High-Density Electron Anions in a Nanoporous Single Crystal: [Ca24Al28O64]4+(4e-)

See allHide authors and affiliations

Science  01 Aug 2003:
Vol. 301, Issue 5633, pp. 626-629
DOI: 10.1126/science.1083842


We removed ∼100% of clathrated oxygen ions from the crystallographic cages in a single crystal of 12CaO·7Al2O3, leading to the formation of high-density (∼2 × 1021 cm3) electrons highly localized in the cages. The resulting electron forms a structure that we interpret as an F+ center and migrates throughout the crystal by hopping to a neighboring cage with conductivity ∼100 siemens per centimeter, demonstrating that the encaged electron behaves as an anion. The electron anions couple antiferromagnetically with each other, forming a diamagnetic pair or singlet bipolaron. The resulting [Ca24Al28O64]4+(4e) may be regarded as a thermally and chemically stable single crystalline “electride.”

Electrides are materials that trap electrons at a stoichiometric concentration in the solid state (1). Chemically, trapped electrons can be viewed as the smallest possible anion, and such materials could serve as strong reducing agents. From a physics standpoint, the stabilization of numerous bound electrons, or F centers, could provide new approaches to preparing conductive materials with unusual optical or magnetic properties. In addition, such materials may find application as low-temperature electron emitters. Most electrides have been either organic species, such as alkali-metal adducts of organic cage compounds (1), or inorganic molecular sieves (24). Generally they are stable only at cryogenic temperatures and are air and water sensitive.

Recently, we have explored the use of the insulator 12CaO·7Al2O3 (C12A7) as a starting material for preparing electrides. This material is an electrical insulator composed of densely packed, subnanometer-sized cages with positive charge. It is thermally stable with a melting point of 1415°C, and a single crystal can be grown directly from a congruent melt (5). The unit cell includes two molecules and 12 cages having a free space of ∼0.4 nm in diameter and can be represented as [Ca24Al28O64]4+ + 2O2. The former denotes the lattice framework, and the latter is called “free oxygen ions” that are loosely bound to the cages to compensate the positive charge of the framework and are coordinated with six Ca2+ ions constituting a part of the cage wall. The cage has “entrances” of ∼0.1 nm in diameter that control mass transport between the inner cages and the outside. The concentration of the cage is 7.0 × 1021 cm3. This feature provides flexibility for replacing the free O2 ions with other anions such as OH (6), F (7), Cl (7), O2 (8), and O (9).

Previously, we found that C12A7 can be converted to a persistent electronic conductor by heat treatment in H2 atmosphere (H2 loading, C12A7:H) and subsequent ultraviolet (UV) irradiation (10). Hydride anions (H) are clathrated in the cage in place of the free O2 ion and ionized by UV irradiation to a pair of H0 and e. The released electron is trapped in the cage and behaves like an F+-center in CaO (11). The F+-like center can hop between cages under an electric field, yielding electrical conductivity up to 0.3 S·cm1.

Further studies with proton implantation showed that conductivities of 10 S·cm1 could be achieved and that conductivities of 0.5 S·cm1 remain even at 600°C (12). Although encouraging, these results suggested that it would be difficult to replace the O2 species stoichiometrically through a hydrogen route. We thus approached the target by a completely different route, i.e., extraction of free O2 ions by a base metal such as Ca via oxide formation. We report the properties of the compound [Ca24Al28O64]4+(4e), synthesized by large (∼7 mm by 4 mm by 0.4 mm) C12A7 single crystals (13). The injected electrons are captured in the cages and behave like an anion with a spherical 1s wave function of an F+-like center. They migrate throughout the crystal as a polaron, which thereby enhances electronic conductivity up to ∼100 S·cm1 at 300 K. The electron anions exhibit spin correlation of antiferromagnetic coupling, forming a diamagnetic pair or singlet bipolaron. These results allow us to regard the [Ca24Al28O64]4+(4e) as an “electride” that is thermally stable, can be exposed to ambient conditions, and can be grown as a large single crystal.

The as-grown C12A7 crystals (sample a) were sealed in silica glass tubes under vacuum together with metal calcium shots, and heated at 700°C for 4 (b), 12 (c), 18 (d), 40 (e), or 240 (f) hours (Ca treatment) (14). After the reactions were complete, x-ray diffraction (XRD) measurements revealed that the sample surfaces were covered with a crystalline CaO layer (15). The surface of the Ca-treated C12A7 was mechanically polished to remove the CaO layers before measurements.

The Ca treatment did not change the basic structure of the lattice framework of C12A7. The sample color changed from colorless to green, and finally to black, with increasing duration of Ca treatment (16, 17). Figure 1A shows the optical absorption spectra of samples a to d. The fundamental absorption edge of C12A7 is reported to be ∼5 eV (9, 12, 18, 19). The apparent absorption edge shifts to ∼4 eV after Ca treatment. The Ca treatment also induces two absorption bands at 2.8 and 0.4 eV (12, 20, 21), the former being responsible for the green coloration. The peak intensities of both bands in samples e and f were too high to measure using our spectrophotometer. Thus, the intensities were estimated by Kubelka-Munk analysis of the diffuse reflectance spectra of powdered samples diluted with KBr powders (Fig. 1B). These results show that the intensities of both bands increase for all of the samples, with an increase in the Ca-treatment duration. The band shape and peak position of the two absorption bands agree exactly with those observed in UV-irradiated C12A7:H samples (10). This indicates that the same center, an F+-like center in which an electron is captured by the positively charged cage, is responsible for the induced optical absorption bands in C12A7:H and in the Ca-treated C12A7.

Fig. 1.

(A) Optical absorption spectra of as-grown C12A7 (sample a) and Ca-treated C12A7 (b, c, and d) single crystals. Photos of samples a to c are shown in the inset. Ca treatment induces absorption bands peaking at 2.8 and 0.4 eV. With an increase in the Ca-treatment duration, the peak intensities increase and sample color turns black via green-yellow and green. Absorbance for samples e and f was too high to measure by the transmission mode. (B) Diffuse reflectance spectra and photos of Ca-treated C12A7 (d, e, and f). The samples were crushed into fine powders and diluted with KBr powder for the measurements. The absorption spectra were obtained by Kubelka-Munk transformation. (C) Log σ (σ, conductivity) versus T1 (T, temperature) plot. (D) Log σ versus T1/4 plot. For samples b and c, log σ is proportional to T1. The activation energies calculated from the slopes are 130 and 160 meV, respectively. For samples d and e, log σ is proportional not to T1 but to T1/4 over a wide temperature range, 300 to 50 K. This characteristic implies that the conduction is controlled by a variable-range hopping-like mechanism. For the highly conductive sample f, log σ is almost constant in this temperature region.

We examined the electrical properties of C12A7 samples using Pt electrodes formed in a four-probe configuration. The conductivity of the as-grown crystal is lower than the detection limit (1010 S·cm1) at room temperature. After the Ca treatment for 4 hours, the conductivity increased to 103 S·cm1. It was enhanced drastically with a further extension of the Ca-treatment duration. After application of the Ca treatment for 1 to 2 weeks, the conductivity reached a saturation level of ∼100 S·cm1. The Seebeck coefficients are negative and decrease with the Ca-treatment duration, indicating that the major carrier is an electron and that its concentration increases with the Ca-treatment duration. Figure 1, C and D, shows the temperature dependence of electrical conductivity (σ) for each sample in the range 300 to 15 K. In the low-conductive samples b and c, log σ is proportional to T 1. The activation energies calculated from the slopes are 130 and 160 meV, respectively. These values are consistent with the simple polaron conduction model (22). In the medium-conductive samples d and e, log σ is proportional not to T 1 but to T 1/4 over a wide range, 300 to 50 K. This characteristic implies that the conduction is controlled by a mechanism similar to the variable-range hopping (23). It is likely that when F+-like centers are randomly distributed, if their concentration is moderately high, they migrate through the distribution of the potential with varied hopping distances. In the high-conductive sample f, log σ is almost constant and is insensitive to temperature.

The above results—the formation of the surface CaO layer, the appearance of the two absorption bands, and the drastic increase in electrical conductivity—suggest that the electrons are injected in the C12A7 lattice by the Ca treatment in place of the free O2 ions. We performed an XRD profile simulation using a code RIETAN-2000 (24) and found that the reduction of the free O2 concentration gives rise to noticeable changes in the XRD pattern. Figure 2A shows XRD patterns of powdered samples before and after the Ca treatment. The Ca treatment induced a reduction of specific peak (e.g., diffraction from 420 planes at ∼32°) intensities by ∼20%. Figure 2B shows the XRD patterns simulated with and without the free O2 ions. In the simulation, crystallographic parameters including space group, lattice constants, fractional coordinates, and temperature factors were from (25). Good agreement between the simulated and observed patterns provides direct evidence for the above speculation that almost all of the free O2 ions are extracted from the crystallographic cages.

Fig. 2.

(A) Observed powder XRD patterns (CuKα) of as-grown (a) and Ca-treated C12A7 (f). The bottom line shows the difference between two patterns. The intensities of the XRD pattern are normalized by the 211 diffraction peak intensity (2θ = 18.1°). (B) Simulated powder XRD patterns of stoichiometric C12A7, [Ca24Al28O64]4+(2O2), and C12A7 lacking the free O2 ions, [Ca24Al28 O64]4+(4e). The difference between the simulated patterns agrees well with that observed in (A), indicating that almost all of the free O2 ion is removed, and the chemical formula can be described as [Ca24Al28 O64]4+(4e). (C) First-derivative EPR spectra of samples d and f (microwave frequency = 9.7 GHz, temperature = 300 K). (D) Temperature dependence of magnetic susceptibility (χm) of sample f measured by SQUID. FC (field cooling) and ZFC (zero-field cooling) curves were measured. The magnetic field was measured at H = 100 Oe. (Inset) M-H hysteresis curve measured at 5 K.

The density and electronic structure of the electrons introduced in the cages were investigated by electron paramagnetic resonance (EPR). The EPR spectrum of the as-grown C12A7 (a) shows a weak O2 signal at a concentration of 1 × 1018 cm3 (8, 9). After the Ca treatment, this signal disappeared, and a new signal with an isotropic Lorentzian shape appeared at g = 1.994 [see (d) in Fig. 2C], which previously was assigned to an F+-like center (10). With an increase in the Ca-treatment duration, the intensity of this signal increased, and the line shape became distorted and finally took on a Dysonian characteristic (26, 27), a feature that is typical of good electronic conductors [see (f) in Fig. 2C]. Spin concentrations of the F+-like centers are determined by using the crushed powders (28). Here, we assumed that unpaired electrons (S = 1/2) are fully localized and isolated. The concentration increased with the Ca-treatment duration to a maximum value of 5 × 1019 cm3 in sample f, showing a saturation tendency against the conductivity, as will be discussed for Fig. 3.

Fig. 3.

Electrical conductivity at 300 K (σ300) versus absorption coefficients at 2.8 eV (α2.8) and 0.4 eV (α0.4), F+-like center concentration (NF+) determined by EPR, and Seebeck coefficient at 300 K (S300). The F+-like center concentration in sample f measured by SQUID is plotted by a filled blue circle. NF+, α2.8, and α0.4 increase linearly with σ300 in the range σ300 < 0.2S·cm1 (b, c, and d). However, the NF+ value shows saturation tendency at ∼1 × 1019 cm3 in the high σ300 range (σ300 > 0.2S·cm1). The α2.8 and α0.4 values increase linearly even in the high σ300 region. Seebeck coefficients are proportional to conductivity, indicating that the carrier mobility is almost independent of carrier concentration.

Figure 2D shows the temperature dependence of molar magnetic susceptibility (χm, for a Ca12Al14O33 molecule) of sample f measured by a superconducting quantum interference device (SQUID) in the range 5 to 300 K at an external magnetic field (H) of 100 Oe. We measured χm under zero field cooling (ZFC) and field cooling (FC). The ZFC process was done by cooling the sample to 5 K at H = 0 Oe, whereas the FC process was done at H = 100 Oe. A clear difference between the FC and the ZFC curves is observed at temperatures below 15 K, suggesting spin freezing. The FC curve obeys the Curie-Weiss equation, χm = C/(T – θ), yielding a Curie constant (C) of 7.9 × 103 emu·K·mol1 and a Weiss temperature (θ) of –7.9 K. From the C value, the effective moment is calculated to be 0.25 μB·mol1B is the Bohr magneton) and the paramagnetic electron concentration to be 3 × 1019 cm3 (provided that the electron has S = 1/2 and is fully localized); this concentration is in reasonable agreement with that determined from the EPR measurement (5 × 1019 cm3). Further, the negative θ value indicates that the magnetic interaction between the spins is antiferromagnetic. A magnetic hysteresis with a residual magnetization of 0.1 emu·mol1 was also observed for the M-H curve below 15 K (Fig. 2D, inset), suggesting that uncoupled electrons form a spin glass state at low temperatures.

Figure 3 summarizes the correlations between σ at 300 K (σ300) and absorption coefficients of the 2.8-eV band (α2.8) and the 0.4-eV band (α0.4). The concentrations of the F+-like center (NF+) determined from the spin concentration obtained by EPR and from the Curie constant are also plotted, together with the Seebeck coefficient. The carrier electron concentration (ne) can be estimated from the α2.8 and α0.4 values if each F+-like center acts as a free carrier; alternatively, ne can be estimated from the Seebeck coefficient, which should have a linear dependence on log ne as long as the carrier mobility is not a function of ne. Both the α values and Seebeck coefficient increase linearly with σ300. This result supports the above supposition, i.e., each F+-like center formed acts as a carrier, and its mobility is almost independent of the ne value at ∼0.1 cm2 (V·s)1. The maximum ne value of 2 × 1021 cm3 was obtained in sample f based on the absorption cross section for the 2.8-eV absorption band. This ne value agrees well with the theoretical maximum value of 2.33 × 1021 cm3—assuming that all the free oxygen anions (O2) are replaced with electrons (e), thereby maintaining the relation O2 → 2e—and indicates that almost all of the free O2 ions in the cages are replaced with electrons by the Ca treatment. However, NF+ increases linearly with the σ300 value when σ300 is <0.2 S·cm1, and the NF+ value agrees with the ne value determined from the α2.8 value in this σ300 range. With a further increase in σ300, the NF+ value shows a saturation tendency at ∼3 × 1019 cm3 in sample f, which is only ∼2% of the carrier concentration ne. This saturation value strongly suggests that the antiparallel spin coupling occurs dominantly, forming diamagnetic electron pairs or a bipolaron (S = 0) above a critical concentration of the F+-like centers. The above results on magnetic susceptibility substantiate this figure.

If we view [Ca24Al28O64]4+ as a large cation, and the four localized electrons that accompany the large lattice distortion as anions, then [Ca24Al28O64]4+(4e) can be viewed as an electride that is thermally and chemically stable in ambient atmosphere and can be grown readily as a single crystal. Figure 4 illustrates schematically the crystal structure of [Ca24Al28O64]4+(4e) electride. The framework is composed of densely packed cages. Because each cage has 12 neighboring cages, the concentration fraction of anionic electron is 1/3 per cage. A simple statistical analysis that assumes random distribution of electrons in a 12-coordinating cage structure shows that ∼99% of the electrons have other electrons in a neighboring cage at the maximum electron concentration. This model is consistent with the observation that the fraction of the paramagnetic spins, observed by EPR or SQUID, is only a few percent of the total electrons in sample f, if the paired electrons have a spin interaction and form a diamagnetic pair or singlet bipolaron.

Fig. 4.

(A) Crystal structure of [Ca24Al28O64]4+ (4e). Yellow region corresponds to a unit cell. (B) The framework of stoichiometric C12A7 is composed of 12 cages. Two out of 12 cages are occupied by O2 ions. (C) In the [Ca24Al28O64]4+(4e ) compound, 4 out of 12 cages are occupied by electrons in place of O2 ions. (D) Each cage is surrounded by 12 neighboring cages. When electrons are randomly distributed in the cages, 99% of the electrons have a neighboring electron.

Further, the energy levels of the clathrated electrons are ∼1 eV below the conduction-band edge with a small work function of ∼3 eV. These features—a high-density, loosely bound electron anion with a small work function— hold promise for applications that require an efficient cold electron field-emitter competing with carbon nanotubes and diamond.

References and Notes

Stay Connected to Science

Navigate This Article