Various embodiments for improved power devices as well as their methods of manufacture, packaging and circuitry incorporating the same for use in a wide variety of power electronic applications are disclosed. One aspect of the invention combines a number of charge balancing techniques and other techniques for reducing parasitic capacitance to arrive at different embodiments for power devices with improved voltage performance, higher switching speed, and lower on-resistance. Another aspect of the invention provides improved termination structures for low, medium and high voltage devices. Improved methods of fabrication for power devices are provided according to other aspects of the invention. Improvements to specific processing steps, such as formation of trenches, formation of dielectric layers inside trenches, formation of mesa structures and processes for reducing substrate thickness, among others, are presented. According to another aspect of the invention, charge balanced power devices incorporate temperature and current sensing elements such as diodes on the same die. Other aspects of the invention improve equivalent series resistance (ESR) for power devices, incorporate additional circuitry on the same chip as the power device and provide improvements to the packaging of charge balanced power devices.

Power semiconductor devices and methods of manufacture

A silicon wafer bonding process is described in which only thermally grown oxide is present between wafer pairs. Bonding occurs after insertion into an oxidizing ambient. It is proposed the wafers are drawn into intimate contact as a result of the gaseous oxygen between them being consumed by oxidation, thus producing a partial vacuum. The proposed bonding mechanism is polymerization of silanol bonds between wafer pairs. Silicon on insulator (SOI) is produced by etching away all but a few microns of one of the bonded pair. Capacitor measurements show a 27 μs minority‐carrier lifetime and no degradation of the SOI‐insulator interface. In addition, there is negligible charge at the bonding interface making the technique attractive for three‐dimensional as well as planar SOI applications.

Wafer bonding for silicon‐on‐insulator technologies

The present invention includes a memory subsystem comprising at least two semiconductor devices (15, 16 , 17), including at least one memory device (15, 16 or 17), connected to a bus (18), where the bus includes a plurality of bus lines for carrying substantially all address, data and control information needed by said memory devices (15, 16 or 17), where the control information includes device-select information and the bus (18) has substantially fewer bus lines than the number of bits in a single address, and the bus (18) carries device-select information without the need for separated device-select lines connected directly to individual devices. The present invention also includes a protocol for master and slave devices to communicate on the bus (18) and for registers in each device to differentiate each device and allow bus requests to be directed to a single or to all devices (15, 16, 17). The present invention includes modifications to prior-art devices to allow them to implement the new features of this invention. In a preferred implementation, 8 bus data lines and an Address Valid bus line carry address, data and control information for memory addresses up to 40 bits wide.

Integrated circuit I/O using a high performance bus interface

General-purpose computing devices allow us to (1) customize computation after fabrication and (2) conserve area by reusing expensive active circuitry for different functions in time. We define RP-space, a restricted domain of the general-purpose architectural space focussed on reconfigurable computing architectures. Two dominant features differentiate reconfigurable from special-purpose architectures and account for most of the area overhead associated with RP devices: (1) instructions which tell the device how to behave, and (2) flexible interconnect which supports task dependent dataflow between operations.
We can characterize RP-space by the allocation and structure of these resources and compare the efficiencies of architectural points across broad application characteristics. Conventional FPGAs fall at one extreme end of this space and their efficiency ranges over two orders of magnitude across the space of application characteristics. Understanding RP-space and its consequences allows us to pick the best architecture for a task and to search for more robust design points in the space.
Our DPGA, a fine-grained computing device which adds small, on-chip instruction memories to FPGAs is one such design point. For typical logic applications and finite-state machines, a DPGA can implement tasks in one-third the area of a traditional FPGA. TSFPGA, a variant of the DPGA which focuses on heavily time-switched interconnect, achieves circuit densities close to the DPGA, while reducing typical physical mapping times from hours to seconds.
Rigid, fabrication-time organization of instruction resources significantly narrows the range of efficiency for conventional architectures. To avoid this performance brittleness, we developed MATRIX, the first architecture to defer the binding of instruction resources until run-time, allowing the application to organize resources according to its needs. Our focus MATRIX design point is based on an array of 8-bit ALU and register-file building blocks interconnected via a byte-wide network. With today's silicon, a single chip MATRIX array can deliver over 10 Gop/s (8-bit ops). On sample image processing tasks, we show that MATRIX yields 10-20$\times$ the computational density of conventional processors.
Understanding the cost structure of RP-space helps us identify these intermediate architectural points and may provide useful insight more broadly in guiding our continual search for robust and efficient general-purpose computing structures. (Copies available exclusively from MIT Libraries, Rm. 14-0551, Cambridge, MA 02139-4307. Ph. 617-253-5668; Fax 617-253-1690.)

Reconfigurable Architectures for General-Purpose Computing

A gate-insulated thin film transistor is disclosed. One improvement is that the thin film transistor is formed on a substrate through a blocking layer in between so that it is possible to prevent the transistor from being contaminated with impurities such as alkali ions which exist in the substrate. Also, a halogen is added to either or both of the blocking layer and a gate insulator of the transistor in order that impurities such as alkaline ions, dangling bonds and the like can be neutralized, therefore, the reliability of the device is improved.

Thin-film transistor

Building on nearly two decades of reported results for MOSFET's fabricated in small-grain polycrystalline silicon, a design methodology is developed that yields devices which have low threshold voltage, high drive current, low leakage current, tight parameteric control, and reduced topology, while requiring no nonstandard materials, processes, and tools. Design criteria and device performance are discussed, grain boundary characterization techniques are described, technological issues pertinent to VLSI implementation are investigated, and long-term device reliability is studied. The potential applications of the polysilicon MOSFET's in high-density dRAM and sRAM are explored. The successful implementation of an experimental stacked CMOS 64K static RAM proves the utility of these devices for three-dimensional integration in a VLSI environment.

Characteristics and three-dimensional integration of MOSFET's in small-grain LPCVD polycrystalline Silicon

DRAM cells and arrays of cells on a semiconductor substrate, together with methods of fabrication, are disclosed wherein the cells are formed in pairs or quartets by excavating a trench or two trenches through the cell elements to split an original cell into two or four cells during the fabrication. The cells include vertical field effect transistors and capacitors along the trench sidewalls with word lines and bit lines crossing over the cells.

Vertical DRAM cell and method

A 1T DRAM cell with both the transistor and the capacitor fabricated on the sidewalls of a deep trench is described. Trench Transistor Cell (TTC) fabrication and characterization are discussed.

http://ieeexplore.ieee.org/iel5/9951/31947/01485625.pdf

A trench transistor cross-point DRAM cell

The described embodiments of the present invention provide structures, and a method for fabricating those structures, which include a memory cell formed within a single trench. A trench is formed in the surface of a semiconductor substrate. The bottom portion of the trench is filled with polycrystalline silicon to form one plate of a storage capacitor. The substrate serves as the other plate of the capacitor. The remaining portion of the trench is then filled with an insulating material such as silicon dioxide. A pattern is then etched into the silicon dioxide which opens a portion of the sidewall and the top portion of the trench down to the polycrystalline capacitor plate. A contact is then formed between the polycrystalline capacitor plate and the substrate. Dopant atoms diffuse through the contact to form a source region on a sidewall of the trench. A gate insulator is formed by oxidation and a drain is formed at the surface of the trench adjacent to the mouth of the trench. Conductive material is then formed inside the open portion of the upper portion of the trench thereby forming a transistor connecting the upper plate of the storage capacitor to a drain region on the surface of the semiconductor substrate.

Dram cell and method

Improvements in polysilicon grain-boundary passivation techniques have made polysilicon MOSFET's increasingly attractive, as vertically stackable circuit components in applications, where high mobility is not a primary requirement. A simple method for the "last step" passivation of grain boundaries in polysilicon MOSFET's is presented. The method involves diffusion of atomic hydrogen at 450°C from a plasma-deposited compressive silicon nitride layer for reaction at silicon grain-boundary dangling bond sites. By use of this technique, ON/OFF current ratios of greater than 106can be achieved with drive currents that are sufficient for many circuit applications.

Ashwin H. Shah

Papers

Characteristics and three-dimensional integration of MOSFET's in small-grain LPCVD polycrystalline Silicon

Vertical DRAM cell and method

A trench transistor cross-point DRAM cell

Dram cell and method

Hydrogen passivation of PolySilicon MOSFET's from a plasma Nitride source