Introduction

The rapid developments observed recently in the field of spintronics have generated the appearance of new related research topics1, which are based on a variety of traditional or more novel materials such as graphene or topological insulators2. Among those new proposals, there is also the exploration of the current-induced spin transfer or thermal gradient induce spin current effects by using ferromagnetic semiconductors or diluted magnetic semiconductors (DMSs) as an alternative approach to control the system magnetization and the spin current3,4,5. These possibilities have encountered the problem of the operability at room temperature (RT), a common dilemma with other proposals. Although most of pure DMSs are good source of spin polarized carriers6,7, their low ferromagnetic ordering temperature (TC) is a serious shortcoming. Similarly, DMSs containing additional inclusions, such as metallic clusters and inhomogeneities are usually avoided because they add an uncontrolled mechanism of transport and magnetism. On the other hand, the presence of ferromagnetism in DMSs, with a magnetic moment value that depends on cluster size inclusions, shows notable TC increase, which has been reported to be above RT8,9. Furthermore, their self-organization can open a promising path for magnetic recording media due to their large positive magnetoresistance8. In recent reports the possibility of spin polarization due to the presence of magnetic clusters within the sample is under debate10,11. Until now, however, no direct evidence on their impact to the magnetic response has been addressed and therefore the understanding of the role of the magnetic clusters into the magnetization in DMSs requires a different approach.

The clustering formation and presence of inhomogeneities are reported to have a direct influence in the high TC in ferromagnetic chalcopyrite-based materials12,13,14,15,16,17,18. Recently, we have shown that the competing interplay between MnAs micro-size clusters in Mn-doped CdGeAs2 lead to unusual pressure tunable metamagnet-like states19. The direct control of magnetic behavior of clusters via an external pressure in these inhomogeneous systems could lead to striking magnetic phenomena. In particular, among all high-TC ferromagnetic chalcopyrites, Mn-doped CdGeP220 is a prototypical material, where the clusters influence on the magnetic response can be clearly characterized. However, consideration of an external pressure and its influence to the magnetic response of such systems has not been yet considered21.

Here we report our first investigation of ferromagnetic MnP clusters in strongly inhomogeneous Mn-doped CdGeP2, using systematic high-pressure measurements (magnetic, volume and transport). We clearly identify, two RT serial transitions, starting from a ferromagnetic (FM) to non-magnetic (NM) state. We show below that this effect is coming from MnP cluster response to the external pressure. Specifically, for the low pressure regime, the results suggest that an enhancement in the stability of the host structure is highly realized as the concentration of Mn doping goes from x = 0.09 to x = 0.25, as can be seen on Fig. 1 b. For small Mn doping levels, the sample pressure response behaves mostly as the undoped host. While at large doping levels, the magnetic response is largely tuned by the presence of MnP clusters, which have been observed at ambient pressure. To offer a light into our magnetic and volume measurements, we correlate the observed cluster changes with theoretical density functional theory (DFT) calculations. In particular we study the pressure-volume dependence of single orthorhombic MnP crystal as well as the effect of substitutional Mn in the magnetic and structural properties of Cd1−xMnxGeP2.

Figure 1
figure 1

High-pressure physical properties of Cd1−xMnxGeP2.

(a) Magnetic ac susceptibility χac as function of pressure for 0.09 ≥ x ≥ 0.225 are obtained for T = 300 K. Inset shows selected χac vs P at the temperature changes between RT and TC (332 K) for x = 0.225. Note that χac in the CdGeP2 host does not develop any peak with pressure. (b) Relative volume ΔV(P)/V0 vs P measured at RT. Dashed lines show the extrapolation to ΔV(P)/V0 from high to zero pressure to evaluate ωs. (c) resistivity ρ/ρ0 and Hall coefficient RH/RH0 vs P normalized to ambient pressure values for 0.135 ≥ x ≥ 0.225. The absolute values at ambient pressure: ρ = 2.75, 3.02, 2, 0.72, 1530 Ω·cm and RH = 73.1, 20, 10, 3, 5000 cm3·C−1 for x = 0, 0.09, 0.135, 0.19 and 0.225, respectively. A giant hysteresis is observed during decompression for x = 0.135 (inset).

Full size image

Results

Magnetic ac-susceptibility

First of all, RT pressure dependence of the magnetic ac-susceptibility χac for synthetized Cd1−xMnxGeP2 samples (0.09 ≥ x ≥ 0.225) was measured. At the studied pressure range (0–5.5 GPa), two bordered peaks, at PC1 = 2.18–2.4 GPa and at PC2 around 3 GPa, for 0.09 ≥ x ≥ 0.225), were noted (Fig. 1a). The first peak in χac as function of composition show remarkable upturn as the Mn doping concentration increases, with a red shift towards high pressures (see also Fig. S2 in the supplement material). This observation clearly marks up a magnetic transition from FM to NM state, which is largely leaded by the MnP clusters. At the end of this pressure interval, the ac-susceptibility goes back to the state of low pressure. At pressures close to 3 GPa, a second less intensive peak develops, showing a second transition with similar magnetic characteristics as the previous one. When doping concentration is low (for x = 0.09), the peak of χac is negligible, suggesting very weak cluster correlation. Interestingly, both peaks happen to be very sensitive to changes in the temperature range between RT and TC. Inset of Figure 1a displays how the first peak, for x = 0.225, is suppressed gradually while the second disappear completely when temperature is increased up to T = 315 K.

Volumetric and transport properties

Now we focus on the less intense second peak at PC2, which happens around 3 GPa. Consistent in situ volumetric and transport high-pressure measurements performed at RT (Fig. 1b, c) reveal a new unexpected behavior. For example, an abrupt change in the pressure-dependent relative volume ΔV(P)/V0 (Fig. 1b) happens to be quite similar to that reported for Cd1−xMnxGeAs222, as well as for Invar-like bulk materials23,24. Furthermore, a large value of the spontaneous volume magnetostriction of ωs = 0.2 ÷ 0.98% for x = 0.09 ≥ x ≥ 0.225 was obtained from the extrapolation from the high to low-pressure regime, similarly to previous results22. This onset of ωs exposes the rapid destabilization of the FM state under pressure that is related to the lattice compressibility and increase of the bulk moduli (see Fig. S3 in the Supplement material). Whereas the abrupt change in the slope around P ≈ 3.2 GPa in the host sample is related to a structural phase transition driven by pressure. This topic will be further discussed with the help of ab initio calculations.

The discussed trend is clearly reflected in the performed transport measurements. In the host, for low doping Mn concentration (x < 0.09), there are only small changes in the samples resistivity ρ and the Hall coefficient RH up to 3 GPa21, see Fig. 1c. In contrast, we found more dramatic changes (between 3–4 orders of magnitude) in the values of ρ(P) and RH(P) for larger doping concentrations. In particular for 0.135 ≥ x ≥ 0.225 samples, a slightly upper PC2 was found (Fig. 1c), which is a typical characteristic of an insulator-metal transition25. The measured RH in all cases specifies that charge carrier and keep with the same sign with pressure (even in a logarithmic normalized scale). The inset in Figure 1c displays a giant hysteresis which appear during decompression on both transport parameters for the x = 0.135 sample. We stress that this hysteretic behavior emerges only at higher Mn doping concentration (0.135 ≥ x ≥ 0.225). Based on all our measurements, we can suggest that the 3 GPa observed feature corresponds to a second transition from a FM to a NM state.

High-pressure zero-field magnetization

In order to ensure that both transitions are in fact originated from a two serial FM to NM states and their cluster dependences; we provide RT zero-field (ZF) magnetization measurements as function of pressure. In Fig. 2a, unusual step-like behavior of ZF magnetization for x = 0.225 demonstrates two areas located at PC1 = 2.4 GPa and PC2 = 3.3 GPa evidencing well defined consecutive pressure-induced transitions. Both of these values are in well agreement the obtained peaks in χac (Fig. 1a). Within the considered measured pressure range the total decrease in ZF magnetization is not exceeded by 10% and does not go to zero as expected for a cluster embedded in a crystal host8. For more accurate evolution of the magnetic behavior, we measured hysteresis loops at very weak magnetic fields, up to 1000 Am−1 (i.e. maximally achievable by coils). As depicted in Fig. 2b, the hysteresis characteristics of FM state for x = 0.225 starting from ambient up to 2.20 GPa and appearing again at 2.75 GPa (inset of Figure 2b) is clearly demonstrated.

Figure 2
figure 2

High-pressure magnetic measurements.

(a) ZF magnetization for x = 0.225 as function of applied pressure. Dashed line corresponds for appearance of second FM state as it further obtained from minor hysteresis loops vs magnetic field measured at pressures up to 2.45 GPa (b) and up to 3.3 GPa (inset). All measurements were performed at RT regime.

Full size image

DFT calculations

In order to strength our observations we now rely on DFT calculations, as implemented in the Vienna ab initio simulation package (VASP)26 (see Methods, Computational details). Due to the fact that the experimentally measured MnP clusters are embedded in a crystalline system and they are not free, we have tried to understand its magnetic properties by studying the effect of the magnetic configuration changes in the crystalline system as a function of hydrostatic pressure, as well as on the substituted Mn atoms on the Cd vacancy16 of the host. All calculated results are summarized in Fig. 3. Besides, we have studied the structural behavior of CdGeP2 under pressure in order to elucidate the volumetric changes of Fig. 1b. Since the understanding of CdGeP2 is the first step to elucidate the pressure behavior of Cd1−xMnxGeP2.

Figure 3
figure 3

Summarized DFT calculations for MnP clusters and substituted Mn atoms in Cd1−xMnxGeP2.

(a) Total energy-volume curves for MnP clusters for non-magnetic (NM) and several FM spin configurations. (b) Enthalpy difference ΔH at T = 0 K as function of pressure for these configurations. The pressure dependence of Mn-Mn and Mn-P interatomic bond distances of MnP are presented in (c) and (d), respectively. Here dashed lines correspond to isostructural transformations in MnP at ≈5.5 GPa and 7.2 GPa. (f) Pressure evolution of lattice parameters and (g) magnetic moment at the specified Mn content when Mn atoms substituted in the host.

Full size image

Figure 3a shows total energy calculation as function of volume for MnP crystalline cell. The crystallized cell with Pnma structure was created starting from the experimental values, as in Reference 27. Several MnP spin configurations were considered: NM, FM-I, FM-II and FM-III, which are obtained by pressurizing the initial FM-I configuration. Here the FM configurations have the magnetic moment from 1.50, 1.20 and 0.53 μB/f.u. for FM-I, FM-II and FM-III states, respectively. Our theoretical results indicate that the FM-I state corresponds to the lowest energy configuration. We then calculate the enthalpy difference ΔH as function of pressure at T = 0 K (Fig. 3b). The volume changes show three sharp consecutive steps FM-I → FM-II → FM-III at 5.5 GPa and 7.2 GPa (see Fig. S4 in the Supplement Material). In detail, this picture can be related to pressure-depended changes in the interatomic bond distances d along Mn-Mn (Fig. 3c) and Mn-P (Fig. 3d) atoms. This analysis stresses how the pressure is switching between different FM configurations (however, slightly overestimated experimental results), suggesting then isostructural transformations occurring in the cation coordination of the structure with Pnma space group, without involving group-subgroup structural changes. The transition between a magnetic and non-magnetic state is obtained around 9 GPa (depending on the used functional), a lower phase transition pressure can be obtained if the total free energy is used by considering vibrational contribution obtained from the quasi-harmonic approximation.

The theoretical results for the atomic Mn substitution in the host provide us additional confirmation that host-doping concentration leads to host structure stability under pressure, only if the role of pressure-induced decomposition and other possible high-pressure phases beyond the chalcopyrite type structure are excluded. This dependence shows no phase transitions up to 20 GPa, as it evident from weak reduction in lattice parameters (Fig. 3e). Furthermore, for both 0.125 mol% (x = 0.005) and 0.25 mol% (x = 0.01) Mn concentrations, we found a magnetic moment of 3.88 μB/Mn at ambient pressure that is then completely depleted around 20 GPa (Fig. 3f).

Figure 1b shows that CdGeP2 and Cd1−xMnxGeP2 undergo a phase transition between 3 and 4 GPa, which is accompanied by a volume reduction, i.e. a first order phase transition. In order to consider possible structural phase transitions, we study the stability of CdGeP2 by considering several structures reported as high-pressure phases of another chalcopyrite-type compounds. Among those, we put special attention on the modified NaCl-type cubic phase with P atoms on (0.25, 0.25, 0.25) position28 and the NaCl-type structure with atoms located on a random distribution29. Our results do not show any phase transition to these phases below 10 GPa. However, it was observed that CdGeP2 is decomposed into several compounds as pressure increases (see Fig. S4 from supplement material). Which could suggest that Cd1−xMnxGeP2 could experience decomposition under pressure, or a phase transition to a NaCl-type structure at pressures much higher than those investigated in this report. Nevertheless the identification of the high pressure phase will be addressed with future experiments.

Discussion

Therefore, our data demonstrates a large magnetic dependence of the sample as function of pressure with large doping concentration. This is due to the presence of MnP clusters, which are strongly inhomogeneous in Cd1−xMnxGeP2. We do not observe any noticeable precipitation that indicates compound segregation for the pressure regimes considered. However, it is unclear how the FM clusters can be influenced by other inhomogeneous NM inclusions. Based on the EDS and SEM investigations, we can identify six different phases, including MnP clusters (70 μm approx.), all present at ambient conditions as noted in Fig. 4a and supplemental material. The presence of Mn could make the sample quite magnetic but our samples show mostly the formation of MnP clusters. Therefore, the effect of only Mn is largely reduced. Even in the case of host Mn substitutions, we find those to be in very low concentration and in general they cannot be attained as responsible for the intrinsic ferromagnetism. One point to have in consideration is that when pressure is applied (Figure 4b), the inhomogeneous phases observed in our samples of Cd1−xMnxGeP2 could separate and react as new subproducts due to the strong pressure-induced decomposition trend. However, this topic should be detailed studied in future experiments.

Figure 4
figure 4

A schematic illustration of strongly inhomogeneous Cd1−xMnxGeP2 with MnP clusters and substituted Mn atoms in the host.

(a) There are six different phases, including MnP clusters at normal conditions. (b) When pressure in applied there is possible separation and appears of new subproducts due to pressure-induced decomposition trend.

Full size image

Next, we briefly comment on the implementation of pressure control of ferromagnetically ordered clusters. Recently, progress in the field of microfluidic technology has proposed to create a static digital control logic using pressure-gain valves30. Rendering this concept into the development of devices based on embedded magnetic clusters, it is possible to imagine that by controlling the cluster size in ferromagnetic chalcopyrites, could be possible then by pressure to manipulate the magnetic and electric response.

In summary, unconventional behavior of ferromagnetically MnP clusters in strongly inhomogeneous Mn-doped CdGeP2 was investigated via effect of hydrostatic pressure and ab initio simulations. We find two serial transitions from FM to NM states at RT, which directly arises from cluster magnetic transitions (magnetic switching). Although, the detail microscopic physics of these phenomena was not completely addressed, future experimental and theoretical efforts should focused in those aspects. Even though, the results here discussed present a pressure controlled mechanism that affect directly on the clusters magnetization. This novel scenario of using Mn-doped chalcopyrites in fabrication devices based on the microfluidic control via pressure-gain28 is a path worth of pursuing.

Methods

Samples

Polycrystalline samples of Cd1−xMnxGeP2 with 0 ≥ x ≥ 0.225 were prepared by the direct fusion method. The procedure of synthesis, as well as magnetization data at ambient pressure is presented elsewhere31. In this work, we focus on Mn concentration of x = 0.225, where the obtained TC was 332 K, to our knowledge the highest value for this compound20. All samples Cd1−xMnxGeP2 with 0 ≥ x ≥ 0.225 were characterized by high-resolution x-ray diffraction (HRXRD), EDS and SEM. According to HRXRD data, the main crystal phase was identified as the tetragonal chalcopyrite structure CdGeP2 with space group I42d. The EDS and SEM analysis point to inhomogeneous distribution of several phases, including magnetically MnP clusters on the micron size. The detailed description of sample characterization can be found in the Supplementary material (see Fig. S1).

High-pressure measurements

In situ volumetric and transport measurements were performed in Toroid type high-pressure apparatus by combination of high-precision strain gauge technique and a six-probe method on samples of 3 × 1 × 1 mm3 in size (see References 32 and 33 for details). High-pressure ac-susceptibility and ZF magnetization measurements were implemented by conventional self-excited frequency and inductance methods respectively. A pure mixture of ethanol-methanol 4:1 was used as a pressure-transmitting medium. Pressure inside the teflon cell with useful volume ≈80 mm3 was continuously monitored by a manganin wire and calibrated on the end points to the Bi phase transition.

Computational details for MnP crystals

Total energy calculations and structural optimizations were performed within the framework of the density functional theory (DFT) and the projector-augmented wave (PAW)34,35 method as is implemented in the Vienna ab initio simulation package (VASP)26,36. We use a plane-wave energy cutoff of 356 eV to ensure a high precision in all our calculations. The GGA of Perdew, Burke and Ernzerhof (PBE)37 based on the projector augmented wave (PAW)34 method was used to treat the exchange-correlation (XC) functional. Monkhorst-Pack scheme was employed to discretize the Brillouin-zone (BZ) integrations38 with a mesh of 4 × 5 × 4 for MnP crystal. In the relaxed equilibrium configuration, the forces are less than 3 meV/Å per atom in each of the Cartesian directions. We have used the antiferromagnetic (AFM, Mn: ↑↑↓↓, ↑↓↑↓, ↑↓↓↑), ferromagnetic (FM, Mn: ↑↑↑↑) and nonmagnetic (NM) configurations for MnP. According to our results the AFM phases are energetically noncompetitive against the FM and NM phases for MnP.

Computational details for Mn substitutional atoms

We have initiated our calculations, by taking the I42d crystal structure of CdGeP2, consisting of 16 atoms (Cd = 4 atoms, Ge = 4 atoms and P = 8 atoms) in the unit cell. To dope with 0.125% and 0.25% of Mn atoms, we have constructed the 2 × 2 × 1 and 2 × 1 × 1 supercells of CdGeP2, respectively. The energy cutoff of 600 eV for plane wave basis and 8 × 8 × 8 K-point mesh, generated by Gamma method for Brillouin zone sampling, were found suitable to ensure the sufficient convergence, until the Hellman-Feyman forces acting on each atom became less than 0.005 eV/Å.