JP4716736B2 - Memory cell using gated diode and method of use thereof, semiconductor structure - Google Patents

Description

  The present invention relates to semiconductor structures, and more particularly to semiconductor memories.

  During the past decades, dynamic random access memory (DRAM) has continued to become slower due to higher density, and static random access memory (SRAM) has continued to become lower density due to higher speed. Yes. The two memory types have been developed separately along their own technology development curves for density and speed. Recently, embedded memories in DRAM and SRAM have begun to appear. Embedded memory is memory that is on the same “chip” as the processor. Currently, each DRAM and SRAM also has its own technical issues along its unique scaling path: standby and active large leakage currents (both sub-threshold and tunnel), threshold voltage Facing problems due to fluctuations and inconsistencies. With embedded memory, new memory cell possibilities suitable for better optimization and adjustment between speed, area, power, hold time, soft error rate and technical parameters such as threshold voltage and leakage current Demand is opened. The embedded memory of the chip is typically located in a specific processor or application specific integrated circuit (ASIC) based on the chip and application requirements.

In particular, if the DRAM is smaller and made substantially faster, the DRAM is advantageous for embedded memory and its cells and is suitable for smaller voltages as the voltage continues to decrease.
U.S. Patent Application No. 10 / 751,713

  Accordingly, there is a need to provide an improved memory cell and memory that uses this cell.

  Exemplary aspects of the present invention provide improved memory cells, memory arrays, and methods of using them.

  In an exemplary aspect of the invention, a memory cell is disclosed. The memory cell is configured to be coupled to a first control line, at least one second control line, and at least one bit line for selectively accessing the memory cell. The memory cell includes a write switch having a control terminal and first and second terminals, the first terminal of the write switch being coupled to the at least one bit line, the control terminal of the write switch being a first terminal. Is connected to the control line. The memory cell also has a two-terminal semiconductor device having first and second input terminals, the first input terminal of the two-terminal semiconductor device being coupled to the second terminal of the write switch; and A second input terminal of the two-terminal semiconductor device is coupled to the at least one second control line. Two-terminal semiconductor devices are used as charge storage devices in memory cells. The two-terminal semiconductor device is configured so that the capacitance when the voltage of the first input terminal with respect to the second input terminal exceeds the threshold voltage is larger than the capacitance when the voltage does not exceed the threshold voltage. And a device.

  The memory cell further comprises a read selection switch having a control terminal and first and second terminals, the control terminal of the read selection switch being coupled to the at least one second control line, A first terminal of the read select switch is coupled to the at least one bit line. The memory cell also has a read switch having a control terminal and first and second terminals, the control terminal of the read switch being the first input terminal of the two-terminal semiconductor device and the second of the write switch. The first terminal of the read switch is coupled to the second terminal of the read select gate, and the second terminal of the read switch is coupled to ground.

  The two terminal semiconductor device may be a gated diode. This gate control diode is a semiconductor device composed of a source and a gate, and when the gate-source voltage (Vgs) exceeds a threshold voltage, charge is accumulated in the inversion layer, otherwise substantially less charge is charged. Either it is accumulated or no charge is accumulated. Specifically, when the gate-source voltage (Vgs) exceeds a threshold, the two-terminal device has a capacitance, and when the gate-source voltage (Vgs) is lower than the threshold voltage, is the capacitance substantially smaller? Or smaller in digit order. The switch is typically implemented as a field effect transistor (FET), the gated diode is typically implemented as a “partial” FET, and the memory cell FETs can be n-type or p-type FETs. Furthermore, the two-terminal semiconductor device can be realized as an n-type or p-type semiconductor device.

  In another exemplary aspect of the invention, an array using the memory cell is disclosed. Such an array can have more than one bit line and several control lines. For example, single port or dual port memory cells can be used. More ports can be used if desired. The control line can be realized in several ways. By way of example, the at least one second control line may be a single control line coupled to both the read selection switch and the second terminal of the two-terminal semiconductor device. Having a single second control line allows, for example, read, write, and read voltage increase (eg, by changing the voltage at the second terminal of a two-terminal semiconductor device by the memory cell) High voltage is created). As another example, the second control line is two control lines: a write control line coupled to the second terminal of the two-terminal device and a read selection control line coupled to the control terminal of the read selection switch. Can be divided. This configuration allows for read, read voltage increase, write, and write voltage increase (eg, increased voltage is stored in the memory cell). In addition, this configuration allows voltage to be applied to the read selection switch so that the leakage current of the read selection switch (eg, FET) is minimized.

  In another exemplary aspect of the invention, a method for accessing a memory cell is disclosed. The cell is read by changing the voltage at the second terminal of the two-terminal semiconductor device. In general, the voltage at the second terminal of a two-terminal semiconductor device is raised from a small voltage (eg, ground) to a large voltage (eg, “VB”). However, this change depends on whether an n-type device or a p-type device is used. If the cell is at a high voltage (eg, a value of data 1), the two-terminal semiconductor device has a large capacitance, a large amount of charge is stored, and the voltage at the first terminal is (eg, data 1 is It is raised by approximately the voltage of the second terminal of the two-terminal semiconductor device (above the stored cell voltage). If the cell is at a low voltage (eg, a value of data 0), the two-terminal semiconductor device has a small capacitance or a non-existing capacitance, accumulates very little or no charge, and The voltage at the first terminal is raised very slightly (eg, above the cell voltage where data 0 is stored). Changing the voltage at the second terminal of the two-terminal gate control device while reading the memory cell is called increasing the read voltage.

  The write voltage can also be increased by increasing the voltage at the second terminal of the two-terminal semiconductor device while writing the value to the memory cell. The write voltage increase allows for a much larger voltage than the predetermined write voltage for data 1, so that a longer storage time is obtained when the value of data 1 is stored in the cell.

  In another exemplary aspect of the invention, the semiconductor has a semiconductor memory device, the semiconductor memory device having a first control line for selectively accessing the semiconductor memory device, at least one second. To the control line and at least one bit line. The semiconductor memory device comprises a write transistor comprising an insulator formed between a gate and a well, a gate, and first and second source / drain diffusion regions formed on the side of the gate, and the first of the write transistors. One source / drain diffusion region is coupled to the at least one bit line and the gate of the write transistor is coupled to the first control line. The semiconductor memory device further includes a two-terminal semiconductor device including at least a source diffusion region at least in contact with an insulator formed between the gate and the well, the gate, and a part of the insulator. The gate is coupled to the second source / drain diffusion region of the write switch, and the source diffusion region of the two-terminal semiconductor device is coupled to the at least one second control line. The semiconductor memory device also includes a read select transistor comprising an insulator formed between the gate and the well, the gate, and first and second source / drain diffusion regions formed on the side of the gate, The gate of the read select switch is coupled to the at least one second control line, and the first source / drain diffusion region of the read select transistor is coupled to the at least one bit line. The semiconductor memory device further comprises a read transistor comprising an insulator formed between the gate and the well, the gate, and first and second source / drain diffusion regions formed on the gate side, the read transistor Is coupled to the first terminal of the two-terminal semiconductor device and to the second source / drain diffusion region of the write transistor, and the first source / drain diffusion region of the read transistor is the second source of the read select gate. The second source / drain diffusion region of the read transistor is coupled to ground.

  A more complete understanding of the present invention, as well as further features and advantages of the present invention, will be obtained by reference to the following detailed description and drawings.

  This disclosure relates to high speed non-destructive read memory cells based on gated diodes, and related arrays and silicon that can be easily implemented with normal logic based bulk silicon and silicon on insulator (SOI) The structure will be described. The target performance of the memory cell can be equal to or even better than the speed of static random access memory (SRAM), from 50% to 70% of the area of the same technology generation SRAM. It occupies only an area of%. Due to the high gain characteristics of the memory cell itself, the memory operates at a very low power supply voltage that is much smaller than that required by conventional dynamic random access memory (DRAM) and SRAM. be able to. An important emphasis here is on the high speed aspects of memory cells and architectures intended for sufficient retention time and low soft error.

  For ease of reference, the following disclosure is divided into the following sections: Introductory and gated diode structure, gated diode circuit, 3T1D memory cell, voltage increase and voltage gain, retention time, leakage and capacitance ratio, write gate retention time and subthreshold leakage control, read operation and voltage increase Divided into methods, write operations and voltage increase methods, 3T1D memory cell structures, and 3T1D gated diode memory arrays.

Introduction and Gated Diode Structure The term “gated diode” as used herein generally refers to a semiconductor device comprised of a source and a gate, where the gate-source voltage (Vgs) When the voltage exceeds the threshold voltage, charge is accumulated in the inversion layer, otherwise substantially less charge is accumulated or no charge is accumulated. A gated diode is an example of a two-terminal semiconductor device. A two-terminal semiconductor device has a capacitance that is (generally large) when the voltage at the first terminal relative to the second terminal is greater than a predetermined voltage, and the voltage at the first terminal relative to the second terminal is Any two-terminal semiconductor device having the property of having a smaller capacitance (generally a much smaller capacitance) when smaller than a predetermined voltage can be used. This predetermined voltage is referred to herein as the threshold voltage, and the increase in capacitance usually only needs to be on a smaller amount of voltage than this threshold voltage. Thus, the capacitance of a two terminal semiconductor device is non-linear. For example, in a gated diode made using n-type field effect transistor (FET) technology, a high voltage above the threshold voltage causes a large amount of charge to accumulate in the inversion layer and A lower voltage will cause a substantially small amount of charge to accumulate several orders of magnitude or no charge at all. In future technologies, as described below, when implementing memory cells and circuits based on gated diodes, this gate source goes beyond bulk silicon and silicon-on-insulator (SOI). Inter-characteristic can be used.

  As shown in the following figure, in the background of a conventional field effect transistor (FET), a gated diode is a three-terminal FET device (n-type or p-type), as shown in (for example) FIGS. The source and gate of either type can be formed, and the drain is floating (eg, not connected or absent). In this exemplary form, the gated diode is implemented in the form of a “partial” FET or “half” FET. As shown in FIGS. 4, 10, 12 and 14 (for example), sometimes the source and drain of such a FET may be connected to each other at the same potential, and two gated diodes connected in parallel and Can be considered. In this disclosure, these two different gated diodes are used interchangeably. And, unless explicitly specified, a gated diode is considered just the first basic form, namely the source and gate of a semiconductor device.

  FIG. 1 shows exemplary symbols used for the first n-type gated diode. Symbol 190 is an exemplary symbol for the first n-type gated diode shown in FIGS. FIG. 2 shows an example of a side view of the first n-type gate control diode 100 formed of a semiconductor. The first n-type gated diode 100 includes a gate insulator 120 formed between a gate 115 (eg, N + doped polysilicon) and a p-well 130, a source diffusion region 110, and two shallow trench isolations. (STI) regions 105 and 125, an optional n isolation band 140, and a p-substrate 135. As will be described below, the threshold voltage of the gated diode 100 is substantially controlled by the dopant concentration of the p-well 130.

  FIG. 3 shows exemplary symbols used for the second n-type gated diode. Symbol 190 is an exemplary symbol for the second n-type gated diode shown in FIGS. The same symbol 190 is used in both FIGS. FIG. 4 shows an example of a side view of the second n-type gate control diode 100 formed of a semiconductor. The second n-type gated diode 200 includes a gate insulator 220 formed between a gate 215 (eg, N + doped polysilicon) and a p-well 230, a source diffusion region 210, and two STI regions 205. 225, optional n isolation band 240, p substrate 235, "drain" diffusion region 245 (eg, second source / drain diffusion region), and electrically coupling source diffusion region 210 and "drain" diffusion region 245 Interconnect 250 is provided. As described below, the threshold voltage of the gated diode 100 is substantially controlled by the dopant concentration of the p-well 230.

  FIGS. 5 and 6 show how the charge stored in the gated diode 100/200 varies with the gate-source voltage (Vgs). A voltage corresponding to a high voltage (for example, data 1) exists in the gate 115/215, and the gate-source voltage (Vgs) is higher than the threshold voltage (Vt) of the gate control diode 100/200. At times, charge is stored in the gate 115/215 (eg, by the inversion layer 126/226 formed under the gate 115/215). A voltage corresponding to a low voltage (low voltage) (for example, data 0) exists in the gate 115/215, and the gate-source voltage (Vgs) is lower than the threshold voltage (Vt) of the gate control diode 100/200 Sometimes, no or little charge is accumulated in the gate 115/215 (eg, no or little charge is accumulated under the gate 115/225 because there is no inversion layer 126/226). . Therefore, the capacitance characteristics of the gated diode with respect to the gate-source voltage (Vgs) are shown in FIG. Capacitance (eg, as determined by the accumulated charge divided by the voltage) is negligible when Vgs is below the threshold voltage, and increases linearly with Vgs when Vgs exceeds the threshold voltage. And settles to a maximum value above a certain value of Vgs. When Vgs exceeds the threshold voltage, a certain amount of charge is stored in the gated diode, which acts as a capacitor. When Vgs is below the threshold voltage, the gated diode 100/200 has negligible capacitance. The area under the graph is the amount of accumulated charge. It is this changing capacitance with respect to Vgs that gives rise to many new and useful circuits in memory and logic circuits.

  The threshold voltage of the gated diode 100/200 is over a relatively wide range of zero Vt, low Vt, standard Vt, and high Vt during manufacturing implants (eg, p wells 130, 230 or wells shown in the following figures). It can be controlled very precisely by the amount of dopant) diffused. Vt increases as the dopant concentration increases, and zero or very low Vt gated diodes require little or no dopant implantation, so that the memory cell used herein The result is a more accurate threshold voltage that is less sensitive to process variations.

  FIG. 6 also shows how the capacitance of the gated diode varies with the size of the gate of the gated diode. The maximum value of the gated diode capacitance obtained when Vgs is substantially higher than the threshold voltage is approximately proportional to the gated diode gate area for a given gate oxide thickness and dielectric constant.

  When the gate-source voltage (Vgs) exceeds the threshold and represents data 1, the gated diode accumulates a sufficient amount of charge in the inversion layer and accumulates little or no charge for data 0. As such, gated diode memory cells (described in more detail below) have inherently high gain (eg, greater than 1) characteristics. Further, when using voltage augmentation techniques, gated diode memory cells are much smaller than the bit line voltage required for DRAM and SRAM (eg, typically 50% with the same technology) Can be written with line voltage. A gated diode, such as gated diode 100/200, forms a memory cell during reading and writing, generally where the memory cell voltage is increased only when writing or reading the value of data 1. The voltage increase has no effect on the opposite data, ie data 0. As a result, the gated diode memory cell achieves voltage gain. This gain characteristic is unique and is one element that differentiates gated diode memory cells from other SRAM cells, DRAM cells, and gain cells.

  In this specification, it is assumed that the voltage corresponding to data 1 is the high voltage (high voltage) of the technology used and the voltage corresponding to data 0 is the low voltage (low voltage) of the technology. However, this is just an assumption and the logic level may be reversed.

  For the purposes of this disclosure, a gated diode is assumed to be n-type unless explicitly stated. In the case of a p-type gated diode, the voltage and operation are complementary to the n-type, and those skilled in the art can easily design it accordingly.

  FIG. 7 shows exemplary symbols used for the first p-type gated diode. Symbol 490 is an exemplary symbol for the first p-type gated diode shown in FIG. The same symbol 490 is used in both FIGS. FIG. 8 shows an example of a side view of the first p-type gate control diode 400 formed of a semiconductor. The first p-type gate control diode 400 includes a gate insulator 420 formed between a gate 415 (eg, P + doped polysilicon) and an n-well 430, a source diffusion region 410, and two STI regions 405. And 425, and a p-substrate 435. The threshold voltage of the gated diode 400 is substantially controlled by the dopant concentration of the n-well 430.

  FIG. 9 shows exemplary symbols used for the second p-type gated diode. Symbol 490 is an exemplary symbol for the second p-type gated diode shown in FIG. FIG. 10 shows an example of a side view of the second p-type gate control diode 500 formed of a semiconductor. The second p-type gated diode 500 includes a gate insulator 520 formed between a gate 515 (eg, P + doped polysilicon) and an n-well 530, a source diffusion region 510, and two STI regions 505. And 525, a p-substrate 535, a “drain” diffusion region 545, and an interconnect 550 that electrically couples the source diffusion region 510 and the “drain” diffusion region 545. The threshold voltage of the gated diode 500 is substantially controlled by the dopant concentration of the n-well 530.

  FIG. 11 shows an example of a side view of a first n-type gated diode 600 formed of SOI. The first n-type gated diode 600 includes a gate insulator 620, a source diffusion region 610, and two STI regions 605 formed between a gate 615 (eg, N + doped polysilicon) and a p-well 630. And 625, and an insulator 635. A p-well 630 is formed on the well boundary 636. The threshold voltage of the gated diode 600 is substantially controlled by the dopant concentration in the p-well 630.

  FIG. 12 shows an example of a side view of a second n-type gated diode 700 formed of SOI. The second n-type gated diode 700 includes a gate insulator 720 formed between a gate 715 (eg, N + doped polysilicon) and a p-well 730, a source diffusion region 710, and two STI regions 705. And 725, an insulator 735, a “drain” diffusion region 745, and an interconnect 750 that electrically couples the source diffusion region 710 and the “drain” diffusion region 745. A p-well 730 is formed on the well boundary 736. The threshold voltage of the gated diode 700 is substantially controlled by the dopant concentration in the p-well 730.

  FIG. 13 shows an example of a side view of a first p-type gate control diode 800 formed of SOI. The first p-type gated diode 800 includes a gate insulator 820 formed between a gate 815 (eg, P + doped polysilicon) and an n-well 830, a source diffusion region 810, and two STI regions 805. And 825, and an insulator 835. An n-well 830 is formed on the well boundary 836. The threshold voltage of the gated diode 800 is substantially controlled by the dopant concentration in the n-well 830.

  FIG. 14 shows an example of a side view of a second p-type gate control diode 900 formed of SOI. The second p-type gated diode 900 includes a gate insulator 920 formed between a gate 915 (eg, P + doped polysilicon) and an n-well 930, a source diffusion region 910, and two STI regions 905. And 925, an insulator 935, a “drain” diffusion region 945, and an interconnect 950 that electrically couples the source diffusion region 910 and the “drain” diffusion region 945. N-well 930 is formed on well boundary 936. The threshold voltage of the gated diode 900 is substantially controlled by the dopant concentration of the n-well 930.

Gated diode circuit This disclosure describes a memory cell that uses a gated diode. In order to understand the operation of the gated diode in the memory cell, this section shows and analyzes the gated diode circuit.

  Memory cells that use gated diodes have signal amplification that utilizes the charge stored in the gate channel inversion layer (eg, inversion layer 126/226). As described above, when a small voltage signal corresponding to a high voltage (eg, data 1) is present at the gate of the gated diode and the gate voltage is higher than the threshold voltage of the gated diode, charge accumulates at the gate. (Eg, via an inversion layer under the gate). When a signal corresponding to a low voltage (eg, data 0) is present at the gate and that voltage is below the threshold voltage of the gated diode, little or no charge is accumulated at the gate.

  Turning briefly to FIG. 16, a gated diode circuit 1100 is shown. The gated diode circuit 1100 is coupled to the signal line 1110 and further includes a gated diode 1130, the gate input (and hence the gate) of the gated diode 1130 is coupled to the signal line 1110 and its source input ( Thus, the source diffusion region) is coupled to the control line 1120. Signal line 1110 has a capacitance CL, which is the lumped capacitance from signal line 1110, the coupling capacitance of the gate, and the total capacitance of the connection circuit to the signal line (if capacitance exists). The capacitive load (CL) is not considered to be part of the gated diode circuit 1110. As shown in FIG. 16, the signal line 1110 is connected to the gate of the gate control diode 1130. The source of the gated diode 1130 is connected to the control line 1120, which is typically ground (GND) in the case of an n-type gated diode, or the supply voltage (in the case of a p-type gated diode). VDD).

  During signal amplification by the gated diode, the control line voltage (Vs) is generally increased. In accordance with the control line voltage, the source voltage (eg, source diffusion region) of the gated diode is also increased to a certain magnitude (denoted VB), generally ranging from 50 percent to 100 percent of the power supply voltage (VDD). However, it is higher for n-type and lower for p-type.

  Attention is now directed to FIG. 15, which shows a graph of amplifier gain when a capacitor is used as a charge storage device in amplifier 1110. In other words, the gated diode 1130 of FIG. 16 has been replaced with a conventional or linear capacitor (ie, a capacitor whose capacitance remains constant with voltage). The first graph Vs shows how the voltage on the control line 1120 changes. The second graph shows how point 1101 changes. As can be seen in FIG. 15, when a capacitor is used instead of the gated diode 1130, the gain of the circuit 1100 is about one. When the signal line 1110 has a high voltage, the output is VB plus a high voltage (data 1). When the signal line 1110 has a low voltage, the output is VB plus a low voltage (data 0). The difference dVin is all the differences existing between the data 1 voltage and the data 0 voltage. Therefore, the gain obtained by dividing dVout by dVin (data 1 voltage minus data 0 voltage) is about 1. That is, a gain of 1 means no voltage gain.

  Referring again to FIG. 16, during signal amplification, the gate control voltage is raised in the n-type case by an amount close to the source voltage increase minus a certain amount depending on the total capacitance CL, and p In the case of a mold, it is pulled down. Here, this total capacitance CL is the sum of the total stray capacitance of the gate coupled to the nearby device, the capacitance of the line, and the total capacitance of the connection circuit (if any). The actual magnitude of the resulting increased voltage output (eg, the output of signal line 1110) can be calculated.

  If the gate signal is data 0, then there is no or little charge stored in the gate (eg, position 1101), and the gate of the gate control diode 1130 is off (gate-source voltage (Vgs)). Therefore, when detecting data 0, there is little increase in the gate voltage of the gated diode, and the output voltage VL remains at 0 or is substantially smaller. This is shown in FIG. In FIG. 17, the gated diode 1130 has a very small capacitance, indicated by reference 1150. Even if Vs rises, the resulting output voltage of signal line 1110 is low. In other words, the voltage transfer between the control line 1120 and the gate (eg, point 1101) is small.

  On the other hand, if the gate signal (voltage measured from GND for n-type or voltage measured from VDD for p-type) is data 1 and is higher than the threshold voltage of the gated diode, the gated diode Is on and initially a substantial amount of charge is stored in the inversion layer of the gated diode. When Vs rises, the gated diode 1130 becomes less biased or turns off. In addition, the charge of the inversion layer is transferred to the total capacitance (CL) connected to the gate, including stray capacitance, line capacitance, and capacitance of nearby devices such as inverter or buffer or switch gates. Also, the source and gate can act as a capacitor. When the total capacitance (CL) is within a certain range compared to the on-capacitance (Cg_gd (on)) of the gated diode (shown in FIG. 18), a large voltage increase occurs at the gate of the gated diode 1130. . Thus, as shown in FIG. 18, when VL is greater than Vt, the gated diode can be regarded as a large capacitor, as indicated by reference 1160. In other words, the voltage transfer between the control line 1120 and the gate (eg, point 1101) is large.

  This creates a large voltage difference between the data 1 signal and the data 0 signal, typically 50 to 150 percent of VDD, at the gate of the gated diode. This is shown in FIG. Here, dVout is a large value. Depending on the signal amplitude and the increased voltage (VB), ie the ratio of the load capacitance (CL) and the gated diode on capacitance, the gain realized will vary and the total load (CL) of a given gated diode will be This gain can be calculated and characterized over a range. In general, the gated diode amplifier 1100 achieves a voltage gain of 2-10 during signal amplification. The output of the gated diode 1130 is itself a full CMOS voltage swing and can drive a typical small inverter buffer or latch.

The following analysis shows typical values for a gated diode amplifier as shown in FIG. Let Cg_gd (on) and Cg_gd (off) be the gate capacitance of the gated diode when the gated diode is on and off, respectively. And
Rc = Cg_gd (on) / CL, and
rc = Cg_gd (off) / CL.

In general operation, the value of the load capacitance (CL) is less than or as large as the on-capacitance of the gated diode (Cg_gd (on)), but CL is the same as that of the gated diode. It is much larger than the off capacitance (Cg_gd (off)). That is,
Cg_gd (on)> CL >> Cg_gd (off).

For example,
Cg_gd (off): CL: Cg_gd (on) = 1: 10: 20, and
Rc = 2, rc = 0.1.

  First consider a logic one signal that is higher than the threshold voltage. In this case, the gated diode is on.

  Assume that VL_high is a logic 1 voltage, VL_low is a logic 0 voltage, and is generally 0 (or ground) for an n-type gated diode.

When the control line Vs increases by a voltage of magnitude VB, the output voltage of the gate is:
Vout (1) = VL_High + VB Rc / (1 + Rc)
^~ VL_HIGH + VB, where Rc >> 1
Vout (0) = VL_Low + VB rc / (1 + rc)
^~ VL_low, where rc << 1.

dVin is the gate voltage difference between 0 and 1 before Vs increases, and dVout is the gate voltage difference between 0 and 1 after Vs increases. The output difference dVout is as follows.
dVout = VL_high + VB Rc / (1 + Rc)-(VB rc / (1 + rc) + VL_low).

The input difference dVin is as follows.
dVin = VL_High-VL_Low.

When VL_low = 0, the gain is:
Gain ^{= dVout / dVin ~ 1+ (VB} / VL_ high) Rc / (1 + Rc) > 1.

Consider the following example:
Example 1. In this case, VB = 0.8V, VL_high = 0.2V, and VL_low = 0.
then,
Using a gated diode, gain = 5, and using a linear capacitor, gain = 1.
Consider another example.
Example 2. In this case, VB = 0.8V, VL_high = 0.1V, and VL_low = 0.
then,
Using a gated diode, gain = 9, and
When using a linear capacitor, gain = 1.

When Rc is small (<1), it can be shown that the gain of gated diode signal amplification is given by:
Gain = 1 + Rc- (Vt_gd / VL_High) Rc ^~ 1 + Rc,
Here, Vt_gd is a threshold voltage of the gated diode.

  FIG. 20 shows an exemplary gated diode memory cell and its voltage for use in FIGS. In FIG. 20, Vg_f is the final voltage of the gate of the gated diode. It should be noted that Vg_i is the initial voltage of the gate of the gated diode.

  FIG. 21 is a table showing the complete charge transfer region and limited charge transfer region of the gated diode memory cell of FIG. Full charge transfer means that the gated diode provides all or all of its charge to the load CL. Limited charge transfer means that the gated diode provides only a portion of its charge to the load CL. FIG. 22 shows the voltage gain of the gated diode under different load ratios Rc. Here, as defined above, Rc = Cg_gd (on) / CL.

  When the signal line has a large capacitive load (CL) greater than the on-capacitance of the gated diode, the gain of the basic gated diode amplifier begins to drop and eventually the gain becomes unity (ie gain) Is gone). In addition, large capacitive loads slow down the amplifier.

For example, the following table shows the gain for different ratios of gated diode on-capacitance (Cg_gd (on)) to load capacitance (CL). Assume the following.
VB = 0.8 V,
VL_ high = 0.2 V,
VL_low = 0, and
rc = 0.1.

  And the table is as follows.

ここで、Rc=Cg_gd（オン）/CL、dVout=VL_ハイ+VBRc/(1+Rc)、dVin=VL_ハイ、および利得=dVout/dVin。

Here, Rc = Cg_gd (ON) / CL, dVout = VL_high + VBRc / (1 + Rc), dVin = VL_high, and gain = dVout / dVin.

  Further details on the gated diode circuit of FIG. 16 can be found in U.S. Patent Application No. Luk et al. Entitled “AMPLIFIFIERS USING GATED DIODE” filed on the same day as this application. It can be found in No. XX / XXX, XXX. As such, the disclosure of this application is incorporated by reference.

3T1D Memory Cell Three transistors (T), one diode (D), gated diode memory cell (3T1D) are composed of one gated diode and three FETs. One FET is used to write the data value to the memory cell, and the other FET acts as a write gate and a read gate, each of which is on the write bit line or read bit line to perform 2-port operation. Connected. The two bit lines can be combined into a single bit line that is shared for reading and writing. Gated diode memory cells and silicon structures will become a future generation silicon technology that operates at low bitline voltages (eg, 0.3-1.0V) in bulk silicon, SOI, and dual gate finFETs Can be applied.

  FIG. 23 shows a 3T1D memory cell 1300, which in this case is a dual port memory cell. The 3T1D memory cell 1300 includes a gated diode (gd) 1330 and three FETs: a write gate (wg) FET 1325, a read gate (rg) FET 1345, and a read select (rs) FET 1340. The 3T1D memory cell 1300 is coupled to a write bit line (BLw) 1305, a read bit line (BLr) 1310, a read word line (WLr) 1335, and a write word line 1320, the read bit line 1310 having a capacitance Cbl 1315. In this example, the gate of the gate control diode 1330 is a storage node, and charges are accumulated in the inversion layer after a high voltage corresponding to data 1 is applied. Illustratively, if data 0 is added, no charge is accumulated. The source of gated diode 1330 connects to read word line (WLr) 1335 for voltage increase during a write or read operation. Details of the method of increasing the voltage of the gated diode in read and write operations that amplify the cell voltage are described in detail below. The voltage increase is important for signal increase and operation of the 3T1D gated diode memory cell 1300. It should be noted that FETs 1325, 1340 and 1345 operate as switches. For write FET (wg) 1325, the gate is connected to write word line (WLw) 1320, the drain is connected to bit line (BLw) 1305, and the source is connected to the gate of gate control diode 1330. When the write word line (WLw) 1320 is high, the cell is selected and the storage node (eg, the gate of the gated diode 1330) is written with the bit line voltage. For the read gate FET (rg) 1345, the gate connects to the storage node to read stored data 0 or data 1. A read select FET (rs) 1340 is enabled by a read word line (WLrs) 1350 during a read operation, and the drain of the read gate (rg) 1345 is used to detect the storage node voltage (eg, Vcell). The read bit line (BLr) 1310 is connected.

  In some situations, both WLr 1335 and WLrs 1350 may be the same control signal. The read select FET 1340 can be enabled using an increase in the source voltage of the gated diode in a read voltage increase mode of operation that will be described in detail later. However, in general, a negative voltage may be applied to the gate of the unselected read select FET 1340 to reduce subthreshold leakage current. This may require WLrs 1350 to be different from WLr 1335. Since the wiring direction is not limited as in the bit line direction, it is feasible to provide more than one word line, such as WLw 1320, WLrs 1350 and WLr 1335. Also, by using separate word lines, the load can be distributed and heavy loads can be reduced.

  In FIG. 23, a 3T1D gated diode memory cell 1300 has two ports for reading and writing. Read word line (WLr) 1335 is connected to the source of gated diode 1330. Another word line (WLrs) 1350 is connected to the gate of the read select FET (rs) 1340 to allow read operation and voltage increase of the storage cell during read. The storage cell voltage can be increased much higher than the stored voltage.

  In FIG. 23, WLr may be an additional write word line (WLw2) (ie, WLr is named WLw2, WLw2 is not connected to WLrs 1350), which is connected to the source of gated diode 1330. This arrangement supports a write voltage increase operation, which uses the write word line (WLw2) during the write operation to increase the storage node voltage much greater than the bit line high voltage (VBLH). Is done. The bit line high voltage (VBLH) corresponds to the voltage for writing data 1 to the memory cell. Moreover, in this configuration (described in more detail below with respect to FIG. 34), WLrs 1350 is generally implemented as a separate control line for read select FET 1340.

  FIG. 24 shows a 3T1D gated diode memory cell 1400 with a single port, where the read and write bitlines are shared by a single bitline 1410.

  Exemplary voltages for the memory cells, wordlines and bitlines of the 3T1D gated diode memory cell 1300 are shown in FIG. 25 for one exemplary manufacturing technique. Although the bit line voltage is less than the technology's power supply voltage (VDD), the inherent gain of the gated diode memory cell enabled by the voltage increase method of read and write operations allows the memory cell to read and write It operates with a signal voltage higher than the bit line voltage during operation, and achieves a better signal-to-noise margin. A typical bit line voltage is 50% of the power supply voltage (VDD), resulting in substantial active power savings compared to conventional DRAM and SRAM.

Voltage Increase and Voltage Gain Next, the gated diode 1330 can be written with a small voltage from the bit line through the write gate (wg) 1325, where the small voltage is the voltage at the source of the gated diode. By raising (called voltage “increase”), it is generally amplified 2 to 3 times. Signal amplification can be performed during either a write operation or a read operation as follows.
Write voltage increase: If the original data 1 voltage was amplified with a voltage increase during the write cycle, the gate of the gated diode 1330 was increased higher for subsequent reading by the read gate (rg) 1345. A voltage is stored in the gated diode 1330 after the write operation. The higher amplified gate voltage results in a larger signal margin and larger gate overdrive of the read gate 1345 during the read operation. For data 0, there is little signal amplification or voltage increase and the gate voltage remains almost zero.
Read voltage increase: If a lower voltage of the original data 1 is stored in the gated diode 1330 during the write operation, the lower voltage is increased higher during the read operation. This results in signal amplification and voltage gain of the original data 1 signal, and thus a larger signal margin and larger gate overdrive of the read gate 1345 during the read operation. In the case of data 0, there is almost no signal amplification due to voltage increase, and the gate voltage remains almost zero.

  The increase in the voltage of the gated diode memory cell 1300 due to either writing or reading causes the signal to noise margin of the memory cell 1300 and the separation margin of data 0 and data 1 below or above the threshold of the read gate 1345. Become very large. This is particularly important for future generations of technology. This is because when small signals are used, the margin is very small due to threshold voltage variations and mismatches. Although higher external signal voltages can be applied (eg, via the bit line voltage), this requires more active power. The 3T1D memory cell 1300 does not require a high bit line voltage to create a high cell signal voltage to obtain a better signal to noise margin. Instead, this higher signal voltage is created by the voltage increase of the gated diode, either in a write operation or a read operation.

In the gated diode memory cell 1300:
Let Cg_gd (on) and Cg_gd (off) be the gate capacitance of the gated diode when the gated diode is on and off, respectively,
Let CL be the total load capacitance connected to the gate of the gated diode (gd). Where CL is generally the gate capacitance of the read gate (rg) 1345 and the stray capacitance of the gate of the gated diode 1330 relative to nearby objects.
then,
Rc = Cg_gd (on) / CL,
rc = Cg_gd (off) / CL.

In a typical situation, the load capacitance (CL) is less than the on-capacitance (Cg_gd (on)) of the gated diode. However, CL is much larger than the off capacitance of the gated diode (Cg_gd (off)).
Cg_gd (on)> CL >> Cg_gd (off).

For example,
Cg_gd (off): CL: Cg_gd (on) = 1: 10: 20,
Rc = 2, rc = 0.1.

Cg_gd (on)> CL,
Vt_gd = 0,
Vt_rd = 0.2V
Vcell_i = 0.4V (initial cell voltage)
If VB = 0.8 V (the magnitude of the increase voltage, Vs = 0 to> 0.8 V), the charge accumulated in the gate control diode is given by the following equation.
Q_accumulation = (Vcell_i-Vt_gd) Cg_gd.

When the voltage at the source of the gated diode is raised, some charge is transferred from the inversion layer of the gated diode to the load CL. The final gate voltage Vcell_f is
Vcell_f> VB + Vt_gd,
The charge for charging CL to VB + Vt_gd is given by the following equation.
Q_Transfer 1 = (VB + Vt_gd-Vt_rg) CL.

The charge to charge both Cg_gd + CL higher than VB + Vt_gd is given by:
Q_Transfer 2 = Q_Store-Q_Transfer 1
= (Vcell_i-Vt_gd) Cg_gd- (VB + Vt_gd-Vt_rg) CL,
= Vcell_i Cg_gd-VB CL + Vt_rg CL-Vt_gd (Cg_gd + CL),
del_V1 = VB + Vt_gd-Vt_rg,
del_V2 = Q_Transfer 2 / (Cg_gd + CL),
= [(Vcell_i-Vt_gd) Cg_gd- (VB + Vt_gd-Vt_rg) CL] / (Cg_gd + CL),
= Vcell_i Rc / (1 + Rc) -VB / (1 + Rc) + Vt_rg / (1 + Rc) -Vt_gd,
Vcell_f = Vt_rg + del_V1 + del_V2,
Vcell_f = (VB + Vcell_i) Rc / (1 + Rc) + Vt_rg / (1 + Rc), (2)
(Vcell_i> Vt_gd, Rc> 1, large Rc)
Gain = Vcell_f / Vcell_i ^to (1 + VB / Vcell_i) Rc / (1 + Rc). (3)

For small Rc <1,
Gain = 1 + Rc
It can be shown that.

If you put a general value,
Vcell_i = 0.4V (bit line voltage VBLH),
VB = 1V (VDD),
Vt_gd = 0,
Vt_rg = 0.2V (Vt of readout gate),
Rc = 10.
Vcell_f = (1 + 0.4) (10) / (1 + 10) + 0.2 / (1 + 10) = 1.29V
Gain = 1.29 / 0.4 = 3.23.

  The following table shows the gain as a function of Rc, Vs, and Vg_i.

If you put a representative value,
Vcell_i = 0.4 (bit line voltage VBLH),
VB = 0.8V (VDD),
Vt_gd = 0,
Vt_rg = 0.2V (RVt, readout gate),
Rc = 10,
Vcell_f = (0.8 + 0.4) (10) / (1 + 10) + 0.2 / (1 + 10) = 1.11V
Gain = 1.11 / 0.4 = 2.78.

  This is shown in the following gain table as a function of Rc, Vs, and Vg_i.

  When the source voltage is increased, the gated diode 1330 partially transfers some charge from the storage cell to a nearby circuit (in this case, the read gate 1345), and a conventional DRAM cell using a capacitive cell and A very large signal is realized compared to the gain cell. The voltage gain is always greater than 1, and generally a gain between 2 and 10 can actually be realized. Indeed, the 3T1D memory cell 1300 achieves a voltage gain that is several times the initial stored voltage from the bit line, but there is no voltage gain in the case of conventional DRAM cells and gain cells. In gain cells, the voltage gain is unity, but in DRAM, cell charge and voltage are lost due to charge sharing and recovered after a read operation is required. In the present invention, this gain of cell voltage can be utilized either during a write operation or a read operation. “Dual gain” is achieved in both the storage cell (voltage gain) and the sensing read gate (current gain) compared to just a single gain in the conventional gate cell (due to the read gate).

Retention Time, Leakage and Capacitance Ratio In order to improve the retention time of the gated diode memory cell 1300, two major leakage currents must be addressed. That is,
A subthreshold voltage current flowing through the source of the write gate 1320 connected to the gate of the gated diode 1330, and a leakage current due to tunneling at the gate of the gated diode 1330.

  Since the subthreshold current is minimized by applying a negative voltage to the write word line (WLw) 1320, all the write gates of the unselected cells have a negative gate voltage and thus a negative gate-source voltage (Vgs). . Further, increasing the threshold voltage of the FET 1325 also reduces subthreshold leakage at about 100 mV per decade of threshold voltage.

  By using a high K gate dielectric with increasing gate insulator thickness, tunneling current through the gate can be reduced. A typical gate oxide thickness of 25 Angstroms may be sufficient to limit the gate tunnel current. As an example, above 90 nanometer (nm) technology, a high K is required to achieve sufficient on-capacitance of gated diode 1330 while maintaining sufficient oxide thickness to keep tunnel current low. Gate dielectric may be required.

  Considering current technology and considering future technology scaling, the total capacitance of the gated diode 1330 and the read gate 1340 is typically approximately 1 to 2 femtofarads (fF). In future generations, the feature size will be smaller, but the gate oxide thickness will be scaled thinner, so the net result is that the gate capacitance is still approximately constant. Up to a point where the gate oxide thickness can no longer be reduced proportionally with scaling, a high K dielectric is an option to keep the capacitance values approximately the same. As such, the design value of the capacitance of the gated diode 1330 is considered to be substantially constant over a period of time. Much smaller capacitance values can be used as long as it is recognized that the ratio of the capacitance of gated diode 1330 and read gate 1345 is given by equations (1), (2), (3) and (4) shown above. It should be pointed out that higher values of the gated diode capacitance are advantageous for stabilizing the cell voltage due to leakage and soft errors, but the larger the capacitance value, the larger Cell area is required. So 2fF to 8fF is a good range for the total capacitance of the gated diode and read gate.

  If it is desirable to use a larger total capacitance for cell voltage stabilization to improve retention and reduce soft error, it is better to use a smaller capacitance ratio Rc. Here, Rc = Cg_gd / CL. This is because the greater the gated diode on capacitance, the more the gated diode requires a larger wordline (WLr) driver to drive. In such cases, using a smaller capacitance ratio provides a good balance between the large total capacitance and the ease of driving the word line. In general, the on-capacitance of the gated diode should be less than 4 pF. For writing and reading data 0, gated diode 1330 does not accumulate charge, so there is no active current that needs to be supplied to the source and gate of gated diode 1330. This is an important advantage of the gated diode 1330 to save active power and minimize bit line and word line voltage fluctuations. Because, on average, 50 percent of data reads and writes are zero.

Retention Time and Write Gate Subthreshold Leakage Control The retention time of the gated diode memory cell 1300 depends on the total leakage current from the gate of the gated diode 1330 to the silicon body and to the drain through the gate tunnel and the write gate. 1325 source to drain and drain to subthreshold leakage current through the body. The cell hold time (T_hold) is determined by the time it loses a certain amount of charge, since the voltage is discharged by the total leakage current (I_leakage) when the cell is storing a high voltage of data 1. . If the holding time is defined as losing some amount of high voltage V_ drop, eg losing 10 percent of the stored voltage,
T_hold = V_decrease Cg_gd (on) / I_leakage.

In general, if T_hold is set to 10 μs (microseconds),
V_decrease = 0.1 V, Cg_gd (on) = 2 fF, and
I_leakage <V_decrease Cg_gd (on) / T_hold
= 0.1 (2e-15) / 10e-6 = 2e-11 = 20 pA.

  In order to achieve such a small leakage current in relation to current silicon technology, the threshold voltage of the write gate needs to be approximately 3-4 decades larger than the standard threshold voltage device. This is equivalent to applying a negative voltage of approximately -0.4 V to -0.3 V to the gate of the write gate of the unselected memory cell for a threshold voltage ramp of approximately 100 mV / decade. Furthermore, the gate oxide thickness must be greater than 25 mm in order to reduce the gate tunnel current below this leakage current requirement.

Method of reading and voltage increase When a voltage corresponding to data 1 is present at the gate of gated diode 1330 and this voltage is higher than the threshold voltage of gated diode 1330, charge is accumulated in the inversion layer of the gate channel. The When a voltage corresponding to data 0 is present at the gate and the voltage is lower than the threshold voltage of the gated diode 1330, no or little charge is accumulated at the gate. The capacitance characteristics of the gated diode with respect to the gate-source voltage (Vgs) are shown in FIGS. As explained above, the capacitance is negligible when Vgs is below the threshold voltage, and when Vgs exceeds the threshold voltage, the capacitance increases linearly with Vgs and reaches a maximum value above a certain value of Vgs. Calm down. When Vgs exceeds the threshold voltage, a certain amount of charge is stored in the gated diode, which acts as a capacitor. When Vgs is below the threshold voltage, the gated diode 1330 has negligible capacitance.

Referring again to FIGS. 16, 17 and 18, in the case of a read operation, the voltage on read word line (WLr) 1335 increases by a magnitude VB. When a voltage corresponding to data 0 is applied to the gate of the gated diode 1330, the voltage is lower than the threshold voltage, so that the charge stored in the gated diode is zero or negligible, The capacitance between is the peripheral and overlap capacitance (Cg_gd (off)) between the gate and source. This capacitance is very small compared to the gate on capacitance (Cg_gd (on)) and the ambient stray capacitance (C_stray). As WLr rises, Cg_gd (off) is very small and the coupling effect is very small, so the voltage increase at the gate of the gated diode is very slight. The data 0 coupling effect results from the voltage divider formed by the gated diode capacitance (eg, almost zero) and the load capacitance (CL), with the capacitance CL being greater, eg, 10: 1. Capacitance CL is comprised of the stray capacitance of the gate of gated diode 1330 and the capacitance of the connected lines and devices, in this case the gate capacitance of read gate 1345. Therefore, the voltage increase of the gate for reading data 0 is very small, about VB / 10. When data 1 is stored in the memory cell 1300, a significant amount of charge (Q_accumulation) is accumulated in the gated diode 1330 and the capacitance across the gated diode (Cg_gd (on)) is large. When the voltage of WLr 1335 rises, this voltage is the same voltage as the source voltage of the gated diode, but the gate voltage (Vg_gd) is increased as:
Vcell_f = VB cc + Vcell_i, cc = Cg_gd / (Cg_gd + CL),
Where cc is the coupling coefficient of the voltage divider formed with respect to the gate of the gated diode 1330 by the gated diode capacitance and the connection node capacitance, and Vcell_i is the initial cell voltage of the gate of the gated diode ( Vcell), and Vcell_f is a cell voltage after the voltage of WLr 1335 increases (referred to as “voltage increase”).

For example,
Cg_gd (off): CL: Cg_gd (on) = 1: 10: 100
VB = 0.8V, VBLH = 0.4V,
The following table shows the voltage gain of the gated diode when reading data 0 and data 1, where gain = (1.13-0.08) / (0.4-0) = 2.63. is there.

  This causes a large voltage difference between reading data 0 and reading data 1. In fact, in the above example, a memory cell 1300 produces a voltage gain of about 2.6. When the gated diode is replaced with a capacitor, the read voltage of 0 and the read voltage of 1 are 0.8V and 1.2V, respectively, and there is no voltage gain (ie, gain = 1).

  FIG. 26 shows a memory portion 1700 comprising a 3T1D memory cell 1300 and a sense amplifier unit 1710. The sense amplifier unit 1710 detects and amplifies the signal on the read bit line (Blr) 1310 during a read operation and outputs it to the global bit line 1711. During the write operation, the sense amplifier unit sends the signal on the global bit line 1711 to the write bit line (Blw) 1305 for later writing to the gated diode storage cell 1330. FIG. 27 shows some example waveforms using the memory portion 1700. These figures are used in the following description.

  FIG. 27 illustrates a method having a read operation of a 3T1D memory cell 1300 with a voltage increase that amplifies the cell voltage during the read operation. FIG. 26 shows a memory cell 1300 having an n-type gated diode 1330. Although a two-port configuration having a read bit line (BLr) 1310 and a write bit line (BLw) 1305 is shown, the method shown in FIG. 27 is also applicable to a memory cell 1300 having a single port or more than two ports. There is one write word line (WLw) 1320 and one read word line (WLr) 1335 connected to the memory cell 1300. Memory cell 1300 is a 3T1D memory cell, which includes a gated diode (gd) 1330, a read gate (rg) 1345, a write gate (wg) 1325, as described in detail above. And a read selection FET (rs) 1340. The bit line is connected to an input / output sense amplifier 1710, which is generally outside the memory cell. Read wordline 1310 and write wordline 1305 are typically driven by a wordline driver (not shown) that is external to the memory cell.

  The voltage on the write word line (WLw) 1320 is a control signal that activates the write cell 1330. The voltage on the word line 1320 is generally between 0 volts (V) (eg, GND) and the power supply voltage (eg, VDD), or preferably to reduce the subthreshold leakage current of the unselected cell. In addition, a negative voltage is used instead of 0 volts. When WLw 1320 is high, the cell 1330 connected to it is activated for writing. The voltage of the read word line (WLr) 1335 is a control signal connected to the source of the gated diode (gd) 1330 for the read operation. Read word line WLr 1335 is used to increase the voltage of gated diode 1330 as described in detail above. The voltage on the read word line is typically between 0 volts and VB, the magnitude of the voltage increase. The low voltage of WLr1335 is slightly negative to increase the signal written to the cell, and can bias the threshold voltage of the gated diode. The voltage of the write bit line (BLw) 1305 is a signal line for transmitting data 0 (generally a low voltage such as 0 V or ground) or data 1 (generally a high voltage such as VBLH) to be written into the cell. The voltage of the read bit line (BLr) 1310 is a signal line for transmitting a signal generated by reading data stored in the cell. Read bitline 1310 is typically precharged to a bitline high voltage (VBLH) and is connected to a sense amplifier for signal detection and output.

  WLr 1335 can be used to control the gate of the read select FET (rs) 1340 to enable the FET 1340 for the bit line 1310 during read. It is also desirable to add another optional control line (WLrs) 1350 without using WLr to control the source of the gated diode 1330 to control the gate of the read select FET (rs) 1340. When WLr 1350 is high, memory cell 1300 is coupled to a read bit line. This is what WLr 1335 does. WLrs 1350 has a negative low voltage to apply to the gate of FET 1340 when no row is selected, which reduces the active subthreshold leakage current when the array is active, No row is selected.

  These exemplary methods and operations of FIG. 27 include an n-type gated diode memory cell 1300 and the read bit line is precharged high (VBLH). This method applies correspondingly to p-type gated diodes and the read bit line is precharged low (GND) and operates in a complementary manner. Details of p-type operation can be determined by one skilled in the art.

  The method of increasing the read voltage is illustrated in FIG. 27 to operate the 3T1D memory cell 1300. In the right part of FIG. 27, there are six voltages over the time shown to explain the operation. The first graph is the voltage on the write word line (WLr) 1320. The second graph is the voltage on the read word line (WLr1335 or WLrs1350). The third graph is the voltage of the memory cell (Vcell). The fourth graph is the voltage of the write bit line (BLw) 1305. The fifth graph is the voltage of the read bit line (BLr) 1310. The sixth graph is the voltage at the sense amplifier output 1711. Both data 0 and data 1 are shown and the writing and reading of data 0 or data 1 is illustrated.

  First, it shows a “write” operation (not increased in voltage). The write word line is activated high and the write bit line data (either data 1 or data 0) is written to the gated diode 1330 via the write gate. In the unselected cell, the gate is connected to a negative voltage at the other WLw 1320 to reduce the subthreshold leakage current.

  For read word line WLr 1335, a small negative voltage can be applied to the read word line as an option during a write operation. This small negative voltage has the effect of lowering the threshold voltage of the gated diode whose source is connected to the read word line 1335 and increases the signal or charge written to the cell.

  Second, a “read” operation is shown. The read word line (WLr) 1335 connected to the source of the gated diode 1330 is activated. As explained in detail above, the voltage is pulled higher by VB to increase the voltage across the gated diode 1330. When cell 1300 stores data 0, the cell voltage at the gate of gated diode 1330 is GND. Since the gate control diode 1330 has almost no charge, the cell voltage remains at GND. The cell voltage at GND appears at the gate of read gate (rg) 1345. Since the cell voltage is lower than the threshold voltage of the read gate 1345, as a result, the read gate is still off and the current through the read gate 1345 is negligible. The bit line voltage remains unchanged and remains the precharged voltage (VBLH) high. If the cell is storing data 1, the cell voltage is VBLH before the voltage increase is utilized. After the voltage increase is utilized, the gate voltage of gated diode 1330 increases by “sVB”. Here, s is a coupling coefficient as described above in the operation of the gated diode. With this one-signal readout of magnitude (VBLH + sVB), a large voltage gain is realized and the readout gate (rg) 1345 is strongly turned on. Since the FET (rs) 1340 is on, a sufficiently large current for discharging the bit line 1310 is generated. The voltage on the bit line 1310 moves towards ground (GND) and its time constant is determined by the RC of the bit line 1310 and the read gate current. When reading data 0, the read gate remains off and the bit line remains high. After a bit line signal corresponding to either data 0 (eg, the bit line voltage remains high) or data 1 (eg, the bit line voltage goes low) is generated on the bit line 1310, the sense amplifier Is activated and outputs a corresponding signal on the global bit line 1711. A single-ended sense amplifier such as a single-ended gated diode sense amplifier can be used.

Write Operation and Voltage Increase Method FIG. 28 illustrates a 3T1D memory cell 1300 method and write operation with voltage increase that amplifies the cell voltage during the write operation. FIG. 26 shows a memory cell having an n-type gated diode 1330. Details of memory cell signals and control have been described above.

  The voltage of the write word line (WLw) 1320 is a control signal that activates the write cell. The voltage is between 0 volts (GND) and the power supply voltage (VDD) or preferably a negative voltage is used instead of 0 volts to reduce the subthreshold leakage current of the unselected cell 1300. Is done. When WLw 1320 is high, the cell 1300 connected to it is activated for writing. The voltage of the write bit line (BLw) 1305 is a signal line for transmitting data 0 or data 1 to be written to the cell. The voltage of the read bit line (BLr) 1310 is a signal line for transmitting a signal generated by reading data stored in the cell. The read bit line (BLr) 1310 is precharged to the bit line high voltage (VBLH) and is connected to a sense amplifier for signal detection and output. WLr 1335 can be used to control the gate of read select FET (rs) 1340 to couple FET 1340 to bit line 1310 during read. As described above, another optional control line (WLrs) 1350 is also used to control the gate of the read select FET (rs) 1340 without using WLr 1335 which controls the source of the gated diode 1330. It is desirable to add. When WLrs 1350 is high, the memory cell 1300 is coupled to the read bit line 1310 as WLr 1335 does. WLr 1350 has a negative low voltage that is typically applied to the gate of FET 1340 when no row is selected. This is because active subthreshold leakage current is reduced when the array is active. However, that particular row is not selected.

  In the method of the write operation having the voltage increase (FIG. 28), the use of the control signal is slightly different from the read operation (FIG. 27). Instead of using WLr, a second additional write word line (WLw2) is connected to drive the source of the gate diode (eg, WLw2 is the same line as WLr1335, but WLw2 is WLr) Used in a different way). This write word line WLw2 is used to increase and amplify the voltage stored in the gated diode memory cell 1300 so that a much larger signal is stored in the cell. A larger stored voltage improves voltage stability and provides longer data retention characteristics and resistance to soft errors due to radiation.

  This exemplary method and operation of FIG. 28 includes an n-type gated diode memory cell and the read bit line 1310 is precharged high (VBLH). This method applies correspondingly to p-type gated diodes and the read bit line is precharged low (GND) and operates in a complementary manner. Such an operation can naturally be determined by those skilled in the art.

  A method for increasing the write voltage for operating the 3T1D memory cell 1300 has been described. During a write operation, the write word line WLw21335 drops by a magnitude VB at the beginning of the cycle. A voltage corresponding to data 0 or 1 of the write bit line (WLw) 1320 is written to the gate of the gate control diode 1330. The voltage written to the cell is either 0 volts for data 0 or VBLH for data 1. At that time, the write word line WLw2 rises higher than the magnitude VB, so that the voltage of the write word line WLw2 returns to the cycle before writing. As a result of this increase in WLw2, the voltage stored in the cell at the gate of the gated diode 1330 is increased by a magnitude “sVB” more than. The final voltage of the cell is approximately (VBLH + sVB). When the gate voltage of the gated diode rises, which is the same as the source voltage of the write gate, the write gate 1320 is automatically disconnected. This is because the source voltage is higher than the gate voltage plus the threshold voltage. Since the write gate is off, such a voltage rise will not cause an active current or disturbance in the bit line 1305. If the data is 0, the gated diode memory cell 1300 has little charge, the cell voltage remains at GND, and the write gate 1320 remains on. There is also no active current flow through the write gate, which is advantageous for low power operation.

  After the write operation, the voltage stored in the gated diode memory cell 1300 is much greater than the original voltage (VBLH) written to the cell at the beginning of the cycle. The voltage gain mechanism and gain magnitude have been described in detail above. The bit line active current and power are much smaller than that by applying a higher voltage to the bit line and when such a voltage is used to write to the memory cell 1300 (in general, (About 50%) is the advantage of this voltage increase during the write operation. Using this method and cell structure, voltage gain is achieved during a write operation, which distinguishes the other DRAM cells, SRAM cells and gain cells from this memory cell 1300 and how it is used. The voltage written to the cell is higher than the bit line voltage (VBLH) and is generally about twice or more, thereby enhancing cell voltage stability, data retention, and resistance to soft errors due to radiation.

  During the read operation, when the read word line WLr 1340 is activated, the read gate 1345 is strongly turned on by the large voltage of the data 1. When the read select FET (rs) 1340 is turned on, a sufficiently large current is generated to discharge the bit line 1310. The bit line voltage moves towards ground (GND) and its time constant is determined by the RC of the bit line 1310 and the read gate current. In reading data 0, the read gate remains off and the bit line remains high. After a bit line signal corresponding to either data 0 (eg, bit line voltage remains high) or data 1 (eg, bit line voltage goes low) is generated on bit line 1310, The sense amplifier is activated and outputs a corresponding signal on the global bit line 1711.

3T1D memory cell structure Gated diode planarization As explained in detail earlier in this disclosure, a gated diode memory cell is the simplest form of a FET "half" with only the gate and source connected. Can be realized. For a planar gated diode, a very small Vt or zero Vt device is advantageous (Vt_gd˜0) to get the data 1 signal high to the cell when data 1 is written to the cell. . Realization of low level dopant n-type FETs or depleted n-type FETs can be chosen as gated diodes. FIG. 29 shows a cross section of an exemplary 3T1D memory cell 1900 implemented in bulk silicon using an n-type gated diode.

  The 3T1D memory cell 1900 includes one or more bit lines 1905, ground 1910, two ground metal contacts 1911 and 1912, and a global ground line runs parallel to the bit line every 2 to 16 cells. (In the figure, the ground line is behind the bit line), read bit line (BLr) 1915 and its metal contacts 1916 and 1917, write word line (WLw) 1920 and its metal contact 1921, write Bit line (BLw) 1925 and its metal contacts 1926 and 1927, drain diffusion region 1975, gate 1978, gate insulator 1983, source diffusion region 1970, MCBar metal contact 1991, shallow trench isolation (STI) region 1965, polysilicon Down gate 1994, and a read word line (WLr) 1945, STI regions 1950 and 1953, p-well 1992, optional n separation zone 1955, p substrate 1960 and insulating layer 1930, which has metal contacts 1933 and 1940. The gated diode 1904 is formed of a gate 1904, a gate insulator 1990, and a source (not shown). The source of gated diode 1904 is coupled to WLr 1945 through metal contact 1933. Portion 1992 separates gate insulator 1990 and gate insulator 1993. In general, portion 1992 may be filled with polysilicon from gate 1994.

  The write FET 1901 is formed by a gate 1978, a gate insulator 1983, and source / drain diffusion regions 1970 and 1975, and is connected to the WLw line 1920 via a metal contact 1921. The read select FET 1902 is formed before the gate 1994, while the read FET 1903 is formed behind the gate 1994. The source (not shown) of the read selection FET 1902 is connected to the drain (not shown) of the read FET 1903 (shown after the gate 1994). The drain (not shown) of the read select FET 1902 is in front of the gate 1904, and the FET channel (not shown) of the read select FET 1902 enters toward the page. The source (not shown) of the read FET 1903 is connected to the GND line 1910 (behind the gate 1994) through a metal contact 1912. The gate (not shown) of the read FET 1903 is behind the gate 1994, and its FET channel (not shown) enters toward the page. MCBar 1991 connects the source diffusion region 1970 to the gate of the read FET 1903. Gate 1983, gate 1994, and gates of read FET 1903 and read select FET (not shown) are generally made of heavily doped (eg, N +) polysilicon. In this example, the write line WLr 1933 and the control line WLrs 1940 are combined.

  As shown in FIG. 29, a source diffusion region 1975 is formed as part of the gated diode memory cell write gate and may be shared between the write gates of the two gated diode memory cells. (Eg, one write gate of a gated diode memory cell is formed by gate 1978, gate insulator 1983, source diffusion region 1975, etc., and the other write gate of another gated diode memory cell is The gate is formed by gate 2521, gate insulator 2583, source diffusion region 1975, write word line 2520, metal contact 2527, and drain diffusion region 2570). This structure allows two adjacent gated diode memory cells implemented next to each other to share a source diffusion region 1975 that may be coupled to the same bit line (BLw). Further, the silicon area is reduced.

  FIG. 30 shows a cross section of an exemplary 3T1D memory cell implemented in SOI using an n-type gated diode. The 3T1D memory cell 2000 comprises the structure shown above in FIG. 30, but is formed on an insulator 2060.

2. Gated Diode Trench Realization FIG. 31 illustrates an exemplary 3T1D gated diode memory cell 2100 configured with the structure shown above in FIG. In addition, the gated diode memory cell 2100 has a gated diode 1904 implemented in the form of a shallow trench, the gate of which is a cylindrical shape surrounded by a thin oxide 2190 and separated by silicon 1950. A polysilicon trench 2110 is formed. It should be understood that the polysilicon trench can have any cross-sectional shape and is not limited to the illustrated cylindrical realization. The region on the silicon surface next to the polysilicon trench is more heavily doped to form the source diffusion of the gated diode, with the polysilicon trench being the gate. This implementation creates a relatively small surface area for the gated diode and produces a relatively large capacitance (Cg_gd), deep for data retention and improved protection against soft error rate (SER). However, it has the advantage that it can operate in a limited charge transfer region (Rc> 10-100). This requires technological development in addition to planar technology, and additional processing steps for embedded memory in standard planar silicon technology.

  FIG. 32 shows a side cross-section of the trench gated diode structure used in FIG. The other gate is formed by a cylindrical polysilicon trench 2210 surrounded by a thin oxide 2290 and separated by silicon 1950. Source diffusion region 2220 is formed as part of the gated diode and is shared by the two gated diodes (one is gate 1994, gate insulator 2190 and 1990, polysilicon trench 2110, and source Formed by diffusion region 2220, the other being formed by gate 2294, gate insulators 2290 and 2291, polysilicon trench 2210, and source diffusion 2220). STI regions 2230 and 2240 isolate the gated diode from other devices on the semiconductor. In this example, the “gate” includes a cap portion 1994 and a trench portion 2110. The structure of FIG. 32 allows two adjacent gated diodes implemented adjacent to each other, sharing a source diffusion region that may be coupled to the same word line (WLr), and silicon The area is reduced.

  29-32, the bit line and the GND line run parallel to the second metal (M2) of the common technology. The word line runs orthogonal to the bit line of the first metal (M1) of common technology. Either a single port memory cell where a single bit line is used for both reading and writing, or a dual port memory cell can be implemented with the structure shown. The metal lines M1 and M2 and metal contacts shown are exemplary arrangements, and the underlying silicon structures of the three transistors and gated diodes shown are planar bulk and SOI silicon technology, 3T1D of trench silicon technology. Those skilled in the art can conceive other structures of 3T1D, such as a p-type device or an equivalent structure of an array, to explain the basic operating principle of a memory cell.

3T1D gated diode memory array FIG. 33 shows a plurality of gated diode memory cells 2310-11 to 2310-NM formed by arranging gated diode memory cells in a two-dimensional array and writing The word line (coupled to drivers 2320-1A to 2320-NA) and the read word line (coupled to drivers 2320-1B to 2320-NB) run in one direction, eg, parallel to the horizontal, and Write bit lines 2330-1 to 2330-M, read bit lines 2340-1 to 2340-M, and ground (GND) lines 2350-1 to 2350-M / 2 run orthogonally, for example, in the vertical direction. The intersection of the bit line and the word line defines the location of the memory cell, ie the bit of data storage. In addition, write bitline drivers 2350-1 to 2350-M and read bitline sense amplifiers 2360-1 and 2360-M are used. It should be noted that since read bit line 2340 is generally precharged during a read operation, “BLPC” in the figure means bit line precharge and “SA” is a read bit line ( BLr) represents a “sense amplifier” used to detect the signal.

  FIG. 34 shows the same structure when write voltage increase is used. One additional word line (WLw2) 2420-1C to 2420-NC connecting to the source of the gated diode is added for each row to achieve a voltage increase during the write operation. Under this method of operation, the word line control signal WLw22420-XC (X = 1,..., N) replaces the connection of the other control signal WLr to the source of the gated diode. In addition, WLrs lines / drivers 2420-1B to 2420-NB are coupled to the gate of the read select FET, while WLw lines / drivers 2320-1A to 2320-NA are coupled to the write FET.

  FIG. 35 shows the array of FIG. 33, except that ground lines 2350-1 to 2350-M / 8 are shared by all such many (8 in this example) bits. The ground line is connected to a global ground line that runs parallel to the bit line, locally in the figure, horizontally for each group of eight memory cells.

  33-35, the read and write bit lines can be separated as different bit lines for reading and writing for each column of cells, thereby enabling dual port read / write operations. In dual port operation, read and write operations can be overlapped to some extent to achieve higher read / write data throughput or shorter read / write cycle times. The read / write bit lines can also be combined into a single bit line for each column of cells. This is a single port read / write operation. In the single port operation, the read operation and the write operation cannot be overlapped. In the case of a single port, as one bit line and one word line, but in the case of a dual port memory cell, two bit lines and two word lines are the memory cell. Pass through a single bit.

  Each horizontal read or write word line drives many memory cells (typically 64 to 256 cells in relatively small high speed arrays, and in some cases 1024 cells in very high density arrays), each bit Lines (read / write) run vertically and generally connect to 128-256 cells. Horizontal word lines and vertical bit lines form a memory array. Since wordlines and bitlines are long lines, such appropriate wordline drivers must be designed to handle wordline loads and RC delays. Also, to achieve the timing objective, an appropriate design must be used to supply sufficient current to drive the bit lines during read and write operations.

  Each bit line is connected to a single-ended sense amplifier. As described in Luk et al., US Patent Application No. XX / XXX, XXX, entitled “AMPLIFIERS USING GATED DIODE” filed on the same day as this application. In addition, an exemplary high speed sense amplifier based on a gated diode can be used. The disclosure of this application has already been incorporated by reference. In a write operation, the sense amplifier writes voltages 0 volts and VBLH to memory cell 1300 for data 0 and data 1, respectively. Prior to the read operation, each bit line is precharged to VBLH. In the complementary mode of operation where a p-type gated diode 1330 is used, the bit line is precharged to GND. For an n-type gated diode, during a read operation, for a data 0, the bit line remains high (VBLH) and therefore the sense amplifier does not switch. When reading data 1, the bit line voltage is discharged and goes to ground, and after a certain time interval after the signal is generated on the bit line, the signal amplifier is activated and the signal amplifier detects the bit line signal. , A value corresponding to either data 0 or data 1 is output.

  For a write word line connected to such a memory cell that is not selected, the write gate 1325 of such a cell that is not selected for read or write has a negative gate voltage equivalent to a few decades less with respect to the threshold voltage, and thus a negative gate. A negative voltage (VWLL) can be applied to the word line to have a source-to-source voltage (Vgs) to reduce the subthreshold current of the write gate 1345 and thus greatly increase cell retention.

  For read word line WLr 1335, a small negative voltage can be applied to the read word line during a write operation. This has the effect of lowering the threshold voltage of the gated diode 1330 whose source is connected to the read word line 1335, increasing the signal or charge written to the cell.

  In relation to the current state of the art, the power supply voltage VDD is about 0.8 to 1.2, and in conventional DRAMs and SRAMs, the gain cell, bit line voltage and amplitude are closely related to VDD. The read and write voltage increase method of gated diode memory cell 1300 requires about half of the VDD voltage to apply to the cell, thus reducing power and coupling noise. As such, VBLH is typically between 0.4 and 0.5V. The voltage increase and voltage gain characteristics of the memory cell 1300 can increase the cell signal during writing or reading to achieve a larger signal-to-noise margin and faster cell speed compared to conventional memory cells. it can. By reducing the bit line voltage by approximately half, the active power is reduced by almost 50 percent.

  Furthermore, because of the lower bit line voltage than conventional memories, the word line voltage can be much smaller. The reason is that no word line voltage increase is required to write a higher bit line voltage to the cell. Compared to the very high (about 50-75 percent higher) word line voltage required in conventional DRAMs, the commonly used word line voltage (VWLH) is the same as VDD. This simplifies the wordline voltage scheme and greatly reduces wordline driver complexity and wordline driver active power. The gated diode memory cell wordline driver saves substantial area, and thus far better array efficiency is achieved.

  For n-type gated diode memory cells, the bit line is precharged to VBLH, and the voltage amplitude on the bit line is typically between 0 V (GND) and VBLH, so this amplitude is the conventional DRAM cell, About half of the SRAM cell and the gain cell. This results in low active state power, low refresh power, and low bit line switching noise.

  Further, when data 0 is written and read during voltage increase, no active current flows because no charge is stored in the gated diode memory cell. This advantage shows a difference from other methods of increasing voltage using conventional capacitors. During a read operation, the bit line with data 0 remains high at VBLH, providing a shielding effect that reduces switching coupling noise between the bit lines.

  A ground (GND) line runs parallel to the read bit line and the write bit line. This ground line provides a shielding effect that reduces switching noise coupling between the bit lines during read and write operations. In addition, the GND line is configured so that the current from the bit line to ground is distributed throughout the bits of the array during a read operation.

  FIG. 36 shows two graphs illustrating various waveforms of a 3T1D gated diode memory cell. The upper graph shows the waveform of the write word line WLw and the waveform of the read word line WLr. Furthermore, corresponding cell voltage waveforms of the read bit line BLr and the write bit line BLw are shown. From the cell voltage waveform (shown in the upper graph of FIG. 36), the cell voltage waveform highlights an important characteristic of the 3T1D memory cell, namely the voltage gain of the memory cell during a read operation. The cell voltage waveform shows a voltage difference of about 0.5V between stored data 1 and data 0, and when read, causes a voltage difference of about 1.3V between data 1 and data 0, thus A voltage gain of about 2.6 is achieved. The lower graph shows the sense amplifier (SA) output. As can be seen in FIG. 36, when reading data 1, the cell voltage is increased and the bit line starts with a high voltage, but ends with a lower voltage than when reading data 1. When reading data 1, the SA output goes high. During the reading of data 0, the cell voltage only experiences a slight increase in voltage. The SA output remains at a low voltage.

  It should be understood that the embodiments and variations shown and described herein are merely illustrative of the principles of the present invention and that those skilled in the art will depart from the scope and spirit of the present invention. And various modifications can be made.

FIG. 6 shows exemplary symbols used for the first n-type gated diode. It is an example of the side view which shows the 1st n-type gate control diode formed with the semiconductor. FIG. 6 illustrates exemplary symbols used for a second n-type gated diode. It is an example of the side view which shows the 2nd n-type gate control diode formed with the semiconductor. FIG. 6 is a graph showing how the dopant concentration affects the threshold of a gated diode. 6 is a graph showing how the capacitance of a gated diode varies with gate-source voltage (Vgs) applied to gated diodes of different sizes. FIG. 4 shows exemplary symbols used for the first p-type gated diode. It is an example of the side view which shows the 1st p-type gate control diode formed with the semiconductor. FIG. 6 shows exemplary symbols used for a second p-type gated diode. It is an example of the side view which shows the 2nd p-type gate control diode formed with the semiconductor. It is an example of the side view which shows the 1st n-type gate control diode formed with the silicon on insulator (SOI). It is an example of the side view which shows the 2nd n-type gate control diode formed by SOI. It is an example of the side view which shows the 1st p-type gate control diode formed with the silicon | silicone on insulator (SOI). It is an example of the side view which shows the 2nd p-type gate control diode formed by SOI. It is a graph which shows the gain of a circuit when a capacitor is used as a charge storage and transfer device. It is a figure which shows the example of a gate control diode circuit. FIG. 17 is a representative circuit illustrating the gated diode circuit of FIG. 16 when the gated diode is turned off. FIG. 17 is a representative circuit showing the gated diode circuit of FIG. 16 when the gated diode is turned on. Fig. 6 is a graph showing the gain of an amplifier when a gated diode is used as a charge storage and transfer device. FIG. 23 shows an example of a gated diode memory cell used in FIGS. 21 and 22. 21 is a table showing full charge transfer and limited charge transfer for the gated diode memory cell of FIG. A graph showing how the gain of a gated diode memory cell varies with capacitive load (CL), especially with a capacitance ratio Rc = Cg_gd (on) / CL, where Cg_gd (on) is Gated diode on capacitance. FIG. 6 illustrates a gated diode dual port memory cell. FIG. 6 illustrates a gated diode single port memory cell. FIG. 24 illustrates the gated diode dual port memory cell of FIG. 23 used in one example. FIG. 4 illustrates a memory portion having a gated diode dual port memory cell and a sense amplifier. FIG. 27 illustrates several waveforms of the memory portion of FIG. 26 having a read voltage increase. FIG. 27 shows several waveforms of the memory portion of FIG. 26 with a write voltage increase. 1 is a side view of a three-transistor, one-diode (3T1D) gate-controlled diode planar memory cell implemented using field effect transistor (FET) technology in bulk silicon. FIG. FIG. 6 is a side view of another 3T1D gated diode planar memory cell implemented using silicon on insulator (SOI). FIG. 3 is a side view of a 3T1D gated diode trench memory cell implemented using FET technology. FIG. 32 is a side cross-sectional view showing the two trench gated diode structures used in FIG. 31 sharing a common connection. 1 is a circuit diagram and overall structure showing several memory arrays using 3T1D gated diode memory cells. FIG. 1 is a circuit diagram and overall structure showing several memory arrays using 3T1D gated diode memory cells. FIG. 1 is a circuit diagram and overall structure showing several memory arrays using 3T1D gated diode memory cells. FIG. Figure 3 is a graph showing certain waveforms of 3T1D gated diode memory cells.

Explanation of symbols

100, 200, 600, 700 n-type gate control diode 400, 500, 800, 900 p-type gate control diode 105, 125, 205, 225, 405, 425, 505, 525, 605, 625, 705, 725, 805, 825, 905, 925, 1965 Trench isolation regions 110, 210, 410, 510, 610, 710, 810, 910, 1970, 1975, 2220 Source diffusion regions 115, 215, 415, 515, 615, 715, 815, 715, 815, 915 Gate 120, 220, 420, 520, 620, 720, 820, 920, 1983, 1990, 1993, 2190, 2290, 2583 Gate insulator 130, 230, 430, 530, 630, 730, 830, 930, 1992 245, 545, 745, 945, 1975, 1970, 2570 Drain diffusion region 250, 550, 750, 950 Interconnect 1130, 1330, 1904 Gate control diode 1120, 1350, 1940 Control line 1300, 1400, 1900, 2000, 2100, 2311-11 to 2310-NM 3T1D gated diode memory cell 1305, 1925, 2330-1 to 2330-M Write bit line (BLw)
1310, 1915, 2340-1 to 2340-M Read bit line (BLr)
1320, 1335, 1920, 2520 Write word line (WLw)
1335, 1340, 1350, 1945 Read word line (WLr)
1325 Write gate FET
1345 Read gate FET
1340, 1902 Read selection FET
1410 single bit line 1901 write FET
1903 Read FET
1910 ground 1994 (gate) cap portion 2110, 2210 cylindrical polysilicon trench 2350-1 to 2350-M / 2, 2350-1 to 2350-M / 8 ground line G gate S source SA sense amplifier STI shallow trench isolation

Claims (11)

A memory cell configured to be coupled to a first control line, a second control line, a third control line and at least one bit line for selectively accessing the memory cell,
A write switch having a control terminal to which a voltage signal composed of a first voltage level indicating selection of writing and a second voltage level indicating non-selection is input, and first and second terminals, A write switch in which the first terminal of the switch is coupled to the at least one bit line and the control terminal of the write switch is coupled to the first control line;
A two-terminal semiconductor device having first and second input terminals, wherein the first input terminal of the two-terminal semiconductor device is coupled to the second terminal of the write switch, and the two-terminal semiconductor device The second input terminal is coupled to the second control line, and the capacitance when the voltage of the first input terminal with respect to the second input terminal exceeds a threshold voltage does not exceed the threshold voltage The voltage signal input to the control terminal of the write switch is transitioned from the second voltage level and is at the first voltage level during a write operation. voltage of said second input terminal is changed between the final voltage of the first input terminal to be stored after the writing, the write voltage applied to the first input terminal 2 and terminal semiconductor device that increases according to the magnitude of the capacitance to respond,
A read selection switch having a control terminal and first and second terminals, wherein the control terminal of the read selection switch is coupled to the third control line, and the first terminal of the read selection switch is the A read select switch coupled to at least one bit line;
A read switch having a control terminal and first and second terminals, wherein the control terminal of the read switch is connected to the first input terminal of the two-terminal semiconductor device and the second terminal of the write switch. A read switch coupled, wherein the first terminal of the read switch is coupled to the second terminal of the read select switch, and the second terminal of the read switch is grounded. ·cell.   The two-terminal semiconductor device comprises a gated diode having a gate input and a source input, the gate input is the first input terminal of the two-terminal semiconductor device, and the source input is the one of the two-terminal semiconductor device. The memory cell of claim 1, wherein the memory cell is a second input terminal.   The write switch, the read select switch, and the read switch are field effect transistors (FETs), each FET having a gate and two source / drain diffusion regions, each gate coupled to a respective control terminal. The memory cell of claim 1, wherein each of the respective source / drain diffusion regions is coupled to one of the respective first or second terminals. A memory array,
A plurality of first control lines;
A plurality of second control lines;
A plurality of third control lines;
Multiple bit lines,
Multiple ground wires,
A group comprising a plurality of memory cells, each of the memory cells comprising:
A write switch having a control terminal to which a voltage signal composed of a first voltage level indicating selection of writing and a second voltage level indicating non-selection is input, and first and second terminals, A write switch in which the first terminal of the switch is coupled to at least one of the bit lines and the control terminal of the write switch is coupled to one of the first control lines;
A two-terminal semiconductor device having first and second input terminals, wherein the first input terminal of the two-terminal semiconductor device is coupled to the second terminal of the write switch, and the two-terminal semiconductor A capacitance when the second input terminal of the device is coupled to one of the second control lines and the voltage of the first input terminal relative to the second input terminal exceeds a threshold voltage. It is configured to be larger than the capacitance when the threshold voltage is not exceeded, and during the write operation, the voltage signal input to the control terminal of the write switch changes from the second voltage level and the second voltage level. 1 of the voltage of said second input terminals while in the voltage level is changed, a final voltage of said first input terminal to be stored after the write, et al in addition to the first input terminal 2 and terminal semiconductor device that increases according to the magnitude of the capacitance corresponding to the write voltage that,
A read selection switch having a control terminal and first and second terminals, wherein the control terminal of the read selection switch is coupled to one of the third control lines, the first of the read selection switches; A read selection switch having one terminal coupled to at least one of the bit lines;
A read switch having a control terminal and first and second terminals, wherein the control terminal of the read switch is connected to the first input terminal of the two-terminal semiconductor device and the second terminal of the write switch. A read switch coupled, wherein the first terminal of the read switch is coupled to the second terminal of the read selection switch, and the second terminal of the read switch is coupled to a ground line; A memory array.   The plurality of second control lines are a plurality of write control lines, the plurality of third control lines are a plurality of read selection control lines, and a predetermined read selection control line is the plurality of memory A predetermined write control line coupled to the control terminal of the read select switch for a subset of a subset of memory cells in a group including cells, wherein a predetermined write control line is for the subset of the memory cells The memory array of claim 4, wherein the memory array is coupled to the second terminal of the gated diode. A predetermined memory circuit configured to be coupled to a first control line, a second control line, a third control line, and at least one bit line for selectively accessing a predetermined memory cell; A method of accessing a cell,
The predetermined memory cell is
A write switch having a control terminal to which a voltage signal composed of a first voltage level indicating selection of writing and a second voltage level indicating non-selection is input, and first and second terminals, A write switch in which the first terminal of the switch is coupled to one of the at least one bit line and the control terminal of the write switch is coupled to the first control line;
A two-terminal semiconductor device having first and second input terminals, wherein the first input terminal of the two-terminal semiconductor device is coupled to the second terminal of the write switch, and the two-terminal semiconductor The second input terminal of the device is coupled to the second control line, and the two-terminal semiconductor device has a voltage when the voltage of the first input terminal with respect to the second input terminal exceeds a threshold voltage. The capacitance is configured to be larger than the capacitance when the threshold voltage is not exceeded, and during the write operation, the voltage signal input to the control terminal of the write switch transitions from the second voltage level. the final voltage of the first voltage of said second input terminal is changed while in the voltage level, said first input terminal to be stored after the write, the first 2 and terminal semiconductor device that increases according to the magnitude of the capacitance corresponding to the write voltage applied to the force terminal,
A read selection switch having a control terminal and first and second terminals, wherein the control terminal of the read selection switch is coupled to the third control line, and the first terminal of the read selection switch is the A read select switch coupled to one of the at least one bit line;
A read switch having a control terminal and first and second terminals, wherein the control terminal of the read switch is connected to the first input terminal of the two-terminal semiconductor device and the second terminal of the write switch. A read switch coupled, wherein the first terminal of the read switch is coupled to the second terminal of the read select switch, and the second terminal of the read switch is coupled to ground; Said method comprises
Changing the voltage of the second control line coupled to the predetermined memory cell of the plurality of memory cells from a first voltage to a second voltage. The second control line is a write control line, the third control line is a read selection control line;
The method further comprises applying a signal to one of the at least one bit lines connected to the write switch;
Said changing step comprises:
Lowering the voltage of the write control line from a predetermined high voltage to a predetermined low voltage;
Raising the voltage of the write control line from the predetermined low voltage to the predetermined high voltage; and
The method of claim 6, further comprising stopping applying the signal to the at least one bit line, wherein the step of stopping applying the signal is performed after the step of reducing voltage. the method of. A semiconductor structure comprising at least one semiconductor memory device, the at least one semiconductor memory device having a first control line for selectively accessing the at least one semiconductor memory device, a second Configured to be coupled to a control line, a third control line, and at least one bit line, the at least one semiconductor memory device comprising:
An insulator formed between a gate and a well to which a voltage signal composed of a first voltage level indicating writing selection and a second voltage level indicating non-selection is input, on the gate and the gate side A write transistor comprising first and second source / drain diffusion regions formed, wherein the first source / drain diffusion region of the write transistor is coupled to at least one bit line; A write transistor having a gate coupled to the first control line;
A two-terminal semiconductor device comprising at least a gated diode insulator formed between a gate and a well, the gate, and a source diffusion region at least in contact with a part of the gated diode insulator, the two-terminal semiconductor device The gate of the write transistor is coupled to the second source / drain diffusion region of the write transistor, and the source diffusion region of the two-terminal semiconductor device is coupled to the second control line, with respect to the source diffusion region The voltage signal input to the control terminal of the write switch during a write operation is configured such that a capacitance when the gate voltage exceeds a threshold voltage is larger than a capacitance when the gate voltage does not exceed the threshold voltage. Transitions from the second voltage level to the first voltage level. Voltage of the source diffusion region is changed between that, the final voltage of the gate to be stored after the write, the 2-terminal semiconductor device that increases according to the magnitude of the capacitance corresponding to the write voltage applied to the gate ,
A read selection transistor comprising an insulator formed between a gate and a well, the gate, and first and second source / drain diffusion regions formed on a side of the gate, the read selection transistor comprising: A read select transistor having a gate coupled to the third control line and the first source / drain diffusion region of the read select transistor coupled to the at least one bit line; and
A read transistor comprising an insulator formed between a gate and a well, the gate, and first and second source / drain diffusion regions formed on the gate side, wherein the gate of the read transistor comprises: Coupled to the gate of the two-terminal semiconductor device and the second source / drain diffusion region of the write transistor, the first source / drain diffusion region of the read transistor is the second source / drain diffusion region of the read select transistor. A read transistor coupled to a source / drain diffusion region and wherein the second source / drain diffusion region of the read transistor is coupled to ground.   The two-terminal semiconductor device is a gated diode, and the gate of the gated diode is at least partially formed in a trench, and the insulator separates the trench from the well and the source diffusion region. A semiconductor structure according to claim 8. The at least one semiconductor memory device comprises a first semiconductor memory device, and the gated diode is a first gated diode having a first source diffusion region;
A second gate formed at least partially in a second trench; a second insulator formed between the second trench and the well; and the second insulator. A second gated diode having a second source diffusion region in contact with the second semiconductor memory device,
The first source diffusion region of the first gated diode is also the second source diffusion region of the second gated diode, whereby the source diffusion region is the first semiconductor memory. 10. The semiconductor structure of claim 9, shared between a device and the second semiconductor memory device. The first gated diode has a first trench, a first gate, and a first insulator;
The first trench is under a first cap portion of the first gate;
The first insulator is formed between a portion of the first cap that is not over the first trench and the well;
The second trench is under a second cap portion of the second gate;
The semiconductor structure of claim 10, wherein the second insulator is formed between a portion of the second cap that is not over the second trench and the well.

Images