Magnetization and antiferromagnetic coupling of the interface between a 20 nm ${\mathrm{Y}}_3{\mathrm{Fe}}_5{\mathrm{O}}_{12}$ film and ${\mathrm{Gd}}_3{\mathrm{Ga}}_5{\mathrm{O}}_{12}$ substrate

We present evidence for and characterization of a $\approx 4\text{−}\mathrm{nm}$-thick ${\left({\mathrm{Y}}_{1-x}{\mathrm{Gd}}_x\right)}_3{\mathrm{Fe}}_5{\mathrm{O}}_{12}$ layer with $x\ge 0.4$ formed at the interface between a gadolinium gallium garnet (GGG) substrate and a sputtered ${\mathrm{Y}}_3{\mathrm{Fe}}_5{\mathrm{O}}_{12}$ (YIG) epitaxial film with nominal thickness of 20 nm. Temperature-dependent polarized neutron reflectometry (PNR) and superconducting quantum interference device (SQUID) magnetometry show antiferromagnetic alignment of this interfacial layer with the bulk of the YIG film at low $T$ that persists to at least 3 tesla. These experiments also show that this interfacial alignment switches from antiparallel to parallel between 100 and 200 K in small applied magnetic fields. Simple modeling suggests correlation of this crossover with the ferrimagnetic compensation point of the mixed garnet that forms this interfacial layer.

Figure 1

(a) Polarized neutron reflectivity data measured at 6 K and 15 mT for both spin analyzed nonspin-flip specular reflectivities, $++$ (blue symbols), and $--$ (red symbols), with fits shown as solid lines. Inset: Schematic view of the sample showing GGG substrate, (Gd-Y)IG interfacial region, and YIG film, with magnetization corresponding to conditions of this measurement shown schematically with black arrows. (b) Nuclear ${\rho }_{\mathrm{N}}$, and magnetic ${\rho }_{\mathrm{M}}$, scattering length-density profiles determined from fits to the data.

Figure 2

PNR spin asymmetry [(a), (b)], magnetization profiles determined from model fitting [(c), (d)], and resulting $M$ vs $T$ measured for ${\mu }_{0}H=15\mathrm{mT}$ (left) and ${\mu }_{0}H=3\mathrm{T}$ (right). The three magnetic components of the sample are shown in background shading in panels (c) and (d), and their $M$ separately plotted in (e) and (f). We provide bulk values of $M\left(T\right)$ for YIG for comparison in panel (e).

Figure 3

PNR spin asymmetry [(a),(b)], modeled magnetization profiles [(c),(d)], and resulting $M$ vs ${\mu }_{0}H$ at fixed $T=7\mathrm{K}$ (e) and $M$ vs $T$ at fixed ${\mu }_{0}H=2\mathrm{T}$ (f). Solid lines in (e) and (f) are guides to the eye. Possible nonmonotonic behavior of the (Gd-Y)IG layer could result from frustrated domains in this interfacial layer. Note however that the negative $M$ at all ${\mu }_{0}H$ for the (Gd-Y)IG indicates very strong antiferromagnetic exchange at the interface.

Figure 4

(a) $\chi =M/H$ vs $T$ for the bare GGG substrate in ${\mu }_{0}H=10\mathrm{mT}$ vs $T$ from 2 K to 300 K. Inset: $1/\chi$ vs $T$. In both the red line is a fit to the Curie-Weiss law with a small negative $\theta$, as expected for GGG. (b) $m$ vs ${\mu }_{0}H$ at 300 K for the GGG substrate and the YIG/(Gd-Y)IG/GGG sample. At this $T$ both the large paramagnetic contribution from the GGG and the essentially ferromagnetic contribution from the YIG are obvious even without background subtraction. (c) $m$ vs ${\mu }_{0}H$ at 10 K, where the paramagnetic background dominates the ferromagnetic signal. The different slopes are due to a slight difference in the substrate volume.

Figure 5

$M$ vs ${\mu }_{0}H$ loops for the YIG/(Gd-Y)IG/GGG sample after subtraction of the matched GGG background for a series of temperatures: (a) 300 K, (b) 200 K, (c) 100 K, (d) 70 K, (e) 50 K, (f) 40 K, (g) 20 K, (h) 10 K, and (i) 2 K. All loops show the expected FM hysteresis and positive remanence, with $H_{\mathrm{c}}$ growing as $T$ drops. Neither the small shift on the ${\mu }_{0}H$ axis nor the shift on the $M$ axis can be clearly linked to the YIG/(Gd-Y)IG, as discussed in the main text.

Figure 6

(a) $M_{\mathrm{s}}$ vs $T$ for the YIG/(Gd-Y)IG films. The blue data points show $M_{\mathrm{s}}$ with volume calculated for the entire nominal film thickness $t_{\mathrm{nom}}=20\mathrm{nm}$. $M_{\mathrm{s}}$ falls below bulk YIG values (dashed-red line) at all $T$. The grey data points show $M_{\mathrm{s}}$ when a simple linear background is assumed, instead of using the measured GGG values. Inset: ${\mu }_0{H}_{\mathrm{c}}$ vs $T$ shows a monotonically decreasing trend. (b) Measured moment $m$ of the YIG/(Gd-Y)IG layer (blue) compared to calculated values for a 15.9-nm-thick layer of bulk YIG (dashed-red line), 3.6-nm-thick layers of ${\left({\mathrm{Gd}}_x{\mathrm{Y}}_{1-x}\right)}_3{\mathrm{Fe}}_5{\mathrm{O}}_{12}$ with $x=1$ (GdIG, dark-yellow line), and $x=0.6$ (green line), and two models subtracting these two contributions to reflect the antiferromagnetic alignment (brown and darker green lines).

Figure 7

Estimated moment of the (Gd-Y)IG interface layer, $-m_{\mathrm{int}}$ vs $T$ from SQUID magnetometry (blue symbols), and PNR (red circle) compared to bulk values of ${\left({\mathrm{Gd}}_x{\mathrm{Y}}_{1-x}\right)}_3{\mathrm{Fe}}_5{\mathrm{O}}_{12}$ for various $x$ (dashed lines). For the $x=0.4$ and $x=0.6$ mixed garnets we also show the moment with reversed sign above $T_{\mathrm{comp}}$, which roughly brackets the measured $-m_{\mathrm{int}}$ from both PNR and SQUID at high $T$. At low $T$, $-m_{\mathrm{int}}$ again falls roughly between the values for $x=0.4$ and $x=0.6$, suggesting that the interface layer has a high Gd content, though is most likely not purely bulk-like GdIG.