Arrays of transistors with back control gates buried beneath the insulating film of a semiconductor-on-insulator substrate

The present invention is related to the field of semiconductor processing equipment for microelectronics, and more precisely to a semiconductor device made on a semiconductor-on-insulator (SeOI) substrate based on components having uniform patterns.

It is known that above certain critical temperatures surfaces of compound semiconductors can be impaired, e.g., by distortions and deformations. To avoid such impairments in lithographic structures that have to be transferred onto a wafer, it is also known in the art to seek to simplify lithography. For example, the prior art recommends avoiding as far as possible corners in the same lithographic plane.

Use of different channel widths of transistors is a common practice. The channel length of an MOSFET transistor is at the present time typically around 30 nm, whereas its width (W) is typically much greater. The width determines the current intensity in the transistor for given source, drain and gate voltages.

Therefore, it is usually desirable to design an electronic circuit with transistors having different channel widths. But, in practice it is difficult to obtain different widths with precision because of the resolution limits of lithography. In fact, although it is relatively easy to produce long strip-like structures by lithography, short strips of very controlled dimensions are particularly difficult to fabricate.

US patent application 2008/0251848 teaches a fabrication process aimed at obviating inhomogeneities in performance between the various transistors of a circuit. To do so, this document proposes to make the influence of the environment on the various transistors uniform. More precisely, this document provides that an array of FET transistors be arranged in the form of long strips. The drain and source regions of any one strip then have the same dimensions, being spaced apart by gate regions of fixed dimensions. This document is further described herein.

The resolution limit of the lithography tends to dictate the use of such long strips of transistors having identical dimensions. However, with long strips, the flexibility in designing electronic circuits is then lost since it is no longer possible to vary the geometric width of the various transistors so as to modulate their performance.

The present invention provides devices and methods that obviate the variability inherent in lithographic pattern transfer, in particular, random variability due to the structuring of the lines leading to line/space/line variability. The present invention also simplifies lithography and saves wafer space by avoiding the need for shallow trench isolation (STI) and by tightening up the rules governing transistor structures (e.g., active zones, gate lines, interconnect lines, etc.). More specifically, this invention provides semiconductor device structures fabricated by lithography that are more regular and more dense.

According to a first embodiment, this invention provides a semiconductor device structure formed on a semiconductor-on-insulator (SeOI) substrate (known to comprise a thin film of semiconductor material separated from a base substrate by an insulating film) including an array of patterns, each pattern defining at least one field-effect transistor having, in the thin semiconductor film, a source region, a drain region, a channel region (the channel region being delimited by the source and drain regions), and a front control gate region formed above the channel region. The patterns of the source and drain regions are arranged in the form of rows, having the same dimensions and spaced apart by front control gate regions of fixed dimensions.

Significantly, the provided device structure further includes at least one pattern defining a FET transistor with a back control gate region formed in the base substrate beneath the channel region. This back gate region is capable of being biased in order to shift the threshold voltage of the transistor to simulate a modification in the channel width of the transistor, or to force the transistor to remain off or on whatever the voltage applied on its front control gate.

Certain preferred, but non-limiting, aspects of this device structure are the following:

Some of the patterns in a row can be formed on one and the same active zone of the thin semiconductor film of the SeOI substrate; An isolation region can delimit the adjacent patterns, which can further include a front isolation gate formed above the active zone and a back isolation gate formed in the base substrate beneath the active zone; A back isolation line can connect the back isolation gates of each of the isolation regions of any one row; This back isolation line can be common to several rows; The back isolation gate can have a conductivity of opposite type to that of the active zone; A back gate line can connect the back gate region(s) of one or more transistors. A back gate line can connect the back gate region(s) to ground or to a nominal supply voltage; A back gate line can connects the back gate region(s) to an analog and (preferably, continuously) adjustable potential; The back gate region can be isolated from the base substrate by a well of opposite conductivity; The back gate region can have a conductivity of the same type as that of the transistor channel; and The back gate region can have a conductivity of opposite type to that of the transistor channel.

According to a second embodiment, this invention provides a method of operating a device of this invention, wherein the back control gate region one or more transistors is biased positively or negatively, preferably by an analog and (preferably, continuously) adjustable potential, in order to shift the threshold voltage of those transistors.

According to an aspect of this method, the invention provides a method of operating a device structure of this invention, wherein the biasing is controlled so that the shift of the threshold voltage is sufficient to cause the transistor to remain off or on whatever the voltage applied to its front control gate. The shift of the threshold voltage can in particular be programmed by means of memory cells that store and deliver predetermined voltages to the back control gates. The memory cells can be interleaved between rows of the pattern (that is one row of memory cells can be between two rows of the pattern).

According to a third embodiment, this invention provides a re-programmable circuit including a device structure according of this invention interleaved with rows of memory cells that store and deliver predetermined voltages to the back control gate regions.

The present invention is described herein for convenience in terms of “rows” and “columns” that reflect a particular and conventional geometric orientation. It should be understood that this invention is independent of this particular orientation, and that this orientation can be arbitrarily rotated without changing the nature of the invention. For example, the term of “row” as used herein could be replaced with the term “column,” and the term of “column” used herein could be replaced with the term “row.”

Further aspects and details and alternate combinations of the elements of this invention will be apparent from the following detailed description and are also within the scope of the inventor's invention.

The present invention will be understood more fully by reference to the following detailed description of the preferred embodiments of the present invention, non-limiting examples of specific embodiments of the invention and the appended figures in which:

FIG. 1 illustrates a prior art circuit;

FIG. 2 illustrates control of the threshold voltage of a transistor by biasing a back control gate;

FIGS. 3 to 7 illustrate various embodiments of an active zone beneath the insulating film of a SeOI substrate;

FIG. 8 illustrate a sectional view of a row of transistors of the circuit of FIG. 7 ;

FIG. 9 illustrates an example of a logic circuit that is reconfigurable to cover all 2-input Boolean functions by means of appropriate voltages applied of the back gates of the various transistors;

FIG. 10 illustrates the truth table of the reconfigurable logic circuit of FIG. 9 ; and

FIG. 11 illustrates an arrangement of logic cells and memory cells used to store the programming of the logic cells and to deliver the appropriate voltages to the back gates of the transistors of the logic cells.

Prior to describing preferred and non-limiting embodiments of the invention, US Patent Application 2008/0251848 is further described herein.

FIG. 1 of this document shows a CMOS electronic circuit according to this document. This prior-art circuit comprises an array of patterns that are formed by at least one field-effect transistor, the patterns being further arranged in the form of rows with the source and drain regions of each transistor of any one row having the same dimensions and being spaced apart by front control gate regions (WL 1 -WL 7 ) of fixed dimensions. The channel width of the various transistors is then uniform, defined by the section between two perpendicular strips.

Thus, during the lithography steps only wide strips (horizontal rows and vertical front control gate regions) need be formed. It should be noted that there are no STI-type isolation trenches between adjacent transistors of any one strip. However, such isolation trenches are indeed present along the strips so as to isolate them from one another.

This prior-art circuit comprises, from left to right in FIG. 1 , the following patterns: a logic gate NOR 2 , three inverters INV 1 , INV 2 and INV 3 and a logic gate NAND 2 . More precisely, the circuit comprises, in this example, nine metal busses, the p-FET transistors being arranged along busses 2 and 3 whereas the n-FET transistors are arranged along busses 7 and 8 . Busses 4 to 6 are used to make the input/output connection to these patterns and to connect these various patterns together. It is clear that other combinations are possible, with 8 to 12 metal busses being a reasonably possible number of busses. These metal busses are preferably at metal interconnect level 2 according to device technologies known in the art, and are referred to also as “metal2” busses. Herein, metal busses are preferably at interconnect level 2 unless otherwise specified.

Metal (preferably at interconnect level 1 , or “metal1”) supply lines BL P1 , BL P2 , BL N1 , BL N2 of this prior-art circuit serve to fix the potential of some of the drain regions of the transistors. Thus, the drain of the first p-FET transistor of the logic gate NOR 2 and the drain of the p-FET transistors of the inverters INV 1 and INV 2 are connected to the line BL P1 , whereas the drain of the p-FET transistor of the inverter INV 3 and the drain of the p-FET transistors of the logic gate NAND 2 are connected to the line BL P2 . The drain of the n-FET transistors of the logic gate NOR 2 and the drain of the n-FET transistors of the inverters INV 1 and INV 2 are connected to the line BL N1 , whereas the drain of the n-FET transistor of the inverter INV 3 and the drain of the first n-FET transistor of the logic gate NAND 2 are connected to the line BL N2 .

The lines BL P1 and BL P2 are used typically to deliver a nominal supply voltage, V DD , whereas the lines BL N1 and BL N2 are typically connected to ground, GND.

The patterns along a row of this prior-art circuit are formed on one and the same active zone of the substrate, and therefore isolation regions are provided between adjacent patterns. These isolation regions, each having a front isolation gate formed above the active zone, bear, in what follows, the references I P1 -I P6 for the isolation regions linked to the p-FET transistors and the references I N1 -I N6 for the isolation regions linked to the n-FET transistors. The front isolation gates of the isolation regions are biased by means of isolation gate supply lines BI P and BI N for the isolation regions between p-FET patterns and between n-FET patterns, respectively. These lines BI P and BI N are typically formed from a polycrystalline semiconductor material (typically polycrystalline silicon).

Turning now to the present invention, circuit structures provided by this invention are fabricated onto a conventional SeOI substrate (comprising a thin film of semiconductor material separated from a base substrate by an insulating film) and have a uniform environment similar to that shown in FIG. 1 . However, unlike FIG. 1 , the invention also provides a back control gate for at least one transistor in the base substrate facing the channel of the at least one transistor. By biasing the back control gate of the at least one transistor positively or negatively (typically by +V DD or −V DD ), the properties of this transistor may be individually modified. In particular, the threshold voltage of this transistor may be shifted. Modifying the threshold voltage is known to be equivalent to modifying the physical width of the transistor's channel.

Thus, in this invention, the physical width of the channel is defined once and for all and for all the transistors. However, it is possible to modify, individually for each transistor, the apparent (effective) width of its channel by choosing the way in which the back control gate is biased (or, in other words, driven). Since the voltage applied to the back control gate may be modified, the invention thus offers the advantage of modifying the effective channel width dynamically.

A transistor in which the channel is of n-type conductivity and a back control gate is of p-type conductivity has a very high threshold voltage. This threshold voltage may then be reduced by applying a positive voltage to the back control gate. In the present description, such a back control gate is said to have a work function.

A transistor in which the channel is of n-type conductivity and a back control gate is of n-type conductivity has a nominal threshold voltage (that is, a threshold voltage expected for a transistor of this type and configuration). This threshold voltage can be reduced by applying a positive voltage to the back control gate. In the present description, such a back control gate is said to have no work function.

This variation in the threshold voltage of the transistor via the back control gate may be formulated as V th =V t0 −αV BG , where V th represents the threshold voltage of the transistor, V BG represents the voltage applied to the back control gate, V t0 represents the nominal threshold voltage (which may be shifted by the work function depending on whether a back control gate of n-type or p-type is used) and α represents a factor associated with the geometry of the transistor.

It will therefore be understood that the doping type of the back control gate associated with a transistor may or may not shift the nominal threshold voltage, but that by biasing the back control gate it is possible to adjust the threshold voltage. It is thus possible to benefit from an increase in the conduction current I ON in the on-state of the transistor (by reducing the threshold voltage) and from a reduction in the small leakage current I OFF in the off-state of the transistor (by increasing the threshold voltage). A contribution may also be made to the isolating function of an isolation region separating the adjacent patterns along a row by reducing the leakage current therein.

As presented in the thesis “Architectures innovantes de mémoire non-volative embarquée sur film mince de silicium (Innovative non-volatile memory architectures on thin silicon films)” defended by Germain Bossu in June 2009 at the Provence, Aix-Marseille University I, the factor α may in particular be approximated as α=3t ox1 /(t Si +3t ox2 ), where t ox1 denotes the thickness of the dielectric gate film separating the front control gate from the channel (typically 1-2 nm), t ox2 denotes the thickness of the insulating film separating the back control gate from the channel (typically 5-20 nm for SeOI substrates) and t Si , denotes the thickness of the thin semiconductor film.

FIG. 2 illustrates the control of the threshold voltage of a transistor produced on a SeOI substrate by biasing a back control gate fabricated in the base substrate, beneath the insulating film and facing the channel of the transistor. In this figure, the central curve C N represents the nominal characteristic log(I D (V G )) (for a transistor with no back control gate).

The upper curve C VT− (that is, C with subscript VT−) represents the characteristic log(I D (V G )) under the effect of a back control gate driven with a voltage below the nominal supply voltage V DD of the circuit. This upper curve illustrates the increase in the threshold voltage and the reduction in the currents I ON and I OFF . The lower curve C VT+ (that is, C with subscript VT+) represents the characteristic log(I D (V G )) under the effect of a back control gate driven with a nominal supply voltage V DD . This lower curve illustrates the reduction in the threshold voltage and the increase in the currents I ON and I OFF .

It will thus be understood that, by varying the back control gate bias, it is possible to cover the entire area between the lower curve C VT− and the upper curve C VT+ , thus modulating the threshold voltage of the transistor and its characteristic currents I ON and I OFF .

The following equation links in particular the current I D in the channel to the channel width W and to the threshold voltage V th :

Consider a first example in which the gate and drain voltages are 0.9 V (V GS =V DS =0.9 V), with λ=0.05, and the threshold voltage is V th =0.3 V. By modifying the threshold voltage of the transistor so that it takes a value between 0.05 V and 0.6 V, it is then possible to simulate a modification in the physical width of the channel (at fixed I D ). Theoretically, the effective width of the transistor channel can be modified between 0.25 and 2.01 times the physical width (W 0 ) of the channel.

Consider a second example with lower gate and drain voltages, namely 0.7 V, and further consider again modifying the threshold voltage of the transistor so that it takes a value between 0.05 V and 0.6 V. Then, the effective width of the channel can be theoretically modified between 0.06 and 2.64 times the physical width of the channel

The invention thus allows a reduction/increase of the effective channel width. This control is particularly important when the supply voltages are low. In this regard, since the tendency in the technical field of the invention is for future generations of electronic components to use ever lower supply voltages, this invention can become, a priori, even more significant for such future generations.

FIGS. 3 to 7 illustrate the active zone beneath the insulating film of a circuit similar to that in FIG. 1 according to various embodiments and aspect of the invention.

In FIG. 3 :

p-doped back control gate regions are associated with: the isolation regions I P1 -I P6 ; the two p-FET and n-FET transistors of the inverter INV 2 ; and with the two n-FET transistors of the logic gate NAND 2 ; and n-doped back control gate regions are associated with the isolation regions I N1 -I N6 and with f the two p-FET and n-FET transistors of the inverter INV 3 .

A back gate line BG P and a back gate line BG N serve for collectively connecting the p-doped back gate regions and the n-doped back gate regions, respectively, to the same potential. The line BG P associated with the p-doped back gates may thus be in a high state (being typically connected to the potential V DD ), whereas the line BG N associated with the n-doped back gates may be in a low state (being typically connected to ground GND).

In such a manner, the isolation regions experience higher threshold voltages and consequently lower leakage currents I OFF , thus helping to maintain better isolation between adjacent patterns along any one row.

As regards the inverter INV 2 , the back control gate of the p-FET transistor and that of the n-FET transistor are in the high state V DD . The conduction current I ON of the n-FET transistor is increased, while the conduction current of the p-FET transistor is reduced. This inverter thus has a low p-FET and a high n-FET. As regards the inverter INV 3 , this inverter has a high p-FET and a low n-FET, because the back control gate of the p-FET transistor and that of the n-FET transistor are in the low state GND.

As regards the logic gate NAND 2 , the p-FET transistors do not have a back control gate and therefore operate in the nominal manner. The n-FET transistors do have back control gates in the high state, and therefore have a higher conduction current.

FIG. 4 shows another embodiment in which four different voltage levels are used to offer more flexibility. In FIG. 4 :

p-doped back control gate regions collectively connected to a back gate line BG PH in the high state are associated with the isolation regions I P1 -I P6 and with the p-FET transistor of the inverter INV 2 ; a p-doped back control gate region connected to a back gate line BG PL in a low state is associated with the p-FET transistor of the inverter INV 3 ; n-doped back control gate regions collectively connected to a back gate line BG NL in a low state are associated with the isolation regions I N1 -I N6 and with the n-FET transistor of the inverter INV 3 ; and an n-doped back control gate region connected to a back gate line BG NH in a high state is associated with the n-FET transistor of the inverter INV 2 .

As in the example in FIG. 3 , the leakage current of the isolation regions is thus reduced.

The inverter INV 2 has a low p-FET transistor (p-doped back control gate in a high state) and a high n-FET transistor (n-doped back control gate in a high state). In contrast, the inverter INV 3 has a high p-FET transistor (p-doped back control gate in a low state) and a low n-FET transistor (n-doped back control gate in a low state).

FIG. 5 shows another embodiment in which a back control gate associated with a transistor is connected to a back gate line which is associated only with that transistor. It will be understood that such an embodiment enables the potential applied to the back gate line dedicated to a single transistor to be modulated. In FIG. 5 :

p-doped back control gate regions are collectively connected to a back gate line BG PH in a high state, and are associated with the isolation regions I P1 -I P6 and with the p-FET transistor of the inverter INV 2 ; a p-doped back control gate region connected to an individual back gate line BG PA , the voltage of which can be adjusted, is associated with the p-FET transistor of the inverter INV 3 ; n-doped back control gate regions are collectively connected to a back gate line BG NL in a low state, and are associated with the isolation regions I N1 -I N6 and with the n-FET transistor; and an n-doped back control gate region connected to an individual back gate line BG NA , the voltage of which can be adjusted, is associated with the n-FET transistor of the inverter INV 2 .

Thus, the inverter INV 2 has a low p-FET transistor (p-doped back control gate in a high state) and an adjustable n-FET transistor (an individual n-doped back control gate addressed via the line BG NA , the potential of which can be adjusted). The inverter INV 3 , this has a low n-FET transistor (n-doped back control gate in a low state) and an adjustable p-FET transistor (individual p-doped back control gate addressed via the line BG PA , the potential of which can be adjusted).

FIG. 6 shows an alternative embodiment of the example of FIG. 4 . In FIG. 6 :

p-doped back control gate regions are collectively connected to a back gate line BG PH in a high state, and are associated with the isolation regions I P1 -I P6 ; p-doped back control gate regions are collectively connected to a back gate line BG PL in a low state, and are associated with the p-FET transistor of the inverter INV 2 and with the p-FET transistor of the inverter INV 3 ; n-doped back control gate regions are collectively connected to a back gate line BG NH in a high state, and are associated with the isolation regions I N1 -I N6 and with the n-FET transistor of the inverter INV 3 ; and an n-doped back control gate region is connected to a back gate line BG NL in a low state, and is associated with the n-FET transistor of the inverter INV 2 .

Thus, the inverter INV 2 has a high p-FET transistor (p-doped back control gate in a low state) and a low n-FET transistor (n-doped back control gate in a low state). The inverter INV 3 has a high p-FET transistor (p-doped back control gate in a low state) and a high n-FET transistor (n-doped back control gate in a high state).

FIG. 7 shows a preferred embodiment in which the isolation regions are associated with back control gate regions having a conductivity of opposite type. The leakage currents in these isolation regions are thus further reduced.

More precisely, in FIG. 7 :

n-doped back control gate regions are collectively connected to a back gate line BG NH in a high state, and are associated with: the isolation regions I P1 -I P6 ; one of the n-FET transistors of the logic gate NOR 2 ; and the p-FET transistor of the inverter INV 2 ; a p-doped back control gate region is connected to an individual back gate line BG PA1 , the voltage of which can be adjusted (to any intermediate value chosen by the skilled person), and is associated with the p-FET transistor of the inverter INV 3 ; p-doped back control gate regions collectively are connected to a back gate line BG PL in a low state, and are associated with the isolation regions I N1 -I N6 , with one of the p-FET transistors of the logic gate NAND 2 and with the n-FET transistor of the inverter INV 3 ; a p-doped back control gate region is connected to an individual back gate line BG PA2 , the voltage of which can be adjusted (to any intermediate value chosen by the skilled person), and is associated with the n-FET transistor of the inverter INV 1 ; an n-doped back control gate region is connected to an individual back gate line BG NA1 , the voltage of which can be adjusted (to any intermediate value chosen by the skilled person), and is associated with the n-FET transistor of the inverter INV 2 ; and an n-doped back control gate region is connected to an individual back gate line BG NA2 , the voltage of which can be adjusted (to any intermediate value chosen by the skilled person), and is associated with the p-FET transistor of the logic gate NAND 2 which is not connected to BG PL .

Thus, the isolation regions I P1 -I P6 have a p-channel with a p-type back control gate in the high state (typically at V DD ). These regions have a maximum threshold voltage and consequently a minimum leakage current. The isolation regions I N1 -I N6 themselves have an n-channel with an n-type back control gate in the low state (typically at GND). These regions have a maximum threshold voltage and consequently a minimum leakage current.

One of the n-FET transistors of the logic gate NOR 2 also has an n-channel, but with an n-type back control gate in the high state. It thus has a minimum threshold voltage and consequently maximum performance characteristics (in terms of conduction current I ON ). The n-FET transistor of the inverter INV 1 has an n-channel with a p-type back control gate, the voltage of which can be adjusted. Thus, this transistor has performance characteristics lying between the minimum performance characteristics and the mean performance characteristics depending on the voltage applied to the individual line BGP A2 .

The p-FET transistor of the inverter INV 1 does not have a back control gate, and therefore it operates in a nominal manner.

The n-FET transistor of the inverter INV 2 has an n-channel with an n-type back control gate, the voltage of which can be adjusted. Thus, this transistor has performance characteristics lying between the mean performance characteristics and the maximum performance characteristics depending on the voltage applied to the individual line BGP NA1 .

The p-FET transistor of the inverter INV 3 has a p-channel with a p-type back control gate, the voltage of which can be adjusted. Thus, this transistor has performance characteristics lying between the minimum performance characteristics and the mean performance characteristics depending on the voltage applied to the individual line BGP PA1 .

The p-FET transistor of the logic gate NAND 2 , which does have an n-type back control gate, the voltage of which can be adjusted via the individual line BG NA2 , has performance characteristics lying between the mean performance characteristics and the maximum performance characteristics. The p-FET transistor of the logic gate NAND 2 , which has a p-type back control gate, the voltage of which is in the low state via the line BG PL , has itself a minimum threshold voltage and maximum performance characteristics.

In FIG. 7 , the references C P and C N represent the wells that isolate the n-type and p-type back control gates respectively. The wells C P are typically in the low state GND, whereas the wells C N are typically in the high state V DD .

The upper part of FIG. 8 shows two metal (preferably at level 2 ) lines, 7 and 8 , along which the FET transistors of the circuit shown in FIG. 7 are placed. The lower part of FIG. 8 shows a sectional view of a portion of metal (preferably at level 2 ) line 8 comprising the isolation regions I N1 -I N3 and the n-FET transistors T 1 -T 4 of the NOR 2 (T 1 and T 2 ), INV 1 (T 3 ) and INV 2 (T 4 ) patterns. In this sectional view, the insulating film bears the reference BOX.

In FIG. 8 , the transistor channels are fully depleted, the source and drain regions being in contact with the insulating film. However, the invention also applies to the partially depleted technology in which the source and drain regions do not extend into the entire thin film. It should be noted in this case that the back control gate is overall less effective as it is further away from the channel region between the source and drain regions.

In FIG. 8 , and as discussed above, the isolation regions I N1 -I N3 have an n-channel and each region has a p + -type back control gate G P1 -G P3 (in the low state via the line BG PL ).

In FIG. 8 , one of the nFET transistors, T 1 , of the logic gate NOR 2 has an n + -type back control gate G N1 (in the high state via the line BG NH ), whereas the other nFET transistor, T 2 , of the logic gate NOR 2 does not have a back control gate.

In FIG. 8 , the nFET transistor T 3 of the inverter INV 1 has a p + -type back control gate G P4 (which can be adjusted via the individual line BG PA2 ). The nFET transistor T 4 of the inverter INV 2 has an n + -type back control gate G N2 (which can be adjusted via the individual line BG NA1 ).

In FIG. 8 , the associated back control gate is localized so as to extend only opposite the channel of the associated transistor. For example, the back control gate is formed by implanting dopants beneath the insulating film BOX. The back control gate is isolated from the base substrate by a well C N1 , C P1 , C N2 , C P2 of opposite conductivity to that of the control gate (n − -type well C N1 , C N2 for a p + -type back control gate G P1 , G P2 , G P3 , G P4 ; p − -type well C P1 , C P2 for an n + -type back control gate G N1 , G N2 ).

The well voltage is chosen so that the diode created by the electrical node between the back control gate and the well is always reverse-biased, the diode isolating the back control gate from the well and from anything that it may contain (especially other back control gates). In fact, it is of course possible to provide a well common to several back control gates of the same type, as is the case for the well C N2 which collectively isolates the back control gates G P2 , G P3 and G P4 .

According to an alternative embodiment (not shown), a second insulating film, placed in the base substrate beneath the insulating film, BOX, may contribute, fully or partly, to isolating a back control gate from the base substrate.

A device according to the invention has the following advantages. A wide performance range may be obtained for the p-FET and n-FET transistors while using only a single physical channel width. Typically, three types of performance are the following:

regular performance, in the absence of a back control gate: with the transistor then being an ordinary SeOI transistor and it is unnecessary to modify the existing circuit designs; boosted performance, with an “ON” back control gate, increasing the conduction of the transistor: with the transistor then operates as if it were wider than it actually is, or it has a smaller footprint per performance unit (speed, I ON ); and reduced performance, with an “OFF” back control gate, reducing the conduction of the transistor: with the transistor then operating as if it were narrower than it actually is. This may prove to be advantageous when performance ratios are desired (in the case of latches, for example, of the flip-flop type) since it is therefore unnecessary to widen another device. The leakage currents are also significantly reduced. This type of performance may thus be used in the off-states so that, in so far as no transistor switches (no ratio to be respected), the leakage current I OFF then being reduced.

However, the invention is in no way limited to these three types of performance. Indeed, the performance characteristics can be dynamically modulated by applying an analog and (preferably, continuously) adjustable bias to the back control gate of a transistor. Any kind of performance in between the above mentioned “reduced performance” and the “boosted performance” can therefore be achieved.

In a particular variant, the back control gate may be biased by any one of multiple predefined voltages. These multiple voltages are typically voltages that are available in the environment of the device, such as Vdd, various multiples or fractions of Vdd (such as Vdd*2, Vdd/4) and various combinations of available voltages (such as Vdd−Vtp, Vdd−Vtn, etc.). Thereby, the ratios between the pFET and nFET transistors may be adjusted without having to modify the channel width.

Thus, it will be understood that a device structure according to the invention obviates the needs encountered in the conventional designs for transistors having different dimensions. Effectively only a single physical width and simplified transistor models and parameters are used. It should be pointed out that the transistor models are actually complex equations with many secondary (parasitic) or tertiary edge effects. Most of these effects depend on the dimensions of the transistor and, for recent technologies, the environment (proximity stress). In this invention, a single topology can be used, and the models can then be greatly simplified (this results in more rapid availability, shorter development times, etc.).

Moreover, a device structure of this invention can have a very low variability because of the high level of regularity, and because only the polysilicon roughness remains, and because, when fully depleted structures are produced, fluctuation in the level of dopants disappears.

Moreover, a device structure of this invention can also be insensitive to edge rounding effects of patterns because all the channels have the same physical width. The active zone strip is effectively a long polygon, having no rounding near a polysilicon connection line. The 90° corner of such a line is also far from the active strip (both at the top and the bottom of the structure) and does not interact with the active strip.

Moreover, a device structure of this invention can be insensitive to cross-coupling effects. This is because, in conventional designs, it often happens that a polysilicon connection is adjacent to a drain of another logic gate. The two nodes are then capacitively coupled and mutually disturbed, generally resulting in communication delays. Since the status of the neighboring cells is unpredictable, this coupling cannot be taken into account in the model of each cell and is therefore discovered relatively late in the design of the application. In this invention, the interaction between active strip and polysilicon connection is the same whatever the status of the neighboring cells, thereby resolving this drawback encountered in conventional designs. In particular, the modelling of the coupling remains valid after circuit production.

Moreover, a device structure of this invention can have a reduced power consumption because of the association of back control gates with the isolation regions, which reduces leakage currents therein. There is the further possibility of dynamically acting on the back control gates associated with the transistors in their off-state, in order to further reduce the leakage currents.

To illustrate the effectiveness of the invention, it will be recalled that a standard CMOS cell library may comprise twelve inverters having different performance characteristics. The invention makes it possible, when applied in fully-depleted technology, to use only three inverters (INV 1 , INV 4 and INV 8 ), because the effective channel width may be modulated between +/−50% of the physical width. In partially-depleted technology, only four inverters are necessary (INV 1 , INV 4 , INV 6 and INV 9 ), because the effective channel width may be modulated between +/−30% of the physical width. It follows that the standard cell library is considerably simplified—in fact it can be generally reduced by a factor of 2. Thus, although around a hundred design rules are used today, the invention makes it possible for only around fifty of to actually be used.

It should be noted in this regard that, in the past, the design rule manual contained approximately 100-200 rules. Since technologies are now generally below 100 nm, many physical effects are appearing that are resulting in new rules that prevent the simple and easy application of the initial set of 100-200 rules. At the 32 nm technology node, the design rule manual comprises about 800 to 1000 rules, in which most of the new rules are complex descriptions of difficult combinations of topologies. This entails a loss of efficiency in terms of footprint. In contrast, the efficiency of the invention remains generally constant across technology line sizes. Thus, if the invention and a conventional approach are generally equivalent in terms of footprint used at the 45 nm node, the invention will become increasingly more efficient at subsequent technology nodes.

Moreover, in so far as the set of design rules used in this invention can be a reduced subset of the usual set of rules, and because each of the rules used are applied in a single and unique context, it proves possible to design transistors below what the lithography might initially by presumed to permit. In particular, it is possible to optimize the width of contacts (for example by replacing two nominal square contacts by a single rectangular, and somewhat narrower, contact), and also to optimize the poly-contact distance.

Moreover, it should be noted that the back control gates have the advantage of being buried beneath the insulating film and consequently have no impact on the footprint. It should also be pointed out that the footprint may be reduced by about 10 to 15% since it is possible to generate strong conduction with the aid of cells “boosted” in accordance with the invention. Finally, it should be noted that the extreme regularity of (what is called) the “front-end” structure of the transistors is particularly suitable for the normal use of (what are called) “back-end” metallization for (what are called) “standard” cells (i.e. cells predesigned for general usage).

This is because standard cells are interconnected (routed) together via metallization levels that are alternately horizontal and vertical with a constant pitch. The reduction in number of design rules enormously facilitates the closing-off of the front-end and back-end constraints (the transistor repeat pitch and the routing pitch are made identical). The usage of standard cells then becomes easier since—by construction—the inputs/outputs are placed on the metal routing grid. In the conventional case, this is not always very easy and it is necessary in particular to take into account the possibility of reflecting the cells along the vertical and/or horizontal axes. However, keeping the inputs/outputs on the routing grid often means increasing the footprint of the standard cell. In the case of the invention, all these considerations are eliminated through the general simplification and pre-placement of the strips of transistors.

Moreover, the invention is not limited to providing the devices described, but rather the invention also provides methods of driving such devices in which the back control gate regions are biased positively or negatively in order to shift the threshold voltage of the transistors. Advantageously, a p-type back isolation gate can be connected to ground, and an n-type back isolation gate can be connected to a nominal supply voltage.

As mentioned above, the application of a back gate voltage on a FDSOI (fully-depleted SOI) transistor modifies the electrical characteristic of the device. Simply described, a higher voltage on the back gate reduces the threshold voltage of an N-channel device and vice-versa. The same applies on P channel devices.

It is possible to saturate this effect by applying very high voltages on the back gates. For instance a (very) high voltage applied on an N-channel transistor can reduce its threshold down to a negative value, making the transistor always ON, despite the voltage level applied on its front gate. On the contrary, a (very) high negative voltage on the back gate increases the threshold above the power supply, Vdd, at a level at which the transistor remains OFF for all operating voltages applied on its front gate (0−Vdd). By symmetry the same apply on the complementary P channel device.

In other words, transistors can be transformed into “open” and “shorts” through an appropriate back gate control voltage. This feature proves to be particularly efficient for creating reconfigurable logic. FIG. 9 represents an example of a reconfigurable logic circuit that can be controlled to perform all 2-input Boolean functions, depending on the voltage applied on the back gate BG 1 -BG 4 of the various transistors T 1 -T 4 , as indicated on the truth table of FIG. 10 . In FIG. 9 , A and B represent the two inputs of the reconfigurable logic circuit, while OUT represents the output of the reconfigurable logic circuit.

In FIG. 10 , Vpp represents a voltage that is sufficiently high to force the threshold voltage either to a negative value (the transistor is always ON) or a value higher than the supply voltage Vdd (the transistor is always OFF) for a N-channel transistor, and respectively to a value either positive or greater than Vdd in absolute value for a P-channel transistor. It will be noted that this principle of reconfigurable logic can easily be extended to 3 or more inputs.

Transforming transistors into “opens” and “shorts” through an appropriate back gate control proves useful for re-programmable circuits, such as FPGAs for instance. In this case, the back gate is not used to modulate the strength of the transistor, but to establish or cut electrical links between a group of transistors. Of interest is that the layout of a re-programmable cell is unique for all functions.

The voltage applied to the different back gates advantageously comes from external, and if possible adjacent, circuits, such as SRAM cells or Flash cells. The programming voltages of the various functions that can be implemented by the re-programmable circuits of the invention can be stored in these circuits. As all Boolean functions can be programmed from this type of cell, then a unique and predefined back-end interconnection can be defined as well. As a result the predefined chip can be fully processed and then programmed by the end user.

FIG. 11 shows an example of a re-reprogrammable logic circuit according to the invention and indicates a possible arrangement of the logic cells, LC, (each logic cell corresponding for instance to the circuit of FIG. 9 ) and of the memory cells, MC, (SRAM or Flash) used to store the programming of the logic cells and to deliver the appropriate voltages to their back gates. In FIG. 11 , strips of logical cells LC are interleaved with rows of memory cells MC. For clarity, the logic cells LC have been schematically represented by a box on FIG. 11 , while the arrows originating from the memory cells MC represent the control of the back gate voltages of the logic cells LC by the memory cells MC.

In FIG. 11 , the memory cells MC use only the metal layers (preferably at levels 1 or 2 ) at relatively relaxed pitches as the dimensions are induced by the transistor pitches in the logic area LC. M 1 -MC represents the metal (preferably at level 1 , or “metal1”) layer used for selecting rows of the memory cells MC, while M 2 -MC represents the metal (preferably at level 2 ) layer used for selecting columns of the memory cells. In particular there is still enough resources at the metal (preferably at level 2 ) level to connect different inputs and outputs of the logic cells LC to a predefined network composed by the upper metal layers.

In FIG. 11 , M 2 -LC represents the metal (preferably at level 2 ) layer used for interconnecting tracks of logic cells LC, while M 3 -LC represents the metal (preferably at interconnect level 3 , or “metal3”) layer used for interconnecting tracks of logic cells LC.

The articles “a” or “an” are used herein in manners that are not limited to “only one,” but can mean “one or more” as appropriate in the context. The words “significant” and “likely” (and similar words of degree) are used here to mean within acceptable and expected limits, usually commercially-acceptable limits. It should be understood that this invention is not limited to commercial uses; intended uses include research uses, special purpose uses, and so forth.

The preferred embodiments of the invention described above do not limit the scope of the invention, since these embodiments are illustrations of several preferred aspects of the invention. Equivalent embodiments are intended to be within the scope of this invention, and indeed, various modifications of the invention in addition to those shown and described herein, such as alternate useful combinations of the elements described, will become apparent to those skilled in the art from the subsequent description. Such modifications are also intended to fall within the scope of the appended claims.