UNITED STATES DISTRICT COURT NORTHERN DISTRICT OF CALIFORNIA SAN JOSE DIVISION
OPTIWAVE SEMICONDUCTOR CORP., a Delaware corporation,
Plaintiff,
v.
LUMOS PHOTONICS, INC., a Delaware corporation,
Defendant.
Case No. 5:22-cv-03417-EJD
Before the Honorable Edward J. Torvelis
PLAINTIFF OPTIWAVE SEMICONDUCTOR CORP.'S OPENING CLAIM CONSTRUCTION BRIEF
Date: December 5, 2022 Time: To Be Set Courtroom: 4, 5th Floor Judge: Hon. Edward J. Torvelis
CALDERWOOD & HARKNESS LLP Alan Ngo (Cal. Bar No. 287,419; USPTO Reg. No. 52,871) Marcus Delgado (Cal. Bar No. 301,553) Omar Siddiqui (Cal. Bar No. 312,078) Christy Bauer (Cal. Bar No. 318,264) One Embarcadero Center, 34th Floor San Francisco, California 94111 Telephone: (415) 772-4000 Facsimile: (415) 772-4001
Email: ango@calderwoodharkness.com
Attorneys for Plaintiff Optiwave Semiconductor Corp.
TABLE OF AUTHORITIES
Cases.
Statutes and Rules
Treatises and Technical Publications
Other Authorities.
I. INTRODUCTION
Plaintiff Optiwave Semiconductor Corp. ("OWS" or "Plaintiff") respectfully submits this Opening Claim Construction Brief pursuant to Patent Local Rule 4-5 and the Court's Case Management Order dated August 15, 2022 (Dkt. No. 45). This brief addresses the construction of eleven disputed claim terms across the four United States patents asserted in this action.
OWS is a leading fabless silicon photonics company headquartered in San Jose, California. Founded in 2014 by Dr. Priya Ramanathan and Dr. Kevin Fong — both pioneers in the field of integrated photonics — OWS has devoted nearly a decade to developing core innovations in integrated silicon photonics for high-speed optical transceivers used in data center interconnects, telecommunications networks, and next-generation computing architectures. OWS holds a portfolio of eighty-seven issued United States patents and fourteen pending applications, reflecting its sustained investment in research and development at the forefront of this rapidly growing technology.
This action concerns four of OWS's most significant patents, each directed to foundational innovations in the silicon photonics transceiver ecosystem:
• U.S. Patent No. 10,234,567 ("the '567 Patent"), entitled "Integrated Waveguide Coupler with Tapered Mode Converter for Silicon Photonics Platforms," directed to low-loss optical coupling between silicon and silicon nitride waveguide layers;
• U.S. Patent No. 10,891,234 ("the '234 Patent"), entitled "Germanium-on-Silicon Photodetector with Graded Buffer Layer for Enhanced Responsivity," directed to high-performance photodetectors with improved responsivity achieved through graded germanium buffer layers;
• U.S. Patent No. 11,045,678 ("the '678 Patent"), entitled "Multi-Channel Optical Transceiver with Integrated Wavelength-Division Multiplexing on Silicon," directed to monolithically integrated wavelength-division multiplexing for compact, high-bandwidth optical engines; and
• U.S. Patent No. 11,312,890 ("the '890 Patent"), entitled "Driver Circuit Architecture for High-Speed Silicon Photonic Mach-Zehnder Modulators," directed to co-integrated driver circuits with pre-emphasis equalization for improved modulation performance.
Defendant Lumos Photonics, Inc. ("Lumos" or "Defendant") manufactures and sells products that practice OWS's patented technologies, including the LumosBridge™ 800G Optical Engine and the LumosEdge™ 400ZR+ Pluggable Transceiver. These products, which generated approximately $148.6 million in revenue in calendar year 2022, incorporate the same core silicon photonics innovations claimed in OWS's patents — adiabatic waveguide couplers, germanium photodetectors with graded buffer layers, monolithically integrated WDM elements, and co-integrated driver circuits with pre-emphasis equalization.
The parties identified eleven disputed claim terms in their Joint Claim Construction and Prehearing Statement filed on October 14, 2022 (Dkt. No. 78), pursuant to Patent Local Rule 4-3. OWS's proposed constructions for each of these terms are grounded in the plain and ordinary meaning of the claim language as understood by a person of ordinary skill in the art ("POSITA"), and are fully supported by the intrinsic record — including the patent specifications and prosecution histories — as well as the expert testimony of Dr. Harold Whitfield, Professor Emeritus of Electrical and Computer Engineering at the University of California, Santa Barbara, a leading authority in integrated photonics with over two hundred publications in peer-reviewed journals and an extensive career spanning both academic research and industrial collaboration in the silicon photonics field.
In contrast, Lumos's proposed constructions are improperly narrow. Across multiple disputed terms, Lumos repeatedly seeks to: (1) import fabrication process limitations into structural product claims; (2) read specific numerical thresholds and performance parameters from isolated examples in the specification into the claims; and (3) confine the claims to single preferred embodiments, in direct contravention of the Federal Circuit's instruction that courts must not "confine the claims to those embodiments." Phillips v. AWH Corp., 415 F.3d 1303, 1323 (Fed. Cir. 2005) (en banc). Lumos's constructions are designed to artificially narrow the claims in a manner that has no basis in the claim language, the specification, or the prosecution history.
For the reasons set forth below and in the accompanying Declaration of Dr. Harold Whitfield (Exhibit A), the Court should adopt OWS's proposed claim constructions for all eleven disputed terms.
II. LEGAL STANDARD FOR CLAIM CONSTRUCTION
Claim construction is a question of law for the Court. Markman v. Westview Instruments, Inc., 517 U.S. 370, 388–91 (1996). The purpose of claim construction is to determine the meaning and scope of the patent claims, which define the boundaries of the patentee's right to exclude. Id. at 373. The Court's construction of disputed claim terms controls the subsequent determination of infringement and validity.
The governing framework for claim construction was established by the Federal Circuit's en banc decision in Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005). Under Phillips, the words of a claim are generally given their ordinary and customary meaning as understood by a person of ordinary skill in the art at the time of the invention, read in the context of the patent specification and prosecution history. Id. at 1312–13. The inquiry is an objective one, focused on how the relevant scientific or technical community would have understood the claim terms. Id. at 1313.
The intrinsic record — comprising the claims themselves, the specification, and the prosecution history — is the primary source of evidence for claim construction. Id. at 1314. Among these sources, the specification occupies a position of central importance. As the Federal Circuit explained in Phillips:
The specification "is always highly relevant to the claim construction analysis. Usually, it is dispositive; it is the single best guide to the meaning of a disputed term." Ridgeway & Co. v. Conceptronic, Inc., 90 F.3d 1576, 1582 (Fed. Cir. 1996).
Phillips, 415 F.3d at 1315 (quoting Vitronics). The specification may reveal that the patentee has defined a term in a particular way, or has disavowed the full scope of a claim term either in the specification or during prosecution. Id. at 1316. The claims must be read in view of the specification, and the specification provides context and meaning that may inform the proper construction of a claim term.
At the same time, however, the Federal Circuit has "repeatedly warned against confining the claims to those embodiments." Phillips, 415 F.3d at 1323. While the specification may describe particular embodiments with specificity, the claims are not limited to the preferred embodiments described therein unless the patentee has clearly and unambiguously limited the scope of the claims through definitional language in the specification or through prosecution history disclaimer. Id. It is improper to read a particular structural feature, numerical parameter, or implementation detail from a preferred embodiment into the claims unless the claim language itself, or the prosecution history, requires it.
The prosecution history also forms part of the intrinsic record and constitutes a public record of the proceedings before the Patent and Trademark Office, including the prior art cited during examination, any rejections by the examiner, and the applicant's amendments and remarks in response. Phillips, 415 F.3d at 1317. The prosecution history may inform claim construction by establishing that the patentee disclaimed certain subject matter to obtain the patent. Id. However, prosecution history disclaimer requires that the alleged disavowal be "clear and unmistakable." Omega Eng'g, Inc. v. Raytek Corp., 334 F.3d 1314, 1325–26 (Fed. Cir. 2003). "Prosecution disclaimer does not apply where the alleged disavowal is ambiguous or where the applicant's statements are susceptible to more than one reasonable interpretation." Id. at 1326. The doctrine of prosecution history estoppel, as set forth in Festo Corp. v. Fenmore & Co., 535 U.S. 722 (2002), limits the scope of the doctrine of equivalents based on narrowing amendments made during prosecution, but it is a doctrine of infringement, not a claim construction tool. A narrowing amendment does not, standing alone, require the Court to construe the literal scope of a claim term more narrowly than its ordinary meaning would support.
Extrinsic evidence, including expert testimony, learned treatises, technical dictionaries, and other sources, may be used to assist the Court in understanding the technology and the meaning of claim terms as they would be understood by a POSITA. Phillips, 415 F.3d at 1318–19. However, extrinsic evidence is subordinate to the intrinsic record and "is unlikely to result in a reliable interpretation of patent claim scope unless considered in the context of the intrinsic evidence." Id. at 1319. Extrinsic evidence may not be used to contradict the meaning of claim terms established by the intrinsic record. Id.
For purposes of these patents, a POSITA is a person with (i) a Ph.D. or equivalent professional experience in electrical engineering, applied physics, or materials science with a focus on photonics or semiconductor devices, and (ii) at least three to five years of experience in the design or fabrication of integrated silicon photonics components. This definition is supported by the Declaration of Dr. Harold Whitfield (¶¶ 8–12) and reflects the level of skill in the art at the time of the inventions disclosed in the patents-in-suit. OWS understands that Lumos does not dispute this definition of a POSITA.
III. TECHNOLOGY BACKGROUND
To assist the Court in understanding the claim terms in dispute, OWS provides the following overview of the relevant technology. A more detailed discussion is set forth in the Declaration of Dr. Harold Whitfield (¶¶ 13–32) and in the Technical Tutorial Presentation submitted concurrently herewith.
Silicon Photonics: An Overview
Silicon photonics is a technology that integrates optical components — including waveguides, modulators, photodetectors, and multiplexers — onto silicon semiconductor substrates using fabrication processes compatible with conventional complementary metal-oxide-semiconductor ("CMOS") manufacturing. The fundamental promise of silicon photonics is to leverage the enormous existing infrastructure and economies of scale of silicon semiconductor fabrication to produce high-performance optical interconnect components at dramatically lower cost and higher volume than is possible with traditional compound semiconductor (e.g., indium phosphide) platforms.
Silicon photonics has emerged as a critical enabling technology for high-speed optical interconnects in modern data centers and telecommunications networks. As data center traffic has grown exponentially — driven by cloud computing, artificial intelligence workloads, and streaming media — the demand for optical transceivers operating at 100 Gbps, 400 Gbps, and now 800 Gbps and beyond has created a large and rapidly expanding market. Silicon photonic integrated circuits ("PICs") form the optical core of these transceivers, integrating the key optical functions needed for high-speed data transmission and reception onto a single chip.
The core building blocks of a silicon photonic transceiver include:
• Silicon waveguides: Optical channels formed in a thin silicon device layer on a silicon-on-insulator ("SOI") substrate. These waveguides confine and route light on the chip with low propagation loss. The high refractive index contrast between silicon (n ≈ 3.48 at 1550 nm wavelength) and the surrounding silicon dioxide cladding (n ≈ 1.44) enables tight optical confinement and compact waveguide bends.
• Optical modulators: Devices that encode electrical data onto an optical carrier wave by modulating the phase or intensity of the light. The Mach-Zehnder modulator ("MZM") is a widely used architecture in silicon photonics, employing phase shifters based on the plasma dispersion effect in silicon to achieve high-speed modulation.
• Photodetectors: Devices that convert received optical signals back into electrical signals. Because silicon is transparent at the telecommunications wavelengths of interest (1310 nm and 1550 nm), germanium — which has strong optical absorption in this wavelength range — is grown epitaxially on the silicon substrate to form germanium-on-silicon ("Ge-on-Si") photodetectors.
• Wavelength-division multiplexing (WDM) elements: Optical components such as arrayed waveguide gratings ("AWGs"), echelle gratings, or ring resonator filters that combine (multiplex) or separate (demultiplex) multiple optical signals at different wavelengths onto or from a single optical fiber, enabling multi-channel operation and dramatically increasing the data capacity of a single fiber connection.
Waveguide Coupling and Tapered Mode Converters
One of the persistent challenges in silicon photonics is efficiently coupling light between optical fibers and the on-chip silicon waveguides, and between different waveguide layers within the PIC. Optical fibers have a relatively large mode field diameter (approximately 10 µm for standard single-mode fiber), while silicon waveguides are typically only a few hundred nanometers in cross-section. This enormous mismatch in optical mode size leads to high coupling losses unless mode conversion structures are employed.
Tapered mode converters address this challenge by gradually transforming the optical mode from one waveguide to another through a taper structure that changes in width (and sometimes height) along its length. The taper gradually reduces the confinement of the optical mode in one waveguide, causing the mode to expand and couple into an adjacent waveguide with a larger mode volume. When this taper is sufficiently gradual, the optical mode evolves without exciting higher-order or radiation modes, and the transition is termed "adiabatic." Adiabatic coupling is highly desirable because it achieves low insertion loss and broadband operation.
Similarly, in multi-layer silicon photonics platforms that incorporate both silicon and silicon nitride waveguide layers, tapered mode converters are used to transfer light between the two layers. Silicon nitride waveguides offer lower propagation loss and better tolerance to fabrication variations than silicon waveguides, making them valuable for passive optical functions, while silicon waveguides are needed for active devices (modulators and photodetectors). The efficient transfer of light between these two waveguide layers through tapered adiabatic couplers is a key design challenge — and a key subject of the '567 Patent.
Germanium-on-Silicon Photodetectors. As noted above, silicon is essentially transparent at the near-infrared wavelengths used in optical communications, necessitating the use of germanium as the absorbing material for on-chip photodetectors. Germanium has a direct bandgap transition at approximately 0.8 eV (1550 nm), making it suitable for detecting light at both the 1310 nm and 1550 nm telecommunications windows.
However, the growth of high-quality germanium on silicon is complicated by the approximately 4.2% lattice mismatch between the two materials. This lattice mismatch leads to the formation of threading dislocations — crystallographic defects that propagate through the germanium film — which degrade the photodetector's performance by increasing dark current (leakage current in the absence of light) and reducing responsivity (the efficiency of converting incident optical power to photocurrent). The use of buffer layers between the silicon substrate and the germanium absorber layer to manage this lattice mismatch and reduce threading dislocation density is a well-established strategy in the field. Graded buffer layers, in which the germanium concentration varies across the buffer thickness, are particularly effective at reducing dislocation density by gradually accommodating the lattice mismatch. The '234 Patent is directed to germanium-on-silicon photodetectors employing such graded buffer layers.
Wavelength-Division Multiplexing on Silicon
Wavelength-division multiplexing allows multiple data channels, each carried on a different wavelength of light, to be transmitted simultaneously over a single optical fiber. This is a cornerstone technique for increasing the aggregate bandwidth of optical interconnects. In silicon photonics, WDM elements such as arrayed waveguide gratings, echelle gratings, and cascaded ring resonator filters can be integrated directly on the silicon PIC, eliminating the need for separate discrete WDM components and enabling more compact and cost-effective transceiver designs.
The monolithic integration of WDM elements on the same silicon die as the modulators and photodetectors is a significant achievement that reduces assembly complexity, improves performance through tighter optical alignment, and lowers overall transceiver cost. The '678 Patent is directed to this monolithic integration approach.
Mach-Zehnder Modulator Driver Circuits
High-speed silicon photonic MZMs require driver circuits that provide the electrical signals to actuate the modulator's phase shifters. As data rates have increased to 50 Gbaud, 100 Gbaud, and beyond, the bandwidth limitations of the modulator and its electrical interconnects become increasingly significant. Pre-emphasis equalization — a signal processing technique that boosts the high-frequency content of the drive signal to compensate for bandwidth roll-off — is essential for achieving adequate signal quality at these data rates.
Traditionally, MZM driver circuits have been fabricated as separate integrated circuits and connected to the modulator through bond wires, interposers, or other package-level interconnects. However, these external interconnects introduce parasitic inductance and capacitance that limit bandwidth and degrade signal integrity. Co-integrating the driver circuit on the same substrate as the modulator reduces these parasitics and enables tighter electrical coupling, resulting in improved modulation bandwidth and signal fidelity. The '890 Patent is directed to such co-integrated driver circuit architectures with pre-emphasis equalization.
OWS's Role as a Pioneer in Silicon Photonics
OWS was founded in 2014 by Dr. Priya Ramanathan and Dr. Kevin Fong, both of whom had extensive experience in silicon photonics research at leading academic institutions and in industry. Since its founding, OWS has developed key innovations in waveguide coupling, germanium photodetection, integrated WDM, and modulator driver circuits, resulting in a portfolio of eighty-seven issued U.S. patents and fourteen pending applications. OWS's products, including the OWS-400ZR and OWS-800LR coherent optical transceiver chipsets — fabricated on Pinnacle Semiconductor Foundry Co., Ltd.'s 45nm SOI process — are widely recognized in the industry as among the highest-performing silicon photonic transceiver solutions available.
For additional technical background, the Court is respectfully referred to the Declaration of Dr. Harold Whitfield (¶¶ 13–32) and the Technical Tutorial Presentation filed concurrently herewith.
IV. THE PATENTS-IN-SUIT
A. U.S. Patent No. 10,234,567 ("the '567 Patent")
The '567 Patent, entitled "Integrated Waveguide Coupler with Tapered Mode Converter for Silicon Photonics Platforms," was filed on March 14, 2017 and issued on March 19, 2019. The named inventors are Dr. Kevin Fong, Dr. Priya Ramanathan, and Hideo Tanaka. The patent is assigned to OWS. OWS asserts Claims 1, 4, 7, 12, and 15 of the '567 Patent in this action.
The '567 Patent is directed to an integrated optical coupler comprising a tapered silicon waveguide adiabatically coupled to a silicon nitride waveguide layer. The invention achieves low-loss optical mode transitions for efficient fiber-to-chip and layer-to-layer coupling on silicon photonics platforms. As the specification explains, prior approaches to coupling between silicon and silicon nitride waveguide layers suffered from high insertion losses due to abrupt mode transitions at the coupling interface, limiting the performance of multi-layer silicon photonic integrated circuits. The '567 Patent solves this problem through a tapered mode converter structure that provides adiabatic coupling — ensuring that the optical mode evolves gradually from one waveguide to the other without significant power loss.
Independent Claim 1 of the '567 Patent recites:
1. An integrated optical coupler comprising:
a tapered silicon waveguide having a tip width of less than 200 nanometers adiabatically coupled to a silicon nitride waveguide layer, wherein the tapered silicon waveguide is configured to convert an optical mode between the silicon waveguide and the silicon nitride waveguide layer with an insertion loss of less than 0.5 dB across a wavelength range of 1260 nm to 1360 nm.
Dependent Claims 4, 7, 12, and 15 add further limitations relating to specific taper geometries, coupling lengths, additional waveguide layers, and integration with fiber-to-chip coupling structures.
Prosecution History. The '567 Patent was subject to two office action rejections during prosecution. The first office action rejected the claims over Tanaka et al. (U.S. Patent No. 9,456,012, issued September 6, 2016) in combination with a secondary reference. In response, the applicant amended independent Claim 1 to include the "adiabatically coupled" limitation and presented arguments distinguishing the claimed invention from the Tanaka prior art. Specifically, the applicant stated in remarks dated July 12, 2018:
"The claimed integrated optical coupler achieves adiabatic coupling, ensuring that the optical mode transitions without significant power loss, unlike the abrupt coupling junction disclosed in Tanaka. In Tanaka, the coupling between the silicon waveguide and the silicon nitride layer occurs at an abrupt interface without gradual mode evolution, resulting in significant scattering losses. The claimed adiabatic coupling, by contrast, ensures smooth mode evolution and minimal power loss."
The examiner subsequently issued a second office action maintaining the rejection on different grounds, which was overcome by further amendment and argument. The patent issued on March 19, 2019. Notably, the applicant's distinction of the Tanaka prior art was based on the adiabatic nature of the coupling — the smooth, low-loss mode evolution — rather than on any specific geometric structure or taper profile. The prosecution history excerpts are included in Exhibit C.
B. U.S. Patent No. 10,891,234 ("the '234 Patent")
The '234 Patent, entitled "Germanium-on-Silicon Photodetector with Graded Buffer Layer for Enhanced Responsivity," was filed on September 8, 2017 and issued on January 14, 2021. The named inventors are Dr. Kevin Fong and Dr. Amara Osei. The patent is assigned to OWS. OWS asserts Claims 1, 3, 5, 8, and 11 of the '234 Patent.
The '234 Patent is directed to a germanium-on-silicon photodetector incorporating a graded buffer layer with a continuously varying germanium concentration. The graded buffer layer reduces threading dislocation density in the germanium absorber, thereby enhancing the photodetector's responsivity for high-speed photodetection in silicon photonic receivers. The specification explains that prior germanium-on-silicon photodetectors suffered from high dark current and limited responsivity due to the high threading dislocation density resulting from the lattice mismatch between germanium and silicon. The '234 Patent addresses this problem through a carefully engineered graded buffer layer.
Independent Claim 1 of the '234 Patent recites:
1. A germanium-on-silicon photodetector comprising:
a silicon substrate;
a graded buffer layer having a continuously varying germanium concentration disposed on the silicon substrate; and
a germanium absorber layer disposed on the graded buffer layer, wherein the graded buffer layer reduces threading dislocation density in the germanium absorber layer for enhanced responsivity.
Dependent Claims 3, 5, 8, and 11 add limitations relating to specific buffer layer thicknesses, germanium concentration profiles, integration with silicon waveguide structures, and particular applications in receiver architectures.
Prosecution History. The '234 Patent was allowed on the first office action following an examiner interview conducted on April 22, 2020. The examiner's interview summary indicates that the claims as originally filed were found to be patentable over the prior art of record after discussion of the claimed graded buffer layer structure. No narrowing amendments or disclaimers were made during prosecution.
Significantly, the specification of the '234 Patent describes embodiments of both "continuously graded" buffer layers (where the germanium concentration varies in a smooth, continuous fashion) and "step-graded" buffer layers (where the germanium concentration changes in discrete steps). Specifically, the specification states at col. 4, ll. 15–42:
"In one embodiment, the buffer layer has a continuously graded germanium concentration profile, in which the germanium atomic fraction varies smoothly from approximately 0% at the silicon substrate interface to approximately 100% at the absorber layer interface. In another embodiment, the buffer layer has a step-graded germanium concentration profile comprising a plurality of sub-layers, each having a substantially uniform germanium concentration, with the germanium concentration increasing from sub-layer to sub-layer."
Despite this description of both continuous and step-graded profiles, the claims use the language "continuously varying," which, as discussed in Section V.B.4 below, encompasses both profiles. No narrowing amendments or disclaimers were made during prosecution to limit this term.
C. U.S. Patent No. 11,045,678 ("the '678 Patent")
The '678 Patent, entitled "Multi-Channel Optical Transceiver with Integrated Wavelength-Division Multiplexing on Silicon," was filed on June 22, 2018 and issued on June 29, 2021. The named inventors are Dr. Priya Ramanathan, Dr. Amara Osei, and Liwei Zhang. The patent is assigned to OWS. OWS asserts Claims 1, 2, 6, 9, and 14 of the '678 Patent.
The '678 Patent is directed to a multi-channel optical transceiver with an integrated WDM element monolithically formed on a single silicon die, enabling compact, high-bandwidth optical engines without discrete WDM components. The specification explains that prior multi-channel transceiver designs required separate, discrete WDM components (such as thin-film filter arrays or silica-on-silicon AWGs) to be assembled with the silicon photonic die, increasing the transceiver's size, cost, and assembly complexity. The '678 Patent eliminates this need by monolithically integrating the WDM element directly on the silicon photonic die.
Independent Claim 1 of the '678 Patent recites:
1. A multi-channel optical transceiver comprising:
a silicon photonic die;
an integrated wavelength-division multiplexing element monolithically formed on a single silicon die, the WDM element configured to multiplex or demultiplex a plurality of optical channels at different wavelengths; and
a plurality of optical modulators and photodetectors formed on the silicon photonic die and optically coupled to the WDM element.
Dependent Claims 2, 6, 9, and 14 add further limitations concerning the type of WDM element (AWG, echelle grating, ring resonators), channel spacing, integration with fiber coupling structures, and specific transceiver architectures.
Prosecution History. The '678 Patent was subject to three office action rejections during prosecution. The primary basis for rejection was a combination of prior art references that disclosed multi-channel transceivers with WDM functionality achieved through hybrid integration — that is, by bonding separately fabricated WDM components onto a silicon photonic die. In response, the applicant amended Claim 1 to add the "monolithically formed" limitation and submitted the following remarks, dated February 8, 2021:
"Applicant respectfully submits that the amended claims are patentable over the cited prior art. The present invention is distinguished from heterogeneous integration approaches, such as that disclosed in the primary reference, where separately fabricated WDM components are bonded onto the silicon die. In contrast, the claimed multi-channel optical transceiver includes a WDM element that is monolithically formed on a single silicon die — i.e., fabricated as an integral part of the silicon die during the semiconductor fabrication process."
OWS submits that this statement distinguishes heterogeneous integration — which involves the bonding of separately fabricated wafers or dies from different material systems — but does not constitute a blanket disclaimer of all forms of integration that may involve any bonding or attachment step during the overall fabrication process on a single die. The statement specifically targets "heterogeneous integration approaches" in which "separately fabricated WDM components are bonded onto the silicon die" — a fundamentally different process from monolithic integration in which the WDM element is defined lithographically and formed during the wafer-level processing of a single die. This distinction and its implications for claim construction are discussed in detail in Section V.C.6 below.
D. U.S. Patent No. 11,312,890 ("the '890 Patent")
The '890 Patent, entitled "Driver Circuit Architecture for High-Speed Silicon Photonic Mach-Zehnder Modulators," was filed on November 15, 2018 and issued on April 26, 2022. The named inventors are Dr. Kevin Fong and Rajesh Iyer. The patent is assigned to OWS. OWS asserts Claims 1, 5, 8, and 13 of the '890 Patent.
The '890 Patent is directed to a driver circuit architecture incorporating a pre-emphasis equalization stage electrically co-integrated with the modulator on a common substrate, enabling improved modulation bandwidth and signal integrity for high-speed silicon photonic MZMs. The specification explains that conventional approaches, in which the driver circuit is fabricated on a separate die and connected to the modulator through bond wires or other package-level interconnects, suffer from significant parasitic effects that limit the achievable modulation bandwidth. The co-integration approach of the '890 Patent addresses this limitation by placing the driver circuit and modulator on the same substrate, minimizing interconnect parasitics.
Independent Claim 1 of the '890 Patent recites:
1. A driver circuit architecture for a high-speed silicon photonic Mach-Zehnder modulator, comprising:
a pre-emphasis equalization stage electrically co-integrated with the modulator on a common substrate, the pre-emphasis equalization stage configured to compensate for bandwidth limitations of the modulator by boosting high-frequency components of a drive signal; and
a differential output stage coupled to the pre-emphasis equalization stage and configured to drive the phase shifter sections of the Mach-Zehnder modulator.
Dependent Claims 5, 8, and 13 add limitations relating to specific equalization architectures, output stage configurations, feedback mechanisms, and integration with transmitter datapaths.
Prosecution History. The '890 Patent was subject to two office action rejections during prosecution. The primary rejection was based on a journal article by Iyer & Patel, "High-Speed Driver Architectures for Mach-Zehnder Modulators," 65 IEEE Transactions on Circuits and Systems 2501 (2018), which disclosed a high-speed driver circuit for an MZM with pre-emphasis equalization. However, in the Iyer & Patel reference, the driver circuit was fabricated on a separate SiGe BiCMOS die and connected to the silicon photonic MZM through wire bonds in a multi-chip module package.
In response, the applicant amended Claim 1 to add the "co-integrated" limitation and submitted remarks dated September 15, 2021, stating:
"The claimed invention is distinguished from the Iyer & Patel reference, in which the driver circuit is fabricated on a separate die and connected to the modulator through wire bonds in a multi-chip module. In contrast, the claimed driver circuit architecture includes a pre-emphasis equalization stage that is electrically co-integrated with the modulator on a common substrate, achieving tighter electrical coupling than separately packaged driver solutions. This co-integration on a common substrate minimizes parasitic inductance and capacitance, enabling superior modulation bandwidth that cannot be achieved with the wire-bonded, separately-packaged approach of Iyer & Patel."
The applicant's remarks thus distinguish the prior art based on the separate packaging and wire-bonded interconnection of the Iyer & Patel approach, as contrasted with the co-integration on a common substrate of the claimed invention. The applicant did not disclaim all bonding or attachment methods; it disclaimed separately packaged driver solutions connected through package-level wire bonds.
V. DISPUTED CLAIM TERMS AND OWS'S PROPOSED CONSTRUCTIONS
A. Disputed Terms of the '567 Patent
1. "adiabatically coupled" (Claims 1, 4, 7, 12, 15). OWS's Proposed Construction: "coupled such that optical mode transitions occur without significant power loss."
Lumos's Proposed Construction: "coupled through a gradual, continuous taper that prevents excitation of higher-order modes throughout the entire transition region."
The Court should adopt OWS's construction because it faithfully reflects the plain and ordinary meaning of "adiabatically coupled" as understood by a POSITA, is consistent with the full scope of the specification, and is not narrowed by the prosecution history. Lumos's construction, by contrast, improperly imports structural limitations from a single preferred embodiment and adds performance requirements not found in the claim language.
a. The Plain and Ordinary Meaning of "Adiabatically Coupled"
The term "adiabatic" has a well-established meaning in the field of photonics. Derived from the thermodynamic concept of an adiabatic process (one in which no energy is exchanged with the environment), the term as applied to photonic waveguide coupling refers to a coupling process in which the optical mode evolves sufficiently gradually that energy remains in the fundamental mode and is not lost to radiation or transferred to higher-order modes. The result of adiabatic coupling is a mode transition that occurs without significant power loss.
Dr. Harold Whitfield, whose expertise in integrated photonics spans over three decades and includes extensive work on waveguide coupling structures, confirms that a POSITA would understand "adiabatically coupled" as a functional descriptor referring to coupling that achieves low-loss mode transitions, rather than as a structural requirement mandating a particular taper geometry. Whitfield Decl. ¶¶ 33–41. As Dr. Whitfield explains, the concept of adiabatic coupling in photonics is defined by the outcome — the preservation of the fundamental mode during the transition — not by the specific physical structure of the coupler. Id. ¶ 35. Various taper geometries, including continuous linear tapers, parabolic tapers, exponential tapers, and multi-segment tapers with discrete coupling sections, can all achieve adiabatic coupling depending on the design parameters. Id. ¶ 36.
b. The Specification Supports OWS's Construction
The specification of the '567 Patent uses the term "adiabatically coupled" to describe the functional result of the coupling — a low-loss mode transition — not a specific geometric structure. At col. 3, ll. 22–45, the specification states:
"The tapered silicon waveguide is adiabatically coupled to the silicon nitride waveguide layer, meaning that the optical mode transitions between the two waveguide layers without significant power loss. The adiabatic nature of the coupling ensures that the fundamental optical mode evolves smoothly from the silicon waveguide to the silicon nitride waveguide, minimizing insertion loss."
This language plainly describes "adiabatically coupled" in terms of the functional outcome — "without significant power loss" — rather than in terms of a particular taper geometry.
Further, the specification discloses multiple embodiments of adiabatic coupling structures, confirming that the term is not limited to a single geometric implementation. At col. 5, ll. 8–30, the specification describes a continuous linear taper (illustrated in Fig. 4A) in which the silicon waveguide width decreases linearly from its initial width to a tip width of less than 200 nm. In addition, the specification describes a multi-segment taper (illustrated in Fig. 4B) comprising a plurality of discrete taper sections with different taper angles and segment lengths. Both embodiments are described as achieving "adiabatic coupling" between the silicon and silicon nitride waveguide layers. The disclosure of multiple taper geometries — continuous and segmented — that achieve adiabatic coupling confirms that the term "adiabatically coupled" is not limited to continuous tapers.
c. Lumos's Construction Improperly Imports Limitations From a Preferred Embodiment
Lumos's proposed construction — "coupled through a gradual, continuous taper that prevents excitation of higher-order modes throughout the entire transition region" — improperly imports two limitations that are not found in the claim language.
First, Lumos adds a structural requirement: "a gradual, continuous taper." This language corresponds to the linear continuous taper embodiment of Fig. 4A but excludes the multi-segment taper embodiment of Fig. 4B, which is expressly described in the specification as achieving adiabatic coupling. The Federal Circuit has repeatedly admonished that claims should not be confined to preferred embodiments. Phillips, 415 F.3d at 1323; Ashford Enterprises v. Dunmore Corporation, 358 F.3d 898, 906 (Fed. Cir. 2004) ("Even when the specification describes only a single embodiment, the claims of the patent will not be read restrictively unless the patentee has demonstrated a clear intention to limit the claim scope using words or expressions of manifest exclusion or restriction.").
Second, Lumos adds a performance requirement: "prevents excitation of higher-order modes throughout the entire transition region." This language does not appear in the claims and imposes an absolute performance standard that goes beyond what "adiabatic" means in the field. As Dr. Whitfield explains, adiabatic coupling in practice is characterized by minimal — not zero — excitation of higher-order modes, and the degree to which higher-order modes are suppressed varies along the transition region depending on local taper geometry and wavelength. Whitfield Decl. ¶ 38. A POSITA would not understand "adiabatically coupled" to require that higher-order modes are never excited at any point throughout the entire transition region; rather, it means that the coupling is sufficiently gradual that the fundamental mode is preserved and the overall transition occurs without significant power loss. Id.
d. The Prosecution History Does Not Support Lumos's Narrow Construction
During prosecution of the '567 Patent, the applicant added the "adiabatically coupled" limitation to distinguish the Tanaka prior art (U.S. Patent No. 9,456,012), which disclosed an abrupt coupling junction between silicon and silicon nitride waveguides. The applicant's remarks stated: "The claimed integrated optical coupler achieves adiabatic coupling, ensuring that the optical mode transitions without significant power loss, unlike the abrupt coupling junction disclosed in Tanaka." This statement drew a distinction between adiabatic coupling (low-loss mode evolution) and abrupt coupling (high-loss due to lack of gradual transition) — not between continuous tapers and segmented tapers. The applicant's amendment and remarks established only that the claimed coupler must achieve adiabatic coupling; they did not limit the term to any particular taper geometry.
Lumos may argue that the applicant's amendment constitutes a prosecution history disclaimer that narrows the scope of "adiabatically coupled." However, prosecution history disclaimer requires a "clear and unmistakable" disavowal of claim scope. Omega Eng'g, 334 F.3d at 1325–26. Nothing in the applicant's remarks during prosecution of the '567 Patent clearly and unmistakably disclaims segmented tapers or other non-continuous taper geometries that achieve adiabatic coupling. The applicant simply distinguished adiabatic coupling from abrupt coupling — a distinction that OWS's proposed construction fully captures.
Moreover, as a doctrinal matter, prosecution history estoppel under Festo Corp. v. Fenmore & Co., 535 U.S. 722 (2002), applies to limit the doctrine of equivalents based on narrowing amendments, but does not operate to narrow the literal scope of claim terms below their ordinary meaning. The fact that "adiabatically coupled" was added during prosecution to overcome prior art does not mean that the Court must construe the term more narrowly than its plain and ordinary meaning. Rather, the Court construes the claim term according to its ordinary meaning and then, at the infringement stage, considers whether prosecution history estoppel limits the application of the doctrine of equivalents. Id. at 733–34.
e. Dr. Whitfield's Expert Testimony Confirms OWS's Construction
Dr. Whitfield opines that a POSITA would understand "adiabatically coupled" as a functional descriptor referring to coupling that achieves optical mode transitions without significant power loss, encompassing any taper geometry — continuous, segmented, or otherwise — that achieves this result. Whitfield Decl. ¶¶ 33–41. Dr. Whitfield's opinion is based on his extensive experience in the field, his familiarity with the relevant technical literature, and his understanding of how the term "adiabatic" is used in the photonics community. Id. This extrinsic evidence is consistent with and reinforces the construction derived from the intrinsic record.
For all of the foregoing reasons, the Court should adopt OWS's proposed construction of "adiabatically coupled": "coupled such that optical mode transitions occur without significant power loss."
2. "tapered silicon waveguide having a tip width of less than 200 nanometers" (Claim 1). OWS's Proposed Construction: Plain and ordinary meaning. No construction is necessary. The term is clear on its face: a silicon waveguide that tapers to a tip width below 200 nanometers.
Lumos's Proposed Construction: "a silicon waveguide with a linearly varying width that narrows to a single pointed tip having a measured width of less than 200 nanometers at its narrowest extent."
The Court should hold that this term needs no construction beyond its plain and ordinary meaning, or, to the extent construction is necessary, should reject Lumos's attempt to import a "linearly varying" limitation and a "single pointed tip" requirement into the claim.
a. The Claim Language Is Clear and Unambiguous
The claim recites "a tapered silicon waveguide having a tip width of less than 200 nanometers." Each element of this phrase is readily understood by a POSITA. A "tapered" waveguide is one whose width varies along its length. The waveguide has a "tip" — the narrow end of the taper — with a "width" that is "less than 200 nanometers." There is no ambiguity in this language, and no construction is needed. The term simply requires a silicon waveguide with a taper, where the tip of the taper is narrower than 200 nm.
b. Lumos's Construction Improperly Imports a "Linearly Varying" Limitation
Lumos proposes to add the requirement that the waveguide have "a linearly varying width." This limitation finds no support in the claim language, which says only "tapered" — not "linearly tapered." The specification of the '567 Patent discloses multiple taper profiles, including linear, parabolic, and exponential tapers. At col. 4, ll. 1–18, the specification states:
"The taper profile of the silicon waveguide may be linear, parabolic, exponential, or any other smooth or piecewise-smooth function that achieves gradual width reduction from the initial waveguide width to the tip width. In various embodiments, the taper profile is optimized to minimize insertion loss and maximize coupling efficiency."
This passage makes clear that the patent does not limit the taper to a linear profile. Lumos's construction would exclude parabolic and exponential tapers expressly disclosed in the specification — a result that is inconsistent with the intrinsic record and improperly narrows the claim. Phillips, 415 F.3d at 1323.
c. Lumos's "Single Pointed Tip" Requirement Is Unsupported
Lumos also proposes to require "a single pointed tip." The claim language says "a tip width of less than 200 nanometers" — a dimensional constraint — not a geometric shape requirement. A "tip" in this context is simply the narrow end of the taper; it need not be geometrically "pointed" in the sense of coming to a sharp point, nor must there be only a single tip. The specification discloses embodiments in which the taper terminates in a blunt tip with a finite width (less than 200 nm), not a geometrically pointed apex. Whitfield Decl. ¶¶ 42–45.
The Court should construe this term according to its plain and ordinary meaning, or, at most, as "a silicon waveguide that tapers to a tip width below 200 nanometers," without any requirement of linear variation or a single pointed tip.
3. "silicon nitride waveguide layer" (Claim 1). OWS's Proposed Construction: "a waveguide layer formed of silicon nitride (SiₓNᵧ)."
Lumos's Proposed Construction: "a waveguide layer composed of stoichiometric silicon nitride (Si₃N₄) deposited by low-pressure chemical vapor deposition (LPCVD)."
The Court should adopt OWS's construction because the claim recites "silicon nitride" without specifying stoichiometry or deposition method, and Lumos's proposed construction improperly imports both a compositional limitation and a process limitation into the claim.
a. "Silicon Nitride" Encompasses the SiₓNᵧ Family
A POSITA would understand "silicon nitride" to refer to the family of silicon nitride compositions, denoted SiₓNᵧ, which includes both stoichiometric silicon nitride (Si₃N₄) and silicon-rich or nitrogen-rich variations. The specification of the '567 Patent references "silicon nitride" generally without limiting the term to stoichiometric compositions. At col. 3, ll. 5–12, the specification states: "The waveguide layer comprises silicon nitride formed on the buried oxide layer of the SOI substrate." There is no reference to stoichiometric Si₃N₄ in the claims or in the passages of the specification describing the waveguide layer composition. Whitfield Decl. ¶¶ 46–49.
In the silicon photonics industry, both stoichiometric Si₃N₄ and silicon-rich silicon nitride (SiRN) are used for waveguide applications, with the choice depending on design requirements such as refractive index, stress, and optical loss. A POSITA reading the claim would not understand "silicon nitride waveguide layer" to be limited to the stoichiometric form. Id. ¶ 47.
b. Lumos's Construction Improperly Imports a Process Limitation
More fundamentally, Lumos's construction adds a deposition process limitation: "deposited by low-pressure chemical vapor deposition (LPCVD)." The claim is a product claim that defines the structure of the waveguide layer — a waveguide layer formed of silicon nitride — not the process by which it is made. Importing a deposition method (LPCVD) into a product claim is contrary to established Federal Circuit precedent. See Abbottott Labs. v. Sandoz, Inc., 566 F.3d 1282, 1293 (Fed. Cir. 2009) (rejecting construction that would convert a product claim into a process claim by importing manufacturing method limitations).
Silicon nitride waveguide layers can be deposited by multiple methods, including LPCVD, plasma-enhanced chemical vapor deposition (PECVD), sputtering, and other techniques. The specification does not limit the deposition method to LPCVD. Lumos's proposed construction would exclude waveguide layers that are compositionally and structurally identical to LPCVD-deposited silicon nitride but formed by a different deposition process — a result that makes no sense from a structural standpoint and finds no support in the claim language.
The Court should adopt OWS's construction: "a waveguide layer formed of silicon nitride (SiₓNᵧ)."
B. Disputed Terms of the '234 Patent
4. "graded buffer layer having a continuously varying germanium concentration" (Claims 1, 3, 5, 8, 11). OWS's Proposed Construction: "a buffer layer in which the germanium concentration changes across the layer thickness."
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."
The Court should adopt OWS's construction because it faithfully reflects the plain and ordinary meaning of the claim language, is consistent with the specification's disclosure of both continuously graded and step-graded buffer layer embodiments, and is not narrowed by the prosecution history. Lumos's construction improperly adds three limitations — monotonicity, absence of step changes, and directionality — that appear nowhere in the claim language and that would exclude embodiments expressly contemplated by the specification.
a. The Plain and Ordinary Meaning of "Continuously Varying"
The claim recites a graded buffer layer "having a continuously varying germanium concentration." A POSITA would understand this language to mean that the germanium concentration in the buffer layer is not uniform — i.e., it varies (changes) across the thickness of the buffer layer. The word "continuously" in this context indicates that the variation occurs across the layer — the concentration is continuously (i.e., persistently, throughout) varying as one moves through the buffer layer — rather than being constant.
Dr. Whitfield explains that in the context of semiconductor buffer layers, a "continuously varying" concentration is understood to distinguish a graded buffer from a uniform-composition buffer, in which the germanium concentration is the same throughout the layer. Whitfield Decl. ¶¶ 50–55. A POSITA would not read "continuously varying" as imposing a strict mathematical requirement of a continuous (i.e., non-step) function in the calculus sense. Id. ¶ 54. Rather, the term describes the overall characteristic of the buffer layer — its concentration varies throughout its thickness — as opposed to having a single fixed composition. Id.
This understanding is reinforced by the fact that the specification describes both "continuously graded" and "step-graded" buffer layers as embodiments of the invention, as discussed below.
b. The Specification Discloses Both Continuous and Step-Graded Buffer Layers
The specification of the '234 Patent expressly describes two types of graded buffer layers: continuously graded buffers (where the germanium concentration varies in a smooth, continuous fashion) and step-graded buffers (where the concentration changes in discrete steps across a plurality of sub-layers). At col. 4, ll. 15–42, the specification states:
"In one embodiment, the buffer layer has a continuously graded germanium concentration profile, in which the germanium atomic fraction varies smoothly from approximately 0% at the silicon substrate interface to approximately 100% at the absorber layer interface. In another embodiment, the buffer layer has a step-graded germanium concentration profile comprising a plurality of sub-layers, each having a substantially uniform germanium concentration, with the germanium concentration increasing from sub-layer to sub-layer."
Both embodiments are described as achieving the invention's objective of reducing threading dislocation density and enhancing photodetector responsivity. The specification does not indicate that one embodiment is preferred over the other, and neither is excluded by the claim language.
If the Court were to adopt Lumos's construction — which requires that the germanium concentration vary "without step changes" — the step-graded buffer layer embodiment expressly described in the specification would be excluded from the claims. This result would be at odds with the well-established principle that a claim construction that excludes a disclosed embodiment is "rarely, if ever, correct." Ridgeway & Co. v. Conceptronic, Inc., 90 F.3d 1576, 1583 (Fed. Cir. 1996). A POSITA would understand that a step-graded buffer layer, in which the germanium concentration changes from sub-layer to sub-layer, is one form of a buffer layer with a "continuously varying germanium concentration" — the concentration varies continuously in the sense that it changes across the layer thickness.
c. Lumos's "Monotonically" Limitation Is Unsupported
Lumos's construction requires that the germanium atomic fraction "increases monotonically." The claim language says "continuously varying," not "monotonically increasing." A "varying" concentration is one that changes — it could increase, decrease, or both. The specification does not limit the germanium concentration profile to a monotonically increasing function. While the primary disclosed embodiments show increasing germanium concentration from the substrate to the absorber, the claims do not impose this restriction, and a POSITA would understand that non-monotonic grading profiles (e.g., those with local maxima and minima) are also within the scope of a "continuously varying" concentration. Whitfield Decl. ¶ 56.
d. Lumos's Directionality Limitation Is Unsupported
Lumos also proposes to add "from the silicon substrate interface to the absorber layer" to the construction. This specifies where the grading begins and ends — a limitation not found in the claim language. The claim says simply "continuously varying germanium concentration" without specifying the direction or endpoints of the variation. While the specification describes embodiments in which the germanium concentration increases from the substrate to the absorber, the claim does not recite this specific directionality. Whitfield Decl. ¶ 58.
e. The Prosecution History Provides No Basis for Lumos's Construction
The '234 Patent was allowed on the first office action following an examiner interview. No narrowing amendments were made, and no disclaimers were entered during prosecution. There is simply no prosecution history basis for Lumos's proposed narrow construction. The absence of any narrowing amendment or disclaimer means that the claim terms should be given their plain and ordinary meaning, unconstrained by prosecution history arguments.
f. Expert Testimony Confirms OWS's Construction
Dr. Whitfield opines that a POSITA would understand "graded buffer layer having a continuously varying germanium concentration" broadly to encompass any buffer layer design in which the germanium concentration is not uniform across the buffer layer thickness — including both smoothly graded and step-graded profiles, with any directionality of variation. Whitfield Decl. ¶¶ 50–60. This expert testimony is consistent with and supports the construction derived from the intrinsic record.
The Court should adopt OWS's construction: "a buffer layer in which the germanium concentration changes across the layer thickness."
5. "enhanced responsivity" (Claim 1). OWS's Proposed Construction: Plain and ordinary meaning: "improved ability of the photodetector to convert incident optical power to electrical current."
Lumos's Proposed Construction: "a responsivity of at least 0.8 A/W at a wavelength of 1550 nm as measured under specified test conditions."
The Court should adopt OWS's construction because "enhanced responsivity" is a well-understood relative term describing an improvement over prior art photodetectors, not an absolute numerical threshold.
a. "Enhanced Responsivity" Is a Relative Term
The claim language recites "enhanced responsivity" — that is, responsivity that has been enhanced (improved) relative to what would be achieved without the graded buffer layer. This is a comparative term describing the result of the claimed structure, not an absolute performance metric. A POSITA would understand "enhanced responsivity" to mean that the graded buffer layer structure results in better responsivity than prior germanium-on-silicon photodetectors without such buffer layers — because the buffer layer reduces threading dislocation density, which in turn reduces dark current and improves the photodetector's ability to convert optical power to photocurrent. Whitfield Decl. ¶¶ 61–64.
b. The Specification Does Not Define a Minimum Numerical Threshold
The specification discusses enhanced responsivity as a functional benefit of the graded buffer layer structure but does not define a specific responsivity value as a minimum requirement. At col. 5, ll. 20–35, the specification states:
"The graded buffer layer reduces threading dislocation density in the germanium absorber layer, resulting in enhanced responsivity compared to germanium photodetectors grown directly on silicon without a graded buffer. The reduced dislocation density improves the crystalline quality of the germanium absorber, leading to higher photocurrent generation and lower dark current."
This passage confirms that "enhanced responsivity" is described in relative, qualitative terms — "enhanced responsivity compared to germanium photodetectors grown directly on silicon without a graded buffer" — not as an absolute numerical value.
c. Lumos Improperly Imports a Specific Value From a Single Example
Lumos's proposed construction of "a responsivity of at least 0.8 A/W at a wavelength of 1550 nm" is drawn from a single experimental example in the specification: Example 3 at col. 8, ll. 10–15, which reports a measured responsivity of 0.85 A/W at 1550 nm for one particular embodiment. Importing this specific numerical value from one experimental example into the claim is precisely the type of improper limitation-from-embodiment reading that Phillips prohibits. Phillips, 415 F.3d at 1323 ("Although the specification often describes very specific embodiments of the invention, we have repeatedly warned against confining the claims to those embodiments."). Different embodiments of the claimed photodetector — with different buffer layer thicknesses, germanium concentration profiles, and device