Sunday, July 01, 2012

New case law from CAFC for memresistor nanowire crossbars

In re Mouttet - CAFC Case No. 2011-1451

This is an interesting new piece of case law from the CAFC based on memory resistor crossbars used in arithmetic processing. The case is based on an obviousness rejection for an electrical memory resistor crossbar designed to perform arithmetic operations. The two primary references were Falk US 5249144 which teaches an optical crossbar configured to perform arithmetic processing and a technical article by Das et al. which teaches an electrical nanowire crossbar formed with resistance switching molecules. The CAFC supported the obviousness of changing an optical arithmetic crossbar into a resistance switching nanowire crossbar based on Falk and Das. A review of the CAFC decision from a legal perspective is given by the blog However, a problem with this case is that from a technical perspective it is difficult to uphold the obviousness claim.

The way optical crossbars (such as Falk’s) work is that you need some sort of optical detector (or detectors) to detect the light intensity at each crossbar intersection region in parallel. For example, in the case of a 2×2 optical crossbar you would need to detect 4 (=2×2) optical states in parallel. In column 3, lines 44-51 Falk describes that each intersection region can have one of three possible values according to the detected light intensity of intersecting light sources (2=both light sources on, 1 = one light source and the other off, 0 = both light sources off). Based on the light source inputs the array of optical intensities at the crossbar intersections are simultaneously detected (i.e. in parallel) and compared to a look-up table to determine the arithmetic result. For example, optically adding binary 10 and 10 (1=light on, 0=light off) would produce a detectable light intensity pattern of

2, 1,
1, 0

in the crossbar intersections which would be converted to 4=100(binary) when compared to a look-up table.

So the problem is how exactly do you do what Falk does electrically? Applied voltages do not work the same way light does. If you connect a row and column wire of an electrical crossbar to the same voltage you get zero current flow instead of the 2 units of light intensity as explained by Falk. There is no teaching in Das of an electrical equivalent for the light detection at each intersection region of the electrical crossbar so how is this achieved? Usually electrical crossbars only have sensing amplifiers at the outputs of each column wire rather than a sensor capable of reading each individual intersection and I am not sure how you would accomplish this without causing multiple current sneak paths which would ruin the workability of the design. Das does not offer a solution to these problems.

This case may help identify the differences between legal obviousness used in patent rejections and technical obviousness used by engineers. There are several examples in telecommunication and computing where it is useful to create optical devices which serve analogous function to electrical devices (e.g. optical transistors, optical logic gates, etc.) This case could potentially be used as legal precedent to support the "obviousness" of converting between electrical and optical components without regard to the engineering reality of the difficulty of such conversion. It will be interesting to see how and when this decision is used by the PTO and the courts.