Report

Bistability in Atomic-Scale Antiferromagnets

See allHide authors and affiliations

Science  13 Jan 2012:
Vol. 335, Issue 6065, pp. 196-199
DOI: 10.1126/science.1214131

Abstract

Control of magnetism on the atomic scale is becoming essential as data storage devices are miniaturized. We show that antiferromagnetic nanostructures, composed of just a few Fe atoms on a surface, exhibit two magnetic states, the Néel states, that are stable for hours at low temperature. For the smallest structures, we observed transitions between Néel states due to quantum tunneling of magnetization. We sensed the magnetic states of the designed structures using spin-polarized tunneling and switched between them electrically with nanosecond speed. Tailoring the properties of neighboring antiferromagnetic nanostructures enables a low-temperature demonstration of dense nonvolatile storage of information.

Nanometer-scale ferromagnets are used as magnetic bits to hold information in mass storage devices. Antiferromagnets have been difficult to switch and sense because of their lack of net magnetic moment, but they offer advantages such as insensitivity to magnetic fields.

In ferromagnetic materials, the magnetic moments of the constituent atoms align, yielding a net magnetic moment. The direction of this magnetization can be changed by the application of a magnetic field or by spin-polarized currents (1). As magnetic devices shrink toward atomic dimensions, new tools to fabricate and probe them with atomic resolution are emerging (24). These have revealed magnetic bistability in ferromagnetic islands (5, 6) and chains (7), having as few as 30 atoms, as well as in metal-organic molecules (810).

Antiferromagnets have neighboring atoms with counteraligned magnetic moments. The absence of a net magnetic moment makes imaging the magnetic structure of antiferromagnets more difficult. Antiferromagnetic (AFM) domains in thin films have been imaged using x-ray scattering (11). On the atomic scale, the spin structure of antiferromagnets has been observed by scanning tunneling microscopy (12, 13) and atomic force microscopy (14). So far, controlled switching of antiferromagnets has required the help of nearby ferromagnetic domains (15), magnetoelectricity (16), or optical pulses (17). We investigated the role that AFM nanostructures can play as candidates for magnetic storage and spintronic devices.

We assembled AFM nanostructures with a low-temperature scanning tunneling microscope (STM) by placing Fe atoms in a regular pattern on a surface (Fig. 1A). The spins of neighboring Fe atoms couple antiferromagnetically by an exchange interaction with strength J = 1.2 meV (18) (fig. S1). The Fe atoms were placed at a binding site on a Cu2N surface, for which Fe has a large magnetic anisotropy field that aligns its spin to the resulting easy axis (19). A magnetic field of up to 6 T was applied in order to make the microscope’s tip spin-sensitive by polarizing its magnetic apex (18) and in order to test the effect of magnetic field on the nanostructures.

Fig. 1

Bistable AFM array of Fe atoms. (A) Schematic of atoms on a surface coupled antiferromagnetically with exchange energy J. Surface-induced magnetic anisotropy fields cause the spins of the atoms to align parallel to the easy magnetic axis, D. A spin-polarized STM tip reads the magnetic state of the structure by magnetoresistive tunneling. A magnetic field applied parallel to D polarizes the tip. (B) Spin-polarized STM image of a linear chain of eight Fe atoms assembled on a Cu2N overlayer on Cu(100). This is a constant-current image using 2 mV and 1 pA. Spins are in Néel state 0. (C) Section through center of chain in (B) with the spin orientation of each Fe atom indicated by colored arrows. (D and E) Same as (B) and (C) but in Néel state 1.

In assemblies of just a few magnetic atoms, the atomic spins often couple to form quantum superposition states (20). For AFM coupling, this results in a singlet ground state, characterized by a wave function in which all spins populate opposing spin states equally (21). In contrast, we find that isolated AFM structures with as few as six Fe atoms exhibit stable Néel states, in which the spin orientation alternates between neighboring atoms. These states can be well described by the classical Ising model (22), in which the spins always point along one axis. Spin-polarized STM images of a linear eight-atom chain (Fig. 1, B to E) can clearly distinguish the two Néel states. The spin-polarized STM tip forms a magnetic tunnel junction in which the conductance alternates between high (parallel alignment of tip and sample spins) and low (antiparallel alignment) as the tip passes from atom to atom along the chain (12, 13, 23). Identical chains built from Mn atoms do not show the magnetic bistability and do not exhibit a spin-polarized contrast along the chain (fig. S2). A key difference between Fe and Mn chains is the strength of the magnetic anisotropy, which is ~50 times stronger in Fe than in Mn on this surface (19). The strong easy-axis anisotropy of Fe evidently stabilizes the two Néel states as observable magnetic states.

The stability of the magnetic states was not affected by imaging them using an applied voltage of <2 mV, but voltages in excess of ~7 mV caused switching. To intentionally switch the magnetic state of the entire antiferromagnet, the tip was held stationary over any Fe atom of the structure, and tunnel current was passed through it at >7 mV until a step, indicating a change in magnetic state, was observed in the current (Fig. 2A). Subsequently, the voltage was lowered to prevent further switching. Near the 7-mV switching threshold, the Néel state in which the spin of the atom under the tip is aligned with the magnetic field was occupied ~90% of the time (Fig. 2A). This directionality offers a path toward controlled directional switching. An alternative process to switch AFM structures, the use of spin transfer torque (1, 5), has previously been proposed (24, 25).

Fig. 2

Switching between Néel states induced by tunneling electrons. (A) Tunnel current as a function of time at 7 mV. The tip is placed over an end atom of a (1×8) Fe chain. The chain switches its magnetic state about twice per second. (B) Same as (A) but with pulsed voltage V to demonstrate fast switching. Pulses of 0.5 V and 10-ns duration were applied every 10 ms (at dashed vertical lines). Between pulses, only 2 mV was applied (below the threshold for switching) to sense the magnetic state. (C) Switching rate versus sample voltage V. Voltage was applied continuously for |V| < 10 mV and as 5- to 1000-ns pulses for |V| > 10 mV. Green circles indicate transitions from low to high current; blue squares indicate high to low. Open symbols were recorded at negative sample voltage. Magnetic field 1 T and temperature 0.5 K for all panels. Tip-sample distance was set at 20 pA and 2 mV for (B) and (C).

We found that the state switched most readily when the tip was placed over an end atom of a chain. The switching between magnetic states was found to occur stochastically, with a uniform probability per unit of time, which we characterized by means of a switching rate (18) (fig. S3). This rate increased rapidly when the tunneling current was increased.

With increasing voltage, the switching rate exceeded the bandwidth of the STM’s current amplifier, so a pulsed-voltage scheme was used to determine the fast switching rates (Fig. 2B). Submicrosecond pulses were applied to the junction (26), and each pulse was followed by a low-voltage window in which the resulting magnetic state was detected (18). The switching rate increased faster than in proportion to the voltage up the highest voltage tested, with switching times of ~20 ns at 0.5 V (Fig. 2C). This demonstrates electrical switching of the AFM nanostructures at high speeds and femtojoule energies.

To investigate the stability of the Néel states, we examined the thermal switching rates of linear chains of Fe atoms with varying length, (1×n), and arrays of two coupled chains, (2×n) (Fig. 3). All structures containing eight or more atoms were found to be stable at the lowest temperature, 0.5 K. Spontaneous flipping between the two Néel states sets in with increasing temperature. Structures with more atoms remain stable to higher temperatures (Fig. 3, A to C) (6, 27).

Fig. 3

Thermal stability of AFM arrays. (A to C) STM images of (2×6) and (2×4) arrays of Fe atoms. (A) 1.2 K. Both arrays have stable Néel states. (B) 3.0 K. The smaller array switched rapidly during the image. (C) 5.0 K. Both arrays switched rapidly. Image size, 7.7 × 7.7 nm. Image was taken at 2 mV and 3 pA, and image acquisition time was 52 s. (D) Schematic of the atomic positions of Fe and Cu2N substrate atoms in (2×n) and (1×n) arrays. Cu atoms, yellow; N atoms, light blue. Ball colors depict the spin alignment of one Néel state, with red being parallel and blue antiparallel with the tip’s spin. (E) Arrhenius plot of the switching rates for the arrays of (A) and a (1×8) and (1×6) chain (fig. S5). The determination of switching rates is explained in fig. S3. Magnetic field was 3 T. Fig. S4 shows comparison to a 1-T field. Fit parameters are given in table S1.

Above ~5 K, the switching rates of the (1×6), (1×8), and (2×6) arrays follow the Arrhenius law with comparable spin reversal barriers, EB ~ 7 to 12 meV, and exponential prefactors, ρ0 ~ 108 s−1 (Fig. 3E and table S1). This prefactor falls in the typical range, 107 to 1014, found for ferromagnetic nanoparticles (5, 28) and magnetic molecules (29). The values for EB are comparable to the threshold for voltage-induced switching (Fig. 2C) and to the energy 2 S2J = 9.6 meV (S = 2 for Fe) needed to create a single Ising domain wall within one of the chains by flipping one or more consecutive spins at the end of a chain (22). This indicates that current- and temperature-induced switching between the two Néel states is accomplished by propagating domain walls along each chain.

Below ~5 K, the switching rates of the (1×6) and (1×8) chains become independent of temperature. Such behavior is consistent with quantum tunneling of magnetization (30), which is typically observed in few-atom molecular magnets (8, 9) and also occurs in magnetic nanoparticles (28). Here, it causes the AFM nanostructures to evolve between the two Néel states, thus limiting their stability even though thermal switching is frozen out. Comparison of the structures of Fig. 3E highlights two avenues to reduce quantum tunneling: First, through increasing the chain length. The addition of two atoms, from (1×6) to (1×8), reduced the tunneling rate 1000-fold. Second, through the coupling of two chains, from (1×6) to (2×6) as shown in Fig. 3D. Even though the spin coupling between chains of J' = 0.03 meV per atom (18) is much weaker than the exchange coupling within a chain, J, it suppresses tunneling markedly. The large difference in the strength of J and J' is linked to the Cu2N surface’s crystal structure and is evidence of a superexchange-mediated interaction in the Cu2N molecular network (1820) (fig. S1).

A different manifestation of quantum tunneling of magnetization can be found in the (2×4) array, which has a much reduced exponential prefactor and energy barrier, 1.5 meV. This energy is comparable to 4 × 2 S2J' = 1.1 meV, the energy required to frustrate the weak coupling between the two short chains. This low barrier and the much-reduced exponential prefactor of only ρ0 = 5 × 103 s−1, indicate a reversal process in which one entire chain switches in a thermally assisted tunneling process (29).

The thermal switching rates were found to be independent of magnetic field (fig. S4), and both Néel states were occupied for equal amounts of time (fig. S3), showing that these AFM arrays are fully spin-compensated. The AFM nanostructures are magnetically stable even in the absence of an external magnetic field (fig. S6).

The (2×6) array is highly stable at low temperatures, where switching was observed so rarely that no tunneling rate could be derived. We experimentally determined a lower limit for the stability of these arrays of less than one switching event per 17 hours at 0.5 K.

A challenge to miniaturizing the bits in ferromagnetic storage media is the interaction of neighboring bits because of their dipolar magnetic fields (31). This would not be present in AFM storage media. At atomic dimensions, however, exchange interactions can still cause undesired coupling between neighboring bits (27). Figure 4A shows an AFM byte, a dense packing of eight (2×6) Fe arrays, with each array representing one bit of information. The structure was engineered to have reduced bit-to-bit exchange interactions. Neighboring bits were staggered in a way that places the atoms of any given bit symmetrically between the atoms of the neighboring bits, resulting in a near-perfect cancellation of bit-to-bit exchange couplings through geometric frustration (Fig. 4B) (23).

Fig. 4

Ultradense AFM data storage. (A) Non–spin-polarized STM image, 24 × 8 nm, of eight (2×6) arrays assembled from Fe atoms. (B) Schematic of the bits in (A), with colors as in Fig. 3D. Jb and Jb': pairwise canceling exchange couplings between atoms in neighboring bits. (C) Information storage in a magnetic byte. A color-coded difference between spin-polarized and spin-averaged images is shown, with red corresponding to higher tip height and blue to lower tip height in the spin-polarized image. (Top) All eight bits in logic 1 state (as defined by the spin orientation of the top two Fe atoms in each bit). (Middle) Alternating pattern of 1 and 0. (Bottom) All 0. More bit patterns are shown in fig. S7.

Each of the eight bits shown in Fig. 4A can be switched without perturbing the state of the other bits. Figure 4C and fig. S7 show short sequences of test arrangements written into the byte. These configurations are stable over a time scale of hours, and readout was achieved by topographic imaging. Each bit occupies an area of only 9 nm2.

The arrangement of Fe atoms that form each bit in the byte is a variant on the (2×6) array (compare Fig. 3D and Fig. 4B), in which the ends of each bit are beveled to give the endmost atoms of each bit the same spin orientation. This provides clarity in viewing the state and shows that the exact arrangement of atoms is not critical for magnetic stability.

Our results demonstrate that switchable nanoscale antiferromagnets are candidates for future memory, storage, and spintronic applications.

Supporting Online Material

www.sciencemag.org/cgi/content/full/335/6065/196/DC1

Materials and Methods

Figs. S1 to S7

Table S1

References (3238)

References and Notes

  1. Supporting material is available on Science Online.
  2. Acknowledgments: We acknowledge B. Melior for expert technical assistance. S.L., C.P.L., and A.J.H. thank the Office of Naval Research for financial support. A patent application regarding information storage in antiferromagnetic nanostructures was filed with the U.S. Patent and Trademark Office.
View Abstract

Navigate This Article