Role of nitrogen split interstitial defects in the magnetic properties of Cu-doped GaN

Cu-doped GaN thin films are grown by plasma-assisted molecular beam epitaxy. With nitrogen plasma only, films phase segregate into GaN and Cu-rich alloys. In contrast, when nitrogen-hydrogen plasma is used, the films are single-phased Ga${}_{1-x}$Cu${}_x$N, with $x$ as high as 0.04. Contrary to earlier studies, however, these films are not ferromagnetic, but rather paramagnetic in nature. First-principles calculations indicate that although each substitutional Cu${}_{\mathrm{Ga}}$ exhibits a moment of 1 ${\mu }_{\mathrm{B}}$/Cu, it can be suppressed by neighboring intrinsic defects such as N split interstitials.

Figure 1

Scanning electron microscopy images of Ga${}_{1-x}$Cu${}_x$N films grown with (a) N${}_2$ plasma, and (b) N${}_2$-H${}_2$ plasma. (c) EDS spectra taken at regions A, B, and C as marked in (a) and (b).

Figure 2

X-ray diffraction patterns for a Ga${}_{0.96}$Cu${}_{0.04}$N film 100-nm-thick grown with N${}_2$-H${}_2$ plasma. Inset: closeup view of the peaks around the GaN (0002) peak at 34.61${}^{\circ }$.

Figure 3

Comparison of the magnetizations for Ga${}_{1-x}$Cu${}_x$N/SiC films grown using N${}_2$-H${}_2$ plasma without ($x$ $=$ 0) and with Cu doping ($x$ $=$ 0.04). (a) Magnetic field dependence of the magnetization measured at 4.2 and 298 K. (b) Temperature dependence of the magnetization with a 500 Oe field applied parallel to the surface of the films. All curves are normalized by the mass of the samples.

Figure 4

Cu local density of states for (a) simple Cu${}_{\mathrm{Ga}}$ (majority and minority) and (b) Cu${}_{\mathrm{Ga}}$ next to a N split interstitial (majority and minority the same). The shaded regions represent the calculated bulk GaN DOS, and the local structure of the N split interstitial is given as the inset to (b).

Figure 5

Calculated XAS and XMCD Cu $L_{\mathrm{III}}$ spectra for the simple Cu${}_{\mathrm{Ga}}$ and next to a N split interstitial. The effect of the core hole was included in the calculations, and the energy position was fixed from the differences in total energies of the systems with and without core holes. (The corresponding calculated $p_{3/2}$ energy in bulk metallic Cu is 931.4 eV compared to the experimental value of 932.5 eV.) The corresponding XAS (solid line) and XMCD (dashed line) spectra for calculations ignoring the core hole (with the energies set by the initial state core eigenvalues) are given in the inset.