Abstract: A pixel cell includes a photodiode disposed in a pixel cell region and proximate to a front side of a semiconductor layer to generate image charge in response to incident light directed through a backside to the photodiode. A cell deep trench isolation (CDTI) structure is disposed in the pixel cell region along an optical path of the incident light to the photodiode and proximate to the backside. The CDTI structure includes a central portion extending a first depth from the backside towards the front side. Planar outer portions extend laterally outward from the central portion. The planar output portions further extend a second depth from the backside towards the front side. The first depth is greater than the second depth. Planes formed by each of the planar outer portions intersect in a line coincident with a longitudinal center line of the central portion of the CDTI structure.

Abstract: A pixel-array substrate includes a floating diffusion region and a first photodiode formed in a semiconductor substrate. A top surface of the semiconductor substrate defines a trench 1A and a trench 1B each (i) extending into the semiconductor substrate away from a planar region of the top surface between the trench 1A and the trench 1B and (ii) having a respective distal end, with respect to the floating diffusion region, located between the floating diffusion region and the first photodiode. In a horizontal plane parallel to the top surface and along an inter-trench direction between the trench 1A and the trench 1B, a first spatial separation between the trench 1A and the trench 1B increases with increasing distance from the floating diffusion region.

Abstract: A pointed-trench pixel-array substrate includes a floating diffusion region and a photodiode region formed in a semiconductor substrate. The semiconductor substrate includes, between a top surface and a back surface thereof, a sidewall surface and a bottom surface defining a trench extending into the semiconductor substrate away from a planar region of the top surface surrounding the trench.

Abstract: A method for forming a pixel includes forming, in a semiconductor substrate, a wide trench having an upper depth with respect to a planar top surface of the semiconductor substrate. The method also includes ion-implanting a floating-diffusion region between the planar top surface and a junction depth in the semiconductor substrate. In a cross-sectional plane perpendicular to the planar top surface, the floating-diffusion region has (i) an upper width between the planar top surface and the upper depth, and (ii) between the upper depth and the junction depth, a lower width that exceeds the upper width. Part of the floating-diffusion region is beneath the wide trench and between the upper depth and the junction depth.

Abstract: A pixel includes a semiconductor substrate, a photodiode region, a floating diffusion region, and a dielectric layer. The substrate has a top surface forming a trench lined by the dielectric layer, and having a trench depth relative to a planar region of the top surface. The photodiode region is in the substrate and includes a bottom photodiode section beneath the trench and a top photodiode section adjacent to the trench, adjoining the bottom photodiode section, and extending toward the planar region to a photodiode depth less than the trench depth. The floating diffusion region is adjacent to the trench and has a junction depth less than the trench depth. A top region of the dielectric layer is between the planar region and the junction depth. A bottom region of the dielectric layer is between the photodiode depth and the trench depth, and thicker than the top region.

Abstract: A pixel includes a semiconductor substrate, an upper surface thereof forming a trench having a trench depth relative to a planar region of the upper surface surrounding the trench, and in a plane perpendicular to the planar region; an upper width between the planar region and an upper depth that is less than the trench depth; and a lower width, between the upper depth and the trench depth, that is less than the upper width. A floating diffusion region adjacent to the trench extends away from the planar region to a junction depth exceeding the upper depth and is less than the trench depth. The photodiode region in the substrate includes a lower photodiode section beneath the trench and an upper photodiode section adjacent to the trench, beginning at a photodiode depth that is less than the trench depth, extending toward and adjoining the lower photodiode section.

Abstract: A photodiode array has buried photodiodes and vertical selection transistors. Trenches are lined with gate oxide and metallic plugs of first material lie within the trenches. Gate contacts of second material contact the metallic plugs, with photodiode diffusion regions adjacent the trenches as sources of vertical transistors, the metallic plugs form gates of the vertical transistors, and buried photodiode regions form sources of the vertical transistors. In embodiments, the first conductive material is tungsten, titanium nitride, titanium carbide, or aluminum and the second conductive material is polysilicon. The array is formed by trenching, growing gate oxide, and depositing first material in the trenches. The first material is etched to define metallic plugs, the second material is deposited onto the metallic plugs then masked and etched; and drain regions implanted. Etching the second material is performed by a reactive ion etch that stops upon reaching the metallic plugs.

Abstract: A pixel cell includes a photodiode disposed proximate to a front side of a semiconductor layer to generate image charge in response to incident light directed through a backside of the semiconductor layer. A cell deep trench isolation (CDTI) structure is disposed along an optical path of the incident light to the photodiode and proximate to the backside of the semiconductor layer. The CDTI structure includes a plurality of portions arranged in the semiconductor layer. Each of the plurality of portions extends a respective depth from the backside towards the front side of the semiconductor layer. The respective depth of each of the plurality of portions is different than a respective depth of a neighboring one of the plurality of portions. Each of the plurality of portions is laterally separated and spaced apart from said neighboring one of the plurality of portions in the semiconductor layer.

Abstract: A CMOS image sensor has an array of photodiode cells, the photodiode cells each include four buried photodiodes coupled by vertical transfer gate transistors to a single floating node diffusion. Each cell also has a reset transistor coupled to the floating node diffusion, a source follower transistor having gate coupled to the floating node diffusion, and a read select transistor coupled to the source follower transistor. The reset transistor, source follower transistor, and read select transistor have predominately gate and shape edges oriented at an angle greater than 30-degrees and less than 60-degrees from a line extending along an entire horizontal row of photodiodes of a photodiode array of the image sensor and are formed vertically above, and in the same integrated circuit as, the photodiodes of the photodiode array.

Abstract: A semiconductor structure for a CMOS image sensor includes a semiconductor substrate having a first side and a second side. A photodiode is disposed in the semiconductor substrate proximate to the first side. The photodiode accumulates image charge photogenerated in the photodiode in response to incident light directed through the second side. A deep trench isolation structure enclosing the photodiode. The deep trench isolation structure extends from the second side toward the first side. The deep trench isolation structure includes a light absorption region disposed at a first end of the deep trench isolation structure toward the first side.

Abstract: A computer-implemented method for distributed synchronous training of a neural network model includes performing, by a worker machine of a plurality of worker machines, a forward computation of a training data set using a plurality of N layers of the neural network model. The forward computation starts at Layer 1 and proceeds through Layer N of the neural network model. The method further includes performing, by the worker machine, a backward computation of the training data set, the backward computation starting at Layer N and proceeding through Layer 1 of the neural network model. The method further includes synchronizing, by the worker machine, a plurality of gradients outputted by the neural network model during the backward computation. The synchronizing of the plurality of gradients is performed with other worker machines of the plurality of worker machines and in parallel with the backward computation.

Abstract: A pixel cell includes a photodiode disposed in a pixel cell region and proximate to a front side of a semiconductor layer to generate image charge in response to incident light directed through a backside to the photodiode. A cell deep trench isolation (CDTI) structure is disposed in the pixel cell region along an optical path of the incident light to the photodiode and proximate to the backside. The CDTI structure includes a central portion extending a first depth from the backside towards the front side. Planar outer portions extend laterally outward from the central portion. The planar output portions further extend a second depth from the backside towards the front side. The first depth is greater than the second depth. Planes formed by each of the planar outer portions intersect in a line coincident with a longitudinal center line of the central portion of the CDTI structure.

Abstract: An image sensor includes photodiodes disposed in a pixel region and proximate to a front side of a semiconductor layer. A backside metal grating is formed in a backside oxide layer disposed proximate to a backside of the semiconductor layer. A deep trench isolation (DTI) structure with a plurality of pixel region portions and an edge region portion is formed in the semiconductor layer. The pixel region portions are disposed in the pixel region of the semiconductor layer such that incident light is directed through the backside metal grating, through the backside of the semiconductor layer, and between the pixel region portions of the DTI structure to the photodiodes. The edge region portion of the DTI structure is disposed in an edge region outside of the pixel region. The edge region portion of the DTI structure is biased with a DTI bias voltage.

Abstract: A computer-implemented method for distributed synchronous training of a neural network model includes detecting gradient sets from a plurality of worker machines, each worker machine generating a gradient set in a current iteration of a training data set, and each gradient set of the gradient sets comprising a plurality of gradients. A lagging gradient set from a lagging worker machine is detected. The lagging gradient set is generated by the lagging worker machine in a prior iteration of the training data set. Aggregated gradients are generated based on the gradient sets and the lagging gradient set. The neural network model is updated based on the aggregated gradients.

Abstract: The present disclosure relates to semiconductor structures and, more particularly, to single diffusion cut for gate structures and methods of manufacture. The structure includes a single diffusion break extending into a substrate between diffusion regions of adjacent gate structures, the single diffusion break filled with an insulator material and further comprising an undercut region lined with a liner material which is between the insulator material and the diffusion regions.

Abstract: The present disclosure generally relates to semiconductor structures and, more particularly, to middle of line gate structures and methods of manufacture. The structure includes: a plurality of adjacent gate structures; a bridged gate structure composed of a plurality of the adjacent gate structures; source and drain regions adjacent to the bridged gate structure and comprising source and drain metallization features; and contacts to the bridged gate structure and the source and drain metallization features.

Abstract: An apparatus and method are provided for review-based machine learning. Included are a non-transitory memory storing instructions and one or more processors in communication with the non-transitory memory. The one or more processors execute the instructions to receive first data, generate a plurality of first features based on the first data, and identify a first set of labels for the first data. A first model is trained using the first features and the first set of labels. The first model is reviewed to generate a second model, by receiving a second set of labels for the first data, and reusing the first features with the second set of labels in connection with training the second model.

Abstract: The present disclosure relates to semiconductor structures and, more particularly, to single diffusion cut for gate structures and methods of manufacture. The structure includes a single diffusion break extending into a substrate between diffusion regions of adjacent gate structures, the single diffusion break filled with an insulator material and further comprising an undercut region lined with a liner material which is between the insulator material and the diffusion regions.

Abstract: A finFET device is disclosed including a fin defined in a semiconductor substrate, the fin having an upper surface and a first diffusion break positioned in the fin, wherein the first diffusion break comprises an upper surface that is substantially coplanar with the upper surface of the fin.

Abstract: The present disclosure relates to semiconductor structures and, more particularly, to semiconductor device with reduced gate height budget and methods of manufacture. The method includes: forming a plurality of gate structures on a substrate; recessing material of the plurality of gate structures to below a surface of an insulator material; forming trenches in the insulator material and underlying material adjacent to sidewalls of the plurality of gate structures; and filling the recesses and trenches with a capping material.