Article pubs.acs.org/IC
Cobalt-Based Single-Ion Magnets on an Apatite Lattice: Toward Patterned Arrays for Magnetic Memories Pavel E. Kazin,*,† Mikhail A. Zykin,‡ Walter Schnelle,§ Yan V. Zubavichus,∥ Konstantin A. Babeshkin,† Victor A. Tafeenko,† Claudia Felser,§ and Martin Jansen§,⊥ †
Department of Chemistry, Moscow State University, 119991 Moscow, Russia Department of Materials Science, Moscow State University, 119991 Moscow, Russia § Max Planck Institute for Chemical Physics of Solids, Nöthnitzer Strasse 40, 01187 Dresden, Germany ∥ National Research Centre “Kurchatov Institute”, 123182 Moscow, Russia ⊥ Max Planck Institute for Solid State Research, Heisenbergstrasse 1, 70569 Stuttgart, Germany ‡
S Supporting Information *
ABSTRACT: Single-ion magnets (SIMs) that can maintain magnetization direction on an individual transition metal atom represent the smallest atomic-scale units for future magnetic data storage devices and molecular electronics. Here we present a robust extended inorganic solid hosting efficient SIM centers, as an alternative to molecular SIM crystals. We show that unique dioxocobaltate(II) ions, confined in the channels of strontium hydroxyapatite, exhibit classical SIM features with a large energy barrier for magnetization reversal (Ueff) of 51− 59 cm−1. The samples have been tuned such that a magnetization hysteresis opens below 8 K and Ueff increases by a factor of 4 and can be further enhanced to the highest values among 3d metal complexes of 275 cm−1 when Ba is substituted for Sr. The SIM properties are preserved without any tendency toward spin ordering up to a high Co concentration. At a maximal Co content, a hypothetical regular hexagonal grid of SIMs with a 1 nm interspacing on the (001) crystal facet would allow a maximal magnetic recording density of 105 Gb/cm2.
■
INTRODUCTION Extensively developed over the past several decades, singlemolecule magnets (SMMs) have proven to be very attractive materials for future ultra-high-density magnetic memory, molecular electronics, and spintronics devices.1 Many polyand mononuclear d and f metal complexes have been shown to exhibit SMM properties, the main one being slow relaxation of magnetization characterized by an energy barrier for the magnetization reversal (Ueff) and by a magnetization blocking (irreversibility) temperature (Tb).2−5 With regard to mononuclear SMMs, also called single-ion magnets (SIMs), the highest Ueff values of 712 and 670 cm−1 have been achieved for DyIII and TbIII complexes, respectively.6,7 Among 3d metal complexes, iron and cobalt compounds reveal superior properties8−11 with Ueff values as high as 246 cm−1 for a 2fold-coordinated methanide complex of iron(I).8b To realize even higher Ueff values, important issues are the considerable angular momentum contributions to the electronic ground state and the pronounced axial component of the ligand field, a linear coordination being considered the best.12 Recently, we have reported on the SIM behavior of Cu-doped alkaline-earth apatite phosphates, in which linear paramagnetic (S = 1) [OCuO]− ions situated in the trigonal channels show a Ueff of ≲100 cm−1, depending on the alkaline-earth cation nature.13,14 © XXXX American Chemical Society
The compounds are chemically inert toward air and water and stable to high temperatures. This suggests principally new opportunities to create SIMs in such a stable inorganic matrix and develop their regular patterns, so that the material (either as a crystal, as a thin film, or as a several-atom-thick layer) may serve as a highly integrated part of a molecular electronic device. Previously, we have shown that Co ions also enter into the trigonal channels of strontium apatite.15 Here we report on the successful formation of arrays of SIMs in Co-doped apatites with Ueff values exceeding, to the best of our knowledge, all those of 3d metal complex SMMs.
■
RESULTS AND DISCUSSION Composition and Crystal Structure. The single-phase polycrystalline samples with apatite-type structure (Figure S1) correspond to the general composition M10(PO4)6(CoxO1−yH1−2x−2y−δ)2. The value of δ means that the oxidation state of cobalt may be above +2 and/or peroxide ions may present in the compound. Compounds for which M = Sr and x = 0.02, 0.05, 0.2, and 0.3 are denoted 1−4, respectively, and that for which M = Ba and x = 0.02 is denoted Received: September 27, 2016
A
DOI: 10.1021/acs.inorgchem.6b02348 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
Arrhenius plots for the calculated relaxation time (τ) are collected in Figure 2.
5. 1 and 2, additionally annealed in oxygen (oxygenated), are denoted 1a and 2a, respectively. The sample denoted as 5a has the same chemical composition as 5 and was prepared by the same procedure (see Experimental Section), with the exception that the annealing was performed at 1300 °C instead of 1400 °C. A single crystal (M = Sr, and x = 0.18) is denoted 6. Attempts to prepare strontium compounds for which x > 0.3 resulted in the formation of CoO along with the apatite phase for which x ∼ 0.3. In the case of barium analogues, the increase of x led to multiphase samples. In the well-known apatite structure of 6, Sr(1) forms linear chains running along the z-axis and Sr(2) constitutes “walls” (marked by lines) of trigonal channels built by the infinite chain of face-shared Sr(2)6 octahedra running along the z-axis. PO4 groups are situated between the Sr(1) and Sr(2) atoms. Consistent with our previous structure estimation from powder diffraction data,15 a Co atom is found in the trigonal channel in a 6-fold-split position. It shifts from the ideal (0,0,0) by 0.45 Å in the direction of O(3) of the PO4 group and has two close contacts with the adjacent intrachannel O(4) atoms, building a [OCoO]2− ion with bent geometry (Figure 1 and Tables S1−
Figure 2. Arrhenius plots for magnetization relaxation time τ: (bottom plots) NSR magnetization under a field of 1.5 kOe and (top plots) VSR magnetization under a field of 1.5 kOe (red and purple symbols) or 0 kOe (green and blue symbols). Black lines are linear fits. Dashed lines are fits to eq 1.
In general, we distinguish three regimes of the magnetization relaxation. Below 7 K, slow magnetization relaxation is observed in the ac susceptibility measurements. This relaxation is induced by a weak magnetic field, covers the major part of the susceptibility, and is characterized by a narrow distribution of τ values. The latter is an inherent property of SMMs. As this kind of relaxation is common for all the samples and correlates with the Co content, we call this regime “normal slow relaxation” (NSR) and attribute it to the [OCoO]2− ion. The part of the magnetization that remains reversible at the highest frequencies applied and corresponds to a value of χs in the generalized Debye model (see the Supporting Information) still comprises 20−30% of the susceptibility. It may be connected to Co ions in a different coordination environment and to other paramagnetic defects like O2−.16 The behavior of this part of the magnetization we consider as the regime of “fast relaxation” (FR). The ln τ values of NSR show a typical behavior, decreasing first slowly with the decrease of inverse temperature (1/T) at low temperatures and then faster at higher temperatures. At a sufficiently low temperature, τ increases from 4 to 1; i.e., at lower Co contents, as one would expect because the paramagnet’s dilution suppresses dipolar interactions, thus diminishing the probability of quantum tunneling (QT) processes that limit τ at low temperatures.1 With an increase in T, the Raman process may contribute to the production of a nonlinear ln τ(1/T). At sufficiently high temperatures, the thermally activated Orbach process [τ = τ0 exp(Ueff/kT)] can prevail, resulting in an approach to the linear ln τ(1/T) dependence. The latter is observed for 1 and 2; their plots almost merge at higher temperatures, thus confirming that the effects of QT and Raman processes are small. The linear fit for 2 (last four points, 5.5−7 K) yields a Ueff of 51.4(5) cm−1 and a
Figure 1. Fragment of the crystal structure of Sr10(PO4)6(Co0.18O0.92H0.48−‑δ)2. The nonbonding separation [Co− O(3)] is indicated by a dashed line.
S3). O(4) represents an averaged position of oxygen atoms belonging to the dioxocobaltate(II), a hydroxide, and a peroxide that is expected to form under preparation conditions.16 The Co K-edge EXAFS analysis of 4 (Figure S2) reveals two O atoms located 1.73 Å from the Co atom. The O−Co−O angle is then calculated to be 150(2)°. The Co atom shift results in a decrease in the Co−O(3) distance to 2.65 Å, which seems still to be too long for the bonding. The distance between the Co atoms cannot be ≲7 Å along the c-direction and ≤9 Å in the a−b plane. This is comparable to the distance between 3d ions in the crystal lattice of metal complex SIMs. The [OCoO]2− ions are embedded in the diamagnetic and dielectric matrix of an apatite-type strontium phosphate. Hence, possible interionic interactions are expected to be very weak and comparable to those in classical SIMs, which allows us to consider our compounds as the array of the magnetically independent [OCoO]2− molecular ions in the solid matrix like the crystalline metal complex SIMs. Relaxation of Magnetization. The frequency ( f) dependence of the ac susceptibility (χac) and the dc magnetization time dependence are shown in Figures S3 and S4, respectively. The B
DOI: 10.1021/acs.inorgchem.6b02348 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry τ0 of 3.9(4) × 10−10 s. The estimated Ueff value may still represent a lower boundary of a real energy barrier height. A more comprehensive analysis involves considering the ln τ(1/ T) dependence over the whole temperature range, taking into account all relaxation processes4b,8b in accord with eq 1 τ −1 = τQTM −1 + BT + CT n + τ0−1 exp( −Ueff /kBT )
and 2a under a non-zero field cannot be fitted at all: at intermediate temperatures, the experimental curve strongly deviates downward from the calculated one. This might be related to additional factors, e.g., the presence of an excited electronic state with an intermediate energy. Equation 1 includes several free parameters, and when a smooth ln τ(1/ T) dependence is being fit, their correlation is large, which produces considerable uncertainty. Therefore, we cannot decide whether the linear fit or a description based on eq 1 gives results closer to the real barrier height. In spite of this slight ambiguity, we believe with a high degree of confidence that the VSR energy barrier lies in the range of 200−250 cm−1. Thus, the Ueff values obtained exceed the highest energy barrier known for Co-based SMMs, which is currently 180 cm−1.11b Ba apatite 5 does not show NSR but along with FR reveals an ∼15% fraction of VSR both under zero and non-zero fields. The dc magnetization decay recorded at 1.8 K and 1.5 kOe corresponds to a relaxation time of 760 s (Figure S5a). In contrast to the case for Sr apatites, upon annealing in oxygen, the VSR fraction disappears. In comparison with that of the Sr apatite samples, the ln τ(1/T) linear fit (H = 1.5 kOe) gives a higher Ueff of 269(16) cm−1 with a τ0 of 3(2) × 10−9 s (Figure S5b). Another Ba apatite sample, 5a, was prepared at a lower annealing temperature assuming that it would be differently oxygenated and not overoxidized, which is thought to take place upon the annealing in oxygen mentioned above. Indeed, the VSR fraction in 5a was increased to 30%. The whole ln τ(1/ T) curve (see Figure 2) cannot be fitted using eq 1 for the same reasons cited for the Sr apatite samples. The linear fit of the high-temperature part of the Arrhenius plot (H = 1.5 kOe) results in a close Ueff of 275(11) cm−1 with a τ0 of 1.9(8) × 10−9 s. To the best of our knowledge, the estimated Ueff values exceed all the values known among 3d metal complexes. However, there is an extended solid material that possibly features a superior parameter, iron(I)-substituted lithium nitride Li3−xFexN: the authors reported for the well-diluted paramagnet properties resembling those of SIMs with a Ueff of ∼300 cm−1.17 The VSR magnetization comprises a solid fraction of the magnetization over a wide temperature range and under different magnetic fields. Therefore, we consider it as the contribution to the magnetization from a Co-based SIM in an apatite lattice, different from that of the observed intrachannel [OCoO]2− ion. dc Magnetization. 4 exhibits mostly NSR magnetization that apparently originates from the [OCoO]2− ions. This sample reveals paramagnetic behavior down to 1.8 K (Figure 3a and Figure S6). The estimated μeff is close to the spin-only value. This justifies the fit to χT(T) using an appropriate analytical expression for an S of 3/2 with the axial zero-field splitting parameter, D.18 The fit yielded a 2D of −52(1) cm−1, a g∥ of 2.222(1), a g⊥ of 2.028(1), and a χTIP of 0.001490(2) cm3/ mol of Co. The negative value of D points to easy-axis anisotropy with ground state MS = ±3/2 doublet. The absolute 2D value is close to the Ueff obtained for the NSR process, in excellent agreement with the theory implying that the magnetic moment reverses through the first excited state (MS = ±1/2). In addition, series of M(T,H) values were fitted using anisofit version 2.0.19 The results comply with the χT(T) analysis described above: 2D = −50.4 cm−1, E = 5.1 cm−1, and giso = 2.178. A large value of E implies strong rhombic distortion and is consistent with the bent geometry of the [OCoO]2− ion.
(1)
where τQTM is the relaxation time of the quantum tunneling process, B = AHm (where H is the magnetic field strength) is the coefficient of the direct process, and C is the coefficient of the Raman process. For the Raman process, an n of 5 was successfully used to fit an iron(I) complex with a d 7 configuration of the central ion.8b The fitting of NSR in 1 and 2 for an n of 5 provides the following results: for 1, Ueff = 59(2) cm−1, τ0 = 7(3) × 10−11 s, B = 16(1) s−1 K−1, and C = 0.07(1) s−1 K−5; for 2, Ueff = 58(2) cm−1, τ0 = 10(5) × 10−11 s, B = 28(2) s−1 K−1, and C = 0.21(3) s−1 K−5. τQTM was large and did not affect the fit and therefore was fixed to 100 s. We consider the practically identical Ueff values obtained using eq 1 as good estimates of the real energy barrier height. Ueff values for 3 and 4 cannot be extracted using eq 1 as the contribution of Raman and direct processes is too large. It is noteworthy that the dc susceptibility (χdc) and χac (0.1 Hz) are close for 4 only. Other samples reveal the prevalence of χdc, growing with a decreasing Co concentration. This implies that along with NSR a very slow relaxation process exists, which affects another part of the magnetization. This relaxation regime is denoted as “very slow relaxation” (VSR). In the ac susceptibility plots, this feature becomes noticeable at 10 K and above, both for 1.5 kOe and zero fields (see, for example, Figure S3j). At 10 K, the VSR susceptibilities comprise 22, 17, 11, and
