Due to the potential applications in spintronics devices, diluted magnetic semiconductors (DMSs) that simultaneously combine the spin and charge degrees of electrons have received extensive attentions^[1-6]. In 1990s, the III–V DMS (Ga,Mn)As was successfully prepared by low temperature molecular beam epitaxy (LT-MBE) method. The Curie temperature TC has reached as high as ~200 K^[7-9], which is much higher than those of some II–VI DMSs, such as (Zn,Mn)Te^[10] and (Cd,Mn)Te^[11]. To explain the magnetic mechanism of III–V DMSs, Dietlet al. proposed a mean-field Zener's model that described the ferromagnetic interactions between the local magnetic moments and carriers, and predicted that TC of some DMSs may reach room temperature if more Mn atoms are doped^[12,13]. Subsequently, Reedet al. reported that TC of (Ga,Mn)N was up to ~370 K^[14]. The application of DMSs in spintronic devices will be practical once the Curie temperature of some DMSs are above room temperature. However, the development of III–V DMSs has also encountered some limitations. For example, in (Ga,Mn)As, magnetic atoms Mn easily enter the interstitial positions of the lattice, which makes it difficult to accurately determine the actual amount of doped Mn atoms. Meanwhile, the substitutions of Mn for Ga introduce magnetic moments and carriers simultaneously, which constrains us to study their individual contribution to the formation of the ferromagnetic ordering. In addition, films cannot be used for microscopic probes that are based on bulk materials, such as nuclear magnetic resonance (NMR), neutron scattering and muon spin relaxation (μSR)^[5,15-17]. On that account, the study of bulk DMSs with decoupled charge and spin doping are worthy of attentions.

Recently, many bulk Zn-based DMSs derived from Fe-based superconductors have been reported. Foremost, according to the pioneer theory work of Maseket al.^[18], Denget al.^[16] prepared a new 111-type DMS Li(Zn,Mn)As, which is isostructural to Fe-based superconductor LiFeAs^[19]. This is the first bulk DMS that carriers and magnetic moments could be controlled independently. In I–II–V semiconductor LiZnAs, doping excessive Li introduces carriers and Mn substitutions for Zn introduce local moments, and the Curie temperature TC is ~50 K. Until now, the highest Curie temperature recorded is a 122-type DMS (Ba,K)(Zn,Mn)₂As₂ with TC up to ~230 K^[20,21], which has surpassed that of (Ga,Mn)As. In addition, Dinget al.^[17] synthesized a 1111-type DMS (La,Ba)(Zn,Mn)AsO with TC up to ~40 K.μSR measurements have proved that the ferromagnetic ordering of these DMSs is intrinsic rather than caused by magnetic clusters or impurities. Moreover,μSR measurements reveal that Li(Zn,Mn)As, (Ba,K)(Zn,Mn)₂As₂ and (La,Ba)(Zn,Mn)AsO share the same mechanism of magnetic origin as that of (Ga,Mn)As^{[16,17,20,22]}. That is to say, searching for more new DMSs materials will help us to understand the general mechanism of magnetic ordering in all DMSs, and direct us to optimize the choices of materials.

In this paper, we report the successful fabrication of a new 1111-type magnetic semiconductor, (Ba_1−xNa_x)F(Zn_1−xMn_x)Sb (x = 0.05, 0.075, 0.1, 0.125, 0.15, and 0.175), via doping Na into Ba sites and Mn into Zn sites in the parent compound BaFZnSb to introduce hole carriers and local magnetic moments, respectively. BaFZnSb shares the same crystal structure with that of LaOZnAs and LaOCuS, which have been reported to be the parent semiconductors of 1111-type DMSs (La,Ba)O(Zn,Mn)As ( TC ~ 40 K)^[17] and (La,Sr)O(Cu,Mn)S ( TC ~ 200 K)^[23], respectively. Comparing with oxides, the ionic radius of fluoride is smaller and its electronegativity is stronger, which results in constituting stronger ionic bonds^[24]. The Weiss temperatureθ of (Ba_1−xNa_x)F(Zn_1−xMn_x)Sb is up to ~ 16 K forx = 0.175, which is a sign of ferromagnetic interaction, and followed by a magnetic glassy transition below Tf ~ 14 K.

We synthesized the polycrystalline specimens of (Ba_1−xNa_x)F(Zn_1−xMn_x)Sb through the solid-state reaction method. High-purity starting materials Ba (99.2%, Alfa Aesar), BaF₂ (99.99%, Aladdin), Na (99.7%, Aladdin), Zn (99.9%, Alfa Aesar), Mn (99.95%, Alfa Aesar) and Sb (99.999%, Prmat) were mixed according to stoichiometric proportions and were placed in alumina crucibles before sealing in evacuated silica tubes. The mixture was heated slowly to 200 °C and held for 10 h, then heated to 750 °C and held for another 10 h, followed by furnace cooling to room temperature. After that, the intermediate products were grounded, pressed into pellets, placed in alumina crucibles, sealed in evacuated silica tubes again and then reheated to 750 °C for 30 h. To protect the contamination from air and H₂O, all operations except the sealing of silica tubes were executed in a glove box filled with high-purity Ar (O₂ and H₂O < 0.1 ppm).

The crystal structures of the samples were characterized at room temperature by a powder X-ray diffractometer (Model EMPYREAN) using monochromatic Cu-K_α1 radiation withλ(K_α1) = 1.540598 Å. The detailed information of lattice constants and unit cell volume was calculated by Rietveld refinement method using the GSAS-II software package^[25]. The DC magnetization measurements were performed on a quantum design magnetic property measurement system (MPMS). The AC susceptibility measurements were measured on a quantum design physical property measurement system (PPMS). The electrical resistivity measurements were conducted on sintered pellets by the four-probe method.

InFig. 1(a), we show the powder X-ray diffraction patterns of polycrystals (Ba_1−xNa_x)F(Zn_1−xMn_x)Sb withx = 0.00, 0.05, 0.075, 0.1, 0.125, 0.15, and 0.175, respectively. The Bragg peaks can be well indexed by a ZrSiCuAs-type tetragonal structure with space group P4/nmm^[26], which is isostructural to the 1111-type superconductor La(F,O)FeAs^[27], indicating that Na substitution for Ba and Mn substitution for Zn have no influence on the tetragonal crystal structure. InFig. 1(b), we show the Rietveld refinement result of (Ba_0.875Na_0.125)F(Zn_0.875Mn_0.125)Sb. The resultant weighted reliable factor Rwp is 7.962%, indicating the samples are in good quality. With doping levelx increasing, some small peaks of ZnSb impurities marked as * inFig. 1(a) were observed. ZnSb impurities are non-magnetic, thus they will not affect the discussion of magnetism in the following.

As shown inFig. 1(b), the parent compound BaFZnSb has a layered crystal structure, which consists of two layers in the ab plane: one is [BaF]⁺ layers (BaF₄ tetrahedra) and the other is [ZnSb]^– layers (ZnSb₄ tetrahedra)^[28]. These layers are conventionally regarded as the tetragonal fluorite and anti-fluorite structure types separately^[28], and they are stacked alternately along thec-axis. The lattice parameters obtained from Rietveld refinement are shown inFig. 1(c). The lattice parameters of parent compound BaFZnSb area = 4.43605 Å andc = 9.81786 Å, which is close to previous reported values ofa = 4.4384 Å andc = 9.7789 Å^[28]. Meanwhile, with doping levelx increasing,a monotonically increases from 4.43720 to 4.44271 Å, andc continuously decreases from 9.80767 to 9.78259 Å, repectively. This should be ascribed to the fact that the ionic radius of Na⁺ (1.02 Å) is smaller than that of Ba²⁺ (1.35 Å), but the ionic radius of Mn²⁺ (0.83 Å) is larger than that of Zn²⁺ (0.74 Å). The monotonical behaviours of lattice parametersa andc with doping levelx demonstrate the successful doping of Na into Ba sites and Mn into the Zn sites, respectively. InFig. 1(d), we find that the unit cell volume doesn't vary monotonically with increasing doping levelx. This is due to the fact that lattice parametersa andc behave in opposite direction.

We show DC magnetization (M) measurements of the parent phase BaFZnSb and (Ba_0.95Na_0.05)FZnSb under field-cooling condition with an external field Bext = 100 Oe inFig. 2(a). We find that the magnetization for both samples is very small, ~10^–8 emu/mg, confirming the characteristic of paramagnetism. By using the Curie-Weiss law to fit the data, we get the values of Weiss temperatures as −0.7 and −0.6 K, respectively. This also demonstrates that both BaFZnSb and (Ba_0.95Na_0.05)FZnSb are in paramagnetic ground state. That is to say, doping only carriers can not introduce any type of magnetic phase transition. This is similar to the situation in SrF(Zn,Cu)Sb^[29], where doping only Cu into SrFZnSb does not induce any magnetic ordering.

InFig. 2(b), we show the temperature-dependent magnetization (M) in both zero-field-cooling (ZFC) and field-cooling (FC) procedures under an applied external field Bext = 100 Oe for (Ba_1−xNa_x)F(Zn_1−xMn_x)Sb (x = 0.05, 0.075, 0.1, 0.125, 0.15 and 0.175). There is an obvious increase for the magnetic moment at low temperature. Also, ZFC and FC curves abruptly split at a temperature defined as Tirr (as marked by the arrow in the inset ofFig. 2(b)) for all doping levels, which is related to the pinning of the domain wall^[30]. InFig. 2(b), for (Ba_0.9Na_0.1)F(Zn_0.9Mn_0.1)Sb, the split between FC and ZFC curves is at Tirr ~ 16 K, and the maximum of ZFC curve is at Tf ~ 13 K. On the other hand, the magnetic moments ( Mbase) at base temperature of 2 K under FC mode increase from 0.0038 μB/Mn forx = 0.05 to 0.0112 μB/Mn forx = 0.10, and then decrease to 0.0092 μB/Mn forx = 0.15. In the inset ofFig. 2(b), we can clearly observe the variation of magnetic moments under low temperature for specimens withx = 0.05, 0.075, 0.1, 0.125, 0.15 and 0.175, respectively. The increasing Mbase forx ≤ 0.1 may be attributed to the formation and enhancement of ferromagnetic interaction. Mbase decreases forx ≥ 0.125, which might be generated by the competition between local ferromagnetic coupling through the indirect exchange interaction and antiferromagnetic coupling caused by direct exchange interaction existing in Mn atoms at nearest-neighbour (N.N.) Zn sites. This is also a piece of evidence that too many Mn atoms doping is detrimental to the ferromagnetic interaction, as shown in many types of DMSs, such as (La,Sr)(Zn,Mn)SbO^[31] and (Ca,Na)(Zn,Mn)₂Sb₂^[32]. The probability that two Mn atoms are at N.N. sites satisfies P=C41x(1−x)3. For example, for the doping level ofx = 0.1, for a Mn atom finding one Mn atom at its N.N. sites is P=C41×0.1×0.93=29.16%, andP increases to 36.85% forx = 0.15. That is, as the Mn doping levelx continues to rise, the enhanced antiferromagnetic interaction suppresses the ferromagnetic interation.

When the temperature is much higher than the Curie temperature, the magnetic susceptibility generally satisfies the Curie-Weiss law, χ=χ0+C/(T−θ), where χ0 is a temperature-independent component,C is the Curie constant andθ is Weiss temperature. InFig. 2(c), we plot 1/(χ−χ0) versus temperature. The intersections of the fitting lines and thex-axis are the Weiss temperaturesθ. All values ofθ are positive, further identifying the establishment of ferromagnetic interaction between Mn atoms. Moreover, Zn/Mn-Sb bond length is longer than that of Zn/Mn-As, and Sb-Zn/Mn-Sb bond angleα is further away from the idealα value in non-distorted ideal tetrahedron as that of As-Zn/Mn-As^[33]. On that account, compared with that of (Ba,K)F(Zn,Mn)As^[34], the Weiss temperature of our samples is lower. In addition, we also apply the formula C=Nμ0μeff2/3kB to obtain effective magnetic moment μeff, and put these numbers inTable 1.

InFig. 2(d), we present the iso-thermal magnetization at 2 K for (Ba_1−xNa_x)F(Zn_1−xMn_x)Sb (x = 0.05, 0.075, 0.1, 0.125, 0.15 and 0.175). Clear hysteresis loops are observed, which indicates the presence of ferromagnetic interaction. With the increasing of doping levelx, the coercive field HC decreases. In particular, the largest HC is up to ~12 000 Oe, which is also reflected in the large differences between the measurements of temperature-dependent magnetization (M) in ZFC and FC condition at Tirr. The large coercive field HC is similar to that in (La,Ba)(Zn,Mn)AsO^[17] and (Ba,K)F(Zn,Mn)As^[34], but much larger than that in Li(Zn,Mn)As^[16] and Cu₂(Zn,Mn)(Sn,Al)Se₄^[35]. Comparing the crystal structures of two kinds of materials mentioned above, we find that the coercive fields seem to be much larger in crystals with layered structure. We list all parameters obtained above inTable 1.

We observe that the magnetization has not even reached saturation at 2 K under the maximum external magnetic field as high as 5 T inFig. 2(d), which is a characteristic of spin-glass systems. Therefore, we perform AC susceptibility with varying frequenciesf under zero field for (Ba_0.9Na_0.1)F(Zn_0.9Mn_0.1)Sb. The real part χ' and imaginary part χ'′ of the AC susceptibility are shown inFigs. 3(a) and3(b), respectively. As the frequency increases, the cusp temperature, Tf, shifts to higher temperature on both real part χ' and imaginary part χ'′, and the overall magnitude of real part χ' slightly decreases. We apply the equation K=ΔTf/[TfΔlogf] to describe thef-dependence of Tf^[36]. The value ofK obtained inFig. 3(c) is ~ 0.077, which is within the range between 0.005 to 0.08 of conventional spin-glass systems^[37], indicating our samples evolve into spin-glass state below Tf. Similar behaviour has also been observed in other DMSs, such as Ba(Zn,Cu,Mn)₂As₂^[38] and (Ba,K)(Zn,Mn)₂Sb₂^[39]. The frustrated magnetic moments of spin-glass behaviour may be ascribed to the competition between ferromagnetic and antiferromagnetic interactions among magnetic Mn atoms^[40].

Electrical transport measurement is another important characterization method of magnetic semiconductors. InFig. 4, we perform temperature-dependent resistivityρ for (Ba_1−xNa_x)F(Zn_1−yMn_y)Sb (x = 0.00,y = 0.00;x = 0.05,y = 0.00;x = 0.05,y = 0.05;x = 0.075,y = 0.075;x = 0.10,y = 0.10;x = 0.125,y = 0.125;x = 0.15,y = 0.15;x = 0.175,y = 0.175). Both of the parent compound BaFZnSb and (Ba_0.95Na_0.05)FZnSb where only carriers are introduced show metallic behaviour. The semiconducting behaviour is displayed for all samples with substitutions of (Ba,Na) and (Zn,Mn) simultaneously. Remarkably, the resistivity increases monotonically with the doping levelx increasing (except forx = 0.175), similar to that in (La,Ca)(Zn,Mn)SbO^[41]. The main cause for that is that carriers are scattered by magnetic fluctuations caused by doped Mn atoms. Similar phenomenon has also been reported in (Ga,Mn)As^[42].

To summarize, we successfully synthesized a new 1111-type fluoride-antimonide magnetic semiconductor (Ba_1−xNa_x)F(Zn_1−xMn_x)Sb (x = 0.05, 0.075, 0.10, 0.125, 0.15 and 0.175) via Na substitutions for Ba supplying carriers and Mn substitutions for Zn introducing local magnetic moments, respectively. Forx = 0.175, the Weiss temperature is up to ~ 16 K. The iso-thermal magnetization measurements at 2 K for (Ba_1−xNa_x)F(Zn_1−xMn_x)Sb show clear hysteresis loops, and the coercive field HC is as large as 12 000 Oe, similar to that in (La,Ba)(Zn,Mn)AsO^[17] and (Ba,K)F(Zn,Mn)As^[34]. Furthermore, the AC susceptibility measurements show that the system evolves into spin-glass state below Tf. The temperature-dependent resistivity measurements demonstrate that all specimens with Na and Mn codoped display semiconducting behaviour. The Zn-based magnetic semiconductor (Ba_1−xNa_x)F(Zn_1−xMn_x)Sb, Fe-based surperconductor LaFeAsO and antiferromagnet LaMnAsO have the same ZrSiCuAs-type tetragonal structure, which is possible to make multifunctional heterojunctions. The magnetic semiconductors in fluoride compounds have shown some difference from oxide compounds, and provide new ideas for the development of new materials.

The work was supported by the Key R&D Program of Zhejiang Province, China (2021C01002) and NSF of China (No. 12074333).