Polishchuk et al.

Nanoscale Research Letters (2018) 13:245

https://doi.org/10.1186/s11671-018-2669-0

NANO EXPRESS Open Access

Magnetic Hysteresis in Nanostructures with

Thermally Controlled RKKY Coupling
Dmytro Polishchuk1,2, Yuliya Tykhonenko-Polishchuk1,2, Vladyslav Borynskyi2, Anatolii Kravets1,2* ,
Alexandr Tovstolytkin2 and Vladislav Korenivski1

Abstract
Mechanisms of the recently demonstrated ex-situ thermal control of the indirect exchange coupling in magnetic
multilayer are discussed for different designs of the spacer layer. Temperature-induced changes in the hysteresis of
magnetization are shown to be associated with different types of competing interlayer exchange interactions.
Theoretical analysis indicates that the measured step-like shape and hysteresis of the magnetization loops is due to
local in-plane magnetic anisotropy of nano-crystallites within the strongly ferromagnetic films. Comparison of the
experiment and theory is used to contrast the mechanisms of the magnetization switching based on the competition
of (i) indirect (RKKY) and direct (non-RKKY) interlayer exchange interactions as well as (ii) indirect ferromagnetic and
indirect antiferromagnetic (both of RKKY type) interlayer exchange. These results, detailing the rich magnetic phase
space of the system, should help enable the practical use of RKKY for thermally switching the magnetization in
magnetic multilayers.
Keywords: Magnetic multilayers, Indirect exchange coupling, Magnetization switching, Magnetic coercivity, Thermo-
magnetic effects

Background We recently demonstrated [8, 9] a new mechanism of

The important discoveries of the indirect exchange ex-situ thermal control of the interlayer RKKY coupling
coupling (IEC) [1] of Ruderman-Kittel-Kasuya-Yosida in magnetic multilayers. The idea is based on the use of
(RKKY) type and of the giant magnetoresistance effect a diluted ferromagnetic alloy with relatively low Curie
[2] have generated a great deal of new basic physics re- temperature (TC*) instead of the nonmagnetic spacer
sults as well as numerous applications [3]. The discov- between strongly ferromagnetic (FM) layers. In the ini-
ered IEC oscillates in magnitude and sign versus the tial design, the Cr spacer in a classical RKKY trilayer
separation of the individual ferromagnetic layers in a Fe/Cr/Fe is replaced with the diluted alloy FexCr100 − x
metallic stack, yielding either parallel (P) or antiparallel (Fig. 1a, b). When the spacer is paramagnetic (PM) at
(AP) magnetic ground states. This RKKY-type inter- T > TC* (Fig. 1a), the trilayers exhibit an antiparallel
action is almost independent of temperature [4, 5] and alignment of the Fe moments due to the antiferromag-
largely insensitive to any other external control post-fab- netic (AFM) indirect exchange coupling (RKKY). The
rication, which limits the use of the effect. Recent at- parallel alignment is enforced by the direct exchange
tempts to enhance the effect of temperature on RKKY coupling when the spacer is FM (T < TC*) (Fig. 1b). As
and use it to control the IEC in Tb/Y/Gd [6] and Co/Pt temperature is varied, these trilayers demonstrate a
[7] multilayers report relatively weak RKKY without dir- parallel-to-antiparallel magnetization switching, with a
ect parallel-to-antiparallel (P-to-AP) thermal switching, rather broad transition of ~ 100 K due to the magnetic
with broad thermal transitions (~ 100 K). proximity effect [10]. In contrast to the trilayers with
the uniform spacer, trilayers with non-uniform, compos-
ite spacers demonstrate a significantly enhanced per-
* Correspondence: anatolii@kth.se
1
Nanostructure Physics, Royal Institute of Technology, 10691 Stockholm,
formance with the thermo-magnetic transition widths
Sweden down to ~ 10 K. Moreover, by tailoring the spacer prop-
2
Institute of Magnetism, NAS of Ukraine and MES of Ukraine, 03142 Kyiv, erties, either an antiparallel (Fig. 1c) or parallel ground
Ukraine

© The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made.
Polishchuk et al. Nanoscale Research Letters (2018) 13:245 Page 2 of 7

Fig. 1 Illustration of magnetic layout of Fe/uniform-spacer/Fe multilayers when the spacer is paramagnetic (PM) (a) or ferromagnetic (FM) (b). c, e
Structures with modified, composite spacers sp1 and sp2 exhibit, respectively, antiparallel and parallel magnetic ground state at low temperature
(T < TC*). d, f Corresponding characteristic temperature variation of remanent magnetization of structures with spacers sp1 and sp2 for different
compositions of spacers’ inner diluted alloy layer. Layer thicknesses are given in parentheses in “nanometers”

state (Fig. 1e) can be obtained at T < TC*. On heating

above TC*, the Fe moments reverse their mutual orien-
tation either into parallel for Fe/sp1/Fe (Fig. 1d) or
into antiparallel for Fe/sp2/Fe (Fig. 1f ). The narrow
thermal transition and the ability to choose the
magnetic regime (P/AP) as well as the operating
temperature interval are all important advantages in
terms of practical implications.
Antiferromagnetic interlayer coupling in our multilayers
is clearly manifested as zero remnant magnetization,
reflecting the antiparallel alignment of the Fe layers’ mo-
ments. Besides the zero remanence, the magnetization
curves M(H) are characterized by a step-like approach to
saturation and hysteresis on reversing the field sweep
(Fig. 2a). The bi-linear exchange coupling model gives
M(H) as a line approaching saturation at the effective field
of the indirect exchange, HJ. A step-like character of the
saturation occurs due to in-plane magnetic anisotropy of
the structure’s ferromagnetic layers, resulting in one step
for the easy-axis anisotropy [11], and two sequential steps
for the four-fold anisotropy [12]. However, our angle-
dependent magnetometric and magnetoresonance studies
of the multilayers reveal no macroscopic in-plane mag-
netic anisotropy. The latter fact warrants a more compre-
hensive analysis of the experimental results, accompanied
by model simulations. In the following, such comprehen-
Fig. 2 a Typical in-plane magnetization curve, M(H), measured by
sive approach is used to contrast the mechanisms of the
MOKE for Fe/sp2/Fe multilayers with antiferromagnetic interlayer
magnetization switching for the two key multilayer de- coupling. Curved arrows show direction of field sweep; horizontal
signs—with uniform (Fig. 1a, b) as against composite spa- arrows denote mutual alignment of Fe magnetic moments. b MOKE
cer layers (Fig. 1c, e). M(H) loops for reference Fe(2)/Cr(10) (bottom Fe) and Cr(10)/Fe(2)
We point out the importance of understanding the (top Fe) bilayers. c Reference-frame schematic of in-plane M1, M2,
and H, with respect to easy axis of twofold magnetic anisotropy of
mechanisms involved in the interlayer exchange in a given
a nano-crystallite
system. The pioneering work on RKKY in multilayers [13]
Polishchuk et al. Nanoscale Research Letters (2018) 13:245 Page 3 of 7

and its extensions to, e.g., bi-quadratic exchange [14, 15] distributed across all in-plane angles (the films were de-
set off a major development in physics and technology posited under in-plane rotation). These assumptions
known as spintronics. The RKKY in the original form, are reasonable for the studied system and produced the
however, is not used today due to the lack of a suitable best fit to the measured M(H) data at various tempera-
switching mechanism, but often plays an assisting role in tures as discussed below.
devices for, e.g., flux-closing reference layers. In this work, The free energy density for our F1/NM/F2 system can
we study such a primary RKKY-switching mechanism and, then be written as
more specifically, analyze the interplay among the interac-
tions leading to thermal on/off switching of RKYY, which
U ¼ UH þ Ua þ U J ¼
in turn controls the efficiency of the P/AP switching of
¼ −MH ½ cosðφ1 −φH Þ þ cosðφ2 −φH Þ

the magnetization of the nanostructure. Based on this
− 1=2MH a1 cos2 φ1 þ 1=2MH a2 cos2 φ2 þ
analysis, we are able to make conclusions about and rec-
þ1=2MH J cosðφ1 −φ2 Þ;
ommendations for optimizing the switching performance
of the Curie-RKKY nanodevices. ð1Þ
where UH, Ua and UJ are, respectively, the Zeeman en-
Methods ergy of the FM layers in field H = (H, φH), uniaxial an-
In this work, we analyze two series of samples: (1) Fe(2)/ isotropy energy, and the interlayer coupling energy of
sp1(x = 30–40 at.%)/Fe(2), where sp1 = N/f/N/f/N, N = bi-linear type [16, 17]. The magnetic moments of the
Cr(1.5), f = Fe(0.25)/FexCr100 − x(3)/Fe(0.25) (Fig. 1c), and FM layers, M1 = (M, φ1) and M2 = (M, φ2), are of the
(2) Fe(2)/sp2(x = 10–20 at.%)/Fe(2), where sp2 = N/f/N, same magnitude, as illustrated in Fig. 2c. Ha1,2 and HJ
N = Cr(dCr), f = FexCr100 − x(d), dtot = (2dCr + d) = 1.5 nm are the effective fields of the uniaxial (twofold) anisot-
(Fig. 1e). Additionally, a number of reference films and ropy and the bi-linear interlayer coupling, respectively.
bi-layers were deposited. The thicknesses in parentheses Conversion to angular variables φm = (φ1 + φ2)/2 and φd
are in “nanometers”. The multilayers were deposited at = (φ1 − φ2) simplifies the expression for the magnetic
room temperature onto Ar pre-etched undoped Si (100) free energy of the system to
substrates using a dc magnetron sputtering system.
Layers of diluted FexCr100 − x binary alloys of varied com-
position were deposited using co-sputtering from separ- U ¼ −2MH cosðφm −φH Þ cosðφd =2Þ
ate Fe and Cr targets. Additional details on the −1=2M½H a1 cos2 ðφm þ δ=2Þ
multilayer fabrication can be found elsewhere [8, 9]. þH a2 cos2 ðφm −δ=2Þ þ 1=2MH J cosφd :
The in-plane magnetic characterization was carried ð2Þ
out using a vibrating-sample magnetometer (VSM)
equipped with a high-temperature furnace (Lakeshore In the following simulations, the magnetization curves,
Inc.) in the temperature range of 295–400 K, and a M(H), are obtained by finding parameters φm and φd,
magneto-optical Kerr effect (MOKE) magnetometer which correspond to the minimum of U in (2) for given
equipped with an optical cryostat (Oxford Instr.) in the φH, H1a, H2a, and HJ, according to
temperature range of 77–450 K. Additionally, ferromag-
netic resonance (FMR) measurements were performed M=M s ¼ ½ cosðφ1 −φH Þ þ cosðφ2 −φH Þ=2
room-temperature using an X-band Bruker ELEXYS ¼ cosðφm −φH Þ cosðφd =2Þ: ð3Þ
E500 spectrometer equipped with an automatic goniom-
eter to measure the in-plane-angle dependence of the
magnetic resonance spectra. Coercivity of Magnetization
The measured M(H) for the structures with AFM ex-
Results and Discussion change coupling are of a step-like shape, with well-de-
Phenomenology of Indirect Exchange Coupling fined coercivity for the reversing field sweep (Fig. 3a).
A phenomenological magnetostatic model used for sim- The above phenomenological model is used to analyze
ulations of magnetization curves for trilayer F1/NM/F2, both the magnetic properties of the ferromagnetic
where F1 and F2 are ferromagnetic layers and NM is layers Fe(2 nm) and the thermally induced magnetic
nonmagnetic spacer, has the following assumptions. transition in the composite spacers, which mediates the
First, the magnetic field is applied in the plane of the interlayer coupling.
films, which corresponds to our experiment and simpli- Epitaxial (100) Fe-based multilayers grown on single-
fies the calculations. Second, the individual grains in crystal substrates are usually characterized by fourfold
the polycrystalline films are characterized by twofold in-plane magnetic anisotropy [12], while substrates of
in-plane anisotropy with the easy axes uniformly other texture [e.g., (211)] can result in twofold anisotropy
Polishchuk et al. Nanoscale Research Letters (2018) 13:245 Page 4 of 7

film plane can result from the deposition on rotating sub-

strates in the case of our samples. Such pattern of mag-
netic anisotropy can then be explained in terms of a
polycrystalline nature of the sputtered multilayers and
in-plane strain variations between the nano-crystalline
grains [18].
M(H) curves for the model system F1/NM/F2, simu-
lated for different strengths of the AFM interlayer ex-
change coupling (effective field HJ) and shown in Fig. 3b,
exhibit all the key features found in the experimental
curves (Fig. 3a). M(H) for Fe/sp1(x = 35 at.%)/Fe under-
goes a significant change with increasing temperature.
The changes are due to the weakening of the interlayer
coupling, which can be directly compared the simulated
M(H) shown in Fig. 3b. All of the changes seen in the
experimental M(H) data, including the enhancement of
the coercivity as the interlayer coupling is weakened, are
in correlate very well with the simulated behavior, which
validates the model. One should note that the model cal-
culations are performed without taking into account the
effect of temperature directly (only via effectively re-
duced HJ), which should reduce magnetic coercivity of
the individual layers. This is the likely cause for some-
what smaller coercivity on the experiment.
The simulated M(H) curves shown in Fig. 3b are ob-
tained by averaging the M(H) calculated for different an-
gles φH between the external field H and the easy axis of
the uniaxial magnetic anisotropy. Figure 3c shows the
curves at selected angles φH for the case HJ/Haav = 2.
Fig. 3 a Measured M(H) curves for a sample from series, Fe/sp1(x = Here, Haav = (Ha1 + Ha2)/2, where Ha1 and Ha2 are the ef-
15%)/Fe, for different temperatures. b Corresponding simulated M(H) fective fields of the in-plane uniaxial anisotropy acting in
curves for model F1/NM/F2 trilayer for different strengths of effective
the F1 and F2 layers, respectively. Ratio Ha1/Ha2 = 0.7,
field HJ of indirect exchange coupling. (Haav = (Ha1 + Ha2)/2, where
Ha1and Ha2 are anisotropy fields of layers F1 and F2. c M(H) curves used in the calculation, corresponds to the value ob-
simulated for selected angles φH, for HJ/Haav = 2. d Transformation tained experimentally (Fig. 2b). The step-like shape and
of local minima of free energy (2) as a function of applied field H, coercivity are well-defined for φH < 60°. As mentioned
for case HJ/Haav = 2 and φH = 15°. Blue lines trace the path connecting above, additional VSM and FMR studies of the reference
energy minima for different φm (φd). Front surface of energy surface is
Fe(2 nm) films and Fe/Cr/Fe trilayers did not reveal any
transparent for visual clarity of illustration
significant in-plane angular dependence in the hysteresis
loops or the resonance spectra. Since VSM and FMR
[11]. The main difference in M(H) between the two cases measure the integral properties of the samples, we con-
is in the presence of two characteristic steps in M vs. H clude that essentially no macroscopic in-plane magnetic
when the anisotropy is fourfold and only one M-vs-H step anisotropy is present. On the other hand, the observed
when it is twofold. Our VSM and FMR studies of the ref- coercivity can be attributed only to an in-plane magnetic
erence Fe(2 nm) films and Fe/Cr/Fe tri-layers (data not anisotropy. Additionally, the shape of the experimental
shown) did not reveal any significant in-plane angular de- M(H) curves is closer to the calculated curves obtained
pendence in the hysteresis loops or resonance spectra, by averaging rather than to any individual curve for a se-
leading us to conclude that essentially no macroscopic lect φH. Therefore, taking into account the polycrystal-
in-plane magnetic anisotropy is present. On the other line nature of our sputtered multilayers, one can
hand, the numerical analysis described above concludes conclude that the Fe(2 nm) layers have a uniform angu-
that the measured one-step-shaped M(H) loops for the lar distribution of the local anisotropy easy axes in the
RKKY-coupled Fe/Cr/Fe trilayers must be due to twofold film plane.
magnetic anisotropy on the scale of the individual crystal- Figure 3d illustrates how the energy U(φm, φd) of Eq. 2
lites forming the polycrystalline films. The uniform angu- changes in response to H. We, again, take HJ/Haav = 2
lar distribution of the local anisotropy easy axes in the and φH = 15°, which corresponds to the second curve in
Polishchuk et al. Nanoscale Research Letters (2018) 13:245 Page 5 of 7

panel (c). The solid thick line in Fig. 3d traces the path
connecting the energy minima for different φm (φd). The
local energy minima are well-defined within this
minimum-value path. The minimum at low field corre-
sponds to the antiparallel orientation of the Fe moments
(φm ≈ 90°, φd ≈ 180°). With increasing H, a second local
energy minimum emerges and deepens, while the first
minimum become shallower and eventually disappears.
This single-minimum state corresponds to the parallel
orientation of the Fe moments (φm ≈ φH, φd ≈ 0°). On
subsequently decreasing H, the system initially is in the
second minimum (parallel magnetic state) until it disap-
pears at lower H and the system ends up in the first en-
ergy minimum (antiparallel state).

Competition Between Direct and Indirect Exchange

Coupling: Temperature Dependence of Magnetic Coercivity
While the first series of trilayers Fe/sp1/Fe exhibits a
thermally induced transition from the low-temperature
AFM interlayer coupling into the high-temperature
decoupled state, the second series shows a transition
from the low-temperature FM to the high-temperature
AFM coupling. For the FM-to-AFM thermal transition
in the second case, no external magnetic field is required
and the magnetization switching is fully reversible—a
key advantage for applications. Fig. 4 a Magnetization versus field measured by the MOKE for sample
Using the model validated through the above analysis of from second series, Fe/sp2(x = 15%)/Fe, for different temperatures. b
the first series of samples, we next focus on investigating Corresponding simulated M(H) curves for model F1/NM/F2 trilayer, for
the competition between the direct and indirect interlayer different effective field of indirect exchange coupling, HJ
exchange coupling in Fe/sp2*(x)/Fe, with uniform spacers
of type sp2* = FexCr100 − x(1.5 nm) and composite spacers
of type sp2* = Cr(dCr)/FexCr100 − x(d)/Cr(dCr), d + dCr = (HJ = 0). Strong FM interlayer coupling, however, always
1.5 nm (sp2* is a derivative from the thickness-fixed spa- results in a single M(H) loop.
cer sp2 = Cr(0.4)/FexCr100 − x(0.7)/Cr(0.4) of the second Coercivity of the partial loops (Hcpart) has a pro-
series). Figure 4 compares the experimental M(H) loops nounced temperature dependence for all samples and
for the structures with sp2 = Cr(0.4)/Fe15Cr85(0.7)/Cr(0.4) increases almost linearly with decreasing temperature.
[panel (a)] and the corresponding M(H) curves simulated Figure 5a shows the temperature dependence of the
with HJ chosen such as obtain the best fit to the experi- coercive field defined as the difference between the
ment. First to be noted is the high similarity between the fields of the two peaks on the magnetization deriva-
calculated loops and those measured, with all the key fea- tive, dM/dH vs H. The series with x = 15% contains
tures reproduced. Secondly, the experiment shows a samples with different thickness of the layers compos-
temperature-induced transition from the FM interlayer ing the spacer: d (dCr) = 3 (6), 7 (4), 9 (3), 11 (2), 15
coupling [low-temperature single-loop in Fig. 4a] to the (0) Å. The last sample [d (dCr) = 15 (0) Å] is the tri-
AFM coupling [high-temperature loop with zero reman- layer with a uniform spacer Fe15Cr85 (1.5 nm). The
ence in Fig. 4a]. The variation in the shape of the simu- samples with d ≤ 7 Å (dCr ≥ 4 Å) show a monotonous
lated loops for various effective coupling field values HJ increase in Hcpart with decreasing temperature. The
(Fig. 4b) additionally confirms the validity of the chosen coercivity of the samples with smaller dCr (< 4 Å)
phenomenological description. Same as in the previous begins to deviate from this slope right below the tran-
section, Ha1/Ha2 = 0.7 was used in the simulations. It sition temperature. The high-temperature part of
should be noted that, even though not the case here, the Hcpart(T), however, is on the general linear trend
step-like M(H) shape taken to be due to the AFM inter- [shown as a thick red line in Fig. 5a]. This linear
layer coupling (for example, loops at 300 K and HJ = slope in the coercive field versus temperature is asso-
0.5Haav) can in principle be caused by different coercive ciated mainly with the change in the intrinsic coerciv-
fields in F1 and F2 in the absence of interlayer coupling ity of the outer Fe(2 nm) layers.
Polishchuk et al. Nanoscale Research Letters (2018) 13:245 Page 6 of 7

Fig. 5 a Temperature dependence of coercivity of partial loops (Hcpart) for structures Fe/sp2(x = 15%)/Fe with different thickness of FexCr100 − x and Cr
layers (d and dCr, respectively) in spacer sp2. Red thick line is linear approximation of high-temperature part of Hcpart(T). b Temperature dependence of
coercivity normalized to linear background. c Coercivity vs. HJ obtained from simulated M(H) curves for two cases: (1) Ha1/Ha2 = 0.7 and (2) Ha1 = Ha2

In our previous work [9] the structures with the spacer indirect AFM exchange, at some temperature compen-
thickness of d ≤ 7 Å (dCr ≥ 4 Å) showed the sharpest sating it such that HJ = 0. This case is well described by
thermo-magnetic switching. We then suggested that the our model, where the F1 and F2 layers have different an-
reason for such narrowing of the magnetic transition isotropy fields [blue curve in Fig. 5c]. In contrast, the Fe
was switching off the direct exchange channel between layers in the structure with the composite spacer are FM
the outer Fe layers. On the other hand, the dependence coupled at low temperature sequentially through Fe/Cr/
of Hcpart* vs T (Fig. 5b), obtained by normalizing FeCr and FeCr/Cr/Fe, with the spacers’ FeCr inner layer
Hcpart(T) to the sloping intrinsic-coercivity background, is in FM state. Since this FeCr layer acts as an addition
shows a noticeable negative deviation only for the struc- exchange link, the spacer transmits exchange in such a
tures with thin Cr spacers (dCr < 4 Å) and essentially no way as to effectively equalize the coercivity of the outer
deviation for dCr ≥ 4 Å. The dependence for x = 20%, dCr Fe layers [black curve in Fig. 5c]. When the FeCr layer is
= 4 Å is shown for comparison because the transition in its paramagnetic state, the system behaves similar to
for x = 15%, dCr = 4 Å (TC* ≈ 140 K) is close to the lowest the one with the uniform spacer [high-temperature part
measurement temperature. The absence of a negative of the Hcpart* vs T dependence in Fig. 5b and the AFM
deviation on Hcpart* vs T for the structures with dCr ≥ 4 Å side of Hcsim vs T (HJ > 0) in Fig. 5c].
can serve as an addition confirmation that the direct
interlayer coupling is fully suppressed. Conclusions
To separate and analyze the part of the dependence In summary, we have described and compared two mech-
Hcpart(T), which is driven by changes in the strength and anisms of temperature-induced magnetization switching
sign of the interlayer coupling (HJ), the coercivity of the in multilayers with different types of interlayer exchange
simulated M(H) is plotted versus HJ in Fig. 5c. Thus, ob- mediating spacers. The switching mechanisms reflect the
tained Hcsim vs T depends on the ratio between the ef- competition of either the direct and indirect exchange
fective anisotropy fields of the F1 and F2 layers, Ha1/Ha2. coupling through a uniform spacer or the all-indirect ex-
The larger the deviation of Ha1/Ha2 from unity, the dee- change coupling of ferromagnetic and antiferromagnetic
per the minimum and bigger its offset from zero field on types through a composite spacer. The key element of the
the FM side of the diagram (HJ < 0). When the anisot- spacer design is the weakly magnetic diluted-alloy layer,
ropy fields are equal (Ha1/Ha2 = 1), the minimum is not the Curie transition of which is transformed into a P-AP
present. This behavior is similar to the difference be- magnetization switching in the structure. Our measured
tween Hcpart*(T) for the structures with uniform and data, supported by detailed theoretical simulations of the
composite spacers with large dCr (≥ 4 nm) [blue and magnetic hysteresis in the multilayer, are explained as due
black curves in Fig. 5b, respectively]. This indicates that to nanograins of uniaxial magnetic anisotropy with its easy
these two types of spacers transmit the interlayer coup- axes uniformly distributed in the plane of the outer ferro-
ling between the two outer Fe layers differently. In the magnetic layers. The temperature dependence of the mag-
uniform spacer, direct FM exchange competes with netic coercivity in the magnetic transition region has a
Polishchuk et al. Nanoscale Research Letters (2018) 13:245 Page 7 of 7

different form for different spacer designs. The specific be- 4. Parkin SSP, Bhadra R, Roche KP (1991) Oscillatory magnetic exchange
havior for the structure with the composite spacer is coupling through thin copper layers. Phys Rev Lett 66:2152–2155. https://
doi.org/10.1103/PhysRevLett.66.2152
found to be a result of the suppressed direct interlayer ex- 5. Toscano S, Briner B, Hopster H, Landolt M (1992) Exchange-coupling between
change channel, such that the relevant P-AP switching ferromagnets through a non-metallic amorphous spacer-layer. J Magn Magn
mechanism is a competition of indirect ferromagnetic and Mater 114:L6–L10. https://doi.org/10.1016/0304-8853(92)90326-J
6. Döbrich KM, Wietstruk M, Prieto JE et al (2008) Temperature-induced
indirect antiferromagnetic (both RKKY type) exchange. reversal of magnetic interlayer exchange coupling. Phys Rev Lett 100:
We thus have shown that the broken channel of direct 227203. https://doi.org/10.1103/PhysRevLett.100.227203
interlayer exchange within the spacer is correlated with 7. Liu ZY, Zhang F, Xu B et al (2009) Thermally induced antiferromagnetic
interlayer coupling and its oscillatory dependence on repetition number in
the sharper thermo-magnetic transition. We furthermore spin-valve Co/Pt multilayers. J Phys D Appl Phys 42:35010. https://doi.org/
have shown that that the thermally driven competition of 10.1088/0022-3727/42/3/035010
the purely indirect interlayer exchange, ferromagnetic 8. Polishchuk DM, Tykhonenko-Polishchuk YO, Kravets AF, Korenivski V (2017)
Thermal switching of indirect interlayer exchange in magnetic multilayers.
RKKY versus antiferromagnetic RKKY, where the proxim- EPL 118:37006. https://doi.org/10.1209/0295-5075/118/37006
ity effect in the spacer is out of action, leads to even better 9. Polishchuk DM, Tykhonenko-Polishchuk YO, Holmgren E et al (2017)
switching performance. These results should be important Thermally induced antiferromagnetic exchange in magnetic multilayers.
Phys Rev B 96:104427. https://doi.org/10.1103/PhysRevB.96.104427
for device applications of the Curie-RKKY nanostruc- 10. Kravets AF, Dzhezherya YI, Tovstolytkin AI et al (2014) Synthetic ferrimagnets
tures in spin-thermo-electronic devices [19, 20]. with thermomagnetic switching. Phys Rev B 90:104427. https://doi.org/10.
1103/PhysRevB.90.104427
Abbreviations 11. Grimsditch M, Kumar S, Fullerton EE (1996) Brillouin light scattering study of
AFM: Antiferromagnetic; AP: Antiparallel; FM: Ferromagnetic; FMR: Ferromagnetic Fe/Cr/Fe (211) and (100) trilayers. Phys Rev B 54:3385–3393. https://doi.org/
resonance; IEC: Indirect exchange coupling; MOKE: Magneto-optical Kerr effect; 10.1103/PhysRevB.54.3385
NM: Nonmagnetic; P: Parallel; PM: Paramagnetic; RKKY: Ruderman-Kittel-Kasuya- 12. Köbler U, Wagner K, Wiechers R et al (1992) Higher order interaction terms
Yosida; VSM: Vibrating-sample magnetometer in coupled Fe/Cr/Fe sandwich structures. J Magn Magn Mater 103:236–244.
https://doi.org/10.1016/0304-8853(92)90195-T
Funding 13. Grünberg P, Schreiber R, Pang Y et al (1986) Layered magnetic structures:
This work was partly financially supported by the Swedish Stiftelse Olle evidence for antiferromagnetic coupling of Fe layers across Cr interlayers.
Engkvist Byggmästare (Grant No. 2013-424/2018-589), the Swedish Research Phys Rev Lett 57:2442–2445. https://doi.org/10.1103/PhysRevLett.57.2442
Council (VR Grant No. 2014-4548), the Department of Targeted Training of 14. Slonczewski JC (1991) Fluctuation mechanism for biquadratic exchange
Taras Shevchenko National University of Kyiv at the National Academy of coupling in magnetic multilayers. Phys Rev Lett 67:3172–3175. https://doi.
Sciences of Ukraine (Grant No. 0117U006356), Volkswagen Foundation (Grant org/10.1103/PhysRevLett.67.3172
No. 90418), and the State Fund for Fundamental Research of Ukraine (Grant 15. Slonczewski JC (1995) Overview of interlayer exchange theory. J Magn
No. F76/34-2018). Magn Mater 150:13–24. https://doi.org/10.1016/0304-8853(95)00081-X
16. Bruno P (1995) Theory of interlayer magnetic coupling. Phys Rev B 52:411–
Availability of Data and Materials 439. https://doi.org/10.1103/PhysRevB.52.411
All data are fully available without restriction. 17. Stiles MD (1999) Interlayer exchange coupling. J Magn Magn Mater 200:
322–337. https://doi.org/10.1016/S0304-8853(99)00334-0
Authors’ Contributions 18. Johnson MT, Bloemen PJH, den Broeder FJA, de Vries JJ (1996) Magnetic
DP, YTP, AK, and VK developed the approaches to fabricate multilayers and anisotropy in metallic multilayers. Reports Prog Phys 59:1409–1458. https://
carried out main works on characterization of the samples. DP and YTP doi.org/10.1088/0034-4885/59/11/002
carried out temperature-dependent magnetic measurements on multilayers. 19. Gerrit E. W. Bauer, Eiji Saitoh, Bart J. van Wees (2012) Spin caloritronics. Nat
DP, VB, and AT analyzed the data of magnetic measurements and made Mater 11(5):391–399
calculations to extract magnetic parameters of the multilayers. All authors 20. A. M. Kadigrobov, S. Andersson, D. Radić, R. I. Shekhter, M. Jonson, V.
contributed to analysis of experimental data and writing manuscript. AK, AT, Korenivski, (2010) Thermoelectrical manipulation of nanomagnets. J Appl
and VK supervised the work and finalized the manuscript. All authors read Phys 107(12):123706
and approved the final manuscript. All authors have the appropriate permissions
and rights to the reported data. All authors have agreed to authorship and order
of authorship.

Competing Interests
The authors declare that they have no competing interests.

Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.

Received: 29 December 2017 Accepted: 15 August 2018

References
1. Grünberg P, Schreiber R, Pang Y et al (1987) Layered magnetic structures:
evidence for antiferromagnetic coupling of Fe layers across Cr interlayers. J
Appl Phys 61:3750–3752. https://doi.org/10.1063/1.338656
2. Baibich MN, Broto JM, Fert A et al (1988) Giant magnetoresistance of
(001)Fe/(001)Cr magnetic superlattices. Phys Rev Lett 61:2472–2475. https://
doi.org/10.1103/PhysRevLett.61.2472
3. Žutić I, Fabian J, Das SS (2004) Spintronics: fundamentals and applications.
Rev Mod Phys 76:323–410. https://doi.org/10.1103/RevModPhys.76.323