**What are typical linewidths ?** Besides output power (for which good strategies exist to reach the required specifications), linewidth is an important issue and determines basically the type of applications where STOs can compete. The linewidth measured in the literature for a large variety of oscillators varies from about 1 GHz to 1 MHz (with the exception of vortex oscillators that can have linewidths down to tens of kHz). In most oscillator configurations and experimental situations, these values are too large and great efforts are currently undertaken in order to understand the origin of linewidth broadening and to define configurations to reduce it (see also sections on Origin of Linewidth Broadening).

**What do we know from theory ?**At zero temperature, and in absence of any noise, the linewidth is expected to be zero (Dirac emission peak), since the spin transfer torque cancels the damping torque. At non zero temperature or in the presence of any other noise, the linewidth is non-zero and depends on [Kim, Slavin] :

- the total noise power
- the total excitation power (and with this the excitation volume)
- the non-linear coupling between frequency and amplitude

In most experiments on spin torque driven excitations, the polarizer is considered as being fixed in its orientation and only excitations of the free layer are studied, where the free layer is a single magnetic thin film element. An increase of the excitation volume and thus a decrease of the linewidth can be expected when using instead of a single layer, two magnetic layers that are coupled via RKKY exchange. Such a coupled structure is part of most spin valves or tunnel junctions as a polarizer in form of a synthetic antiferromagnet where the two magnetic layers are aligned antiparallel. Via the reciprocal effect of spin transfer, spin torque driven excitations can also be induced in the SAF for opposite current and field signs as the excitations in the free layer.

**Linewidth of a synthetic antiferromagnet**

- We have studied the spin torque driven excitations of a spin valve oscillator, with a synthetic antiferromagnetic polarizer and a single free layer [1]. The SAF is exchange biased by an antiferromagnet. The antiparallel configuration of the SAF is stable in a field range of -1kOe to +1 kOe.
- We find excitations of the free layer for positive field and negative current, as well as for negative field and positive current. We also find excitations of the SAF for positive field and positive current, see Fig. 1.

- Comparing the linewidth of the FL and SAF excitations, see Fig. 2, we see clearly that
__the linewidth of the SAF excitations is reduced by a factor of 10 as compared to the free layer__excitations. This confirms the possibility to improve the linewidth characteristics in a SAF structure

**Influence of dynamic exchange on the frequency**

Another interesting property that we have observed for the SAF excitations [1], and that have been confirmed by simulations [2], is the change of the slope of the frequency versus current, see Fig. 3. In low fields, this slope is negative, corresponding to a frequency redshift, as for a single free layer. In higher fields, this slope is positive, corresponding to a frequency blueshift. This blueshift has no analogy in the single free layer (when considering an in-plane precession mode). It is due to the dynamic RKKY exchange interactions [2]. When for increasing current or applied field the precession amplitude increases, the magnetizations of the two SAF layers approach on certain points along their trajectory. This costs a lot of exchange energy. In order to reduce this dynamical exchange, the trajectory distorts (compare trajectories (1) and (2) in Fig. 3) and results in a change of the frequency.

**Influence of dynamic exchange on the frequency**

Another interesting property that we have observed for the SAF excitations [1], and that have been confirmed by simulations [2], is the change of the slope of the frequency versus current, see Fig. 3. In low fields, this slope is negative, corresponding to a frequency redshift, as for a single free layer. In higher fields, this slope is positive, corresponding to a frequency blueshift. This blueshift has no analogy in the single free layer (when considering an in-plane precession mode). It is due to the dynamic RKKY exchange interactions [2]. When for increasing current or applied field the precession amplitude increases, the magnetizations of the two SAF layers approach on certain points along their trajectory. This costs a lot of exchange energy. In order to reduce this dynamical exchange, the trajectory distorts (compare trajectories (1) and (2) in Fig. 3) and results in a change of the frequency.

**SPINTEC Publications**

[1] Appl. Phys. Lett. **96**, 072511 (2010), D. Houssameddine, et al ;

[2] Phys. Rev. B **79**, 104406 (2009), D. Gusakova et al;

**Main contributors :**

- Dimitri Houssameddine, PhD, STO microwave characterization
- Juan Sierra, Postdoc, STO microwave characterization
- Daria Gusakova, Liliana Buda-Prejbeanu, simulation
- Bertrand Delaët, nanofabrication
- Marie-Claire Cyrille
- Bernard Dieny
- Ursula Ebels

**References**

[Kim] Phys. Rev. Lett. **100**, 017207 (2008), J.-V. Kim et al;

[Slavin] IEEE Trans. Magn. **45**, 1875 (2009), A. N. Slavin et al;