Blog entry

What is a Magnetic Tunneling Junction?

By José Luis Sambricio (University of Manchester)

A Magnetic Tunneling Junction (MTJ from now on) is a relatively simple device: it consists on a spacer, typically Aluminium Oxide or Magnesium Oxide sandwiched by two ferromagnets, as depicted in figure 1:


Figure 1. Simple sketch of a MTJ:  a spacer between two ferromagnets.

In a MTJ both ferromagnets must behave differently. In particular, it must be possible to magnetise them in an antiparallel configuration, or, in other words, their reversal field -denoted by Hc in figure 2- has to be different. Let us consider a situation where a high field is applied initially, so that both ferromagnets will be oriented in the same direction, in the so called parallel configuration. As the magnitude of the field is reduced, the magnetization of one of the ferromagnets will change direction –our device has entered the antiparallel configuration. Upon further reduction of the field, the second ferromagnet will switch as well, returning to the parallel configuration. 


Figure 2. Hysteresis loop of a typical ferromagnet. Hc denotes the reversal field, the field under which the magnetization changes direction.

The most important characteristic of a MTJ is that the electrical resistance of this device depends on the specific configuration (parallel/antiparallel), as shown in figure 3.

Figure 3. The resistance of a MTJ is higher in the antiparallel configuration (denoted by the antiparallel arrows).

The fundamental origin of this change in resistance is complicated and still remains as an important area of study. The principles of all current theories rely on different tunnelling probabilities for the majority and minority spins from one ferromagnet to the other, through the spacer.  In particular, in the first theory that explained the different tunnelling probabilities (Julliere, 1975) two basic assumptions were made:

• During the tunnelling, the spin of every single electron is conserved.

• The tunnelling probability is proportional to the product of the densities of states of both ferromagnets for such spin.

The first assumption implies that spins stay as up or down during the tunnelling process, not switching from one to the other. The second assumption implies that the tunnelling probability is high if the spin is a majority spin in both ferromagnets, intermediate if the spin is majority in one ferromagnet and minority in the other,and low if it the spin is minority on both ferromagnets.

At this point the reader may confuse majority/minority spins with spins up/down. Basically, spins up are those pointing in a certain direction (which we choose arbitrarily) and those pointing in the opposite direction are called spins down. In contrast, majority spins are those pointing in the direction of the magnetization of a ferromagnet and minority spins are those pointing in the opposite direction. Remember that the direction of the magnetization of a certain ferromagnet may or may not coincide with that of spins up or down!

In figure 4 you can observe how the tunnelling probability will be different in the parallel configuration or in the antiparallel:


Figure 4. a) Parallel configuration. b) Antiparallel configuration.

In the case of a parallel configuration (figure 4a) the spins down (blue arrows) are minority spins in both ferromagnets and, according to the second assumption have a low tunnelling probability, while the spins up -majority on both ferromagnets, red arrows- have a high tunnelling probability. In the case of the antiparallel configuration (figure 4b) spins up and down are majority in one ferromagnet and minority in the other; therefore the tunnelling probabilities for both are normal.

This means that in a parallel configuration tunnelling are more common than in the antiparallel configuration, resulting in the lower resistance in the parallel state encountered in figure 3.

This difference in the resistance, depending on the magnetization of two ferromagnets is being used in many commercial hard drives and solid state memories to store information. Basically, the parallel/antiparallel states constitute a “bit” of information (1/0) which can be read due to the difference in resistance. 

José Luis Sambricio (U. Manchester)