silicon substrate without heterogeneous bonding or attachment of separately fabricated optical components."

Analysis. The Court's construction is largely aligned with Lumos's proposal but expands the processing scope to encompass both front-end-of-line (FEOL) and back-end-of-line (BEOL) processing sequences, which is slightly broader than Lumos's original proposal (which referenced only FEOL). This modification reflects the Court's recognition that standard monolithic semiconductor fabrication includes both FEOL (device formation) and BEOL (interconnect formation) steps. However, the critical exclusion — "without heterogeneous bonding or attachment of separately fabricated optical components" — aligns precisely with Lumos's proposed exclusion of hybrid and heterogeneous integration techniques.

The Court's reasoning was grounded primarily in the prosecution history:

First, the Court found that the prosecution history of the '678 Patent was dispositive. The '678 Patent underwent three office action rejections during prosecution. The first two rejections were based on prior art references disclosing multi-channel WDM transceiver designs using hybrid integration approaches — specifically, designs in which WDM elements fabricated on separate substrates were bonded to silicon photonic dies. In response to the third office action, the applicants amended independent Claim 1 to add the "monolithically formed on a single silicon die" limitation, accompanied by remarks explicitly stating:

"The present invention is distinguished from the heterogeneous integration approaches of the cited references. In the claimed invention, the WDM element is monolithically formed — that is, fabricated as an integral part of the silicon die itself through standard semiconductor processing, rather than being separately fabricated and subsequently attached, bonded, or otherwise integrated onto the die through heterogeneous means."

The Court found these remarks to constitute a clear, unambiguous, and unequivocal disavowal of heterogeneous integration techniques. The Court held that having explicitly distinguished the claimed invention from "heterogeneous integration approaches" during prosecution, the applicants were bound by that distinction, and the Court would construe "monolithically formed" accordingly.

Second, the Court examined the specification and found it consistent with the prosecution history. The specification of the '678 Patent describes the advantages of monolithic integration as including "eliminating the alignment tolerances, yield losses, and packaging complexity associated with hybrid and heterogeneous integration approaches." The Court found that this language further supported the construction excluding heterogeneous techniques.

Third, the Court credited Dr. Ng's testimony that in the silicon photonics industry, "monolithic" integration is universally understood to be distinct from "heterogeneous" integration, and that a POSITA would not understand "monolithically formed" to encompass designs in which separately fabricated components are bonded or attached to the die.

Impact on Infringement Theory: Fatal. Lumos's 800G platform — the LumosBridge™ 800G Optical Engine — utilizes a heterogeneous integration architecture in which micro-transfer-printed III-V gain elements are bonded to the silicon photonic integrated circuit (PIC). This is precisely the type of heterogeneous integration technique that the Court's construction (and the prosecution history) explicitly excludes. The Lumos design involves separately fabricating III-V semiconductor components (specifically, indium phosphide-based gain elements) and then transferring and bonding them to the silicon PIC using micro-transfer printing technology — a well-known heterogeneous integration method. Under the Court's construction, which requires that the WDM element be "fabricated entirely within the same front-end-of-line and back-end-of-line semiconductor processing sequence on a single silicon substrate without heterogeneous bonding or attachment of separately fabricated optical components," Lumos's design is squarely outside the literal claim scope.

As with the '567 Patent, prosecution history estoppel under Festo bars any DOE argument. The "monolithically formed" limitation was added by amendment during prosecution specifically to distinguish heterogeneous integration approaches — the exact category of technology employed by Lumos. The Festo presumption of surrender applies, and none of the three recognized exceptions is viable. Heterogeneous integration was well known in the art at the time of the amendment (indeed, it was the subject of the prior art references being overcome), so the unforeseeability exception does not apply. The amendment directly addressed the distinction between monolithic and heterogeneous integration, so the tangential relation exception does not apply. And there is no "other reason" that would rebut the presumption.

2. Other '678 Patent Terms

The Court construed two additional terms from the '678 Patent:

"plurality of optical channels at distinct wavelengths" (Claim 1). OWS proposed a construction of "two or more optical channels, each operating at a different wavelength." Lumos proposed a construction of "four or more optical channels operating at wavelengths separated by at least 4.5 nm." The Court adopted a modified construction: "two or more optical channels, each operating at a wavelength that is spectrally distinct from the other channels." This construction was closer to OWS's position than to Lumos's, as the Court rejected Lumos's attempt to import a specific numerical minimum for the number of channels and the wavelength spacing from a preferred embodiment. This term was counted in our scorecard as the one "modified construction" that corresponds to neither party's verbatim proposal, though it is substantively closer to OWS's position.

"single silicon die" (Claim 1). Lumos proposed that this term required a die fabricated from a "bulk silicon" or "silicon-on-insulator" wafer, excluding any die that incorporates non-silicon substrate materials. OWS proposed that the term required only "a die having a silicon substrate." The Court adopted Lumos's construction, finding that "single silicon die" refers to a die fabricated from a silicon or silicon-on-insulator wafer and does not encompass dies with compound semiconductor substrates.

Neither of these secondary constructions independently affects the infringement analysis, which is controlled by the dispositive construction of "monolithically formed." The favorable modified construction on "plurality of optical channels at distinct wavelengths" is immaterial if the "monolithically formed" requirement cannot be met.

3. Net Assessment — '678 Patent

Infringement viability: Untenable. The Court's prosecution history-driven construction of "monolithically formed" categorically excludes Lumos's heterogeneous integration architecture, in which micro-transfer-printed III-V gain elements are bonded to the silicon PIC. Prosecution history estoppel prevents DOE arguments, as the limitation was added specifically to distinguish heterogeneous approaches during prosecution.

Recommendation: Abandon infringement theories under the '678 Patent.

D. U.S. Patent No. 11,312,890 (the '890 Patent) — "Driver Circuit Architecture for High-Speed Silicon Photonic Mach-Zehnder Modulators"

The '890 Patent is directed to a driver circuit architecture for high-speed Mach-Zehnder modulators (MZMs) used in silicon photonic transceivers, in which a pre-emphasis equalization stage is co-integrated with the modulator on a common substrate to achieve improved signal integrity at high baud rates. The patent claims priority to a non-provisional application filed on April 3, 2020, and was issued on May 17, 2022. The named inventor is Dr. Amara Osei. OWS asserted Claims 1, 6, 10, and 13 of the '890 Patent in this action, all of which depend from independent Claim 1. The Court construed three disputed terms from the '890 Patent.

1. Term: "pre-emphasis equalization stage electrically co-integrated with the modulator on a common substrate" (Claim 1)

This is the critical term for the '890 Patent. Claim 1 recites, inter alia, "a pre-emphasis equalization stage electrically co-integrated with the modulator on a common substrate, the equalization stage configured to compensate for frequency-dependent losses in the modulator's electrical and optical response."

OWS's Proposed Construction: "an equalization circuit located on the same substrate as the modulator."

OWS proposed a broad, location-based construction that would require only physical co-location on the same substrate, without imposing requirements on how the equalization circuit is fabricated or electrically connected to the modulator. OWS argued that "co-integrated" simply means "integrated together" and that the claim language does not specify the fabrication method or interconnection technique.

Lumos's Proposed Construction: "an equalization circuit fabricated in the same semiconductor process as the modulator, sharing the same doping profiles and process layers, and electrically connected without bond wires, interposers, or package-level routing."

Lumos proposed a narrow construction imposing three distinct requirements: (1) same-process fabrication, (2) shared doping profiles and process layers, and (3) direct electrical connection without bond wires, interposers, or package-level routing. Lumos argued that "co-integrated" is a term of art in semiconductor design denoting that two circuit elements are fabricated together in the same process, sharing the same layers and doping steps, and that a mere physical co-location (e.g., through flip-chip bonding of separately fabricated dies) does not constitute "co-integration."

Court's Construction: The Court adopted Lumos's proposed construction in full: "an equalization circuit fabricated in the same semiconductor process as the modulator, sharing the same doping profiles and process layers, and electrically connected without bond wires, interposers, or package-level routing."

Analysis. The Court's adoption of Lumos's construction in its entirety — without any modification — is notable and represents another decisive adverse ruling. The Court's reasoning proceeded as follows:

First, the Court examined the prosecution history of the '890 Patent. During prosecution, the claims were rejected twice, first under § 102 and then under § 103, based on prior art references disclosing driver circuits that were separately fabricated and then bonded to modulator dies. In response, the applicant amended independent Claim 1 to add the "co-integrated" limitation, accompanied by remarks distinguishing the prior art on the ground that "in the prior art, the driver circuit is a separately fabricated component that is subsequently bonded to the modulator die, whereas in the present invention, the pre-emphasis equalization stage is co-integrated with the modulator — i.e., fabricated as part of the same semiconductor process on the same substrate." The Court found that these remarks, while not as explicit as the prosecution history of the '567 and '678 Patents, nonetheless supported the narrow construction requiring same-process fabrication.

Second, the Court examined the specification and found that every described embodiment of "co-integration" involved an equalization circuit that was fabricated in the same process as the modulator, sharing the same doping profiles, and directly electrically connected without bond wires or other package-level interconnects. The specification described the advantages of co-integration as including "eliminating parasitic inductance and capacitance associated with bond wires and package-level interconnects," which the Court found further supported the requirement of direct electrical connection without package-level routing.

Third, the Court credited Dr. Ng's testimony that "co-integration" is a term of art in semiconductor design that is consistently understood to mean fabrication within the same process, and is distinguished from "co-packaging" or "hybrid integration" approaches in which separately fabricated dies are assembled together.

Impact on Infringement Theory: Severely weakened but not categorically foreclosed. Unlike the other three patents, the '890 Patent presents a slightly different posture — one in which literal infringement is untenable but a narrow DOE theory is not entirely foreclosed, though it is assessed as weak.

Lumos's driver architecture for the accused products employs a separate CMOS driver die that is flip-chip bonded to the silicon photonic modulator die via copper pillar micro-bumps. Under the Court's construction, this design falls outside the literal claim scope for two independent reasons:

(a) Different semiconductor processes and doping profiles. The Lumos driver is fabricated in a standard CMOS process (believed to be a 16nm or 12nm FinFET CMOS node), while the modulator is fabricated in a silicon photonic process with entirely different doping profiles and process layers. The two dies do not share the same semiconductor process, doping profiles, or process layers.

(b) Copper pillar micro-bump interconnection. The driver and modulator dies are electrically connected through copper pillar micro-bumps — a form of flip-chip bonding that constitutes package-level interconnection. This falls squarely within the category of "bond wires, interposers, or package-level routing" that the Court's construction requires be absent.

However, the prosecution history of the '890 Patent, while supporting the narrow construction, is slightly less definitive on the scope of DOE foreclosure than the prosecution histories of the '567 and '678 Patents. Specifically:

• The applicant's prosecution remarks distinguished the prior art on the basis that the prior art driver was "separately fabricated" and "subsequently bonded," but the remarks did not explicitly enumerate or address every possible bonding or integration technique that was being disclaimed. By contrast, the prosecution remarks for the '567 Patent explicitly addressed "segmented coupling" and the '678 Patent explicitly addressed "heterogeneous integration."

• The amendment adding "co-integrated" was made to overcome prior art disclosing wire-bonded driver configurations. Copper pillar micro-bump flip-chip bonding is a different — and arguably more advanced — interconnection technology than wire bonding. There is a theoretical argument that the scope of prosecution history surrender was limited to wire bonding and did not extend to all forms of flip-chip interconnection, although this argument is weak.

• Under Festo, the patentee bears the burden of rebutting the presumption of surrender. The "unforeseeability" exception might theoretically be invoked if copper pillar micro-bump technology were unforeseeable at the time of the amendment, but this argument is difficult to sustain given that flip-chip bonding with copper pillars was well known in the semiconductor industry by 2020 when the application was filed.

The residual DOE theory would proceed under the function-way-result test, arguing that Lumos's separate CMOS driver die connected via copper pillar micro-bumps performs substantially the same function (pre-emphasis equalization of the modulator drive signal), in substantially the same way (by applying a compensating transfer function to the electrical drive signal in close proximity to the modulator), to achieve substantially the same result (improved signal integrity at high baud rates with reduced frequency-dependent losses). While there is some plausibility to this argument at a high level of abstraction, it requires the factfinder to accept that a separately fabricated CMOS driver connected through package-level copper pillar interconnects is "equivalent" to a co-integrated driver fabricated in the same process with direct on-chip electrical connections — a proposition that faces substantial technical challenges, given the fundamental differences in process technology, parasitic characteristics, and interconnect architecture.

Estimated probability of success of the DOE theory at trial: 10–15% at best. This assessment reflects the difficulty of overcoming the prosecution history estoppel presumption, the technical dissimilarities between Lumos's design and the claimed co-integration, and the risk that a jury or court would view the DOE argument as an impermissible attempt to recapture subject matter surrendered during prosecution.

2. Other '890 Patent Terms

The Court construed two additional terms from the '890 Patent:

"pre-emphasis equalization" (Claim 1). OWS proposed a construction of "equalization applied to a signal before modulation to compensate for expected distortion." Lumos proposed a construction of "a feed-forward equalization technique in which the frequency response of the drive signal is pre-distorted based on a fixed, predetermined transfer function to compensate for known frequency-dependent losses." The Court adopted Lumos's construction, finding that the specification consistently described pre-emphasis equalization as a feed-forward, fixed-transfer-function technique, and that the claim should be limited accordingly. While this construction imposes additional requirements on the type of equalization that is claimed, it does not independently affect the infringement analysis because Lumos's driver does employ a pre-emphasis equalization technique with a predetermined transfer function, and this limitation would likely be met even under the narrower construction.

"common substrate" (Claim 1). OWS proposed a construction of "a single semiconductor substrate on which both the modulator and the equalization stage are present." Lumos proposed a construction of "a single monolithic semiconductor substrate in which both the modulator and the equalization stage are fabricated." The Court adopted OWS's proposed construction, finding that "common substrate" simply means a single substrate on which both components are present and does not independently require monolithic fabrication (which is already required by the "co-integrated" limitation).

OWS's victory on the "common substrate" term represents one of the two terms on which OWS prevailed overall. However, this victory does not rehabilitate the infringement case. Even under OWS's broader construction of "common substrate," the independent "co-integrated" construction requires same-process fabrication and direct electrical connection without package-level routing — requirements that Lumos's design does not meet.

3. Net Assessment — '890 Patent

Infringement viability: Significantly weakened. Literal infringement is untenable under the Court's construction of "co-integrated." A DOE theory is theoretically possible but assessed as weak, with an estimated probability of success of no more than ten to fifteen percent. The DOE argument would require expert testimony establishing that copper pillar micro-bump flip-chip bonding of a separately fabricated CMOS driver die is equivalent to true monolithic co-integration in terms of function, way, and result — a technically difficult proposition given the fundamental differences in process technology, doping profiles, and interconnect architecture. This argument would also need to overcome the Festo presumption of prosecution history estoppel, which, while slightly less definitive for the '890 Patent than for the other three patents, remains a significant obstacle.

The '890 Patent is the only patent with any residual infringement theory, however slim. Standing alone, it cannot support the damages case originally contemplated, which was premised on all four patents and a 7.5% reasonable royalty rate.

Recommendation: If the decision is made to continue litigation, the '890 Patent is the only patent on which any infringement theory survives. However, as discussed in Sections IV and VI below, the expected value of proceeding on this theory alone is negative. We do not recommend continuing litigation solely on the '890 Patent DOE theory.

IV. Impact on Damages Theories

The Markman Order fundamentally undermines OWS's damages case. This section analyzes the impact on each element of the pre-Markman damages framework.

A. Reasonable Royalty Analysis

Prior to the Markman Order, the damages framework prepared for presentation through Dr. Lawrence Kessler of Apex Claim Analytics LLC was premised on the following elements:

• A reasonable royalty rate of 7.5%, derived from a hypothetical negotiation analysis incorporating all fifteen Georgia-Pacific factors and reflecting the combined value of the four asserted patents;

• A royalty base of approximately $148.6 million, representing Lumos's combined accused product revenue for the LumosBridge™ 800G Optical Engine and the LumosEdge™ 400ZR+ Pluggable Transceiver for calendar year 2022;

• Pre-suit damages of approximately $62.3 million, representing accused product revenue for the first and second quarters of 2022 (prior to the June 3, 2022 filing date) on which OWS sought pre-suit damages based on the March 15, 2022 cease-and-desist letter providing constructive notice;

• A willfulness enhancement of up to three times under 35 U.S.C. § 284, based on Lumos's continued sales after receiving the cease-and-desist letter; and

• A total claimed damages figure exceeding $75 million.

With three of the four patents-in-suit effectively eliminated by the Markman Order, this damages framework collapses. The reasonable royalty rate of 7.5% was predicated on the combined value of four patents covering different aspects of the accused products' technology — waveguide coupling, photodetector architecture, WDM integration, and driver circuits. A single surviving patent claim with a weak DOE theory cannot support a royalty rate anywhere near 7.5%. Under the Georgia-Pacific framework, the hypothetical negotiation would be fundamentally different: a willing licensor and willing licensee negotiating over a single patent with questionable infringement coverage would arrive at a dramatically lower royalty rate.

Based on our preliminary reassessment, a royalty rate supportable by a single weak patent infringement theory would likely be in the range of 0.5% to 1.0% at most. Moreover, the royalty base would need to be reconsidered, as the '890 Patent covers only the driver circuit architecture — not the waveguide coupling, photodetector, or WDM integration aspects of the accused products — and an apportionment analysis would be required to isolate the value attributable to the driver circuit component. A reasonable apportioned royalty base might be in the range of $25 to $40 million, reflecting the fractional value of the driver circuit relative to the overall accused product.

Combining a reduced royalty rate (0.5%–1.0%) with a reduced, apportioned royalty base ($25–$40 million), the maximum reasonable royalty recovery on the '890 Patent alone would be in the range of $125,000 to $400,000 — before consideration of the low probability of establishing infringement through the DOE theory.

The Georgia-Pacific factor analysis (draft prepared by Dr. Ke

PRIVILEGED AND CONFIDENTIAL ATTORNEY WORK PRODUCT — ATTORNEY-CLIENT COMMUNICATION
This memorandum is a privileged and confidential attorney work product prepared in anticipation of litigation and constitutes a protected attorney-client communication. This document and its contents are not to be disclosed to any person or entity outside the engagement team and approved distribution list without the express written authorization of the Responsible Partner. Any unauthorized disclosure may result in waiver of privilege.
February 21, 2023
INTERNAL MEMORANDUM
To: File — Matter No. 1014-00003
From: Alan Ngo, Responsible Partner Omar Siddiqui, Lead Associate Christy Bauer, Associate Intellectual Property Group, Calderwood & Harkness LLP
Re: Post-Markman Impact Assessment — Optiwave Semiconductor Corp. v. Lumos Photonics, Inc., Case No. 5:22-cv-03417-EJD (N.D. Cal.)
Distribution: Limited to engagement team members Alan Ngo, Marcus Delgado, Omar Siddiqui, and Christy Bauer. This memorandum may be shared with client representatives Sandra Ellison (General Counsel, Optiwave Semiconductor Corp.), Ted Marchetti (Chief Financial Officer, Optiwave Semiconductor Corp.), and Dr. Priya Ramanathan (Chief Technology Officer and Co-Founder, Optiwave Semiconductor Corp.), as well as representatives of Vantage Point Capital Partners, including Douglas Fairbanks III (Managing Director) and Rebecca Torres (Principal), at the discretion of the Responsible Partner.
Purpose: This memorandum provides a comprehensive assessment of the Markman Order issued by the Honorable Edward J. Torvelis on February 17, 2023, in the above-captioned matter. The memorandum analyzes the impact of the Court's claim constructions on Optiwave Semiconductor Corp.'s ("OWS") infringement theories for all four patents-in-suit, evaluates the remaining viability of those theories, assesses appellate options, and sets forth recommendations for the path forward, including a cost-benefit analysis of continued litigation versus settlement.
Right-click to update Table of Contents
I. Executive Summary
On February 17, 2023, the Honorable Edward J. Torvelis of the United States District Court for the Northern District of California issued a thirty-four-page Markman Order construing all eleven disputed claim terms across the four patents asserted by OWS in this action: 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"). The Order followed a six-and-a-half-hour claim construction hearing held on January 30, 2023, at which both parties presented technical tutorials and oral argument.
Overall Scorecard. The Court adopted Defendant Lumos Photonics, Inc.'s ("Lumos") proposed construction — or a construction substantially aligned with Lumos's position — on eight of the eleven disputed claim terms. The Court adopted Plaintiff OWS's proposed construction on two terms, and adopted a modified construction that did not correspond verbatim to either party's proposal on one term. This outcome represents a decisive defeat for OWS on the claim construction issues that are most material to its infringement theories.
Bottom-Line Assessment. The Court's constructions render OWS's infringement theories untenable as to three of the four patents-in-suit — the '567 Patent, the '234 Patent, and the '678 Patent — and significantly weakened as to the fourth, the '890 Patent. Specifically, the constructions adopted for the four critical claim limitations at issue — "adiabatically coupled" (the '567 Patent), "continuously varying germanium concentration" (the '234 Patent), "monolithically formed" (the '678 Patent), and "co-integrated" (the '890 Patent) — effectively exclude the accused Lumos products, namely the LumosBridge™ 800G Optical Engine and the LumosEdge™ 400ZR+ Pluggable Transceiver, from the literal scope of the asserted claims. In each instance, the Court's construction introduces specificity and narrowness that maps precisely to technical distinctions between OWS's patented designs and Lumos's accused products.
For the three patents where infringement theories are untenable, the prosecution history underlying the Court's constructions further forecloses any viable doctrine-of-equivalents ("DOE") theory, as the narrowing limitations were in each case added by amendment during prosecution to overcome prior art, triggering the presumption of prosecution history estoppel under Festo Corp. v. Fenmore & Co., 535 U.S. 722 (2002). For the '890 Patent, while a narrow theoretical DOE argument may survive, it is assessed as weak, with an estimated probability of success at trial of no more than ten to fifteen percent.
Recommendation. Based on the analysis set forth in this memorandum, we recommend that OWS initiate settlement discussions with Lumos at the earliest opportunity. We further recommend that this impact assessment be presented to the OWS Board of Directors and to Vantage Point Capital Partners at a Board meeting to be scheduled for February 24, 2023, at which formal settlement authority should be requested. Our recommended settlement approach is a cross-license with a nominal royalty payment from Lumos as the opening position, with authority to accept a walk-away settlement with mutual releases as an acceptable fallback outcome.
II. Background
A. The Litigation
Optiwave Semiconductor Corp. commenced this action by filing a Complaint for Patent Infringement on June 3, 2022, in the United States District Court for the Northern District of California, San Jose Division. The Complaint asserted infringement of four United States patents owned by OWS, each directed to different aspects of high-speed optical transceiver technology utilized in data center interconnect applications:
•  U.S. Patent No. 10,234,567 (the "'567 Patent"), entitled "Integrated Waveguide Coupler with Tapered Mode Converter for Silicon Photonics Platforms," which covers integrated waveguide coupling architectures employing tapered mode converters for efficient optical coupling in silicon photonic integrated circuits;
•  U.S. Patent No. 10,891,234 (the "'234 Patent"), entitled "Germanium-on-Silicon Photodetector with Graded Buffer Layer for Enhanced Responsivity," which covers germanium-on-silicon photodetector architectures utilizing graded buffer layers to achieve enhanced responsivity at telecommunications wavelengths;
•  U.S. Patent No. 11,045,678 (the "'678 Patent"), entitled "Multi-Channel Optical Transceiver with Integrated Wavelength-Division Multiplexing on Silicon," which covers multi-channel WDM integration architectures in which wavelength-division multiplexing elements are monolithically formed on a single silicon die; and
•  U.S. Patent No. 11,312,890 (the "'890 Patent"), entitled "Driver Circuit Architecture for High-Speed Silicon Photonic Mach-Zehnder Modulators," which covers MZM driver circuit architectures in which a pre-emphasis equalization stage is co-integrated with the modulator on a common substrate.
The Complaint identified two Lumos products as infringing each of the four patents-in-suit:
•  LumosBridge™ 800G Optical Engine — a co-packaged optics engine designed for deployment in hyperscale data center switches. This product first shipped to customers in the first quarter of 2022 and carries a unit price in the range of $1,200 to $1,800, depending on configuration and volume; and
•  LumosEdge™ 400ZR+ Pluggable Transceiver — a QSFP-DD form factor pluggable transceiver module designed for 400ZR+ coherent data center interconnect applications. This product has been in production since the third quarter of 2021 and carries a unit price in the range of $600 to $900.
Combined accused product revenue for Lumos for calendar year 2022 was approximately $148.6 million, based on publicly available financial disclosures and information obtained through fact discovery. OWS's damages claim, as originally formulated, exceeded $75 million, premised on a 7.5% reasonable royalty rate applied to the accused product revenue base, inclusive of pre-suit damages for the period in which the accused products were sold prior to the filing of the Complaint and a willfulness enhancement of up to three times under 35 U.S.C. § 284.
Lumos filed its Answer and Counterclaims on August 5, 2022, denying infringement and asserting counterclaims for declaratory judgment of non-infringement and invalidity of all four patents-in-suit under 35 U.S.C. §§ 102, 103, and 112.
B. Claim Construction Proceedings
The claim construction proceedings in this action followed the Patent Local Rules for the Northern District of California and proceeded on the following timeline:
•  September 12, 2022: The parties exchanged Preliminary Claim Construction Contentions pursuant to Patent Local Rule 4-1, identifying the claim terms each party contended required construction and setting forth preliminary proposed constructions.
•  October 3, 2022: The parties exchanged Responsive Claim Construction Contentions pursuant to Patent Local Rule 4-2.
•  October 14, 2022: The parties filed a Joint Claim Construction and Prehearing Statement pursuant to Patent Local Rule 4-3, identifying the eleven disputed claim terms across all four patents-in-suit that would require resolution by the Court, together with each party's proposed constructions and supporting evidence.
•  October 31, 2022: Claim construction discovery closed. During the discovery period, both parties deposed the named inventors of the patents-in-suit, including depositions of Dr. Kevin Fong (co-inventor on the '567, '234, and '678 Patents), Hideo Tanaka (co-inventor on the '567 Patent), and Dr. Amara Osei (named inventor on the '890 Patent). These depositions addressed, among other things, the understanding of a person of ordinary skill in the art regarding the disputed claim terms, the prosecution history of each patent, and the state of the art at the time of the inventions.
•  December 5, 2022: OWS filed its Opening Claim Construction Brief (45 pages), setting forth OWS's proposed constructions for all eleven disputed terms, supported by intrinsic evidence (specification and prosecution history), extrinsic evidence (technical literature, dictionary definitions, and expert declarations), and the declaration of Dr. Harold Whitfield.
•  December 19, 2022: Lumos filed its Responsive Claim Construction Brief (45 pages), setting forth Lumos's proposed constructions and opposing OWS's proposals, supported by comparable evidence and the declaration of Dr. Raymond Ng.
•  January 9, 2023: OWS filed its Reply Brief in support of its proposed constructions (15 pages).
•  January 19, 2023: Lumos filed a Sur-Reply Brief by leave of Court (15 pages), responding to new arguments raised in OWS's Reply.
•  January 30, 2023: The Court held a Markman hearing lasting approximately six and a half hours. Each party presented a technical tutorial to the Court, followed by oral argument on each disputed term. Dr. Harold Whitfield of Whitfield Technical Consulting Group, Inc. (Palo Alto, California) testified as OWS's technical tutorial witness, providing an overview of silicon photonics technology and explaining OWS's proposed constructions. Dr. Raymond Ng, Of Counsel at Brannick Howell & Stein LLP and former Professor of Electrical Engineering at the University of Colorado Boulder, testified as Lumos's technical tutorial witness. The Court posed extensive questions to both tutorial witnesses and to counsel for both parties throughout the hearing.
•  February 17, 2023: The Court issued its Markman Order (34 pages), construing all eleven disputed claim terms. The Order is the subject of this memorandum.
C. The Engagement Team and Expert Advisors
The following personnel are engaged on this matter:
Calderwood & Harkness LLP Litigation Team:
•  Alan Ngo — Responsible Partner. Mr. Ngo leads the firm's semiconductor and photonics patent litigation practice and has over twenty-five years of experience in complex patent litigation in the Northern District of California and the International Trade Commission. Mr. Ngo is the primary point of contact for the client and for Vantage Point Capital Partners.
•  Marcus Delgado — Billing Partner. Mr. Delgado serves as the billing partner on this engagement and provides strategic oversight on litigation budgeting and resource allocation.
•  Omar Siddiqui — Lead Associate. Mr. Siddiqui manages the day-to-day conduct of the litigation, including discovery, claim construction briefing, and expert coordination. Mr. Siddiqui holds a B.S. in Electrical Engineering from Stanford University and a J.D. from Georgetown University Law Center.
•  Christy Bauer — Working Associate. Ms. Bauer assists with legal research, brief drafting, document review, and prosecution history analysis. Ms. Bauer holds a B.S. in Physics from MIT and a J.D. from the University of Virginia School of Law.
Technical Expert:
•  Dr. Harold Whitfield — Retained through Whitfield Technical Consulting Group, Inc. (Palo Alto, California). Dr. Whitfield holds a Ph.D. in Electrical Engineering with a specialization in photonics from the University of California, Berkeley. He is Professor Emeritus of Electrical and Computer Engineering at the University of California, Santa Barbara, where he directed the Integrated Photonics Laboratory for over twenty years. Dr. Whitfield has authored or co-authored over 200 publications in the field of integrated photonics and has served as a testifying expert in over thirty patent cases involving semiconductor and photonics technologies.
Damages Expert:
•  Dr. Lawrence Kessler — Retained through Apex Claim Analytics LLC (Chicago, Illinois). Dr. Kessler holds a Ph.D. in Economics from Northwestern University and specializes in damages quantification in patent infringement matters, with particular expertise in reasonable royalty calculations for semiconductor technologies.
Defendant's Counsel:
•  Victoria "Tori" Ashworth — Lead Partner, Brannick Howell & Stein LLP (San Francisco office). Ms. Ashworth leads Lumos's defense team and has a well-established reputation as one of the most effective patent defense litigators in the Northern District of California.
•  Daniel Kowalski — Associate Team Lead, Brannick Howell & Stein LLP (San Francisco office). Mr. Kowalski serves as the associate team lead on the defense side and has handled the majority of claim construction briefing for Lumos.
III. Term-by-Term Analysis of the Markman Order
This section provides a detailed analysis of the Court's claim construction rulings as set forth in the February 17, 2023 Markman Order, organized by patent. For each patent, we analyze the critical disputed claim term that is dispositive of OWS's infringement theory, address any secondary terms construed by the Court, and provide an overall assessment of the viability of OWS's infringement position. A complete summary table covering all eleven disputed terms is provided in Appendix A.
A. U.S. Patent No. 10,234,567 (the '567 Patent) — "Integrated Waveguide Coupler with Tapered Mode Converter for Silicon Photonics Platforms"
The '567 Patent is directed to integrated waveguide coupling structures that use tapered mode converters to efficiently couple light between optical fibers and silicon photonic waveguides. The patent claims priority to a provisional application filed on March 14, 2017, and was issued on March 19, 2019. The named inventors are Dr. Kevin Fong and Hideo Tanaka. OWS asserted Claims 1, 4, 7, 12, and 15 of the '567 Patent in this action, all of which depend from independent Claim 1. The Court construed two disputed terms from the '567 Patent.
1. Term: "adiabatically coupled" (Claim 1)
This is the critical term for the '567 Patent. Claim 1 of the '567 Patent recites, inter alia, "a first waveguide and a second waveguide that are adiabatically coupled through a tapered mode converter region." The meaning of "adiabatically coupled" was the most extensively briefed and argued term in the Markman proceedings, occupying approximately ninety minutes of the January 30 hearing.
OWS's Proposed Construction: "coupled such that optical mode transitions occur without significant power loss."
OWS argued that the term "adiabatically coupled" should be given its broadest reasonable meaning consistent with the specification, which OWS contended was a functional definition focused on the performance characteristic of the coupling — namely, that optical mode transitions occur without significant power loss. OWS relied on the ordinary meaning of "adiabatic" in physics (a process occurring without exchange of energy with the environment) and on portions of the specification describing the goal of the invention as "achieving efficient, low-loss coupling" between waveguides. OWS further argued that the specification disclosed multiple embodiments of structures that could achieve adiabatic coupling, and that the term should not be limited to any single structural implementation.
Lumos's Proposed Construction: "coupled through a gradual, continuous taper that prevents excitation of higher-order modes throughout the entire transition region."
Lumos argued that "adiabatically coupled" had a specific structural meaning to a person of ordinary skill in the art in the field of silicon photonics, requiring a gradual and continuous taper — not merely a low-loss coupling. Lumos relied heavily on the prosecution history of the '567 Patent, pointing to a narrowing amendment made during prosecution in which the applicants added the "adiabatically coupled" limitation to overcome a rejection based on the Tanaka prior art reference, U.S. Patent No. 9,456,012, issued September 6, 2016. That prior art reference — ironically co-invented by Hideo Tanaka before he joined OWS — disclosed a non-continuous, segmented coupling approach. Lumos argued that by adding "adiabatically coupled" to distinguish over the Tanaka reference, the applicants disclaimed non-continuous coupling designs and adopted a construction limited to gradual, continuous tapers. Lumos further relied on the testimony of Dr. Raymond Ng, who opined that a person of ordinary skill in the art in silicon photonics would understand "adiabatic" coupling to require a continuous, gradual taper structure that avoids exciting higher-order modes at any point along the transition.
Court's Construction: "coupled through a gradual, continuous taper designed and configured to prevent excitation of higher-order modes over the entire transition length."
Analysis. The Court adopted Lumos's proposed construction with only minor wording modifications. Specifically, the Court substituted "designed and configured" for "that prevents" and "over the entire transition length" for "throughout the entire transition region." These modifications are stylistic rather than substantive; the constructed meaning is functionally identical to Lumos's proposal. In its analysis, the Court made the following key findings:
First, the Court examined the prosecution history and found it dispositive. During prosecution of the '567 Patent, the claims were rejected twice over the Tanaka prior art (U.S. Patent No. 9,456,012). In response to the second rejection, the applicants amended independent Claim 1 to add the "adiabatically coupled" limitation and submitted remarks arguing that the claimed invention was "fundamentally distinguished from Tanaka's segmented coupling approach" because "the present invention employs an adiabatic transition in which the taper is continuous and gradual, ensuring that no higher-order modes are excited at any point during the transition." The Court found that these remarks constituted a clear and unambiguous disavowal of non-continuous coupling structures.
Second, the Court examined the specification and found that every embodiment described in the specification consistently depicted adiabatic coupling in terms of continuous, gradual tapers. While OWS argued that the specification's description of the "goal" of adiabatic coupling (low-loss transitions) supported a functional construction, the Court found that the structural descriptions in the specification were consistent with — and reinforced — the narrower construction.
Third, the Court credited Dr. Ng's testimony that a person of ordinary skill in the art in the field of silicon photonics, as of the priority date, would understand "adiabatic" coupling to refer specifically to coupling through a continuous, gradual geometric transition that prevents mode conversion, and would not understand the term to encompass segmented or discontinuous coupling approaches that might also achieve low loss through other mechanisms.
Fourth, the Court declined to credit Dr. Whitfield's broader construction. While acknowledging Dr. Whitfield's credentials and expertise, the Court noted that the intrinsic evidence — particularly the prosecution history — was clear and unambiguous, and that resort to extrinsic evidence to broaden the claim term beyond what the intrinsic evidence supports is impermissible under Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005) (en banc).
Impact on Infringement Theory: Fatal. The Markman construction of "adiabatically coupled" is dispositive of OWS's infringement theory for the '567 Patent. It is undisputed — and confirmed through OWS's own reverse engineering analysis of the accused products — that the LumosBridge™ 800G Optical Engine employs a segmented (non-continuous) taper design with discrete coupling sections. Specifically, the LumosBridge™ waveguide coupler utilizes a multi-segment inverse taper structure in which the waveguide width changes in discrete steps across multiple lithographically defined sections, rather than a single continuous taper. Under the Court's construction, which requires a "gradual, continuous taper," Lumos's segmented design falls squarely outside the literal scope of Claim 1.
Moreover, a doctrine-of-equivalents argument is foreclosed by prosecution history estoppel. Because the "adiabatically coupled" limitation was added by amendment during prosecution specifically to overcome the Tanaka prior art — which disclosed a non-continuous coupling approach — the Festo presumption applies. Under Festo Corp. v. Fenmore & Co., 535 U.S. 722 (2002), a narrowing amendment made to satisfy any requirement of the Patent Act gives rise to a presumption that the patentee has surrendered all equivalents between the original claim scope and the amended claim scope. Lumos's segmented taper design is precisely the type of structure that was disclaimed during prosecution; applying DOE to recapture it would directly contradict the prosecution history. None of the three recognized exceptions to the Festo presumption — unforeseeability, tangential relation, or "some other reason" — is applicable here, as segmented coupling structures were well known in the art at the time of the amendment and were the specific subject matter the amendment was designed to overcome.
2. Other '567 Patent Terms
The Court also construed one additional term from the '567 Patent: "tapered mode converter region" (Claim 1). The parties disputed whether this term required a specific geometric taper profile (Lumos's position, proposing a construction requiring a "linear or adiabatic taper profile") or whether it encompassed any region in which mode conversion occurs through a geometric transition (OWS's position). The Court adopted OWS's proposed construction on this term, finding that the specification described multiple taper profiles and that the claim language did not require a specific profile type.
While OWS prevailed on this secondary term, the victory is immaterial. Because the "adiabatically coupled" limitation appears in independent Claim 1 and is present in or inherited by every asserted claim (Claims 1, 4, 7, 12, and 15), the Court's construction of "adiabatically coupled" is dispositive for the entire set of asserted claims from the '567 Patent. A favorable construction of "tapered mode converter region" cannot rescue an infringement theory that fails on the "adiabatically coupled" limitation.
3. Net Assessment — '567 Patent
Infringement viability: Untenable. The Court's construction of "adiabatically coupled" as requiring a "gradual, continuous taper designed and configured to prevent excitation of higher-order modes over the entire transition length" excludes the accused products from the literal scope of every asserted claim of the '567 Patent. The LumosBridge™ product's segmented taper design is categorically outside this construction. Prosecution history estoppel under Festo further forecloses any meaningful DOE theory, as the limitation was added specifically to overcome prior art disclosing the very type of non-continuous coupling that Lumos employs.
Recommendation: Abandon infringement theories under the '567 Patent. In the event OWS elects to pursue settlement discussions, the '567 Patent should not be identified as a patent on which OWS intends to press its infringement claims.
B. U.S. Patent No. 10,891,234 (the '234 Patent) — "Germanium-on-Silicon Photodetector with Graded Buffer Layer for Enhanced Responsivity"
The '234 Patent is directed to germanium-on-silicon (Ge-on-Si) photodetector architectures that employ a graded buffer layer between the silicon substrate and the germanium absorber layer to reduce threading dislocation density and enhance photodetector responsivity. The patent claims priority to a provisional application filed on July 22, 2018, and was issued on January 12, 2021. The named inventor is Dr. Kevin Fong. OWS asserted Claims 1, 3, 8, and 14 of the '234 Patent in this action, all of which depend from independent Claim 1. The Court construed two disputed terms from the '234 Patent.
1. Term: "graded buffer layer having a continuously varying germanium concentration" (Claim 1)
This is the critical term for the '234 Patent. Claim 1 recites, inter alia, "a graded buffer layer disposed between the silicon substrate and the germanium absorber layer, the graded buffer layer having a continuously varying germanium concentration."
OWS's Proposed Construction: "a buffer layer in which the germanium concentration changes across the layer thickness."
OWS proposed a broad construction that would encompass any buffer layer in which the germanium concentration is not uniform — that is, any layer in which the concentration "changes" from one level to another, whether by continuous gradient, step gradient, or any other variation. OWS argued that the specification described both "continuously graded" and "step-graded" buffer layer embodiments and that the claim language was broad enough to encompass both. OWS further argued that a POSITA would understand "continuously varying" to mean "varying across the layer," rather than "varying without interruption," relying on certain dictionary definitions of "continuously" that emphasize duration or extent rather than smoothness.
Lumos's Proposed Construction: "a buffer layer in which the germanium atomic fraction increases monotonically and without step changes from the silicon substrate interface to the absorber layer."
Lumos argued that "continuously varying" had a clear and well-understood technical meaning to a POSITA in the field of semiconductor epitaxy, requiring a smooth, monotonic change in germanium concentration without discrete steps. Lumos pointed to the claim language itself — "continuously varying" — and contrasted it with the "step-graded" embodiments described in the specification but not claimed. Lumos argued that if the applicant had intended to encompass step-graded designs, the applicant could have used broader language (such as "varying" or "graded") rather than "continuously varying."
Court's Construction: "a buffer layer in which the germanium atomic fraction increases monotonically and without step changes from the silicon substrate interface to the absorber layer."
The Court adopted Lumos's proposed construction verbatim. This represents the most straightforward and complete adverse ruling of the eleven disputed terms.
Analysis. The Court's reasoning proceeded along two principal lines:
First, the Court examined the claim language and found that "continuously varying" has a well-understood plain technical meaning. The Court noted that in the field of semiconductor epitaxy and materials science, "continuous" variation is routinely distinguished from "step" or "discrete" variation, and that a POSITA would understand "continuously varying germanium concentration" to require a smooth gradient without discrete steps. The Court cited several technical references from the intrinsic and extrinsic record confirming this usage, including the IEEE Photonics Technology Letters article by Chen et al. (2015) and the textbook by Kasper and Lyutovich, Properties of Silicon Germanium and SiGe:Carbon (INSPEC, 2000), both of which distinguish "continuously graded" from "step-graded" SiGe buffer layers.
Second, and critically, the Court addressed OWS's argument that the specification's description of both continuously graded and step-graded embodiments supported a broader construction. The Court rejected this argument, holding that while the specification described step-graded designs as part of the background and as alternative embodiments, the claims used only "continuously varying" language. The Court emphasized that it is a fundamental principle of claim construction that claims are not broadened to encompass unclaimed embodiments merely because those embodiments are described in the specification. The Court cited Phillips, 415 F.3d at 1323, and Ashford Enterprises v. Dunmore Corporation, 358 F.3d 898, 906 (Fed. Cir. 2004), for the proposition that "the specification may describe unclaimed embodiments without expanding the scope of the claims." The Court found that the applicant made a deliberate drafting choice to claim only "continuously varying" concentrations and not step-graded designs, and the Court would enforce that choice.
The Court's reasoning on this point deserves particular attention because it illustrates a patent drafting deficiency that cannot be remedied at this stage of the litigation. The specification of the '234 Patent describes step-graded buffer layers in some detail, and the named inventor, Dr. Kevin Fong, testified at his deposition that he considered both continuous and step-graded approaches to be within the scope of his invention. However, the claim language as drafted is limited to "continuously varying" concentrations, and the Court properly held that the claims control.
Impact on Infringement Theory: Fatal. The accused Lumos products employ a step-graded SiGe buffer layer with four discrete composition steps, in which the germanium atomic fraction increases from 0.25 to 0.50 to 0.75 to 1.0 at defined interfaces. This step-graded structure is precisely the type of design that the Court's construction excludes through the "without step changes" language. There is no colorable literal infringement argument under this construction.
A DOE argument is likewise foreclosed. To argue that a four-step graded buffer is equivalent to a continuously varying (i.e., no step changes) buffer would directly vitiate the "without step changes" limitation, which is impermissible. Under Ridgeway & Co. v. Conceptronic, Inc., 90 F.3d 1576, 1583 (Fed. Cir. 1996), and its progeny, a DOE theory that would effectively eliminate a claim limitation is barred by the doctrine of claim vitiation. Here, applying DOE to encompass a design with four discrete composition steps would render the "without step changes" requirement meaningless, and no court would permit such an argument.
2. Other '234 Patent Terms
The Court also construed one additional term from the '234 Patent: "enhanced responsivity" (Claim 1). OWS proposed a construction defining this as "responsivity greater than that of a comparable photodetector without a graded buffer layer," while Lumos proposed a construction requiring "responsivity of at least 0.9 A/W at 1310 nm," based on a specific performance figure recited in the preferred embodiment. The Court adopted OWS's proposed construction, finding that the specific responsivity figure cited by Lumos was from a preferred embodiment and should not be imported as a limitation into the claims.
This was the second of the two terms on which OWS prevailed in the Markman proceedings. However, as with the "tapered mode converter region" term in the '567 Patent, this victory is immaterial because the dispositive "continuously varying germanium concentration" construction eliminates OWS's infringement theory at the threshold of independent Claim 1, from which all asserted claims depend.
3. Net Assessment — '234 Patent
Infringement viability: Untenable. The Court's verbatim adoption of Lumos's proposed construction for "graded buffer layer having a continuously varying germanium concentration," requiring a monotonic increase "without step changes," categorically excludes Lumos's step-graded photodetector design. No viable literal infringement or DOE theory survives this construction.
The particular significance of this outcome merits emphasis: the specification of the '234 Patent itself described step-graded buffer layer embodiments, but the applicant chose claim language limited to "continuously varying" concentrations. This drafting choice — which may have been inadvertent — was enforced by the Court as a deliberate claim scope limitation. This underscores the importance of claim drafting precision and the risk of relying on specification disclosures that are not captured in the claim language.
Recommendation: Abandon infringement theories under the '234 Patent. As with the '567 Patent, the '234 Patent should not be a basis for continued infringement assertions in any settlement negotiation.
C. U.S. Patent No. 11,045,678 (the '678 Patent) — "Multi-Channel Optical Transceiver with Integrated Wavelength-Division Multiplexing on Silicon"
The '678 Patent is directed to multi-channel optical transceiver architectures in which a wavelength-division multiplexing (WDM) element is monolithically integrated on a single silicon die alongside other photonic components, enabling higher-density channel integration and reduced packaging complexity. The patent claims priority to a non-provisional application filed on November 8, 2019, and was issued on June 29, 2021. The named inventors are Dr. Kevin Fong and Dr. Amara Osei. OWS asserted Claims 1, 5, 9, 11, and 18 of the '678 Patent in this action, all of which depend from independent Claim 1. The Court construed three disputed terms from the '678 Patent.
1. Term: "integrated wavelength-division multiplexing element monolithically formed on a single silicon die" (Claim 1)
This is the critical term for the '678 Patent. Claim 1 recites, inter alia, "an integrated wavelength-division multiplexing element monolithically formed on a single silicon die, the WDM element configured to combine or separate a plurality of optical channels at distinct wavelengths."
OWS's Proposed Construction: "a WDM component fabricated as part of the same semiconductor fabrication process on a single die."
OWS proposed a construction focused on the fabrication process, arguing that "monolithically formed" required only that the WDM component be fabricated as part of the same semiconductor process flow used to create the rest of the photonic integrated circuit on the die. OWS argued that this construction was consistent with common usage of "monolithic" in the semiconductor industry and would encompass designs where some components (such as III-V gain elements) are integrated onto the silicon die during the fabrication process, even if different materials are involved.
Lumos's Proposed Construction: "a WDM component formed during front-end-of-line processing within the same silicon substrate, excluding hybrid integration, flip-chip bonding, or any heterogeneous integration technique."
Lumos proposed a narrow construction that would limit "monolithically formed" to components created within the silicon substrate itself during standard semiconductor processing, explicitly excluding any form of hybrid integration, flip-chip bonding, or heterogeneous integration in which separately fabricated components are attached to the silicon die. Lumos relied heavily on the prosecution history of the '678 Patent, which involved three office action rejections and a narrowing amendment.
Court's Construction: "fabricated entirely within the same front-end-of-line and back-end-of-line semiconductor processing sequence on a single silicon substrate without heterogeneous bonding or attachment of separately fabricated optical components."
Analysis. The Court's construction is largely aligned with Lumos's proposal but expands the processing scope to encompass both front-end-of-line (FEOL) and back-end-of-line (BEOL) processing sequences, which is slightly broader than Lumos's original proposal (which referenced only FEOL). This modification reflects the Court's recognition that standard monolithic semiconductor fabrication includes both FEOL (device formation) and BEOL (interconnect formation) steps. However, the critical exclusion — "without heterogeneous bonding or attachment of separately fabricated optical components" — aligns precisely with Lumos's proposed exclusion of hybrid and heterogeneous integration techniques.
The Court's reasoning was grounded primarily in the prosecution history:
First, the Court found that the prosecution history of the '678 Patent was dispositive. The '678 Patent underwent three office action rejections during prosecution. The first two rejections were based on prior art references disclosing multi-channel WDM transceiver designs using hybrid integration approaches — specifically, designs in which WDM elements fabricated on separate substrates were bonded to silicon photonic dies. In response to the third office action, the applicants amended independent Claim 1 to add the "monolithically formed on a single silicon die" limitation, accompanied by remarks explicitly stating:
"The present invention is distinguished from the heterogeneous integration approaches of the cited references. In the claimed invention, the WDM element is monolithically formed — that is, fabricated as an integral part of the silicon die itself through standard semiconductor processing, rather than being separately fabricated and subsequently attached, bonded, or otherwise integrated onto the die through heterogeneous means."
The Court found these remarks to constitute a clear, unambiguous, and unequivocal disavowal of heterogeneous integration techniques. The Court held that having explicitly distinguished the claimed invention from "heterogeneous integration approaches" during prosecution, the applicants were bound by that distinction, and the Court would construe "monolithically formed" accordingly.
Second, the Court examined the specification and found it consistent with the prosecution history. The specification of the '678 Patent describes the advantages of monolithic integration as including "eliminating the alignment tolerances, yield losses, and packaging complexity associated with hybrid and heterogeneous integration approaches." The Court found that this language further supported the construction excluding heterogeneous techniques.
Third, the Court credited Dr. Ng's testimony that in the silicon photonics industry, "monolithic" integration is universally understood to be distinct from "heterogeneous" integration, and that a POSITA would not understand "monolithically formed" to encompass designs in which separately fabricated components are bonded or attached to the die.
Impact on Infringement Theory: Fatal. Lumos's 800G platform — the LumosBridge™ 800G Optical Engine — utilizes a heterogeneous integration architecture in which micro-transfer-printed III-V gain elements are bonded to the silicon photonic integrated circuit (PIC). This is precisely the type of heterogeneous integration technique that the Court's construction (and the prosecution history) explicitly excludes. The Lumos design involves separately fabricating III-V semiconductor components (specifically, indium phosphide-based gain elements) and then transferring and bonding them to the silicon PIC using micro-transfer printing technology — a well-known heterogeneous integration method. Under the Court's construction, which requires that the WDM element be "fabricated entirely within the same front-end-of-line and back-end-of-line semiconductor processing sequence on a single silicon substrate without heterogeneous bonding or attachment of separately fabricated optical components," Lumos's design is squarely outside the literal claim scope.
As with the '567 Patent, prosecution history estoppel under Festo bars any DOE argument. The "monolithically formed" limitation was added by amendment during prosecution specifically to distinguish heterogeneous integration approaches — the exact category of technology employed by Lumos. The Festo presumption of surrender applies, and none of the three recognized exceptions is viable. Heterogeneous integration was well known in the art at the time of the amendment (indeed, it was the subject of the prior art references being overcome), so the unforeseeability exception does not apply. The amendment directly addressed the distinction between monolithic and heterogeneous integration, so the tangential relation exception does not apply. And there is no "other reason" that would rebut the presumption.
2. Other '678 Patent Terms
The Court construed two additional terms from the '678 Patent:
"plurality of optical channels at distinct wavelengths" (Claim 1). OWS proposed a construction of "two or more optical channels, each operating at a different wavelength." Lumos proposed a construction of "four or more optical channels operating at wavelengths separated by at least 4.5 nm." The Court adopted a modified construction: "two or more optical channels, each operating at a wavelength that is spectrally distinct from the other channels." This construction was closer to OWS's position than to Lumos's, as the Court rejected Lumos's attempt to import a specific numerical minimum for the number of channels and the wavelength spacing from a preferred embodiment. This term was counted in our scorecard as the one "modified construction" that corresponds to neither party's verbatim proposal, though it is substantively closer to OWS's position.
"single silicon die" (Claim 1). Lumos proposed that this term required a die fabricated from a "bulk silicon" or "silicon-on-insulator" wafer, excluding any die that incorporates non-silicon substrate materials. OWS proposed that the term required only "a die having a silicon substrate." The Court adopted Lumos's construction, finding that "single silicon die" refers to a die fabricated from a silicon or silicon-on-insulator wafer and does not encompass dies with compound semiconductor substrates.
Neither of these secondary constructions independently affects the infringement analysis, which is controlled by the dispositive construction of "monolithically formed." The favorable modified construction on "plurality of optical channels at distinct wavelengths" is immaterial if the "monolithically formed" requirement cannot be met.
3. Net Assessment — '678 Patent
Infringement viability: Untenable. The Court's prosecution history-driven construction of "monolithically formed" categorically excludes Lumos's heterogeneous integration architecture, in which micro-transfer-printed III-V gain elements are bonded to the silicon PIC. Prosecution history estoppel prevents DOE arguments, as the limitation was added specifically to distinguish heterogeneous approaches during prosecution.
Recommendation: Abandon infringement theories under the '678 Patent.
D. U.S. Patent No. 11,312,890 (the '890 Patent) — "Driver Circuit Architecture for High-Speed Silicon Photonic Mach-Zehnder Modulators"
The '890 Patent is directed to a driver circuit architecture for high-speed Mach-Zehnder modulators (MZMs) used in silicon photonic transceivers, in which a pre-emphasis equalization stage is co-integrated with the modulator on a common substrate to achieve improved signal integrity at high baud rates. The patent claims priority to a non-provisional application filed on April 3, 2020, and was issued on May 17, 2022. The named inventor is Dr. Amara Osei. OWS asserted Claims 1, 6, 10, and 13 of the '890 Patent in this action, all of which depend from independent Claim 1. The Court construed three disputed terms from the '890 Patent.
1. Term: "pre-emphasis equalization stage electrically co-integrated with the modulator on a common substrate" (Claim 1)
This is the critical term for the '890 Patent. Claim 1 recites, inter alia, "a pre-emphasis equalization stage electrically co-integrated with the modulator on a common substrate, the equalization stage configured to compensate for frequency-dependent losses in the modulator's electrical and optical response."
OWS's Proposed Construction: "an equalization circuit located on the same substrate as the modulator."
OWS proposed a broad, location-based construction that would require only physical co-location on the same substrate, without imposing requirements on how the equalization circuit is fabricated or electrically connected to the modulator. OWS argued that "co-integrated" simply means "integrated together" and that the claim language does not specify the fabrication method or interconnection technique.
Lumos's Proposed Construction: "an equalization circuit fabricated in the same semiconductor process as the modulator, sharing the same doping profiles and process layers, and electrically connected without bond wires, interposers, or package-level routing."
Lumos proposed a narrow construction imposing three distinct requirements: (1) same-process fabrication, (2) shared doping profiles and process layers, and (3) direct electrical connection without bond wires, interposers, or package-level routing. Lumos argued that "co-integrated" is a term of art in semiconductor design denoting that two circuit elements are fabricated together in the same process, sharing the same layers and doping steps, and that a mere physical co-location (e.g., through flip-chip bonding of separately fabricated dies) does not constitute "co-integration."
Court's Construction: The Court adopted Lumos's proposed construction in full: "an equalization circuit fabricated in the same semiconductor process as the modulator, sharing the same doping profiles and process layers, and electrically connected without bond wires, interposers, or package-level routing."
Analysis. The Court's adoption of Lumos's construction in its entirety — without any modification — is notable and represents another decisive adverse ruling. The Court's reasoning proceeded as follows:
First, the Court examined the prosecution history of the '890 Patent. During prosecution, the claims were rejected twice, first under § 102 and then under § 103, based on prior art references disclosing driver circuits that were separately fabricated and then bonded to modulator dies. In response, the applicant amended independent Claim 1 to add the "co-integrated" limitation, accompanied by remarks distinguishing the prior art on the ground that "in the prior art, the driver circuit is a separately fabricated component that is subsequently bonded to the modulator die, whereas in the present invention, the pre-emphasis equalization stage is co-integrated with the modulator — i.e., fabricated as part of the same semiconductor process on the same substrate." The Court found that these remarks, while not as explicit as the prosecution history of the '567 and '678 Patents, nonetheless supported the narrow construction requiring same-process fabrication.
Second, the Court examined the specification and found that every described embodiment of "co-integration" involved an equalization circuit that was fabricated in the same process as the modulator, sharing the same doping profiles, and directly electrically connected without bond wires or other package-level interconnects. The specification described the advantages of co-integration as including "eliminating parasitic inductance and capacitance associated with bond wires and package-level interconnects," which the Court found further supported the requirement of direct electrical connection without package-level routing.
Third, the Court credited Dr. Ng's testimony that "co-integration" is a term of art in semiconductor design that is consistently understood to mean fabrication within the same process, and is distinguished from "co-packaging" or "hybrid integration" approaches in which separately fabricated dies are assembled together.
Impact on Infringement Theory: Severely weakened but not categorically foreclosed. Unlike the other three patents, the '890 Patent presents a slightly different posture — one in which literal infringement is untenable but a narrow DOE theory is not entirely foreclosed, though it is assessed as weak.
Lumos's driver architecture for the accused products employs a separate CMOS driver die that is flip-chip bonded to the silicon photonic modulator die via copper pillar micro-bumps. Under the Court's construction, this design falls outside the literal claim scope for two independent reasons:
(a) Different semiconductor processes and doping profiles. The Lumos driver is fabricated in a standard CMOS process (believed to be a 16nm or 12nm FinFET CMOS node), while the modulator is fabricated in a silicon photonic process with entirely different doping profiles and process layers. The two dies do not share the same semiconductor process, doping profiles, or process layers.
(b) Copper pillar micro-bump interconnection. The driver and modulator dies are electrically connected through copper pillar micro-bumps — a form of flip-chip bonding that constitutes package-level interconnection. This falls squarely within the category of "bond wires, interposers, or package-level routing" that the Court's construction requires be absent.
However, the prosecution history of the '890 Patent, while supporting the narrow construction, is slightly less definitive on the scope of DOE foreclosure than the prosecution histories of the '567 and '678 Patents. Specifically:
•  The applicant's prosecution remarks distinguished the prior art on the basis that the prior art driver was "separately fabricated" and "subsequently bonded," but the remarks did not explicitly enumerate or address every possible bonding or integration technique that was being disclaimed. By contrast, the prosecution remarks for the '567 Patent explicitly addressed "segmented coupling" and the '678 Patent explicitly addressed "heterogeneous integration."
•  The amendment adding "co-integrated" was made to overcome prior art disclosing wire-bonded driver configurations. Copper pillar micro-bump flip-chip bonding is a different — and arguably more advanced — interconnection technology than wire bonding. There is a theoretical argument that the scope of prosecution history surrender was limited to wire bonding and did not extend to all forms of flip-chip interconnection, although this argument is weak.
•  Under Festo, the patentee bears the burden of rebutting the presumption of surrender. The "unforeseeability" exception might theoretically be invoked if copper pillar micro-bump technology were unforeseeable at the time of the amendment, but this argument is difficult to sustain given that flip-chip bonding with copper pillars was well known in the semiconductor industry by 2020 when the application was filed.
The residual DOE theory would proceed under the function-way-result test, arguing that Lumos's separate CMOS driver die connected via copper pillar micro-bumps performs substantially the same function (pre-emphasis equalization of the modulator drive signal), in substantially the same way (by applying a compensating transfer function to the electrical drive signal in close proximity to the modulator), to achieve substantially the same result (improved signal integrity at high baud rates with reduced frequency-dependent losses). While there is some plausibility to this argument at a high level of abstraction, it requires the factfinder to accept that a separately fabricated CMOS driver connected through package-level copper pillar interconnects is "equivalent" to a co-integrated driver fabricated in the same process with direct on-chip electrical connections — a proposition that faces substantial technical challenges, given the fundamental differences in process technology, parasitic characteristics, and interconnect architecture.
Estimated probability of success of the DOE theory at trial: 10–15% at best. This assessment reflects the difficulty of overcoming the prosecution history estoppel presumption, the technical dissimilarities between Lumos's design and the claimed co-integration, and the risk that a jury or court would view the DOE argument as an impermissible attempt to recapture subject matter surrendered during prosecution.
2. Other '890 Patent Terms
The Court construed two additional terms from the '890 Patent:
"pre-emphasis equalization" (Claim 1). OWS proposed a construction of "equalization applied to a signal before modulation to compensate for expected distortion." Lumos proposed a construction of "a feed-forward equalization technique in which the frequency response of the drive signal is pre-distorted based on a fixed, predetermined transfer function to compensate for known frequency-dependent losses." The Court adopted Lumos's construction, finding that the specification consistently described pre-emphasis equalization as a feed-forward, fixed-transfer-function technique, and that the claim should be limited accordingly. While this construction imposes additional requirements on the type of equalization that is claimed, it does not independently affect the infringement analysis because Lumos's driver does employ a pre-emphasis equalization technique with a predetermined transfer function, and this limitation would likely be met even under the narrower construction.
"common substrate" (Claim 1). OWS proposed a construction of "a single semiconductor substrate on which both the modulator and the equalization stage are present." Lumos proposed a construction of "a single monolithic semiconductor substrate in which both the modulator and the equalization stage are fabricated." The Court adopted OWS's proposed construction, finding that "common substrate" simply means a single substrate on which both components are present and does not independently require monolithic fabrication (which is already required by the "co-integrated" limitation).
OWS's victory on the "common substrate" term represents one of the two terms on which OWS prevailed overall. However, this victory does not rehabilitate the infringement case. Even under OWS's broader construction of "common substrate," the independent "co-integrated" construction requires same-process fabrication and direct electrical connection without package-level routing — requirements that Lumos's design does not meet.
3. Net Assessment — '890 Patent
Infringement viability: Significantly weakened. Literal infringement is untenable under the Court's construction of "co-integrated." A DOE theory is theoretically possible but assessed as weak, with an estimated probability of success of no more than ten to fifteen percent. The DOE argument would require expert testimony establishing that copper pillar micro-bump flip-chip bonding of a separately fabricated CMOS driver die is equivalent to true monolithic co-integration in terms of function, way, and result — a technically difficult proposition given the fundamental differences in process technology, doping profiles, and interconnect architecture. This argument would also need to overcome the Festo presumption of prosecution history estoppel, which, while slightly less definitive for the '890 Patent than for the other three patents, remains a significant obstacle.
The '890 Patent is the only patent with any residual infringement theory, however slim. Standing alone, it cannot support the damages case originally contemplated, which was premised on all four patents and a 7.5% reasonable royalty rate.
Recommendation: If the decision is made to continue litigation, the '890 Patent is the only patent on which any infringement theory survives. However, as discussed in Sections IV and VI below, the expected value of proceeding on this theory alone is negative. We do not recommend continuing litigation solely on the '890 Patent DOE theory.
IV. Impact on Damages Theories
The Markman Order fundamentally undermines OWS's damages case. This section analyzes the impact on each element of the pre-Markman damages framework.
A. Reasonable Royalty Analysis
Prior to the Markman Order, the damages framework prepared for presentation through Dr. Lawrence Kessler of Apex Claim Analytics LLC was premised on the following elements:
•  A reasonable royalty rate of 7.5%, derived from a hypothetical negotiation analysis incorporating all fifteen Georgia-Pacific factors and reflecting the combined value of the four asserted patents;
•  A royalty base of approximately $148.6 million, representing Lumos's combined accused product revenue for the LumosBridge™ 800G Optical Engine and the LumosEdge™ 400ZR+ Pluggable Transceiver for calendar year 2022;
•  Pre-suit damages of approximately $62.3 million, representing accused product revenue for the first and second quarters of 2022 (prior to the June 3, 2022 filing date) on which OWS sought pre-suit damages based on the March 15, 2022 cease-and-desist letter providing constructive notice;
•  A willfulness enhancement of up to three times under 35 U.S.C. § 284, based on Lumos's continued sales after receiving the cease-and-desist letter; and
•  A total claimed damages figure exceeding $75 million.
With three of the four patents-in-suit effectively eliminated by the Markman Order, this damages framework collapses. The reasonable royalty rate of 7.5% was predicated on the combined value of four patents covering different aspects of the accused products' technology — waveguide coupling, photodetector architecture, WDM integration, and driver circuits. A single surviving patent claim with a weak DOE theory cannot support a royalty rate anywhere near 7.5%. Under the Georgia-Pacific framework, the hypothetical negotiation would be fundamentally different: a willing licensor and willing licensee negotiating over a single patent with questionable infringement coverage would arrive at a dramatically lower royalty rate.
Based on our preliminary reassessment, a royalty rate supportable by a single weak patent infringement theory would likely be in the range of 0.5% to 1.0% at most. Moreover, the royalty base would need to be reconsidered, as the '890 Patent covers only the driver circuit architecture — not the waveguide coupling, photodetector, or WDM integration aspects of the accused products — and an apportionment analysis would be required to isolate the value attributable to the driver circuit component. A reasonable apportioned royalty base might be in the range of $25 to $40 million, reflecting the fractional value of the driver circuit relative to the overall accused product.
Combining a reduced royalty rate (0.5%–1.0%) with a reduced, apportioned royalty base ($25–$40 million), the maximum reasonable royalty recovery on the '890 Patent alone would be in the range of $125,000 to $400,000 — before consideration of the low probability of establishing infringement through the DOE theory.
The Georgia-Pacific factor analysis (draft prepared by Dr. Ke

PRIVILEGED AND CONFIDENTIAL ATTORNEY WORK PRODUCT — ATTORNEY-CLIENT COMMUNICATION

INTERNAL MEMORANDUM

Record

Parties

Dates

Amounts