• Making much smaller electron beam ic circuit features in parallel could

    From Treon Verdery@21:1/5 to All on Sun Aug 14 01:59:11 2022
    I think photolithography techniques can be used to etch a plurality of physical beamguides 80 nm long and 24 nm wide such that one big powerful electron beam is split into 16 others, an ebeam magnetic location shifter then angles the beam to a space
    above the waveguide and travels back to the waveguide sequence beam introduction port, and, if the e-beams are pulses, makes 256 e- beams, these pulses are are recirculated this way 8 times to make a pulsed ebeam pattern with 2^32, or about 5 billion
    separate (or purposefully overlapping e-beams. The velocity of some electrons is about 90% of the speed of light, but they can be slowed to 1 meter per second. That gives very wide duration of effect latitude on a billion minibeams of lasers, produced
    from a grating, on recombining electron beams, or making patterns of electrons in the waveguides that block them. That means its possible to make something that looks like a 2D QR barcode, or even a 3D QR-code like energy arrangement at a cube of
    traveling electron beams. Now when making bulk transistor layouts in parallel rather than make a 2D or 3D QR code like beam, a semiconductor (Si, Ge, indium phosphide, graphene) transistor layer etching 5 billion 2^32 plural electron beams simultaneously
    could be used. That writes/etches entire CPU, GPU, AI processing units 5 billion times faster than a solitary raster scanned e-beam. If 2^40th beamsplit and recirculated electron beam patterns are possible then an wafer can be immediately written to
    make a trillion transistor single processing unit structure such as exist during 2021, or 100k 10 billion transistor phone CPUs, with a 2021 fiscal value of about $3 each, or each etched wafer being worth about 1/3 of a million dollars and be etchable
    and CVD layerable 2-3 times per 24 hours for production in 8 or fewer days. The velocity of producing the parallel electron beam written chips can go up 4-8 times if the E-beam is mathematically treated as a depthy self-focused wave effect 4-8 already
    applied at the computer chip surface layers of stacked semiconductors, much like etching out the interior of a 4-8 layer cake with electron beams that have different purposed overlap. This deep 3D etching may be especially well suited to electron beams (
    or muon beams). That causes the entire wafer with 100K $3 chips to be rapidly made as a 4 or 8 CVD stack, in 2-4 hours at 1/2 an hour per layer, and then a massively multiparallel electron beamsplitted transistor array template etched in less than 8-24
    minutes. Then there would be a fresh oxide layer and another 4-8 layer e-beam etchable stack put on top of the first stack to make 8-16 layer computer chips, at a velocity of 4-6 wafers made per 24 hours, so as much as the e-beam massively plural
    beamsplitter along with the chemical vapor deposition machine together, at $1 million per machine earns $1.2-2 million $ for every 24 hours of continuous usage. This makes an ic fab that earns 400-700 million $ each 365 days with just 2 maker machines,
    and a chip sawing, microwire attaching and packaging machine. That $3-4 million $ fab and building and materials is 249-300 times cheaper than freshly built 2020 IC Fabs. Again, the speedup factor that makes it so chips are produced 72 times faster is
    beamsplitter originated 2D and 3D full wafer 5 billion to one trillion plural electron beam depthy etching of what could be called uncustomized FPGA-like repeat transistor arrays. It is possible to improve this further by periodically halving the feature
    size, and doubling the density of the plural electron beam etched features. Electrons, perhaps attometers or even zeptometers big are physically very much smaller than the 7-20 picometer diameters of the atoms they are a part of. The internet says the
    smallest feature size visible with an electron microscope, a kind of electron beam machine is 1/10th of a nanometer, 20-30 times smaller than actual 2-3 nm feature size 2020 produced phone chips, and 10 times smaller than IBMs 1 nanometer feature chips.
    Imaginably a way to make a super high energy electron beam of 10-20 picometer diameter (1 carbon atom diameter) is to use observed as particle ballistic electrons traveling through a monoatomic or duoatomic conductive polymer filament, such as PEDOT, or
    the Se element version of pedot. Placed in either a circular winding with a driving electromagnet, or a 1-11 cm long mass driver/railgun winding around a vastly multiparallel string-cheese columnar pedot rod, or at wafer size, cylindrical disk, electrons
    are accelerated to the distal tip of the monomolecular or duomolecular pedot, it is possible ballistic electrons spread out less, heightening beam resolution, and it is possible making the distal most surface atoms of the pedot rod be made of highly
    electronegative atoms like phosphorus, bromine, or electron saturated at time of manufacture making electret pedot are all approaches to making a 7-29 picometer wide electron beam. That beam can of course be electromagnetically lensed and accelerated.
    Electrons and electron beams of sufficient energy fling atoms around when they etch. And, encouragingly at a completely different system where laser microbeams from a grating etch NiFe invar alloy, single atom thicknesses can be ejected during the
    drilling process. 2020 Electron microscopes are able to image, perhaps even change their depth of focus to look at metal coated features 1000 nm deep. That suggests the ability to use massively multiparallel string cheese pedot or Se pedot rods or disks
    to make 14-30 nm wide grooves and features, at a drillability depth of 100 nm, 300-600 times deeper than they are wide. That supports making 2D and depthy 3D QR-code like shapes to make vertically stacked transistors at the one size fits all, fast and
    cheap to chemically vapor deposit, layers of FPGA-ish semiconductor material. A massively multiparallel electron beam lithography depthy transistor writes does more than produce ics faster than multiple single step photolithography, it shrinks feature
    size on chips, between 29 picometers and IBM's 1 nanometer process there is 2^6 (64) chip density increase from 6 doublings, typically occuring during 9 years of engineering and production. Some technologies could velocitize those computer improvements
    even more: research and construction of negative refractive index electron beam superlensing and refractive index matching, is a velocitizing electron beam technology, in a way where i don't know how it works, but it increases the hardness of metals, 12-
    24 hour pre etch cryogenic treatment of wafers could cause finer feature etching and higher resolution, making the EM steering, compressing, collimating part of the layers at chip base or middle during massively parallel electron beam etching, so
    printing electromagnetic windings and cobalt/Fe or ferrite pole pieces at/into the base layer makes all kinds of etch resolution heightening bumps, fields, annuluses and ££@magnetic 3D structured fields that look and perhaps act like waveguides but are
    not.

    Wikipedia ghost imaging of electron beams could heighten beam resolution to smaller than the published 100 picometersachinentroduction

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)