Paramagnetic $\mathrm{GaN}:\mathrm{Fe}$ and ferromagnetic $(\mathrm{Ga},\mathrm{Fe})\mathrm{N}$: The relationship between structural, electronic, and magnetic properties

We report on the metalorganic chemical vapor deposition of $\mathrm{GaN}:\mathrm{Fe}$ and $(\mathrm{Ga},\mathrm{Fe})\mathrm{N}$ layers on $c$-sapphire substrates and their thorough characterization via high-resolution x-ray diffraction, transmission electron microscopy (TEM), spatially resolved energy dispersive x-ray spectroscopy (EDS), secondary-ion mass spectroscopy (SIMS), photoluminescence (PL), Hall-effect, electron-paramagnetic resonance (EPR), and magnetometry employing a superconducting quantum interference device (SQUID). A combination of TEM and EDS reveals the presence of coherent nanocrystals presumably ${\mathrm{Fe}}_x\mathrm{N}$ with the composition and lattice parameter imposed by the host. From both TEM and SIMS studies, it is stated that the density of nanocrystals and, thus the Fe concentration increases towards the surface. According to Hall effect measurements, electrons from residual donors are trapped by midgap Fe acceptor states in the limit of low iron content $x\lesssim 0.4\%$, indicating that the concentration of ${\mathrm{Fe}}^{2+}$ ions increases at the expense of Fe ions in the $3+$ charge state. This effect is witnessed by PL measurements as changes in the intensity of the ${\mathrm{Fe}}^{3+}$-related intraionic transition, which can be controlled by codoping with Si donors and Mg acceptors. In this regime, EPR of ${\mathrm{Fe}}^{3+}$ ions and Curie-like magnetic susceptibility are observed. As a result of the spin-orbit interaction, ${\mathrm{Fe}}^{2+}$ does not produce any EPR response. However, the presence of Fe ions in the $2+$ charge state may account for a temperature-independent Van Vleck–type paramagnetic signal that we observe by SQUID magnetometry. Surprisingly, at higher Fe concentrations, the electron density is found to increase substantially with the Fe content. The coexistence of electrons in the conduction band and Fe in the $3+$ charge state is linked to the gradient in the Fe concentration. In layers with iron content $x\gtrsim 0.4\%$ the presence of ferromagnetic signatures, such as magnetization hysteresis and spontaneous magnetization, have been detected. A set of precautions has been undertaken in order to rule out possible sources of spurious ferromagnetic contributions. Under these conditions, a ferromagneticlike response is shown to arise from the $(\mathrm{Ga},\mathrm{Fe})\mathrm{N}$ epilayers, it increases with the iron concentration, it persists up to room temperature, and it is anisotropic—i.e., the saturation value of the magnetization is higher for in-plane magnetic field. We link the presence of ferromagnetic signatures to the formation of Fe-rich nanocrystals, as evidenced by TEM and EDS studies. This interpretation is supported by magnetization measurements after cooling in and without an external magnetic field, pointing to superparamagnetic properties of the system. It is argued that the high temperature ferromagnetic response due to spinodal decomposition into regions with small and large concentration of the magnetic component is a generic property of diluted magnetic semiconductors and diluted magnetic oxides showing high apparent Curie temperature.

Figure 1

In situ ellipsometry: ellipsometric angle $\Psi$ at $2\mathrm{eV}$ as a function of time for a standard deposition sequence. Particular regions relative to the growth process and marked (a)–(g) are discussed in the main text.

Figure 2

(Color online) In situ XRD kinetic measurement: FWHM of the $\left(11\overline{2}4\right)$ $\mathrm{GaN}$ reflection vs deposition time for two samples grown with different ${\mathrm{Cp}}_2\mathrm{Fe}$ flow rates. The evolution of the FWHM during the growth of the $\mathrm{GaN}$ buffer is reproducibly independent of the sample, whereas the broadening during the deposition of the Fe-containing layer is a function of the magnetic ions content.

Figure 3

(Color) In situ XRD space maps in the (11$\overline{2}$4) reflection of $\mathrm{GaN}$ and broadening in the inverse of the lattice parameter along the growth direction (left panels) for two values of ${\mathrm{Cp}}_2\mathrm{Fe}$ flow rates, 50 and $350\mathrm{sccm}$ (upper and lower panel, respectively).

Figure 4

HRTEM lattice fringe image of the (0002) planes demonstrating that low Fe doping (sample No. 465) does not cause detectable strain, the lattice parameter $(d_{0002}=0.25\pm 0.01\mathrm{nm})$ being the same as that of undoped $\mathrm{GaN}$.

Figure 5

EDS spectrum showing the presence of Fe in the measured regions of the $\mathrm{GaN}$ layers doped with a low concentration of Fe ions (sample No. 465).

Figure 6

Bright-field transmission electron micrographs of $(\mathrm{Ga},\mathrm{Fe})\mathrm{N}$ demonstrating the presence of precipitates as indicated by the white triangles (a) close to the surface and (b) in the middle of the epilayer. Some precipitates (c) are embedded into the surface and others, (c) and (d) sink to the dislocations.

Figure 7

EDS spectra for sample No. 466 (left panel) taken around the precipitate (right panel) showing that the concentration of Fe is significantly enhanced in the region of the precipitate (recorded as spectrum 3) as compared with the surrounding matrix (recorded as spectrum 4).

Figure 8

(a) Elongated precipitate observed in an HRTEM image of Moiré fringes contrast. (b) SADP pattern acquired in the region around the precipitate along the $\left(10\overline{1}0\right)$ zone axis; (c) the corresponding schematic graph for indexing of the diffraction spots.

Figure 9

(a) HRTEM image of a precipitate residing on the surface of an highly Fe-doped sample (No. 466); (b) the corresponding SADP pattern along the $\left(10\overline{1}0\right)$ zone axis.

Figure 10

Electron concentration $n$ from Hall effect measurements as a function of the inverse temperature for a $\mathrm{GaN}$ layer and a series of $\mathrm{GaN}:\mathrm{Fe}\text{−}(\mathrm{Ga},\mathrm{Fe})\mathrm{N}$ samples with differing Fe content controlled by the ${\mathrm{Cp}}_2\mathrm{Fe}$ flow rate.

Figure 11

Electron concentration $n$ from Hall effect measurements as a function of the ${\mathrm{Cp}}_2\mathrm{Fe}$ flow rate for two $\mathrm{GaN}$ reference layers and for two series of $(\mathrm{GaN}:\mathrm{Fe}\text{−}\mathrm{Ga},\mathrm{Fe})\mathrm{N}$ samples (triangles and squares, respectively) at room temperature. Reduced electron concentrations in the low flow rate region points to a semi-insulating character of Fe-doped layers in this regime.

Figure 12

Electron Hall mobility $\mu$ as a function of temperature for a $\mathrm{GaN}$ layer and a series of $(\mathrm{Ga},\mathrm{Fe})\mathrm{N}$ samples with differing Fe content controlled by the ${\mathrm{Cp}}_2\mathrm{Fe}$ flow rate.

Figure 13

Photoluminescence spectra at $10\mathrm{K}$: comparison between samples with low (No. 409) and high (No. 466) Fe concentration.

Figure 14

Near-band-edge emission as a function of the Fe precursor flow rate.

Figure 15

Near band edge photoluminescence in $\mathrm{GaN}:\mathrm{Fe}$. The positions of $D^0{X}_A$ and the two electron satellite $D^0{X}_{A,n=2}$ are indicated by arrows, as well as those of the LO phonon replicas for the bound- and free-exciton $\left(F{X}_A\right)$.

Figure 16

Intensity dependence of the ${\mathrm{Fe}}^{3+}$ internal transition as a function of codoping with Mg.

Figure 17

Intensity dependence of the ${\mathrm{Fe}}^{3+}$ internal transition as a function of codoping with Si.

Figure 18

EPR spectra—first derivative of the microwave absorption with respect to the magnetic field—of ${\mathrm{Fe}}^{3+}$ in $\mathrm{GaN}$ for the magnetic field direction close to the [0001] $\mathrm{GaN}$ axis. The contribution of ${\mathrm{Cr}}^{3+}$ and ${\mathrm{Er}}^{3+}$ contamination of the sapphire substrate is easily seen. The lower spectrum was taken at a higher microwave power than the upper one (since the EPR signal of ${\mathrm{Er}}^{3+}$ is not as easily saturated as that of ${\mathrm{Cr}}^{3+}$) and a background due to microwave absorption in the cryostat is observed.

Figure 19

Angular dependence of the EPR line positions for sample No. 406. The solid lines represent the fit to experimental data with the Hamiltonian and parameters given in the text.

Figure 20

(Color online) Concentration of ${\mathrm{Fe}}^{3+}$ estimated from the EPR signal intensity (circles) as compared to the total Fe concentration measured by SIMS (triangles) vs the iron precursor flow rate. Full and open symbols identify two different sample series, grown at the same conditions.

Figure 21

(Color online) EPR line broadening (left $y$ axis) as a function of the iron precursor flow rate (hollow circles) for ${\sigma }_0=12\mathrm{G}$. The SIMS concentration (right $y$ axis) at low flow rates, denoted by full triangles, is shown for comparison.

Figure 22

(Color online) Magnetic susceptibility as a function of temperature, showing the temperature-independent and the temperature-dependent components assigned to the Van Vleck and Curie paramagnetism, respectively. The ferromagnetic and substrate contributions (temperature independent and linear in the field) have been subtracted.

Figure 23

(Color online) Magnetization from SQUID measurements at $200\mathrm{K}$ for a series of $(\mathrm{Ga},\mathrm{Fe})\mathrm{N}$ samples with different Fe content (solid symbols). The substrate contribution (temperature independent and linear in the field) has been subtracted in this and in the figures that follow. Data obtained by numerical reflection, if only a half of the hysteresis cycle has been measured, are shown by open symbols.

Figure 24

(Color online) Magnetization at $5\mathrm{K}$ for a series of $(\mathrm{Ga},\mathrm{Fe})\mathrm{N}$ samples with various Fe contents (solid symbols). Data obtained by numerical reflection, if only a half of the hysteresis cycle has been measured, are shown by open symbols.

Figure 25

(Color online) Magnetization at $200\mathrm{K}$ for a series of $(\mathrm{Ga},\mathrm{Fe})\mathrm{N}$ samples with various Fe content (solid symbols). Inset: above room temperature $M\left(H\right)$ for sample No. 408. Data obtained by numerical reflection, if only a half of the hysteresis cycle has been measured, are shown by open symbols.

Figure 26

(Color online) Determination of the spontaneous magnetization at various temperatures obtained by plotting the square of the magnetization vs the ratio of the magnetic field to the magnetization (Arrot plot, sample No. 408). The inset shows the established spontaneous magnetization as a function of temperature, whose extrapolation leads to the apparent Curie temperature of $500\mathrm{K}$ for the same sample.

Figure 27

(Color online) Magnetization measurements for sample No. 408 (solid symbols), indicating that the saturation value of magnetization is greater for the magnetic field $H_{\Vert }$ along the surface of the sample (perpendicular to the $c$ axis). Data obtained by numerical reflection, if only a half of the hysteresis cycle has been measured, are shown by open symbols.

Figure 28

(Color online) Magnetization of sample No. 466 measured at $H_{\mathrm{appl}}=50\mathrm{Oe}$ after cooling the sample in absence of the external field (ZFC) and during cooling in the same $H_{\mathrm{appl}}$. Qualitatively the same results are obtained at $H_{\mathrm{appl}}=200\mathrm{Oe}$.

Figure 29

Saturation magnetization of the ferromagnetic component extracted from $M\left(H\right)$ curves at $200\mathrm{K}$ for the No. 466 - No. 473 layer series versus the Fe precursor flow rate.

Figure 30

(Color online) Typical examples of magnetic moment vs magnetic field $m\left(H\right)$ recorded for the layers presented in this paper. Left panels: raw data. Right panels: after subtracting the major linear component. The high slope lines on the left panels indicate the uncompensated $m\left(H\right)$ data. Data obtained by numerical reflection, if only a half of the hysteresis cycle has been measured, are shown by open symbols.