by Dennis Crouch
Sigray, Inc. v. Carl Zeiss X-Ray Microscopy, Inc., — F.4th — (Fed. Cir. 2025)
The Federal Circuit established an important precedent regarding inherent disclosure and implicit claim construction in this IPR appeal, holding that the Board’s purported interpretation of the prior art was rather an implicit claim construction that neither party requested. The decision, went rely heavily on the Federal Circuit inherency decision in SmithKline Beecham Corp. v. Apotex Corp., 403 F.3d 1331 (Fed. Cir. 2005), allowing for inherency finding even for miniscule or undetectable features found in the prior art. As I dug into the case though, I was surprised to find that the PTAB did not cite or reference SmithKline, nor did any of the briefs filed in the appeal. As Judge Dyk wryly stated at oral arguments: “but of course, nobody showed them the SmithKline case.”
This case involved x-ray optics, with claims requiring “magnification of the projection x-ray stage is between 1 and 10 times.” Although neither party ever asked for claim construction, the appellate panel went ahead to determine that this claim scope – based upon its “plain meaning” – “includes tiny, even undetectable, magnification” rejecting the PTAB’s implicit construction that seemed to require a meaningful divergence from parallel.
Background: The dispute centers on U.S. Patent No. 7,400,704, which claims X-ray imaging systems incorporating projection magnification. The key technical concept involves geometric magnification, where diverging X-ray beams spread out as they travel from source to detector. The mathematical relationship is expressed as M = (L(s) + L(d))/L(s), where M is magnification, L(s) is the source-to-sample distance, and L(d) is the sample-to-detector distance. Under basic geometric optics, magnification equals exactly 1 only when L(d) = 0 (when the sample is touching detector), and any diverging beam with L(d) > 0 necessarily produces magnification greater than 1.
Sigray petitioned for inter partes review of all claims, arguing that Jorgensen—a 1998 scientific paper describing collimated X-ray microtomography of rodent organs—anticipated the claims. The parties agreed that Jorgensen explicitly disclosed almost all claim limitations. The one limitation in dispute is “magnification of the projection x ray stage is between 1 and 10 times,” which the PTAB concluded was not found in the reference.
Notably, a key goal of Jorgensen was to have parallel rather than diverging beams (i.e., collimated) — and the way to do that was to substantially eliminate the magnification (i.e., make the magnification M = 1). And, everyone seems to agree that if M=1, it is not “between 1 and 10.” Of course, parallel beams are a theoretical ideal but are never actually possible. Jorgensen indicated its result was “reasonably pure,” sufficient enough so that they could use math that applied parallel approximation. Still, the patent challenger Sigray argued for inherent disclosure, contending that any remaining divergence would necessarily produce magnification within the claimed range. Zeiss countered that the collimation rendered the beam sufficiently parallel to eliminate meaningful projection magnification.
I mentioned above that any actual system will include error from the M=1 goal – a statistical certainty. Although it required some implicit theoretical assumptions, the court concluded that the error was always one sided where M>1 because of the optics geometry, the . This was true because of the optics geometry that would require a negative distance from sample to detector – something that is physically impossible with the setup.
Despite all of this, the Board sided with the patentee Zeiss, finding that Sigray failed to show that “the collimated X-ray beam in Jorgensen diverges enough to result in projection magnification ‘between 1 and 10 times,’ as claimed.” The result – no anticipation.
On appeal: The Federal Circuit’s analysis began by addressing whether the Board had engaged in claim construction, despite the Board’s statements to the contrary. The court found that the Board’s use of “diverges enough” in describing required divergence “reflects that it considered a certain level of divergence as outside the claim” and that “[n]arrowing the claim scope in this way is in fact claim construction.”
To reach its conclusion, the court relied upon Google LLC v. EcoFactor, Inc., 92 F.4th 1049 (Fed. Cir. 2024) for the proposition that the court should look to substantive outcomes to see whether the claims had been construed – even if procedural labels were missing. Google involved a dispute over whether the five enumerated inputs must be entirely distinct from one another, or whether one input could be calculated based on another input. Like in Sigray, the PTAB in Google expressly stated it was “not constru[ing] any of the claim limitations,” but then went on to conclude that the inputs were “distinct components” that must be “used distinctly from the other listed inputs.” The Federal Circuit held that this analysis constituted implicit claim construction creating boundaries around the patented subject matter regardless of the Board’s disclaimers about its process–emphasizing that the determinative factor in identifying claim construction is whether the tribunal’s analysis establishes the scope and meaning of claim terms, not whether it explicitly acknowledges doing so.
Ultimately, in Sigray, the court concluded that the claim language of “between 1 and 10” should be strictly interpreted according to the formula 1<M<10. For beams close to parallel (M≈1), any positive divergence from absolute 1 qualifies as within the claim, even if not “meaningfully greater” such that it would be measurable, or require different mathematics to model.
Inherent disclosure: Having adjusted the claim construction, the Federal Circuit then found that Jorgensen did, in fact, inherently disclose magnification M>1 because of the physical geometry discussed above.
In its analysis, the court drew heavily on SmithKline Beecham Corp. v. Apotex Corp., 403 F.3d 1331 (Fed. Cir. 2005). In SmithKline, the Federal Circuit faced a pharmaceutical patent claiming PHC hemihydrate, a crystal form containing bound water molecules that was serendipitously discovered in 1985 while attempting to manufacture PHC anhydrate according to a prior art patent from the 1970s. The court held that this earlier patent inherently anticipated the later hemihydrate patent, even though the prior art made no mention of the hemihydrate form and the compound wasn’t even known to exist when the prior art was created. The court explained that inherent anticipation requires only that “the natural result flowing from the operation as taught would result in” the claimed invention, rejecting the district court’s more demanding standard that would have required proof that it was “impossible” to practice the prior art without producing the claimed compound. The court found that manufacturing PHC anhydrate according to the prior art patent “necessarily and naturally resulted in derivative production of at least trace amounts of PHC hemihydrate,” and emphasized that inherent disclosure does not require the prior art’s authors to recognize or understand the presence of the inherently disclosed feature. This case thus established the principle that even unknown, undetectable trace amounts of a claimed compound can constitute inherent anticipation if they flow naturally from practicing the prior art as taught.
But see, In SmithKline, the Federal Circuit distinguished In re Seaborg, 328 F.2d 996 (C.C.P.A. 1964), by emphasizing the difference between mere possibility and actual evidence of inherent production. In Seaborg, the court found no inherent anticipation of americium isotope because there was “no positive evidence” that the claimed isotope was inherently produced by the prior art nuclear reaction process conducted by Fermi—only a “possibility” that the element “may have been produced,” which the court held was insufficient for anticipation. Seaborg involved speculation about what might happen, while SmithKline involved undisputed facts about what actually does happen when the prior art process is practiced.
Applying SmithKline‘s inherency standard to the X-ray optics context, the Federal Circuit in Sigray found that Jorgensen’s collimated beam system inherently disclosed the claimed magnification range, even though Jorgensen’s explicit goal was to minimize magnification through collimation. The court emphasized that perfect collimation—producing truly parallel beams with magnification of exactly 1—is physically impossible as a practical matter, and that any remaining divergence, however minute, necessarily produces magnification greater than 1.
Just as in SmithKline, where trace amounts of hemihydrate inevitably resulted from the prior art process regardless of intent, the Sigray court found that some degree of magnification necessarily flowed from Jorgensen’s described system, making the “only evidence-supported conclusion…that Jorgensen contains a beam diverging at the sample, which necessarily produces projection magnification over 1.”
Theoretical Physics Meets Patent Reality: I’ll go on a bit of a tangent here about underlying assumptions about the nature of physics reality.
The court’s analysis assumes classical geometric optics throughout, treating X-ray propagation as simple ray tracing governed by Euclidean geometry. However, at the microscopic scales relevant to X-ray wavelengths (typically 0.01 to 10 nanometers), quantum mechanical effects and near-field phenomena can become significant. When distances approach wavelength scales, classical geometric relationships break down, and phenomena like quantum interference, diffraction, and near-field effects can substantially alter magnification relationships. The court’s mathematical certainty that “any diverging beam with L(d) > 0 necessarily produces magnification greater than 1” ignores these quantum-scale complexities that could theoretically permit exceptions to classical geometric predictions. 
Equally problematic is the court’s treatment of Jorgensen’s disclosure as describing a theoretical, perfectly constructed system rather than an actual, physical machine. But Jorgensen discloses and describes the result of the actual X-ray systems built and operated by the authors. These real systems suffer from construction tolerances, thermal expansion, mechanical vibration, component wear, and measurement uncertainties. Although Jorgensen admits to using a cone-beam geometry algorithm for image reconstruction, the majority of the document indicates the potential for otherwise: the “collimated X-ray beam”, specimen size designed for no magnification, and teaching against magnification in the document. And, the reconstruction algorithm works for both diverging and parallel beams. Looking at this together makes me think the result is not sufficient to reach the “necessary” requirement of true inherency.
The De Novo Review Preference: Perhaps most concerning is the Federal Circuit’s apparent preference for resolving complex technical disputes through claim construction—a legal question reviewed de novo—rather than through detailed factual analysis of prior art technology subject to substantial evidence review. By recharacterizing what could have been a factual dispute about the technical operation of Jorgensen’s system into a claim construction question, the court avoided the deference typically accorded to the PTAB’s technical expertise. This approach allows the Federal Circuit to substitute its own technical judgment.
The opinion was authored by E.D.Pa. Chief District Judge Mitchell Goldberg, sitting by designation, and joined by Circuit Judges Dyk and Prost. Alan Grayson Laquer argued for Sigray, with Craig Summers and James Youngblood also on the briefs, all from Knobbe Martens. Kurt Glitzenstein argued for Zeiss, with Ryan O’Connor, John Phillips, and Charles Reese also on the briefs, all from Fish & Richardson.