The present invention relates generally to integrated circuit devices and, more particularly, to a piezo-effect transistor device and applications thereof.
Complementary Field Effect Transistors (FETs) support the standard computer architecture (CMOS) currently used in logic and memory. FETs exploit high channel mobility to control few-carrier currents electrostatically. However, limitations in this highly successful technology are appearing at current and future device scales.
More specifically, difficulties in scalability arise from short channel effects and from few-dopant fluctuation effects. The HfO2 gate oxide short channel solution brings about mobility limitations which are slowing clock speeds (Moore's Law scaling becomes negative). The unfavorable FET geometry wherein the gate capacitance corresponds to gate area, but wherein current corresponds to channel width/channel length (resulting in a speed ˜1/L2), means that the FET is a relatively high impedance device. Hence undesirably large-area FETs are required in “power hungry” applications, such as programming a PCM memory, driving long wires, or shutting down power to inactive circuit blocks.
It is desirable but very complex to build multi-layer structures in CMOS, due to the need for all FETs to be formed in single crystal silicon. A new technology in which straightforward lithographic processes can build multilayer structures could open up significant new applications such as high capacity multilayer memories and combinations of logic and memory at different levels optimized to reduce wiring length.
In an exemplary embodiment, a piezo-effect transistor (PET) device, includes a piezoelectric (PE) material disposed between first and second electrodes; and a piezoresistive (PR) material disposed between the second electrode and a third electrode, wherein the first electrode comprises a gate terminal, the second electrode comprises a common terminal, and the third electrode comprises an output terminal such that an electrical resistance of the PR material is dependent upon an applied voltage across the PE material by way of an applied pressure to the PR material by the PE material.
In another embodiment, a piezo-effect, electronic memory storage element includes a first piezo-effect transistor (PET) device coupled to a second PET device in a latch configuration, with the first and second PET devices each comprising a piezoelectric (PE) material disposed between first and second electrodes, a piezoresistive (PR) material disposed between the second electrode and a third electrode, wherein the first electrode comprises a gate terminal, the second electrode comprises a common terminal, and the third electrode comprises an output terminal such that an electrical resistance of the PR material is dependent upon an applied voltage across the PE material by way of an applied pressure to the PR material by the PE material.
In another embodiment, a digital logic gate includes two or more piezo-effect transistor (PET) devices each comprising a piezoelectric (PE) material disposed between first and second electrodes, a piezoresistive (PR) material disposed between the second electrode and a third electrode, wherein the first electrode comprises a gate terminal, the second electrode comprises a common terminal, and the third electrode comprises an output terminal such that an electrical resistance of the PR material is dependent upon an applied voltage across the PE material by way of an applied pressure to the PR material by the PE material.
In still another embodiment, a method of forming a piezo-effect transistor (PET) device includes forming a first electrode; forming a piezoelectric (PE) material over the first electrode; forming a second electrode over the PE material; forming a piezoresistive (PR) material over the second electrode; and forming a third electrode over the PR material; wherein the first electrode comprises a gate terminal, the second electrode comprises a common terminal, and the third electrode comprises an output terminal such that an electrical resistance of the PR material is dependent upon an applied voltage across the PE material by way of an applied pressure to the PR material by the PE material.
Referring to the exemplary drawings wherein like elements are numbered alike in the several Figures:
a) and 1(b) are schematic diagrams of a PET device shown in an n-type configuration and a p-type configuration, respectively, along with a three-terminal symbolic representation thereof;
a) illustrates the molecular structure of a photoconductive, porphyrin derivative known as ZnODEP;
b) is a graph illustrating photocurrent as a function of distance during the compression of a ZnODEP film;
a) and 6(b) illustrate a schematic diagram of a PET-based flip-flop device, in accordance with a further embodiment of the invention;
a) and 7(b) are schematic diagrams illustrating the bistability of a PET-based flip-flop device, such as shown in
a) is a schematic diagram of a PET-based inverter, in accordance with a further embodiment of the invention;
b) is a graph illustrating Vout/V0 versus Vi/V0 for two values of μ of the inverter of
a) is a schematic diagram of a PET-based NAND gate, in accordance with a further embodiment of the invention;
b) is a schematic diagram of a PET-based NOR gate, in accordance with a further embodiment of the invention;
a)-(c) illustrate a mechanical software pressure simulation for an exemplary PET device;
a) through 15(h) are cross sectional views illustrating an exemplary method of forming a PET device.
Disclosed herein is a piezo-effect transistor device and applications thereof that provide a novel solution to fast, lithography-capable and scalable switching needs. The embodiments disclosed herein arise from the observation that good channel conductance should be obtainable from high carrier density materials, even if the carriers do not have high mobility. This is especially true if the current flow were transverse to the device plane, thus escaping from the unfavorable FET geometry. Because certain materials of this type are pressure sensitive, i.e., piezoresistive (PR), their conducting behavior can thus be controlled via pressure produced by a voltage-controlled piezoelectric (PE) element.
A piezoelectric (PE) material either expands or contracts, depending on the polarity of the voltage applied across it. A piezoresistive (PR) material has a high or low resistance depending on its compression. As described in further detail herein, the juxtaposition of a PE material and a PR material in a way that allows the expansion and contraction of the PE material to compress and decompress the PR material results in a sensitive switch in which the resistance in the PR material can be controlled by varying the voltage across the PE material. More specifically, a three-terminal device, with one terminal connected to a thin metallic layer between the PE and PR, another to the far side of the PE and a third to the far side of the PR forms a transistor-like switch that may be used for logic and memory functionalities. Hereinafter, such a device is referred to as a Piezo-Effect Transistor or PET.
Referring now to
In the three-terminal, 5-layered PET device 100 shown in
In total, an exemplary height of the PET device 100 is about 35-120 nm, with dimensions of about 45-90 nm in the x-y plane. Furthermore, the PET device 100 is scalable and many of the problems associated with conventional FET scaling are absent. For example, carrier transport is enhanced by the favorable geometry of the PET, in that current flows transversely through the thin channel film (instead of longitudinally as in the FET). In addition, there are no short-channel effects, as the input is screened from the output by the common electrode. Because the PET does not have a dopant nonuniformity problem, it should be less impurity/geometry sensitive than FETs, due to short mean free paths and efficient screening by the high density of carriers. The PET should have theoretically similar performance to that of FETs (as described in more detail below), and is capable of low ON impedance at very small scales.
The use of continuous transition materials, such as SmSe are expected to pressurize reversibly and their transition speed may be controlled essentially by the velocity of sound, while their materials degradation due to cycling should be minimal. However, the use of materials with a discontinuous transition is also expected to be effective. Still other examples of possible PR materials that experience an insulator-to-metal transition under applied pressure include, but are not limited to: EuNiO3, Ni(S,Se), hexagonal BaTiO3-δ, InSb, and (2,5 DM-DCNQI)2Cu.
With respect to suitable PE materials contemplated for use in the disclosed PET device embodiments, well-known piezoelectric materials include, for example lead-zirconate-titanate (PZT), strontium-doped lead-zirconate-titanate (PSZT), PSN—PMN—PNN—PSZT, PZNT 91/9 and PMNT 70/30 [Y. J. Yamashita and Y. Hosono, Jap. J. Appl. Phys. 43, 6679-6682 (2004)] with piezoelectric coefficients (d33) lying in the range of about 200-1500 pm/V.
Referring now to
Considering for simplicity only the z-component of electric field and stress/strain, and assuming rigid mounting of the top and bottom surfaces of the device 400 in
wherein E denotes the Young's modulus of the given element (EPR or EPE), t denotes film thickness of the given element (tPR or tPE) parallel to the z-axis, A denotes surface area of the given element (APR or APE) normal to the z-axis, and d33 denotes the zz-component of the piezoelectric coupling coefficient of the PE material. Using exemplary values of EPR=EPE≈40 GPa, tPR/tPE≈1/5, an area ratio APR/APE≈1/4, d33=0.6 nm/V, and a reasonable electric field of 0.02 V/nm, the pressure rise is about 1 GPa. The applied voltage will be about 1 volt, with a PE thickness of tPE=50 nm. As will be seen from the more detailed simulations discussed below, 0.6 GPa may be reached (drive voltage 1.6 V), using a piezoelectric with d33=0.37 nm/V, which will scale up to 1.5 GPa using a PE material such as PSN—PMN—PNN—PSZT with d33=0.94 nm/V. In contrast, using an organic PR material such as ZnODEP, only pressures on the order of about 0.22 GPa are needed, and low-power operation at drive voltages as low as 0.24 V is possible.
The intrinsic rate of response in PE films to voltage change depends on the piezo resonance frequency, which is comparable to the sound crossing rate in the film. Although data is difficult to find at the required high GHz frequencies, scaling upward from low frequency data is one possible analysis. In a publication entitled “High Frequency PZT Composite Thick Film Resonators,” Integrated Ferroelectronics, 2004, Vol. 63, pp. 27-33, Duval et al. discuss a linear scaling between the piezo resonance frequency and the inverse piezo film thickness tPE. Inputting bulk data gives 0.25 GHz for tPE=9 micrometers (μm). Their actual data on 7 μm liquid-phase made samples are similar. If extrapolated to a device thickness that would be of interest for digital electronics, such as tPE=50 nm, for example, then the extrapolated piezo frequency is f≈50 GHz. The characteristic time scale is τ=1/(2πf)=5 ps. While this scale is acceptable for a digital device, it is desirable to verify the scale experimentally.
The estimated speed of sound in, for example, SmSe is about 2×105 cm/s. For a 10 nm thick PR film, the transmission time is thus about 5 ps. Electrically, the response time can be estimated as the time for the device to charge the capacitance presented by the input of a similar device (or several such devices). It is useful to compare the response times for the FET and the PET estimated on this basis. The FET time scale τFET, given usually as the time to charge one gate of another similar device, is τFET=2 L2/(μSVD), where μS=mobility, VD=drain voltage, and L=channel length. For the PET, τPET=tPR ρminσ, where tPR=channel thickness, ρmin=minimum PR resistivity, σ=piezo material capacitance/unit area. The ratio of the time scales is therefore:
τFET/τPET=2 L2/(tPRρminσ μS VD)˜1,
where ρmin=3×10−3 Ω·cm, tPR=10 nm, μ=200 cm2/V·s, L=50 nm; and where σ is estimated using a piezo thickness of tPE=50 nm, and a dielectric constant of 4000. Electrically, the speeds of the silicon-based FET and the estimated speeds of the SmSe PET are seen to be comparable.
In absolute terms, the time to charge four (4) similar FET (in this case PET) gates is often used as a standard. The expression for this time constant τ4 can be written as:
τ4≈4ρminεε0 tPR/tPE seconds,
where ρmin is the minimum PR resistivity, and ε the dimensionless dielectric constant. If SmSe is compressed to 2.6 GPa, its resistivity goes down to ρmin˜3×10−3 Ω·cm, giving a time constant of 0.8 ps at ε=4000.
The sound transmission times are seen to limit the response speed of the PET device, based on the foregoing assumptions for an SmSe channel, more than the electrical times.
With respect to the concern of static “OFF” current of PET devices, it is noted that non-switching devices will dissipate static power due to the finite conductance of the OFF state of the PR material. Assuming a total of about 109 devices on a chip, with maximum PR resistivity ρmax=3×103 Ω·cm, so that the resistance of a 40 nm×40 nm device with a PR thickness of 10 nm is about 2×108Ω, then the static chip dissipation at 1V is about 5 W, which is acceptable.
With respect to the dimensionless gain parameter of a PET device, it is assumed for the sake of simplicity that the pressure-dependence of the resistivity in the PR element takes the exponential form:
ρ=ρ0exp(−γpPR),
where ρ0 is the resistivity at zero pressure, and the coefficient γ=−d ln ρ/dpPR is determined from the slope of the curve shown in
where during switching Voc is typically about one half the line voltage V0 and
in the foregoing notation. d ln ρ/dpPR≈−5.5 GPa−1 for SmSe, while dpPR/dV≈1.5 GPa/V with the foregoing parameters and with d33=0.94 nm/V. A voltage of V0=1.7V, is required to reach the maximum pressure pPR=2.6 GPa, when the gain is μ=7 (the gain μ is essentially 1.15×log10(ρmax/ρmin)).
With respect to ON/OFF ratio for a PR element, since the maximum resistivity is set by the static OFF current at ρmax=3×103 Ω·cm, and an ON resistivity of about ρmin=3×10−3 Ω·cm is required for reasonable device speed, it is seen that the ON/OFF ratio needs to be about ρmax/ρmin=106.
An electronic flip-flop (latch) device is generally characterized as having two stable states that can be used to serve as one bit of memory. In order to meet the condition for bistability of a flip-flop, a simple circuit analysis demonstrates that the critical region of gain for bistability is μ>1. As such, a PET device as disclosed above is a suitable building block for a flip-flop device as in the case of the SmSe based PET it is estimated that μ=7 satisfies the bistability constraint.
Referring now to
In operation, the PET that is in the “ON” state has a voltage (of appropriate polarity) close to V0 in magnitude across its PE, while the PET that is in the “OFF” state has only a small voltage across its PE. For the Bit 1 state shown in FIG. 6(a), the voltage applied to the R/W wire is nearly as high as V0 such the voltage drop across the PE portion of the PPET 604 is high, the PE is expanded and the resulting high pressure on the PR portion of the PPET gives it a low resistance. Thus, the PPET 604 is in an ON or conductive state. At the same time, the voltage drop across the PE portion of the NPET 602 is low (i.e., V0 minus the voltage drop across the ON PPET 604). As a result, the PE portion of the NPET 602 is contracted and the resulting low pressure on the PR portion of the NPET 602 gives it a high resistance. Thus, the NPET 602 is in an OFF or nonconductive state. The low resistance of the ON PPET 604 combined with the high resistance of the OFF NPET 602 results in stabilizing the voltage of the commonly tied read/write control terminal 606 high.
For the Bit 0 state shown in
Accordingly, under equilibrium conditions one of the PETs is OFF while the other is ON. Because the PETs are connected in series, similar to the NFET/PFET combination of a CMOS inverter, the current drain on the supply is only that from the OFF PET. As discussed above, provided that the OFF resistivity of the PR is large enough, the resulting static power consumption will be acceptable. The electrical time constant RC, where R is the PR resistance and C the piezo gate capacitance will be ¼ of that estimated for supplying 4 gates above; i.e., 0.2 ps (assuming ρmin=3×10−3 Ω·cm, and a dielectric constant of 4000), which is a negligibly short time. As also indicated above, the lattice response times of the PR and PE, estimated above as a few picoseconds, dominate the switching time scale.
a) shows another schematic of PET-based flip-flop device 600, which specifically illustrates the labeling of variable resistance values R1 and R2, and intermediate control terminal voltage v1 indicated. Considering again the model where R is exponentially dependent on pressure and therefore on voltage (as the case for the exemplary material SmSe): R═R0exp(−αVgc), where α=2μ/V0. Thus for the PPET device, R1═R0exp(−αv1) and for the NPET device, R2═R0exp(−α[V0−v1]).
From Kirchoff's law:
(V0−v1)/v1=R1/R2=exp(αV0)exp(−2αv1)
log(V0/v1−1)=2α(V0/2−v1).
This is plotted at the top right portion of
α>2/V0 or μ>1.
From SmSe data, μ=7, and thus the condition for bistability is satisfied.
In a more practical embodiment of a bistable flip-flop memory cell, at least one additional device is desirable. An example of such a three-transistor (3T), PET-based memory cell 800 is illustrated in
In operation, the 3T cell 800 is isolated whenever the row enable line 810 is at −v potential and the W/S line 812 is at zero (ground) potential, meaning that NPET 808 is OFF and the voltage at the read/write control terminal 806 is stable at either v or −v. In order to write a “0” bit into the cell, the data write/sense (W/S) line 812 is lowered to −v potential while the row enable line 810 is pulsed to a voltage of about 3 v. This turns NPET 808 strongly on and either maintains terminal 806 at −v or pulls terminal 806 from v down to −v, depending on the previous state of the cell. In a memory array, other such cells 800 coupled to the data write/sense (W/S) line 812, but in different rows, would remain stable since their respective row enable lines would remain at the low rail potential.
Conversely, to write a “1” bit into the cell 800, the data write/sense (W/S) line 812 is raised to v potential while the row enable line 810 is pulsed to a voltage of about 3 v. This also turns NPET 808 strongly on and either maintains terminal 806 at v or pulls terminal 806 from −v up to v, depending on the previous state of the cell. Finally, in order to sense or read the state of the cell 800, the data write/sense (W/S) line 812 is coupled to a sense amplifier (not shown) and grounded while the row enable line 810 is pulsed to a voltage of about 3 v. This turns NPET 808 strongly on and charges the W/S line 312 to the voltage of terminal 806, with is either at −v or v down to −v, depending on the state of the cell. Again, this operation is isolated from other cells along the same column but in different rows.
Referring now to
Thus, in this embodiment, the flip-flop portion of the cell does not drive a sense line directly, but instead through a fourth PET 914 (e.g., a PPET). As further shown in
In operation, the 4T cell 900 is isolated whenever the row enable line 910 is at −v potential and the W and S lines 912, 916 are at zero (ground) potential, meaning that both NPET 908 and PPET 914 are OFF and the voltage at the read/write control terminal 906 is stable at either v or −v. The writing of the cell 900 is similar to that of cell 800, in that the row enable line 910 is pulsed to about 3 v while either a +v or a −v voltage is applied to the data write (W) line 912. To sense the state of the cell, the write line (W) 912 is raised to +v, while the sense line (S) 916 is grounded. The row enable line 910 is only pulsed to +v in this instance. Whereas NPET 908 remains OFF, the PPET 914 will turn ON if the cell voltage at the read/write control terminal 906 is at −v. In such a case, the voltage on the sense line (S) 916 will charge up to +v. On the other hand, the PPET 914 will remain OFF if the cell voltage at the read/write control terminal 906 is at +v, in which case, the voltage on the sense line (S) 916 will remain at ground.
In addition to the novel circuit topologies for the above discussed bistable flip flop, 3T cell and 4T memory cell embodiments, other types of existing logic circuits topologies can be formed by using the PET devices. For example,
Additional basic PET-based logic gates are illustrated in
In
As will thus be appreciated, piezo-effect transistors (PETs) have significant advantages over conventional FET transistors. For one, PETs are highly scalable, as the structures are simple, and many FET scaling problems are absent. There are no short-channel effects, as the input is screened from the output by the common electrode. The PET does not have a dopant nonuniformity problem. Moreover, the PET has a low impedance in its ON state as carrier transport is enhanced by the favorable geometry, wherein current flows transversely through the thin channel film (instead of longitudinally as in the FET). PETs should also be less impurity and geometry sensitive than FETs due to short mean free paths and efficient screening. In certain embodiments, a PET SRAM has only one half the transistor count of a corresponding FET SRAM. Furthermore, a PET manufacture process allows for multiple layers of devices, since there is no requirement for a high-mobility monocrystalline substrate. In principle, the PET can operate at low voltages as there is no intrinsic limitation on voltage, the piezo device is linear, and PR behavior is limited only by material properties.
The PET may be used in general computing applications. The sound propagation delay of a few picoseconds appears to be a main limitation on speed. Speed competitiveness needs to be judged in the context of actual achievable speed in the next litho technology generation. Due to its low ON output impedance the PET (even in the minimal device configuration) can sustain a high fanout, which can probably be used in a logic redesign approach to mitigate the intrinsic sound propagation delay.
The PET has clear advantages in the power driver context due to its low impedance. Examples include, but are not limited to, driving long wires, programming PCM memory, and control of static power in temporarily unused circuit blocks by switching off the power supply thereto.
With optimization of PR and PE materials to increase the gain μ, there is the possibility to reduce voltage and thus power/heat, which goes as the square of the voltage. The PET is not limited by the constraints which make it very difficult to further reduce FET voltages.
Referring now to
The nitride clamp structure 1206 forms a rigid frame so that the electrically induced displacement of the PE material 1212 is mechanically coupled to (and focused primarily towards) the PR material 1214. Tungsten forms the conducting electrodes (leads not shown), and is also mechanically rigid, while the low-K buffer structure 1204 (being a soft material) does not impede the operating displacements significantly.
b) shows the stress distribution of the simulated structure 1200 when 1.6 V is applied to the PE material 1212 with a resulting electric field of 0.02 V/nm. It is noted that a contraction (positive pressure) of the PE element 1212 results in an expansion (negative pressure) of the PR element 1214 and vice-versa. It will be seen from
a) through 15(h) are cross sectional views illustrating an exemplary method of forming a PET device. As shown in
d) illustrates the formation of the piezoelectric (PE) layer 1510 (e.g., PSZT) within dielectric layer 1508 and atop the bottom metal contact 1506. The PE layer formation may be by a blanket layer, by part of via fill prior to planarization, or by an extra patterning and etch step to form a discrete piezoelectric pad, for example. In addition, polling of the PE layer may 1506 be implemented at this point. Next, another dielectric layer 1512 is formed over the PE layer 1510 and M1 stud 1509 in
As further shown in
Finally,
While the invention has been described with reference to a preferred embodiment or embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof without departing from the scope of the invention. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the invention without departing from the essential scope thereof. Therefore, it is intended that the invention not be limited to the particular embodiment disclosed as the best mode contemplated for carrying out this invention, but that the invention will include all embodiments falling within the scope of the appended claims.