Quantum dot memory

Abstract

A method of making a quantum dot memory cell, the quantum dot memory cell including an array of quantum dots disposed between a first electrode and a second electrode, includes obtaining values for a tunneling current through the quantum dot memory cell as a function of a voltage applied to the quantum dot memory cell and selecting parameters of the quantum dot memory cell such that the tunneling current through the quantum dot memory cell exhibits a bistable current for at least some values of the voltage applied to the quantum dot memory cell. The values for the tunneling current are determined on the basis of a density of states of the array of quantum dots.

Description

FIELD OF THE INVENTION

This invention relates to a quantum dot memory.

BACKGROUND

Resonant tunneling diodes (RTDs) are used as memory devices due to the bistable behavior that results from a hysteresis intrinsic to the DC current-voltage characteristics of the RTDs. This bistability is predicted to exist in nanoscale devices such as single electron transistors and single molecule transistors. Tunneling current through degenerate states of a single quantum dot or molecule leads to a switching effect only in the case of an attractive electron-electron interaction which is mediated by the electron-phonon interaction. It has been proposed, considering only a single energy level, that the hysteresis of I-V characteristics can be observed in a single molecule junction with an effective attractive Coulomb interaction on the basis of the Hartree approximation and the polaron effect. Hysteretic tunneling current in a polaron model has also been observed beyond the Hartree approximation.

Theoretical studies have predicted the existence of hysteresis in a quantum dot or molecular junction, although conclusive experimental support for the predictions has not been achieved. The tunneling current through a carbon nanotube quantum dot exhibits a periodic oscillatory behavior with respect to an applied gate voltage, which arises from an eightfold degenerate state. A periodic oscillatory differential conductance also arises as a result of a tunneling current through a single spherical PbSe quantum dot having a sixfold degenerate state. However, neither of these situations exhibits a bistable tunneling current, indicating that electron-phonon interactions in nanotube quantum dots and PbS quantum dots are not sufficient to yield the strong effective electron-electron interactions necessary for the existence of a bistability. In general, bistable current in existing memory systems arises from a phase transition of a bulk material, which phase transition vanishes on the nanoscale.

Semiconductor quantum dot arrays can be chemically fabricated to form a superlattice. The size of the quantum dots and the lattice constant of the superlattice are controllable via nanoscale manipulation, enabling charges in the quantum dot array to be tuned in either the Coulomb blockade regime or the semiconducting regime. Consequently, quantum dot arrays are promising candidates for the investigation of strongly correlated systems as well as for use as integrated electronic devices.

SUMMARY

In a general aspect, a method of making a quantum dot memory cell, the quantum dot memory cell including an array of quantum dots disposed between a first electrode and a second electrode, includes obtaining values for a tunneling current through the quantum dot memory cell as a function of a voltage applied to the quantum dot memory cell and selecting parameters of the quantum dot memory cell such that the tunneling current through the quantum dot memory cell exhibits a bistable current for at least some values of the voltage applied to the quantum dot memory cell. The values for the tunneling current are determined on the basis of a density of states of the array of quantum dots.

Embodiments may include one or more of the following. The quantum dot memory cell includes a quantum well positioned between the array of quantum dots and the first electrode. The values for the tunneling current are obtained on the basis of an occupation number of the quantum well, the occupation number of the quantum well representing a number of electrons in a subband of the quantum well.

The values for the tunneling current are obtained on the basis of a first tunneling rate between the array of quantum dots and the first electrode and a second tunneling rate between the array of quantum dots and the second electrode. The values for the tunneling current are obtained on the basis of a first Fermi distribution function of the first electrode and a second Fermi distribution function of the second electrode. The values for the tunneling current are obtained on the basis of a first occupation number of the array of quantum dots, the first occupation number representing a number of electrons in orbitals coupling the quantum dots to the first electrode and the second electrode. The orbitals coupling the quantum dots to the first electrode and the second electrode are localized around each quantum dot. The values for the tunneling current are obtained on the basis of a second occupation number of the array of quantum dots, the second occupation number representing a number of electrons in one of a plurality of energy levels in extended states of the array of quantum dots.

The parameters include one or more of the following: a composition of the quantum dots, a separation between adjacent quantum dots in the array of quantum dots, a size of the quantum dots, a number of quantum dots in the array of quantum dots, a composition of a material in which the array of quantum dots is embedded, a composition of the first electrode and a composition of the second electrode, a composition of the quantum well.

The tunneling current arises at least in part due to charge swapping between localized states in the array of quantum dots and extended states in at least one of the first electrode and the second electrode. The quantum dot memory cell exhibits bistable current when the voltage applied to the quantum dot memory cell is less than about 100 mV. The quantum dot memory cell is capable of operating with a switching speed of about 1 Terahertz.

In another aspect, a quantum dot memory cell includes a first electrode, a second electrode, and an array of quantum dots disposed between the first electrode and the second electrode. Parameters of the quantum dot memory cell are selected such that a tunneling current through the quantum dot memory cell exhibits a bistable current for at least some values of a voltage applied across the quantum dot memory cell, the tunneling current determined on the basis of a density of states of the array of quantum dots.

Embodiments may include one or more of the following. The quantum dot memory cell includes a quantum well positioned between the array of quantum dots and the first electrode. The array of quantum dots includes at least 50 quantum dots. The array of quantum dots includes a single layer of quantum dots in substantially all regions of the array of quantum dots. The array of quantum dots is substantially two-dimensional. The array of quantum dots is substantially one-dimensional.

The tunneling current arises at least in part due to charge swapping between localized states in the array of quantum dots and extended states in at least one of the first electrode and the second electrode. The quantum dot memory cell exhibits bistable current when the voltage applied to the quantum dot memory cell is less than about 100 mV. The quantum dot memory cell is capable of operating with a switching speed of about 1 Terahertz. The quantum dots include a semiconductor. The quantum well includes a semiconductor.

A quantum dot memory cell as described herein has a number of advantages, including the following. Bistable states of such a quantum dot memory cell are used as a nanoscale memory operable at room temperature; a bistable current response of the memory cell is used to read out information stored in the nanoscale memory. For instance, such a quantum dot memory cell has low power consumption and is capable of operating with an applied bias of less than 100 mV. Due to the quantum nature of the memory cell, the switching time is comparable to the tunneling rate through the quantum dots in the memory cell, which is on the order of a picosecond. Thus, a memory clock rate on the order of 1 THz is attainable. Additionally, a quantum dot memory cell as described herein has a storage density as high as about 1 TB/in², which is significantly higher than the storage density of existing nonvolatile memory devices. Furthermore, a nanoscale memory composed of multiple quantum dots does not encounter problems related to background charge, which problem significantly limits the applications of single electron transistors and molecular transistors.

BRIEF DESCRIPTION OF DRAWINGS

FIG. 1A depicts a quantum dot memory cell.

FIG. 1B depicts another embodiment of a quantum dot memory cell.

FIGS. 2A, 2B, and 2C depict an array of quantum dots.

FIGS. 3A and 3B are plots of auxiliary functions as a function of an occupation number of a quantum dot for various values of a bias applied to a quantum dot memory cell.

FIG. 4 is a plot of an occupation number of a quantum dot as a function of a bias applied to a quantum dot memory cell.

FIG. 5A is a plot of a tunneling current across a quantum dot memory cell as a function of a bias applied to the memory cell for various values of a Coulomb interaction in a quantum dot of the memory cell.

FIG. 5B is a plot of an occupation number of a quantum dot as a function of a bias applied to a quantum dot memory cell for various values of a Coulomb interaction in the quantum dot.

FIG. 6 is a plot of a tunneling current across a quantum dot memory cell as a function of a bias applied to the memory cell for various arrangements of an array of quantum dots in the memory cell.

FIG. 7 is a plot of a tunneling current across a quantum dot memory cell as a function of a bias applied to the memory cell for various tunneling rate ratios through the memory cell.

DETAILED DESCRIPTION

Referring to FIG. 1A, a quantum dot memory cell 100 has a nanoscale memory structure capable of high density storage that operates on the basis of electronic interactions among quantum dots in the memory cell. Quantum dot memory cell 100 includes an array 102 of quantum dots 104 embedded in an insulator matrix layer 106, which is positioned between a bottom electrode 108 and a top electrode 110, both of which are metals such as copper. Bottom electrode 108 sits on a substrate 112, which is made of, for instance, silicon. A bias V_a is applied across quantum dot memory cell 100; insulator layer 106 containing the array 102 of quantum dots 104 acts as a tunneling barrier between bottom electrode 108 and top electrode 110.

In some embodiments, the material of insulator matrix layer 106 embeds quantum dots 104 but is not capable of forming a good quality connection with bottom electrode 108 and top electrode 110. In this situation, a first insulating spacer layer 114 is positioned between insulator matrix layer 106 and bottom electrode 108 and a second insulating spacer layer 114 is positioned between insulator matrix layer 106 and top electrode 110. The material of first and second insulating spacer layers 114 and 116 is chosen to form a smooth connection with bottom and top electrodes 108, 110.

Quantum dots 104 may be composed of any semiconductor material, including but not limited to GaAs, InAs, and CdSe. In some examples, quantum dots 104 have a radius of around 2-3 nm such that only one localized electronic state exists in each quantum dot. In other examples, quantum dots 104 are large enough to each support multiple localized electronic states. The quantum dots preferably have a shape that is approximately spherical. Quantum dots 104 in array 102 are separated from each other by a distance that is sufficiently short so as to allow electronic states of adjacent dots to electronically couple. For instance, in some embodiments quantum dots 104 are separated by about 1-2 nm, or by less than about 3 nm. The degree of coupling between quantum dots 104 is also affected by the size and composition of quantum dots 104 and the dielectric constant of insulator matrix layer 106, among other parameters.

Referring to FIG. 1B, in another embodiment, a quantum dot memory cell 150 includes a quantum well 118. Quantum well 118 is positioned between insulator matrix layer 106 and first insulating spacer layer 114, as shown, or between insulator matrix layer 106 and second insulating spacer layer 116. Quantum well 118 is made of a semiconductor material, such as GaAs or any other semiconductor commonly used for quantum wells.

Array 102 of quantum dots 104 is formed by any standard method of quantum dot assembly. In some embodiments, array 102 is generated by self-assembly of previously synthesized quantum dots. For instance, array 102 is formed via a directed self-assembly process in which a grid is defined on bottom electrode 108 and quantum dots 104 are guided to self-assemble in the grid into array 102. In other embodiments, array 102 of quantum dots is formed directly on bottom electrode 108 by an epitaxy process (e.g., molecular beam epitaxy) or a chemical vapor deposition process. Referring to FIG. 2A, quantum dots 104 of array 102 are arranged in a lattice having a first lattice constant a along an x axis and a second lattice constant b along a y axis. A z axis projects out of the page of the figure. A unit cell of array 102 is defined on the basis of the arrangement of quantum dots 104 in array 102. In the array 102 shown in FIG. 2A, a unit cell 202 is defined such that a quantum dot 104 is located at each corner of the unit cell 202. In other arrangements of quantum dots in 104 in array 102, other unit cells may be defined. For instance, some arrangements of quantum dots 104 may be described by a unit cell having a quantum dot 104 at each corner of the unit cell and one quantum dot located in the center of the unit cell.

Referring to FIGS. 2A, 2B, and 2C, electronic states (i.e., orbitals) of quantum dots 104 are localized at each quantum dot in the array; electronic states in bottom electrode 108 and top electrode 110 are extended throughout each electrode. FIGS. 2A and 2B show array 102 of quantum dots in the x-y plane; FIG. 2C shows array 102 of quantum dots in the x-z plane. Each quantum dot 104 has a pair of p_x orbitals 204 oriented along the x axis, a pair of p_y orbitals 206 oriented along the y axis, and a pair of p_z orbitals 208 oriented along the z axis. p_x orbitals 204, p_y orbitals 206, and p_z orbitals 208 are six-fold degenerate prior to any coupling effect; i.e., all six orbitals have the same energy. An electron hopping between p_x orbitals 204 of adjacent quantum dots 104 along the x axis has a hopping rate υ_i; an electron hopping between p_x orbitals 204 of adjacent quantum dots 104 along the y axis has a hopping rate υ_t. An electron hopping between p_y orbitals 206 of adjacent quantum dots 104 along the x axis has a hopping rate υ_t′; an electron hopping between p_y orbitals 206 of adjacent quantum dots 104 along they axis has a hopping rate υ_i′.

p_x and p_y orbitals 204, 206 (also called p_xy orbitals) of adjacent quantum dots 104 overlap (i.e., hybridize) in the x-y plane, forming two-dimensional (2-D) conduction bands in array 102 of quantum dots 104. Array 102 includes a sufficient number of quantum dots 104 to form a band; in general, array 102 contains at least about 50-100 quantum dots. The separation between adjacent quantum dots 104 is small enough to allow coupling between p_x and p_y orbitals 204, 206. The height of a potential barrier between adjacent quantum dots also affects the degree of coupling between p_x and p_y orbitals 204, 206. For instance, for a potential barrier of 0.5 eV, coupling occurs between p_x and p_y orbitals of adjacent quantum dots having an interdot hopping energy of 20 meV when the quantum dots are separated by less than about 3 nm. In other instances, depending on factors such as the composition of quantum dots 104 and insulating matrix layer 106 and the temperature, coupling occurs between p_x and p_y orbitals of adjacent quantum dots separated by a distance of less than about 2 nm.

In the embodiment shown in FIG. 1B, extended band states exist in quantum well 118. In this case, p_x and p_y orbitals 204, 206 of adjacent quantum dots 104 do not need to overlap to form conduction bands in array 102. Thus, the spacing between adjacent quantum dots in array 102 is less critical. Charge neutrality in quantum dot memory cell 150 is maintained by having a sufficient quantity and areal density of quantum dots 104. Because the spacing between quantum dots does not need to be precisely controlled, quantum dot memory cell 150 including quantum well 118 is easier to fabricate than quantum dot memory cell 100.

p_z orbitals 208 remain localized in the x-y plane at each quantum dot 104; that is, coupling between p_z orbitals of neighboring quantum dots is negligible. If quantum dots 104 are positioned too close together, p_z orbitals 208 of adjacent quantum dots may couple, destroying the localization of the p_z orbitals. During operation of the quantum dot memory cell, p_z orbitals 208 of quantum dots 104 couple electronically with the extended electronic states of bottom electrode 108 and top electrode 110 and with the 2-D conduction bands in array 102 via an on-site Coulomb interaction. In the example shown in FIG. 1B, p_z orbitals 208 couple with the extended band or sub-band states of quantum well structure 118 rather than with the extended states of array 102 of quantum dots. The thickness and composition of quantum well 118 is designed such that the energy of a sub-band state of the quantum well is close to the energy of localized orbitals of quantum dots 104, which are not restricted to the p_z orbitals 208. In this case, the existence of the p_x and p_y orbitals is not critical, and the shape of the quantum dots is not restricted to being approximately spherical.

An expression for the extended Anderson Hamiltonian, H=H₀+H_T+H_d, derived from a combination of the Falicov-Kimball model and the Anderson model, is used to mathematically describe quantum dot memory cell 100.

The first term of the extended Anderson Hamiltonian, H₀, describes electronic states in bottom electrode 108 and top electrode 110. H₀ is written as

$H_0=\sum _{k,\sigma ,\beta ,\ell }{\varepsilon }_k{a}_{k,\sigma ,\beta }^{†}{a}_{k,\sigma ,\beta },$ where α_k,σ,β^‡(α_k,σ,β) creates (destroys) an electron of wave vector k and spin σ having an energy ∈_k in the β electrode (i.e., in bottom electrode 108 or top electrode 110).

Referring to FIGS. 1, 2A, and 2B, the second term of the extended Anderson Hamiltonian, H_T, describes coupling between electronic states of bottom and top electrodes 108, 110 and p_z orbitals 208 of quantum dots 104:

$\begin{array}{cc}{H}_T=\sum _{k,\sigma ,\beta ,\ell }{V}_{k,\beta ,\ell }{a}_{k,\sigma ,\beta }^{†}{d}_{\ell ,\sigma }+\sum _{k,\sigma ,\beta ,\ell }{V}_{k,\beta ,\ell }^{*}{d}_{\ell ,\sigma }^{†}{a}_{k,\sigma ,\beta },& \left(1\right)\end{array}$ where d_l,σ^‡ (d_l,σ) creates (destroys) an electron with spin σ in the p_z orbital 208 of a particular quantum dot 104 located at a site l in array 102 and V_k,β,l describes the coupling between continuous electronic states in bottom and top electrodes 108, 110 and p_z orbitals 208 of quantum dots 104. In the expression for H_T, it is assumed that the coupling between the bottom and top electrodes 108, 110 and the p_x,y orbitals 204, 206 of array 102 of quantum dots 104 is negligible because p_x,y orbitals 204, 206 are localized along the z axis.

The third term of the extended Anderson Hamiltonian, H_d, describes interactions among electronic states of quantum dots 104 in array 102:

$\begin{array}{cc}\begin{array}{c}{H}_d=\sum _{\ell ,\sigma }{E}_{p_z}{d}_{\ell ,\sigma }^{†}{d}_{\ell ,\sigma }+\sum _{p,\lambda }\left({\varepsilon }_{p,\lambda }+U\left(N_c-N_{\lambda }\right)\right)c_{p,\lambda }^{†}{c}_{p,\lambda }+\\ \sum _{\ell ,\sigma }{U}_{\ell }{d}_{\ell ,\sigma }^{†}{d}_{\ell ,\sigma }{d}_{\ell ,-\sigma }^{†}{d}_{\ell ,-\sigma }+\frac{U_{dc}}{N_{QD}}\sum _{\ell ,p,p^{\prime },\sigma }{c}_{p,\lambda }^{†}{c}_{p^{\prime },\lambda }{e}^{i\left(p-p^{\prime }\right)·R_{\ell }}{d}_{\ell ,\sigma ,\lambda }^{†}{d}_{\ell ,\sigma },\end{array}& \left(2\right)\end{array}$ where d_l,σ^‡ (d_l,σ) creates (destroys) an electron in p_z orbital 208 (having an energy E_pz=E_P) of a particular quantum dot 104 located at a site l in array 102 and c_p,λ and c_p,λ^‡ represent creation and annihilation operators, respectively, of an electron having a wave vector p in a band λ of the 2-D conduction band of array 102. The second term of Eq. (2) describes 2-D conduction bands of array 102 arising from the coupling of p_x,y orbitals 204, 206 in array 102, where ρ labels 2-D wave vectors of array 102, R_l denotes the position of a particular quantum dot 104 located at a site l in array 102, λ labels the conduction bands (including spin), and U denotes an on-site Coulomb interaction between two electrons in p_x,y orbitals 204, 206 of a quantum dot 104. Ignoring quadrupole and higher order terms in the expansion of 1/r₁₂, the Coulomb repulsion integrals between two electrons in any of the three degenerate p-like orbitals 204, 206, 208 are the same. N_λ is the occupation number per unit cell 202 for the λ-th conduction band of array 102, and

$N_c=\sum _{\lambda }{N}_{\lambda }$ is the total occupation number per unit cell 202 in p_x,y orbitals 204, 206 forming the conduction bands of array 102. The second term of Eq. (2) is obtained using a mean-field theory (which is justified for extended states) applied to the 2-D conduction bands of array 102. The last two terms of Eq. (2)

$\left(i.e.,\sum _{l,\sigma }{U}_l{d}_{l,\sigma }^{†}{d}_{l,\sigma }{d}_{l,-\sigma }^{†}{d}_{l,-\sigma }\mathrm{and}\frac{U_{\mathrm{dc}}}{N_{\mathrm{QD}}}\sum _{l,p,p^{\prime },\sigma }{c}_{p,\lambda }^{†}{c}_{p^{\prime },\lambda }{e}^{i\left(p-p^{\prime }\right)·R_l}{d}_{l,\sigma ,\lambda }^{†}{d}_{l,\sigma }\right)$ involve U_l=U and U_dc, which denote, respectively, an on-site repulsive Coulomb energy in the p_z orbital 208 of a particular quantum dot 104 located at a site l in array 102 and a Coulomb interaction energy between p_x,y orbitals 204, 206 and p_z orbitals 208 of a quantum dot 104 in array 102. N_QD denotes the number of quantum dots 104 in array 102. In Eq. (2), the focus is on the p_z orbitals 208 of quantum dots 104 rather than the ground state orbitals because, in the range of applied bias considered, the ground state energy level of a quantum dot 104 is deeply below the Fermi levels of both bottom electrode 108 and top electrode 110 and electron tunneling from bottom electrode 108 to top electrode 110 (or vice versa) through the ground state of quantum dots 104 is therefore blocked. Carriers in the ground state of a quantum dot 104 merely cause a constant shift to all the p orbitals 204, 206, 208 of the quantum dot 104.

Eq. (2) is the same as the Falicov-Kimball (FK) model if the self energy term, U(N_c−N_λ), is neglected. The FK model has been used extensively to study the semiconductor-metal transition in transition metals and to study rare-earth oxides containing both localized and delocalized orbitals. The total number of carriers in localized orbitals and delocalized orbitals is conserved in the FK model. However, in the case of quantum dot memory cell 100, the total number of carriers (i.e., electrons) is not conserved, as carriers can be injected from and allowed to tunnel out of quantum dots 104 through bottom electrode 108 and top electrode 110.

Using Keldysh's Green's function technique, a tunneling current J_l,σ through a quantum dot 104 at site l in array 102 is expressed as

$\begin{array}{cc}{J}_{l,\sigma }=\frac{-e}{\hslash }\int \frac{d\varepsilon }{\pi }\left[f_L\left(\varepsilon \right)-f_R\left(\varepsilon \right)\right]\frac{{\Gamma }_{l,L}{\Gamma }_{l,R}}{{\Gamma }_{l,L}+{\Gamma }_{l,R}}{\mathrm{Im}G}_{l,l}^r\left(\varepsilon \right),& \left(3\right)\end{array}$ where f_L=f(∈−μ_L) and f_R=f(∈−μ_R) are Fermi distribution functions for bottom electrode 108 and top electrode 110, respectively, and μ_L and μ_R represent a chemical potential of bottom electrode 108 and top electrode 110, parameters which are determined by the composition of the electrodes. The chemical potential difference between bottom electrode 108 and top electrode 110 is related to applied bias V_a by μ_L−μ_R=eV_a. The electron charge is represented by e and Planck's constant is represented by h. Γ_l,L(∈) and ∈_l,R(∈) denote tunneling rates from p_z orbitals 208 of quantum dots 104 to bottom electrode 108 and top electrode 110, respectively, where

${\Gamma }_{l,\beta }=2\pi \sum _k{V_{l,\beta ,k}}^2\delta \left(\varepsilon -{\varepsilon }_k\right).$ In the wide-band limit, tunneling rates ∈_l,L(∈) and ∈_l,R(∈) are approximately energy-independent. Therefore, the calculation of the tunneling current J_l,σ is entirely determined by the spectral function A=lmG_l,l^r(∈), which is the imaginary part of the retarded Green's function G_l,l^r(∈).

Using the equation of motion for G_l,l^r(∈), the following expressions are obtained:

$\begin{array}{cc}\left(\varepsilon -E_p+i\Gamma \right)G_{i,j}^r\left(\varepsilon \right)={\delta }_{i,j}+U\langle n_{i,-\sigma }{d}_{i,\sigma }{d}_{j,\sigma }^{†}\rangle +\sum _{p^{″},p^{\prime },\sigma }{g}_{p^{″},p^{\prime }}\langle c_{p^{″},{\sigma }^{\prime }}^{†}{c}_{p^{\prime },{\sigma }^{\prime }}{d}_{i,\sigma }{d}_{j,\sigma }^{†}\rangle ,& \left(4\right)\\ \left(\varepsilon -E_p+i\Gamma \right)G_{i,p}^r\left(\varepsilon \right)=U\langle n_{i,-\sigma }{d}_{i,\sigma }{c}_{p,\sigma }^{†}\rangle +\sum _{p^{″},p^{\prime },\sigma }{g}_{p^{″},p^{\prime }}\langle c_{p^{″},{\sigma }^{\prime }}^{†}{c}_{p^{\prime },{\sigma }^{\prime }}{d}_{i,\sigma }{c}_{p,\sigma }^{†}\rangle ,& \left(5\right)\\ \left(\varepsilon -{\varepsilon }_{p^{\prime },\lambda }-U\left(N_c-N_{\lambda }\right)\right)G_{p^{\prime },p}^r\left(\varepsilon \right)={\delta }_{p^{\prime },p}+\sum _{i,p^{″},\sigma }{g}_{p^{″},p^{\prime }}\langle \left(n_{i,\uparrow }+n_{i,\downarrow }\right)d_{p^{″},\sigma }{c}_{p,\sigma }^{†}\rangle ,\mathrm{and}& \left(6\right)\\ \left(\varepsilon -{\varepsilon }_{p^{\prime },\lambda }-U\left(N_c-N_{\lambda }\right)\right)G_{p^{\prime },j}^r\left(\varepsilon \right)={\delta }_{p^{\prime },p}+\sum _{i,p^{″},\sigma }{g}_{p^{″},p^{\prime }}\langle \left(n_{i,\uparrow }+n_{i,\downarrow }\right)d_{p^{″},\sigma }{d}_{j,\sigma }^{†}\rangle .& \left(7\right)\end{array}$ Here,

$\Gamma =\left({\Gamma }_{l,L}+{\Gamma }_{l,R}\right)/2\mathrm{and}{g}_{p,p^{\prime }}=\frac{U_{\mathrm{dc}}}{N}{e}^{i\left(p-p^{\prime }\right)·R_i}.$ In Eqs. (4)-(7), four one-particle Green's functions have been introduced: G_i,j^r(∈)=d_i,σd_j,σ^‡, G_i,p^r(∈)=d_i,σc_p,σ^‡, G_p′,p^r(∈)=c_p′,σc_p,σ^‡ and G_p′,j^r(∈)=c_p′,σd_j,σ^‡. These four one-particle Green's functions are coupled with two-particle Green's functions via U and U_dc. The equation of motion for the two-particle Green's function (defined as n_i,−σd_i,σd_j,σ^‡, n_i,−σd_i,σc_p,σ^‡c_{p″,σ′cp′,σ′}d_i,σd_j,σ^‡c_p″,σ′^‡c_p′,σ′d_i,σc_p,σ^‡(n_i,↑+n_i,↓)d_p″,σc_p,σ^‡, and (n_i,↑+n_i,↓)d_p″,σd_j,σ^‡ are coupled to three-particle Green's functions. In order to terminate this hierarchy of equations of motion, the Hartree-Fock approximation method is used to decouple terms involving U_dc. Meanwhile, in the calculations for n_i,−σd_i,σd_j,σ^‡ and n_i,−σd_i,σc_p,σ^‡, the coupling terms between localized states and the bottom and top electrodes 108 and 110 are determined following a scheme described in “Tunneling Current Spectroscopy of a Nanostructure Junction Involving Multiple Energy Levels,” published in Physical Review Letters, vol. 99, article 086803; and “Theory of charge transport in a quantum dot tunnel junction with multiple energy levels,” published in Physical Review B, vol. 77, article 245412, both of which are incorporated herein by reference, which scheme is valid for a regime in which Coulomb blockade occurs.

Solving Eqs. (4)-(7), the following is obtained:

$\begin{array}{cc}{G}_{p,p^{\prime },\lambda }^r\left(\varepsilon \right)=\frac{{\delta }_{p,p^{\prime }}}{\varepsilon -{\varepsilon }_p-{\Delta }_{\lambda }},\mathrm{where}{\Delta }_{\lambda }=U_{\mathrm{dc}}\left(N_{d,\sigma }+N_{d,-\sigma }\right)+U\left(N_c-N_{\lambda }\right)\mathrm{and}& \left(8\right)\\ G_{l,l}^r\left(\varepsilon \right)=\frac{1-N_{d,-\sigma }}{\varepsilon -E_p-{\Delta }_c+i\Gamma }+\frac{N_{d,-\sigma }}{\varepsilon -E_p-U-{\Delta }_c+i\Gamma },& \left(9\right)\end{array}$ where Δ_c=U_dcN_c denotes the self energy of a quantum dot 104 due to the Coulomb interaction of the local orbital of the quantum dot with the band and where N_d is the occupation number of p_z orbitals 208 in each unit cell 202 in array 102.

To reveal the tunneling current behavior across quantum dot memory cell 100, the occupation numbers N_d,σ and N_c are solved self-consistently by the following equations:

$\begin{array}{cc}{N}_{d,\sigma }=-\int \frac{d\varepsilon }{\pi }\frac{{\Gamma }_L{f}_L\left(\varepsilon \right)+{\Gamma }_R{f}_R\left(\varepsilon \right)}{{\Gamma }_L+{\Gamma }_R}{\mathrm{Im}G}_{l,l}^r\left(\varepsilon \right)\mathrm{and}& \left(10\right)\\ N_c=-\sum _{p,\lambda }\int \frac{d\varepsilon }{\pi }\frac{{\Gamma }_{L,c}{f}_L\left(\varepsilon \right)+{\Gamma }_{R,c}{f}_R\left(\varepsilon \right)}{{\Gamma }_{L,c}+{\Gamma }_{R,c}}{\mathrm{Im}𝒢}_{p\lambda .p\lambda }^r\left(\varepsilon \right)/N_{\mathrm{QD}},& \left(11\right)\end{array}$ where G_pλ,pλ^r(∈)=1/(∈−∈_p,λ−U(N_c−N_λ)−U_dc(N_d,↑+N_d,←)+i(Γ_L,c+Γ_R,c)/2). As mentioned above, the coupling between the bottom and top electrodes 108 and 110 and the p_x,y orbitals 208 of array 102 of quantum dots 104 is negligible (i.e., Γ_c=(Γ_L,c+Γ_R,c)/2), where Γ_L,c(Γ_R,c) denotes the tunneling between the bottom (top) electrode 108 (110) and the p_x,y orbitals, is small). Therefore, the imaginary part of Eq. (8) can be written as ImG_pλ,pλ^r(∈)≈πδ(∈−∈_p,λ−U(N_c−N_λ)−U_dc(N_d,↑+N_d,↓)).

The range of applied bias used for the operation of quantum dot memory cell 100 is not sufficient to overcome the charging energy of U+Δ_c. Thus, the second term in Eq. (9) is ignored, giving G_l,l^r(∈)=(1−N_d,−σ)/(∈−E_p−Δ_c+iΓ). The occupation numbers at zero temperature are then calculated by

$\begin{array}{cc}{N}_{d,\sigma }=\frac{\left(1-N_{d,-\sigma }\right)}{\pi }\frac{{\Gamma }_L}{{\Gamma }_L+{\Gamma }_R}{\int }_{-\infty }^{E_F+{\mathrm{eV}}_a}d\varepsilon \frac{\Gamma }{{\left(\varepsilon -E_p-\alpha {\mathrm{eV}}_a-{\Delta }_c\right)}^2+{\Gamma }^2}\mathrm{or}& \left(12\right)\\ \frac{{\Gamma }_L}{{\Gamma }_L+{\Gamma }_R}{N}_d/\left(1-N_d\right)=\frac{1}{\pi }{\cot }^{-1}\left(\frac{E_F+{\mathrm{eV}}_a-E_p-{\mathrm{\alpha eV}}_a-U_{\mathrm{dc}}{N}_c}{\Gamma }\right),& \left(13\right)\end{array}$ in which the αeV_a, term arises from the applied bias across quantum dot memory cell 100, where α is a dimensionless scaling factor determined by the environment of quantum dot memory cell 100, and

$\begin{array}{cc}{N}_{\lambda }=\frac{{\Gamma }_{L,c}}{{\Gamma }_{L,c}+{\Gamma }_{R,c}}{\int }_{-\infty }^{E_F+{\mathrm{eV}}_a}d\varepsilon D_{\lambda }(\varepsilon -U_{\mathrm{dc}}\left(N_{d,\uparrow }+N_{d,\downarrow }\right)-U\left(N_c-N_{\lambda }\right),& \left(14\right)\end{array}$ where

$D_{\lambda }\left(\varepsilon \right)=\sum _p\delta \left(\varepsilon -{\varepsilon }_{p,\lambda }\right)/N_{\mathrm{QD}}$ denotes the density of states (DOS) per unit cell 202 of the λth conduction band. These equations are solved given the following features of quantum dot memory cell 100: (1) The energy level of p_z orbital 208 is always above the Fermi energy E_F of top electrode 110 (in the range of bias considered). Thus, the electron injection from right electrode 110 can be ignored in Eqs. (12) and (14). (2) The Hamiltonian is spin independent, such that the occupation numbers for spin-up and spin-down states are equivalent in Eqs. (11), (13), and (14). That is, N_d,↑=N_d,↓.

In the simple case of a square lattice in which N_x=N_y, the DOS is approximated by a square pulse functionD_x(∈)=D_y(∈)=1/W for E_b∈<E_b+W,  (15)where E_b denotes the bottom of the conduction band of array 102 and W is the bandwidth of the conduction band. Such an approximation allows Eq. (11) to have a simple analytic solution of the formN_λ=g−cN_d,  (16)with g=[E_F+(1−α)eV_a−E_b]/(γW+3U) and c=2U_dc/(γW+3U), where γ=(Γ_L,c+Γ_R,c)/Γ_L,c. Substituting this expression into Eq. (13) gives rise to a simple transcendental equation which can be solved numerically. The resulting equation generates a maximum of three roots, of which only two are stable.

For a more realistic DOS derived using a 2-D tight-binding model, the coupled transcendental equations given in Eqs. (13) and (14) are solved numerically. The tight-binding model assumes that p_x,y orbitals 204, 206 are arranged on a rectangular lattice having lattice constants a and b. With this model, the band structure for the p_x band of array 102 is given byE_x(k)=E_p−2υ₁ cos(k_xa)−2υ_t cos(k_yb),  (17)where υ_i denotes the (ppσ) interaction within the band and υ_t denotes the (ppπ) interaction within the band. For the p_y band, the band structure is given byE_y(k)=E_p−2υ_l′ cos(k_yb)−2υ_t′ cos(k_xa).  (18)The DOS per unit cell 202 for the p_x band, for υ_l>υ_t, is given by

$\begin{array}{cc}{D}_x\left(\varepsilon \right)=\{\begin{array}{cc}\frac{1}{{\pi }^2}{\int }_0^{\pi }d{\eta \left[\begin{array}{c}{\left(2{\upsilon }_l\right)}^2-\\ {\left(\begin{array}{c}2{\upsilon }_l+\\ 2{\upsilon }_t\left(1-\cos \eta \right)-\tilde{\varepsilon }\end{array}\right)}^2\end{array}\right]}^{-1/2}\theta \left(\tilde{\varepsilon }-2{\upsilon }_t\left(1-\cos \eta \right)\right)& \mathrm{for}0<\tilde{\varepsilon }<4{\upsilon }_t\\ \frac{1}{{\pi }^2}{\int }_0^{\pi }d{\eta \left[\begin{array}{c}{\left(2{\upsilon }_t\right)}^2-\\ {\left(\begin{array}{c}\tilde{\varepsilon }-2{\upsilon }_t-\\ 2{\upsilon }_l\left(1-\cos \eta \right)\end{array}\right)}^2\end{array}\right]}^{-1/2}\theta \left(\tilde{\varepsilon }-2{\upsilon }_l\left(1-\cos \eta \right)\right)& \mathrm{for}4{\upsilon }_t<\tilde{\varepsilon }<4{\upsilon }_l\\ \frac{1}{{\pi }^2}{\int }_0^{\pi }d{\eta \left[\begin{array}{c}{\left(2{\upsilon }_l\right)}^2-\\ {\left(\begin{array}{c}2{\upsilon }_l+\\ 2{\upsilon }_t\left(1+\cos \eta \right)-\stackrel{\_}{\varepsilon }\end{array}\right)}^2\end{array}\right]}^{-1/2}\theta \left(\stackrel{\_}{\varepsilon }-2{\upsilon }_t\left(1+\cos \eta \right)\right)& \mathrm{for}0<\stackrel{\_}{\varepsilon }<4{\upsilon }_t,\end{array}& \left(19\right)\end{array}$ where {tilde over (∈)}=∈−E_p+2υ_l+2υ_t and ∈=E_p+2υ_l+2υ_t−∈. If υ_l<υ_t, then the roles of υ_l and υ_t are reversed in the expression for the density of states (i.e., Eq. (19)). The DOS described by Eq. (19) includes the van Hove singularities. A similar expression D_y(∈) holds for the density of states of the p_y band, with hopping parameters υ_l and υ_t=υ_t′ replaced by υ_l′=υ_l and υ_t′, respectively. By varying the hopping parameters υ_l, υ_l′, υ_t, and υ_t′ (e.g., by fixing lattice constant a and tuning lattice constant b, or vice versa), the behavior of the bistable tunneling current through quantum dot memory cell 100 can be studied for both one-dimensional systems and two-dimensional systems.

In general, a bistable tunneling current across quantum dot memory cell 100 is due to charge swapping between a layer with extended states (i.e., bottom electrode 108 or top electrode 110) and a layer having localized states (i.e., array 102 of quantum dots 104). The interplay of on-site Coulomb interactions between p_z orbitals and the delocalized nature of conduction band states derived from the hybridization of p_x,y orbitals gives rise to bistability. The model presented herein is valid in a regime defined by 4W>U.

As an example of the application of the approach described above, the coupled nonlinear equations given in Eqs. (13) and (14) are solved numerically using the following parameters: U_de=U=50 meV, Γ_L=1 meV (Γ_L,c=Γ_L/10), and Γ_R=1 meV (Γ_R,c=Γ_L/10). In general, the following parameters are used throughout the calculations presented herein: T=0K, υ_t′=5 meV, υ_l=20 meV, α=0.5, and E_F+V₀=E_p, where V₀ is a reference bias for V_a. The calculations and method presented herein are not bound to these parameters and apply to any quantum dot memory cell described by the equations presented above. The tight-binding parameters used in solving the above equations are assumed to scale according to 1/R², where R is the separation between two quantum dots 104 in array 102. Thus, υ_l′=υ_l(a/b)² and υ_t=υ_t′(a/b)². When a=b, the square lattice case is obtained.

Referring to FIGS. 3A and 3B, the auxiliary functions

$f_1=\frac{{\Gamma }_L+{\Gamma }_R}{{\Gamma }_L}{N}_d/\left(1-N_d\right)$ and f₂, which are the left-hand side and the right-hand side, respectively, of Eq. (13), are plotted versus N_d, the occupation number of p_z orbitals of quantum dots 104, for various values of the applied bias V_a. The results obtained using a constant DOS and a realistic DOS (i.e., using a tight-binding model) are shown in FIG. 3A and FIG. 3B, respectively. The two sets of curves are similar, except that the results in FIG. 3B obtained using a realistic DOS contain sharp features 302 caused by van Hove singularities in the density of states. The occupation number per spin component in a given conduction band, N_λ as a function of N_d is also shown in both FIG. 3A and FIG. 3B.

The intersection of f₁ and f₂ in FIG. 3 gives the solution to N_d. For low applied bias, only one root 304 to the equation f₁=f₂ exists. Similarly, for high applied bias, only one root 306 exists to the equation f₁=f₂. At moderate bias (in this example, for an applied bias between about 80 mV and 100 mV), three roots exist to the equation f₁=f₂. For instance, for an applied bias of 100 mV, the equation f₁=f₂ is solved by three roots: a first root 308, a second root 310, and a third root 312. Root 310 is an unstable solution; roots 308 and 312 are stable. The occupation number N_λ is also plotted as a function of N_d for various values of applied bias. Although van Hove singularities in the tight-binding density of states give rise to sharp features 302 in the plots of f₂ in FIG. 3B, the structure of N_λ does not exhibit abnormal features. This is because the p_z orbitals 208 of each quantum dot 104 are correlated with the p_x,y orbitals 204, 206 of the array 102 of quantum dots 104 via N_c, which is related to the integral over the density of states. This result remains similar if the constant density of states approximation is used (as described in Eqs. (15) and (16)) with the same bandwidth, W=4(υ_l+υ_t).

Referring to FIG. 4, the occupation number, N_d, is plotted as a function of applied bias V_a at zero temperature for a constant DOS as a curve 402 and for a realistic DOS using the tight-binding model as a curve 404. In both cases, array 102 of quantum dots is assumed to be a square lattice (i.e., a=b); other parameters are the same as those of FIGS. 3A and 3B. It is evident from this graph that the occupation number has bistable roots 406, 408; that is, for a range of applied bias V_a, a hysteresis region 410 exists in which two occupation numbers N_d are obtained. The fact that bistable roots are obtained using both a constant DOS and a more realistic DOS indicates that the hysteresis behavior of quantum dot memory cell 100 is not sensitive to the detailed shape of the band structure or the density of states of array 102 of quantum dots.

Once the occupation numbers are solved, the tunneling current is obtained by the relation

$J=\frac{e}{\hslash }{\Gamma }_R{N}_d\equiv J_0{N}_d,$ which is valid at zero temperature and when the carrier injection from top electrode 110 can be ignored. Consequently, the curve for N_d versus applied bias shown in FIG. 4 also illustrates the tunneling current across quantum dot memory cell 100. In the range of bias where multiple roots are allowed (i.e., in hysteresis region 410), turn-on and turn-off tunneling currents are determined by selecting either root 406 or root 408, depending on the value of the tunneling current corresponding to the previous value of V_a.

As the applied bias is turned on or off, the quantum dot memory cell selects the high conductivity state (larger N_d) or the low conductivity state (smaller N_d). Referring to FIG. 5A, the tunneling current through quantum dot memory cell 100 is plotted as a function of applied bias for various strengths of U_dc, the Coulomb interaction within a quantum dot 104. The tunneling current is in units of J₀=eΓ_R/h. Other parameters are the same as the parameters for FIGS. 3A, 3B, and 4. Curves 502, 504, and 506 correspond to values of U_dc of 50 meV, 30 meV, and 20 meV, respectively. The size of quantum dots 104 affects the value of U_dc; smaller quantum dots have a stronger Coulomb interaction and thus have larger values of U_dc. Curves 502 and 504 exhibit hysteresis regions 503 and 505, respectively, in which two tunneling currents are supported. For instance, curve 502 supports two tunneling currents 520 and 522. For weak Coulomb interactions (i.e., small values of U_dc as shown in curve 506), the bistable current vanishes. A large value of U_dc causes the bistability to be more robust against changes in temperature and broadening due to stray fields.

The critical level of Coulomb interaction needed to maintain a bistable current across the quantum dot memory cell depends on physical parameters of the quantum dot memory cell. These physical parameters include, for instance, tunneling rates between array 102 of quantum dots 104 and electrodes 108 and 110, temperature, bandwidth of the two-dimensional conduction band of array 102 of quantum dots, and charging energy of quantum dots 104. Although the calculations presented herein were performed at a temperature of 0 K, the bistable behavior of quantum dot memory cell 100 is maintained at higher temperatures, e.g., at room temperature, provided a charging energy of quantum dots 104 is sufficiently large (i.e., quantum dots 104 are sufficiently small) so as to overcome the thermal energy. The bistable current through the quantum dot memory cell does not exhibit negative differential conductivity, in contrast to bistable current through quantum well systems, which typically display negative differential conductivity.

Referring to FIG. 5B, the behavior of N_c/4, the total occupation number per unit cell 202 of the conduction band of array 102, is shown as a function of applied bias for the same set of Coulomb interaction strengths shown in FIG. 5A. Understanding the behavior of N_c is useful for clarifying the bistable behavior of N_d in the quantum dot memory cell. During the “turn-on” process, as the Fermi level in bottom electrode 108 is gradually raised from below the resonance level E_p (as shown in a region 510), the allowed solution to the occupation number N_d remains small (i.e., corresponding to tunneling current 520 in FIG. 5A) as a result of inter-level Coulomb blockade, and the energy of p_z orbitals 208 in array 102 of quantum dots is raised by an amount N_cU_dc, where the two-dimensional conduction band occupation number N_c is appreciable. Once the applied bias V_a reaches a critical value at which the inter-level Coulomb blockade is overcome (at a point 512), N_d increases quickly to a value around ⅓ (i.e., corresponding to tunneling current 522 in FIG. 5A). At this point, charges accumulate in the localized p_z orbitals 208 of quantum dots 102, leading to an increase of the self-energy of the conduction band states of array 102 by an amount 2U_dcN_d, hence reducing N_c to a much smaller value (i.e., reducing N_c from N_c2 to N_c3). This reduction in N_c in turn causes the energy of p_z orbitals 208 to decrease during the “turn-off” process (region 514) while N_d remains around ⅓ in order to keep the transcendental equation self-consistent. When the applied bias V_a reaches a lower critical value (at a point 516), N_c switches from N_c4 to N_c1 and N_d rapidly returns to the lower value (i.e., corresponding to tunneling current 520), which becomes the only self-consistent solution to the transcendental equation.

FIG. 5B demonstrates that bistable tunneling current arises from the repulsive Coulomb interactions U_dc within quantum dots 104. That is, the bistable tunneling current is due to an interplay of a Coulomb blockade effect for localized states of individual quantum dots 104 with the self-energy correction to the delocalized 2-D conduction band states formed by hybridized p_x,y orbitals in array 102 of quantum dots. In polaron models, a bistable behavior arises from an attractive potential having the form

$\frac{-2{\lambda }_p^2}{w_0}{d}_{l,\sigma }^{†}{d}_{l,\sigma }{d}_{l,-\sigma }^{†}{d}_{l,-\sigma }^{†},$ where λ_p and W₀ are an electron-phonon interaction strength and a phonon frequency, respectively. When the intralevel Coulomb interaction is contained within a single quantum dot (or, similarly, within a single molecule), the attractive potential giving rise to the bistable behavior is written as

$\left(U-\frac{-2{\lambda }_p^2}{w_0}\right)d_{l,\sigma }^{†}{d}_{l,\sigma }{d}_{l,-\sigma }^{†}{d}_{l,-\sigma }^{†}.$ Due to the large repulsive Coulomb interaction U, a net attractive electron-electron interaction mediated by phonons is difficult to achieve, which difficulty may explain why a bistable tunneling current through a single quantum dot junction has not yet been observed.

Referring to FIG. 6, the tunneling current through the p_z orbitals 208 of quantum dots 104 is shown as a function of applied bias for various ratios of b/a, which are the lattice constants for array 102 of quantum dots 104 as shown in FIGS. 2A and 2B. Other parameters are the same as those for FIG. 4. As the b/a ratio increases, the array 102 of quantum dots 104 in memory cell 100 approaches a one-dimensional (1-D) limit. Other parameters are the same as those used in the determination of the curves of FIG. 4. For a b/a ratio of 1 (shown as a curve 602), a b/a ratio of 2 (shown as a curve 604), and a b/a ratio of 3 (shown as a curve 606), the bistable behavior of the quantum dot memory cell is sustained. For b/a=1, a hysteresis region 603 is present; for b/a=2, a hysteresis region 605 is present; and for b/a=3, a hysteresis region 607 is present. However, the bistability becomes weaker and the hysteresis region becomes smaller as the 1-D limit is approached (i.e., as the b/a ratio increases). That is, hysteresis region 603 is larger in area than hysteresis region 605, which region is larger than hysteresis region 607. When the b/a ratio is greater than 3, the bistable current behaves essentially in the same manner as shown for the b/a=3 situation 606. Thus, b/a≈3 sufficiently represents a 1-D quantum dot memory cell capable of supporting a bistable current in a range of applied bias. However, approaching the limit of isolated quantum dots (i.e., as the bandwidth of the conduction band of array 102 decreases), the bistability disappears. Bistable current in 1-D quantum dot arrays is essential for the realization of high density nanoscale memory structures.

For a one-dimensional quantum dot memory cell, taking a=3 nm and b=9 nm (i.e., b/a=3) and considering 50 quantum dots 104 in order to establish a band-like behavior, the density of the quantum dot memory device is around 1/1350 nm²≈0.5 TB/in², which is significantly higher than the density of existing nonvolatile memory devices. A 1-D quantum dot memory cell is useful for specialized applications, such as an aligned memory. A 2-D memory is more generally applicable, however. For instance, a 2-D quantum dot memory with cells arranged in a checkerboard-like pattern is useful for dynamic random access memory (DRAM) applications. If 50 quantum dots were arranged in a close-packed 2-D array (i.e., b/a=1), the band-like behavior for the p_x,y states of the array would remain, thus increasing the memory density of the quantum dot memory cell by a factor greater than two.

Referring to FIG. 7, the tunneling current across quantum dot memory cell 100 is shown as a function of applied bias for various tunneling rate ratios Γ_L/Γ_R for the one dimensional limit (b/a=3), where Γ_L is the tunneling rate between bottom electrode 108 and array 102 of quantum dots and Γ_R is the tunneling rate between array 102 of quantum dots and top electrode 110. Other parameters are the same as those for FIG. 4. For a symmetrical quantum dot memory cell with Γ_L/Γ_R=1, the tunneling current is shown in curve 704, which displays a hysteresis region 705. When Γ_L is larger than Γ_R, such as shown in curve 706 for Γ_L=1 meV and Γ_R=0.1 meV (i.e., a shell-filling condition), the effect of U_dc is enhanced, leading to a larger hysteresis region 708 and a wider voltage range for bistable current. Conversely, the bistability behavior disappears for Γ_L=0.1 meV and Γ_R=1 meV (i.e., a shell-tunneling condition), as shown in curve 702, which does not have a hysteresis region. In the shell-tunneling condition, charges are unable to accumulate in p_z orbitals 208 of quantum dots 104. Consequently, the effect of U_dc is suppressed. Thus, it is possible to control the bistable current by adjusting the ratio of incoming to outgoing tunneling rate (i.e., the ratio of the tunneling rate from bottom electrode 108 to array 102 to the tunneling rate from array 102 to top electrode 110).

It is to be understood that the foregoing description is intended to illustrate and not to limit the scope of the invention, which is defined by the scope of the appended claims. Other embodiments are within the scope of the following claims.

Claims

1. A method of making a quantum dot memory cell, the quantum dot memory cell including an array of quantum dots disposed between a first electrode and a second electrode, the method comprising: obtaining values for a tunneling current through the quantum dot memory cell as a function of a voltage applied to the quantum dot memory cell, the values for the tunneling current determined on the basis of a density of states of the array of quantum dots;selecting parameters of the quantum dot memory cell such that the tunneling current through the quantum dot memory cell exhibits a bistable current for at least some values of the voltage applied to the quantum dot memory cell.

2. The method of claim 1, wherein the quantum dot memory cell further includes a quantum well positioned between the array of quantum dots and the first electrode.

3. The method of claim 1, wherein the values for the tunneling current are obtained on the basis of a first tunneling rate between the array of quantum dots and the first electrode and a second tunneling rate between the array of quantum dots and the second electrode.

4. The method of claim 1, wherein the values for the tunneling current are obtained on the basis of a first Fermi distribution function of the first electrode and a second Fermi distribution function of the second electrode.

5. The method of claim 1, wherein the values for the tunneling current are obtained on the basis of a first occupation number of the array of quantum dots, the first occupation number representing a number of electrons in orbitals coupling the quantum dots to the first electrode and the second electrode.

6. The method of claim 5, wherein the orbitals coupling the quantum dots to the first electrode and the second electrode are localized around each quantum dot.

7. The method of claim 5, wherein the values for the tunneling current are obtained on the basis of a second occupation number of the array of quantum dots, the second occupation number representing a number of electrons in one of a plurality of energy levels in extended states of the array of quantum dots.

8. The method of claim 1, wherein the tunneling current arises at least in part due to charge swapping between localized states in the array of quantum dots and extended states in at least one of the first electrode and the second electrode.

9. The method of claim 1, wherein the parameters include a composition of the quantum dots.

10. The method of claim 1, wherein the parameters include a separation between adjacent quantum dots in the array of quantum dots.

11. The method of claim 1, wherein the parameters include a size of the quantum dots.

12. The method of claim 1, wherein the parameters include a number of quantum dots in the array of quantum dots.

13. The method of claim 1, wherein the parameters include a composition of a material in which the array of quantum dots is embedded.

14. The method of claim 1, wherein the parameters include a composition of the first electrode and a composition of the second electrode.

15. The method of claim 2, wherein the parameters include a composition of the quantum well.

16. The method of claim 2, wherein the values for the tunneling current are obtained on the basis of an occupation number of the quantum well, the occupation number of the quantum well representing a number of electrons in a subband of the quantum well.

17. The method of claim 1, wherein the quantum dot memory cell exhibits bistable current when the voltage applied to the quantum dot memory cell is less than about 100 mV.

18. The method of claim 1, wherein the quantum dot memory cell is capable of operating with a switching speed of about 1 Terahertz.

19. A quantum dot memory cell comprising: a first electrode;a second electrode; andan array of quantum dots disposed between the first electrode and the second electrode,parameters of the quantum dot memory cell selected such that a tunneling current through the quantum dot memory cell exhibits a bistable current for at least some values of a voltage applied across the quantum dot memory cell, the tunneling current determined on the basis of a density of states of the array of quantum dots.

20. The quantum dot memory cell of claim 19, further comprising a quantum well positioned between the array of quantum dots and the first electrode.

21. The quantum dot memory cell of claim 19, wherein the array of quantum dots includes at least 50 quantum dots.

22. The quantum dot memory cell of claim 19, wherein the array of quantum dots includes a single layer of quantum dots in substantially all regions of the array of quantum dots.

23. The quantum dot memory cell of claim 19, wherein the array of quantum dots is substantially two-dimensional.

24. The quantum dot memory cell of claim 19, wherein the array of quantum dots is substantially one-dimensional.

25. The quantum dot memory cell of claim 19, wherein the tunneling current arises at least in part due to charge swapping between localized states in the array of quantum dots and extended states in at least one of the first electrode and the second electrode.

26. The quantum dot memory cell of claim 19, wherein the quantum dot memory cell exhibits bistable current when the voltage applied to the quantum dot memory cell is less than about 100 mV.

27. The quantum dot memory cell of claim 19, wherein the quantum dot memory cell is capable of operating with a switching speed of about 1 Terahertz.

28. The quantum dot memory cell of claim 19, wherein the quantum dots comprise a semiconductor.

29. The quantum dot memory cell of claim 20, wherein the quantum well comprises a semiconductor.