Device with air-gaps to reduce coupling capacitance and process for forming such

    公开(公告)号:US11574910B2

    公开(公告)日:2023-02-07

    申请号:US16457677

    申请日:2019-06-28

    Abstract: A device is disclosed. The device includes a plurality of capacitors, a transistor connected to each of the plurality of capacitors, and a first dielectric layer and a second dielectric layer on respective adjacent sides of adjacent capacitors of the plurality of capacitors. The first dielectric layer and the second dielectric layer include a top portion and a bottom portion, the top portion of the first dielectric layer and the top portion of the second dielectric layer extend from respective directions and meet at a top portion of a space between the adjacent capacitors, the bottom portion of the first dielectric layer and the bottom portion of the second dielectric layer extend from respective directions and meet at a bottom portion of a space between the adjacent capacitors. The device also includes one or more air-gaps surrounded by the first dielectric layer and the second dielectric layer on respective adjacent sides of the adjacent capacitors, the top portion of the first dielectric layer and the second dielectric layer between the adjacent capacitors, and the bottom portion of the first dielectric layer and the second dielectric layer between the adjacent capacitors.

    In-memory analog neural cache
    7.
    发明授权

    公开(公告)号:US11502696B2

    公开(公告)日:2022-11-15

    申请号:US16160800

    申请日:2018-10-15

    Abstract: Embodiments are directed to systems and methods of implementing an analog neural network using a pipelined SRAM architecture (“PISA”) circuitry disposed in on-chip processor memory circuitry. The on-chip processor memory circuitry may include processor last level cache (LLC) circuitry. One or more physical parameters, such as a stored charge or voltage, may be used to permit the generation of an in-memory analog output using a SRAM array. The generation of an in-memory analog output using only word-line and bit-line capabilities beneficially increases the computational density of the PISA circuit without increasing power requirements. Thus, the systems and methods described herein beneficially leverage the existing capabilities of on-chip SRAM processor memory circuitry to perform a relatively large number of analog vector/tensor calculations associated with execution of a neural network, such as a recurrent neural network, without burdening the processor circuitry and without significant impact to the processor power requirements.

    Adhesion structure for thin film transistor

    公开(公告)号:US11482622B2

    公开(公告)日:2022-10-25

    申请号:US16214706

    申请日:2018-12-10

    Abstract: A transistor structure includes a layer of active material on a base. The base can be insulator material in some cases. The layer has a channel region between a source region and a drain region. A gate structure is in contact with the channel region and includes a gate electrode and a gate dielectric, where the gate dielectric is between the gate electrode and the active material. An electrical contact is on one or both of the source region and the drain region. The electrical contact has a larger portion in contact with a top surface of the active material and a smaller portion extending through the layer of active material into the base. The active material may be, for example, a transition metal dichalcogenide (TMD) in some embodiments.

Patent Agency Ranking