The Fermi level dependence of the optical and magnetic properties of Ga1−xMnxN grown by metal–organic chemical vapour deposition

The suppression of the ferromagnetic behaviour of metal–organic chemical vapour deposition grown Ga1−x Mnx N epilayers by silicon co-doping, and the influence of the Fermi level position on and its correlation with the magnetic and optical properties of Ga1−x Mnx N are reported. Variation in the position of the Fermi level in the GaN bandgap is achieved by using different Mn concentrations and processing conditions as well as by co-doping with silicon to control the background donor concentration. The effect on Mn incorporation on the formation of defect states and impurity induced energy states within the bandgap of GaN was monitored by means of photoluminescence absorption and emission spectroscopy. A broad absorption detected around 1.5 eV is attributed to the presence of a subband introduced by Mn induced energy states due to temperature independent transition energies and linewidths. The intensity and the linewidth of the absorption band correlate with the Mn concentration. Similarly, the magnitude of the magnetization decreases as the Fermi level approaches the conduction band, as the Fermi energy is increased above the Mn(0/−) acceptor state. Silicon concentrations >1019 cm−3 caused the complete loss of ferromagnetic behaviour in the epilayer. The absorption band at 1.5 eV is also not observed upon silicon co-doping. The observed spectroscopic data favour a double-exchange-like mechanism rather than an itinerant free carrier mechanism for causing the ferromagnetism. This behaviour significantly differs from the properties reported for widely studied (Ga, In)MnAs. 0953-8984/06/092615+08$30.00 © 2006 IOP Publishing Ltd Printed in the UK 2615 2616 M Strassburg et al There has been increased interest in transition metal (TM) doped wide bandgap materials, such as Ga1−xMnxN, that was triggered by theoretical predictions suggesting that ferromagnetism of diluted magnetic semiconductors with Curie temperatures above room temperature can be obtained [1]. The theoretical work by Dietl et al [1] states that the ferromagnetism is facilitated by interaction between the Mn2+ ions and holes from the GaN valence band. This assumption suggests that the preparation of p-type material would be required to shift the Fermi level further towards the valence band to increase the observed ferromagnetism [2]. The theory still predicts the observation of ferromagnetism in n-type material but much weaker (more than four times) than that observed in p-type material. Experimental identification of the Mn ion charge state is enabled by electron spin paramagnetic resonance (EPR) and optical spectroscopy [2–5]. The existence of room temperature ferromagnetism in Ga1−x MnxN has also been predicted by other theoretical studies which suggest that a Mn induced impurity band provides effective-mass transport within the band [6, 7]. Sato et al [7] stated that the incorporation of Mn3+ facilitates the formation of a sharp E impurity band and a broader T2 impurity band, altering the electronic structure in the bandgap of Ga1−xMnx N. In this model, broadening of the partially filled T2 band stabilizes the ferromagnetism via the double-exchange interaction [8, 9], provided the Fermi level is located in the defect band. This mechanism differs significantly from that theoretically predicted and experimentally confirmed for Ga1−x Mnx As [1, 10]. Reed et al have recently shown a strong correlation between the observed magnetic signature and the position of the Fermi level in Ga1−x MnxN [11]. Further energy states in the GaN bandgap due to the Mn incorporation have been reported from optical spectroscopy causing photoluminescence (PL) bands predominantly in the blue, and in the yellow spectral range [4, 12–15]. However, implantation induced defect states cannot be completely ruled out as a cause for these PL bands, since most of the structures are ion-implanted Ga1−xMnx N epilayers. Correlation of these magnetic observations with optical, electronic and structural investigations may improve the fundamental understanding of the nature of the ferromagnetic interaction in these materials. In this paper, an investigation of the Fermi level dependence and of the magnetization and corresponding optical properties in MOCVD grown Ga1−x MnxN is presented. Room temperature ferromagnetism was observed in samples with Mn concentrations between 1 and 2%. The magnitude of the magnetization scaled with the Mn concentration and showed a strong dependence on the position of the Fermi level; additionally, it varied with the silicon co-doping. Complete suppression of the ferromagnetic behaviour was observed upon silicon co-doping to >1019 cm−3. A broad absorption band detected around 1.5 eV is assigned to a Mn induced impurity band and showed a dependence on doping similar to that observed for the magnetization dependence. These experimental findings rather favour a double-exchange-like interaction causing the observed ferromagnetic behaviour in Ga1−x Mnx N, in that variable range hopping between mixed-valence deep level transition metal impurities is likely the source of ferromagnetism. This mechanism is different to that typically reported valence band carrier mediated mechanism for Ga1−x Mnx As underlining the significant differences between TM doped nitrides and arsenides. Epitaxial Ga1−x MnxN films with Mn concentration up to ∼2% were grown on GaN template layers in a specifically modified VEECO D125-MOCVD reactor. Biscyclopentadienyl manganese (Cp2Mn) and silane (SiH4) were used as the manganese and n dopant sources respectively. A more detailed description of the growth is given elsewhere [16, 17]. Subsequent characterization of these thin films included x-ray diffraction (XRD), secondary-ion mass spectroscopy (SIMS), SQUID magnetometry, photoluminescence (PL) and transmission spectroscopy. The PL was excited by a frequency-doubled titanium– sapphire laser; transmission was effected using the red and infrared spectrum of a halogen lamp. The emitted and transmitted light was detected by a photomultiplier attached to a 0.24 m Optical and magnetic properties of Ga1−x Mnx N 2617 Figure 1. Magnetization spectra of Ga0.985Mn.015N and Ga1−x Mnx N:Si for two different Si concentrations recorded at 5 K. The latter are also shown in the inset on a smaller scale. The observed hysteresis indicates the ferromagnetic behaviour of the Mn doped sample. Upon Si co-doping, the Fermi level shifts towards the bottom of the conduction band, reducing the ferromagnetic interaction. For the highest Si concentration, the ferromagnetic contribution is completely suppressed. The diamagnetic behaviour is due to the sapphire substrate. monochromator with a spectral resolution of better than 1 nm for emission and better than 6 nm for transmission experiments. The crystalline quality of Ga1−xMnx N and the absence of secondary phases in the layers were confirmed by XRD analysis described in more detail elsewhere [16, 17]. The Mn concentration was varied between 0.3% and 1.5% as confirmed by various techniques. Slight variations in Mn concentrations and thicknesses across the wafer were established on the basis of the gas flow and the temperature gradients in the MOCVD growth. Ferromagnetic behaviour, indicated by hysteresis in the magnetization spectra, was detected in the Ga1−x Mnx N samples at doping levels of 1.5%, as shown in figure 1. Two significant results were obtained. First, the saturation magnetization (the size of the hysteresis loop) scales with the Mn concentration confirming that Mn doping causes the observed ferromagnetic properties. Hence, mandatory contributions from the Cr3+ impurities in the substrate as observed and confirmed by means of EPR for other samples [18] can be ruled out as the primary source of the ferromagnetic behaviour. Secondly, with increasing silicon concentration, the size of the hysteresis loop decreases significantly (by more than one order of magnitude). Although the results are shown at 5 K, the magnetic hysteresis persists and an almost identical magnetization curve is observed at room temperature. The Curie temperature was not measured due to experimental limitations, but there is only a little drop-off in the observed magnetization between 5 and 300 K indicating that Tc is well above room temperature, as reported elsewhere [15]. At high Si concentration ([Si] > 1019 cm−3), ferromagnetic behaviour is suppressed and the magnetization spectra reveal an overall diamagnetic behaviour for this sample due to the contribution from the sapphire substrate. The suppression is attributed to the shift of the Fermi level towards the conduction band and above the Mn2+/3+ acceptor level, when the Mn acceptors are finally overcompensated by silicon that introduces shallow donor states in GaN. EPR investigations confirm the presence of the Mn2+ state for the Si co-doped samples [18]. Si co-doping in GaN normally increases the free carrier concentration providing electrons in the conduction band. According to the Zener mean field model based theories, an increasing n-type carrier concentration would also yield an increase of the magnetization, which was not observed in this work (figure 1). However, it is questionable whether the carrier and Mn concentrations are high enough in this case to render the mean field carrier mediated model applicable. Unlike that 2618 M Strassburg et al of Ga1−x MnxAs, the magnetization behaviour of Ga1−xMnx N can be understood assuming the formation of a partially filled Mn induced band. The ferromagnetism is thus stabilized by the double-exchange-like interaction of electrons in this band. In the case of low concentration Si co-doping, not every Mn acceptor state is compensated. Hence, holes are present in this band; the magnetic ordering in optimally MOCVD grown Ga1−x MnxN is attributed to exchange between electrons localized on the levels lying deep in the forbidden energy gap [19]. The reduction of available free states for the double-exchange-like interaction (holes) in the T2 band holds for the observed smaller magnetization [7]. The measured carrier type in all of these samples is n type in all cases, similar to what has been reported for ferromagnetic Ga1−xMnxN in the literature [20]. However, it is likely that this is merely a function of the template layer used to grow the samples, especially in the samples without Si co-doping. To achieve high structural quality for the Ga1−xMnx N layer, these epilayers were grown on 1 μm thick GaN template layers; for the Si co-doped samples, these layers were doped n type. The measured Hall concentrations of the as-grown, unintentionally doped Ga1−x MnxN films were around n = 5 × 1016 cm−3, which is very close to the measured background carrier concentration of the unintentionally doped template layer. Similarly increasing the Si doping within the Ga1−xMnx N layer does not result in an increase in measured carrier concentration from the n = 8×1017 cm−3 that was almost exactly the measured carrier concentration in the n-type template layer on which these samples were grown. Thus, all of the measured n-type behaviour is due to parallel conduction through the virtual template, and the observed ‘n-type’ character that is often reported in these systems may be solely due to the template. In order to clarify this point, layers were grown from the substrate with Mn in the virtual template layers. For these samples, the resistivity was too high to measure and the contact resistance increased by several orders of magnitude, consistent with predictions and other observation of Mn as a deep impurity level. It should be noted that the layer grown on the Mn-containing template has a reduced quality and does not exhibit strong ferromagnetism even though it exhibits a strong reddish tint indicative of Mn2+ incorporation. To check whether the magnetic behaviour is due to second phases or clusters, the concentration per magnetic element was calculated. Assuming a uniform growth rate and Mn concentration as measured using SIMS across the wafer, the measured saturation magnetization corresponds to a contribution of 2.9 μB/Mn for the 1% doped sample and 1.2 μB/Mn for the 1.5% doped sample. This compares favourably with the predicted contribution based on first-principles band structure calculations of 4 μB/Mn when the Fermi level is located in the centre of the Mn impurity band [6, 21]. The predicted contributions for the mean field model are also around 4 μB/Mn for hole mediated ferromagnetism, though experimentally for Ga1−x MnxAs these values are often much lower than anticipated especially at high doping levels due to defect cluster and/or defect compensation [22]. For comparison, the predicted magnetic moment per Mn4N cluster is 17 μB/Mn [23]. Thus, if the ferromagnetism was due only to this atomic configuration, more than half of the Mn would have be tied up in these clusters, which should be observable through structural characterization techniques. However, since a phase separation based FM has been observed in ferromagnetic chalcopyrite materials with similar magnetization strength, further work is required to unambiguously rule out clusters as the origin of FM in MOCVD grown Ga1−xMnx N [24, 25]. To clarify whether isolated Mn ions (as observed for bulk Ga1−x MnxN [26] and Ga1−xMnx As) or the formation of an Mn induced impurity energy band [6, 7] are the primary cause for the observed ferromagnetism, transmission and PL emission studies were performed. The transmission spectra of four samples are shown in figure 2. Both the Mn content and/or the donor concentration were varied during the MOCVD growth. The latter was predominantly achieved by Si co-doping during the growth. Despite high Si concentrations, Optical and magnetic properties of Ga1−x Mnx N 2619 1.4 1.5 1.6 1.7 1.8 1.9 2.0 Ga 0.985 Mn 0.015 N Ga 0.99 Mn 0.01 N GaN:Mn