Enhanced magnetocaloric effect in Ni-Mn-Sn-Co alloys with two successive magnetostructural transformations

High magnetocaloric refrigeration performance requires large magnetic entropy change ΔSM and broad working temperature span ΔTFWHM. A fourth element doping of Co in ternary Ni-Mn-Sn alloy may significantly enhance the saturation magnetization of the alloy and thus enhance the ΔSM. Here, the effects of Co-doping on the martensite transformation, magnetic properties and magnetocaloric effects (MCE) of quaternary Ni47−xMn43Sn10Cox (x = 0, 6, 11) alloys were investigated. The martensite transformation temperatures decrease while austenite Curie point increases with Co content increasing to x = 6 and 11, thus broadening the temperature window for a high magnetization austenite (13.5, 91.7 and 109.1 A·m2/kg for x = 0, 6 and 11, respectively). Two successive magnetostructural transformations (A → 10 M and A → 10 M + 6 M) occur in the alloy x = 6, which are responsible for the giant magnetic entropy change ΔSM = 29.5 J/kg·K, wide working temperature span ΔTFWHM = 14 K and large effective refrigeration capacity RCeff = 232 J/kg under a magnetic field of 5.0 T. These results suggest that Ni40.6Mn43.3Sn10.0Co6.1 alloy may act as a potential solid-state magnetic refrigerant working at room temperature.

Over the last decade, Ni-Mn-X (X = Sn, In and Sb) metamagnetic shape memory alloys (MSMAs) have attracted significant interests due to their potential applications as magnetic refrigeration materials near room temperature. These materials, based on magnetocaloric effect (MCE), are considered to be environmentally friendly and cost-effective refrigerants because of their large refrigeration capacity (RC) which is comparable to those of expensive rare-earth-containing MCE compounds 1 . In Ni-Mn-X-based (X = Sn, In and Sb) alloys, the metamagnetic structural transition from the weak-magnetic martensite to ferromagnetic austenite under an applied magnetic field leads to an inverse MCE [2][3][4] , whereas the magnetic transition of the austenite phase is responsible for a conventional MCE [5][6][7] . Clearly, the metamagnetic transition is driven by the Zeeman energy E zeeman = μ 0 HΔM, where ΔM and μ 0 H represent the magnetization difference between martensite/austenite phases and the applied magnetic field, respectively. So, enhanced ΔM is in favor of large E zeeman , which is responsible for the high magnetic entropy change (ΔS M ) and wide working temperature span (ΔT FWHM, defined as the full width at half maximum of the magnetic entropy peak) 8 , where the ΔT FWHM is crucial for magnetic refrigeration applications in the case of inverse MCE 9, 10 .
In Ni-Mn-X off-stoichiometric MSMAs, the ΔM can be enhanced by composition tuning 11,12 , heat treatment 13,14 and doping 15,16 . It has been reported that high-Mn content Ni-Mn-X alloys exhibited enhanced ΔM because the magnetization of the austenite was mainly attributed to the ferromagnetic interaction between the neighboring Mn-Mn atoms 17 . For instance, the austenite of Mn-rich Ni 43 Mn 46 Sn 11 alloy 18 shows a higher magnetization of 68 A·m 2 /kg under a magnetic field of 5.0 T, compared to that of 39 A·m 2 /kg for Ni 46 Mn 43 Sn 11 alloy 19 . Heat treatment is also a useful method to modulate the grain constraint in order to control the microstructure, martensite transformation (MT) temperature and magnetic properties of the ferromagnetic MSMAs (FMSMAs) 20,21 . Doping of non-ferromagnetic elements [22][23][24] such as Cu, Al and Ti can enhance ΔM effectively by tuning the valence electron concentration e/a or changing the unit cell volume, which mainly changes the MT temperatures. Of note is that the doping of ferromagnetic elements such as Fe and Co can effectively improve the ΔM. For instance, the substitution of Fe for Ni can simultaneously enhance the magnetic-field-induced reverse , 4) Co atoms at Ni site contribute a much larger magnetic moment (~1.0 µ B ) compared to that of Ni (~0.3 µ B ) in the austenite 35 .
The enlargement of ΔT FWHM may be realized by two successive magnetostructural transformations, which have been found in some ferromagnetic alloys. For example, rare earth containing low temperature magnetic refrigeration compounds, such as TbMn 2 Si 2 36 , ErGa 37 , DyB 2 38 , HoPdIn 39 , may exhibit a two successive magnetic transitions behavior deriving from the coupling of spin-reorientation temperature (TSR) and Curie temperature (T c ); Two successive ΔS M peaks with the same sign, associated with a first-order martensite transformation (MT) and an intermediate martensite transformation (IMT), have also been discovered in some Ni-Mn-X alloys after composition tuning or under external pressure [40][41][42] . These two adjacent transformations lead to a partially overlap of the refrigerant temperature intervals, yielding an improved refrigerating capacity. Here, we reported the MT, magnetic transition and MCE of Ni 47−x Mn 43 Sn 10 Co x (x = 0, 6, 11) alloys. Two successive magnetostructural transformations from austenite phase to two different modulated martensite phases, i.e. A → 10 M and A → 10 M + 6 M, were demonstrated in the Ni 40.6 Mn 43.3 Sn 10.0 Co 6.1 (Co6) alloy induced by an external magnetic field. As a consequence, a large magnetic entropy change ΔS M of 29.5 J/kg·K with a wide ΔT FWHM of ~14 K attributed to the occurrence of successive magnetostructural transformations and a strong metamagnetic transition behavior were revealed in the Co6 alloy.

Results and Discussion
Microstructure of the Ni 40.6 Mn 43.3 Sn 10.0 Co 6.1 (Co6) alloy. The typical microstructure of the martensite phase of the Co6 alloy is shown in Fig. 1, where the sample was firstly kept in the ice water for several minutes to reach the martensite state. The microstructure is very similar to those modulated martensite structures reported in some ferromagnetic shape memory alloys 43 . Many cracks distributed along the grain boundaries were observed, as shown in Fig. 1a, implying the brittleness of the polycrystalline alloy. In addition, the martensite morphology is plate-like, which can be recognized by the straight twin boundaries of each plate. Fine twins with thickness of 2-10 μm exist inside the broad martensite plates, as shown in Fig. 1b. Previous report 44 showed that different martensite structures possess different morphologies for Ni-Mn-Sn alloys, where the 10 M martensite exhibited broad plate morphology, 14 M martensite a fine form and the unmodulated structure (L1 0 ) in between. The co-existence of the martensite twins with different widths in the present work implied the existence of different martensite structures in Co6 alloy at room temperature (RT). Detailed martensite structure of the present alloy will be analyzed by XRD and discussed in the later sections.
Martensite transformation of Ni 47−x Mn 43 Sn 10 Co x (x = 0, 6, 11) alloys. Figure 2 shows the DSC curves of Ni 47−x Mn 43 Sn 10 Co x (x = 0, 6, 11) alloys in the temperature range 220-500 K. The strong exothermic and endothermic peaks produced by the forward and reverse MT correspond to the starting and finishing temperatures M s DSC , M f DSC , A s DSC and A f DSC , respectively, determined by the intersection of the base-line and tangent line. These temperatures are summarized in Table 1 together with the values of the valence electrons per atom (e/a). With the addition of Co element, a monotonic decrease in the MT temperatures occurs, as shown in Table 1. Generally, the change in MT temperatures can be interpreted from the following two aspects. Firstly, the e/a dependence of MT temperatures has been found to increase monotonically in Ni-Mn-based Heusler alloys 45,46 . It has been supposed that the L2 1 austenite is stabilized because its Fermi surface just touches the Brillouin zone boundary 47 . With e/a increasing and thus the Fermi surface overlapping the Brillouin zone, the L2 1 austenite structure becomes instable, which induces the occurrence of the MT 47 . Here, the values of e/a are the concentration-weighted sum of s, d, and p valence electrons, i.e. 10 (3d 8 4s2) for Ni, 7 (3d 5 4s2) for Mn, 4 (5s 2 5p2) for Sn and 9 (3d 7 4s 2 ) for Co, respectively. The calculated e/a decreases with increasing Co substitution of Ni mainly because the valence electrons of Co is less than Ni, thus decreasing the MT temperatures of the alloy, which is in good agreement with the e/a dependence rule 45,46 . On the other hand, the atom size effect, originating from the unit cell expansion of the austenite, is not favorable to the occurrence of MT due to changes of the relative positions between the Fermi surface and Brillouin zone 29 . The substitution of larger Co atoms (atomic radius r = 0.126 nm) for smaller Ni atoms (r = 0.125 nm) causes a slight expansion of the austenitic unit cell. Both factors stabilize the austenite phase and therefore decrease the MT temperatures.
A further feature of weak shoulder thermal effect (T C A ) is also observed for sample Co6 and Co11, as shown in the inset in Fig. 2b,c, respectively, which corresponds to the magnetic transition from ferromagnetism to paramagnetism of the austenite during cooling process. But the T C A point was not found for Co0 probably because the magnetic transition occurs below MT temperatures leading to the extremely low magnetization of the austenite (as shown in Fig. 3a). Therefore, it can be deduced that the substitution of Co for Ni in Ni-Mn-Sn alloys effectively increases the T C A , which is consistent with refs 29,30 . The comparison of the transformation entropy changes during heating (ΔS tr endo ) and cooling (ΔS tr exo ) process are also listed in Table 1 . ΔH endo and ΔH exo , denote as the enthalpy changes during the inverse and direct MT, was obtained from the area between the DSC peaks and the baseline. The results show that the difference between ΔS tr endo and ΔS tr exo are very small (≤2.6 J/kg·K). Figure 3 shows the field-heating (FH) and field-cooling (FC) magnetization vs temperature plots (M-T) of Ni 47−x Mn 43 Sn 10 Co x (x = 0, 6, 11) alloys measured under magnetic fields of 0.02 and 5.0 T. The magnetization change (ΔM) in Co0 between austenite and martensite phases during MT is very small (Fig. 3a) because the phase transformation temperatures are higher than the magnetic transition point (T tr > T C A ) and thus MT occurs in a weak-magnetic state. Furthermore, the metamagnetic transition behavior is also weak in Co0 alloy, i.e. its MT  temperatures shift little to lower temperatures with increasing magnetic field (ΔA s /ΔH = −0.2 K/T, as shown in Table 2 based on Fig. 3a), which is unfavorable to MCE 2 . In contrast, both Co6 and Co11 alloys undergo the transformation between a weak-magnetic martensite and a ferromagnetic austenite (Fig. 3b,c), leading to higher ΔM.  , this implies that the friction resistance during MT in Co11 alloy is larger than that in Co6 alloy. It can also be seen from Table 2 that, during heating, the magnetization of the martensite M M changes little but that of the austenite M A increases significantly with increasing Co content. As a result, ΔM in Co6 and Co11 alloys reaches 85.7 and 96.1 A·m 2 /kg under a magnetic field 5.0 T, respectively, which is much higher than ΔM = 9.2 A·m 2 /kg in Co0. The enhancement of ΔM favors the magnetic entropy change ΔS M according to the Maxwell equation and broadens the ΔT FWHM due to the increase in Zeeman energy 45 . Furthermore, the percentage of MT caused by field-induced metamagnetic transition mainly relies on dT M /dH 40 Table 2) and ΔS tr determined by DSC (Table 1), the calculated ΔM/ΔS tr is −2.5 and −5.4 K/T for Co6 and Co11, respectively. By contrast, the  In order to investigate the martensite transformation behavior of the Co6 alloy, X-Ray diffraction (XRD) measurements were performed in the temperature range 343-173 K during cooling process. The data were recorded at a temperature interval 10 K (except for 223 K and 173 K), as shown in Fig. 4. The diffraction peaks marked with five-pointed star are associated with the sample platform (Fig. 4). The martensite peaks were indexed as 10 M and 6 M martensite phases according to the theoretical calculation results 50 and is displayed in the Supplementary Materials Fig. S2. The alloy exhibits a L2 1 -type austenite structure at high temperatures. With decreasing temperature from 343 to 283 K, the peaks corresponding to the 10 M martensite occur starting at 303 K, as shown in Fig. 4a. As the temperature further decreases, several peaks associated with the 6 M martensite appear at 273 K (much lower than M f = 304 K from Table 1), as shown in Fig. 4b. It can also be seen that the peak intensity of 6 M martensite becomes stronger with decreasing temperature. On the other hand, the peak intensity of 10 M martensite still keeps increasing with decreasing temperature after the occurrence of the 6 M martensite, implying the co-existence of the A, as shown 6 M transitions and the mixed A + 10 M + 6 M phase state during the MT process of the Co6 alloy. Such mixed state of 10 M and 6 M martensite phases has also been found in Ni 43 Co 7 Mn 39 Sn 11 51 and Ni 37 Co 11 Mn 43 Sn 9 50 alloys. It is worth noting that the peak (220) A of the L2 1 -type austenite can still be observed at temperature as low as 173 K (see Fig. 4b), indicating that an incomplete MT with residual austenite even at 173 K. This phenomenon is consistent with some other Ni-Mn-based alloys 52,53 , implying that the complete MT might be difficult to be achieved supposing that slightly inhomogeneous composition distribution exists in the bulk alloy. Furthermore, the two successive magnetostructural transition processes (A → 10 M and A → 10 M + 6 M) in the present alloy are essentially different from the two successive martensite (MT) and intermartensite transition (IMT) occur in some other types of FMSMAs such as Ni 55 55 alloys. In addition, the XRD pattern measured at 343 K (A f = 326 K) is shown in Supplementary Materials Fig. S1b. Single austenite phase is detected without any secondary phases, which is consistent with the back scattered electron (BSE) micrograph observations (Supplementary Materials Fig. S1a)  , denote as the 10 M and 6 M transformation starting temperatures, respectively, decrease with increasing magnetic field (i.e. the 10M and 6M change from 321 and 315 K under μ 0 H = 1.0 T to 308 and 301 K under μ 0 H = 5.0 T, respectively). This phenomenon probably indicate that the two successive magnetostructural transformations (A → 10 M + 6 M and A → 10 M) can be induced by an external magnetic field, which is still an open question and will be studied in the near future. Compared to the XRD patterns during cooling from 343 to 173 K (see Fig. 4), the magnetic-field-induced inverse martensite transformation is much easier than the temperature-field-induced one. The isothermal magnetization M-H loops for Co6 alloy at temperatures 290-330 K are shown in Fig. 6a. The plots were recorded under magnetic field up to 5.0 T and at a temperature interval ΔT = 2 K covering the A s and A f . At a temperature of 295 K, far below As VSM (314 K in Table 1), no austenite was induced by magnetic field up to 5.0 T. When test temperatures approach As, the magnetic-field-induced austenite (MIA) transformation gradually occurs, with MIA critical magnetic field decreasing with increasing temperatures. At 307 K, a large jump in magnetization at about 3.4 T (Fig. 6a), due to the occurrence of MIA, was observed. After removing the magnetic field, the sample transformed back to martensite, but with a small amount of field-induced austenite remaining because the test temperature 307 K was slightly higher than M f DSC (304 K in Table 1) 59 . The dM/dH curve measured at 309 K shows two successive inflection peaks centered at 4.0 and 4.5 T (as shown in Fig. 6b), implying the appearance of the two successive magnetostructural transformations. The first inflection peak at H p1 = 4.0 T corresponds to the field-induced inverse martensite transformation 6 M + 10 M → A, and the second  peak at H p2 = 4.5 T to the 10 M → A. The fact that dM/dH under 4.0 T is larger than that under 4.5 T implies that the phase transition resistance of the former is smaller than the latter one.
During isothermal magnetization M-H cycling, the hysteresis loss (HL) was determined by integrating the areas between magnetization and demagnetization branches (purple shaded area in Fig. 6a). The temperature dependence of HL is plotted in Fig. 6c. The average hysteresis loss (AHL), calculated by averaging the integral area under the temperature range of the full width at half maximum of the hysteresis peak Based on isothermal magnetization data with a field interval of ΔH i = 0.05 T at various temperatures (shown in Fig. 6a), the above partial derivative and integration may be numerically approximated to the equation: The ΔS M related to the FOMT for the Co6 alloy under ΔH = 0.5-5.0 T is plotted as a function of temperature displayed in Fig. 7a. The positive ΔS M is attributed to inverse MCE originating from a magnetic-field-induced metamagnetic transition from the weak-magnetic martensite to the ferromagnetic austenite phase 2 . Additionally, the ΔS M peak temperature exhibits a field-dependent behavior, i.e. shifting to lower temperatures with increasing magnetic fields.
As depicted in Fig. 6a, the magnetization saturates at lower fields with increasing temperature above A s . According to the Maxwell equation, further increase of the magnetic field contributes little contribution to ΔS M when the saturation magnetization plots intersect, such as the temperatures of 317, 319 and 321 K in Fig. 6a.   68 . While ΔS M max of the Ni 45.5 Co 4.5 Mn 37 In 13 single crystal does not saturate up to 7.0 T using "curved modelization" adopted by Bourgault et al. 69 . Both approaches are equally consistent with the experimental data at a low magnetic field range (<4.0 T), whereas the "curved modelization" is better in reflecting S M max Δ at high fields. The MCE tends to reach a saturation value for fields above 5.0 T in Ni 51 Mn 33.4 In 15.6 alloy by quasi-directly from isofield calorimetric measurements according to Stern et al. 70 Fig. 7b shows ΔS M -T plots at different temperatures in the reverse MT range. At low temperatures (i.e. T < 316 K), the field-induced fraction of austenite phase increases with the rise of temperature, which is responsible to a similar increase in ΔS M . At higher temperatures, where both martensite and austenite phases coexist at zero field, the field-induced austenite fraction and ΔS M minify as the temperature increases. Furthermore, the negative slope of ΔS M which can be clearly seen in Fig. 7b for high magnetic fields at 320 and 327.5 K, which is due to the contribution of this conventional direct MCE in the vicinity of T C A . The refrigeration capacity (RC), which represents the amount of heat transferring between the cold and hot reservoirs in a thermodynamic cycle, is defined as where T cold and T hot are the corresponding temperatures at ΔT FWHM of ΔS M max . By numerically integrating the area under the ΔS M -T curves between T cold and T hot (shaded area in Fig. 7a), the RC = 322 J/kg under a magnetic field change of 5.0 T was obtained in Co6 alloy. However, for evaluating the effective magnetic refrigeration capacity, the AHL should be deducted 18,45,71 . Therefore, a more reasonable criterion for assessing the cooling efficiency, the effective refrigeration capacity (RC eff ) defined as RC eff = RC-AHL, was adopted here. For comparison, the RC eff values under a field of 5.0 T for the present work and some most studied magnetocaloric materials 1,18,45,63,[72][73][74][75][76][77][78][79][80][81][82][83][84] are summarized in Fig. 8. Noteworthily, the RC eff of the present Co6 alloy reaches 223 J/kg, which is significantly larger than those of the stoichiometric Ni-Mn-based alloys 18,45,76,77,79,[81][82][83][84] and is comparable to those of some Co-doped Mn-rich Ni-Mn-based compounds 1,63,[78][79][80]83 . Besides, the achieved RC eff is also comparable to those rare-earth containing La(FeSi) 13 -based 72,73 and Gd 5 (SiGe) 4 -based 74,75 compounds. Of note is that the working temperature of the present alloy is slightly above 300 K, which is very encouraging for room-temperature magnetic refrigeration applications.

Conclusion
The effect of Co doping on the martensite transformation, magnetic properties and magnetocaloric effects were investigated in Ni 47−x Mn 43 Sn 10 Co x (x = 0, 6, 11) alloys. The main conclusions may be drawn as follows: 1) The martensite transformation temperatures decreased and Curie temperature of austenite phase increased significantly with increasing Co concentration, broadening the temperature window for a higher magnetization of austenite: 13.5, 91.7 and 109.1 A·m 2 /kg for x = 0, 6 and 11, respectively. 2) Two successive magnetostructural transformations, i.e. A → 10 M and A → 10 M + 6 M, existed in the Ni 40.6 Mn 43.3 Sn 10.0 Co 6.1 alloy whose magnetic-field-induced shift of austenite finishing temperature ΔA f / ΔH reached −2.6 K/T. 3) A sizable maximum magnetic entropy change ΔS M max 29.5 J/kg·K, a wide working temperature span ΔT FWHM 14 K and a high effective refrigeration capacity RC eff 223 J/kg were obtained in the Ni 41 Mn 43 Sn 10 Co 6 alloy under a magnetic field of 5.0 T.

Methods
Sample and heat treatment. Ni 47−x Mn 43 Sn 10 Co x (x = 0, 6, 11, atomic percent) ingots, denoted as Co0, Co6 and Co11, respectively, were prepared by induction melting pure Ni (99.99%), Co (99.99%), Mn (99.9%) and Sn (99.99%) in an argon atmosphere and casting in a copper mold. The ingots were sealed in a quartz tube accompanied with Mn powder (to create a Mn vapor atmosphere in the tube) and Ti sheet (acted as oxygen getter), evacuated under vacuum at 10 −2 Pa, annealed at 1173 K for 24 h and then quenched in iced-water. The composition of the specimens were determined by a Zeiss-SUPRA SEM equipped with an Oxford EDS, using 20 kV voltage, 97 µA emission current, 10 mm working distance, 50 µA probe current and >60 s data acquisition time duration. The composition measurement precision of the EDS was calibrated with chemical analysis results (ICP-OES) to be less than 0.5%.
Composition and martensite transformation tests. The actual and nominal compositions of the alloys were shown in Table 1. The microstructure of the specimens was examined by a Zeiss-SUPRA SEM and an Olympus PMG3 optical microscope (OM) with a polarizing filter. The martensite transformation temperatures and Curie points were measured by differential scanning calorimetry (DSC) at heating and cooling rates of 10 K/ min. XRD analyses were performed using Cu Kα radiation with a low-temperature chamber.

Magnetic property and magnetocaloric effects (MCE) evaluation. Magnetization measurements were
performed using a vibrating sample magnetometer (VSM) in a commercial Magnetic Property Measurement System (MPMS) of Quantum Design. Magnetization vs temperature (M-T) curves were recorded under magnetic fields 0.02 and 5.0 T with heating/cooling rates of 5 K/min and temperature range 200-400 K. Isothermal magnetization (M-H) curves were measured at different test temperatures (T test ) from temperatures 290-330 K under an external magnetic field up to 5.0 T. In order to rule out the temperature and field history effects and thus avoid the spurious spike, the so-called loop process 82 was performed before each M-H test. The detailed loop process was as follows: 1) The sample was initially zero-field-cooled down to 200 K to ensure a full weak-magnetic martensite state prior to recording each M-H cycle at a constant temperature, 2) Zero-field-heated to (T test −10) K at 10 K/min, then heated to T test at 1 K/min and finally maintained at the T test temperature for 2 min before starting the M-H cycle.
Data availability statement. The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.