Emergence of ferromagnetism through the metal-insulator transition in undoped indium tin oxide films

We present a detailed study of the emergence of bulk ferromagnetism in low-carrier-density samples of undoped indium tin oxide. We use annealing to increase the density of oxygen vacancies and change the sample morphology without introducing impurities through the metal insulator transition (MIT). We utilize a novel and highly sensitive “Corbino-disk torque magnetometry” technique to simultaneously measure the thermodynamic and transport effects of magnetism on the same sample after successive annealing. With increased sample granularity, the carrier density increases, the sample becomes more metallic, and ferromagnetism appears as the resistance approaches the MIT. Ferromagnetism is observed through the detection of magnetization hysteresis, the anomalous Hall effect (AHE), and hysteretic magnetoresistance. A sign change of the AHE as the MIT is approached may elucidate the interplay between the impurity band and the conduction band on the weakly insulating side of the MIT.

Figure 1

(a) Schematic of torque magnetometry where Hall currents in a Corbino disk induce the magnetic moment $\mu$; adapted from [32]. (b) Picture of the ITO Corbino disk cantilever. The center contact is connected to an underlying ground plane. The metal pad on the end of the cantilever is used for optical alignment.

Figure 2

An example of extracting the even component of $\delta f_0(\pm V,B)$ and ${\sigma }_{xy}$ for nonmagnetic metallic ITO; adapted from [32]. (a) A full data set of $\delta f_0(\pm V,B)$ including the linear in $B$ term in $\delta f_0(\pm V,B)$ due to misalignment. (b) The even component of $\delta f_0(\pm V,B)$ found by averaging $\delta f_0(\pm V,-B)$ and $\delta f_0(\pm V,+B)$. (c) Extracted Hall conductivity from the even component of $\delta f_0(\pm V,B)$ and Eq. (6).

Figure 3

Characterization of the granularity of ITO with annealing. The initially large patches of ITO or smooth ITO break down into granular areas after annealing. (a, b) Atomic force microscopy profile of the ITO Hall bar before and after annealing. Sample height fluctuations and granularity increase after annealing. (c, d) SEM images of the ITO Hall bar before and after annealing. (e, f) SEM images of the Corbino disk ITO before and after annealing.

Figure 4

Behavior of ${\sigma }_{xy}$ and ${\sigma }_{xy}{\rho }_{xx}$ at 5 T measured on a series of annealed ITO samples through the MIT. The outlying sample very close to the MIT may be strongly influenced by the magnetic behavior as discussed in Secs. 3b2 and 3b3. Additional data obtained through a Hall bar approach and discussed in Sec. 3c are included as a reference. Resistive and nonmagnetic points above the MIT are highlighted in (b).

Figure 5

Hysteresis in magnetization $\mathrm{\Delta }m$ for three annealed ITO resistivities. We calculate the hysteretic magnetic moment per “formula unit” (f.u.), approximated as the unit cell volume of crystalline ${\mathrm{In}}_2{\mathrm{O}}_3$. The amplitude of the hysteresis and saturating field increase with longer annealing as reported in Table 2. Ramping of the magnetic field to $\pm 7$ T was performed in 0.7-T steps.

Figure 6

Fits to the even component of $\delta f_0(\pm V,B)$ assuming a sigmoidal kink in the anomalous Hall signal at $B_s$. For comparison, $B_s$ as fit from the closing of the $\mathrm{\Delta }{f}_0$ hysteresis in Fig. 5 is shown by the vertical line. Low field cutoffs are imposed to avoid obscuring data with high error-bar points.

Figure 7

Hysteresis in transport or $\delta f_0(\pm V,B)$ for 32- and 17-$\mathrm{k}\mathrm{\Omega }/\square$ annealed devices. Peaks near the closing of the hysteresis loops are highlighted in boxes. The Corbino disk voltage was $\pm 0.3$ V. Hysteresis could appear due to both the magnetoresistance for $B\lesssim B_s$ and the anomalous Hall effect as $B\approx B_s$.

Figure 8

(a) Magnetoresistance for a lightly annealed and a more strongly annealed ITO sample. (b) Hysteresis in magnetoresistance after two stages of annealing for a vertically and ${20}^{\circ }$ mounted Hall bar.

Figure 9

Comparison of the grain sizes between annealed and nonannealed ITO samples by SEM. Fluctuations in SEM brightness changed from local noise to distinct larger grains with annealing.

Figure 10

Voltage-independent $\mathrm{\Delta }{f}_0$ for four cooldowns. There is a rapid increase in $\mathrm{\Delta }{f}_0$ in the yellow-shaded region of $B$ near 0 T seen in all cooldowns regardless of the sample magnetization. The low-field $\mathrm{\Delta }{f}_0$ is therefore not attributed to the ITO and is not fit in Fig. 5.