PHYSICAL REVIEW B 102, 024206 (2020)
Spin-glass phase transition revealed in transport measurements
Guillaume Forestier , Mathias Solana, and Cécile Naud
Institut Néel, Centre National de la Recherche Scientifique, B.P. 166, 38042 Grenoble Cedex 09, France
Andreas D. Wieck
Lehrstuhl für Angewandte Festkörperphysik, Ruhr-Universität, Universitätstraße 150, 44780 Bochum, Germany
François Lefloch
Institut Nanosciences et Cryogénie, Commissariat à l’Énergie Atomique, 17 avenue des Martyrs, 38054 Grenoble Cedex 09, France
Robert Whitney
Laboratoire de Physique et Modélisation des Milieux Condensés, B.P. 166, 38042 Grenoble Cedex 09, France
David Carpentier
Laboratoire de Physique, École Normale Supérieure de Lyon, 47 allée d’Italie, 69007 Lyon, France
Laurent P. Lévy and Laurent Saminadayar*
Institut Néel, Centre National de la Recherche Scientifique, B.P. 166, 38042 Grenoble Cedex 09, France
and Université Grenoble-Alpes, B.P. 53, 38041 Grenoble Cedex 09, France
(Received 4 May 2018; revised 5 May 2020; accepted 14 May 2020; published 31 July 2020)
We have measured the resistivity of magnetically doped Ag:Mn mesoscopic wires as a function of temperature
and magnetic field. The doping has been made using ion implantation, allowing a distribution of the dopants in the
middle of the sample. Comparison with an undoped sample, used as a reference sample, shows that the resistivity
of the doped sample exhibits nonmonotonic behavior as a function of both magnetic field and temperature,
revealing the competition between the Kondo effect and the RKKY interactions between spins. This proves that
transport measurements are still a reliable probe of the spin-glass transition in nanoscopic metallic wire doped
using implantation.
DOI: 10.1103/PhysRevB.102.024206
Spin glasses are one of the most fascinating states of matter. They have attracted the interest of a large community for
several decades, as it is one of the most fundamental problems
in condensed matter physics [1]. A spin glass appears when
magnetic atoms are randomly diluted in a nonmagnetic metallic host. As the spatial distribution of the spins is random,
the Ruderman-Kittel-Kasuya-Yoshida (RKKY) interactions
between the spins [2], which depend on the distance between
them, are also random: This leads to frustration between the
magnetic moments. It is this interplay between disorder and
frustration which is at the basis of the formation of a spin glass
below a transition temperature Tsg . It has been shown recently,
using a very tricky experiment, that this subtle scenario for
the formation of a spin glass is actually the real one [3]. The
very nature of the ground state is still heavily debated and may
consist of an unconventional state of matter with remarkable
behaviors [4]. Let us mention, however, that spin glasses
may also appear in insulating systems [5,6]: This peculiar
situation will not be discussed in this paper as it is focused on
*
saminadayar@neel.cnrs.fr
2469-9950/2020/102(2)/024206(7)
transport measurements. Finally, it is worth mentioning that
spin glass models have also been applied to sociology, biology
[7], economy [8,9], and games [10] and are deeply connected
to mathematics (topology) [11].
The spin-glass transition itself is quite subtle to detect. In
contrast with other phase transitions, there is, for example, no
divergence of the specific heat at the transition temperature,
but rather a cusp in the (low-frequency) magnetic susceptibility [6]. This has been the most commonly used technique
to detect the transition down to very low concentrations of
magnetic impurities [12]. More surprising is the onset of
irreversibility in the glassy phase: magnetization measured
during a cooldown under magnetic field [field cooled (FC)
procedure] is completely different from the magnetization
obtained when the magnetic field is applied after cooling
the sample [zero field cooled (ZFC) procedure]. This discrepancy between these two measurements appears exactly
at the spin-glass transition temperature [13,14]. This onset of
irreversibility is one of the most characteristic signatures of
the spin-glass transition.
Another way to probe the spin glass state and the spin
glass transition consists of measuring transport properties.
024206-1
©2020 American Physical Society
GUILLAUME FORESTIER et al.
PHYSICAL REVIEW B 102, 024206 (2020)
Historically, people were interested in measuring Kondo effect as a function of the concentration of magnetic impurities.
For high doping levels, they observed deviations from the
well known logarithmic Kondo-like increase of the resistivity:
Below a certain temperature, the resistivity exhibits a broad
maximum followed by a subsequent decrease [15–18]. Such
a behavior was interpreted as a transition to a spin-glass state
as spin-spin interactions become predominant. This interpretation was supported by theoretical work [19] which relates
the maximum in the resistivity at a temperature Tg to the
spin-glass transition. Another natural way to detect changes
in the spin configuration of a metal consists of measuring the
anomalous Hall effect [20,21]. As interactions between spins
become predominant, anomalous Hall resistance exhibits a
broad maximum followed by a rapid decrease as the number
of free (unfrozen) spins becomes smaller and smaller [22]. A
more subtle and difficult way to probe the spin-glass phase
consists of measuring the resistance noise of a metallic spinglass sample [23]. Finally, it should be stressed that the most
striking feature of the Kondo physics has also been observed
in transport measurements: The onset of irreversibility has
been observed in field-cooled (FC)/zero-field cooled (ZFC)
resistivity measurements [24].
The question of the existence of a spin-glass transition
for small systems has been largely addressed theoretically
using numerical simulations. Using state-of-the-art computing
facilities [25], it has been shown that samples as small as
30 × 30 × 30 spins already exhibit a spin-glass transition
[26–28]. Experimentally, the situation is much more complex:
In a mesoscopic sample (i.e., containing few spins), magnetic
signals become too small to be detected, even using the
most sensitive techniques like SQUID measurements. It is thus
tempting to use transport measurements to detect the spinglass transition, as resistivity measurement can easily be performed on nanometer-size samples. This technique has been
used in the past to measure the temperature dependance of the
resistivity of Kondo systems [29]: The idea was to detect the
effect of the finite size of the sample on the development of
the Kondo cloud and thus on the screening of the magnetic
impurities by the conduction electrons.
Metallic spin glasses are usually obtained by dilution of
magnetic impurities in the host metal. Below a critical concentration at which an alloy is formed, impurities will end
up in interstitial positions and the crystal structure of the
host metal is preserved. The main problem of this technique,
however, is that impurities may move during the evaporation
and form clusters. Determination of the actual concentration
and actual spin of the impurities are then sometimes delicate.
An alternative technique consists of implanting ions in the
metallic host [30]. This technique allows us to control perfectly the dose implanted [31] and to avoid any clustering
as energy barriers forbid any displacement of ions at room
temperature. Two points, however, differ between these two
techniques: During an implantation, the concentration of the
ions through the thickness of the sample presents a gaussian distribution. Very few ions are present at the surface
and spin glass physics should not be modified by surface
effects as it has been sometimes alleged. Moreover, ions are
implanted at high energy: Damages are thus induced in the
crystalline structure of the host metal. As RKKY interactions
FIG. 1. Electron micrograph of the sample. The wire is 18 μm
long, 200 nm wide, and 50 nm thick. Several voltage probes are set
along the wire distant by 2 μm.
are mediated by conduction electrons, spin glass transition is a
many impurities problem: The equivalence between these two
implantation techniques, from a physical point of view, is an
interesting and sensible question.
In this paper, we report on transport measurements on
nanometer-size mesoscopic Ag:Mn spin glasses at very low
doping level. Doping has been obtained by ion implantation
technique. Measurements have been performed down to very
low temperature (50 mK) and up to high magnetic field (8 T).
We show that the behavior of the resistivity can be perfectly
explained by an interplay between the Kondo effect and a
spin-glass phase transition. This proves that our mesoscopic
samples exhibit a phase transition similar to what is observed
for macroscopic samples and that the implantation process,
although producing crystalline defects, does not change the
spin glass physics in such confined geometries. The observed
transition temperature is identical to the one detected previously on macroscopic samples doped by standard dilution
techniques and with the same doping level using magnetic
measurements.
Samples have been fabricated on a silicon/silicon oxide wafer using standard electron-beam lithography on
polymethyl-methacrylate resist. Geometry of the sample consists of a long (length L ≈ 18 μm) and thin (width w ≈
200 nm, thickness t ≈ 50 nm) wire (see Fig. 1). Several contacts have been put along the wire, in order to measure the
resistance over different lengths and to thermalize properly
the electrons along the wire [32]. Silver has been evaporated
using a dedicated electron gun evaporator and a 99.9999%
purity source with no adhesion layer. Samples have then been
implanted with Mn2+ ions of energy 70 keV. This energy has
been chosen after numerical simulations based on calculations
using the SRIM software [31] in order to ensure that ions will
end up in the sample following a gaussian distribution whose
maximum lies in the middle of the sample thickness. This
technique of implantation allows us to avoid clustering or
migration of the Mn2+ ions as no further annealing has been
performed on the samples. The ion dose, measured via the
current of the implanter, has been chosen in order to give
a final ion concentration in the wire of 500 ppm. For this
024206-2
SPIN-GLASS PHASE TRANSITION REVEALED IN …
PHYSICAL REVIEW B 102, 024206 (2020)
FIG. 2. Resistivity as a function of temperature for the pure Ag
sample.
√ Inset: low temperature part (T 1 K) plotted as a function
of 1/ T . The red line is a fit of the data using Eq. (1).
concentration of magnetic impurities, the number of spins in
a section of the wire is approximately 450. This number is
quite small (typically a 15 × 30 network of spins) and can be
reasonably compared with the dimensions of state-of-the-art
numerical simulations. Finally, note that some samples have
been left unimplanted (pure) in order to be used as reference
samples.
Samples have been cooled down in a dilution fridge whose
base temperature T is ≈50 mK and equipped with a two-axis
superconducting coil of maximum magnetic fields Bz = 8 T
(out-of-plane field) and Bxy = 1.5 T (in-plane field). Transport measurements have been carried out using an ac lock-in
technique at a frequency of 11 Hz in a bridge configuration
and a very low ac current in order to avoid any overheating of
the sample. In particular, we have kept eVds kB T with Vds
the drain-source voltage, e the charge of the electron, and kB
the Boltzmann constant. We have seen no dependance of the
resistivity on the frequency of the excitation up to our experimentally accessible possibility (≈100 kHz). The signal is amplified using an ultra-low noise
√ homemade voltage amplifier
(voltage noise Sv = 500 pV/ Hz) at room temperature. All
the measuring lines connecting the sample to the experimental
setup consist of lossy coaxes which ensure a very efficient
radio-frequency filtering and thus a good thermalization of the
electrons in the sample [33–35]. At 4.2 K, the resistance of
our samples is R ≈ 10 for the pure samples and R ≈ 30
for the doped ones.
The resistivity of the pure sample as a function of temperature is depicted in Fig. 2 (a small magnetic field of ≈1000 G
is applied in order to cancel the weak localization correction
[36]). At high temperature, the resistivity is dominated by
electron-phonon interaction and thus decreases rapidly with
decreasing temperature. At low temperature, one observes a
slight increase of the resistivity; this is typical of mesoscopic
samples, in which electron-electron interactions modify the
density of states at the Fermi energy [36]. The resistance R of
the sample as a function of the temperature T is then given by:
R(T ) = 0.782λσ
√
R2 LT
α
=√ ,
RK L
T
(1)
where LT = h̄D/kB T is the thermal length, D the diffusion
coefficient, Rk = h/e2 the quantum of resistance with h the
FIG. 3. Resistivity as a function of temperature under zero field
of the Ag:Mn spin glass sample doped at 500 ppm. Inset: The black
curve is the same data on a logarithmic scale, while the green curve
has had the electron-electron interaction contribution subtracted.
Planck constant and e the charge of the electron, and λσ a
constant related to the screening parameter of the Fermi liquid
theory [36]. For silver, λσ ≈ 3.1. The inset of Fig. 2√shows
the resistance of the pure sample as a function of 1/ T . As
expected, we observe a nice linear behavior down to 50 mK,
proving that the electrons are indeed cooled down to the
lowest temperature [37].
The same measurements on the doped wire are depicted in
Fig. 3. The behavior is clearly different: At low temperature,
one observes a pronounced minimum, followed by an increase
of the resistivity and an abrupt decrease at very low temperature. The temperature dependence of the resistivity can be
separated into four terms:
δρ(T ) = δρe-ph (T ) + δρe-e (T ) + δρe-mag (T ) + δρwl (T ) (2)
corresponding, respectively, to the electron-phonon interaction, electron-electron interaction, and electron-magnetic impurity contribution to the resistivity (which is absent for the
pure sample). The term δρwl (T ) corresponds to the quantum
correction to the conductivity (weak localization).
This magnetic contribution arises from two distinct phenomena. The first one is the well-known Kondo effect while
the second corresponds to the formation of a spin-glass state.
It has to be noted that those two effects do not emerge from
the same processes. The Kondo effect is due to the scattering
of conduction electrons off up-down degenerated magnetic
impurities in the single impurity limit. The typical energy for
this process is the Kondo temperature TK . Such a behavior has
been extensively studied and leads to a logarithmic temperature dependence of the resistivity [38] as the temperature is
reduced.
In order to fit the data, we use Hamann’s law [39] which is
a high (T ≫ TK ) temperature expansion for the resistivity in
the Kondo regime:
0
ρKondo
ln (T /TK )
ρKondo (T ) =
1−
(3)
2
ln2 (T /TK ) + π 2 S(S + 1)
024206-3
GUILLAUME FORESTIER et al.
PHYSICAL REVIEW B 102, 024206 (2020)
case, this leads to:
ρ(T ) =
Tsg
A
1
−
α
,
S
T
ln2 (T /TK )
(5)
where A is a constant and αS a constant which depends on
the spin S (for S = 5/2, αS = 2.33). The temperature of the
maximum in the ρ(T ) curve, corresponding to the transition
between the high temperature phase and the spin-glass one, is
then given by:
FIG. 4. Resistivity of the spin-glass sample as a function of
temperature and after subtraction of the electron-electron interactions
contribution to the resistivity. Left panel: orange line is a fit using
Hamann’s law [Eq. (3)]. Black line is the same fit but adding a
power law to take into account the phonon scattering at “high” T .
Right panel: same data but restricted to the 0.05–5.5 K temperature
range. Green line is a fit using Vavilov’s equation [Eq. (5)] which
describes the emergence of the spin-glass phase and the corresponding decrease of the resistivity. The different arrows indicate TK as the
Kondo temperature, Tsg as the spin-glass transition temperature, and
Tm as the temperature of the maximum in the ρ(T ).
0
with S the spin of the magnetic impurities and ρKondo
a
constant given by:
0
ρKondo
=
4π h̄cimp
ne2 kF
(4)
with n the electronic density, cimp the concentration of magnetic impurities, e the charge of the electron, and kF the
Fermi wave vector. Resistivity of the sample as a function of
temperature is depicted in the left panel of Fig. 4. The orange
line is a fit for T > 4 K using equation (3), with the fitting
giving S = 5/2 and TK = 40 mK. These values are in good
agreement with those found in the literature [40] for diluted
manganese ions in silver.
At “high” temperature (from 4 K to 15 K) the main contribution to the resistivity is due to the electron-phonon
scattering. To take into account for this scattering, we have
added to Hamann’s law a power law αT n . Remarkably, this
combination of Kondo effect and electron-phonon scattering
describes the experimental data very well between 4 K and
15 K.
At “low” temperature (below 4 K), the resistivity deviates
from this simple Kondo description as the spin-glass behavior
starts to be prominent. Indeed, at those temperatures, RKKY
interactions between magnetic impurities are no more negligible, leading to a progressive lift of the up-down spin degeneracy. The density of spins cimp involved in the Kondo processes
is thus lowered, leading to a decrease of the resistivity. In this
regime, the diffusion mechanisms are completely different
and much more complex to calculate. This is related to the
appearance of the spin-glass state which is fundamentally very
complicated to apprehend. Recent theoretical works have been
able, however, to obtain an analytical expression for ρ(T )
by simplifying the RKKY interactions to interactions within
impurity pairs [41,42]. In the limit where Tsg ≫ TK , as in our
Tm ≃
Tsg
αS
Tsg ln
.
2
TK
(6)
In the right panel of Fig. 4, we have plotted the low temperature part of the resistivity of the spin-glass sample. The
green line is a fit using equation (5) with TK = 40 mK. The
fit is in rather good agreement, keeping in mind that equation
(5) is valid only in the proximity of Tsg . From this we obtain
that two parameters are A = 0.124 and Tsg = 500 mK. This
value of Tsg is in perfect agreement with those obtained using
magnetization measurements on macroscopic samples at the
same doping level [12] (see especially Fig. 3 in this reference).
This suggests that even for such small systems, a spin-glass
transition appears at the same temperature as for macroscopic
systems. It must be stressed, however, that only a complete
study including measurements of thermodynamic quantities
(specific heat, magnetization) could unambiguously lead to
this conclusion. Such measurements are, unfortunately, very
difficult on such small samples. The two arrows on the right
panel of the Fig. 4 indicate the spin-glass transition temperature Tsg (left arrow) and the temperature corresponding to the
maximum of the ρ(T ) curve Tm (right arrow). Using equation
(6) with the parameters determined above, we obtain the
theoretical temperature of the maximum Tm ≈ 1.5 K, which
is precisely what is observed experimentally. Considering
Fig. 4, one can see that these resistivity measurements can
be very well described considering the system as a metallic
spin glass at low temperature and as a Kondo system at
high temperature. It should be noticed than between 2.5 and
3.5 K, experimental data deviate from both fits. This region
corresponds to the range of temperature where Kondo effect
and RKKY interactions compete with almost equal strength
and no theoretical approach is able to completely describe this
mixed behavior.
Finally, in order to characterize the effect of the magnetic
field on the resistivity of the sample, we have performed high
field magnetoresistance measurements at different temperatures. For this purpose, we use the zero field cooled (ZFC)
protocol: The sample is cooled down under zero field and the
magnetic field is applied at low temperature. The relative variation of the resistivity, ρ/ρ0 = (ρ(B) − ρ(B = 0))/ρ(B =
0), has been plotted in Fig. 5 for temperatures down to
200 mK (i.e., T ≪ Tsg ) and up to 20 K (i.e., T ≫ Tsg ). At
high temperature (T ≫ 10 K), curves are perfectly superimposed and the magnetoresistivity is perfectly quadratic in B,
as can be seen in Fig. 6. In this regime, the resistivity is
dominated by electron-phonon scattering, and one recovers
the classical quadratic magnetoresistance of metals. In order
to extract the magnetic contribution to the magnetoresistivity,
we have subtracted the high temperature contribution in Fig. 6
from the data in Fig. 5. The result, plotted as a function of
024206-4
SPIN-GLASS PHASE TRANSITION REVEALED IN …
PHYSICAL REVIEW B 102, 024206 (2020)
FIG. 5. Relative magnetoresistivity of a 500 ppm spin glass sample for different temperatures.
the normalized magnetic field μB B/kB T , with μB the Bohr
magneton, is depicted in Fig. 7.
For temperatures larger than 1.5 K, all the curves are rather
well superimposed. This means that, in this range of temperature, the resistivity depends only on the ratio μB B/kB T ,
which is proportional to the polarization, i.e., the number of
free spins [19,43] (Curie’s law). As mentioned above such
behavior is typical of the Kondo effect as it is a single impurity
process.
Below 1 K, however, we observe strong deviations to this
behavior, showing that Kondo physics is not the good description anymore. Note that this is the temperature at which the
ρ(T ) exhibits a maximum (see Fig. 4). It is quite surprising
that deviations from Kondo physics appear at such high temperature as compared to the spin glass transition temperature.
We would like to stress, however, that recent measurements of
universal conductance fluctuations on mesoscopic spin glass
wires suggests that spin-spin interactions already play a role
well above Tsg [30]. Whether this is due to the nature of
the transition itself or to the reduce dimensionality of these
mesoscopic samples is still an open and intriguing question.
Similar behavior can be observed on the temperature dependance of the resistivity under magnetic field. This measurement is depicted in Fig. 8. As we have seen previously,
FIG. 6. Relative magnetoresistivity of a 500 ppm spin glass sample for temperatures larger than 10 K plotted as a function of B
(left panel) and B2 (right panel) (same data as Fig. 5). Blue curve
represents data measured at 10.1 K when the relative magnetoresistance starts to be positive (see Fig. 5). Green curve represents data
measured well above 10, i.e., far in the saturated regime (it has been
shifted for a sake of clarity).
FIG. 7. Relative magnetoresistivity of a 500 ppm spin glass sample as a function of the normalized magnetic field for different
temperatures. The classical B2 contribution has been subtracted.
under zero field, the competition between Kondo physics
and spin-spin interactions leads to a broad maximum in the
ρ(T ) curve. Under magnetic field, the maximum observed
in the ρ(T ) curve is progressively suppressed and shifted
towards higher temperatures. This is due to the fact that the
field progressively suppresses the up-down degeneracy and
thus the Kondo effect. The slope of the logarithmic increase
in ρ(T ) becomes weaker: The maximum amplitude in the
resistivity thus decreases [43]. Moreover, the shift of the ρ(T )
maximum is explained by the fact that, since Kondo effect
is reduced, spin-spin interactions become dominant at higher
temperatures; note that this still applies even when magnetic
field becomes larger than the typical spin-spin interaction
energy μB B ≫ kB Tsg .
As a final point, we would like to stress that such a
characterization of nanoscopic spin glass wires fits naturally
into recent works on mesoscopic spin glasses. In this domain,
the goal is to probe magnetic configuration of the spins via the
dephasing induced on the coherent conduction electrons [36].
The most striking result, reproduced in several experiments
[30,44–47], of this exploration of mesoscopic systems is that
the configuration of spin glasses are unexpectedly robust
against the application of high magnetic fields B (μB B ≫
kB Tsg ): Upon cycling samples under magnetic field of several
Tesla, universal conductance fluctuations are perfectly reproducible, meaning that the spins are in the exact same configuration before and after the application of the field. Our work,
FIG. 8. Resistivity of a 500 ppm spin glass sample as a function
of temperature under different magnetic fields.
024206-5
GUILLAUME FORESTIER et al.
PHYSICAL REVIEW B 102, 024206 (2020)
if it does not shed new light on this strange and remarkable
experimental fact, gives some hints for its interpretation as it
excludes the hypothesis that it may be due to an absence of
spin glass transition in nanoscopic samples.
As a conclusion, we have measured the magnetoresistivity of a 500 ppm Ag:Mn mesoscopic spin glass wire as a
function of temperature and magnetic field. We have explored temperatures ranging from 50 mK (T ≃ TK ≪ Tsg )
up to 10 K (T ≫ Tsg ≫ TK ) and magnetic fields up to 8 T
(μB B ≫ kB Tsg ≫ kB TK ). Despite the small number of spins
in the section of the wire (typically 15 × 30 spins), we observe a signature of the spin glass phase transition revealed
by a maximum in the ρ(T ) curve. Around this maximum,
data can be fitted using Vavilov-Glazman’s law for the low
temperature (spin-glass) part of the curve and Hamann’s
law for the high temperature (Kondo) part. Moreover, the
Tsg and TK extracted from these fits are in agreement with
those obtained by magnetic measurements on macroscopic
samples. These results are validated by magnetoresistance
measurements in which both the dominance of Kondo physics
and spin-glass behavior are observed in different range of
temperature. This, combined with the recent observation of
irreversibility in the resistivity of mesoscopic samples and
numerical simulations, proves that, even in such small samples, a spin glass phase transition appears when the temperature is lower than the typical strength of the spin-spin
interaction.
[1] Les Houches summer school, Session LXXXV, edited by J.-P.
Bouchaud, M. Mézard, and J. Dalibard (Elsevier, Amsterdam,
2006).
[2] C. Kittel, Introduction to Solid State Physics (John Wiley &
Sons, New York, 2005).
[3] M. Ali, P. Adie, C. H. Marrows, D. GreigBryan, and J. Hickey,
Nat. Mater. 6, 70 (2007).
[4] É. Vincent, J. Hammann, and M. Ocio, J. Stat. Phys. 135, 1105
(2009).
[5] K. H. Fisher and J. A. Hertz, in Spin Glasses: An Experimental Introduction (Taylor & Francis Ltd, London,
1993).
[6] J. A. Mydosh, in Spin Glasses (Cambridge University Press,
Cambridge, 1991).
[7] K. B. Storey and J. M. Storey, Sci. Am. 263, 62 (1990).
[8] D. Sornette, A. Johansen, and J. P. Bouchaud, J. Phys. I Fr. 6,
167 (1996).
[9] U. Krey, arXiv:0812.3378.
[10] A. K. Hartmann, arXiv:1312.1839. See http://www.compphys.
uni-oldenburg.de/en//61304.html.
[11] R. Rammal, G. Toulouse, and M. A. Virasoro, Rev. Mod. Phys.
58, 765 (1986).
[12] G. Frossati, J. L. Tholence, D. Thoulouze, and R. Tournier,
Physica B&C 84, 33 (1976).
[13] D. Sherrington and B. W. Southern, J. Phys. F: Met. Phys. 5,
L49 (1975).
[14] S. Nagata, P. H. Keesom, and H. R. Harrison, Phys. Rev. B 19,
1633 (1979).
[15] O. Laborde and P. Radhakrishna, J. Phys. F: Met. Phys. 3, 1731
(1973).
[16] O. Laborde, B. Loegel, and Radhakrishna, J. Phys. Colloques
35, C4-247 (1974).
[17] P. J. Ford and J. A. Mydosh, Phys. Rev. B 14, 2057 (1976).
[18] U. Larsen, Phys. Rev. B 14, 4356 (1976).
[19] M.-T. Béal-Monod and R. A. Weiner, Phys. Rev. 170, 552
(1968).
[20] H. Vloeberghs, J. Vranken, V. Vanhaesendonck, and Y.
Bruynseraede, Europhys. Lett. 12, 557 (1990).
[21] N. Nagaosa, J. Sinova, S. Onoda, A. H. MacDonald, and N. P.
Ong, Rev. Mod. Phys. 82, 1539 (2010).
[22] T. Taniguchi, K. Yamanaka, H. Sumioka, T. Yamazaki, Y.
Tabata, and S. Kawarazaki, Phys. Rev. Lett. 93, 246605 (2004).
[23] M. B. Weissman, Rev. Mod. Phys. 65, 829 (1993).
[24] T. Capron, A. Perrat-Mabilon, C. Peaucelle, T. Meunier, D.
Carpentier, L. P. Lévy, C. Baüerle, and L. Saminadayar,
Europhys. Lett. 93, 27001 (2011).
[25] L. R. Walker and R. E. Walstedt, Phys. Rev. B 22, 3816 (1980).
[26] K. Binder and A. P. Young, Rev. Mod. Phys. 58, 801 (1986).
[27] G. Parisi, Physica A 386, 611 (2007).
[28] R. Alvarez Baños, A. Cruz, L. A. Fernandez, J. M. GilNarvion, A. Gordillo-Guerrero, M. Guidetti, A. Maiorano, F.
Mantovani, E. Marinari, V. Martin-Mayor, J. Monforte-Garcia,
A. Muñoz Sudupe, D. Navarro, G. Parisi, S. Perez-Gaviro,
J. J. Ruiz-Lorenzo, S. F. Schifano, B. Seoane, A. Tarancon,
R. Tripiccione, and D. Yllanes (JANUS collaboration), J. Stat.
Mech. (2010) P06026.
[29] G. Neuttiens, J. Eom, C. Strunk, V. Chandrasekhar, C. Van
Haesendonck, and Y. Bruynseraede, Europhys. Lett. 34, 617
(1996).
[30] T. Capron, G. Forestier, A. Perrat-Mabilon, C. Peaucelle,
T. Meunier, C. Baüerle, L. P. Lévy, D. Carpentier, and L.
Saminadayar, Phys. Rev. Lett. 111, 187203 (2013).
[31] J. F. Ziegler and J. P. Biersack, http://www.srim.org/.
[32] B. Huard, H. Pothier, D. Estèeve, and K. E. Nagaev, Phys. Rev.
B 76, 165426 (2007).
[33] A. B. Zorin, Rev. Sci. Instrum. 66, 4296 (1995).
[34] D. C. Glattli, P. Jacques, A. Kumar, P. Pari, and L. Saminadayar,
J. Appl. Phys. 81, 7350 (1997).
[35] S. Mandal, T. Baütze, R. Blinder, T. Meunier, L. Saminadayar,
and C. Baüerle, Rev. Sci. Instrum. 82, 024704 (2011).
[36] É. Akkermans and G. Montambaux, Mesoscopic Physics of
Electrons and Photons (Cambridge University Press, Cambridge, 2007).
[37] F. Mallet, J. Ericsson, D. Mailly, S. Unlübayir, D. Reuter, A.
Melnikov, A. D. Wieck, T. Micklitz, A. Rosch, T. A. Costi,
L. Saminadayar, and C. Baüerle, Phys. Rev. Lett. 97, 226804
(2006).
[38] A. C. Hewson, The Kondo Problem to Heavy Fermions, Cambridge Studies in Magnetism (Cambridge University Press,
Cambridge, 1997).
We are indebted to H. Pothier and H. Bouchiat for the use
of their Joule evaporators. We thank T. Costi, L. Glazman, E.
Orignac, H. Bouchiat, É. Vincent, A. Rosch, and H. Alloul for
fruitful discussions.
024206-6
SPIN-GLASS PHASE TRANSITION REVEALED IN …
PHYSICAL REVIEW B 102, 024206 (2020)
[39] D. R. Hamann, Phys. Rev. 158, 570 (1967).
[40] C. Rizzuto, Rep. Prog. Phys. 37, 147 (1974).
[41] M. G. Vavilov and L. I. Glazman, Phys. Rev. B 67, 115310
(2003).
[42] M. G. Vavilov and L. I. Glazman, A. I. Larkin, Phys. Rev. B 68,
075119 (2003).
[43] P. Monod, Phys. Rev. Lett. 19, 1113 (1967).
[44] F. Schopfer, C. Baüerle, W. Rabaud, and L. Saminadayar, Phys.
Rev. Lett. 90, 056801 (2003).
[45] P. G. N. de Vegvar, L. P. Lévy, and T. A. Fulton, Phys. Rev. Lett.
66, 2380 (1991).
[46] D. Petit, L. Fruchter, and I. A. Campbell, Phys. Rev. Lett. 88,
207206 (2002).
[47] M. Solana et al. (in preparation).
024206-7