Electrically actuated switch

Electric switch

[0001] Government interest statement

[0002] The present invention was completed with government support under the agreement number HR0011-0503-0001 granted by the Defense Advanced Research Projects Agency. The government has certain rights in the invention.

Technical field

[0003] The present invention relates to electronic switches, and particularly to electric switches.

Background technique

[0004] In the past decade, the prospect of manufacturing electronic circuits based on molecular or molecular-scale components has attracted the interest of many researchers. The potential of molecular electronics has brought the capabilities of synthetic chemistry to the electronics industry by making special devices that can at least partially assemble themselves into useful circuits. This expectation has led to a large amount of research on the electronic properties of molecules and evidence-based demonstrations of certain principles of memory and logic circuits.

[0005] Research on switches in nano-scale cross-wire devices has been previously reported, which can be reversibly switched and has an on-off conductance ratio of 10. These devices have been used to construct crossbar circuits and provide a promising way for the production of ultra-high-density nonvolatile memories. Series-connected crossbar switches that can be used to make latches have been demonstrated; such latches are an important component for logic circuits and for communication between logic and memory. A new logic series has been described, which can be constructed entirely from a crossbar array or as a hybrid structure composed of switches and transistors. These new logic series have the following potential: to greatly improve the computational efficiency of CMOS circuits, which can achieve an order of magnitude performance improvement without having to shrink transistors; or even for some applications, it can replace CMOS if necessary. However, it is desired to improve the performance of the currently manufactured devices, especially to improve their cyclability.

Summary of the invention

[0006] An electric switch includes a first electrode, a second electrode, and an active region arranged between them _oThe active region may include: at least one primary active region, which includes at least one material for transporting and hosting (host) ions used as a dopant to control the flow of electrons through the switch; and a secondary active region, which includes At least one material used to provide a source of ion dopants for one (or more) main active regions. A method of operating the switch is also provided.

Description of the drawings

[0007] FIG. 1A is a perspective view of an example of a solid state switch connecting two different crossover wires;

[0008] FIG. 1B is a view similar to FIG. 1A, showing the array of switches of FIG. 1A, also called a crossbar;

[0009] FIG. 2 is a perspective view of an example of a solid-state switch connecting two segments;

[0010] FIG. 3A is a perspective view schematically showing the structure of an exemplary electric switch based on the sectional line configuration of FIG. 2; [0011] FIG. 3B is a schematic representation of the structure of an exemplary electric switch based on the cross-line configuration of FIG. 1A [0012] FIG. 4 is a perspective view of a schematic representation of the start process of driving (or switching (toggling)) an exemplary switch to an ON state based on the segment line configuration of FIG. 3A ;

CN 101548403 Β

[0013] FIG. 5 is a perspective view of a schematic representation of a starting process for driving (or switching) a switch to an OFF state;

[0014] FIGS. 6A-6C are schematic diagrams of an exemplary switch in an off state, wherein FIG. 6A is a view of the switch, FIG. 6B is an equivalent circuit diagram of the state of the switch, and FIG. 6C is an energy band diagram associated with the switch ;

[0015] FIGS. 7A-7C are schematic diagrams of an exemplary switch in an intermediate state, wherein FIG. 7A is a view of the switch, FIG. 7B is an equivalent circuit diagram of the state of the switch, and FIG. 7C is an energy band diagram associated with the switch;

[0016] FIGS. 8A-8C are schematic diagrams of an exemplary switch in an on state, in which FIG. 8A is a view of the switch, FIG. 8B is an equivalent circuit diagram of the state of the switch, and FIG. 8C is an energy associated with the switch With pictures; and

[0017] FIGS. 9A-9C are schematic diagrams of another embodiment of a switch, which depicts the movement of dopants.

Detailed ways

[0018] These figures depict embodiments of the invention for illustration purposes only. Based on the following discussion, those skilled in the art will readily realize that other embodiments of the structures and methods described herein can be adopted without departing from the principles of the invention described herein.

[0019] Definition

[0020] As used herein, the term "self-aligned" as applied to "junctions" means that no matter where two wires cross each other, a switch and/or other electrical connection is formed between the two wires. The knot, because it is this cross-action that produces the knot, any of the wires can be coated or functionalized.

[0021] The term "self-assembly" as used herein refers to a system that naturally adopts a certain geometric pattern due to the identity of the system components; the system adopts this configuration to at least achieve its energy partial Minimal value.

[0022] The term singly configurable means that a switch can only change its state once through an irreversible process (such as an oxidation or reduction reaction); such a switch can, for example, become the basis of a programmable read-only memory (PROM).

[0023] The term "reconfigurable" means that a switch can change its state multiple times through a reversible process (such as oxidation or reduction); in other words, the switch can be turned on and off multiple times, such as random access memory ( RAM) storage bits.

[0024] The term "configurable" means "single configurable" or "reconfigurable".

[0025] The micron size refers to a size ranging from 1 micrometer to several micrometers.

[0026] Sub-micron size refers to a size ranging from 1 micron down to 0.04 microns.

[0027] Nano-scale dimensions refer to dimensions ranging from 0.1 nanometers to 50 nanometers (0.05 micrometers).

[0028] Micron and sub-micron wires refer to rods or ribbon conductors with a width or diameter of 0.04-10 microns, a height that can vary from a few nanometers to a micron, and a length of several microns and longer. semiconductor.

[0029] A crossbar is a switch array that can connect each line in a set of parallel lines to each member of a second set of parallel lines that intersect the first set (usually the two sets of lines are perpendicular to each other, but this is not Necessary conditions).

[0030] As used herein, the functional size of the device is measured in nanometers (generally less than 50 nm), but the lateral dimension can be nanometer, submicrometer, or micrometer.

[0031] Background on nanowire junctions

[0032] FIG. 1A shows an example of a solid-state switch 100 connecting two different crossover wires 102,104. The switch junction 106 can be used to connect these two different wires 102.104. The switch junction 106 can include at least one material 106a, specifically a switchable molecule (that is, a switchable segment or part of a switch that has relatively stable energy in two different states). Numerator); this kind of

CN 101548403 Β

Examples of subs are now well known in the field of molecular electronics (moletronic) switching. One or two wires 102.104 can be metal or semiconductor. In many cases, both wires are metal, specifically clamps, and a thin titanium layer 106b is formed on the molecule 106a, and then a platinum top wire is formed on the titanium layer 104_o[0033] FIG. 1B shows a crossbar array 110 employing a plurality of solid-state switches 100 shown in FIG. 1A. As shown in FIG. 1B, the approximately parallel lines 102 of the first layer 112 are overlapped by the approximately parallel lines 104 of the second layer 114. The second layer 114 is substantially perpendicular to the lines of the first layer 112 in orientation, but the orientation angle between these layers may vary. These two layers of lines form a lattice or crossbar, and each line 104 of the second layer 114 is superimposed on all the lines 102 of the first layer 112 and intersects with the first line at the intersection of the lines representing the closest contact between the two lines. Each line of the layer is in close contact. The switch junction 106 shown is arranged between the lines 102,104. (Three such switch junctions are shown so as not to mess up the figure; it should be understood that the switch junction 106 is formed at each intersection of the line 102 and the line 104.) Such a crossbar can be made of micron, sub-micron or nano-scale The thread is made, it depends on the application.

[0034] Although each line in the figure is shown as having a square or rectangular cross-section, the lines can also have a circular, elliptical or more complex cross-section. These lines can also have many different widths or diameters and aspect ratios or eccentricities. The term "nanowire crossbar" may refer to a crossbar having one or more layers of sub-micrometer-scale wires, micrometer-scale wires, or larger-sized wires in addition to nanowires.

[0035] FIG. 2 is an example of a solid-state switch connecting two segments of nanowires. Here, two segments of 102.104 nanometer wires are switched

106 is connected. The switching junction 106 includes a switching layer 106a and a thin titanium layer 106b, both of which are as described above.

[0036] It is well known that such a switch can be used as the basis of a memory (such as a storage device for information bits of 1 or 0), as a close or open switch in a cross-point memory for logic similar to a field programmable gate array The configuration bits in the circuit, or used as the basis of the wiring logic programmable logic array. These switches can also be used in a variety of other applications.

[0037] Electric switch

[0038] According to the teachings herein, an improved electric switch is provided, which can be manufactured at the micrometer or nanometer scale and used as a component in a variety of electronic circuits. The switch can be used to connect two different wires, such as a pair of cross wires as shown in Figure 1A-1B. The two different wires can be two wires in a larger crossbar array; or the switch can be used in-line In order to allow or block the flow of current along the line, as shown in Figure 2.

[0039] The improved switch can be used as the basis for the memories, switches, and logic circuits and functions described above.

[0040] These switches may have the following characteristics:

[0041] (1) The main active layer or zone of the switch includes a thin film material that is electronically semi-conductive or nominally electrically insulated but is also a weak ion conductor. The primary active material can transport and host ions used as dopants to control the flow of electrons through the switch. The basic mode of operation is to apply a sufficiently large electric field (drift field, which can exceed a certain threshold for allowing ions to move in the host material) on the switch so that ion species (species) are transported to the host via ion transport. In or out of the main material. The ion species is specifically selected from those ions used as the electrical dopant of the main material, so that the electrical conductivity of the material is changed from low electrical conductivity (ie undoped semiconductor or insulator-switch-off configuration) To high conductivity (doped to provide higher conductivity-switch-on configuration) or change from high conductivity to low conductivity (switch-on to switch-off). Moreover, the host material and dopant species are selected so that it is possible but not too easy for ions to drift into or out of the host material, thereby ensuring that the switch remains in any state in which it is set to be equivalent. For a long time, it may be kept at room temperature for many years. This ensures that the switch is non-volatile, that is, after removing the drift field, the switch

CN 101548403 Β

Keep its state. The switch is a two-terminal device-applying a high bias to the switch will cause both electronic current and ion current to flow, while at low bias the flow of ion current is negligible, which allows the switch to maintain its resistive state.

[0042] (2) The secondary active layer or region includes a thin film material as a dopant source of the main material. These dopants can be impurity atoms such as hydrogen or some other cationic species, such as alkali metals or transition metals, which are used as electron donors of the host material; or these dopants can be anion vacancies, which are charged in the host material Is therefore also the donor of the lattice. It is also possible to drive anionic species into the host host material, which will become an electron acceptor (or hole donor).

[0043] (3) The main active material may be a thin film (thickness is generally less than 50 nm), and in many cases it is nanocrystalline, nanoporous or amorphous. The mobility of dopant species in such nanostructured materials is much higher than that in bulk crystalline materials, because diffusion may occur through grain boundaries, pores, or through local structural defects in amorphous materials. In addition, because the film is so thin, the amount of time required to drift enough dopants into or out of a local area of the film to substantially change its conductivity is relatively fast (for example, the diffusion process The required time t varies with the square of the covered distance, so the time required to diffuse one nanometer is one millionth of the time required to diffuse one micron).

[0044] (4) The switching materials (primary active material and secondary active material) are contacted by metal electrodes or wires on each side, or contacted by a semiconductor on one side and metal on the other side. The contact of the metal to the switching material depletes the free charge carriers of the semiconductor, so in fact the material has a net charge, which depends on the characteristics of the dopant-positive in the case of the donor and positive in the case of the acceptor Is negative. The metal-semiconductor contact area is electrically similar to a Schottky barrier. The traditional description of the metal-semiconductor Schottky barrier is modified by the fact that these materials are structured at the nanometer level, so the structural properties and electrical properties are not even over large distances, while the theory of semiconductor-metal contact is Developed over a long distance.

[0045] (5) The conduction of electrons through the main active material is through the quantum mechanical tunneling effect of electrons. When the semiconducting material is intrinsic in nature, the tunneling barrier is high and wide, so the conductivity through the switch is very low (off state). When a considerable amount of dopant species is injected into the semiconductor, the potential of the charged species reduces the width and perhaps height of the tunneling barrier. This leads to an increase in the conductivity of the switch (on state).

[0046] (6) If one of the interfaces between the connection switch and the metal or semiconductor electrode is non-covalently bonded, the ability of charged species to diffuse into and out of the main material is greatly improved. Such an interface may be caused by voids in the material or it may be caused by an interface containing a molecular material that does not form a covalent bond with the electrode, the main switch material, or both. This non-covalently bonded interface reduces the activation energy for the rearrangement of atoms required for drifting ionic species in the host material. This interface is essentially an extremely thin insulator, and the total series resistance of the switch increases little.

[0047] As mentioned above, the master active material has certain properties for practicing the present invention. One of these material properties is that it is a weak ion conductor. The definition of weak ionic conductor is based on the application for which the switch is designed. The mobility and diffusion constant of the species in the crystal lattice are directly proportional to each other via the "Einstein relationship". Therefore, if the mobility of ionized species in the crystal lattice is high, the diffusion constant is also high. Generally speaking, it is expected that the switching device stays in a certain state (on or off) for an amount of time, which can vary from a fraction of a second to several years, depending on the application. Therefore, in the embodiment, the diffusion constant of such a device is low enough to ensure the desired stable level, so as to avoid the device from inadvertently changing from on to off or from off to on through the diffusion of ionized species, Instead of deliberately setting the state of the switch with voltage pulses. Therefore, a "weak ion conductor" is a type in which the ion mobility and thus the diffusion constant are small enough to ensure that the on or off state of the device is stable under the desired conditions for a necessary long time (for example, the device will not be due to the dopant The spread of

CN 101548403 Β

State) conductor. "Strong ion conductors" have high mobility of ionized species and are therefore unstable for diffusion.

[0048] FIG. 3A shows the structure of an embodiment of the switch 200' of the present invention, which adopts the in-line configuration of FIG. 2. FIG. 3B shows the structure of another embodiment of the switch 100' of the present invention, which adopts the cross-point configuration of FIG. 1A. [0049] FIGS. 3A-3B are illustrations of the structure of an exemplary electric switch. The active area 306 of the switch has two main parts: the main active layer or area 306, which is a material that is both an electrically semiconducting and weak ion conductor layer. This layer can be used as interstitial, vacancy or The electron donor of the impurity is doped; and the sub-layer or region 310, which serves as a source and sink of the doped species. Alternatively, the material of the main active region 308 may be both a nominally electrically insulating and a weak ion conductor. As discussed below, the active region 306 may optionally include a molecular layer 106a as an interface for non-covalent bonding; this molecular layer may include molecules that may be switchable or may not be switchable.

[0050] In the embodiment of the switch 200' depicted in FIG. 3A, the two electrodes 102, 104 are both clamp (Pt) metal, and the semiconducting ion conductor 308 is Ti0.₂, The dopant source 310 is Ti0₂__x, And the optional non-covalent interface includes a thin molecular layer 106a between the electrode 102 and the main switching material 308. The chemical formula TiC^p indicates that the oxygen atom of titanium dioxide is slightly insufficient in the crystal structure; the position where the oxygen (0) atom is missing will be a positively charged vacancy. In this case, the dopant species is oxygen vacancy. Another embodiment of such a switch is shown in FIG. 3B as a switch 100, which has the same active area switch structure 306 as in FIG. 3A.

[0051] By increasing or decreasing the number of donor dopant sites in the semiconductor layer 308 respectively to turn on (that is, increase the conductivity, FIG. 4) or turn off (decrease the conductivity, FIG. 5) operation via donor injection or discharge switch. These donor sites may be interstitial cation species in the host lattice of the semiconductor, anion vacancies in the host lattice, interstitial impurities such as H, or other interstitial or substitute impurities as electron donors of the semiconductor. Since the thickness of the semiconductor layer 308 is smaller than the depletion width of carriers in a moderately highly doped semiconductor (for example, a thickness of about 100 nm), this means that the thin semiconductor is completely depleted of free carriers The reason is that the thin semiconductor is in close contact with the metal on at least one side. These electrons only reside in the metal contacts 102, 104 on each side of the semiconductor 308. For example, every cm³ 10¹⁸The doping concentration of one dopant atom (which is "moderately high") corresponds to a depletion layer thickness of 30 nm. 10²¹cm-³The doping concentration (this is a very high doping, but is entirely possible for many oxides) corresponds to a depletion thickness of 3nm.

[0052] FIG. 4 is a schematic diagram of a starting process for driving (or switching) a switch to an on state. In this case, by applying a positive bias 420 to the junction, a positively charged donor (such as the oxygen vacancies in Ti) is driven from the source/sink material 310 into the semiconductor material 308, where the positive The bias voltage exceeds the threshold field for drifting ionized species. Since the semiconductor layer 308 is completely depleted of charge carriers, this means that the semiconductor layer 308 obtains a net positive charge. The net charge neutrality is maintained by the electrons residing in the metal layer 102.104. For the example discussed above, use stoichiometric Ti0₂The fabricated layer 308 forms the initial state (off state) of the switch. The potential difference between the metal contact and the main semiconductor forms a tunneling barrier to the current between the Pt electrodes. The TiOgp layer 310 contains a relatively high concentration of vacancies, and therefore is a very good conductor. After applying a positive voltage to the electrode 104, the electrode 104 becomes the anode of the electrochemical cell. Oxygen vacancies are driven out of Ti0₂__xLayer 310 (which is oxidized) and enters Ti0₂Layer 308 (which is restored), as indicated by arrow 422. To Ti0₂__xThe influence of the conductivity of layer 310 is relatively small, because a small number of vacancies are pushed out of this area, but Ti. ? The conductivity of layer 308 increases sharply (switch on) due to Ti0₂The layer 308 changes from a state with no vacancies to a state with some vacancies.

[0053] In the off state, the thin semiconductor layer 308 is essentially intrinsic-for example, there are very few dopants in the crystal lattice. In this case, there are essentially Schottky barriers on both sides of the semiconductor 308, and the semiconductor energy band is relatively

CN 101548403 Β

The alignment to the Fermi energy level of the metal is about the mid-gap, which depends on various physical and chemical issues. For e.g. Ti0₂__xThe oxide semiconductor (which is also a weak ion conductor), the off state of the device is the "full oxidation" state, for example, the stoichiometry of the material is as close to Tit as possible)? And there is a state with little impurity or defect doping . For a semiconductor with a suitably wide band gap, there is a large tunnel barrier for the flow of current, which makes the conductivity of the turn-off switch very low.

[0054] FIG. 5 is a schematic diagram of a starting process for driving (or switching) a switch to an off state. By applying a negative bias 420 on this junction, positively charged donors are driven from the semiconductor material 308 into the source/sink material 310, where the negative bias 420 exceeds that used to drift the ionized species (which may be vacancies) The threshold field. The semiconductor layer 308 loses its net positive charge and becomes neutral again. For the example discussed above, there is a slight deviation from the stoichiometric TTI.₂__δThe conduction on state of the switch is formed because the net positive charge in this layer narrows and reduces the tunneling barrier for current flowing between the electrodes 102.104. The TiOgp layer 310 contains a relatively high concentration of vacancies, so it is still a fairly good conductor. After applying a negative voltage to the electrode 104, the electrode 104 becomes the cathode of the electrochemical cell. Oxygen vacancies are driven out of ΤΪ0₂__δLayer 308 (which is oxidized again to Ti0₂) And back to Ti0₂__xThe dopant source layer 310 (which is reduced), as indicated by arrow 424. To Ti0₂__xThe conductivity of layer 310 is relatively small, but Ti0₂The conductivity of layer 308 drops sharply (switch off).

[0055] Without subscribing to any particular theory, conduction through the host material appears to be through the quantum mechanical tunneling effect of electrons. When the semiconductor material is intrinsically intrinsic, the tunneling barrier will be very high, so the conductivity through the switch is very low (off state). When a considerable amount of dopant species is injected into the semiconductor, the potential of the charged species reduces the tunneling barrier. This increases the conductivity of the switch (on state).

[0056] FIGS. 6A-6C, FIGS. 7A-7C, and FIGS. 8A-8C are schematic diagrams of the switch 200' in different switching states. In each group of diagrams, the "A" diagram is the real space diagram or diagram of the switch 200, the "B" diagram depicts the equivalent circuit diagram, and the "C" diagram is the energy band diagram.

[0057] In this embodiment, the exemplary thickness τ of the switch is 2 nm and the TiO₂/TiO₂__xExemplary total width w of layers 308, 310₀It is 3nm.

[0058] In FIG. 6A, for example, the switching materials 308 and 310 may be anatase (polytypes of titanium dioxide) nanocrystals sandwiched between two Pt electrodes 102 and 104. For the purpose of this discussion, it is assumed that the size of this nanocrystal is 1.9nmX1.9nmX2.9nm and contains a total of 75 anatase unit cells or 900 atoms, which shows that it is very small for the energy band structure to be developed. But it is great for quantum chemistry calculations. At this stage, the energy band method can be used to outline the qualitative picture of the electronic state of the system, but this should not be taken too seriously. A single impurity atom or vacancy in this nanocrystal will produce 10²⁰cm-³And it can be assumed that the nanocrystal can contain up to 2% oxygen vacancies (for example ~12). The metal-semiconductor contact area is electrically similar to the Schottky barrier modified by a nanoscale system, so the electrical properties will not be averaged over a large distance, and the theory of semiconductor-metal contact is developed over a large distance.

[0059] Ti0 adjacent to the non-covalent region 106a₂Region 308 is stoichiometric and therefore highly resistive, as indicated by the large tunneling barrier. Ti0 adjacent to the second electrode 104₂__xRegion 310 is highly oxygen-depleted, so this region is highly conductive. In the energy band diagram at the top of Figure 6C, interface 60K603 is both Schottky barriers. On the left is the metal-insulator-semiconductor interface 601 and on the right is the interface 603 between the metal and the highly-donor-doped semiconductor. The band gap of the highly-donor-doped semiconductor has dropped a little from the value of the stoichiometric Tit)?.

[0060] Pt-TiO on the left hand side of the diagram of FIG. 6C₂The interface 601 represents a non-covalent bonding interface between the metal electrode 102 and the semiconductor 308, which is mediated by the intervening molecular layer 106a. The approximate energy band diagram depicted indicates that there is a potential barrier at this interface 601. This area of the film is designed and made to be stoichiometric Ti0_2OAnatase

CN 101548403 Β

The electron affinity is estimated to be about 4. OeV, and the work function of Pt is 5.64 eV, so the barrier height of 1.6 eV can be estimated at this interface 601. The Pt Fermi level should be near the intermediate band gap region of anatase. The bulk band gap of anatase is 3.2 eV, but it may be wider for nanocrystals due to quantum confinement. Even for 10²¹cm-³The depletion length in anatase is ~3nm, so the nanocrystals are completely depleted and are therefore positively charged. These electrons are located in the narrow area of the Pt contact.

[0061] On the right-hand side of the illustration of FIG. 6A, there is a second Pt contact 104 with the nanocrystal. In an embodiment, the system is designed and made so that this region of titanium dioxide has a considerable concentration of oxygen vacancies. Therefore, when moving from left to right in Figure 6A, Ti0₂__xThe value of X changes from zero to a value as high as 0.04. This is a high enough concentration of oxygen vacancies to generate a large number of donor states that significantly broaden the density of states near the edge of the conduction band and thus effectively narrow the band gap. This is shown in the energy band diagram of Fig. 6C as the potential barrier at this interface 603 has dropped compared to the left-hand side. This is due to the narrowing of the band gap caused by the extremely high concentration of positively charged donor states and the significant energy The belt is bent.

[0062] In the as-fabricated state shown in FIG. 6C, the barrier 601 represented by the anatase nanocrystals is large and therefore the conductivity of the switch in this state is very low; this is switch-off Off state. If the left-hand Pt electrode 102 is grounded and a positive potential is applied to the right-hand electrode 104 (the right-hand electrode 104 becomes the anode of a nanoscale electrochemical cell), the positively charged oxygen vacancies will be forced to drift toward the ground electrode (which is the cathode). The fact that there is already a large concentration of vacancies near the anode means that there is no need to create vacancies in order to turn on the switch. The formation of oxygen vacancies in nanocrystalline anatase is much greater than the activation energy of vacancy jumping, so it has a structure in which there are pre-existing oxygen vacancies (or other appropriate positively charged dopants) near the anode of the switch. This reduces the amount of energy consumed during the preconditioning of the switch to form electrical vacancies.

[0063] In FIGS. 7A-7C showing the initial stage of turning on the switch, moving the oxygen vacancy away from the anode means that Ti(U) is oxidized, that is, the value of X is reduced by a small amount of ε. Since the vacancy concentration near the anode 104 is initially larger Therefore, this micro-oxidation has little effect on the conductivity of the region 310. On the other hand, the initial stoichiometry or oxidized material near the cathode is reduced, for example, the stoichiometry is now TiO-. This has an impact on the energy band in the middle of the nanocrystal The influence of is quite large, as shown in Figure 7C. The total effective barrier width of electron tunneling through the nanocrystal is reduced and the conductivity is increased.

[0064] If the bias is increased or applied for a longer time, more oxygen vacancies will drift to the left and narrow the tunneling barrier even more, as shown in FIGS. 8A-8C. The switch is now fully on, and more vacancies are now in 308, so the stoichiometry of this area is marked as Τΐ0₂__δοReversing the polarity of the voltage on the right-hand electrode 104 reverses the sense of the anode and cathode and causes the oxygen vacancy to drift back to the right-hand side, restoring the state of Figures 6A-6C and if this reverse polarity bias is applied long enough The time to make the switch back to off.

[0065] Without agreeing to any particular theory, the molecules in the non-covalent region 106a of the device seem to only play a passive role. They form a thin insulating region between the first Pt electrode 102 and the nanocrystal 308, which can assist the formation of the Schottky barrier. The fact that the interface 601 of the nanocrystal is not covalently bonded to the Pt electrode may make it easier for oxygen vacancies to drift toward this interface because there is essentially an internal free surface that can accommodate lattice distortion. Finally, although the crystal 308.310 introduced here is considered to be a nanocrystal, the main function of the molecule 106a may be to ensure that the main active material (such as titanium dioxide) 308 is actually amorphous.

[0066] As shown in FIGS. 6B, 7B, and 8B, the equivalent circuit is just two series resistors: a resistor, which represents the wiring of the switch and the series resistance of any component that does not change with time; and the series resistance of any component that does not change with time. Variable resistor R_v(t), the change depends on the applied voltage and the current flowing through it. Assuming that Ohms law applies (instead of tunneling resistance that actually exists, tunneling resistance complicates the mathematics drastically),

[0067] V (t) =1 (t) [R_s+R_v (t) ]. 1.

CN 101548403 Β

The variable resistance is related to the stoichiometric Ti present in the nanocrystals. ? 308 is proportional to the width w(t), R_v(t) = P w(t)/A,

2.

The material after the drift front of the vacancy is considered to have a resistance of essentially zero, and P is the

[0068]

[0069]

[0070]

[0071] The resistivity of doped titanium dioxide, and A is the area of the nanocrystal orthogonal to the direction of current flow. For the initial condition in which the switch is off and a bias voltage is applied to turn it on, the width change rate of undoped titanium dioxide is just the vacancy drift rate in the electric field applied over the undoped width, dw

[0072] ---=- E(t) = -jUypI(t)/A, dt3.

[0073] where eye is the mobility of oxygen vacancies in the titanium dioxide, E(t) is the electric field on the undoped material, which is exactly the voltage drop Pw(t)I(t)/A divided by the width w(t), The negative sign appears because the width of the undoped region is decreasing. Integrating equation 3, this becomes:

[0074] W(t)=W°

<img file="CN101548403B_D0001.tif" />

dt, Eightο4.

[0075] where w₀Is the width of undoped titanium dioxide at t = 0, and w(t) is defined so that it cannot be negative. on

The entire equation of V and I becomes:

[0076] v(t) = I(t) [R_s + Hp(w.-policej I(t) dt)].

^wo ^A i5.

[0077] Thus, the assumptions made above produce quite complex mathematical expressions, where the voltage of the switch at time t involves the integration of current and/or voltage over time, which introduces path dependence to the switch. Equation 5 illustrates a wider variety of descriptive equations for switching operations, and the present invention includes ion drift switching mechanisms described by other equations. For example, if space charge is important for ion drift, the functional dependence of current in Equation 5 varies from the integral of the current first power (I) to the second power (I²) Points. The most efficient circuit model for such a system may require a memristor. As can be seen in the sequence of changing from FIG. 6B to the equivalent circuit of FIG. 7B to FIG. 8B, the switch is closed, so w(t) and thus R,t) become smaller over time; that is, The resistance drops to turn on the switch.

[0078] FIG. 6C shows an energy band diagram of the off state of a switch with metal contacts 102 and 104 on both sides. Because there is substantially no doping in the semiconducting region 308, there is a relatively high barrier at the two metal-semiconductor junctions. Note that the switch is not strictly a binary operation-the conductance of the switch can vary over a wide range, depending on how many dopant species are injected into the host material or how many dopant species are generated in the host material. The donor species are moved by applying a bias on the junction (alternatively, they can be formed by chemical methods, for example by reducing Ti0 with metal.₂To form TiO2p or remove by reacting with molecular oxygen). Thus, there are no free carriers in the semiconductor layer, so it acquires a net positive charge. The overall effect of this positive charge is to bend the energy band of the semiconductor downward relative to the Fermi level of the metal (Figure 7C), which in turn reduces the tunnel barrier of the junction and thus increases the conductivity. The production or implantation of these dopants is greatly enhanced by having a suitable source layer 310 adjacent to the semiconductor 308. The material 310 may be, for example, an ion conductor, which can serve as a source or sink of atoms of the semiconductor 308 (for example, different dopant types or 0 atom sinks to enable the formation of oxygen vacancies in the oxide semiconductor layer).

[0079] FIG. 7C shows that more dopant sites are injected into the semiconductor 308 to form a TTI.₂__εThe energy band of the system afterwards. The energy band of the semiconductor is further bent downward due to the increase of positive charge in the depleted semiconductor, further reducing the tunneling

CN 101548403 Β

Potential barrier and increase the conductivity. At Ti0₂__xIn the case of, the oxygen (0) vacancy in the crystal lattice produces a state very close to the conduction band, so at a sufficiently high doping level, the conduction band of the nanocrystal will be pulled down to the Fermi energy of the metal contact level. At this stage, the tunnel barrier at each metal junction is getting thinner, and the conductivity of the system is increasing. This is the initial stage of turning on the switch.

[0080] In FIG. 8C, even more dopant donors are implanted into the semiconductor 308, thereby forming a TTI.₂__δο The semiconductor conduction band is further bent downward due to the increase of positive charge in the depleted semiconductor, thereby reducing the width and height of the tunnel barrier and increasing the conductivity of the layer. This is the switch of the system-fully on state.

[0081] As long as the doping level of the semiconductor film 308 is not so large that the resistance drops to substantially zero, it is possible to reverse the drift of the donor dopant ions by reversing the polarity of the voltage applied to the switch and to reverse the semiconductor Exhaust donor dopant ions.

[0082] The doping of the semiconductor layer 308 is a function of both voltage (any energy barrier that exceeds ion drift) and time (the longer the system maintains the voltage, the more dopants are accumulated), or the integral of the current.

[0083] In the above description, these switches are designed and made to have specific on/off polarities. The switch is turned on by applying a positive bias to an electrode with too many positively charged dopant species, and the switch is turned off by applying a negative bias to the same electrode. However, it is also possible to design and fabricate switches with configurable polarities, for example to determine the switch's on/off polarity during an electronic configuration process performed after the circuit containing the switch is fabricated. This configuration step is also called field programming. This can be done by creating a structure where the material adjacent to the two metal electrodes is not doped, and the material has too much dopant in the middle. Thus, during the initial programming phase of the circuit, a negative bias for attracting positively charged dopants can be applied to those electrodes that one expects to be on electrodes for the applied positive bias, and vice versa. A positive voltage that repels the positively charged dopant can be applied to those electrodes that one expects to be off electrodes with respect to the applied negative bias voltage. In this way, the switch array can be configured such that the on/off polarity of the upper electrode of the switch alternates. For example, if an array of latches is to be manufactured, this particular configuration is particularly useful. It is also possible to reverse the on/off polarity of any switch disclosed in this application by applying a positive bias to the on electrode for a long enough time so that all the dopants drift to the opposite electrode of the switch.

[0084] FIGS. 9A-9C depict the configurable on/off polarity switch described above, in which the dopant source layer 310 is sandwiched between two main active layers 308, 308'. In this embodiment, the thickness of the switch is represented by Κ and Ti0₂/Ti0₂__x/Ti0₂The total width of the layers 308, 310, 308' is denoted by L.

[0085] In FIG. 9A, by applying an appropriate bias voltage to set the on/off polarity of the switch, the ionized dopant originally located in the dopant source layer 310 moves to the two main active layers 308, 308 , Any layer in.

[0086] A positive bias is applied to the electrode 104 (switched as shown in FIG. 9A) so that the positively charged dopant (in the TiO₂/TiO₂__xIn the case of the system, the oxygen vacancy) drifts to the left. This will set the on/off polarity of the switch so that a subsequent negative bias applied to the electrode 104 will turn on the switch and a positive voltage applied to the electrode 104 will turn off the switch. Figure 9B depicts the resulting structure.

[0087] Conversely, applying a negative bias to the electrode 104 (switched as shown in FIG. 9A) makes the positively charged dopant (in Ti0₂/Ti0₂__xIn the case of the system, the oxygen vacancy) drifts to the right. This will set the on/off polarity of the switch so that a subsequent positive bias applied to the electrode 104 will turn on the switch and a negative voltage applied to the electrode 104 will turn off the switch. Figure 9C depicts the resulting structure. Therefore, Figures 9B and 9C represent switches with opposite on/off polarities.

[0088] Making an electric switch

[0089] The switches disclosed herein can be fabricated using a variety of material deposition and processing techniques. First, use

CN 101548403 Β

Conventional technology (such as photolithography or electron beam lithography), or through more advanced technology (such as imprint lithography), to make the initial line 102 (metal or semiconductor). For example, this may be the bottom line 102 of the cross-wire pair 100 as shown in FIG. 1, or it may be the line 102 created in a via to create a connection 200 orthogonal to the plane of the circuit shown in FIG.

[0090] In the solution described here, the next component of the switch to be made is the non-covalent interface material 106a, and can be omitted if greater mechanical strength is required, at the cost of switching at a higher applied voltage Slower. As mentioned above, this is an optional part of switch 200' (or switch 100'). In this case, a layer 106a of some inert material is deposited. This can be a monolayer formed by the Langmuir-Blodgett (LB) process or it can be a Self-Assembled Monolayer (SAM) Generally speaking, this molecular layer 106a It is possible to form only a weak van der Waals type bond with the bottom electrode 102 and the main active material 308 of the switch. Alternatively, the layer 106a may be a thin layer of ice deposited on the cooling substrate. The material forming this ice can be an inert gas such as nitrogen (Ar) or it can be a species such as carbon dioxide (COJ). In this case, the ice prevents strong chemical bonding between the bottom electrode and the switching material The sacrificial layer and disappear from the system by heating the sample later in the processing sequence to sublime the ice. Those skilled in the art can easily conceive other ways for forming a weak bonding interface between the electrode 102 and the switch 306.

[0091] Next, the material of the main active layer 308 is deposited. This can be accomplished by a variety of conventional physical and chemical techniques, including evaporation from Knudsen cell, electron beam evaporation from clay pot, and target material. Sputtering or different forms of chemical vapor or beam growth from reactive precursors. The thickness of the film can be in the range of 1 to 30 nanometers (nm), and it can be grown without dopants. In this case, the resulting structure is switch-off. Depending on the thickness of the membrane 308, it can be nanocrystalline, nanoporous or amorphous in order to increase the speed at which ions can drift in the material, thereby achieving doping by ion implantation or achieving non-doping by expelling ions from 308 . Appropriate growth conditions (such as deposition rate and substrate temperature) can be selected to obtain the desired chemical composition and local atomic structure for this initial insulating or low-conductivity film 308.

[0092] The next layer is the dopant source layer for the primary switching material 308 or the secondary active layer 310, which can also be deposited by any of the techniques mentioned above. This material is selected to provide an appropriate doping species for the main active material. This secondary material 310 is selected to be chemically compatible with the main material 308, for example, the two materials should not chemically irreversibly react with each other to form the third material. If a switch to be configured later is made, another layer of primary material 308 is deposited on the secondary material 310.

[0093] As mentioned above, one example of a pair of materials that can be used as the primary active switching layer 308 and the secondary active switching layer 310 is Ti, respectively. ? And Ti0₂__xO Ti. ? Is a semiconductor with a band gap of approximately 3.2 eV. It is also a weak ion conductor. Ti0₂The thin film will produce the tunnel barrier needed to produce the switch-off configuration, and Ti0₂__xForm an ideal source of oxygen vacancies to dope Ti0₂And make it conductive. If it is desired to make a configurable on/off polarity switch, the main active material Ti can be deposited. ? The second film 308'.

[0094] Finally, a second metal electrode 104 is fabricated on the secondary active layer 310 or the second primary active layer 308 of the switch in a manner similar to that of generating the first line 102. If the system is a cross wire device 100, an etching process is used to remove the deposited switching material not below the top wire in order to isolate the device. If necessary, a planarization process can be added after depositing the bottom line group or after the second line group to provide a flat surface on the switch.

[0095] One problem with making switching devices is to carefully control the dopant concentration in the material, regardless of whether the dopant is present in the crystal lattice of different types of atoms or lacks specific atoms (for example, vacancies). Types of dopants can be introduced into

CN 101548403 Β

The amount of or a component in the growth medium can be reduced to cause vacancies. Another solution is to generate a high-quality and pure main material layer, and then directly deposit a carefully determined amount of type on the main material. This can be the dopant species, which will then diffuse into the host material; or it can be a reactive material, which will chemically react with one of the host materials components to be in the host lattice Cause vacancies. An example of this approach is to deposit a small amount of aluminum on a high-quality and pure titanium dioxide layer. This aluminum partially reacts with the oxygen of the titanium dioxide to form some aluminum oxide and leave vacancies in the titanium dioxide. This is a process similar to "delta doping", and delta doping is currently implemented in the semiconductor industry to provide very thin layers of highly doped semiconductor materials.

[0096] Material combination for switching

[0097] There are various materials that exhibit the desired combination of properties: electrically semiconducting or insulating "weak" ion conductors to enable dopant species to be implanted into and from the active region of the switch by drift discharge. Generally speaking, good switching materials are composite semiconducting oxides and nitrides, but semiconducting sulfides, phosphides, chlorides, selenides, monuments and deserts also provide switching. Generally speaking, any semiconducting material that is also a weak ion conductor capable of electrically doping the semiconductor will be suitable for practicing the present invention and its various embodiments. In other words, possible switch compounds are semiconducting compounds that have a large ionic contribution to bonding. A good combination is an undoped and stoichiometric (thus a good insulator) primary active material combined with a secondary source/sink combination of the same and related parent material, the parent material either containing a large concentration of anion vacancies or Other types of dopants that can drift into the host material under the applied bias voltage. The idea is that the source/sink system 310 is highly conductive, so changing the doping concentration has relatively little effect on the conductivity of this secondary material, but since the main material 308 is essentially intrinsic, even a small amount of dopant Will have a great influence on the conductivity of the material.

[0098] The switching compounds used in the practice of the present invention are usually oxides, sulfides, selenides, nitrides, phosphides, monuments, chlorides and deserts of transition metals and rare earth metals, in which alkaline earth metals are often present In the compound Ο In addition, there are various alloys of compounds similar to each other (like compounds with each other), which can have various compositions if they are mutually soluble in each other. Then there are mixed compounds in which there are two, three or more different metal atoms combined with a certain number of negatively charged elements. These dopants may be elements of different valences or anion vacancies doped into the host material.

[0099] Materials involving the elements Ti, Zr, and Hf are particularly attractive for switches compatible with Si integrated circuit technology. Since the main oxidation state of all three metals is +4, which is the same as Si, these elements No unintentional doping of Si occurs. These compounds are also called titanium oxide, zirconium oxide, and zirconium dioxide, respectively, and can also be referred to as other names specific to the different polytypes of each.

[0100] Another embodiment includes these three oxides in pairs or alloys in which all three are present at the same time (such as Ti_xZr_yHf_z0₂, Where x+y+z = 1). The related group of compounds includes titanate, cocoate, and cocoate, which are exemplified by specific examples of SrTiO₃Representative, where Sr is a divalent element saw. There are a variety of such compounds, in which Ca, Ba, and other divalent elements (eg, Mg, Zn, Cd) can be used instead of Sr, and Zr and Hf can be used instead of Ti. These compounds can be expressed as AB0₃The compound also has a perovskite structure, where A is at least one divalent element and B is one of Ti, Zr, and Hf.

[0101] It is also possible to use alloys of these different compounds, such as Ca_aSr_bBa_cTi_xZr_yHf_z0₃, Where a+b+c = 1 and χ+y+z = Ιο There are also a variety of other oxides of transition metals and rare earth metals with different valences, which can be used alone or as more complex compound of. In each case, the ionizing dopant species may be aliovalent (for example, different valence) elements or oxygen vacancies that are doped into the host material.

CN 101548403 Β

[0102] Yet another embodiment of the compound includes sulfides and selenides of transition metals with certain ionic bonding characteristics, which are essentially s analogs and Se analogs of the above-mentioned oxides.

[0103] Still another embodiment of the compound includes semi-conductive nitrides, such as AIN, GaN, ScN, YN, LaN, rare earth nitrides, alloys of these compounds and more complex mixed metal nitrides.

[0104] Yet another example of the compound includes phosphides and tombstones of various transition metals and rare earth metals (eg, Sc, Y, La, etc.).

[0105] Another embodiment of the compound includes semi-conductive halides such as CuCKCuBr and AgCL.

[0106] In all cases, anion vacancies or heterovalent elements can be used as the mobile dopant species in the above compounds.

[0107] The primary active layer or the secondary active layer can also be made of sublayers of different materials or compounds selected from the above examples.

[0108] The dopants used in different embodiments of the practice are selected from the group consisting of hydrogen, alkali (alkali), alkaline earth cations, transition metal cations, rare earth cations, oxygen anions or vacancies, chalcogens Compound anion or vacancy, nitrogen anion or vacancy, phosphorus group element compound anion or vacancy, or halide anion or vacancy.

[0109] Specific examples of the combination of the main material 308 and the secondary material 310 and the kind of dopant used for each combination are illustrated in the following table. Based on the teachings herein, those skilled in the art will obviously be able to develop other combinations of primary and secondary materials that provide the benefits of the teachings.

[0110] Table · Compatible primary materials and secondary materials and a list of examples of dopant types

[0111]

<td>Main material</td><td>Secondary material</td><td>Types of dopants</td>

<td>Ti0₂</td><td>Ti0₂__x</td><td>Oxygen vacancy</td>

<td>Zr0₂</td><td>Zr0₂__x</td><td>Oxygen vacancy</td>

<td>Hf0₂</td><td>Hf0"</td><td>Oxygen vacancy</td>

<td>SrTiO₃</td><td>SrTiO₃__x</td><td>Oxygen vacancy</td>

<td>GaN</td><td>Gag</td><td>Nitrogen Vacancy</td>

<td>CuCl</td><td>CuC-</td><td>Chlorine Vacancy</td>

<td>GaN</td><td>GaN:S</td><td>Divalent sulfide ion</td>

[0112] Other considerations

[0113] As discussed above, the conductance of the switch can be varied over a wide range. The embodiments described below are aimed at the fact that the switch described herein is actually continuous-for example, from the off state to the on state and back to the off state, the resistance of the switch is continuous but changes a bit sharply. This situation is clearly shown in the schematic diagrams of the switch equivalent circuit (FIGS. 6B, 7B, 8B), which shows a variable resistor in series with a fixed resistor.

[0114] There are many applications for variable or analog switches in electronic circuits.

[0115] In one embodiment, an electronically settable variable resistor can be used as a trimming resistor to adjust the circuit's

CN 101548403 Β

impedance. There is a certain amount of variability in all electronic circuits, which means that the produced circuit will have slightly different operating parameters in terms of its operating speed and the delay time for the signal to propagate through the circuit. In the past, this problem was solved by installing a mechanically adjustable "trim pot" in the circuit board. Expert engineers are required to test and tune the circuit in order to optimize its properties-this is by inspecting the circuit board with an oscilloscope This is done by using a screwdriver to adjust the tuning potentiometer, in other words changing the resistance of the variable resistor to make the impedance of the circuit reach its optimal performance. This is done by testing the circuit for the first time to ensure It is required when it works normally, and as time changes, various components in the circuit may change, so experts need to retest and retune the circuit on site. As the integration of the circuit becomes higher, this tuning requires It becomes unnecessary, because the repeatability of integrated circuit components is improved. Moreover, there is no place on the integrated circuit for placing the tuning potentiometer, so it cannot be tuned for most integrated circuits. As the feature size becomes more Small, the variability of component properties is inevitably increasing, precisely because a single atom determines the lower limit of the uncertainty of the feature size. Therefore, the next few generations of circuits will need to have a tuning potentiometer that can tune and optimize the circuit. In fact, with As the device variability becomes greater, it is likely that if the circuit cannot be tuned, the circuit will not be tuned at all. jobs. Therefore, we will use switches in different positions within the integrated circuit and the feedback circuit so that the circuit can not only be tuned when it is first turned on, but also continuously optimize the performance of the circuit during operation. Therefore, circuits with such switches and feedback elements should be able to continuously improve over time. When a particular component in the circuit fails, the circuit uses switches to reconfigure and reoptimize itself. Therefore, this circuit should have the ability to degrade moderately with age, rather than suddenly stop operating when a single component fails.

[0116] In another embodiment, in addition to the circuit needs to be tuned, it is also desirable to change the operating characteristics of certain circuits to match the specific signal input used to make various measurements. For example, many transducers or measurement systems perform best when there is a "bridge circuit" in which the resistance of the variable resistor is adjusted to balance the differential circuit for very accurate measurements. One example is measuring accurate voltage differences for thermocouples, but there are many other examples of bridge circuits for measurement.

[0117] In yet another embodiment, the application of the analog switch is in a learning circuit or a neural network. In these circuits, the training process is used to teach the circuit how to respond to specific inputs. This is done by comparing the attributes of the component with the expected attributes and using feedback methods to reset the value of the component in response to the input value. Today, most neural networks are implemented in software using conventional integrated circuits and processing units-real neural networks implemented in hardware with adjustable analog switches will be much more efficient.

[0118] In yet another embodiment, other analog circuits and mixed analog and digital circuits can be improved by having a continuously adjustable resistance that is non-volatile-for example, the reference value of the resistor is in the initialization phase Set during the period, and then adjust the resistance during operation in order to respond to the operating conditions of the circuit. This can be used, for example, to compensate for changes in the basic operating characteristics of the circuit with temperature or power supply voltage fluctuations or other changes in environmental conditions.

[0119] For the sake of explanation, the foregoing description uses specific terminology to provide a thorough understanding of the invention. However, it is obvious to those skilled in the art that no specific details are required in order to practice the invention. The foregoing description of specific embodiments of the present invention has been given for the purpose of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise form disclosed. Obviously, many modifications and changes are possible in view of the above teachings. These embodiments are shown and described in order to best explain the principles of the present invention and its practical application, so that those skilled in the art can best utilize the present invention and various embodiments with various modifications, which are suitable for For the specific purpose under consideration. It is intended that the scope of the present invention is defined by the appended claims and their equivalents.

CN 101548403 Β