Magnetic memory effect in the complex metallic alloy T-Al-Mn-Pd

(1)

Magnetic memory effect in the complex metallic alloy T-Al-Mn-Pd

M. Feuerbacher

¹,

M. Heggen

¹,

J. Dolinsek

²,

J. Slanovec

³,

Z. Jagliˇci´c

³

1 IFF-8: Microstructure Research

2 J. Stefan Institute, University of Ljubljana, Jamova 39, SI-1000 Ljubljana, Slovenia

3 University of Ljubljana, Jadranska 19, SI-1000 Ljubljana, Slovenia

The Taylor-phase T-Al3Mn,T-Al3(Mn,Pd) and T- Al3(Mn,Fe) series of complex intermetallic com- pounds, belong to the class of magnetically frustrated spin systems that exhibit rich out- of-equilibrium spin dynamics in the nonergodic phase below the spin–freezing temperature Tf. We observe a memory effect in these materials:

the spin structure of the material stores informa- tion on isothermal aging steps carried out during zero-field cooling, which is detected by measur- ing the magnetization of the sample.

The most prominent example of magnetically frustrated systems are spin glasses (SGs). A SG is a site-disordered spin system that is frustrated and spatially disordered in the sense that the spins are positioned randomly in the sample. These two prop- erties lead to highly degenerate free-energy land- scapes with a distribution of barriers between differ- ent metastable states, resulting in broken ergodicity.

Typical SGs are dilute magnetic alloys of noble metal hosts (Cu, Ag, Au) with magnetic impurities (Fe, Mn), the so-called canonical spin glasses. In this paper we show that pronounced broken-ergodicity phenomena are present also in a class of ordered complex intermetallic Taylor phases [1] T-Al3Mn, T-Al3(Mn,Pd) and T-Al3(Mn,Fe).

We have investigated samples grown by means of the Bridgman technique. Compositions were Al73Mn27for the binary basic phase and ternary ex- tensions with Pd and Fe substituting 2, 4 and 6 at.%

Mn. Magnetic measurements were conducted in a Quantum Design SQUID magnetometer equipped with a 50 kOe magnet, operating in the temperature range 2 – 300 K.

Fig. 1 shows low-temperature measurements of the magnetic susceptibilityχ for samples with Pd contents between 2 and 6 at.% at constant field of 8 Oe. The curves show a pronounced maximum, the freezing temperatureTf, which represents the tran- sition from the non-ergodic to the ergodic regime.

AboveTf, the curves show regular Curie-Weiss behaviour typical for a paramagnetic state. BelowTf, the curves show splitting between the field cooled (fc) and zero-field cooled (zfc) susceptibility, which is a fingerprint behaviour of a SG. Similar curves are found for the binary T-phase and that containing Fe.

The labels fc and zfc refer to the conditions under

which the sample was cooled down to the lowest temperature of the experiment, before the actual susceptibility measurement is taken. Under fc conditions, the sample is cooled down in a constant field, while under zfc conditions no field is present.

FIG. 1: Low-temperature susceptibilityχfor samples with Pd contents of 2 (T-AMP2), 4 (T-AMP4), and 6 at.% Pd (T- AMP6) at constant field H = 8 Oe. Below the freezing temperature the shows clear splitting between the fc and zfc susceptibility, which is typical for a spin glass.

We have then carried out the following measuring procedure: The samples are zero-field-cooled contin- uously from the starting temperature of 100 K into the nonergodic regime with a cooling rate of 2 K/min. At a temperatureTa, cooling is temporarily stopped and the spin system is let to age isothermally for a certain time tw, after which continuous cooling is resumed down to 2 K. At this temperature, a small magnetic field of 2 Oe is applied and the magnetizationM is measured in a subsequent heating run to a temperature aboveTf.

Fig. 2a shows corresponding magnetization measurements for the binary T-Al3Mn sample [2]. Here the isothermal aging stop was carried out atT1= 12 K fortw=10 min, 1 h, and 4 h. A reference run with no stop (tw= 0) atT1was also performed. For nonzero aging times we find a dip in the magnetization, which

(2)

FIG. 2: The memory effect: if an isothermal aging step is carried out during zfc cooling, the magnetization shows a dip at the aging temperature, which increases in depth with annealing time.

is located at the aging temperature, and the depth of which increases with aging duration. The material is thus able to store information on the aging temperature and duration, which we refer to as memory effect (ME). Fig. 2b shows an expanded portion of the curves in the vicinity ofT1. In Fig. 2c, the normalized difference∆M between the reference curve and the curves with aging is displayed.∆M resembles a res- onant curve, peaked at the aging temperature and smeared over a finite temperature interval of about

±2 K.

Similar experiments were carried out on the T-phases containing Fe and Pd. In all cases we consistently find the signatures of the ME effect regardless of the differences in the samples composition and structure.

The aging temperature Ta can be varied within the nonergodic regime, i.e. between 2 and 24 K for the present material. Aging at any temperature in this interval leads to a clear dip in the magnetization curve.

The memory imprint can be erased by a positive temperature cycle within the nonergodic regime. If the temperature is increased to aboutTa+2 K the memory is erased, i.e. the spin system is rejuvenated. In a zfc experiment after a positive temperature cycle, the magnetization is again that of the unaged system with no memory imprint, corresponding to the reference curve withtw= 0. Memory is erased thermally, regardless of the presence or absence of a small magnetic field. Any heating aboveTf into the ergodic phase erases the memory as well. The spin system is ready to memorize isothermal aging again just after the memory has been erased by a positive temperature cycle within the nonergodic phase. Memory erase has no effect on the subsequent memory imprint.

The effects observed can be discussed in terms of a “spin-droplet model”. During aging atTa, the mo- bile spins at that temperature try to equilibrate in an energetically favourable configuration among them- selves and with those spins already frozen. The degree of quasi-equilibration depends on the aging time the spin system is subjected to a given temperature under constant external conditions. In this way, quasi-equilibrated "spin droplets" are formed. Due to the predominant antiferromagnetic type coupling, the magnetization of the droplets tends to zero. Mag- netically quasi-ordered spin droplets are in a more stable configuration than the rest of the spin-glass matrix, so that higher thermal energy is needed to reverse a spin within a droplet. Resuming continuous zero-field-cooling after the isothermal aging, magnetic order within the droplets is partially frozen, whereas weaker-coupled spins gradually freeze in a spin-glass configuration at lower temperature. At the lowest temperature of the zfc run, all spins with re- orientational energies differing from kbTa are in a spin-glass configuration, whereas those with energies of about kbTa form regions with more stable quasi-equilibrated configurations. In a subsequent heating run, the magnetization linearly builds up ex- cept in the vicinity ofTa, where higher thermal energy is needed to reorient the spins in the more stable quasi-ordered droplets, and consequently a diminu- tion atTarelative to the no-aging case is found.

[1] M. A. Taylor, Acta Cryst.. 14,1961, 84

[2] J. Dolinsek, J. Slanovek, Z. Jaglicic, M. Heggen, S. Balanetskyy, M. Feuerbacher, and K. Urban. Phys.

Rev. B 77, 2008, 064430.