Ferromagnetic systems obtained by doping transition metals into semiconductors1 have generated extensive studies since early 1990s2 because of their potential use for spin-sensitive electronics (spintronics) devices3. In proto-typical systems based on III–V semiconductors, such as (Ga,Mn)As and (In,Mn)As, substitution of divalent Mn atoms into trivalent Ga or In sites leads to severely limited chemical solubility. Because of this, the specimens are chemically metastable, available only as thin films1, and their material quality exhibits high sensitivity on preparation methods and heat treatments4. This substitution dopes hole carriers together with magnetic atoms, which prohibits electron doping to obtain n-type systems necessary for formation of spintronics p–n junction devices. To overcome these difficulties, Masek et al.5 theoretically proposed systems based on a I–II–V semiconductor LiZnAs, wherein magnetism due to isovalent (Zn,Mn) substitution may be decoupled from carrier doping with excess/deficient Li concentrations.

LiZnAs is a direct-gap semiconductor that has a cubic crystal structure (Fig. 1a)6,7 and a band gap (1.61 eV) similar to that of GaAs (1.52 eV)8. Monovalent Li and divalent Zn together substitute the role of trivalent Ga (ref. 9). The local density approximation calculation of Masek et al.5 obtained the band structure very close to that of GaAs. LiMnAs is an antiferromagnet with TN exceeding 393 K (refs 10,11) that has a tetragonal crystal structure (Fig. 1b). Via (Zn,Mn) substitution and carrier doping with excess/deficient Li concentrations, Li1+y(Zn1−xMnx)As with a small x is expected to become ferromagnetic analogous to (Ga,Mn)As (ref. 5). Owing to isovalent (Zn,Mn) substitutions, Li1+y(Zn1−xMnx)As may be obtained as chemically stable bulk material for a wide range of x.

Figure 1: Crystal structures of Li(Zn,Mn)As and X-ray and magnetization results.
figure 1

(ac): crystal structures of (a) cubic LiZnAs (direct-gap semiconductor) and Li(Zn,Mn)As (ferromagnetic metal, as demonstrated in the present work), (b) tetragonal LiMnAs (antiferromagnet with TN>393K), and (c) tetragonal LiFeAs (non-magnetic superconductor with Tc~20K). All these structures include square-lattice As layers with the matching of the lattice constant within ~10%. Primitive cells for the tetragonal crystals of (b) and (c) are shown by the black lines. (d,e): X-ray scattering results of (d) intensity profile of Li1.1(Zn,Mn)As with various Mn concentrations x, and (e) lattice constant of Li1.1(Zn1−xMnx) As for various x (blue symbols, bottom horizontal axis), and Li1+yZnAs (red symbols, top horizontal axis) for various Li deficiency/excess y. (f,g): Magnetization M(H,T) results of Li1.1(Zn1−xMnx)As with x=0.0–0.15 showing (f) the T dependence of M in H=2 kOe (no difference in FC and ZFC procedures) and (g) M at T=2 K in various values of external field H. The grey symbols show a hysteresis loop in x=0.03 system plotted for small field regions (top horizontal axis), which demonstrate a very small coercive field of 30–100 Oe.

Here we report the synthesis of bulk specimens of a new ferromagnet Li1+y(Zn1−xMnx)As based on a I–II–V semiconductor, together with characterization of the structural, transport and magnetic properties. Initial values of the ferromagnetic transition temperature TC reach up to 50 K, and the Li excess specimens exhibit p-type carriers, which may be due to excess Li atoms occupying substitutional Zn sites. Future efforts in optimizing growth conditions and heat treatments, as well as the replacement of alkali metal and pnictogen atoms, may allow the development of systems with higher TC and synthesize long-awaited n-type electron doped compounds that exhibit spontaneous magnetic order.


Synthesis and structure

Polycrystalline specimens of Li1+y(Zn1−xMnx) As were prepared by solid-state reaction with high-purity elements at the Institute of Physics (IOP) in Beijing. The main difficulty was the high chemical activity and volatility of lithium. Initially, excess lithium was used, but it was difficult to account for the exact amount. A sealed metal tube solved this problem. Starting materials with amounts proportional to the nominal element concentrations were loaded into a Ta tube with Ar under 1 atm pressure. Then, the sealed Ta tube was sealed into an evacuated quartz tube. The samples were heated to 700 °C and held for several hours before the temperature was decreased slowly. The powder samples were characterized by X-ray diffraction and dc magnetization at IOP, and sintered pellet samples were examined by electrical resistivity and Hall effect at IOP and Tsinghua and muon spin relaxation (μSR) measurements at TRIUMF, in Vancouver, Canada.

The X-ray measurements showed that Li1+y(Zn1−xMnx)As has a crystal structure with the space group F-43 m (Fig. 1a–d). This structure can be viewed as zinc-blend GaAs, in which Ga is replaced by Li and Zn, with Li atoms located at the As tetrahedral interstitial sites. The lattice parameters for Li1+y(Zn1−xMnx)As are shown in Figure 1e: a=5.940(1) Å for LiZnAs is consistent with the reported value7, and the monotonic behaviour with increasing x suggests successful (Zn,Mn) solid solution. The equilibrium solubility of Mn atoms in III–V compounds is less than 1%, and specimens of (Ga,Mn)As are available only in the form of chemically metastable thin films made by molecular beam epitaxy1. In contrast, chemically stable bulk crystals with the cubic structure of LiZnAs can be obtained for Mn concentrations x up to at least 0.15 and Li excess content y ranging −0.1≤y≤0.2. For systems with 0≤x≤0.15, excess Li with 0≤y does not alter the lattice constants, as demonstrated in Figure 1e with the results of x=0, which will be discussed later.

The dc-magnetization M was measured for specimens with nominal Li concentrations 0.9, 0.95, 1.0, 1.05, 1.1 and 1.2 (that is, −0.1≤y≤0.2) and Mn concentration x=0, 0.02, 0.03, 0.05, 0.1 and 0.15. Ferromagnetism was found only in compounds with excess Li (y≥0.05). Figure 1f shows the temperature dependence of M at 2 kG (no difference between field cooling (FC) and zero-field cooling (ZFC) procedures), whereas Figure 1g shows the field dependences of M at T=2 K, for Li 1.1 systems with various x values. Clear signatures of ferromagnetic order are seen in these figures. In Table 1, we list the values of Curie temperature Tc and the spontaneous magnetization M(T=2 K; H=2 kG) for several combinations of Li and Mn concentrations. The Tc values show a saturation at ~50 K for x=0.1–0.15, whereas the ordered moment values per Mn atom show a maximum for low Mn concentrations and decreases monotonically with increasing x. This is presumably due (partly) to an effect of the antiferromagnetic coupling for Mn pairs in the nearest neighbour Zn sites. The lack of history dependence in M and a very thin hysteresis loop in Figure 1g both indicate a very small coercive field (30–100 Oe), which is promising for spin manipulation. In systems with x≤0.1, the magnetic susceptibility χ at high temperatures above Tc exhibits a simple history-independent Curie–Weiss behaviour with the effective paramagnetic moment values of 5.0–6.0 Bohr magnetons per Mn, as expected for fully magnetic individual Mn2+ moments.

Table 1 Ferromagnetic transition temperature and ordered moment size.


Results of transport studies are shown in Figure 2a–d. Semiconducting behaviour of resistivity of undoped LiZnAs (Fig. 2a) changes to T-independent behaviour for Li deficient systems, wheras much smaller resistivity and definite metallic behaviour were found for Li excess samples with y≥0.05. Figure 2b shows the resistivity curves of the Li1.1 systems with x=0, 0.02, 0.03, 0.05 and 0.10. The resistivity increases monotonically with increasing Mn concentration, suggesting that Mn acts as a scattering centre. The onset of magnetic order reduces this scattering rate, as can be seen in the negative magnetoresistance of Li1.1Mn0.1 system in Figure 2c below Tc.

Figure 2: Results of transport measurements.
figure 2

Transport results of sintered polycrystalline specimens of Li1+y(Zn1−xMnx)As: (a) resistivity of Li1+yZnAs, showing metallic behaviour of Li deficient (y<0) and Li excess (y>0) compounds. (b) resistivity of Li1.1(Zn1−xMnx)As, showing the effect of increasing charge scattering with increasing Mn concentration x. (c) resistivity of Li1.1(Zn0.9Mn0.1)As in various external field H, which exhibits negative magnetoresistance below Tc~50 K. (d) Hall resistivity of Li1.1(Zn0.95Mn0.05)As at T=2 K, which exhibits p-type carriers with concentrations of n~1020 cm−3 together with the anomalous Hall effect due to spontaneous magnetization at H=0.

Because an excess Li (positive y) at an interstitial site is expected to serve as an electron donor, we initially expected n-type carriers for systems with y≥0.05. However, the Hall effect measurements clearly showed p-type carriers for Li excess systems with all the Mn concentrations ranging from x=0.0–0.15. Figure 2d shows representative Hall results for the Li1.1 Mn0.05 system, which exhibits anomalous Hall effect at H=0 below Tc and p-type carrier concentration n~1020 cm−3. The p-type carrier for excess Li can be expected when the excess Li occupies substitutional Zn sites, and the substituted Zn atoms either escape from the system or remain neutral without ionization. This possibility is supported by flattening of the lattice constant with increasing y in Figure 1e and chemical analyses of the end products that indicate that Zn concentrations are lower than the starting nominal concentrations by about 5–10% for Li excess systems with y~0.05–0.1.

Magnetic characterization

Earlier μSR studies of (Ga,Mn)As (ref. 12) demonstrated that ferromagnetism develops in the full volume fraction with sharp transitions for films made by proper heat treatments. To examine volume fraction and the ordered moment size, μSR measurements were performed on Li1.1(Zn0.95Mn0.05)As at TRIUMF in Vancouver, Canada. The time spectra observed in zero field (Fig. 3a) clearly exhibit an increase of the relaxation rate below T~25 K (Fig. 3b), and measurements in longitudinal fields confirmed that this is due to static magnetic order. The volume fraction of regions with static magnetism (Fig. 3c) indicates a sharp transition at Tc~25–30 K, and magnetic order with the full volume fraction achieved below T~10 K. Although the transition is very sharp, some coexistence of paramagnetic and ordered volumes can be seen between T=15–25 K. Future studies on fabrication methods and heat treatments might help improve spatial homogeneity of the transition. Figure 3d compares the present result on Li1.1(Zn0.95Mn0.05)As with those of (Ga,Mn)As with various different Mn concentrations12 in a plot of the low-temperature relaxation rate versus Tc. A good agreement of the ratio of these two parameters suggests that ferromagnetism of these two systems are very similar in terms of the ordered moment size and concentration (represented by the relaxation rate) and the magnitude of ferromagnetic exchange interaction (which determines Tc). See Methods for details of the μSR data analyses.

Figure 3: Results of muon spin relaxation measurements.
figure 3

Results of μSR measurements in sintered polycrystalline specimens of Li1.1(Zn0.95Mn0.05)As: (a) time spectra in zero field that exhibit onset of extra relaxation below T~30 K. The solid lines represent fits to the relaxation function for dilute spin systems in zero field for the static case (often used for dilute-alloy spin glasses13), which exhibits a fast relaxation, plus a non-relaxing paramagnetic component (Methods); (b) the relaxation rate a of the signal that exhibits fast relaxation; (c) the volume fraction of the magnetically ordered region, derived from the amplitude of the fast relaxing signal; (d) comparison between the present results (red symbol) and those from (Ga,Mn)As12 in a plot of the relaxation rate (which is proportional to the individual ordered moment size multiplied by the moment concentration) versus Tc (which is a measure of the effective average ferromagnetic interaction). A factor 4/3 is multiplied to the parameter a to adjust the difference from the simple exponential decay rate Λ adopted in ref. 12. The good agreement implies that the plotted systems have common mechanisms for their ferromagnetism.


Availability of bulk crystal specimens is a major advantage of Li(Zn,Mn)As over existing ferromagnets based on III–V semiconductors. This already enabled us to start NMR measurements and obtain promising preliminary results, and plan neutron scattering studies to study spin dynamics. As demonstrated in Figure 1a–c, square lattice As layers are commonly included in cubic crystals of semiconducting LiZnAs and ferromagnetic Li1+y(Zn1−xMnx)As and tetragonal crystals of antiferromagnetic LiMnAs and superconducting LiFeAs (refs 14,15,16). The difference of the diagonal As–As distances is within about 10%, and even better lattice matching may be expected by replacing Li with other alkali metal elements such as Na and/or replacing As with other pnictogen elements such as P and Sb. This feature could open possibilities of making junctions between various combinations of these systems through the As layers in efforts to develop spintronics devices showing novel functionalities. Thus, we hope that the present demonstration of ferromagnetism in Li(Zn,Mn)As could lead to a new generation of research on magnetic semiconductors and relevant devices.


Analysis of muon spin relaxation spectra

For the analysis of the zero-field (ZF) μSR time spectra in polycrystalline sample of Li1.1(Zn0.95Mn0.05)As, we employed the two component function

is the relaxation function for a static magnetic field with Lorentzian distribution:

where γμ=2π×1.3554×104 s−1 G−1 is the gyromagnetic ratio of the muon spin, and a/γμ represents a field amplitude for the half-width at half-maximum. The Lorentzian field distribution is expected for dilute Mn moments randomly substituting Zn sites. was given by Kubo17 as:

and observed in dilute-alloy spin glasses.13 We assumed that β is a temperature-independent parameter. The first and second terms of equation (1) represent the magnetically ordered and paramagnetic volumes, respectively. The ZF spectra above 30 K can be fitted to the second term of equation (1) without the first term. The ZF spectra below 10 K can be fitted to the first term alone. As shown in Figure 4, however, both terms are needed to fit spectra in the temperature region between 15 and 25 K, suggesting coexistence of the paramagnetic and magnetically ordered volumes.

Figure 4: μSR time spectra in zero field.
figure 4

ZF μSR time spectra in Li1.1(Zn0.95Mn0.05)As at T=20 K (open circles). The solid line represents the best fit to equation (1). The black and green broken lines show the first and the second terms, respectively, of this fitting function.

The series expansion for equation (3) in terms of at is

The Maclaurin series for an exponential decay function Gz(t)=exp(−Λt), which was employed for the analysis of ZF-μSR spectra in (Ga,Mn)As,12 is

By comparing the first-order terms of equations (4) and (5), we notice that Λ corresponds to .

Figure 5 shows μSR spectra in a weak transverse field (WTF) of 30 Oe. The WTF μSR measurements provide direct information of the magnetic volume fraction18 If the internal magnetic field is much larger than the applied field, the oscillation amplitude can be an indicator of the paramagnetic volume. The good agreement of the ordered fraction derived from the measurements in ZF and those in WTF, shown in Figure 3c, supports the validity of our analysis of ZF-μSR spectra, using the two-component function.

Figure 5: μSR time spectra in the WTF.
figure 5

μSR time spectra in the WTF of 30 Oe in Li1.1(Zn0.95Mn0.05)As. The oscillation amplitude corresponds to the paramagnetic volume faction.

Additional information

How to cite this article: Deng, Z. et al. Li(Zn,Mn)As as a new generation ferromagnet based on a I-II-V semiconductor. Nat. Commun. 2:422 doi: 10.1038/ncomms1425 (2011).