Introduction

GaN and related (In,Al,Ga)N alloys have become technologically important materials over the past fifteen years due to their unique physical properties, which include a wide direct bandgap, remarkable mechanical strength and high melting temperature. Commercial use of nitrides spans a broad field from optoelectronics to microwave technology. Besides being super-efficient light emitters on their own account, nitride semiconductors also prove to be promising hosts for visible- and infrared-emitting trivalent rare-earth (RE3+) ions1,2. The luminescence of RE3+ ions is an attractive source of emission which is relatively independent of the ambient temperature and host material3. The use of semiconductors as hosts for RE ions has attracted much interest recently for the potential integration of RE emission with existing semiconductor devices. The discovery by Favennec et al.4 that wide bandgap hosts reduce the temperature quenching of RE luminescence led to a widespread interest in doping of III-nitride semiconductors with RE3+ ions. More recently, attempts to exploit another property of the 4f-shell, namely the uncompensated spin of electrons, in order to produce dilute magnetic semiconductor, have been reported5,6.

Doping GaN with Eu is particularly attractive for the realization of red light-emitting devices7 due to problems related to the growth of InGaN with the moderately high InN fractions, required to produce intrinsic red emission from this alloy8. The first Eu-doped GaN electroluminescent devices, operating at a bias of ~50 V, were reported by Heikenfeld et al.9. Recently, the first current-injection light emitting diode (LED), operating at voltages as low as 3 V, was realised by one of the present authors10. Stimulated emission from Eu-doped GaN has also been reported11.

Interestingly, the observation of a ferromagnetic phase in Eu-doped GaN epilayers was first reported nearly a decade ago12. The presence of the ferromagnetism was attributed to the coexistence of magnetic Eu2+ species along with EuGaN since Eu3+ is non-magnetic. Hite et al.13 reconfirmed the existence of a ferromagnetic response from Eu-doped GaN. However, no second phase formation was observed in the samples13 and it was pointed out that no Eu phases are ferromagnetic at room temperature; the secondary phases considered include EuGa, EuGa2, EuGa4, Eu3Ga8, Eu5Ga8, metallic Eu, EuN, EuO, Eu3O4 and Eu2O3. On the other hand, two of the present authors recently reported the Zeeman splitting of emission lines at ~622 nm assigned to the 5D0→7F2 transition of Eu3+ centres in GaN with the magnetic field applied along the c-axis14. The observed Zeeman splittings, with g-factors ~2, were suggested to be associated with the magnetic moment of the final 7F2 state of the 5D0→7F2 transition, since the 7F0 ground state and 5D0 excited state of Eu3+ ions do not have any associated magnetic moments.

In this work we report the detection of the induced magnetic moment associated with the 7F2 state of Eu3+ ions which is involved in optical emission. We used X-ray magnetic circular dichroism (XMCD), which is the difference between the X-ray absorption spectra (XAS) for left- and right-circularly polarized X-rays. XMCD probes the magnetic moments of the states involved in the element-specific X-ray absorption process utilizing the helicity of the circularly polarized photons parallel and antiparallel to the sample magnetization direction15. To measure the induced magnetic moment of Eu3+ ions, the bright red Eu-related luminescence (with photon energy ~2 eV) emitted following X-ray excitation, referred to as X-ray excited optical luminescence (XEOL), was monitored with the X-ray energy scanned across the Eu M4 (~1158 eV) and M5 (~1128 eV) absorption edges (3d3/2,5/2→4f transitions). Conventional XMCD measurements were also taken, by detecting secondary X-ray fluorescence (XRF) and the drain current caused by Auger and secondary electrons, commonly referred to as total electron yield (TEY) detection.

Results

Sample structure and cathodoluminescence spectroscopy

The sample studied was GaN doped with Eu in-situ during growth by organometallic vapor phase epitaxy (OMVPE)16. The thickness of the doped layer was 400 nm. The concentration of Eu, estimated by Rutherford backscattering spectrometry (RBS), was 0.11 at%17 corresponding to 9.6×1019 cm–3. Fig. 1 shows the cathodoluminescence (CL) spectrum of the sample at room temperature. Several emission lines are observed, identified as 5D1→7F1, 5D0→7F1, 5D0→7F2 and 5D0→7F3 intra-4f shell transitions at ~543 nm, 600 nm, 622 nm and 663 nm, respectively, according to the Dieke diagram18,19. Energy states for RE3+ ions are labelled following the Russell-Saunders scheme3 as 2S+1LJ, in which S is the total spin angular momentum, L is the total orbital angular momentum and J = L+S is the total angular momentum. The total momenta S and L are vectorial sums of individual electron spins and orbital moments, respectively. For energy levels with J = 0, the orbital moment compensates the spin moment, so that Eu3+ has no magnetic moment in the 7F0 ground state. The ‘hypersensitive' emission line at 622 nm is by far the brightest transition. Note that GaN band edge emission in the 350–400 nm spectral region is not observed: it is very much weaker than Eu3+ emission which dominates the entire CL spectrum of GaN:Eu.

Figure 1
figure 1

Cathodoluminescence spectroscopy.

(a) Room temperature CL spectrum obtained with an electron beam energy of 6 kV and beam current of 10 nA. The electron beam was defocused to a spot size of 10 μm. (b) Energy level diagram with emission lines due to intra-4f shell transitions for Eu3+ ions in GaN.

Full size image

X-ray magnetic circular dichroism

Fig. 2 shows the XMCD and corresponding XAS recorded by TEY and XEOL. Due to the extremely low Eu concentration the XAS are dominated by the L2,3 absorption from Ga atoms. The XMCD feature corresponding to the M5 absorption peak due to the 3d5/2→4f transition appears at 1128.3 eV and 1129.8 eV in TEY and XEOL, respectively. When the magnetic field is reversed the XMCD features in both TEY and XEOL reverse in sign at the same energy positions as expected, thereby confirming the reliability of the observation. Note that the difference in energy of the XMCD peaks for TEY and XEOL is 1.5 eV. The M4 absorption peak corresponding to the 3d3/2→4f transition is present in TEY at ~1157 eV and not resolved in XEOL. No XMCD features were observed using secondary XRF detection.

Figure 2
figure 2

X-ray absorption and X-ray magnetic circular dichroism.

XAS and XMCD spectra measured by detecting the total electron yield and X-ray excited optical luminescence at T = 1.5 K. The vertical dashed lines with arrows indicate the shift of the XMCD peak in TEY and XEOL. The XAS is dominated by the Ga L2,3 absorption edges. Spectra are offset vertically for clarity.

Full size image

The experimentally observed difference in the XMCD peak positions for TEY and XEOL can be explained by taking into account the depth sensitivity and underlying physics of each detection method. TEY is a surface sensitive, where the signal probes the sample surface with an attenuation length of ~5 nm. X-ray photoelectron spectroscopy, with an even smaller probing depth, indicates that Eu3+ undergoes a valence-state transition into Eu2+ near the surface of GaN20. The presence of a 10 nm thick GaN cap in the sample should therefore reduce the bulk TEY signal by 86%; it is also possible that Eu ions diffuse through the GaN cap. In any case, the XMCD measured by TEY clearly indicates the presence of Eu2+ ions at the surface of GaN.

On the other hand, visible photons emitted by Eu3+ are collected from the entire depth of the epilayer. Since the visible luminescence of the sample is dominated by the emission line at 622 nm due to 5D0→7F2 intra-4f shell transitions with a magnetic final state, the XMCD spectrum measured by XEOL detection uniquely probes the induced magnetic moment of the excited 7F2state of Eu3+ ions in the bulk of the epilayer. The description of the processes underlying the optical detection of XMCD is as follows:

1) X-rays generate Eu3+ XEOL by direct excitation following the absorption of an X-ray photon by an atomic level and indirectly through the generation of electron-hole pairs.

2) The 7F2 final state of the 5D0→7F2 transition is populated for ~150 μs after the excitation19, long enough for the next (probing) X-ray photon to arrive.

3) The X-ray energy is scanned across the Eu M4,5 edges and any resulting change in the Eu3+ luminescence is monitored.

Note that in our experiment the X-ray beam is used both as a pump and as a probe. Interestingly, XEOL detection of X-ray absorption spectra was used successfully by Sham et al.21 to explore the origin of luminescence in porous silicon.

The absence of an XMCD signal in XRF can be explained by the fact that the X-ray photon flux incident on the sample (~1012 photons per second) may not be sufficient to populate the 7F2 states to an adequate detection level, or by the fact that not all Eu3+ ions are involved in the process of light emission17 whereas all Eu3+ ions are expected to emit secondary X-ray fluorescence photons. This is to say that the emission of the visible photons following X-ray absorption is intrinsically a more complex process than the emission of X-ray fluorescence photons. Due to the absence of information about the relaxation from higher to lower states of the 7FJ manifold, however, it is difficult to judge what the primary cause may be.

Atomic multiplet calculations

Theoretical XAS and XMCD spectra for the M4,5 absorption of trivalent and divalent Eu ions were calculated using atomic multiplet theory22. Spectra for Eu2+ 8S7 as well as Eu3+ 7F0 and 7F2 are shown in Fig. 3. In the calculation the XMCD peak for 7F2 is shifted by 2.2 eV towards higher energy compared to that for the Eu2+ 8S7 state. This energy shift is reproduced very well in the experimentally observed positions of the XEOL and TEY peaks.

Figure 3
figure 3

XAS and XMCD calculated by atomic multiplet theory.

Theoretical spectra for Eu 4fn→3d94fn+1 transitions. Vertical dashed lines indicate the shift of the calculated M5 XMCD peaks for Eu2+ 8S7 and Eu3+ 7F2 states. Spectra are offset vertically for clarity.

Full size image

X-ray absorption near edge structure

In order to check for the presence of Eu2+ in the bulk of the epilayer, the X-ray absorption near edge structure (XANES) of the Eu L3 absorption edge at 6977 eV23 was measured using Lα1 secondary X-ray fluorescence detection at 5845.7 eV. XANES is a sensitive tool for the determination of the valence states of a particular element in a sample by analyzing the shape and position of the resonance peak, or so called white line24,25,26,27. In the case of Eu2+ and Eu3+ the resonance is due to 2p3/2→5d electronic transitions. The Eu2+ and Eu3+ valence states can be easily distinguished by the energy position of their white lines. If both Eu2+ and Eu3+ are present, XANES typically displays two peaks with the stronger one belonging to the dominant oxidation state. The white line for Eu2+ species is 7–8 eV lower in energy than that of the Eu3+ species24,25,26,27. The shift in the white line energy position is primarily due to a lower binding energy of the core electrons in Eu2+ through shielding of the nucleus by the additional 4f electron. Generally, the tabulated value of Eu L3 X-ray absorption edge energy position of 6977 eV23 separates the white lines for Eu2+ and Eu3+ species24,25,26,27. Fig. 4 shows the analysis of the XANES spectrum for the sample studied. The white line is located at 6980.5 eV with no other peak resolved below 6977 eV. This is a clear indication that no Eu2+ is present in the bulk of the epilayer within the sensitivity of XANES measurements. Note that our sample is very dilute with a peak concentration of ~9.6×1019 cm−3 corresponding to an average distance between the Eu3+ ions of ~2 nm.

Figure 4
figure 4

X-ray absorption near edge spectroscopy.

XANES spectrum was taken at room temperature. Vertical red dashed line indicates the position of the Eu L3 absorption edge at 6977 eV. Blue short-dashed lines show fitting with a combination of Lorentzian and arctangent functions, which simulate the resonance peak and absorption edge step, respectively. Energy positions of Lorentzian peaks are 6980.5 eV and 6986.5 eV.

Full size image

Discussion

We report here for the first time the detection of an induced magnetic moment of Eu3+ ions in the 7F2 state, which is associated with the 5D0→ 7F2 intra-4f shell optical emission line at 622 nm. The unique features of Eu3+ ions are the non-magnetic 7F0ground state and magnetic 7F1,2,3 states, as displayed in Fig. 1, which can be populated optically or electrically in an LED structure. It is interesting to speculate whether the magnetic moments of the 7FJ manifold can play a role in the reported room temperature ferromagnetism of Eu-doped GaN. For instance, the 7F1 state is ~70 meV above the 7F0 ground state, so that at room temperature, according to Boltzmann statistics, ~7% of the Eu3+ ions will be in the 7F1 state. The sample studied showed ferromagnetic response measured by Superconducting Quantum Interference Device (SQUID) magnetometry, however, the interpretation of this observation and its relation to the optically induced magnetic moment is a matter for future investigations. Another unique property of Eu3+ is its non-magnetic 5D0 excited state. Apart from inducing a magnetic moment by forcing the Eu3+ ion to emit a photon, one can switch off the induced magnetic moment by shining light with an energy specific to the transitions from the 7FJ manifold to the 5D0 state to which GaN is transparent. This opens up a very interesting prospective for electrical or optical manipulations28 with the excited-state magnetic moments of Eu3+ in GaN, which is not feasible with Gd for example since the magnetic moment of Gd cannot be easily switched off. The well-established processing technology for nitride semiconductors should certainly facilitate potential applications of Eu-doped GaN in novel magneto-optoelectronic devices.

Methods

Organometallic vapor phase epitaxy

The growth procedure is described in Ref. 16. Eu-doped GaN layers were grown on a sapphire substrate by OMVPE in Taiyo Nippon Sanso SR-2000 system. Trimethylgallium (TMGa) and ammonia (NH3) were used as precursors. Tris(dipivaroylmethanato)europium referred to as Eu(DPM)3was used as a Eu source. The Eu(DPM)3 source temperature was set at ~135°C, while keeping the dopant line temperature at ~145°C to prevent vapor condensation during transport. The growth process consisted of the following steps. First, a 20 nm thick GaN buffer layer was deposited on (0001) sapphire substrate. Then a 4 μm thick undoped GaN layer was grown followed by the deposition of 400 nm thick Eu-doped GaN layer. The pressure in the reactor was 10 kPa. The TMGa and NH3 flow rate were fixed at 25.5 μmol/min and 8.1 mmol/min respectively. The III/V ratio flow rate was 320. The growth temperature for the sample was 1000°C. Finally, a 10 nm thick GaN cap layer was grown.

Cathodoluminescence spectroscopy

Experiments were carried out using a modified Cameca SX100 electron probe micro-analyser (EPMA). A built-in optical microscope, coaxial and confocal with the electron beam, allows optical monitoring of the sample during measurements. Oriel 0.125 m spectrograph with changeable gratings and a cooled silicon 1024×256 CCD array were inserted in the light path of the optical microscope in order to acquire room temperature CL spectra. A 400 lines/mm diffraction grating was used in the spectrograph. The overall spectral resolution of the system was 0.5 nm.

X-ray magnetic dichroism measurements

XMCD spectra were measured on beamline I06 at Diamond Light Source, UK. Circularly polarized X-rays were produced by an APPLE II undulator operating in third harmonic. Polarization rate was close to 80%. The beamline operated at its highest possible X-ray photon flux of 1012 photons/sec into a spot size of 200 μm (horizontal) by 20 μm (vertical). XMCD spectra were recorded at 1.5 K in a 6 T superconducting magnet using total electron yield and X-ray fluorescence yield detection. X-ray excited optical luminescence was recorded using a Si diode outside the ultrahigh vacuum of the magnet system through a quartz window.

X-ray absorption near edge structure measurements

XANES measurements of Eu L3 edge were performed on beamline B18 at Diamond Light Source. X-rays produced by a bending magnet were collimated by a Pt-coated collimating mirror and tuned to the desired energy by a Si(111) double crystal monochromator. The X-ray beam was focused to a spot size of 500 μm (horizontal) by 300 μm (vertical) using a Pt-coated double-bend focusing mirror. A pair of Pt coated harmonic rejection mirrors were used for harmonic rejection. Eu Lα1 X-ray fluorescence line at 5845.7 eV was detected by a 9-element monolithic Ge detector with XSPRESS-II readout electronics. The Si(111) monochromator was calibrated using the Cr K-edge at 5989 eV prior to the measurements. XANES measurements were done at room temperature. XANES spectra were analyzed by fitting with a combination of Lorentzian and arctangent functions, which simulate the resonance peak and the absorption edge step, respectively.

Atomic multiplet theory

The wave functions of the initial- and final-state configurations for the excitation 4f n → 3d94f n+1 were calculated in intermediate coupling using Cowan's atomic Hartree-Fock (HF) code with relativistic correction22,29. The HF values of the Slater parameters were reduced to 80% to account for screening effects, as was previously found to be the optimal value for the rare-earths M4,5 X-ray absorption spectra. Since such an 80% scaling has become standard for rare earths30, there are essentially no free parameters in the calculation22.