Li(Cd,Mn)P: a new cadmium based diluted ferromagnetic semiconductor with independent spin & charge doping.

We report a new diluted ferromagnetic semiconductor Li1+y(Cd,Mn)P, wherein carrier is doped via excess Li while spin is doped by isovalence substitution of Mn2+ into Cd2+. The extended Cd 4d-orbitals lead to more itinerant characters of Li1+y(Cd,Mn)P than that of analogous Li1+y(Zn,Mn)P. A higher Curie temperature of 45 K than that for Li1+y(Zn,Mn)P is obtained in Li1+y(Cd,Mn)P polycrystalline samples by Arrott plot technique. The p-type carriers are determined by Hall effect measurements. The first principle calculations and X-ray diffraction measurements indicate that occupation of excess Li is at Cd sites rather than the interstitial site. Consequently holes are doped by excess Li substitution. More interestingly Li1+y(Cd,Mn)P shows a very low coercive field (<100 Oe) and giant negative magnetoresistance (~80%) in ferromagnetic state that will benefit potential spintronics applications.

Introduction

Diluted ferromagnetic semiconductors have received extensive attention because of their potential applications for spintronics devices[1-3]. For the prototypical III-V diluted ferromagnetic semiconductors, such as (Ga,Mn)As, substitution of divalent Mn2+ into trivalent Ga3+ results in coupled spin & charge doping, which makes individual control of spin and charge difficult[4]. Additionally, the heterovalent substitution of Mn2+ into Ga3+ also leads to severely limited chemical solubility, and results in the specimens only available as thin films and sensitive to preparation methods and annealing treatments[5,6]. The coupled spin and charge is an obstacle not only for fundamental understanding of ferromagnetic mechanism but also for effective improvement of controllable Curie temperature (T).

Recently a series of new generation diluted ferromagnetic semiconductors, e.g. “111” type Li(Zn,Mn)As and Li(Zn,Mn)P, “122” type (Ba,K)(Zn,Mn)2As2 and “1111” type (La,Ca)(Zn,Mn)SbO, have been discovered to overcome the aforementioned difficulties[7-16]. In these new types of diluted ferromagnetic semiconductors spin is doped via isovalent substitution of (Zn2+,Mn2+), while charge is provided by off-stoichiometry of Li in the “111” type diluted ferromagnetic semiconductors or by heterovalent substitution of cations in the “122” and “1111” diluted ferromagnetic semiconductors. With the advantage of independent spin and charge doping, a record of controllable T of 230 K is achieved in (Ba,K)(Zn,Mn)2As2[10,17]. Furthermore, a number of progresses of these new generation diluted ferromagnetic semiconductors have been made on both fundamental studies and potential applications[18-22]. Large size single crystals and single-phase thin films of (Ba,K)(Zn,Mn)2As2 have been grown[23-25]. Based on single crystal samples, a Mn-impurity band is identified by angle-resolved resonance photoemission spectroscopy, demonstrating strong hybridization between Mn 3d- and As 4p-orbitals[26]. The magnetic pair distribution function measurements discover presence of robust nearest-neighbor ferromagnetic alignment of Mn spins along the c axis even well above T[27]. Andreev reflection junction is fabricated with single crystal (Ba,K)(Zn,Mn)2As2. The obtained spin polarization rate of 66% is comparable to the prototypical diluted ferromagnetic semiconductors, suggesting large potentials of application in these new generation diluted ferromagnetic semiconductors[28-30].

Improvement of T is always a fundamental issue for diluted ferromagnetic semiconductor materials. As well known, T depends on the p-d exchange between carriers and Mn ions. Generally, greater p-d exchange can be reached by shortening bond length in Mn-ligand. However among the “111” diluted ferromagnetic semiconductor materials, Li(Zn,Mn)P has lower hole concentration and lower T than those of Li(Zn,Mn)As. Generally, ferromagnetism in a diluted magnetic semiconductor is mediated by itinerated carriers. Thus a higher T is expected with further carrier doping in Li(Zn,Mn)P. However, further improvement of carrier concentration by changing the Li concentration is invalid in Li(Zn,Mn)P. To overcome the difficulty we make a new diluted ferromagnetic semiconductor compound Li(Cd,Mn)P, where the extended Cd 4d-orbitals lead to more itinerant characters of Li(Cd,Mn)P than that of analogous Li(Zn,Mn)P. Li(Cd,Mn)P is expected to have larger hole concentration and consequently higher T than Li(Zn,Mn)P. In this article we report the synthesis and characterizations of the Cd-based diluted ferromagnetic semiconductor, Li(Cd,Mn)P.

Results and Discussion

Crystal structure

Figure 1(a) shows powder X-ray diffraction (XRD) patterns for the samples Li1.1(Cd1−xMnx)P (x = 0.025, 0.05, 0.075 and 0.1). Parent phase LiCdP and doped samples Li(CdMn)P crystallize into a zinc-blende like structure with the space group of F-43m, as shown in the inset of Fig. 1(a). The lattice parameter a = 6.089(2) Å for parent phase LiCdP is consistent with the previous report[31] and larger than LiZnAs (a = 5.940(2) Å) and LiZnP (a = 5.756(1) Å).

Figure 1

The structure characterizations. (a) Powder XRD patterns of Li1.1(Cd1−xMnx)P with various Mn, inset: crystal structures of cubic LiCdP. (b) Lattice constant of Li1.1(Cd1−xMnx)P (blue symbols, bottom horizontal axis) and Li1+yCdP (red symbols, top horizontal axis).

Homogeneity and real atomic ratios are studied with energy dispersive analysis (EDX). Except the light element Li, the atomic ratios of Cd, Mn and P are measured. The results show that all the detected elements are well distributed and their ratios are consistent with the nominal compositions (see Supplementary Fig. S1 and Tables S1 and S2). Thus the nominal compositions are used in this article for simplification. The high resolution transmission electron microscope (TEM) measurements were performed on a typical sample, the heaviest doped sample Li1.1Cd0.9Mn0.1P. The results don’t show any trace of defect or cluster (Fig. S3). The changes of lattice parameters with varying Li and Mn concentration in Li1+y(Cd1−xMnx)P are shown in Fig. 1(b). The monotonic changes of lattice parameter with increasing Li and Mn suggest successful chemical doping. The lattice shrinks with Mn substitution due to smaller radius of Mn2+ (0.66 Å) than that of Cd2+ (0.78 Å)[32]. The decrease of lattice parameter with increasing Li concentration will be discussed later.

Magnetic properties

Parent phase LiCdP is diamagnetic. Single excess Li- and Mn-doping just makes the compound Pauli paramagnetic and paramagnetic (see Supplementary Fig. S2). Ferromagnetism emerges only in Li and Mn co-doped compounds in which excess Li and (Zn,Mn) substitution provide carrier and spin, respectively. The temperature dependence of magnetization (M(T))for Li1.1(Cd1−xMnx)P present upturns on lowering temperature, clear signatures of ferromagnetic transition, as shown in Fig. 2(a). Note that no visible difference is detected between field cooling (FC) and zero field cooling (ZFC) modes. Among all possible impurities, only Mn3O4 has a comparable T = 42 K. In contrary, T of Li(Cd,Mn)P compounds changes from 15 K to 45 K according to their composition, ruling out the possibility of Mn3O4. The hysteresis curve (M(H))of Li1.1(Cd0.925Mn0.075)P at T = 6 K is plotted in Fig. 2(b) as a typical example. It exhibits ferromagnetic behavior with a small linear field-dependent component, which should be due to remaining paramagnetic spins[9]. Because Mn cations are randomly distributed in material, some Mn cations which locate far away from each other cannot be mediated by carries to form long range ferromagnetic order and consequently lead to local paramagnetic spins. The inset of Fig. 2(b) shows a small coercive field (H < 100 Oe) that is promising for spin manipulation. After subtracting the linear field-dependent component, the magnetic hysteresis loops of Li1.1(Cd1−xMnx)P (x = 0.025, 0.05, 0.075 and 0.10) specimens at 6 K are shown in Fig. 2(c). Because the magnetizations doesn’t saturate even up to 1 T, after subtraction the magnetizations at 6 K and 1 T (M6K,1T) is used to approximately represent saturation magnetizations in following discussions. The M6K,1T = 0.94, 0.83, 0.69, 0.62 μB/Mn for samples of x = 0.025, 0.05, 0.075 and 0.10 respectively. They are smaller than that of (Ga,Mn)As[4] and Li(Zn,Mn)As[7] possibly due to the competition between the long range ferromagnetic ordering and the short range antiferromagtic interactions. The tendency that the M6K,1T reduces with increasing Mn doping levels has also been found in many magnetic ions doped materials[33]. One rational reason is the competition between ferromagnetic interaction of Mn mediated by carriers and antiferromagnetic coupling of Mn pairs in the nearest neighbor sites, as discussed in “111” and “122” diluted ferromagnetic semiconductors[7-9,12].

Figure 2

Magnetism of Li1.1(Cd1−xMnx)P with x = 0.025–0.10. (a) M(T) under applied field H = 2 kOe with ZFC and FC procedures. (b) M(H) of Li1.1(Cd0.925Mn0.075)P at T = 6 K, measured under applied field H up to 1 T, showing hysteresis loop before and after subtraction of linear field-dependent component. Inset: hysteresis in small field regions. (c) M(H) curves after subtraction of the field-dependent component of Li1.1(Cd1−xMnx)P with x = 0.025–0.10. Inset: corresponding M(H) curves in small field regions.

In the paramagnetic region, the susceptibility can be well fitted by the Curie-Weiss law, as shown in Fig. 3(a). The obtained Weiss temperature is 17.8(2) K, demonstrating the ferromagnetic interaction between Mn. To precisely determine the T, Arrott plot method is performed. In Fig. 3(b), H/M versus M2 is plotted over the temperature range of 5–65 K. The isotherm at the Curie point is supposed to be a straight line passing through the origin. In this way, T is determined as 45 K for Li1.1(Cd0.9Mn0.1)P. The T for the other samples are obtained in similar way. Table 1 lists T and M6K,1T for various compositions of Li1+y(Cd1−xMnx)P. When Mn concentration is fixed, one can note that excess Li initially improves both T and M6K,1T within low Li concentration but then suppresses ferromagnetic order when y = 0.2. On the other hand, except samples with overdoped Li (y = 0.2), increasing Mn enhances the T but declines the average local moments on Mn (M6K,1T).

Figure 3

The ferromagnetic ordered temperature of the diluted magnetic semiconductor. (a) Inverse susceptibility dependence of temperature for Li1.1(Cd0.9Mn0.1)P (black line) and Curie-Weiss fit (red line). (b) Arrott plots at various temperatures above and below Tc for Li1.1(Cd0.9Mn0.1)P, shows the ferromagnetic transition at Tc = 45 K.

Table 1

The Curie temperature (T) and M6K,1T to approximately represent saturation magnetizations of Li1+y(Cd1−xMnx)P.

TC (K)/M6K,1T (µB/Mn)	y = 0.05	y = 0.1	y = 0.2
x = 0.025	23/0.84	21/0.94	15/0.71
x = 0.05	27/0.72	27/0.83	16/0.37
x = 0.075	31/0.60	35/0.69	16/0.18
x = 0.10	36/0.56	45/0.62	15/0.26

Electrical transport properties

Figure 4(a) shows the temperature-dependent resistivity (ρ(T)) for a series of Li1+yCdP (y = 0, 0.05 and 0.1). Consistent with ref.[31], the parent compound LiCdP present semiconducting conduction[31]. Excess Li doping decrease resistivity, suggesting the effective doping of carriers. Figure 4(b) shows that at the entire temperature range the resistivity of Li1.1(Cd1−xMnx)P increases with increasing Mn concentration, caused by localization effect. Although carriers introduced by excess Li doping are originally itinerant, they will be weakly bound to the Mn local spin moments and then partially lose their mobility[34]. The similar conducting behavior has been observed in (Ba,K)(Zn,Mn)2As2. The localization of the carriers has been confirmed by Mn K β x-ray emission spectroscopy measurements which indicates that Mn local spin moments traps holes and in turn is declined by hole doping[20].

Figure 4

Transport properties of Li1+y(Cd1−xMnx)P: (a) ρ(T) of Li1+yCdP with y = 0, 0.05 and 0.1. (b) ρ(T) of Li1.1(Cd1−xMnx)P with x = 0, 0.025, 0.05, 0.075 and 0.1. (c) Magnetoresistance of Li1.1(Cd0.9Mn0.1)P at different temperatures. (d) Correlation between T and the hole concentration for several “111”, “122” new types of diluted ferromagnetic semiconductors and typical III-V diluted ferromagnetic semiconductors.

A giant negative magnetoresistance (MR) is observed in all of the ferromagnetic Li(CdMn)P samples. Figure 4(c) shows the MR of Li1.1(Cd0.9Mn0.1)P specimen as a typical example. The value of negative MR (about −80%) is at least twice larger than those of analogues Li(Zn,Mn)As and Li(Zn,Mn)P[7,8]. The negative MR may result from reduction of spin-dependent scattering by aligning the spins under applied field. The Hall effect measurements of all the samples show p-type carriers. Hole concentration (n) of the parent phase is ~1017 cm−3 at 200 K. More holes are doped by excess Li substitution in Li1.1CdP which has n = 3.8 × 1019 cm−3 at 200 K. In ferromagnetic sample Li1.1(Cd0.95Mn0.05)P, 5% Mn doping slightly decreases to n = 2.7 × 1019 cm−3 at 200 K. At lower temperature the resistivity was too large, any small misalignment of the two Hall contacts picks up a longitudinal resistivity signal and this leads to great difficulty in the Hall effect measurement (see Supplementary Fig. S4). The relationships between hole concentration and Curie temperature of “111” diluted ferromagnetic semiconductors and other diluted ferromagnetic semiconductor systems are plotted in Fig. 4(d)[2,7-10,35]. From Li(Zn,Mn)P to Li(Cd,Mn)P, the hole concentration is considerably increased, and consequently the T is improved from 34 K to 45 K. As the Zener model predicted, the ferromagnetism is mediated by carriers, and the Curie temperature is positive correlated with hole concentration.

Theoretical analysis

In order to have insight into the origin of hole carrier, we performed calculation on electronic structures with density functional theory (DFT). Calculation shows that the band structure of LiCdP is very similar to that of LiZnP. The band structure of parent phase LiCdP is shown in Fig. S5. The obtained direct energy gap is 0.59 eV. With the quantum Monte Carlo (QMC) simulation of the Anderson impurity model, the impurity band level of Mn is determined as −0.35 eV.

Two possible different sites for excess Li are discussed in the calculations, (i) the interstitial site LiI and (ii) the Cd-substitutional site LiCd. The former can provide n-type carrier and the latter will serve as a hole donor. We calculate the formation energy for the two excess Li-sites, respectively. Since Mn at Cd-substitutional site MnCd does not introduce any carriers, we study the excess Li in Li1+yCdP for simplification. According to previous work[36], the formation energy is given by Eformation = ET − nLiμLi − nCdμCd − nPμP, where ET is the total energy of the supercell, nx is the number of x atoms in the supercell, and μx is the atomic chemical potential. It has μLi + μCd + μP = μLiCdP(bulk). Table 2 shows formation energy for two extreme conditions, i.e., the Li-rich plus Cd-rich limit (μLi = μLi(bulk), μCd = μCd(bulk)) and the Li-rich plus P-rich limit (μLi = μLi(bulk), μP = μP(bulk,black)). Under both conditions, compounds with LiCd have lower formation energy. The experimental condition must be between these two extreme conditions. It recalls the reduction of lattice parameter with excess Li doping found in XRD measurements. The excess Li at interstitial site is supposed to stretch the lattice. In contrary, substitution of Li into Cd should shrink the lattice due to the smaller Li+ radius (0.59 Å) than Cd2+ (0.78 Å)[32]. Thus we argue that in Li(Cd,Mn)P the excess Li atoms prefer to occupy Cd-substitutional sites LiCd, and thus create the p-type carriers.

Table 2

Formation energy for excess Li atom at different sites, obtained by DFT calculations[36].

LiCdP with excess Li	Formation energy (Li-rich and Cd-rich limit)	Formation energy (Li-rich and P-rich limit)
Interstitial Li (supercell Li28Cd27P27)	2.13 eV	2.13 eV
Li at Cd site and Cd is removed (supercell Li28Cd26P27)	0.22 eV	−1.14 eV

Conclusions

In this work, “111” type Cd-based Li(Cd,Mn)P has been designed to achieve high Curie temperature. In Li(Cd,Mn)P replacement of Zn by Cd successfully compensates for the high hole binding energy in the Li(Zn,Mn)P by increasing Mn-P bond length. As a result, Li(Cd,Mn)P has enlarged the hole concentration, amplified the effective p-d exchange and more importantly a Curie temperature of 45 K which is one third higher than T of Li(Zn,Mn)P. In addition, the observed properties in Li(Cd,Mn)P, such as the low coercive field and the giant negative magnetoresistance, are favorable for future applications. In short, the successful prediction and fabrication of Li(Cd,Mn)P open an new boulevard to tailor the ferromagnetism in diluted magnetic semiconductors.

Method

Polycrystalline specimens of Li(Cd,Mn)P were prepared by solid state reaction with high purity elements. The stoichiometric ratios of starting materials were well mixed and pressed into pellets. All the processes were performed under the protection of high-purity Argon due to the air sensitivity of precursors and products. The pellets were sealed in Ta tubes under 0.5 bar of Argon before being sealed into evacuated quartz tubes. The samples were heated at 470 °C for 48 h. Then the products were reground and sintered at 680 °C for 48 h, followed by a quick quenching to room temperature. The recovered samples were characterized by X-ray powder diffraction (XRD) with a Philips X’pert diffractometer using CuKα radiation. Real compositions of the heavy elements (i.e. Cd, Mn and P) were determined by using energy dispersive analysis (EDX) on a commercial Scanning Electron Microscope (SEM). Microstructure was studied by high resolution Transmission Electron Microscope (TEM). The dc magnetic properties were examined by using Superconductivity Quantum Interference Device (SQUID, Quantum design), and transport properties were examined by Physical Property Measurement System (PPMS, Quantum design). We calculated the electronic structures by using the density functional theory (DFT) implemented in the code QUANTUM ESPRESSO[36]. The exchange-correlation interactions are described by the Perdew-Burke-Ernzerhof generalized gradients approximation (GGA), and the electronion interactions are represented by the Vanderbilt ultrasoft pseudopotentials.

The Supporting Materials