Exchange bias — the loop shift of a ferromagnet (F) in contact with an antiferromagnet (AF) — degrades with repeated field cycling, a phenomenon called the training effect whose microscopic origin is still debated. Dobrynin et al. (2014) proposed, on the basis of X-ray magnetic linear dichroism (XMLD) measurements on Co/CoO bilayers, that training arises from the formation of an antiferromagnetic exchange spring — a continuously twisted AF spin structure — at the frustrated F/AF interface. Their XMLD evidence, however, only averages the AF orientation over the interface. The open problem is to establish microscopically and quantitatively whether such an AF spring actually forms during training: to determine its layer-resolved rotation profile and spatial extent, and the exchange and anisotropy energetics that stabilize it and account for the distinct field-cooled and trained states, the loop-by-loop relaxation of the exchange bias, and its recovery after zero-field warm-up to the blocking temperature — ideally through direct element- and depth-resolved measurement or an equivalently discriminating quantitative test. No published study has resolved the mechanism, which remains contested, with competing explanations still being published through 2019–2021.

Background

Exchange bias (EB) is the shift of a ferromagnet's hysteresis loop along the field axis that occurs when the ferromagnet (F) is grown in contact with an antiferromagnet (AF) and the system is cooled in a magnetic field below the AF's ordering temperature. EB is used as the reference-layer pinning mechanism in spin-valve readers (the giant-magnetoresistance read heads of hard-disk drives) and in magnetic random-access memory. On repeated cycling of the field through consecutive hysteresis loops, the EB field and the coercivity relax toward steady values — the training effect — signaling that the AF is not in equilibrium after the initial field cooling. The blocking temperature TBT_B is the temperature above which the AF no longer pins the F (typically well below the AF Néel temperature — the bulk antiferromagnetic ordering temperature — for thin AF layers); Dobrynin et al. showed that warming the trained Co/CoO bilayer up to TBT_B and cooling it back in zero field fully restores the untrained, high-bias state.

X-ray magnetic linear dichroism (XMLD) is the dependence of the x-ray absorption intensity on the relative orientation of the linearly polarized x-rays and the AF spin axis; measured at the Co L-edge of CoO it probes the orientation of the AF order, but in the conventional grazing-incidence geometry used by Dobrynin et al. it yields an average over the probed AF region rather than a depth-resolved profile. An exchange spring is a smoothly rotating spin structure: in ordinary F/AF-coupled systems, the interfacial spins are pulled by the F while the AF anisotropy anchors the bulk, producing a twist analogous to the ferromagnetic exchange springs of hard/soft magnets. Antiferromagnetic exchange springs and related spiraling AF structures had been demonstrated before 2014: Yang and Chien (2000) inferred a spiraling spin structure in an exchange-coupled FeMn layer in permalloy/FeMn/Co trilayers, Scholl et al. (2004) imaged a planar AF domain wall (the wall-like limit of an AF spring) with XMLD-based photoemission electron microscopy, Wang et al. (2014) used thickness-dependent XMLD and X-ray magnetic circular dichroism (XMCD) measurements together with tunneling anisotropic magnetoresistance (a tunneling resistance that depends on the magnetization direction) to reveal an asymmetric AF spin rotation in a perpendicular CoO-based spring, and theoretical work had shown that magnetization reversal in exchange-coupled F/AF bilayers proceeds through irreversible exchange-spring processes (Guo et al., 2008). The specific proposal of the source paper is that in Co/CoO the formation and progressive modification of such an AF spring at the frustrated F/AF interface is the cause of the training effect and of its recovery on re-cooling from TBT_B.

The training-effect mechanism, however, remains unsettled. Competing or complementary pictures include: analytic models of a disordered, metastable AF interface spin population (Binek, 2004); rearrangement of polycrystalline AF grains (Brems, Temst, and Van Haesendonck, 2007); the role of interfacial AF domain walls evidenced by polarized neutron reflectometry — a depth-sensitive technique that measures the spin-dependent reflection of neutrons from a layered structure (Hauet et al., 2006); the domain-state model, in which training follows from irreversible changes of the AF domain configuration in a diluted or granular AF, treated by Monte Carlo and micromagnetic simulation (Biternas, Chantrell, and Nowak, 2014); and, for granular Co-CoO made by oxygen implantation, the perpendicular anisotropy of rotatable uncompensated interface spins — interfacial AF moments not compensated by the opposite sublattice that can rotate with the ferromagnet (Menéndez et al., 2019), who state explicitly that the origin of training is not uniquely established. Wu et al. (2016) separated distinct antiferromagnetic spin-motion modes during training of Co/CoO with in-plane anisotropy, consistent with AF rearrangement but not depth-resolving it. Depth-sensitive probes that could settle the question exist — polarized neutron reflectometry, soft-x-ray resonant magnetic reflectivity, which recovers magnetization depth profiles from the magnetic modulation of resonant soft-x-ray reflectivity (demonstrated for a perpendicular exchange-coupled system by Tonnerre et al., 2008), standing-wave-enhanced dichroism, and vector-resolved XMLD geometries — but they have not been applied to depth-profile the AF twist in a trained Co/CoO bilayer. What is missing is a direct, layer-resolved test of the twisted-AF-profile picture together with a quantitative account of the energetics — interfacial exchange, AF magnetocrystalline anisotropy, and interface frustration — that would stabilize the spring and account for the field-cooled and trained states, the relaxation of the exchange-bias field across consecutive loops, and its recovery after zero-field warm-up to TBT_B.

Problem Statement

Establish microscopically and quantitatively whether an antiferromagnetic exchange spring — a continuously twisted antiferromagnetic spin structure — forms at the Co/CoO interface during exchange-bias training, as proposed by Dobrynin et al. (2014), and characterize it sufficiently to decide whether it is the origin of the training effect and of the training recovery after zero-field warm-up to the blocking temperature TBT_B. Concretely: (1) determine the layer-resolved rotation profile and spatial extent of the antiferromagnetic twist (how many CoO layers participate and how the AF orientation varies with depth) in the field-cooled versus trained states, going beyond the interface-averaged X-ray magnetic linear dichroism (XMLD) evidence — for example by element- and depth-resolved dichroism, vector-resolved XMLD geometries, standing-wave or resonant-reflectivity techniques, or polarized neutron reflectometry; and (2) establish the energetics — how interfacial exchange, CoO magnetocrystalline anisotropy (including its temperature dependence toward TBT_B), and interface frustration combine to stabilize the spring — and account quantitatively for the two distinct states (high unidirectional anisotropy after field cooling, low unidirectional anisotropy after training), for the relaxation of the exchange-bias field across consecutive hysteresis loops during the initial field cycling, and for the recovery of the field-cooled state after warming to TBT_B and re-cooling in zero field. A full resolution requires both (1) and (2); a partial answer addressing only one of them does not resolve the problem. A valid resolution must discriminate the twisted-AF-spring picture against competing mechanisms (irreversible AF domain rearrangement, rotatable uncompensated interface spins, and interfacial domain walls, a training mechanism distinct from the spring's wall-like limit) rather than merely demonstrate consistency with averaged observables; a direct depth-resolved refutation showing that the proposed antiferromagnetic spring is not the reason of the training effect also resolves the problem.

References

  1. A. N. Dobrynin, F. Maccherozzi, S. S. Dhesi, R. Fan, P. Bencok, P. Steadman, 'Antiferromagnetic exchange spring as the reason of exchange bias training effect', Applied Physics Letters 105, 032407 (2014). DOI: 10.1063/1.4891651, https://doi.org/10.1063/1.4891651
  2. F. Y. Yang, C. L. Chien, 'Spiraling Spin Structure in an Exchange-Coupled Antiferromagnetic Layer', Physical Review Letters 85, 2597 (2000). DOI: 10.1103/PhysRevLett.85.2597, https://doi.org/10.1103/PhysRevLett.85.2597
  3. A. Scholl, M. Liberati, E. Arenholz, H. Ohldag, J. Stöhr, 'Creation of an Antiferromagnetic Exchange Spring', Physical Review Letters 92, 247201 (2004). DOI: 10.1103/PhysRevLett.92.247201, https://doi.org/10.1103/PhysRevLett.92.247201
  4. Y. Y. Wang, C. Song, G. Y. Wang, F. Zeng, F. Pan, 'Evidence for asymmetric rotation of spins in antiferromagnetic exchange-spring', New Journal of Physics 16, 123032 (2014). DOI: 10.1088/1367-2630/16/12/123032, https://doi.org/10.1088/1367-2630/16/12/123032
  5. C. Binek, 'Training of the exchange-bias effect: A simple analytic approach', Physical Review B 70, 014421 (2004). DOI: 10.1103/PhysRevB.70.014421, https://doi.org/10.1103/PhysRevB.70.014421
  6. S. Brems, K. Temst, C. Van Haesendonck, 'Origin of the Training Effect and Asymmetry of the Magnetization in Polycrystalline Exchange Bias Systems', Physical Review Letters 99, 067201 (2007). DOI: 10.1103/PhysRevLett.99.067201, https://doi.org/10.1103/PhysRevLett.99.067201
  7. T. Hauet, J. A. Borchers, Ph. Mangin, Y. Henry, S. Mangin, 'Training Effect in an Exchange Bias System: The Role of Interfacial Domain Walls', Physical Review Letters 96, 067207 (2006). DOI: 10.1103/PhysRevLett.96.067207, https://doi.org/10.1103/PhysRevLett.96.067207
  8. A. G. Biternas, R. W. Chantrell, U. Nowak, 'Dependence of training effect on the antiferromagnetic structure of exchange-bias bilayers within the domain-state model', Physical Review B 89, 184405 (2014). DOI: 10.1103/PhysRevB.89.184405, https://doi.org/10.1103/PhysRevB.89.184405
  9. E. Menéndez, L. E. S. Silva, G. Johann, J. Sort, T. Dias, 'Unraveling the origin of training in granular Co-CoO exchange bias systems with buried antiferromagnetic constituents', Journal of Magnetism and Magnetic Materials 478, 170 (2019). DOI: 10.1016/j.jmmm.2019.01.104, https://doi.org/10.1016/j.jmmm.2019.01.104
  10. R. Wu, C. Yun, S. L. Ding, X. Wen, S. Q. Liu, C. S. Wang, J. Z. Han, H. L. Du, J. B. Yang, 'A separation of antiferromagnetic spin motion modes in the training effect of exchange biased Co/CoO film with in-plane anisotropy', Journal of Applied Physics 120, 053902 (2016). DOI: 10.1063/1.4960092, https://doi.org/10.1063/1.4960092
  11. A. N. Dobrynin, P. Warin, A. Vorobiev, D. Givord, 'On the origin of positive exchange bias and coercivity enhancement in proximity to the blocking temperature', Journal of Magnetism and Magnetic Materials 520, 166707 (2021). DOI: 10.1016/j.jmmm.2020.166707, https://doi.org/10.1016/j.jmmm.2020.166707
  12. J. M. Tonnerre, M. De Santis, S. Grenier, H. C. N. Tolentino, V. Langlais, E. Bontempi, M. García-Fernández, U. Staub, 'Depth Magnetization Profile of a Perpendicular Exchange Coupled System by Soft-X-Ray Resonant Magnetic Reflectivity', Physical Review Letters 100, 157202 (2008). DOI: 10.1103/PhysRevLett.100.157202, https://doi.org/10.1103/PhysRevLett.100.157202
  13. Z. W. Liang, Z.-H. Wang, B.-G. Shen, L. C. Wang, F. P. Wang, P. Wu, 'Exchange bias training in La0.7Sr0.3FeO3/Ni79Fe21 films with native surface oxide: First-order reversal curve analysis', Physical Review B 104, 064440 (2021). DOI: 10.1103/PhysRevB.104.064440, https://doi.org/10.1103/PhysRevB.104.064440
  14. G.-h. Guo, G.-f. Zhang, S.-y. Song, D. W. Wang, G. J. Bowden, P. A. J. de Groot, 'Irreversible magnetic exchange-spring processes in antiferromagnetic exchange-coupled bilayer systems', Applied Physics Letters 93, 102505 (2008). DOI: 10.1063/1.2972028, https://doi.org/10.1063/1.2972028