MagSil v. Hitachi, full-scope enablement, and a memory claim that reached for infinity.

A patent claim is a boundary drawn around a piece of the future. The enablement requirement is the rule that the boundary can only extend as far as you have actually taught someone to go. Most of the time that limit is invisible, because the claim and the disclosure roughly match. MagSil Corp. v. Hitachi (Fed. Cir. 2012) is the case where they came apart spectacularly — and they came apart over a number, a percentage change in electrical resistance, which is why it is the natural companion to the Teva problem I wrote about earlier. Teva asked who decides what a measurement means. MagSil asks how far you can claim past a measurement you cannot yet reach. Both put a physical quantity at the center of a patent’s life or death, and this one sits squarely in my own corner of device physics.

The device

The patent, owned by MIT and exclusively licensed to MagSil, claimed a magnetic tunnel junction — the read element at the heart of a hard-drive head and, later, of magnetic memory. The structure is three layers: two ferromagnetic electrodes separated by an insulating barrier thin enough that electrons tunnel through it quantum-mechanically. When the magnetizations of the two electrodes point the same way, the junction sits in a low-resistance state; when they point opposite ways, resistance is high. In magnetic memory, that difference can encode the bit; in a hard-drive read head, it converts the disk’s magnetic pattern into a readable electrical signal. The figure of merit is the size of the swing between those two states, the tunneling magnetoresistance, expressed as a percentage change in resistance. A larger swing is a more readable signal. Anyone who has worked on resistance-based memory will recognize the shape of the problem at once: the whole device is a controlled difference between two resistance states, and the engineering is a fight to make that difference large, stable, and repeatable.

The claim with no ceiling

The limitation that decided the case asked for “a change in the resistance by at least 10% at room temperature.” Read it the way an engineer reads a spec: a lower bound, and no upper bound. The claim covered a 10% swing, a 100% swing, a 600% swing — everything from ten percent up to infinity. The trouble was what the patent had actually disclosed. At the 1995 filing, the specification enabled a resistive change of about 11.8%. It pointed to an idealized value around 24%, but had not taught how to reach it at room temperature; during prosecution, the inventors believed 100% was the ceiling of the possible. By the time the case was litigated, the field had reached resistive changes well above 600%. The claim, in other words, reached for a performance regime that no one — the inventors included — knew how to build when the application was filed.

What the court held

The Federal Circuit, in an opinion by Chief Judge Rader, affirmed the claims invalid for lack of enablement. The principle is “full scope” enablement: the specification must teach a skilled person to make and use the entire claimed range without undue experimentation, and it must do so as of the filing date. A disclosure that enables only a small subset of an open-ended range does not buy the whole range. And the gap cannot be closed after the fact. That the field eventually achieved 600% changes was beside the point — those advances came from work done years later, on electrode metals, barrier materials, and processes the 1995 specification never described. You do not get to claim the territory because you were first to point at it; you get the territory you taught others to occupy. The open-ended drafting that created the exposure earned no leniency, either: choosing expansive language only enlarges what you are obligated to enable.

The engineer’s reading

What makes the case clarifying, from inside the device world, is that the legal defect and the physical one are the same defect. “A change in resistance of at least 10%, with no ceiling” is not really a description of a device. It is a description of a direction — make the swing as large as you can — and the materials physics that gets you to 12% is not the physics that gets you to 600%. The history bears this out almost too neatly. The low-double-digit swings of the 1990s came from amorphous aluminum-oxide tunnel barriers; the leap to several hundred percent arrived only when the field moved to crystalline magnesium-oxide barriers about a decade later — a different materials system, different interfaces, different processing, found by other people. The 1995 patent claimed performance that depended on materials systems and process control the specification did not yet disclose. Enablement is the doctrine that notices this, and what it enforces is close to a reproducibility standard: a result you cannot teach anyone to reproduce is not yet a result you own. The reward tracks disclosure, not prophecy.

There is an irony here that an engineer feels more sharply than a lawyer might. The inventors were, in a sense, right — large resistance changes turned out to be possible, far larger than they imagined. But being right about the destination is not the same as having built the road, and the patent system, sensibly, declines to pay for the former. An open-ended performance claim is a bet that the future will cooperate. Full-scope enablement is the house rule that you can only collect on the part of the future you actually delivered.

Scope as a conserved quantity

Set beside Teva, the pair maps the two ways a measurement can become the hinge of a patent. Teva was about authority — when the meaning of a technical term turns on evidence, which court’s reading governs. MagSil is about reach — when a claim is written around a performance number, how much of the number-space the inventor is allowed to own. The first is a question of who decides; the second, of how far a right may extend past the thing that justifies it. The second is the more recognizably engineering of the two, because it amounts to a conservation law: you cannot take out, in the scope of your monopoly, more than you put in, in the content of your teaching. The claim is permitted to be as broad as the disclosure that earns it, and not broader in a way that requires undue experimentation.

That is the patent bargain stated precisely. Disclosure is the consideration; scope is the payment; enablement is the clause that keeps the two equal. MagSil is a clean instance because the imbalance is quantitative and visible — about 11.8% taught, infinity claimed — and because the quantity in dispute is the same one my own field spends its time trying to control. It is a useful reminder that the boundaries patents draw are not free-floating legal artifacts. In the end they are bounded by what someone could actually make work in a lab.

References

  • MagSil Corp. v. Hitachi Global Storage Technologies, Inc., 687 F.3d 1377 (Fed. Cir. 2012).
  • U.S. Patent No. 5,629,922 (assignee: Massachusetts Institute of Technology; exclusive licensee: MagSil Corporation).
  • 35 U.S.C. § 112 (enablement requirement).