Abstract: Configurable three-dimensional neural network array. In an exemplary embodiment, a three-dimensional (3D) neural network array includes a plurality of stacked synapse layers having a first orientation, and a plurality of synapse lines having a second orientation and passing through the synapse layers. The neural network array also includes synapse elements connected between the synapse layers and synapse lines. Each synapse element includes a programmable resistive element. The neural network array also includes a plurality of output neurons, and a plurality of select transistors connected between the synapse lines and the output neurons. The gate terminals of the select transistors receive input signals.

Abstract: A method of storing information or data in a nonvolatile memory device with multiple-page programming. The method, in one aspect, is able to activate a first drain select gate (“DSG”) signal. After loading the first data from a bit line (“BL”) to a nonvolatile memory page of a first memory block in response to activation of the first DSG signal during a first clock cycle, the first DSG signal is deactivated. Upon activating a second DSG signal, the second data is loaded from the BL to a nonvolatile memory page of a second memory block. The first data and the second data are simultaneously written to the first memory block and the second memory block, respectively.

Abstract: A memory device includes a static random-access memory (“SRAM”) circuit and a first nonvolatile memory (“NVM”) string, a second NVM string, a first and a second drain select gates (“DSGs”). The SRAM circuit is able to temporarily store information in response to bit line (“BL”) information which is coupled to at the input terminal of the SRAM circuit. The first NVM string having at least one nonvolatile memory cell is coupled to the output terminal of the SRAM. The first DSG is operable to control the timing for storing information at the output terminal of the SRAM to the first nonvolatile memory. The second NVM string having at least one nonvolatile memory cell is coupled to the output terminal of the SRAM. The second DSG controls the timing for storing information at the output terminal of the SRAM to the second nonvolatile memory string.

Abstract: A 3D NAND array with divided string architecture. In one aspect, an apparatus includes a plurality of charge storing devices connected to form a cell string. The apparatus also includes one or more internal select gates connected between selected charge storing devices in the cell string. The one or more internal select gates divide the cell string into two or more segments of charge storing devices. Selectively enabling and disabling the one or more internal select gates during programming operates to isolate one or more selected segments to reduce program-disturb to remaining segments. In another embodiment, a method is provided for programming a memory cell of a cell string having internal select gates that isolate the memory cell to reduce the effects of program-disturb. In another embodiment, multiple memory cells of a cell string having internal select gates are programmed with reduced program-disturb.

Abstract: A CMOS anti-fuse cell is disclosed. In one aspect, an apparatus includes an N? well and an anti-fuse cell formed on the N? well. The anti-fuse cell includes a drain P+ diffusion deposited in the N? well, a source P+ diffusion deposited in the N? well, and an oxide layer deposited on the N? well and having an overlapping region that overlaps the drain P+ diffusion. A control gate is deposited on the oxide layer. A data bit of the anti-fuse cell is programmed when a voltage difference between the control gate and the drain P+ diffusion exceeds a voltage threshold of the oxide layer and forms a leakage path from the control gate to the drain P+ diffusion. The leakage path is confined to occur in the overlapping region.

Abstract: A compact CMOS anti-fuse memory cell. In one aspect, an apparatus includes an N-well and an anti-fuse cell formed on the N-well. The anti-fuse cell includes a lightly doped drain (LDD) region deposited in the N-well, an oxide layer deposited on the N-well and having an overlapping region that overlaps the LDD region, and a control gate deposited on the oxide layer, wherein a bit of the anti-fuse cell is programmed when a voltage difference between the control gate and the LDD region exceeds a voltage threshold of the oxide layer and forms a leakage path from the control gate to the LDD region, and wherein the leakage path is confined to occur in the overlapping region.

Abstract: A memory device is able to store data using both on-chip dynamic random-access memory (“DRAM”) and nonvolatile memory (“NVM”). The memory device, in one aspect, includes NVM cells, word lines (“WLs”), a cell channel, and a DRAM mode select. The NVM cells are capable of retaining information persistently and the WLs are configured to select one of the NVM cells to be accessed. The cell channel, in one embodiment, is configured to interconnect the NVM cells to form a NVM string. The DRAM mode select can temporarily store data in the cell channel when the DRAM mode select is active.

Abstract: A CMOS anti-fuse cell is disclosed. In one aspect, an apparatus includes an N? well and an anti-fuse cell formed on the N? well. The anti-fuse cell includes a drain P+ diffusion deposited in the N? well, a source P+ diffusion deposited in the N? well, and an oxide layer deposited on the N? well and having an overlapping region that overlaps the drain P+ diffusion. A control gate is deposited on the oxide layer. A data bit of the anti-fuse cell is programmed when a voltage difference between the control gate and the drain P+ diffusion exceeds a voltage threshold of the oxide layer and forms a leakage path from the control gate to the drain P+ diffusion. The leakage path is confined to occur in the overlapping region.

Abstract: A memory system is configured to store information using a hybrid volatile and nonvolatile memory device. The memory system, in one aspect, includes memory components, a drain select gate (“DSG”) transistor, and a capacitor component. Each memory component, in one example, includes a source terminal, a gate terminal, a drain terminal, and a nonvolatile cell. The memory components are organized in a string formation and the components are interconnected between source terminals and drain terminals. The drain terminal of DSG transistor is coupled to the source terminal of a memory component and the gate terminal of DSG transistor is coupled to a DSG signal. The drain terminal of the capacitor is coupled to the source terminal of the first DSG transistor. The capacitor component is configured to perform a dynamic random-access memory (“DRAM”) function.

Abstract: Self-learning for neural network arrays. In an exemplary embodiment, a method includes determining input voltages to be applied to one or more input neurons of a neural network, and determining target output voltages to be obtained at one or more output neurons of the neural network in response to the input voltages. The neural network also includes a plurality of hidden neurons and synapses connecting the neurons, and each of a plurality of synapses includes a resistive element. The method also includes applying the input voltages to the input neurons, and applying the target output voltages or complements of the target output voltages to the output neurons to simultaneously program the resistive elements of the plurality of synapses.

Abstract: A memory device includes a static random-access memory (“SRAM”) circuit and a first nonvolatile memory (“NVM”) string, a second NVM string, a first and a second drain select gates (“DSGs”). The SRAM circuit is able to temporarily store information in response to bit line (“BL”) information which is coupled to at the input terminal of the SRAM circuit. The first NVM string having at least one nonvolatile memory cell is coupled to the output terminal of the SRAM. The first DSG is operable to control the timing for storing information at the output terminal of the SRAM to the first nonvolatile memory. The second NVM string having at least one nonvolatile memory cell is coupled to the output terminal of the SRAM. The second DSG controls the timing for storing information at the output terminal of the SRAM to the second nonvolatile memory string.

Abstract: Three-dimensional neural network array. In an exemplary embodiment, a three-dimensional (3D) neural network includes a plurality of input conductors forming a plurality of stacked input layers having a first orientation, and at least one output conductor forming an output layer having the first orientation. The three-dimensional (3D) neural network also includes a plurality of hidden conductors having a second orientation. Each hidden conductor includes an in-line threshold element. The three-dimensional (3D) neural network also includes synapse elements coupled between the hidden conductors and the input conductors and between the hidden conductors and the output conductor. Each synapse element includes a programmable resistive element.

Abstract: A high-density neural network array. In an exemplary embodiment, an apparatus includes a three-dimensional (3D) structure having a plurality of layers forming a neural network. Each layer comprises one or more conductors forming neurons with each neuron having neuron inputs and neuron outputs. The apparatus also includes synapse elements coupled between the neurons outputs and the neuron inputs of neurons in adjacent layers. Each synapse element comprises a material that applies a selected weight to signals passing between neurons connected to that synapse element.

Abstract: A SONOS byte-erasable EEPROM is disclosed. In one aspect, an apparatus includes a plurality of SONOS memory cells forming an EEPROM memory array. The apparatus also includes a controller that generates bias voltages to program and erase the memory cells. The controller performs a refresh operation when programming selected memory cells to reduce write-disturb on unselected memory cells to prevent data loss.

Abstract: A memory device includes a static random-access memory (“SRAM”) circuit and a first nonvolatile memory (“NVM”) string, a second NVM string, a first and a second drain select gates (“DSGs”). The SRAM circuit is able to temporarily store information in response to bit line (“BL”) information which is coupled to at the input terminal of the SRAM circuit. The first NVM string having at least one nonvolatile memory cell is coupled to the output terminal of the SRAM. The first DSG is operable to control the timing for storing information at the output terminal of the SRAM to the first nonvolatile memory. The second NVM string having at least one nonvolatile memory cell is coupled to the output terminal of the SRAM. The second DSG controls the timing for storing information at the output terminal of the SRAM to the second nonvolatile memory string.

Abstract: A CMOS anti-fuse cell is disclosed. In one aspect, an apparatus includes an N? well and an anti-fuse cell formed on the N? well. The anti-fuse cell includes a drain P+ diffusion deposited in the N? well, a source P+ diffusion deposited in the N? well, and an oxide layer deposited on the N? well and having an overlapping region that overlaps the drain P+ diffusion. A control gate is deposited on the oxide layer. A data bit of the anti-fuse cell is programmed when a voltage difference between the control gate and the drain P+ diffusion exceeds a voltage threshold of the oxide layer and forms a leakage path from the control gate to the drain P+ diffusion. The leakage path is confined to occur in the overlapping region.

Abstract: A memory device is able to store data using both on-chip dynamic random-access memory (“DRAM”) and nonvolatile memory (“NVM”). The memory device, in one aspect, includes NVM cells, word lines (“WLs”), a cell channel, and a DRAM mode select. The NVM cells are capable of retaining information persistently and the WLs are configured to select one of the NVM cells to be accessed. The cell channel, in one embodiment, is configured to interconnect the NVM cells to form a NVM string. The DRAM mode select can temporarily store data in the cell channel when the DRAM mode select is active.

Abstract: A CMOS anti-fuse cell is disclosed. In one aspect, an apparatus includes an N? well and an anti-fuse cell formed on the N? well. The anti-fuse cell includes a drain P+ diffusion deposited in the N? well, a source P+ diffusion deposited in the N? well, and an oxide layer deposited on the N? well and having an overlapping region that overlaps the drain P+ diffusion. A control gate is deposited on the oxide layer. A data bit of the anti-fuse cell is programmed when a voltage difference between the control gate and the drain P+ diffusion exceeds a voltage threshold of the oxide layer and forms a leakage path from the control gate to the drain P+ diffusion. The leakage path is confined to occur in the overlapping region.

Abstract: A memory device is able to store data using both on-chip dynamic random-access memory (“DRAM”) and nonvolatile memory (“NVM”). The memory device, in one aspect, includes NVM cells, word lines (“WLs”), a cell channel, and a DRAM mode select. The NVM cells are capable of retaining information persistently and the WLs are configured to select one of the NVM cells to be accessed. The cell channel, in one embodiment, is configured to interconnect the NVM cells to form a NVM string. The DRAM mode select can temporarily store data in the cell channel when the DRAM mode select is active.

Abstract: A dual function hybrid memory cell is disclosed. In one aspect, the memory cell includes a substrate, a bottom charge-trapping region formed on the substrate, a top charge-trapping region formed on the bottom charge-trapping region, and a gate layer formed on the top charge trapping region. In another aspect, a method for programming a memory cell having a substrate, a bottom charge-trapping layer, a top charge-trapping layer, and a gate layer is disclosed. The method includes biasing a channel region of the substrate, applying a first voltage differential between the gate layer and the channel region, injecting charge into the bottom charge-trapping layer from the channel region based on the first voltage differential. The method also includes applying a second voltage differential between the gate layer and the channel region and injecting charge from the bottom charge-trapping layer into the top charge-trapping layer based on the second voltage differential.