1: \documentclass[prb,twocolumn,showpacs,superscriptaddress,preprintnumbers,amssymb]{revtex4}
2: \usepackage{graphicx}% Include figure files
3:
4:
5: \begin{document}
6: \title{Quantum Wells in Polar-Nonpolar Oxide Heterojunction Systems}
7: \author{C.-C Joseph Wang}
8: \affiliation{Department of Physics, The University of Texas at Austin, Austin, Texas, 78712-0264, USA}
9: \author{Bhagawan Sahu}
10: \affiliation{Microelectronics Research Center, The University of Texas at Austin, Austin Texas 78758, USA}
11: \author{Hongki Min}
12: \affiliation{Department of Physics, The University of Texas at Austin, Austin, Texas, 78712-0264, USA}
13: \author{Wei-Cheng Lee}
14: \affiliation{Department of Physics, The University of Texas at Austin, Austin, Texas, 78712-0264, USA}
15: \author{Allan H. MacDonald}
16: \affiliation{Department of Physics, The University of Texas at Austin, Austin, Texas, 78712-0264, USA}
17:
18:
19: \date{\today}
20: \begin{abstract}
21: We address the electronic structure of quantum wells in polar-nonpolar oxide heterojunction systems
22: focusing on the case of non-polar BaVO$_3$ wells surrounded by polar LaTiO$_3$ barriers.
23: Our discussion is based on a density functional
24: description using the local spin density approximation
25: with local correlation corrections (LSDA+U). We conclude that a variety of quite different two-dimensional
26: electron systems can occur at interfaces between insulating materials depending on band line-ups and on
27: the geometrical arrangement of polarity discontinuities.
28: \end{abstract}
29:
30: \pacs{73.20.-r, 73.21.Cd, 75.70.-i}
31: \maketitle
32:
33: \section{Introduction}
34:
35: Complex oxide materials exhibit a wide variety of exotic phenomena from
36: high temperature superconductivity\cite{HS}, to colossal magnetoresistance\cite{CM},
37: to electronic and orbital order\cite{MIT}.
38: Recently, inspired by advances in epitaxial growth techniques,
39: there has been increasing interest\cite{Huang,sc,Brinkman,Ahn,Spaldin,Pickett,WC} in making progress toward
40: the goal of controlling oxide properties by designing layered structures at the
41: atomic level. The class of artificial materials which has been studied most
42: extensively is ABO$_3$ perovskites. Already there is evidence for
43: the emergence of superconductivity from non-superconducting materials\cite{sc}
44: and magnetism from nonmagnetic materials\cite{Brinkman}. One element which distinquishes
45: oxide heterojunction systems from weakly correlated semiconductor heterojunctions
46: is the possibility of exploiting gradients in polarity\cite{Sawatzky} to help control electronic properties.
47: In this article we study the electronic properties of oxide quantum wells in which the
48: barrier and well materials have very different polarities, using the case of
49: non-polar BaVO$_3$ wells surrounded by polar LaTiO$_3$ barriers as an example.
50:
51: Our paper is organized as follows.
52: In Section II, we present a qualitative discussion of four possible polar-nonpolar
53: quantum well geometries. We have tested conjectures made in Section II
54: by performing local spin-density-approximation
55: with local-correlation corrections (LSDA+U) calculations for BVO
56: quantum wells in LTO and LTO quantum wells in BVO.
57: Throughout this paper, we use LTO as shorthand for the perovskites associated with LaO and TiO$_{2}$ layers and BVO as shorthand
58: for the perovskites associated with BaO and VO$_{2}$ layers.
59: In Section III, we briefly detail the LSDA+U electronic structure
60: calculation methods we have employed to model LTO/BVO/LTO or BVO/LTO/BVO quantum well systems.
61: The combination (LTO/BVO) was chosen because LTO is polar and BVO is not and
62: because this material combination should not lead to large strains at the interface.
63: Partly for convenience and partly because multi-quantum-well
64: systems are likely to be of equal experimental interest, we assume periodically repeated quantum wells so that
65: we have a superlattice in the growth direction. In bulk both LTO and BVO are $d^{1}$ Mott
66: insulators. We focus on heterojunction systems composed
67: of materials which have the same $d$-band occupancies in the bulk because more possibilities
68: exist for interface electronic structure compared to the case of heterojunctions between
69: $d^{1}$ and $d^{0}$ materials such as LaTiO$_3$/SrTiO$_3$, for which interface metallicity appears
70: to be inevitable. In Section IV we present the results of our {\em ab initio} electronic structure calculations for
71: the two representative
72: quantum well systems and compare with the conjectures we made in section II.
73: Finally in Section V we briefly summarize our results.
74:
75: \section{Polar-Nonpolar quantum well structures}
76:
77: In this section, we discuss four different possible geometries for a polar-nonpolar
78: quantum well system. The main points we wish to make are shown in Fig.~\ref{fig:Fig1}
79: which illustrates approximate charge transfers and electric potentials. Throughout this discussion we
80: use LTO to represent a prototypical $d^{1}$ polar perovskite and BVO to represent a prototypical
81: $d^{1}$ nonpolar perovskite. We identify two possible geometries for both
82: LTO/BVO/LTO (case (a) and case (c))and BVO/LTO/BVO (case (b) and case (d))quantum well systems.
83: At an interface between two strongly correlated bulk materials
84: we should expect a complex
85: interplay between the ionic potential energy, kinetic energy, and correlation
86: energies of the two materials.
87: By using density functional theory with local correlations (LSDA+U)
88: we can take into account those energies approximately with a single electron picture.
89: The electronic structure is characterized most crudely by an ionic electron counting picture which neglects
90: hybridization between oxygen and transition metal orbitals and ignores the finite thickness of the various atomic layers.
91: Using this picture, the La$^{+3}$O$^{-2}$ planes in LTO
92: are positively charged and the Ti$^{+3}$O$_{2}^{-4}$ are negatively charged.
93: In bulk BVO on the other hand, the Ba$^{+2}$O$^{-2}$ and V$^{+4}$O$_{2}^{-4}$ planes are neutral in the bulk.
94: The transition-metal ions in both materials then have one electron in the $d$ orbitals to
95: maintain charge neutrality.
96: When we combine polar perovskites with nonpolar ones,
97: the transition-metal layer will distribute the $d$-charges in a different way to compensate
98: for the nonuniform distribution of
99: ionized layers in this system, as illustrated in Fig.~\ref{fig:Fig1}.
100: The distributions labeled {\em ideal conjecture} in this
101: figure assign the charge deficiency or excess to the two transition metal layers adjacent to the quantum well whereas
102: the distributions labeled {\em realistic conjecture} assume additional charge redistribution $\delta \rho$ due to the difference
103: between V and Ti $d$-orbital energies.
104: We discuss only cases in which the center of the quantum well is a center of inversion symmetry.
105: In each case the number of layers which would have a positive charge in a bulk polar crystal differs from the
106: number of layers which would have a negative charge.
107:
108: \begin{figure}
109: \includegraphics[width=2.50in]{Fig1a}
110: \hfill
111: \includegraphics[width=2.45in]{Fig1b}
112: \hfill
113: \includegraphics[width=2.45in]{Fig1c}
114: \hfill
115: \includegraphics[width=2.45in]{Fig1d}
116: \caption{(color online) Approximate charge density, electric field, and electric potential distributions in
117: different polar-nonpolar quantum well systems with inversion symmetry in the quantum well center. The arrows (black on line)
118: indicate the expected charge densities $\rho$ centered on different atomic layers, the piece-wise constant thick lines lines (blue on line)
119: indicate the expected electric fields $E$ between the atomic layers, and the cusped lines (red on line) indicate the
120: expected electric potential $-V(z)$ distribution.
121: The signs $+$ and $-$ stand for the direction of electric field $E$, the value for $\rho$, and the potential $-V(z)$.
122: }\label{fig:Fig1}
123: \end{figure}
124:
125:
126: In case (a), we have two BaO layers and one VO$_{2}$ layer inserted in the
127: unit cell of a superlattice with six LaO$^{+}$ layers but seven TiO$_{2}$ layers.
128: The charges in the supercell cannot be neutral if we assume that each TiO$_{2}$ is negatively charged as in
129: the bulk LTO states. Instead,
130: we expect that the extra electrons will be distributed evenly
131: among the TiO$_{2}$ layers near the central VO$_{2}$ layer.
132: The easiest way to accomplish this is to have the
133: TiO$_{2}$ layers next to BaO layers to be Ti$^{+3.5}$O$_{2}^{-4}$
134: rather than Ti$^{+3}$O$_{2}^{-4}$, {\em i.e.} that they
135: contain 0.5 $d$-electrons per Ti rather than 1.0 $d$-electrons
136: per titanium. This idealized conjecture ignores $d-p$ hybridization as well
137: as the difference between the $d$-orbital energies on Ti and V sites. In particular a
138: V$^{+4}$ ion should have a lower intrinsic atomic levels than the Ti$^{+3}$ ion.
139: We should expect that charges might react locally
140: to this difference by transferring additional electrons $\delta \rho$ to the VO$_{2}$ well from nearby TiO$_{2}$ layers.
141: This additional charge transfer partially screens the original atomic energy level difference.
142: These considerations are illustrated schematically in Fig.~\ref{fig:Fig1}.
143: The $d$-electron system electronic structure reconstructs in this way near the interface in reaction to the
144: electrostatic potentials produced by the bulk crystal layer polarizations.
145: Qualitatively similar considerations would apply to wider BVO quantum wells.
146:
147: In case (b), we have two LaO$^{+}$ layers and one TiO$_{2}$ barrier layer in the middle of the supercell. By the same arguments as case (a),
148: the simplest conjecture is that the VO$_2$ layers nearest the TiO$_2$ layers will be
149: negatively charged (V$^{+3.5}$O$_{2}^{-4}$) to keep the system neutral.
150: The charge configuration Ti$^{+3}$O$_{2}^{-4}$ is maintained by the protection of two surrounding LaO$^{+}$ layers.
151: Therefore, there will be 1.5 $d$-electrons on V sites in the V$^{+3.5}$O$_{2}^{-4}$ layers and 9 $d$-electrons in the supercell.
152: However, the local physics is different in this case. The electron can respond locally by transferring electrons from the barrier TiO$_{2}$ to the VO$_{2}$ well layers. The reconstructed potential should therefore be lower in the well region in this case.
153:
154: In case (c), we switch the ordering of layers from case (a) (BaO $\leftrightarrow$ VO$_{2}$ and LaO $\leftrightarrow$ TiO$_{2}$).
155: There are 7 positive charged LaO$^{+}$ layers and 6 Ti$^{+3}$O$_{2}^{-4}$ layers. To compensate for the charge difference,
156: VO$_{2}$ layers have to be negatively charged as V$^{+3.5}$O$_{2}^{-4}$ in the idealized conjecture illustrated.
157: Additional charges can
158: transfer from adjacent Ti$^{+3}$O$_{2}^{-4}$ to the $V^{+3.5}$O$_{2}^{-4}$ layers.
159: The reconstructed potential should therefore be higher
160: in the well region.
161:
162: Similarly in case (d), the structure has been changed from case (b) by switching the ordering
163: BaO $\leftrightarrow$ VO$_{2}$ and LaO $\leftrightarrow$ TiO$_{2}$. There is only one LaO$^{+}$ layer in this case.
164: The abutting TiO$_{2}$ layers should have the configuration Ti$^{+3.5}$O$_{2}^{-4}$ to keep charge neutrality,
165: resulting in a lower reconstructed local potential in the center of the cell.
166: Notice that the magnitude of the additional charge transfer and its
167: spatial distribution expected in the
168: realistic pictures needs to be determined by self-consistent microscopic calculations.
169: In the next section, we will focus on cases (a) and (b) to check the degree to which
170: the qualitative pictures explained here agrees with the results of microscopic DFT calculations and to
171: obtain further insight into the nature of the two-dimensional electron systems which
172: these electronic reconstructions enforce. To avoid confusion to the readers, in microscopic calculation, the $d$ electrons should be
173: interpreted as the electrons occupying the bands, which hybridized with $s$, $p$ orbitals, with dominant $d$ character at transition-metal sites.
174:
175:
176: \section{Electronic Structure Calculations}
177: In our {\em ab initio} electronic structure calculations we have considered symmetric quantum
178: well geometries only, use the experimental atom positions\cite{Koreans,data} for LTO,
179: and neglect atomic relaxation in the BVO layers. (The room temperature crystal structure of BVO is not known
180: experimentally.)
181: The electronic structure calculations were performed using DFT with LSDA+U as implemented in the software package VASP
182: \cite{VASP1,VASP2,VASP3}.
183: Projector-augmented wave pseudopotentials\cite{PAW} are used to describe the
184: electron-ion interaction. We sample the full superlattice Brillouin
185: zone (BZ) with a 5 $\times$ 5 $\times$ 3 mesh and used an energy cutoff of
186: 410 eV which we found to be sufficient to reproduce
187: bulk LTO properties.
188: We used PAW pseudopotentials supplied by the VASP code for La, O, Ba, Ti and V but for Ba, Ti and V, we used the one with semicore states treated as valance states.
189: The screened local correlation $U$ in the partially filled $d$
190: bands of the transition metal elements were treated using the rotationally invariant LSDA+U method due to Dudarev et al.{\cite{LSDA+U}}.
191: Since the La $f$ bands lie much higher in energy in experiments\cite{La, Spaldin}
192: we impose a much larger $U_{La}=11$ eV on La $f$ states to prevent their mixing with $d$ bands at low energy.
193: We chose $U_{Ti} = 3.2$ eV and $J_{Ti} = 0.9$ eV since these values reproduces the Mott-insulating nature of bulk LTO\cite{EPL} and
194: we take $U_{V}$ and $J_{V}$ to be the same as $U_{Ti}$ and $J_{Ti}$. We have
195: varied $U_{V}$ to check the robustness of our calculations
196: and find that our conclusions are independent of this LSDA+U parameter.
197: \begin{figure}
198: \includegraphics[width=3in]{Fig2}
199: \caption{(color online) Half of the inversion symmetric unit cell.
200: The quantum well consists of two layers of BaO and a monolayer of VO$_{2}$ and is
201: embedded in bulk LaTiO$_{3}$. The light-orange shadowed region indicating the narrow
202: active region in which the electronic interface reconstruction (EIR)
203: driven by the polar discontinuity takes place.
204: Oxygen atoms are located at the intersections
205: of the checkered lines and form octahedra
206: around the Ti and V atoms. The numbers specify labels for different layers.
207: The downward arrow indicates the location
208: of the VO$_{2}$ inversion plane. }\label{fig:Fig2}
209: \end{figure}
210:
211: \section{LSDA+U quantum-well electronic structure}
212: \subsection*{Case A : (LaTiO$_{3}$)$_{3}$/(BaO)$_{2}$VO$_{2}$/(LaTiO$_{3}$)$_{3}$TiO$_{2}$}
213: The unit cell of the superlattice consists of a single VO$_2$ layer at the center of the well surrounded by
214: two BaO layers as illustrated in Fig.~\ref{fig:Fig2}. Because of the smaller $p$-$d$ energy separation in bulk
215: BVO compared to LTO, this material combination should form a quantum
216: well centered on the VO$_{2}$ layer, assuming that the $p$-bands of the two materials are nearly aligned at the
217: interface. In our superlattice the BVO quantum well layers are separated
218: by 7 TiO$_2$ transition metal layers, each in turn separated by a LaO layer.
219: Half of the inversion symmetric unit cell
220: is shown in Fig.~\ref{fig:Fig2} in which the VO$_{2}$ inversion plane is marked by an arrow.
221: Ti and V in bulk LTO and BVO respectively both have nominal valence charge $d^{1}$.
222: In the quantum well structure we study here the number of LaO layers, which are
223: positively charged in the polar LTO crystal, is one smaller than the number of
224: TiO$_2$ layers. It follows that the total $d$ charge per transition metal
225: summed over all layers is expected to be reduced by $1$. Unless the inversion symmetry is
226: broken we should expect the missing $d$-charge to be symmetrically distributed around the
227: quantum well center. The simplified assumptions of the previous section would
228: divide this valence change between the two TiO$_2$ layers adjacent to the quantum well.
229:
230: Fig.~\ref{fig:Fig3} shows the $d$-projected DOS at a Ti site and a V site in each TiO$_{2}$ and VO$_{2}$ plane.
231: The states below $- 1.5$ eV are the $d-p$ bonding states between Ti, V atoms and their neighboring oxygens
232: which have dominant $p$-character,
233: whereas the states near the Fermi level (E $=$ E$_{F}$) are the anti-bonding states which have dominant $d$-character.
234: As we move through the superlattice from TO$_{2}$ barrier layers to the VO$_{2}$
235: we see the DOS evolve from the typical bulk LTO form (layers $1$ to $3$) to
236: a new form near the heterojunction. In particular,
237: there is a large depletion away from atomic-like $d^{1}$-charges
238: in layer 4.
239: The electronic reconstruction which accommodates the band offsets and the
240: polarity discontinuity is evident very strongly in the layers 5 and 6 surrounding
241: the quantum-well layer 4 (layer 6 is not shown in Fig.~\ref{fig:Fig3} to avoid redundancies because of the inversion symmetry of the system).
242: We also observe that the $p$-$d$ separation is maintained at the
243: VO$_2$ layer as expected after the electronic reconstruction.
244: The total electron transfer from layer 4 is larger than in the {\em ideal conjecture}, as
245: expected, and more remote layers participate in the transfer because of oxygen mediated coherence between transition
246: metal orbitals centered on different layers.
247:
248: \begin{figure}
249: \includegraphics[width=3in]{Fig3}
250: \caption{(color online) Layer-resolved $d$-projected density of states at Ti and V sites versus energy.
251: The layer labeling is defined in Fig.~\ref{fig:Fig2}.}\label{fig:Fig3}
252: \end{figure}
253: Most Mott insulators are orbitally and magnetically ordered. In order
254: to gain some insight into the way in which the magnetic order is disturbed by the
255: quantum well structure, we plot in Fig.~\ref{fig:Fig4} the spin-resolved $d$-projected DOS in
256: all layers. In our calculations we find G-type antiferromagnetic (AFM) order in which neighboring transition metals
257: ions have opposite spins both within and between layers for the bulk like TiO$_{2}$ layers. (layers 1, 2, 3 in Fig.~\ref{fig:Fig4}).
258: Layer 4 on the other hand has an overall magnetic
259: moment of around 0.1 $\mu_{B}$ per Ti ion indicating a ferromagnetic layer.
260: Because of the electronic interface reconstruction (EIR), the magnetic order in the VO$_{2}$ plane changes from AFM to ferromagnetic.
261: The projected moments on each Ti ions in the AFM bulk-like TiO$_{2}$ layers
262: is calculated to be 0.74 $\mu_{B}$ inside the Ti ionic radius which is fixed in our calculation by the
263: pseudopotential construction. We believe that by enlarging the Ti ionic radius, we can get $d$-orbital integrated moments
264: close to the 1 $\mu_{B}$ value expected for $d^{1}$ Mott insulators.
265: Note that ferromagnetism is induced in the VO$_2$ layer by the two adjacent strongly depopulated and
266: ferromagnetic TiO$_2$ layers.
267:
268: Fig.~\ref{fig:Fig5} shows the superlattice band structure along high symmetry directions in the Brillouin zone.
269: The bands (five majority (red) spin bands and five minority (blue) spin bands), which
270: lie far below the E$_{F}$ (in the energy range between
271: between $- 1$ eV and $- 0.6$ eV), can be identified as bulk LTO $d$-charge character
272: bands by comparing with Fig.~\ref{fig:Fig4} and
273: calculating band-decomposed local charge density contributions. In the supercell, we have 5 bulk TiO$_{2}$ layers (layer 1, 2, 3, 7, 8)
274: in our superlattice corresponding in the antiferromagnetic state to 10 full bands (5 spin-up and 5 spin-down).
275: Therefore, there should be ten full bulk like LTO $d$-bands, identified
276: by the solid circle, in Fig.~\ref{fig:Fig5}.
277: Two full majority spin bands and one full minority spin band (near -1 eV) highlighted by two dashed circles are associated
278: with the $d$-charges in the VO$_2$ layers.
279: The two majority bands close to the interface states in energy and are hybridized along the layer-growth direction and delocalized
280: (This interpretation has been verified by separating the charge density contribution of those bands) .
281: Because of this hybridization, the total $d$-charges (2.4053 $e^{-}$) at the V sites is smaller
282: than the 3 $e^{-}$ value which would be implied by
283: a local interpretation of the band structure.
284: It can in general be misleading to estimate the number of charges at transition metal ions by counting the number
285: of reconstructed bands with a particular dominant orbital character.
286: The three partially full bands (one minority band and two majority bands) near the fermi surface E$_{F}$ are the bands
287: responsible for the charge depletion at the TiO$_{2}$ layers adjacent to the
288: quantum well (layer 4 and 6).
289: The momentum k$_{z}$ dependence of the electronic states is shown in the right panel of Fig.~\ref{fig:Fig5} (from momentum (0 , 0, 0) at $\Gamma$ to (0, 0, $k_z$) at Z). It shows that, the states at the Fermi energy form a two-dimensional electron gas
290: centered on three layers, the active region.
291: Notice that there also appears a flat band between M and X, which corresponds to the direction away from the bonding direction between Ti and O atoms
292: in charge depleting the TiO$_{2}$ layers, due to the fact there is negligible tunneling
293: of electrons along this direction.
294: The two dimensional system is metallic in this case and has three partially filled bands.
295: \begin{figure}
296: \includegraphics[width=3in]{Fig4}
297: \caption{(color online) Spin and layer resolved densities-of-states versus energy. The labels on top are layer labels as in Fig.~\ref{fig:Fig2}.
298: Since we allow G-type AFM order in the barrier material, there are two distinct transition metal atoms in each layer.
299: The DOS for majority spins and minority spins are represented by
300: red and blue lines respectively and the Fermi level is marked with a horizontal dark line at $E=E_{F}$.}\label{fig:Fig4}
301: \end{figure}
302:
303: \begin{figure}
304: \includegraphics[width=4.5in]{Fig5}
305: \caption{(color online) Spin-resolved band structure at the Brillouin zone. The horizontal axis represents superlattice crystal momentum.
306: The red solid lines represents the majority-spin bands and the blue ones are the minority-spin bands.
307: The momentum dependence along the layer direction $k_{z}$ is indicated in the region between the $\Gamma$ and Z points.
308: The full circle marks the $d^{1}$ bands for Ti atoms in LTO bulk states. The dashed circles represents the $d$ bands for V atoms in VO$_{2}$.
309: The bands associated with EIR are flat along k$_{z}$ demonstrating their quasi-two-dimensional character.
310: The shadowed blue square represents the projection of the superlattice Brillouin zone in the plane of the quantum well.
311: }\label{fig:Fig5}
312: \end{figure}
313: Since we have discussed the local potential for the electron in case (a) in Fig.~\ref{fig:Fig1} based on qualitative model
314: in which we assume the charges at each layer are concentrated in a single layer along the layer direction, we would like to examine how
315: the the crude model prediction compares with actual DFT predictions.
316: In Fig.~\ref{fig:Fig6}, the DFT plane-averaged Hartree potential for the electrons is shown.
317: First TiO$_{2}$ layer starts at $z = 0$ and layers repeat with a spatial period of 3.96 ${\AA}$ and with the
318: central layer ($z = 15.84$ ${\AA}$) of the supercell substituted by the VO$_{2}$ layer.
319: First LaO layer is located at $z = 1.98$ $\AA$ and these layers repeat
320: with a period of 3.96 ${\AA}$ except that the two layers surrounding the central VO$_{2}$ layer
321: correspond instead to BaO layers.
322: By inspecting the variation of the local minima in the potential, we can see that the qualitative behavior
323: anticipated in case (a) in Fig.~\ref{fig:Fig1} is valid, as shown by the black-dashed line. The
324: electric field in the qualitative model should be considered as a macroscopically averaged electric
325: field between potential minima in the microscopic calculations.
326: Notice that the potential of the VO$_{2}$ layer is somewhat lower than anticipated.
327:
328: In Fig.~\ref{fig:Fig7}, we show the linear electron number density $\lambda(z)\equiv 0.5 \int \rho(x,y,z) dx dy$ along the layer-growth direction.
329: The charge number spatial distribution $\rho(x,y,z)$ is normalized per transverse unit cell which is doubled because of the system's
330: antiferromagnetic order, and includes the total number of electrons contributed by the $d$-bands between
331: with energies up to $1.5$ eV below the Fermi level. (See Fig.~\ref{fig:Fig5}.)
332: The factor $0.5$ is used to convert the normalization to $d$-electrons per transition metal ion,
333: to facilitate comparison with the qualitative picture.
334: The dashed lines represent the boundaries for the transition-metal oxide layers used in this
335: electron number construction. The area under the curve $\lambda(z)$ represent the
336: total number of electrons from all $d$-bands.
337: We see an apparent charge depletion at the TiO$_{2}$ layers centered at $z = 11.88$ ${\AA}$ and $z = 19.8$ ${\AA}$ along with large charge accumulation at the VO$_{2}$ layer. There is only minor charge depletion at second-nearest TiO$_{2}$ layers ($z = 7.92$ ${\AA}$ and $z = 23.76$ ${\AA}$) away from VO$_{2}$ layer.
338: The total charges from the $d$-bands, 7 $e^{-}$, is obtained by adding the number of electrons at each layers
339: and agrees with the qualitative predictions for case (a).
340: The number of electrons in the depleted TiO$_{2}$ layers is $0.3793$. The other layers have small deviation from one electron.
341: The calculation shows a clear deviation from ideal conjecture we made for this case and is closer
342: to the {\em realistic conjecture}.
343: In general, the charges in the well are mainly transferred from the adjacent TiO$_{2}$ layers but there is also
344: a minor contribution from
345: other TiO$_{2}$ layers as shown from the number of electrons in each layers.
346:
347: \begin{figure}
348: \includegraphics[width=3in]{Fig6}
349: \caption{(color online) Plane averaged Hartree potential along the layer-growth direction $z$.
350: The blue curve represents the Hartree potential experienced by electrons $-e V(z)$.
351: The dashed line indicates the net potential variation between charge-concentrations centers
352: in each layers.}\label{fig:Fig6}
353: \end{figure}
354:
355: \begin{figure}
356: \includegraphics[width=3in]{Fig7}
357: \caption{(color online) Linear electron number density $\lambda(z)$ along layer-growth direction z. The dashed lines represent
358: the boundaries used for the partitioning of density into contributions from different layers.
359: The total number of electrons in each layer contributed by the $d$-character bands in Fig.~\ref{fig:Fig5} is
360: indicated by the numerical values
361: between dashed lines. In this case, the total number of electrons in the supercell is seven.}\label{fig:Fig7}
362: \end{figure}
363:
364: \subsection*{Case B : (BaVO$_{3}$)$_{3}$/(LaO)$_{2}$TiO$_{2}$/(BaVO$_{3}$)$_{3}$VO$_{2}$}
365: In this section, we consider the quantum well structure shown in case (b). In this case, the number of LaO layers, which are
366: positively charged in the polar LTO crystal, is greater than the number of
367: TiO$_2$ layer by one. It follows that the total $d$-electrons per transition-metal
368: summed over all layers is expected to be increased by $1$. Unless the inversion symmetry is
369: broken we should expect the extra $d$-charge to be symmetrically distributed around the
370: quantum well center.
371:
372: Fig.~\ref{fig:Fig8} shows the $d$-projected DOS at a Ti site and a V site in each TO$_{2}$ and VO$_{2}$ plane.
373: The states below $-1$ eV are the $p-d$-bonding states between the transition-metal atoms (Ti, V) and the $p$ orbitals of the neighboring oxygens
374: and have dominant $p$ character.
375: The states near the Fermi level (E $=$ E$_{F}$), on the other hand, have dominant $d$-character.
376: As we move through the superlattice from the VO$_{2}$ layers to the TiO$_{2}$ barrier layer
377: we see the DOS evolve from the typical bulk BVO layer form (layers $1$ to $3$) to
378: a new form near the heterojunction. In particular, we see a strong peak in DOS
379: in layer 4 near the Fermi energy $E_F$ with 1.6 $e^{-}$ per V ion in agreement with our conjectures.
380: \begin{figure}
381: \includegraphics[width=3in]{Fig8}
382: \caption{(color online) Layer-resolved $d$-projected density of states at Ti and V sites versus energy.}
383: \label{fig:Fig8}
384: \end{figure}
385: As far as magnetic order is concerned (Fig.~\ref{fig:Fig9}),
386: we find G-type AFM order in which neighboring transition metals
387: ions have opposite spins both within and between layers for the bulk like VO$_{2}$ layers (layers 1 and 2).
388: The magnetic orders in layer 3 and 5 are reconstructed to be ferromagnetic with the magnetic moment 0.9755 $\mu_B$ per V ion and the moment 0.4750 $\mu_B$ per Ti respectively.
389: The magnetic order in layer 4 is ferrimagnetic with a net magnetic moment 0.2170 $\mu_{B}$. In
390: the corresponding band structure illustrated in Fig.~\ref{fig:Fig10}, we
391: identify three partially filled
392: two-dimensional bands that cross
393: the Fermi level, and sixteen full $d$-bands.
394: These bands accommodate 9 $d$-electrons obtained by taking the average $d$-charges per transition metal ion and summing the charges of all the layers,
395: as illustrated in Fig.~\ref{fig:Fig12} discussed below.
396: The two majority spin partially-filled bands, and the single minority-spin partially-filled band, are
397: responsible for the extra charge at V sites in layer 4 and 6 as shown in Fig.~\ref{fig:Fig9}.
398: The three full minority-spin bands, which are lower in energy than the partially-filled minority-spin band
399: but closer to the Fermi level than other full minority-spin bands
400: mainly contribute to minority-spin charge concentrated in layers 4, 5, and 6.
401: The wave functions for those bands are
402: strongly hybridized within those layers.
403: The lowest majority-spin band in Fig.~\ref{fig:Fig10}
404: is mainly of V character in layer 4 but lower in energy due to the
405: more attractive local potential at this site.
406: The peak of the DOS for the minority spin at the V site seems to shift higher toward
407: the Fermi energy than expected.
408: This is due to the fact those states are strongly hybridized with the states at the
409: Ti sites which are shifted toward higher energy, an effect not included in the qualitative analysis.
410: (See the potential profile of the realistic conjecture of case (b) in Fig.~\ref{fig:Fig1}).
411: Other full bands are responsible for BVO bulk states away from the region in which
412: the electronic interface reconstruction occurs.
413:
414: \begin{figure}
415: \includegraphics[width=3in]{Fig9}
416: \caption{(color online) Spin and layer resolved densities-of-states versus energy. The labels on top are the layer labels
417: indicated in Fig.~\ref{fig:Fig2}.
418: Since we allow G-type AFM order in the barrier material, there are two distinct transition-metal ions in each layer.
419: The DOS for majority spins and minority spins are represented by
420: red and blue lines respectively.}\label{fig:Fig9}
421: \end{figure}
422:
423: \begin{figure}
424: \includegraphics[width=3.5in]{Fig10}
425: \caption{(color online) Spin-resolved band structure. The horizontal axis represents superlattice crystal momentum.
426: The red solid lines represent the majority-spin bands while the blue lines represent minority-spin bands.
427: The momentum dependence along the layer direction $k_{z}$ is captured in the region between the $\Gamma$ and Z points.
428: The three partially occupied bands associated with EIR are flat along k$_{z}$, demonstrating
429: their quasi-two-dimensional character.}\label{fig:Fig10}
430: \end{figure}
431:
432: In Fig.~\ref{fig:Fig11}, we plot plane-averaged Hartree potentials.
433: The first layer starts with VO$_{2}$ at $z = 0$. Transition metal layers repeat with a spatial period of 3.96 ${\AA}$
434: with the center ($z = 15.84$ ${\AA}$) layer occupied by TiO$_{2}$ layer.
435: At $z = 1.98$ $\AA$, the first BaO layer starts and repeats also at a period of 3.96 ${\AA}$; the local minima closest to TiO$_{2}$
436: correspond instead to LaO layers.
437: By inspecting the local minima in the Hartree potential, we can see that the
438: DFT calculations largely verify the scenario of case (b) in Fig.~\ref{fig:Fig1} except that the dashed line wiggles in the region of bulk BVO. The wiggling may due to the slight difference of the strength of ionic bonding between Ba-O and V-O along
439: the layer direction, which causes the small dipole modification of charge distribution in BVO ionized layers.
440:
441:
442: In Fig.~\ref{fig:Fig12}, we show the corresponding electron linear number density $\lambda(z)\equiv 0.5 \int \rho(x,y,z) dx dy$ along the layer-growth direction.
443: This charge number spatial distribution $\rho(x,y,z)$ include the total charges contributed from the bands between 0 and - 0.9 eV with respect to Fermi level in Fig.~\ref{fig:Fig10}.
444: We see an apparent charge accumulation (1.5144 $e^{-}$) at the TiO$_{2}$ layers centered at $z = 11.88$ ${\AA}$ and $z = 19.8$ ${\AA}$ along with much less charge (0.6295 $e^{-}$) at the TiO$_{2}$ layer.
445: Remote transition metal layers have small deviations from a $d^{1}$ valence.
446: The total amount of charges from the $d$-bands is 9 $e^{-}$ by adding the charges at each layer in agreement with our prediction in case (b).
447: The calculation shows small deviations from the ideal conjecture we made for this case and almost the realistic conjecture.
448: In general, the charges in the well are transferred from the TiO$_{2}$ layer to adjacent VO$_{2}$ layers as discussed in section II.
449:
450: \begin{figure}
451: \includegraphics[width=3in]{Fig11}
452: \caption{(color online) Plane averaged Hartree potential along the layer-growth direction $z$.
453: The solid dashed line indicates the net potential variation between charge-concentrated centers at each layer.}\label{fig:Fig11}
454: \end{figure}
455:
456: \begin{figure}
457: \includegraphics[width=3in]{Fig12}
458: \caption{(color online) Linear electron number density $\lambda(z)$ along layer-growth direction z. The dashed lines represent
459: the boundaries between different layers used in constructing the layer decomposition of this charge density.
460: The numerical numbers
461: between dashed lines in each layers are the total number of electrons from $d$-derivative bands in Fig.~\ref{fig:Fig10}.
462: In this case, the total number of electrons is nine.}\label{fig:Fig12}
463: \end{figure}
464:
465:
466: \section{Summary and Discussion}
467: In this paper we have explored some possibilities
468: for the realization of artificial two-dimensional electron systems
469: by growing quantum wells consisting of a few atomic layers of one layered
470: oxide surrounded by another material. In particular we consider systems
471: which involve heterojunctions between polar insulating perovskites like LTO and
472: nonpolar insulating perovskites like BVO. Perovskites can be viewed as consisting of
473: alternating AO and BO$_2$ layers where B is a transition metal.
474: The main idea of this paper is that inversion symmetric
475: quantum wells will necessarily have a different number of
476: AO and BO$_2$ layers, and that this will necessarily lead to
477: an electronic reconstruction which induces
478: two-dimensional electron systems near the quantum well when
479: one material is polar and one is nonpolar.
480: Four different cases can be identified, in which the quantum well
481: material is either polar or non-polar and in which it either has an
482: excess AO layer or an excess BO$_2$ layer.
483: The excess layers induce either a unit increase or a unit decrease in the total
484: $d$-orbital occupancy per transition metal site when summed over
485: all layers. This charge must be localized close to the quantum well to avoid
486: large band shifts due to polarization driven electrostatic potentials.
487: As long as inversion symmetry around the quantum well center
488: is not-broken spontaneously, half of this charge must appear on each half
489: of the quantum well center. Because of the strong-correlation character of
490: the oxide electronic structure, these charges will tend to be localized
491: in half-filled LSDA+U bands which are weakly hybridized so that both cross the
492: Fermi energy.
493:
494: We have tested and elaborated on this qualitative picture of the electronic structure of polar-nonpolar heterojunction
495: quantum wells in perovskites by performing self-consistent LSDA+U electronic structure calculations for
496: a representative system composed of BVO and LTO without any atomic relaxation at the interface.
497: The LSDA+U calculations capture the interplay between long range polar electrostatic energy and local physics driven by
498: energetic offsets between different transition metal ions and by $p-d$ hybridization.
499: We find that in the two cases we have examined one additional two-dimensional bands is induced by the energetic offsets.
500: In all cases the quantum well structure creates a two-dimensional metal inside a three dimensional Mott insulator,
501: and alters the character of the magnetic order.
502: The reconstructed magnetic order is ferromagnetic in the active well region and antiferromagnetic in bulk region.
503: The appearance of two-dimensional ferromagnetism in intimate exchange contact with antiferromagnetism
504: is intriguing from the point of view of spintronics.
505:
506: %{\bf Allan: Check this carefully. All OK?}
507: Because of the complexity of oxide materials, our calculation should be viewed as an exploration of
508: possibilities rather than as predictive in any detailed sense. Among the many sources of uncertainty,
509: the Hubbard U interaction used in our LSDA+U is a phenomenological parameter for which we have
510: chosen a standard bulk values. This value should likely be renormalized near the
511: quantum well. More challenging to account for is the role of inevitable lattice distortions,
512: oxygen stoichiometry variations, and other defects near the
513: interfaces. The interfaces between the thin film oxide and the substrate on which it is grown
514: can also play a complicating role if either the barrier material or the substrate is polar.
515: For a few cases we have explored the influence of
516: lattice distortion effects by allowing the atoms to relax along the layer growth direction in the active region,
517: maintaining $U_{V}$=$U_{Ti}$$=3.2$ eV as before.
518: What do we find is that the amount of charge transfer to or from particular layers and
519: the details of the magnetic orders are highly sensitive to ionic relaxation.
520: These complications make the accurate prediction of the electronic reconstruction and the
521: associated magnetic order highly nontrivial. It is also desirable to understand if exotic symmetry breaking phases
522: such as superconductivity may appear in the interface .
523: Further progress in understanding
524: detailed properties of the two-dimensional electron systems we envision which almost certainly
525: require inputs from experiments. The microscopic calculations reported on here can be used as guidance
526: in constructing phenomenological models which can be fit to any data which might emerge from future work and serve
527: as the theoretical basis for possible exotic phases in the interface.
528: The present paper has addressed $d^{1}$ transition metal systems.
529: Considerations similar to those explained here also apply to
530: transition-metal oxides with heavy transition metal elements (Cu, Ni) in which the $t_{2g}$ bands are fully
531: occupied far below the Fermi level, and low-energy excitations
532: are expressed within the $e_{g}$ bands.
533:
534: \section*{Acknowledgements}
535: This work was supported by the National Science Foundation under grant
536: DMR-0606489, by the Welch Foundation
537: (Houston, TX) under Grants No. F-1473 and No. F-0934,
538: by the Texas Advanced Computing Center (TACC), University
539: of Texas at Austin. B.S. thanks SRC-NRI
540: (SWAN) for financial support.
541: C.-C Joseph Wang gratefully
542: acknowledge helpful conversations with Peter Abbamonte.
543: *Electronic address: joseph@physics.utexas.edu
544:
545: \begin{thebibliography}{99}
546: \bibitem{HS} Patrick A. Lee, Naoto Nagaosa, and Xiao-Gang Wen,
547: Rev. Mod. Phys. {\bf 78}, 17 (2006).
548: \bibitem{CM} A. P. Ramirez, J. Phys.: Condens. Matter {\bf 9}, 8171-8199 (1997).
549: \bibitem{MIT} M. Imada, A. Fujimori, and Y. Tokura, Rev. Mod. Phys.
550: {\bf 70}, 1039 (1998).
551: \bibitem{Huang}
552: A. Ohtomo, D.A. Muller, J.L. Grazul and H.Y. Hwang,
553: Nature {\bf 419}, 6905 (2002);
554: A. Ohtomo and H.Y. Hwang, Nature {\bf 427}, 423 (2004);
555: H.Y. Hwang, MRS Bulliten {\bf 31}, 28 (2006).
556: \bibitem{sc}
557: N. Reyren , S. Thiel , A. D. Caviglia , L. Fitting Kourkoutis
558: , G. Hammerl , C. Richter , C. W. Schneider ,
559: T. Kopp, A.-S. Retschi , D. Jaccard , M. Gabay ,
560: D. A. Muller , J.-M. Triscone and J. Mannhart , Science
561: {\bf 317}, 1196 (2007).
562: \bibitem{Brinkman}
563: A. Brinkman , M. Huijben , M. Van Zalk , J. Huijben ,
564: U. Zeitler , J. C. Maan , W. G. Van der Wiel , G. Rijnders
565: , D. H. A. Blank and H. Hilgenkamp , Nature Mat.
566: {\bf 6}, 493 (2007).
567: \bibitem{Ahn}
568: C. H. Ahn, A. Bhattacharya, M. Di Ventra, J. N. Eckstein, C. Daniel Frisbie,
569: M. E. Gershenson, A. M. Goldman, I. H. Inoue, J. Mannhart, Andrew J. Millis,
570: Alberto F. Morpurgo and Douglas Natelson, Rev. Mod. Phys. {\bf 78}, 1186 (2006).
571: \bibitem{Spaldin} Satoshi Okamoto, Andrew J. Millis, and Nicola A. Spaldin, Phys. Rev. Lett. 97. 056802 (2006).
572: \bibitem{Pickett}
573: Rossitza Pentcheva and Warren E. Pickett, Phy. Rev. Lett. {\bf 99}, 016802 (2007).
574: \bibitem{WC}
575: W.-C. Lee, A.H. MacDonald, Phys. Rev. B 74, 075106 (2006).
576: ; Phys. Rev. B {\bf 76}, 075339 (2007).
577: \bibitem{Sawatzky} S. Altieri, L.H. Tjeng, F.C. Voogt, T. Hibma and G.A.
578: Sawatzky, Phys. Rev. B {\bf 59} R2517 (1999).; J. van den Brink
579: and G.A. Sawatzky, Europhys. Lett. {\bf 50} 447 (2000).
580: \bibitem{Koreans} Hyo-Shin Ahn, Do Duc Guong, Jaichan Lee, and Seungwu Han,
581: J. Korean Phy. Soc. {\bf} 49, 1536 (2006).
582: \bibitem{data} M. Cwik, T. Lorenz, J. Baier, R. M$\ddot{u}$ller, G. Andr$\acute{e}$, F. Bour$\acute{e}$e, F. Lichtenberg,
583: A. Freimuth, R. Schmitz, E. M$\ddot{u}$ller-Hartmann and M. Braden, Phys. Rev. B {\bf 68},
584: 060401(R) (2003).
585: \bibitem{VASP1}
586: G. Kresse, and J. Furthmüller, Phys. Rev. B {\bf 54}, 11169 (1996).
587: \bibitem{VASP2}
588: G. Kresse, and J. Furthmüller, Comput. Mater. Sci. {\bf 6}, 15 (1996).
589: \bibitem{VASP3}
590: G. Kresse, and D. Joubert, Phys. Rev. B {\bf 59}, 1758 (1999).
591: \bibitem{PAW}
592: G. Kresse, and J. Hafner, Phys. Rev. B {\bf 47}, 558 (1993).
593: \bibitem{LSDA+U}
594: S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys and A. P. Sutton, Phys. Rev. B {\bf 57}, 1505 (1998).
595: \bibitem{EPL} S. Okatov, A. Poteryaev, and A. Lichtenstein, Europhys. Lett. {\bf 70}, 499 (2005).
596: \bibitem{La}
597: M. T. Czyzyk and G.A. Sawatzky, Phys. Rev. B {\bf 49}, 14211 (1994).
598: \bibitem{Huang2}
599: N. Nakagawa, H. Y. Hwang, and D. A. Muller, Nature mat. {\bf 5}, 204 (2006).
600: \end{thebibliography}
601: \end{document}
602: