Non-collinear spin-valve

Sensoren, Geräte und Komponenten

Ref.-Nr.: 1201-5998-BC

A classical spin valve (SV) is a device in which the electrical resistance depends on the relative position of the magnetic moments of two ferromagnetic (FM) layers: with parallel (P) and antiparallel (AP) ordering, the SV resistance is minimum and maximum, respectively. Nowadays SVs are actively used in magnetic sensors, hard disk read heads and magnetic random-access memory. Thus, the SV, analogous to a transistor in conventional electronics, is a binary (digital) device. Here we offer a new technology that allows fixing the direction of the moments of M1 and M2 at any arbitrary non-collinear alignment between AP and P states, which makes it possible to create a non-binary spin valve. Such a multiple state resistor can be used, for example, as a multi-state memory cell, or a synapse in an artificial neural network.


The design is based on the SV scheme shown in Fig. 1, with two FM layers (F1, F2) with different types of magnetic anisotropy (uniaxial and unidirectional), and a non-magnetic spacer N2. The FM layer F1 is deposited on the substrate SU with an optional non-magnetic buffer layer N1. The F1 layer has an easy magnetic axis due to uniaxial magnetic anisotropy, which can be induced by the choice of substrate, the growth conditions, or by the shape anisotropy of the device. The FM layer F2 forms an exchange bias pair with the antiferromagnet AF. The Néel temperature TN of the AF is chosen to be smaller than the Curie temperature Tm of the FM layers. The EB effect occurs between the neighboring AF and F2 layers at T<TN. It induces a unidirectional anisotropy (UD) in the F2 layer, such that the direction of M2 can only be altered by a strong field larger than the exchange bias field HEB. The latter is in the order of tens of Oe to several kOe, depending on the layer materials. At T>TN the exchange bias and the unidirectional anisotropy disappear, such that M2 can align parallel to a small field HCL. The direction of the UD axis is defined by the magnetization vector M2 during cooling below TN, which in turn is defined by the field HCL applied during cooling.

Defining or writing the NC magnetization state with angle Da requires the following steps: (i) The AF layer is heated above TN to remove EB. (ii) A small field HCL is applied at an angle Da with respect to the easy axis EA of F1. (iii) The system is cooled below TN with HCL applied. M2 will stay parallel to HCL during cooling, and the direction of UD will be defined by M2. (iv) HCL is removed. M2 is now stabilized by the exchange bias, and stays in the direction previously defined by HCL. In this latter step after removing HCL, M1 aligns along the easy axis EA, such that in this remnant state (T<TN, H=0) the angle between M1 and M2 corresponds to the requested Da. The angle Da affects the electrical resistance of the layer system, such that the readout of the state is performed by measuring the resistance of the device, either in current-in-plane or current-perpendicular-to-plane mode.

In summary, the above-mentioned procedure allows to create a non-volatile device with a well-defined degree of non-collinearity. Another way to alter Da, albeit volatile, is to apply a magnetic field H, which is smaller than Heb of F2 layer. This will rotate the direction of M1 from EA towards H, while the direction of M2 will still be pinned along the HCL. In particular, this small field H can also be used to invert (rotate by 180°) the direction of F1 with respect to the EA, such that in the field free remnant state M1 can be aligned both parallel and antiparallel to EA.

The advantage of this method is that the non-collinear angle Da between M1 and M2 can be set at arbitrary values between -90° and 90° in a reproducible and accurate way. The angle Da is changed by, for example, heating above TN, followed by cooling in the field with the required orientation. Another advantage of the method is a wide range of materials for the F1(2), N1(2) and AF layers, which allows the system to be used in various applications. By choosing an AF layer with proper TN, the operation temperature of the device is adapted to specific applications. For example, by choosing IrMn3 with TN>300K as AF layer allows for preparation of devices working at room temperature. By choosing superconductors for the N1 and/or N2 layers, superconducting NC spin valves or triplet Josephson junctions with variable and arbitrary Da can be designed.

Fig. 1. Scheme (a) and principle of operation of the non-collinear spin-valve. The ferromagnetic layer F1 deposited on the substrate Sb with an optional buffer layer N1 has uniaxial anisotropy with the direction of the easy axis shown by the black arrow with caption EA. The ferromagnetic layer F2 and the antiferromagnetic layer AF form an exchange bias pair. (b) Above the Néel temperature TN of the AF layer an external magnetic field HCL is applied at an angle Da with respect to EA. The system then is cooled below TN in the magnetic field HCL. After release of the magnetic field to zero, the magnetization vector M1 of the layer F1 will turn towards the direction of EA while the magnetization M2 of the layer F2 will stay along the direction of HCL due to the exchange bias. (c) Thus, the remnant state will be characterized by the non-collinear alignment of M1 and M2 with the angle Da.

The figure is courtesy of Reiner Müller; FRM II/TUM


  • Possibility to create a non-volatile SV with adjustable resistance in the range between Rmin and Rmax corresponding to the parallel and antiparallel state respectively.
  • Wide range of materials for the F1(2), N1(2) and AF layers, which allows the system to be used in various applications both at low and at room temperatures.
  • Compatibility with existing technologies for the preparation of spintronic devices.

Patent Information

Patent filed.

PDF Download


Senior Patent- & Lizenzmanager

Dr. Bernd Ctortecka, M. Phil.


Telefon: 089 / 29 09 19-20

Patent- & Lizenzmanager

Dr. Andreas Vogler


Telefon: 089 / 29 09 19-36