Patent Coverage of Impossible Embodiments: A Cautionary Tale for Patent Owners, and Lingering Questions

July 19, 2019

Scott I. Forman, Charles T. Steenburg

(as published by the New York Law Journal)

Patent owners asserting their patents in litigation will often seek broad claim constructions to cover accused products and forestall defendants’ non-infringement defenses. However, a recent decision illustrates potential pitfalls of this strategy even aside from traditional concerns about reading claims so broadly as to cover prior art.

In Trustees of Boston University v. Everlight Electronics Co., 896 F.3d 1357 (Fed. Cir. 2018), the U.S. Court of Appeals for the Federal Circuit invalidated a patent claim to a semiconductor device based on a claim construction that the patent owner proposed. The construction covered six permutations for the composition of the device, one of which was indisputably impossible to make using the method taught in the patent. A patent must be enabling, such that it teaches one of skill in the art to make and use the full scope of the claimed invention without undue experimentation. The court found the claim was not enabled because it encompassed the unattainable permutation. This case demonstrates that patent owners must be careful not to propose a claim construction so broad as to cover embodiments not enabled by the patent disclosure. However, the exact contours of this doctrine remain uncertain.

Plaintiff Trustees of Boston University (BU) owned a patent covering LED semiconductor devices. The devices incorporated gallium nitride (GaN), a composition that emits blue light and was notoriously difficult to fabricate as part of an LED. LEDs typically include a substrate and several layers. The solid-state materials comprising the layers may be monocrystalline (arranging all material in a single crystal structure), polycrystalline (combining multiple smaller crystal structures), or amorphous (non-crystalline).

One common method for fabricating semiconductor layers, called epitaxy, involves depositing semiconductor material onto a substrate. The deposited layer attempts to mimic the crystal lattice structure of the substrate as it grows. However, defects can arise if the lattice structures of the deposited layer and the substrate do not match. BU recognized that it was difficult to grow monocrystalline GaN on a substrate using epitaxy given a mismatch in the structure of the GaN and the substrate. BU solved the problem by first depositing a “buffer layer” of an amorphous GaN film on the substrate, then raising the temperature to crystallize the amorphous layer, and finally growing monocrystalline GaN on the crystallized layer.

BU received a patent on a device fabricated using this process. The relevant claim recited a semiconductor device comprising a substrate, “a non-single crystalline buffer layer” of GaN grown on the substrate, and “a growth layer grown on the buffer layer,” the growth layer comprising GaN. Accusing various defendants of infringement, BU argued that “non-single crystalline buffer layer” meant the layer was “not monocrystalline.” The district court interpreted the claim in line with this proposal, finding that the buffer layer could be polycrystalline, amorphous, or a mixture of polycrystalline and amorphous. The court found that the claimed semiconductor device comprised six permutations: a monocrystalline growth layer formed (1) indirectly (i.e., with an intervening layer) on a polycrystalline layer; (2) indirectly on a layer that is a mixture of polycrystalline and amorphous; (3) indirectly on an amorphous layer; (4) directly on a polycrystalline layer; (5) directly on a layer that is a mixture of polycrystalline and amorphous; or (6) directly on an amorphous layer.

A jury found the patent valid and infringed. In challenging the verdict, the defendants argued that the asserted claim was invalid for lack of enablement, because the patent did not teach one of the six permutations—that is, fabricating an LED with a monocrystalline growth layer formed directly on an amorphous layer. However, the district court denied the renewed JMOL motion and concluded that it was unnecessary for the patent to enable a device embodying this permutation.

The Federal Circuit reversed and stressed that it was indisputably impossible to fabricate this permutation using epitaxy, the only method taught in the patent. As a patent must enable the full scope of the claimed invention, failure to enable one of the permutations covered by the claim rendered the claim invalid. The court noted that “to some extent, BU created its own enablement problem” with its proposed claim construction. Having obtained the construction it sought, which included a physically impossible embodiment, “BU then needed to successfully defend against an enablement challenge as to the claim’s full scope.” The decision reminds patent owners that they must carefully consider enablement issues when crafting claim constructions to target accused infringers.

One notable aspect of this decision is the potential tension with an earlier line of cases holding that patent claims do not necessarily lack enablement solely because they embrace inoperative embodiments, provided artisans in the field would know how to determine, without undue experimentation, which embodiments would work. See Atlas Powder Co. v. E.I. du Pont De Nemours & Co., 750 F.2d 1569, 1576-77 (Fed. Cir. 1984). Given the parties’ agreement that it would be impossible to grow a monocrystalline film directly on an amorphous layer using epitaxy, one perhaps could have reasonably argued that artisans would have known that this embodiment was inoperative and could thus practice the invention without undue experimentation.
Atlas involved a patent claiming an emulsion useful as a blasting agent for mining and construction. The accused infringer argued that the patent was not enabled because it listed numerous salts, fuels, and emulsifiers that could form thousands of emulsions without teaching which combinations would work. The Federal Circuit disagreed and concluded that artisans would have known know how to select a salt and a fuel, determine a proper emulsifier, and ultimately achieve an operable combination.

One could potentially square BU with Atlas by reading an “inoperative embodiment” to be one that can be physically made, but is not useful for its intended purpose, as opposed to a permutation or embodiment that is physically impossible to make. In In re Cook, 439 F.2d 730 (C.C.P.A. 1971), the Federal Circuit’s predecessor held that claims covering lenses with specified parameters were not invalid for lack of enablement despite encompassing “inoperative embodiments” of lenses with parameters within the recited ranges, but that were not useful as zoom lenses as intended because “it would be obvious to those skilled in lens design whether a given embodiment within the indicated ranges, once conceived, would or would not be useful as a zoom lens.”

Of course, there have been other cases since Atlas invalidating patents for failure to enable the full scope of claims. E.g., Liebel-Flarsheim Co. v. Medrad, 481 F.3d 1371 (Fed. Cir. 2007) (claims covering injectors with and without a pressure jacket not enabled where patent taught only injectors with a pressure jacket). Some, like Liebel-Flarsheim, even involved situations in which the patentee intentionally crafted broad claims that proved to be the patent’s downfall. However, these cases generally involved claims covering embodiments that were not necessarily impossible to achieve (even though the specifications failed to enable them). BU seems to break new ground in that five out of six permutations were indisputably possible to make, and the last was indisputably impossible.

One question that arises is whether an otherwise non-enabled claim could be enabled if one of ordinary skill would understand the claim to exclude impossible embodiments. Some cases suggest this is a viable approach to avoid inoperative embodiments. E.g., In re of Geerdes, 491 F.2d 1260, 1265 (C.C.P.A. 1974) (“Having stated the objective (foamed product) together with the process steps, use of materials which might prevent achievement of the objective (by rendering the process inoperative) can hardly be said to be within the scope of the claims.”). Generally, validity does not factor into claim construction unless the claim is otherwise ambiguous. See Phillips v. AWH, 415 F.3d 1303, 1327 (Fed. Cir. 2005) (en banc). The question, then, is whether a court would have accepted an argument by BU that one of ordinary skill would understand the claim to exclude the permutation of a monocrystalline layer formed directly on an amorphous layer because it was known to be impossible—not because this construction is necessary to preserve validity, but because it is how one of ordinary skill would understand the claim.

Another question is whether a broad claim embracing a vast number of possible embodiments could be enabled while a narrower claim covering a finite set of permutations could lack enablement. For example, consider a first claim to “a pharmaceutical formulation comprising micronized ibuprofen” and a second claim to “a pharmaceutical formulation comprising micronized ibuprofen with a particle size of 2-10 micrometers.” Suppose it is physically impossible to make micronized ibuprofen with particles of less than 5 micrometers. Is the first claim enabled because it does not specify the size of the particles, while the second is not enabled? It seems counterintuitive that a broader claim encompassing a potentially limitless number of embodiments could be enabled while a narrower claim with a finite number of embodiments could lack enablement.

To date, no case has cited BU for its implications as to enablement of claims covering impossible embodiments. These questions remain open for resolution another day.