US 10718641

I'll search for this specific patent number across USPTO and litigation sources.

Let me search specifically for litigation and PTAB/CAFC activity.

Let me do a final targeted check for any Federal Circuit or PTAB proceedings.

US Patent 10,718,641 — Verified Summary

Search scope note: I ran targeted searches against patent databases (Google Patents, Justia, companyprofiles.justatic) and general litigation/docket sources (Docket Alarm, Dockets.Justia, PACER Monitor, CourtListener, RPX/Lex Machina 2025–2026 litigation trend reports). I did not find any district-court case, PTAB (IPR/PGR/CBM) proceeding, or Court of Appeals for the Federal Circuit docket — including any 2026 CAFC docket — in which US 10,718,641 is identified as the patent-in-suit. I cannot state with certainty that no such proceeding exists; absence of a hit in the searched sources is not proof of absence. Treat the "no litigation found" conclusion as a search result, not an authoritative negative.

⚠️ Identifier caution — do not conflate these with the patent:

• UK company number 10718641 appears in a UK Gazette filing (VOW TO LOVE WEDDINGS LTD / V O'REGAN LIMITED, April 2026). Unrelated to this patent.
• EP application 10718641.3 (Current 2 Current Limited, "Generation Apparatus," PCT/GB2010/000724) is a European application number, not this US patent.

Bibliographic Data (from the authoritative full text)

Abstract (verbatim): "A physical quantity measurement device includes a physical quantity measurement sensor and a metal block that is a holding member that includes block surfaces that serve as a plurality of holding portions for holding the physical quantity measurement sensor, and the block surface has a groove formed in a region facing the physical quantity measurement sensor."

Plain-Language Overview of the Independent Claims

Claim 1 — the device (the core independent claim). A physical-quantity measurement device (e.g., a multi-axis accelerometer module) built on a base plate, with a rectangular-parallelepiped holding block on top of it. Three of the block's surfaces run along the X-, Y-, and Z-axes, and each of those three surfaces has its own groove. Each groove runs in the same direction as its own surface (X groove on the X surface, etc.) and spans the entirety of the face — i.e., it is a full-length channel, not a partial notch. Three physical-quantity sensors are glued (adhesive) onto the three faces, and each sensor is positioned so it straddles its groove. Key structural/functional limits: (a) the three grooves are discrete from one another (not interconnected); (b) adhesive sits both under each sensor and in the groove; (c) each sensor contains a movable member, a fixed member, and a joint between them, and the joint runs parallel to the groove direction; (d) the groove is actually a pair of parallel grooves, and adhesive is applied only on the land between the pair; and (e) the adhesive's thermal expansion coefficient differs from that of the sensors. Practical effect: excess glue flows into the groove channels rather than spreading under the sensor's corners, mitigating thermal-stress concentration at the corners and preserving package airtightness/accuracy.

Claim 5 — electronic apparatus. An electronic apparatus containing the claim-1 device, plus a case housing it, where each of the three sensors detects a physical quantity of the apparatus, and the apparatus corrects its own shake based on those detected quantities (i.e., an image/camera-shake correction application, consistent with the video camera 300 and mobile phone 400 embodiments).

Claims 6, 7, and 8 — further electronic-apparatus claims. These are drafted in independent form but each simply re-invokes the apparatus of claim 5 while pointing back to the corresponding device claim: 6 → claim 2, 7 → claim 3, 8 → claim 4, each retaining the "case" and "shake corrected in response to the detected physical quantities" limitations. Practically, they are apparatus claims tied to the narrower device structures of claims 2/3/4.

Dependent device claims (for context):

• Claim 2: each groove intersects at least part of an edge of its sensor in plan view.
• Claim 3: the bonded region of the block overlaps the center of the sensor in plan view (stable bonding attitude).
• Claim 4: each sensor has a sensor element whose movable member extends from the fixed member via the joint member, and the movable-member extension direction intersects the groove direction.
• Claim 9: each sensor has a vibrating beam configured to measure the physical quantity (the piezoelectric resonator-based measurement described in the spec).
• Claim 10: the groove directions intersect the directions of the corresponding vibrating beams in plan view.

Claim count: 10 total (1–4 and 9–10 device; 5–8 electronic apparatus).

Points of Uncertainty

1. Litigation/PTAB/CAFC: No 2026 CAFC docket, PTAB petition, or district-court case tied to this patent number surfaced. I cannot rule out sealed, unindexed, or very recent filings. I also did not have authenticated, direct USPTO PatentCenter/PACER API access — findings rest on indexed web sources.
2. Litigation-source caveat: Near-miss search returns (e.g., Flick Intelligence, LLC v. Seiko Epson Corp., W.D. Tex. 7:25-cv-00449; American Patents LLC v. Seiko Epson, E.D. Tex. 4:21-cv-00718) involve Seiko Epson but the available records do not identify US 10,718,641 as an asserted patent. Do not assume association.
3. Assignee identity: Seiko Epson Corporation is listed as both original and current assignee; no reassignment away from Seiko Epson is recorded in the source text.
4. Claim-scope note: The claim set here differs meaningfully from the "Application Examples" in the specification (e.g., the granted claim 1 requires three orthogonal faces, full-length paired grooves, adhesive-in-groove, and the differing-thermal-expansion-coefficient limitation) — a reminder that the prosecution-history-informed claim scope, not the summary section, governs.

I'll search multiple litigation databases and sources for this specific patent number.

Let me search more targeted litigation databases.

Let me try the Unified Patents litigation portal and other case databases directly.

Litigation Search Results for U.S. Patent No. 10,718,641

Bottom line: No litigation known for this specific patent.

Based on my searches of publicly available litigation-related sources, I found no record of any lawsuit asserting U.S. Patent No. 10,718,641 ("Physical quantity measurement device and electronic apparatus," Seiko Epson Corp.). The only case-related references returned were for other, unrelated patents.

What I searched and what came back

The Google Patents page for US10718641B2 shows citations of prior art and subsequent citing documents, but no "Litigation" / "Cases" section, which is consistent with the patent never having been asserted in a filed suit (or at least none indexed).

Important caveats (stated honestly)

1. I could not directly query the primary sources you named. My searches did not return data from the Unified Patents litigation caselist (portal.unifiedpatents.com/litigation/caselist), PACER, or the CAFC docket for this patent number. The tool results were limited to general web hits. Therefore I cannot rule out a suit that is only indexed in those closed/portal databases.

2. This is a 2020 patent with a short assertion window so far. It issued July 21, 2020, and has a listed adjusted expiration of Feb. 17, 2038. It is a Seiko Epson product patent (physical quantity/acceleration sensors used in cameras, phones, etc.), typically enforced defensively rather than through an NPE campaign, which reduces the likelihood of a publicly tracked assertion.

3. Do not rely on this as a legal conclusion. Before treating "no litigation" as definitive, verify directly in:

   • Unified Patents Portal — https://portal.unifiedpatents.com/litigation/caselist
   • PACER / CourtListener RECAP — docket search on "10,718,641"
   • CAFC docket search (for any appeal)
   • Lex Machina / Docket Navigator (commercial, most complete)

Summary table

If you need a definitive negative, I'd recommend a paid Docket Navigator or Lex Machina query, or a direct PACER docket search — those are the only sources I'd rely on to certify that no case exists.

I'll verify the "no PTAB activity" default against public sources before finalizing.

Let me run additional targeted checks against PTAB/court docket aggregators and the Family/related JP patent.

Let me run a final couple of checks to be thorough.

Proceedings overview

Zero AIA trial proceedings. The USPTO Open Data Portal structured record ("PTAB proceedings on file") returns no IPR, PGR, or CBM proceeding naming US 10,718,641 — 0 active, 0 with claims invalidated, 0 with claims sustained, 0 settled, 0 institution denials. My independent web checks corroborated this: searching the patent number against PTAB-facing and litigation-facing sources returned no petition, institution decision, Final Written Decision, or Federal Circuit appeal at all.

Bottom line for a defendant: the '641 patent is completely untested at the PTAB. There is no FWD holding any claim unpatentable, no estoppel running against anyone, and no Board precedent construing the claim terms for you. This is not a "hardened" patent that survived attack — it is simply an unexamined one. The corollary is that there is also no roadmap: you would be the first petitioner, and you would bear the cost of developing the invalidity case from scratch. Concretely, that means a strong § 102/§ 103 challenge remains fully available, but an IPR-based defense strategy cannot be justified by "someone else already did the work."

No proceedings to enumerate

There are no proceedings to list. Stating this explicitly rather than padding the section:

Source of the negative: the structured "PTAB proceedings on file" block supplied in this prompt (USPTO ODP ingest). That block is authoritative for the AIA-trial question and is the reason I am not treating this as merely a failed search.

⚠️ Disambiguation — Seiko Epson is in active PTAB litigation, but not on this patent

My searches surfaced live PTAB activity involving Seiko Epson, and it is important that no one conflates it with the '641 patent:

• [BOE Technology Group Co., Ltd.](/litigations/by-plaintiff/BOE%20Technology%20Group%20Co.%2C%20Ltd.) v. 138 East LCD Advancements Ltd. — IPR petitions on Seiko-Epson-origin LCD display patents (identified in the filings as the '146 and '512 patents), addressing terminal-area/contact-resistance subject matter. These are LCD display patents from Epson's divested display business, sold/transferred to an IPValue-affiliated entity (138 East). Petitioner is BOE. None of the challenged patents is US 10,718,641. (Petitioner brief, PTAB, related papers)
• Flick Intelligence, LLC v. Seiko Epson Corporation et al., No. 7:25-cv-00449 (W.D. Tex.) — a Seiko Epson defendant-side case; no record identifies the '641 patent as asserted. (PacerMonitor docket)

These are Epson patents/IP assets in PTAB, but not this patent. Treat the distinction as material — see the cross-reference flags below.

Identifier trap (carried forward): UK company number 10718641 (V O'Regan Limited, per the London Gazette) and EP application 10 718 641.3 (Current 2 Current Limited, PCT/GB2010/000724 — a hydropower/generation apparatus) are unrelated to this US patent. Any search that "finds" one of these and reports it as the patent is wrong.

Strategic summary

Claim status — all 10 claims UNTESTED. Because no AIA trial has ever been instituted, the claim-level picture is simply the as-issued claim set: independent device claim 1, dependent device claims 2, 3, 4, 9, 10, and electronic-apparatus claims 5, 6, 7, 8. Nothing is canceled, nothing is confirmed by the Board, and nothing has been construed in an AIA forum. Contrast this with a patent that has been through IPR: here there is no § 318(b) certificate canceling anything, and no claim has been narrowed by amendment. The consequence worth stressing to a defendant is that the issued scope governs — including the fairly specific limitations in claim 1 (three orthogonal faces with per-face full-length paired grooves; grooves discrete from each other; adhesive both under each sensor and in the grooves, applied only on the land between each pair; each sensor's joint member running parallel to its groove; adhesive CTE differing from the sensor CTE). Those are the limitations an invalidity case would have to meet.

Estoppel landscape — a blank slate. 35 U.S.C. § 315(e)(2) estops only a petitioner (and its real parties in interest and privies) that has been through an IPR that resulted in a final written decision. There has been no such FWD here, so no estoppel attaches to anyone. Every § 102/§ 103 ground based on patents and printed publications remains available to any defendant, and there is no risk that a prior petitioner has already burned a reference set. Equally, there is no § 315(a)(1) bar and no § 315(b) one-year clock running — because there is no identified complaint alleging infringement of this patent (consistent with the previously-generated litigation summary finding no assertion).

Pattern signals — none of the usual tells. No petitioner has filed on this patent at all, so there is no repeat-petitioner pattern. The patent owner has never had occasion to defend at the Board, so there is no history of aggressive PTAB appeal practice (nothing to appeal). No defensive aggregator — Unified Patents, RPX, AST, LOT — appears anywhere in the chain; the previously-generated assignment analysis found exactly one recorded conveyance (Reel 044854/0018, inventors Kameta and Nishio → Seiko Epson, executed 2018-01-09), leaving title with the original operating manufacturer. That ownership fact is itself the best structural explanation for the empty PTAB docket: an operating company that is not running an assertion campaign generates no IPR petitions against itself.

What the absence signals. A well-asserted patent — particularly one pushed by an NPE across multiple defendants — reliably attracts IPRs, often multiple, because the economics of a § 315(b)-triggered filing become compelling for each accused infringer. The '641 patent has none. The most parsimonious reading: this patent has not been commercially asserted (or at least not in any indexed, publicly visible campaign). For a defendant, that cuts both ways — it suggests a low probability that the patent is weaponized against you, but also means that if a demand letter does arrive citing it, you would be first, and the absence of any prior challenge tells you nothing about the strength of the claims.

Recommended next steps

If you are a defendant facing assertion of this patent:

1. There is no FWD, so there is nothing to cite and no cancellation to leverage. Do not look for a PTAB decision to brandish — none exists. Your invalidity case would be built from the ground up against the issued claims.
2. The § 315(b) clock has not started on the available record. Confirm whether you have been served with a complaint alleging infringement of US 10,718,641 specifically — not merely a demand letter, which does not trigger the one-year bar. See [35 U.S.C. § 315(b)].
3. Claim 1's dense limitation set is your target. The granted claim is considerably narrower than the specification's "Application Examples" — it requires, among other things, three orthogonal faces, full-length paired grooves per face, grooves discrete from each other, adhesive in the grooves and only on the land between each pair, and an adhesive CTE differing from the sensors. Each of those is an independent non-infringement and invalidity lever, and none has been construed by the Board.
4. Check the prior art cited on the face of the patent as a starting point for a § 102/§ 103 ground: EP 0644586 A2 (Manneskmann Kienzle — cited with an English translation as a non-patent citation), JP 2004-037105 A (Akashi, the background reference), US 2006/0130584 A1 (Fujitsu Media Devices), US 2007/0102810 A1 (Mitutoyo), US 8,256,287 B2 (AAI Corp), US 2014/0123754 A1 and JP 2014-119412 A (both Seiko Epson), and US 2017/0248553 A1 (RMCI). Note that § 325(d) discretion is a real risk where you rely on art the Examiner already considered — weigh that before making a prosecution-art-only petition the centerpiece.
5. Timing on any petition. With no parallel litigation identified, a Fintiv-based discretionary denial is unlikely on the current record, but that changes the moment a suit is filed — the Board's timing factors track the district court's investment at the institution stage. If you intend to petition, filing early, before a case matures, preserves optionality.

If no PTAB activity exists and you simply want the negative certified: it is worth being precise about what I can and cannot support.

• What is supported: the ODP ingest shows no AIA trial proceeding for US 10,718,641, and my searches found none. Verify directly and cheaply at the USPTO PTAB E2E / Patent Trial and Appeal Board proceedings search and the PTAB Decisions database by entering patent number 10,718,641.
• What is not covered by that negative: the ODP "proceedings" block addresses AIA trials (IPR/PGR/CBM). It is not a certificate that no ex parte reexamination, inter partes reexamination (legacy), reissue, or derivation proceeding exists, and I did not independently verify those channels in this session. Nor does it speak to foreign proceedings — notably JP 6819338 B2, the Japanese family member, where a mukō shinpan (invalidation trial) or opposition is theoretically possible and unknown to me. If any of those matter to your strategy, they need a separate, targeted check.
• Caveat on method: my searches surfaced only general web/indexed results; I did not have authenticated API access to PTAB E2E, and I could not complete every planned query before running out of search steps. Treat "no proceedings found" as strongly supported but not a certified negative until confirmed against the PTAB E2E docket directly.

Cross-reference flags for the earlier sections

• Consistency with the Patent Summary and Litigation Summary. Both prior sections concluded "no litigation found" for this patent. The zero-PTAB result is mutually corroborative, and the causal link is the assignment record: with title never having left Seiko Epson (the sole conveyance is Reel 044854/0018), the patent was never handed to an NPE assertion vehicle of the kind that drives IPR volume. No contradiction between this section and the earlier three.
• One point to sharpen, not contradict. The assignment section noted Seiko Epson does hold and maintain multi-axis physical-quantity sensor/IMU products (e.g., the M-G3xx IMU line) and that the '641 embodiments target camera-shake correction in video cameras and mobile phones. That commercial-context observation is fully compatible with the empty PTAB docket — an operating company enforcing its own product patents defensively, or not at all, produces exactly this record. Nothing in the PTAB search changes the assignment verdict of "Insufficient data (only the original assignment)."
• New caution to add to the prior sections' disambiguation list. The Seiko-Epson-related PTAB activity surfaced above (BOE v. 138 East, on the LCD '146/'512 patents) is a fresh conflation trap. It involves Seiko Epson and involves the PTAB, so a keyword search can surface it next to this patent — but the patents, the technology (LCD display terminals), the parties (BOE as petitioner; an IPValue affiliate as owner), and the posture are entirely different. Do not attribute any of it to US 10,718,641.
• Identifier warning stands. UK company 10718641 and EP application 10 718 641.3 remain unrelated, as flagged in both prior sections.

I'll verify the assignment record for US 10,718,641 against the USPTO Assignment Center and cross-reference sources.

Let me verify the assignment reel/frame details and check for any Seiko Epson patent transfer activity.

I hit the tool-call ceiling mid-verification. Below is what the records support; one field (correspondent) I could not retrieve, and I say so rather than guess.

Inventors

Unusual-pattern check — clean:

• The two inventors assigned to the same entity on the same date, six days before the family's US filing (2018-02-07) and roughly a year after the JP priority filing (2017-02-13). This is the ordinary corporate-employee invention-assignment pattern, not a split-rights or retained-rights arrangement.
• No evidence of inventor departure from Epson within 12 months of filing, and no second assignment from either inventor to any other entity appears in the record. Note the limit of this finding: the Assignment Center corpus records transfers, not employment, so a departure would only be visible here if the inventor had assigned rights elsewhere — which they did not.
• Both inventors are named only on the US case and its foreign counterparts (JP 2018-132311 / JP 6819338 B2; CN 108427010 A/B) — no orphaned individual-owned siblings surfaced.

Original assignee

Seiko Epson Corporation (セイコーエプソン株式会社), 4-1 Nishi-Shinjuku 2-chome, Shinjuku-ku, Tokyo, Japan. Listed as both original and current assignee; the assignment record shows no conveyance away.

• Primary line of business: operating diversified electronics manufacturer — inkjet printers and printing systems, projectors/display, quartz-crystal (QMEMS) sensing devices and inertial measurement units, timepiece movements, and industrial/robotics equipment. Publicly listed on the Tokyo Stock Exchange. I did not retrieve a current TSE filing in this session, so treat the listing/exchange status as high-confidence background rather than a verified primary-source cite.
• Current status: operating, not acquired, not dissolved, not in bankruptcy. Seiko Epson remains an independent, listed operating company. The 4th-year maintenance fee was paid 2024-01-03 and the patent is listed Active, expires 2038-02-17.
• Product practice: the claims are directed to a multi-axis physical-quantity measurement device (three quartz vibrating-beam sensors bonded to orthogonal faces of a metal block, with adhesive-relief grooves). That is squarely within Epson's quartz-crystal inertial sensor product line. I would characterize this as strongly consistent with an embodied commercial line rather than certified — I did not obtain a product teardown or datasheet mapping the claim elements to a specific SKU in this session.
• The three foreign family members (JP 6819338 B2, CN 108427010 B) likewise remain with Seiko Epson in the available records — no leakage of the family to a third party.

Assignment timeline

One (1) recorded assignment in the chain — the original inventor-to-corporation assignment. There are no post-issuance assignments.

• 2018-01-09 (executed) / recorded 2018-02-07 — Reel 044854/0018
  • Conveyance: Assignment (original, inventor → company)
  • Assignor: KAMETA, TAKAHIRO; NISHIO, SHINJI
  • Assignee: SEIKO EPSON CORPORATION (Japan)
  • Correspondent: ⚠️ Not retrievable. The structured legal-events record for this patent exposes the reel/frame, the assignors, the free-format text ("ASSIGNMENT OF ASSIGNORS INTEREST; … REEL/FRAME:044854/0018") and the effective date, but not the correspondent of record. I was unable to open the underlying Assignment Center abstract for 044854/0018 in this session. I am not going to infer the correspondent name. The only Seiko Epson recording-conveyance data point that surfaced anywhere in my searches was a 2005-era Epson assignment (Reel 017431/0184, correspondents G. Gregory Schivley / Harness, Dickey & Pierce, P.L.C.) — that is a different patent, 13 years earlier, and is not evidence about this one. Flagging it only so you know the query was run.
  • Context: Original corporate assignment, executed on the eve of filing and recorded the same day the application was filed (2018-02-07) — the standard "assignment filed with the application" workflow for a company-filed case, not a purchase.

If you need the correspondent field for the repeat-player check, it has to come from a live Assignment Center abstract query on reel/frame 044854/0018 — that is the one record in the chain, and it is the only place the data exists.

Timeline diagram

timeline
    title Ownership of US 10718641
    2017 : JP priority application filed by Epson
    2018 : Inventors assign rights to Seiko Epson
         : US application filed same day
         : Assignment recorded at reel 044854 frame 0018
    2020 : US patent 10718641 issues
    2021 : JP family member issues
         : CN family member issues
    2024 : Fourth year maintenance fee paid

NPE / troll-pattern signals

Also checked and negative: the name-adjacency trap flagged in the earlier sections — UK company number 10718641 (V O'Regan Limited) and EP application 10718641.3 (Current 2 Current Limited) — are unrelated identifiers and do not appear as assignees/assignors here. Neither is in the chain.

Verdict

Insufficient data — on the constrained taxonomy, because the record consists of only the original assignment.

Read that verdict carefully, because it means the opposite of "we don't know who owns it." The chain is not blank — it contains exactly one link, recorded at Reel/Frame 044854/0018, executed 2018-01-09 and recorded 2018-02-07, conveying the invention from Kameta and Nishio to Seiko Epson Corporation, and nothing since. That is the "no post-issuance assignments" finding the brief anticipates: the original assignee still owns the patent. There is no NPE chain to grade, no asserter to name, and no correspondent recurrence to test, because no second party has ever touched this patent. The exercise is not "operating-company assertion" either, since no suit naming US 10,718,641 was identified in the earlier litigation section — the patent is being held, not enforced.

What would change the verdict: a single recorded conveyance out of Seiko Epson to a licensing entity (that alone would flip signals 1, 4 and likely 5), or a documented Epson→NPE transfer surfaced via a 10-K/8-K or an RPX/Unified asserter-directory hit. None exists today.

Verify at: USPTO Assignment Center and the legacy search at assignment.uspto.gov — query patent number 10718641 or app 15/890,591; the single hit should be reel/frame 044854/0018. That abstract is also where the missing correspondent field sits, and pulling it is the one open thread in this reconstruction.

⭐ Caveat on my two hard limits: (1) I could not retrieve the correspondent of record for 044854/0018 — that is a genuine gap, not a negative finding, and the earlier sections' "no litigation" conclusion carries the same "absence of evidence" caveat. (2) I did not have live Assignment Center abstract or PEDS/ODP field-level access in this session; the reel/frame and dates above come from the patent's own structured legal-events record as mirrored on Google Patents, which is consistent with but not a substitute for the Assignment Center abstract.

I'll research each cited reference to provide accurate prior-art analysis.

Let me research the remaining references.

Prior-Art Analysis — US Patent 10,718,641 B2

Patent under analysis: US 10,718,641 B2, "Physical quantity measurement device and electronic apparatus" (Seiko Epson; filed 2018-02-07; priority JP 2017-023837, 2017-02-13; issued 2020-07-21).

Scope / method note. The authoritative full text lists exactly eight patent citations and one non-patent citation on the face of the patent (the examiner-cited "Citations (8)" set). I reconstructed the citations and dates from the patent's own front page and cross-checked each against the primary patent documents (Google Patents, EPO publication server, Justia) where retrievable. Three references — US 2014/0123754 A1, JP 2014-119412 A, and US 2017/0248553 A1 — I could only partially verify before hitting my research limit; for those I give the bibliographic data from the patent face and title-level description, and I flag where the substance is unverified rather than guessing. This builds on the earlier sections (bibliographic, litigation, PTAB, assignment) and does not repeat them.

⚠️ Identifier caution carried forward from earlier sections: UK company number 10718641 and EP application 10 718 641.3 are unrelated to this US patent and are not prior art here.

1. The cited references at a glance

All are § 102(a)(1)/(a)(2) art because this patent is post-AIA (effective filing 2017-02-13) and every reference predates that date.

Critical observation drawn from the record: the examiner cited an English translation of EP 0644586 (Lindner) as a non-patent citation. That is the classic signature of the primary reference — the examiner evidently relied on Lindner for the inventive concept (a groove that limits/controls the adhesive footprint) and used JP 2004-037105 for the multi-axis mounting architecture. Everything else reads as secondary/background.

2. Reference-by-reference analysis

Reference 1 — EP 0644586 A2 (Lindner, Mannesmann Kienzle) — the most relevant reference

• Full citation: EP 0 644 586 A2, "Anordnung zum Befestigen eines mikromechanischen Sensors auf einem Träger durch Kleben" ("Arrangement to attach a micromechanical sensor on a support by glueing"), Mannesmann Kienzle GmbH; inventors Schreiber-Prillwitz, Plankenhorn, Lindner. Priority 1993-09-17; filed 1994-08-19; published 1995-03-22.
• Description: A glass-silicon-glass micromechanical acceleration sensor glued to a ceramic/metal carrier. To stop capillary creep of the adhesive and prevent clamping stress on the bending oscillator, two grooves (Nute 31, 32) are cut into the glass layer to bound a precisely defined adhesive surface (Klebefläche 33); groove 32 acts as an overflow stop for excess glue. The stated problem is exactly the '641 patent's problem — thermal-expansion mismatch between adhesive and sensor producing stress and measurement error. The patent's own "Related Art" section does not cite Lindner; the examiner added it.
• §102 relevance: This is the closest single reference to the core inventive concept of claim 1 — a groove in the region facing the sensor into which extra adhesive flows. Under § 102 it discloses the groove/adhesive-control element and the CTE-differential motivation.
• Anticipation assessment: Does not anticipate claim 1 in full (it is a single sensor on a single flat carrier — no rectangular-parallelepiped holding member, no three orthogonal faces, no full-length paired grooves spanning an entire face, no "adhesive only between the pair"). It could anticipate a hypothetical claim drawn to "a sensor bonded via adhesive with a groove in the facing support region" — i.e., the specification's Application Example 1 concept. Against the granted claims it is 103 material, not 102.

Reference 2 — JP 2004-037105 A (Akashi Corp) — the closest architecture reference

• Full citation: JP 2004-037105 A, "Physical quantity detector and method of manufacturing physical quantity detector," Akashi Corp. (JP app. 2002-190636,) filed 2002-06-28; published 2004-02-05.
• Description (verified from the document): A capacitive physical-quantity detector in which a sensor chip is attached to each of two or more wall portions orthogonal to one another of a mounting member (取付部材); wiring via a flexible substrate. Claims recite the sensor chip detecting displacement/velocity/acceleration via a movable electrode plate on an elastic portion facing a fixed electrode plate. This is the very reference the '641 specification identifies and distinguishes as "Related Art" ("JP-A-2004-37105").
• §102 relevance: Discloses the multi-axis mounting concept central to claim 1 — a holding member with multiple holding portions holding sensors so that orthogonal physical quantities can be measured.
• Anticipation assessment: Does not anticipate claim 1 — it has no groove in the facing region, no adhesive-in-groove, no paired full-length grooves, and no differing-CTE adhesive limitation. It is highly relevant to the "holding member having a plurality of holding portions on orthogonal faces" element, and is the primary obviousness partner when combined with Lindner. Against the electronic-apparatus claims (5–8) it is background only.

Reference 3 — US 2006/0130584 A1 (Fujitsu Media Devices; Tanaka et al.)

• Full citation: US 2006/0130584 A1, "Acceleration sensor," Fujitsu Media Devices Ltd.; pub. 2006-06-22 (priority 2004-12-16). Family includes US 7,299,695 B2 and EP 1 672 374 A1.
• Description (verified): A "parasol-type" MEMS acceleration sensor with a support portion, a beam portion bearing a detection element, and a suspended weight, plus a stopper substrate fixed by adhesive to the support/frame, limiting overswing. One dependent aspect expressly recites "a groove between the extension of the support portion and the pad."
• §102 relevance: Relevant to claim 1's "movable member, fixed member, and joint member between them" element, and to the general use of a groove adjacent a bonded region. Its family member describes acceleration sensors used to prevent hand shake in a video camera — relevant to the shake-correction apparatus claims 5–8.
• Anticipation assessment: Does not anticipate claim 1 — no multi-face orthogonal sensor mounting and no adhesive-in-groove-bounding-the-bond-region teaching. Relevant to the sensor-element structure and to the apparatus claims 5–8 as prior use context.

Reference 4 — US 2007/0102810 A1 (Mitutoyo; Matsumiya et al.)

• Full citation: US 2007/0102810 A1, "Sensor block," Mitutoyo Corp.; filed 2006-11-07; pub. 2007-05-10 (priority JP 2005-322565, 2005-11-07). Status: abandoned.
• Description (verified): A sensor block comprising a common base block (glass substrate) on which sensor chips are mounted on the X, Y and Z faces to measure accelerations in three axes; base block described as "a substantially regular hexahedron having at least one face bonded to the semiconductor substrate." It expressly discusses the accuracy problem arising from "a difference in thermal expansion coefficient between an adhesive and an element adhered thereto" and thermal stress at bonding.
• §102 relevance: Strongly relevant to claim 1's rectangular-parallelepiped holding member with three orthogonally-oriented sensor-bearing faces, and its discussion of adhesive CTE mismatch is directly on point for claim 1's differing-CTE limitation.
• Anticipation assessment: Does not anticipate claim 1 — it uses anodic bonding with no adhesive (explicitly to avoid adhesive CTE problems) and has no groove. Its multi-face hexahedral mounting makes it a strong 103 partner with Lindner. (Note: the same assignee's later US/EP filings on position sensors surfaced in search are unrelated.)

Reference 5 — US 8,256,287 B2 (AAI Corporation)

• Full citation: US 8,256,287 B2, "Sensor shield retention device and system," AAI Corporation; provisional 2009-04-15; filed 2010-04-15; issued 2012-09-04. (Pub. US 2010/0263444 A1, 2010-10-21.)
• Description (verified): A snap-lock shield/windscreen retention device for microphones and other sensors, with flex petals, latches, and "adhesive traps" / concentric undercuts that capture adhesive so a foam windscreen bonds to the device. It is a mechanical retention invention for acoustic/field sensors, not a multi-axis MEMS package.
• §102 relevance: The only doctrinal hook is the generic notion of trapping adhesive in recesses/undercuts to control where glue goes.
• Anticipation assessment: Does not anticipate any of claims 1–10. Wrong field (microphone windscreens), no holding member with orthogonal sensor faces, no paired full-length grooves, no movable/fixed/joint sensor element. This reads as a citation of convenience for the "adhesive trap" idea; low value.

Reference 6 — US 2014/0123754 A1 (Seiko Epson) — substantively unverified this session

• Full citation: US 2014/0123754 A1, "Physical quantity detecting device, electronic apparatus, and moving object," Seiko Epson Corporation; priority 2012-11-07; published 2014-05-08.
• Description: Same assignee/family context as the patent. Title-level: a physical-quantity detection device packaged for an electronic apparatus / moving object. I could not verify the specific disclosure (whether it recites grooves, multi-face mounting, or adhesive control) before my research limit; treat the description as title-level only. Note it is same-assignee art, which is common in a dense Seiko Epson sensor portfolio (see also JP 2014-119412 below).
• §102 relevance: Cannot be assessed reliably without the document. Given the assignee and title, it may disclose the sensor-package/multi-axis context rather than the groove feature.
• Anticipation assessment: Cannot responsibly state it anticipates any claim. Flagged for retrieval.

Reference 7 — JP 2014-119412 A (Seiko Epson) — substantively unverified this session

• Full citation: JP 2014-119412 A, "Multi-axis physical quantity detection apparatus, manufacturing method of multi-axis physical quantity detection apparatus, electronic apparatus and moving body," Seiko Epson Corp.; priority 2012-12-19; published 2014-06-30.
• Description: Title-level: a multi-axis physical-quantity detection apparatus with a manufacturing method — i.e., directly in the field of claim 1's three-axis sensor arrangement.
• §102 relevance: Likely relevant to the multi-axis mounting architecture of claim 1 and to the electronic-apparatus/vehicle context of claims 5–8. Same-assignee art.
• Anticipation assessment: Cannot responsibly state it anticipates any claim without the document. Given its title and date, it is a plausible 103 partner on the multi-axis architecture. Flagged for retrieval.

Reference 8 — US 2017/0248553 A1 (RMCI, Inc.)

• Full citation: US 2017/0248553 A1, "Sensor mounting pad with secondary restraint feature," RMCI, Inc.; priority 2016-02-25; published 2017-08-31.
• §102 basis: Because it published after the 2017-02-13 priority date, it is not § 102(a)(1) prior art; it qualifies, if at all, only as § 102(a)(2) art (US application publication, effectively filed 2016-02-25, i.e., before the priority date).
• Description: Title-level: a sensor mounting pad with a secondary restraint (i.e., adhesive/mechanical mounting of a sensor). Substance unverified this session.
• §102 relevance: Potentially relevant to the sensor-to-holding-member mounting/adhesive aspect broadly.
• Anticipation assessment: Cannot responsibly state it anticipates any claim. Given the title, it does not appear to disclose grooves or three-face orthogonal mounting; likely low relevance. Flagged for retrieval.

3. Anticipation matrix — mapping references to the granted claims

Because no single reference discloses all elements of claim 1 (three orthogonal faces + full-length paired grooves + adhesive-in-groove + movable/fixed/joint member with joint parallel to groove + adhesive only on the land between the pair + differing CTE), I record no full § 102 anticipation of claim 1. The table identifies where each reference bears on a claim element — i.e., § 102 single-reference potential only for narrow sub-concepts, and § 103 combinations otherwise.

Bottom line on anticipation: None of the eight cited references anticipates claim 1 (or claims 5–8) in a single-reference § 102 sense. Their combined teaching — multi-face orthogonal sensor mounting (JP 2004-037105 / US 2007/0102810) plus adhesive-bounding grooves (EP 0644586) — is an obviousness case, which is consistent with the prosecution history reported earlier (an active 102/103 fight with a reopen-after-final before allowance). The claim limitations that the examiner apparently used to grant — full-length paired grooves spanning the entirety of each face, with adhesive applied only on the land between the pair, and the differing-CTE glue limitation — are precisely the elements absent from this art set, which is why the granted claims survived.

4. Ranking: the most relevant prior art

1. EP 0644586 A2 (Lindner) — most relevant to the inventive concept. Groove-bounded adhesive region, overflow stop, and the identical CTE-mismatch rationale. Examiner cited its English translation as NPL — a strong signal it was the primary reference.
2. JP 2004-037105 A (Akashi) — most relevant to the claim architecture. Multi-face orthogonal sensor mounting; expressly the patent's own distinguished "Related Art."
3. US 2007/0102810 A1 (Mitutoyo) — hexahedral base block with X/Y/Z-face sensors; explicit adhesive-CTE discussion. Strong second architecture reference.
4. US 2006/0130584 A1 (Fujitsu/Tanaka) — support/beam/weight sensor element, adhesive-bonded stopper, groove-adjacent-pad aspect; video-camera hand-shake context.
5. JP 2014-119412 A (Seiko Epson) and US 2014/0123754 A1 (Seiko Epson) — same-assignee multi-axis/sensor-package art (substance unverified this session — retrieve before relying).
6. US 2017/0248553 A1 (RMCI) — mounting-pad restraint; marginal, and only § 102(a)(2).
7. US 8,256,287 B2 (AAI) — windscreen adhesive traps; lowest relevance, different field.

5. Caveats and flags (stated honestly)

1. Three references are title-level only in this report. US 2014/0123754 A1, JP 2014-119412 A, and US 2017/0248553 A1 were not substantively verified before my research limit. Do not treat the § 102 assessments for refs 6–8 as complete. Retrieve the full documents to finalize.
2. This is the examiner-cited set, not an exhaustive prior-art search. "Prior art for the patent" here = the references of record. A validity challenge would need art outside this set to avoid a § 325(d) discretionary denial (a point already developed in the earlier PTAB section).
3. No reference anticipates claim 1 as granted. The single-reference § 102 analysis above is element-level; full anticipation of the granted claim (with its paired full-length grooves, adhesive-only-on-the-land, and differing-CTE limitations) is not made out by any cited reference.
4. Post-AIA framing. Because this patent's effective filing date is 2017-02-13, § 102(a)(1)/(a)(2) govern; EP 0644586 and the Japanese publications are § 102(a)(1) printed publications, and US 2017/0248553 is only potentially § 102(a)(2) art.
5. Citation-fidelity note: the reference is spelled JP 2004-37105 / JP 2004-037105 A in the record; the patent's specification calls it "JP-A-2004-37105." Both denote the same Akashi Corp. document. I have not altered either form.

Verify all three flagged references at Google Patents / EPO / JPO before finalizing any invalidity opinion or petition ground.

I'll ground the § 103 analysis in the actual disclosures of the cited references before mapping them to the claim limitations.

Obviousness Analysis — US 10,718,641 B2 under 35 U.S.C. § 103

Basis and scope. This analysis uses the prior-art reference set appearing on the face of US 10,718,641 B2 (the "Prior Art section of this page") plus the specification's background admissions. It builds on the earlier sections (patent summary, litigation, PTAB, assignment) rather than repeating them. Where I could not independently verify a reference's disclosure in this session, I say so explicitly. The header date given in the task (2026-04-26) differs from my system date; nothing below turns on the difference.

Working assumptions: (i) the '641 claims are entitled to the 2017-02-13 JP priority date (JP 2017-023837) — if any claim is not, the prior-art date window shifts and more art becomes available; (ii) all references were publicly available before the critical date unless noted.

1. The claim-1 limitation set (the target)

Claim 1 is dense. The independent claim requires, in substance:

The specification is explicit that L14 describes the known problem, not a discovery: the "Related Art" section states that in JP-A-2004-37105 the sensor chip's airtightness "is destroyed due to the difference in the thermal expansion coefficient between a bonding member, such as an adhesive, and the sensor chip." That admission is a motivation-to-combine anchor.

2. The reference set — what each actually teaches

EP 0 644 586 A2 (Mannesmann Kienzle; Lindner et al.) — the primary groove/adhesive reference. Verified against Google Patents (EP0644586A2). It discloses an arrangement for fastening a micromechanical sensor (glass-silicon-glass bending oscillator) to a carrier by gluing, in which grooves bound a precisely defined adhesive surface: "At the boundaries of this adhesive surface, adhesive creep caused by capillary forces is prevented on the one hand, on the other hand, the adhesive requirement for this surface can be dosed relatively precisely, i.e. in the case of a strip-shaped adhesive surface, preferably in the form of a caterpillar or as a row of drops." It further states the purpose: the sensor "is not fixed over its entire surface … and thus also a clamping effect, by means of which stresses in the area of the bending oscillator can be caused, is avoided." It explicitly attributes the problem to "differences in the temperature coefficients of the sensor and its support" and to "the connection technology of the sensor and support." It also notes the grooves "can be made in the course of the manufacturing process." This reference teaches the core inventive concept of the '641 (a groove that limits adhesive spread and thereby decouples thermal stress from the sensing region) — 24 years before the '641 priority date.

JP-A-2004-37105 (Akashi Corp.) — the multi-axis mounting reference (the admitted background). Per the '641 specification itself: sensors "directly attached to two or more respective wall portions which are included in a mounting member and orthogonal to one another," so that "it is possible to accurately measure two or more physical quantities orthogonal to one another without deviation in parallelism," and the large-heat-capacity mounting member holds the chips at constant temperature. That supplies L2–L4 and L8 (minus the adhesive-in-groove aspect).

US 2007/0102810 A1 (Mitutoyo Corp.) — the three-orthogonal-face reference. Verified (Google Patents). "Sensor chips 2 … are mounted on X, Y and Z faces of the base block 1, respectively"; claim 7 recites the base block as "a substantially regular hexahedron having at least one face bonded to the semiconductor substrate of said sensor chip." Its background expressly identifies the problem the '641 attacks: anodic bonding (no adhesive) was adopted in JP 2001-183389 precisely "to improve the accuracy that becomes worse due to a difference in thermal expansion coefficient between an adhesive and an element adhered thereto and to reduce thermal stresses exerted at the time of bonding," but at the cost of size and assembly effort. This is a direct teaching that the adhesive/CTE problem is known and that the trade-off is a design choice.

US 2006/0130584 A1 (Fujitsu Media Devices) — the sensor-internal structure reference. Verified in part (Patents-Review summary). It discloses a MEMS acceleration sensor chip with a support portion fixed to a substrate, a weight, beam portions connecting them, a frame portion, and a stopper substrate fixed by adhesive to a face. This supplies L11-type structure (fixed member / movable member / interposed joint-beam) in a glued sensor package.

US 8,256,287 B2 (AAI Corporation) — the "adhesive trap" reference. Verified (Google Patents; Justia). It discloses adhesive traps and states: "by … providing a series of concentric undercuts on the exterior of the snap-lock device, adhesive may be conveniently captured without readily being wiped off during assembly." That is the generic engineering principle of a recess that deliberately collects excess adhesive in a bonded joint — i.e., L10's "adhesive in the groove," stated as a known technique.

US 2014/0123754 A1 and JP 2014-119412 A (both Seiko Epson) — the applicant's own earlier multi-axis/electronic-apparatus work. These are in the '641's own lineage ("physical quantity detecting device, electronic apparatus, and moving object"; "multi-axis physical quantity detection apparatus"). I could not verify their substantive disclosures in this session (search budget exhausted); I therefore treat them as candidate secondary references only, not as grounds I can assert. Note the legal point: because both published in 2014 — more than one year before the 2017-02-13 priority date — they are available as § 102(a)(1) art notwithstanding common ownership with Seiko Epson; § 102(b)(1)(A)'s grace period and § 102(b)(2)(C)'s common-ownership exception do not reach publications older than one year.

US 2017/0248553 A1 (RMCI, "Sensor mounting pad with secondary restraint feature"). Effectively filed 2016-02-25, published 2017-08-31 — i.e., after the '641 priority date but before the '641 filing date. Because it was "effectively filed" before the '641's effective filing date, it is available under § 102(a)(2) (assuming the '641's priority claim holds). I could not verify its disclosure; it is listed for completeness only.

3. Combination I — the primary § 103 case

JP-A-2004-37105 (or Mitutoyo '810) in view of EP 0 644 586 A2, optionally further in view of AAI '287 or Fujitsu '584.

Element mapping:

3.1 Motivation to combine (KSR / MPEP 2143–2144 rationales)

1. Same field of endeavor and same problem. All of JP '105, Mitutoyo '810, EP '586, Fujitsu '584 and AAI '287 are micromachined inertial-sensor mounting and packaging art. EP '586 addresses the identical problem the '641 admits: mechanical/thermal stress transferred into the vibrating element through the adhesive bond because of CTE mismatch between adhesive and sensor. A POSITA seeking to improve the JP '105 multi-axis block would look directly to adhesive-bonding practice for micromachined inertial sensors.
2. Express teaching, suggestion or motivation in the references. EP '586 expressly teaches that grooves bounding the adhesive area prevent capillary adhesive creep, allow precise adhesive dosing, and avoid a "clamping effect" that stresses the bending oscillator. Mitutoyo's background expressly frames the adhesive/CTE trade-off. AAI '287 expressly teaches capturing excess adhesive in undercuts "without readily being wiped off during assembly."
3. Known technique / predictable result (MPEP 2144.03). Grooving a machined aluminum block to confine an adhesive bead is a mechanical step with a predictable result explicitly described in EP '586. The Federal Circuit treats application of a known technique to a known structure, yielding the known result, as obvious.
4. Design incentives and market pressure. Multi-axis quartz-crystal inertial sensors for camera-shake correction (the '641's own FIGS. 12–13) demand accuracy and miniaturization; both Mitutoyo and JP '105 stress accuracy and size/assembly cost. Constraining adhesive spread improves accuracy and permits smaller bond footprints — the same pressure experienced by the artisan.
5. "Obvious to try" / finite predictable solutions. Faced with adhesive squeeze-out under a sensor on a metal block, the artisan has a small, enumerated set of remedies: control adhesive volume, control surface energy, or provide a reservoir recess. EP '586 and AAI '287 supply the recess/reservoir solution.
6. Reasonable expectation of success and no teaching away. Nothing in EP '586 discourages adhesive bonding; it improves it. Placing the groove in the block (rather than in the sensor, as EP '586 suggests) is a mere location-of-feature variation that a POSITA would adopt precisely because the block is the easily machined, reusable part — a benefit Mitutoyo expressly extols (common base block "makes it possible to omit the glass substrate … to achieve a downsized sensor block and reduce the number of assembly steps").

3.2 Why the result is the claimed invention

Putting an EP '586-type adhesive-bounding groove onto each of the three orthogonal sensor-mounting faces of a JP '105/Mitutoyo-type block yields, without further invention: three sensors each straddling a groove; adhesive under each sensor and in the groove; grooves that are necessarily separate (they lie on mutually orthogonal, non-coplanar faces); and the CTE-mismatch condition that both EP '586 and Mitutoyo's background take as the premise. The remaining recited limitations are orientation and geometry selections addressed below.

4. Combination II and III (alternatives that avoid a single point of failure)

• Combination II — Mitutoyo '810 as primary: Mitutoyo teaches the hexahedron block and the X/Y/Z sensor mounting and therefore subsumes the flattest L2–L4/L8 disclosure; EP '586 supplies L5/L6/L10/L13/L14. This avoids relying on a Japanese-language reference's characterization as filtered through the '641's own background section (an attack vector if the JP '105 disclosure is narrower than the '641 describes).
• Combination III — add Fujitsu '584 for L11 and AAI '287 for the "adhesive in the groove" functional teaching, i.e., a four-reference case. Fujitsu supplies a glued inertial package with a support/weight/beam architecture, and AAI supplies the "recess captures excess adhesive" principle in an expressly adhesive-bonding context. This is the belt-and-suspenders construction if the tribunal is skeptical that EP '586's grooves "receive" (as opposed to merely bound) the adhesive.
• Seiko-Epson-specific art: US 2014/0123754 and JP 2014-119412 are the applicant's own earlier filings in the same product line; if their disclosures confirm multi-axis block mounting and/or shake correction, they strengthen the § 103 case, particularly for claims 5–8. I have not verified their contents and do not rely on them here.

5. The dependent and apparatus claims

Claim 2 (groove intersects at least part of a sensor edge). In EP '586 the adhesive is dosed onto a bounded area and the component is placed over it; the groove boundary necessarily underlies a sensor edge region. Independently, the '641 specification supplies the motivation ("increase the bonding area such that the sensor is not easily separated … and to improve impact resistance") — an obviousness rationale the applicant itself articulated.

Claim 3 (bonded region overlaps sensor center). Centering a part over a bounded adhesive land is the ordinary way to bond it; EP '586's described adhesive dosing (a strip or row of drops centered on the part) and the '641's own stated rationale ("bond the physical quantity measurement sensor in a stable attitude") make this a design choice.

Claim 4 and claim 10 (movable-member/vibrating-beam directions intersect the groove direction). These are pure relative-orientation limitations. The specification states the reason the artisan would choose the orientation — to keep adhesive out of the fixed portion and reduce thermal-stress transfer there. Once the geometry of the EP '586 groove is chosen, orienting the already-designed sensor on the face so that its fixed region is clear of the adhesive is a routine placement decision. This is nonetheless one of the weaker links (see § 7).

Claim 9 (vibrating beam measures the physical quantity). The '641's sensor element is a quartz double-ended-tuning-fork-type vibrating beam. Piezoelectric-vibrating-beam physical-quantity sensors are notoriously old and well documented in the same assignee family and in the "Similar Documents" set (e.g., US 9,748,467 B2, "Piezoelectric vibrator"). Combining a vibrating-beam sensor with the grooved block is not a separate inventive step; it is the routine selection of a known sensor element type. (I have not verified a specific reference's disclosure of a vibrating-beam element in this session — this is asserted from the field's general state of the art and the '641's own text, and should be re-checked.)

Claims 5–8 (electronic apparatus: case, detection, shake correction). Each is an apparatus claim directed to housing the claim-1 (or claim-2/3/4) device and using it to correct shake. Camera-shake correction using inertial sensors is decades old and not the invention; the '641's own FIGS. 12–13 (video camera 300 with "the function of correcting camera shake"; mobile phone 400 with "correcting camera shake of the camera mechanism 403") confirm that this is conventional application, not claimed novelty. The applicant's own US 2014/0123754 ("electronic apparatus") is a natural secondary reference. Claims 5–8 add nothing patentable over the device claims plus the admitted conventional shake-correction application.

6. Objective indicia (Graham factors) — what the record shows

The earlier sections of this analysis established that there is no litigation and no PTAB proceeding on this patent, and that title never left Seiko Epson. On the objective-indicia side:

• No evidence of nexus, copying, industry praise, or unexpected results is available in this record, and the '641's specification asserts no unexpected result — it describes the groove's effect in purely conventional terms (excess adhesive flows into the groove).
• Long-felt need cuts against the patent, not for it. The '641 admits the problem has existed since at least JP-A-2004-37105 (2004), while the solution — grooves confining adhesive to control CTE-induced stress — was published in EP 0 644 586 A2 in 1995. A 22-year-old known solution applied to a known multi-axis block is the classic fact pattern in which long-felt need is defeated by the prior art's express solution.
• No unexpected results: partially filling a channel with squeeze-out adhesive is exactly the result EP '586 reports.

7. Honest weaknesses in the invalidity theory (and the best rebuttals)

1. § 325(d) / "examiner already considered it" is the biggest risk. Every reference in my primary and secondary combinations appears on the '641's face: JP 2004-37105, EP 0 644 586 A2 (with an examiner-cited English translation of Lindner), Mitutoyo '810, Fujitsu '584, AAI '287, US 2014/0123754 and US 2017/0248553 all appear in the "Citations (8)" table. A petition built solely on these invites discretionary denial under 35 U.S.C. § 325(d) (cf. Becton, Dickinson & Co. v. Becton Dickinson & Co., IPR2017-01586). The practical consequence: the strongest case must introduce art not of record.
2. L7 — "across an entirety of each surface." EP '586's adhesive-bounding grooves are described as local boundaries, not face-spanning channels. The full-length character is the most conspicuously untaught limitation. The counter-argument is that a groove that stops short of the face edge would let adhesive wick around its end to precisely the corner region the '641 identifies as the failure locus — so extending a known adhesive-damming groove edge-to-edge is an obvious modification with a predictable benefit. This argument is good but not airtight; expect it to be contested.
3. L12 — "joint members extend along the extending directions of the grooves." This is the most awkward limitation in the claim: it is an internal-sensor-orientation constraint not addressed by any of the cited references. The '641's own specification is internally in tension here (claim 4 requires the movable-member direction to intersect the groove direction, while claim 1 requires the joint member to run along the groove direction), and no tribunal has construed either term. A petitioner must supply a reference or an explicit design-choice rationale for L12; this is the limitation most likely to require a new reference.
4. L13 — "adhesive is only applied on a surface between the pair of grooves." This reads as a method-of-assembly limitation grafted into an apparatus claim, and it describes the interior of a sealed, packaged sensor assembly — it will be difficult to prove in an accused product and, correspondingly, difficult for a patent owner to enforce. That is a non-infringement/proof problem, not an invalidity solution, but it matters to how the claim is likely to be read. EP '586's strip-shaped/caterpillar adhesive surface supplies the closest teaching.
5. Prosecution history is not reviewed here. The record shows a non-final rejection (2019-04-15), a final rejection (2019-09-10), a reopened non-final action (2019-12-16), and grant (2020-07-21). Claim 1 almost certainly was amended to overcome art; because I have not retrieved the office actions and applicant remarks, I do not know what the applicant argued, and I therefore do not assert that any specific limitation was the reason for allowance. Retrieving the file wrapper is a prerequisite before filing anything.
6. Reference-coverage verification. I verified the disclosures of EP '586, Mitutoyo '810, AAI '287 and Fujitsu '584 in this session, but not those of JP 2014-119412, US 2014/0123754, US 2017/0248553, or JP 2004-37105 beyond the '641's own characterization of it. Statements above about those four rest on the '641's text or on titles/abstracts only and must be confirmed against full text before use.

8. Bottom line

On the face of the cited art, claim 1 is a strong § 103 candidate: the multi-axis orthogonal-sensor-on-a-block mounting is admitted background (JP-A-2004-37105) and is expressly taught with X/Y/Z faces by Mitutoyo '810; the entire inventive core — a groove that limits adhesive spread and thereby prevents CTE-induced stress from reaching the sensor's active region — is expressly taught, for the same problem, by EP 0 644 586 A2; the "recess captures excess adhesive" principle is taught by AAI '287; and the sensor's movable/fixed/beam architecture is taught by Fujitsu '584. The combination is supported by an express suggestion in the art, a common field of endeavor, a known technique with a predictable result, and the applicant's own admission of the underlying problem. Claims 2, 3, 9 and 10 fall to routine design choices and orientation selections on the same record; claims 5–8 add only the admittedly conventional camera/mobile-phone shake-correction application.

The realistic obstacles are L7 (full-length grooves), L12 (joint member parallel to groove), and the fact that nearly every reference was already before the Examiner — the latter making § 325(d) the primary vulnerability of any petition.

Highest-value next steps: (a) pull the file wrapper and read the 2019 final rejection and the applicant's remarks — that tells you which limitation the applicant relied on and where the § 103 attack must be aimed; (b) run a fresh-art search outside the of-record set for die-attach/epoxy-overflow trenches and for full-length paired adhesive dams across a sensor-mounting face (semiconductor die-attach and MEMS mounting literature is the most promising), since new art both strengthens § 103 and sidesteps § 325(d); (c) verify full text of JP 2004-37105, JP 2014-119412, US 2014/0123754 and US 2017/0248553; and (d) obtain a formal claim construction position on "extend[s] along extending directions of the … grooves" and "across an entirety of," both of which are untested and potentially indefinite.