Impact of fast-solidification on all-d-metal NiCoMnTi based giant magnetocaloric Heusler compounds

-metal

Different optimization strategies have been applied for this all-dmetal NiCoMnTi based Heusler system.For example, by optimizing the heat treatment of the normal arc-melting compounds the value of magnetic entropy change (Δs T ), based on the calculation of isofield magnetization data, can reach 38 Jkg − 1 K − 1 for Δμ 0 H = 2 T [17].Rapid solidification (melt spinning) can also be used to enhance the GMCE performance in the NiCoMnTi based Heusler system, where Δs T can reach about 27 Jkg − 1 K − 1 for Δμ 0 H = 2 T [16].However, the large changes in microstructure result in an increased ΔT hys of 19 K, which will significantly degrade the performance.This signifies that tuning the solidification conditions by carefully modifying the microstructure can be an efficient way to manage its GMCE.Compared with melt-spinning (with cooling rates above 10 5 K/s), suction casting (with cooling rates of 10 2 K/s) provides a relatively mild modification of the microstructure due to the slower cooling rates (about 100 K/s), and can therefore be used to design different materials microstructures [21].In the present study, arc-melted and suction-cast Ni 37 Co 13 Mn 35 Ti 15 compounds have been produced and the thermodynamic, GMCE, microstructural properties have been reported.In comparison with the arc-melted sample, it is found that the suction-cast sample exhibits a sharper FOMT, while a low ΔT hys is maintained at 5.5 K.As a result, a very competitive reversible Δs MH T of 21.8 Jkg − 1 K − 1 for Δμ 0 H = 5 T is obtained for the suction-cast sample from the calculation of isothermal magnetization data in different fields.The responsible mechanism has been investigated from its microstructural aspects.It is found that the favorable properties can result from the reduced concentration in lattice defects, which suggests that defect engineering can be used to control the GMCE in these Heusler compounds.Our current studies elucidate that fast solidification can efficiently regulate the magnetocaloric properties in all-d-metal NiCoMnTi Heusler materials.

Experimental procedure
High-purity (99.9 %) raw materials are used to prepare polycrystalline samples with a nominal composition of Ni 37 Co 13 Mn 35 Ti 15 using high-vacuum (< 10 − 6 mbar) arc-melting and suction-casting under Ar atmosphere.The samples are melted for 5 times for good homogeneity.To compensate for evaporation losses of Mn during melting, 4 at.%extra Mn is introduced.Suction rods with 8 mm diameter are cast as SC sample, and the corresponding suction-cast rod is produced out of the arc-melted sample directly.Then the as-cast AM and SC samples were sealed in quartz tubes under Ar atmosphere and annealed for 6 days at 1173 K in a vertical oven [22].Subsequently, these samples were rapidly quenched from the annealing furnace into cold water, denoted as AM sample and SC sample, respectively.
Zero-field differential scanning calorimetry (DSC) measurements were carried out using a commercial TA-Q2000 DSC calorimeter.X-ray diffraction (XRD) patterns in the high-temperature austenite state (structural phase-transition temperature T str + 100 • C) were collected using an Anton Paar TTK450 temperature chamber and a PANalytical Xpert Pro-diffractometer with Cu K α radiation.Note that the residual strain inside the materials has been released by annealing the powders at 500 • C for 5 h under Ar atmosphere.The above XRD patterns were analysed using Fullprof's implementation of the Rietveld refinement method [23].Moreover, the temperature-dependent magnetization (M-T) and the field-dependent magnetization (M-H) curves for all samples were measured in a superconducting quantum interference device (SQUID, Quantum Design MPMS 5XL) magnetometer.The isothermal M-H measurements at different temperatures were performed by the so-called loop method [24].The samples were initially cooled down to 180 K (complete martensite region) and then subsequently warmed up to the target temperature at a rate of 2 K/min.Doppler broadening positron annihilation spectroscopy (DB-PAS) was employed to study the lattice defects in the samples.DB-PAS depth profiles were measured at room temperature with positron implantation energies ranging from 0.1 to 25 keV, using the variable energy positron beam facility at the Reactor Institute Delft [25].A high-purity Ge (HPGe) detector cooled by liquid nitrogen with an energy resolution of 1.2 keV at 511 keV was used to detect the energy distribution of the 511 keV γ-rays produced by the annihilation of positrons and electrons.The intensity of the low-energy positron beam at the sample position is about 10 4 e + /s, and the full width at half maximum (FWHM) of the beam diameter is about 8 mm.The depth profiles of the extracted shape (S) and wing (W) parameters were fitted by the VEPFIT program [26].The actual compositions for the AM and SC samples have been determined by scanning electron microscopy (SEM) with energy dispersive X-ray spectroscopy (EDX), as shown in Table S1 (Supplementary Information).The EDX maps in Fig. S1 (Supplementary Information) indicate that there is no obvious phase segregation within the matrix.In addition, the basic morphology of samples was investigated by optical microscopy and SEM (FEI Quanta TM 450 FEG).The TEM specimens were prepared with a focused ion beam system (FIB) using the lift-out method.The HRTEM analysis for these two samples was performed using a JEM-F200 and digital image processing was performed by the DigitalMicrograph software (GMS 3, Gatan Inc).

Results and discussion
The heat flow as a function of temperature measured by zero-field DSC for the AM and SC samples is presented in Fig. 1a.Compared with the AM sample, it is observed that the endothermic and exothermic peaks shift to lower temperatures and become sharper for the SC sample, which phenomenologically indicates the first-order phase transition will be sharper.From the DSC results, the corresponding characteristic temperatures (T M s , T M f , T M , T A s , T A f , T A ), the thermal hysteresis ΔT hys and entropy change of the transition (Δs DSC T ) are extracted and summarized in Table 1.The thermal hysteresis is defined as 2 and the entropy change of the transition as ) − 1 dT, where Q is the heat flow and Q base the baseline for the heat flow [27,28].For these two samples, it is found that ΔT hys is low and Δs DSC T is comparable with other good all-d-metal NiCoMnTi GMCE materials [18].
Additionally, M-T curves are measured in an applied magnetic field of 1 T for the AM and SC samples, as shown in Fig. 1b.A strong FOMT from the low-temperature weak magnetism to high-temperature strong ferromagnetism is observed, with phase transitions located near room temperature.Compared with the AM sample, the transition temperatures of the SC sample were lowered, i.e., T str heating moved from 287.2 to 263.2 K.It is known that in comparison to the traditional arc-melting process, the suction-casting process produced a stronger texture in the grain orientation [29,30].For NiMn-X (X = Al, Ga, Sn, In, Sb, Ti) based alloys with a strong magneto-structurally coupled transition, the stress-induced martensitic transition is significantly affected by the texture of the alloy [31].The orientation relations of the austenite grains formed at high temperature with grain boundaries will determine the amount of martensite formed [32,33].Meanwhile, the grain refinement by fast solidification (suction-casting) will increase the short-range atomic chemical disorder, which hinders the transition and thereby reduces the martensite start temperature [30].A similar reduction in the structural transition temperature has been found in Ni-Mn-In intermetallic Heusler compounds for microalloying with boron and Ni-Mn-Ga compounds for nano-structuring [34,35], which is mainly caused by the grain refinement.Additionally, it has been found that T C of the SC sample decreased simultaneously.The ΔT hys values for the AM and SC samples by magnetization measurement are only about 5.5 K.These very low ΔT hys values will further contribute to the reversibility of MCE [22,36].Note that the lower ΔT hys values for the SQUID (2 K/min) in comparison to the DSC (10 K/min) measurements should be ascribed to the slower heating/cooling rates, as shown in Table 1.Because the first-order transitions are driven by nucleation and growth, and the time dependent transitions need certain response time for external fields [37].
To further evaluate the GMCE performance of these two samples, Fig. 2a  Considering the application scenarios (using a commercial electromagnet) [38] the reversible part of Δs T during cycling should be considered.As presented in Fig. 2b, the isothermal magnetic field-cycled M-H loops (1st cycle is indicated by the solid symbols and the 2nd cycle by the open symbols) were carefully collected in the temperature range from 251 to 263 K.It is observed that the M-H curves do not overlap with each other during the 1st increase and 2nd decrease of the magnetic field cycles.This could result from the presence of some residual austenite that does not contribute to the transformation during the field cycles.This effect has been observed before in the same material system [22].However, after 2nd cycle no residual phases are observed, thus cycling is reversible [22].Therefore, the so-called transformation fraction (TF_MH) method is applied assuming that the total magnetization originates from the ferromagnetic austenite phase fraction [36,39].Consequently, the reversible Δs MH T (Δs MH rev T ) can be determined by combining Eqs. ( 1) and (2): where f aus (T,H) is the austenite fraction at certain temperature and field, M(H) is the measured magnetization of the sample, and M m (H) is the magnetization of pure martensite (see the extrapolated black dotted lines in Fig. 2b), M a (H) is the magnetization of pure austenite, Δs DSC T refers to the entropy change of the transition derived from DSC (see Table 1), Δf is the transformed austenite fraction.About the difference between Δs MH T and Δs DSC T , in general, the Δs MH T is calculated from the field-dependent magnetization measurements (isothermal protocol), whereas the Δs DSC T is from the DSC measurements and represents the entropy change of the full and complete phase transition.Note that the 2nd M-H cycle at 263 K is chosen as the reference for pure austenite to avoid an underestimation of the austenite fraction at low fields [22].For the SC sample the corresponding f aus of the 2nd field cycle at different constant temperatures have been derived, as demonstrated in Fig. 2c.Because of the rapid change in magnetization at low fields, the magnetization values below 1 T (shadow area in Fig. 2c) are excluded to guarantee the validity of data [40,41].As shown in Fig. 2d, the sharper FOMT and the low ΔT hys , results in a maximum reversible Δs MH T for the SC sample of 21.8 Jkg − 1 K − 1 at 256 K for Δμ 0 H = 5 T.This value is in good agreement with the data (marked as "◆") extracted from the overlap region where the Δs T determined on cooling and heating are crossing, which therefore corresponds to the reversible Δs T [28,42,43].This excellent maximum Δs MH rev T value (21.8 Jkg − 1 K − 1 ) for Δμ 0 H = 5 T is      [22].It can be noted that the peak of Δs T for the first cycle obtained from the Maxwell relation and TF_MH methods is in good agreement with each other, as found in previous studies [36,41].And it is worth mentioning that compared with the 1st M-H cycle the value of Δs MH T is obviously degraded from 30.7 to 21.8 Jkg − 1 K − 1 , which corresponds to a 29% decrease even though the sample has a very low ΔT hys (5.5 K).For samples that show a strong FOMT, it is important to further investigate the reversibility of the MCE, as this defines their potential in real applications [36,41,[46][47][48].By applying the equation ΔT ad = − T Cp Δs T , where C p is the specific heat at certain temperature, the reversible adiabatic temperature change |ΔT ad | for the SC sample is estimated at 10.7 K for Δμ 0 H = 5 T.
To better understand the transformation mechanism, the microstructures for the AM and SC samples have been obtained, as shown in Fig. 3.The HRTEM image in Fig. 3a exhibits the highly ordered austenite area, as there is no obvious lath-like microstructure observed, that is characteristic for the martensite phase [15].In Fig. 3b the corresponding fast Fourier transform (FFT) for the selected region in Fig. 3a demonstrates its face-centered cubic austenitic structure (Fm-3 m; space group 225).The main crystal planes have been indexed successfully.By applying the inverse fast Fourier transform (IFFT) it is found in Fig. 3c, that abundant dislocations are distributed within the main matrix, which might be introduced by the solidification and casting processes.The high density of lattice defects, such as these random dislocations, may influence the development of internal stresses during the martensitic phase transition [49,50].Moreover, from Fig. 3d the atomic vacancies and vacancy-cluster defects have directly been observed.Edge dislocations are located at the periphery of these point defects.It has been proposed that such quenched-in vacancies through the interaction of clustered vacancies with partial dislocations can prevent the nucleation of martensite [51].This might be the reason that for almost the complete NiCoMnTi series of compounds made by arc-melting with quenching [20,[52][53][54][55][56][57] (without optimized heat-treatment) the transition cannot easily be tuned towards a sharp transition, compared with other traditional NiMn-X (X = Al, Ga, In, Sn, Sb) Heusler compounds.However, only fast solidification [16] or higher annealing temperatures [17] can sufficiently regulate the magneto-structurally coupled transition in this materials family.Here we find that this is correlated with the formation and control of lattice defects.
As mentioned before the expected grain refinement is clearly observed for the SC sample, as shown in Fig. 3e.In comparison to the regular arc-melting technique where a cooling rate of about 10 K/s is achieved [29,58], for suction casting [29] and melt spinning [59,60] cooling rates of 10 2 and 10 5 K/s are achieved, respectively.It is found that melt spinning can drastically reduce the grain size down to below 1  μm [16,61,62].However, at the same time ΔT hys is increased to 19 K, which is not beneficial for the reversibility [16].In contrast, the moderate cooling speed for suction casting provides a more gentle modification of the microstructure.As exhibited in Fig. 3e, the average grain size for the SC sample is about 86 μm.Simultaneously, a clear difference can be seen between the AM and the SC sample, as shown in Fig. 4. Optical microscopy shows that the average grain size undergoes a reduction of about a factor 5 from about 370 μm (AM) to about 80 μm (SC).The above findings suggest that by well controlling the microstructure the suction casting with its lower cooling rate can achieves a better GMCE performance with maintaining a low ΔT hys .The electron diffraction in Fig. 3f proves that the cubic austenite phase has been retained after suction casting.Interestingly, compared with the arc-melted sample, a lower concentration of lattice dislocations is found   for the SC sample in Fig. 3g.This phenomenon may be closely related to the refined microstructure of the SC sample because a finer grain structure provides more grain boundaries in comparison with the regular coarse microstructure.According to the model proposed by Grabski et al. [63], the grain boundaries can act as sinks for dislocations.This means that the existing dislocations can be incorporated in the grain boundaries by the formation of grain-boundary defects.High-temperature annealing helps the release of strain energy carried by the dislocations [64,65], which accumulates during casting.Once dislocation movement occurs, individual dislocations will interact with neighboring ones, which results in coalescence and annihilations of dislocations at grain boundaries.These processes are diffusion limited and will be influenced by diffusion length and the pinning/trapping with other types of defects.For example, in pure Ni and α-Fe systems it is found that an edge dislocation will strongly interact with vacancies (pipe diffusion) [66].Hence, it is believed that the higher concentration of vacancies and longer diffusion length for AM sample conduct more dislocations than SC sample.For AM and SC samples, the defect concentrations during heat-treatment are different because of the change in crystallite size resulted from different solidification rates.In addition, it is known that one of the most important functions for fast-quenching treatments is the retention of phases stable at high annealing temperatures [67].And it also can be used for maintaining the high-temperature vacancy-type point defects (so-called quenched-in vacancies).For example, in specific cases such as in engineering structural materials, there will be a tendency that the vacancies can coalesce into vacancy clusters, and some clusters collapse into dislocation loops which can grow by absorbing more vacancies [68,69].The fast-quenching process can preserve the dislocations at room temperature in crystal materials, as experimentally observed before [70].As shown in Fig. 3h some short-range disordered nanostructures (with a diameter of about 10 nm) are distributed within the crystal matrix.The inset (right) is the relevant FFT result, which proves the disordered structure.Abundant disordered structures will be introduced during the suction-casting or other rapid solidification process [71].Surprisingly, annealing at high temperature (1173 K) did not result in a complete crystallization of these disordered structures.When the annealing temperature is above 0.5T m , where T m is the melting point (around 1380 K for the current samples) [17], it is expected that recrystallization will occur smoothly [72][73][74].Actually, the recrystallization can be considered to be a nucleation and growth phenomenon, which is controlled by thermally activated processes [75].It is known that the process of nucleation and growth will be greatly influenced by lattice defects, for example, non-equilibrium defects such as excess vacancies and dislocations can be recognized as suitable nucleation sites for the heterogeneous nucleation of crystals [68].From above experimental results, compared with the AM sample, the SC sample contains lower concentrations of vacancies and dislocation defects, which will further inhibit the nucleation and growth processes of recrystallization.In addition, even grain boundaries are also favored sites for nucleation, however, compared with the nanoscale disordered region, the large crystallite size (~86 μm) of the SC sample provides fewer nucleation sites.This will lead to a lower nucleation rate, and the recrystallization becomes slower [75].Therefore, even at relatively high annealing temperature (T ≃ 0.8T m ) the recrystallization of limited amount of disordered structures are still maintained.For instance, similar nano-disordered structures have been noticed in the high-temperature annealed melt-spun ribbons [76].Interestingly, it is also found that the nano-disordered structures within specific nanocrystal matrix (e.g.Al-Mg alloys) can have robust stability up to melting temperatures [77].The degree of atomic ordering is found to correlate with the shift in transition temperature for Heusler compounds [49,[78][79][80].Furthermore, according to theoretical calculations the T C of austenite shows a strong correlation with the electron-to-atom ratio (e/a) in the NiCoMnTi system [17].The e/a was calculated from the actual chemical formula of the compound and it is equal to the concentration weighted sum of the number of 3d and 4 s electrons of Ni (3d 8 4s 2 ), Co (3d 7 4s 2 ), Mn (3d 5 4s 2 ) and Ti (3d 2 4s 2 ) [81].For the AM and SC samples, the e/a values are calculated as 7.908 and 7.910, respectively, which are almost the same.In other words, the significant changes of transition temperature should mainly be attributed to changes of microstructures between the AM and SC samples rather than the electronic properties.Additionally, it is known that rapid solidification like melt-spinning can introduce disordered structures and a lower ordering which can further decrease transformation and Curie temperature, which had been observed before [78,82].With regard to the magnetic properties, it is worth mentioning that these disordered structures did not present a magnetic transition at low temperature, as shown in Fig. S2 (Supplementary Information).This is different compared to the reentrant spin-glass type transition found in other Heusler materials [83][84][85].Note that at room temperature in the HRTEM the martensite phases co-exist, which indicates that some residual phases exist, as presented in Fig. S3 (Supplementary Information).
To understand the lattice defects from a microscopic perspective, positron annihilation spectroscopy (PAS) is employed for the AM and SC samples.It is well-known that PAS is a non-destructive method and can efficiently obtain information of open-volume defects, such as dislocations and vacancies with a high sensitivity due to the strong affinity between positrons and defects resulting in a positron-electron pair annihilation process [86][87][88][89].As shown in Fig. S4 (Supplementary Information), XRD results suggest that these two samples have a high purity.In Fig. 5, the Doppler broadening characterized by the S and W parameters is presented for these two samples as a function of the positron implantation energy (probing the depth profiles).The S parameter represents the positron annihilation with valence electrons, which is sensitive to the electronic structure and open-volume defects.The W parameter reflects the positron annihilation with (semi-)core electrons, which is responsible for the local chemical environment of the positron trapping sites [90].As shown in Fig. 5a (orange area), the S parameter of the AM and SC samples rapidly decreased in the low energy range up to about 2.5 keV, corresponding to a mean positron implantation depth less than 20 nm.The observed decrease in S (increase in W) is ascribed to epi-thermal positron remission and positrons annihilating at the surface.When the energy is above 10 keV, the region indicated in blue in Fig 5, a plateau in the S and W parameters is reached, where positrons are primarily stopped and annihilated in the bulk matrix, at positron mean implantation depths greater than 200 nm.The greater increase in S and decrease in W for the AM sample means that there are more open-volume lattice defects, such as dislocations and vacancies distributed in the matrix.This is in good agreement with the HRTEM observations in Fig. 3. Combined with the HRTEM data the above PAS results suggest that the presence of lattice defects significantly influences the thermodynamics and kinetics of the strong first-order phase transition in these materials [91].As a consequence, the transition in the SC sample is sharper in the NiCoMnTi Heusler compounds.Similarly, in NiTi shape memory alloys driven by a martensitic transition it is found by molecular dynamics simulations that defects like vacancies and vacancy clusters have a pronounced effect on the inhibition of the transition [92].This means that controlling the defect concentration provides a way to control the strong magneto-structurally coupled transition in magnetic Heusler compounds.

Conclusions
In summary, arc-melted and suction-cast Ni 37 Co 13 Mn 35 Ti 15 compounds have successfully been synthesized, and their basic thermodynamic, magnetic and microstructural properties have been investigated utilizing DSC, SQUID, HRTEM and PAS.Compared with the AM sample, it is found that fast solidification (suction casting) can significantly increase Δs T as the FOMT is enhanced while ΔT hys remained low.Combining the HRTEM and the PAS measurements, the positive enhancement of the GMCE for the SC sample could be attributed to the lower concentration of lattice defects in comparison to the AM sample.Interestingly, the grain refinement and the short-range disorder in the nanostructures could result in a decrease of the transition temperature for the SC sample.Due to the low ΔT hys and the well-tuned FOMT, a good reversibility of the GMCE has been found for the SC sample, resulting in a very competitive reversible Δs MH T of 21.8 Jkg − 1 K − 1 for Δμ 0 H = 5 T. Our present results indicate that the fast solidification can efficiently tailor the GMCE properties of these all-d-metal NiCoMnTi Heusler compounds by a modification of the structure and lattice defects, which enhances the prospects for future applications.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

F.
Zhang et al.
presents the calculated Δs T (up to a field change of 0-5 T) as a function of temperature in the vicinity of the martensitic transition.The magnetic entropy changes Δs T with different magnetic fields Δμ 0 H are calculated based on the Maxwell relation Δs T (T, H) and the SC sample shows the largest Δs T peak (for heating) at 29.4 Jkg − 1 K − 1 for a field change of Δμ 0 H = 5 T. Compared with the AM sample, the maximum value of Δs T for the SC sample shows a remarkable improvement by 67% from 18.4 to 29.4 Jkg − 1 K − 1 for Δμ 0 H = 5 T.

Fig. 1 .
Fig. 1.(a) Heat flow derived from DSC experiments for the AM and SC samples upon warming and cooling processes at a rate of 10 K/min.(b) Iso-field M-T curves for the AM and SC samples in a field of 1 T. The inset shows the derived temperature dependence of dM/dT from the M-T curves.
), the structural temperature upon cooling (T str cooling ), the structural temperature upon heating (T str heating ), the thermal hysteresis ΔT hys , the entropy change Δs DSC T , the change in magnetization ΔM and the field dependence of the transition temperature dT str /μ 0 dH for the AM and SC samples (measured in DSC and SQUID).

Fig. 2 .
Fig. 2. (a) Magnetic entropy change Δs T of the AM and SC samples as a function of temperature for both heating and cooling determined for field changes Δµ 0 H ranging from 1 to 5 T. (b) Iso-thermal M-H curves for the SC sample measured during the 1st (solid symbols) and 2nd (open symbols) cycle at different temperatures.The black dotted lines denote the extrapolated magnetization for pure martensite at low temperature.(c) Fraction of austenite f aus for the 2nd cycle at different constant temperatures determined from (b).(d) Δs MH T for Δμ 0 H = 5 T obtained from the TF_MH method during the 1st and 2nd cycles.In the 2nd cycle Δs MH T is reversible.The red point ("◆") is from the overlap between cooling and heating curves (Δµ 0 H = 5 T).

F
.Zhang et al.

Fig. 3 .
Fig. 3. (a) HRTEM image for the AM sample.(b) Corresponding FFT patterns in the regions indicated by the red box in (a) for the AM sample.The inset is the selected diffraction spots for IFFT.(c) Corresponding IFFT pattern of (b) consisting of numerous dislocations marked with "T" along the (422) direction.(d) False-color image of the partial enlarged HRTEM image from the red region in (a).(e) SEM image of the SC sample showing the grain structure.(f) HRTEM image for the SC sample.The inset is the corresponding FFT pattern in the regions indicated by the red box.(g) Corresponding IFFT pattern of (f) consisting of less dislocations marked with "T" along the (420) direction.(h) HRTEM image for the SC sample.The blue circled area shows short-range disordered lattice: (left) partial enlarged image; (right) FFT patterns for the selected disorder region.

Fig. 4 .
Fig. 4. (a)-(b) Optical microscope images for the etched AM sample.(c)-(d) Optical microscope images for the etched SC sample.The average diameter of grain is about 370 and 80 μm for the AM and SC sample, respectively.

F
.Zhang et al.

Fig. 5 .
Fig. 5. (a) S parameter and (b) W parameter as a function of the positron implantation energy extracted from positron Doppler broadening (DB) depth profiles for the AM (blue) and the SC (red) sample.