Method of Making a Magnetic Solid Material, Magnetic Solid Material, Magnet, and

Method of Making a Magnet

The invention relates to providing a method of making a solid material having an L1 ₀ structure, a method of making a magnet, a magnetic solid material and a magnet. In particular, the invention relates to a method of forming a magnetic material comprising Fe-Ni in the L1 ₀ structure, preferably in which the magnetic material comprises tetrataenite.

Background of the Invention

Rare-earth magnets are used in a large variety of technical applications, as their superior magnetic fields and high resistance to demagnetisation provide significant advantages over conventional ferrite magnets. The production of rare-earth magnets such as Nd-Fe-B on an industrial scale, however, requires significant amounts of expensive rare-earth elements like Nd, Tb, Dy, etc.

One of the most attractive possible solutions to fulfil the demand for permanent magnets may be based on the tetragonal crystals of Fe-Ni, as the component elements of these materials are much cheaper and more readily available.

The permanently magnetic form of Fe-Ni has a particular crystal structure, known as L1 ₀ .

The L1 ₀ structure forms in approximately equiatomic compounds, and consists of alternating layers of the two constituent elements stacked parallel to the tetragonal c-axis, creating a natural superlattice. The Strukturbericht designation of this superlattice structure is L1 ₀ .

The ordered L1 ₀ phase of Fe-Ni is typically thought to form through a disorder-order transformation from a parent phase, which is expected to be a chemically disordered, cubic close-packed (ccp) solid solution of the two elements Fe and Ni. The formation of the L1 ₀ structure reduces the symmetry of the unit cell and leads to a tetragonal distortion of the unit cell in which the edge (lattice parameter) parallel to the c-axis is no longer the same length as the edges parallel to the a-axis or -axis.

In the equiatomic ccp structure, the probability of occupation of a given atomic site by either of the two component elements Fe and Ni is equal. In the ordered L1 ₀ structure, which is the thermodynamically stable phase below a critical order-disorder temperature, there are two types of atomic site, one preferably occupied by Fe atoms, the other preferably occupied by Ni atoms. In the prior art, L1 ₀ Fe-Ni has been long believed to possess a low chemical ordering temperature of 320 °C (593 K), indicating that the disorder-to-order transformation in Fe-Ni is only possible below 320 °C, and is kinetically limited on account of low atom mobilities at such low temperatures.

Fe-Ni in the L1 ₀ structure (tetrataenite) exhibits a high magnetisation (1.6 T - equivalent to Nd Fei B) and high anisotropy, which make it particularly suitable for permanent-magnet applications. The lattice parameters of ordered Fe-Ni were estimated as a = b = 0.3582 nm and c = 0.3607 nm (c/a ratio of 1.007) by synchrotron radiation-based powder X-ray diffraction (XRD) in M. Kotsugi et at. “Structural, magnetic and electronic state characterization of L1 ₀ -type ordered FeNi alloy extracted from a natural meteorite” Journal of Physics: Condensed Matter 26 (2014) 064206. The ferromagnetic Curie temperature was reported to be 823 K in P. Wasilewski “Magnetic characterization of the new magnetic mineral tetrataenite and its contrast with isochemical taenite” Physics of the Earth and Planetary Interiors, 52 (1988) 150-158. The most important parameter for application as a permanent magnet is the uniaxial monocrystalline anisotropy. The estimated value for tetragonal Fe-Ni is 1 .1-1 .3 MJ-rrr ³ . Together with the high saturation magnetisation, the theoretical energy product {BH _{m ax} ) value reaches as high as 56 MGOe (445 kJ-rrr ³ ) close to the performance of the best-known permanent magnet.

In nature, the bulk form of chemically ordered Fe-Ni phase has been found in certain meteorites. The composition of the ordered Fe-Ni phase found in meteorites was reported to contain 40-50 at.% of Ni.

Significant experimental efforts have been given to create tetrataenite through synthetic routes, but despite these efforts, bulk production of the tetragonal form of Fe-Ni remains a challenge.

Following the accepted understanding of the literature, previous experimental attempts to form tetrataenite have typically aimed to first form disordered ccp forms of Fe-Ni, and then to attempt to force a first-order transformation to the L1 ₀ structure to proceed by nucleation and growth of L1 ₀ ordered regions within the disordered ccp matrix, while maintaining the alloy at a temperature below the accepted order-disorder temperature of 320 °C.

One such attempt to prepare an Fe-Ni alloy having the L1 ₀ crystal structure otherwise known as tetrataenite is disclosed in US2014/0210581 A1 by Lewis et al. According to US2014/0210581A1 , a melt of composition Fe ₍ o _.5-a) Ni ₍ o _.5-b) X _(a+b) (where X is Ti, V, Al, S, P, B or C and wherein 0 < (a+b) < 0.1) is prepared and then rapidly solidified by a melt spinning process. The melt-spun solid ribbons are then mechanically milled to reduce the solid form to a plurality of nanoparticles. Lewis states that mechanical milling is an essential step which creates point defects in the structure to facilitate the A1 (ccp) to L1 ₀ transformation in Fe-Ni, so that the milling step induces chemical ordering in nominally equiatomic Fe-Ni so that an L1o structure forms. During milling, the temperature is kept below the order-disorder temperature of 320 °C, for example by cryomilling.

Although US2014/0210581 A1 discusses the theory behind the L1 ₀ structure and claims a broad range of compositions and processing parameters, US2014/0210581 A1 seems to be purely theoretical, as no details of specific examples or experimental results are disclosed.

Another attempt to prepare tetrataenite is disclosed in WO2016/036856A1 , also by Lewis et al. Similarly to US2014/0210581 A1 , WO2016/036856A1 states that a melt of composition Fe ₍ o _. 5- _a) Ni ₍ o _.5-b) X _(a+b) (where X is Ti, V, Al, B or C and wherein 0 < (a+b) < 0.1) is prepared and then solidified, preferably by a melt-spinning process. The melt-spun ribbons are subjected to a severe plastic deformation process below 320 °C (as this is believed to be the chemical ordering temperature of the L1 ₀ phase), to produce deformed particles of Fe-Ni alloy material with a ccp crystal structure. The deformed Fe-Ni alloy is then annealed at a temperature below 320 °C for a period of time from hours to months, whereby WO2016/036856A1 states that the L1 ₀ structure is formed to yield the magnetic Fe-Ni ordered compound.

Although WO2016/036856A1 speculates on possible compositions and processing parameters which may be used to yield tetrataenite, the only examples evidenced in WO2016/036856A1 related to melts with nominal compositions Fe ₅ oNi ₅ o and Fe ₉ Ni ₄₉ Ti ₂ , which were synthesised by drop-casting and confirmed to have a ccp structure prior to milling and annealing. The alloys were annealed for 100 hours at 500 °C to ensure that homogeneous ccp structures were achieved. The samples were then plastically deformed by cold-rolling, before annealing at 290 °C for 6 weeks. Lewis etal. conclude that the critical step of forming the L1 ₀ structure takes place during the final low-temperature annealing step.

The approach conventionally taken in the prior art is therefore based on a first step of intentionally forming a ccp-structured alloy, followed by further processing steps carried out below the order-disorder temperature of 320 °C. In prior art by Lewis etal., processing steps of plastically deforming the alloy to introduce point defects in the crystal structure, and annealing the deformed alloy for long periods are always used to convert ccp solid material into the desired L1 ₀ structure. Summary of the Invention

The invention provides a method of making a solid material having an L1 ₀ structure, a method of making a magnet, a magnetic material and a magnet, as defined in the appended independent claims, to which reference should now be made. Preferred or advantageous features of the invention are set out in dependent subclaims.

According to a first aspect of the invention, there is provided a method of making a solid material by solidification from a melt, the melt containing Fe, Ni, and one or more additives Z, such that the solid material comprises a phase having an L1 ₀ structure, in which the proportion of the one or more additives Z in the L1 ₀ phase is lower than the proportion of Z in the melt.

The method may be described as a method of making a magnetic solid material, or a method of making a solid material having an L1 ₀ structure.

The phase having an L1 ₀ structure, which may be termed the L1 ₀ phase, is a crystalline phase, and the solid material may comprise the crystalline L1 ₀ phase dispersed in a metallic matrix or one or more other solid phases, which may include an amorphous metallic phase.

The phase having an L1 ₀ structure is Fe-Ni with an L1 ₀ structure, otherwise known as tetrataenite. As solid Fe-Ni having an L1 ₀ structure is magnetic, the method of the first aspect of the invention is preferably a method of making a magnetic material, particularly a method of making tetrataenite.

The L1 ₀ structure of Fe-Ni is illustrated in Figure 2d, and is well known in the art. The prototype structure for L1 ₀ is CuAu I. The presence of a crystalline L1 ₀ phase in the solid material is identifiable by X-ray diffraction (XRD) examination, electron diffraction and high- resolution scanning TEM, as described in relation to the figures below. In particular, these characterisation techniques can be used to identify the tetragonal distortion of the c/a ratio in the Fe-Ni unit cell that is characteristic of the L1 ₀ crystal structure. The L1 ₀ crystal structure is assigned the tetragonal space group 123 P4/mmm.

The solid material made by the present method is an alloy, and may comprise the L1 ₀ phase and possibly further phases not having an L1 ₀ structure. In the present method, the phase having an L1 ₀ structure is formed by solidification from a melt. In other words, the phase having an L1 ₀ structure may advantageously be formed straight from a melt in a single solidification step, without requiring any subsequent processing.

The melt is solidified by cooling the melt to a temperature below its freezing point, so that a solid material is formed.

The solid material is an alloy of the ingredients of the melt, and may comprise a plurality of different solid phases, at least one of which is a crystalline phase having an L1 ₀ structure.

As the melt is cooled to a temperature below its freezing point, a first solid phase is formed. As solidification progresses and more of the melt solidifies, solute partitioning occurs at an interface between the first solid phase and the melt, such that, at the interface between the first solid phase and the melt, a portion of the one or more additives Z is rejected from the first solid phase into the melt by solute partitioning. The first solid phase is preferably a single phase which forms before any other solid phase. On solidification, the first solid phase may have an L1 ₀ structure that is retained as the solid material cools to room temperature, or alternatively the first solid phase may initially have a ccp structure that later orders into the L1 ₀ structure as the first solid phase cools to room temperature.

Regardless of the order in which solid phases are formed during the solidification process, the solidification results in a solid material containing an L1 ₀ phase which contains a lower proportion of additive(s) Z than is present in the melt.

At a temperature of 25 °C, for example, the solid material comprises a phase having an L1 ₀ structure.

Surprisingly, the inventors have found that the crystalline phase having an L1 ₀ structure may be formed from the melt in a single solidification step. This has been achieved by cooling the melt such that, at an interface between the first crystalline solid and the melt, a portion of the one or more additives Z is rejected from the first solid phase into the melt by solute partitioning. This solute partitioning causes the additives to be naturally rejected from the first crystalline solid phase, so that the first crystalline solid phase contains the Fe, Ni, and a small proportion of the one or more additives Z. Without the need for any further processing steps, a crystalline phase having an L1 ₀ structure is formed in the first crystalline solid phase.

This differs significantly from the teaching in the prior art that L1 ₀ Fe-Ni can only be formed by first solidifying a ccp solid from the melt, and then by plastically deforming and annealing the ccp solid below 320 °C to force a phase transformation to L1 ₀ to take place. In the present invention, the phase having an L1 ₀ structure is formed during solidification from the melt, and/or during cooling of the solid material. In other words this L1 ₀ phase is formed directly from the melt, without requiring further processing steps to be carried out on the solid material. Even if further processing steps are carried out for some other reason, this is optional, as the L1 ₀ phase is formed before those further processing steps take place. This is different from prior art methods in which a melt is solidified into a solid amorphous material or a solid ccp material not containing the L1 ₀ phase, and subsequently further processing steps such as heat-treatment or plastic deformation are performed on the solid material to form an L1 ₀ phase.

The inventors have found that this method is usable to form permanently magnetic L1 ₀ phases from a variety of initial melt compositions, and that the method does not require rapid solidification techniques having high cooling rates.

The method may additionally comprise the step of separating the phase having an L1 ₀ structure from the solidified or partly solidified melt, and preferably also the step of fabricating a magnet from the L1 ₀ phase.

The Melt

The composition of the melt may be expressed with the formula Fe _a NipZ _Y , in which Z is one or more additives. For example, the melt may have the composition Fe _a Ni _p Z _Y . Preferably (a + b + Y) = 1 .

The one or more additives Z preferably comprise a non-metal or metalloid additive. The one or more additives Z may be one or more non-metal or metalloid additives.

The one or more additives Z may comprise one or more additives selected from the list: B, C, Si, P, S, Ge, As, Sb, Te.

The one or more additives Z preferably comprise phosphorus, P. The melt may contain P and one or more further additives Z, or in other embodiments P may be the only additive in the melt, such that the melt contains only Fe, Ni, and P.

In addition to P, or alternatively to P, the one or more additives Z may comprise one or more of C, B, Si, S, Al, V, Cu, Mn and Ti, or one or more of C, S, Al, V, Mn and Ti. In preferred embodiments, the melt may comprise P and also one or more of these additives, for example P and C.

In preferred embodiments, the melt does not comprise Cu. In preferred embodiments, the melt does not comprise Si, and/or the melt does not contain B.

In preferred embodiments, the composition of the melt is Fe _a Ni _p Z _Y and in which a ³ 0.4, preferably 0.4 £ a £ 0.55. In other words, the melt preferably comprises 40 at.% Fe or more, preferably 40 at.% to 55 at.% Fe.

Preferably b ³ 0.3, particularly preferably in which 0.3 < b < 0.45. In other words, the melt preferably comprises greater than or equal to 30 at.% Ni, particularly preferably 30 at.% to 45 at.% Ni.

The content of additive in the melt Fe _a Ni _p Z _Y is defined by Z _Y , and preferably 0.005 < g < 0.25, particularly preferably 0.01 < g < 0.25, or 0.02 < g £ 0.22, or 0.05 < g < 0.20. In other words, the melt preferably comprises greater than or equal to 0.5 at.% additive(s) Z, particularly preferably 1 at.% to 25 at.% additive(s) Z, 2 at.% to 22 at.% additives Z, or 5 at.% to 20 at.% additives Z.

Preferably, the additives Z comprise P and optionally C, as the inventors have found that melts containing phosphorus and optionally carbon form the desired L1 ₀ phase on solidification. Particularly preferably the additives Z consist of P and optionally C, so that the melt has the formula Fe _a Ni _p P _n Ce, wherein h + Q = y.

The presence of an additive such as P and optionally C in the melt is understood to be a key factor in the formation of the permanent-magnet L1 ₀ phase on solidification from the melt. The present inventors consider it possible that the presence of the additive in the melt may be an important factor in enabling or accelerating the required chemical ordering during cooling, which allows the L1 ₀ phase to form straight from solidification, or during the natural cooling thereafter. Alternatively, the additives may have the effect of dramatically raising the order-disorder temperature of the composition, to a temperature at which the material can readily adopt the L1 ₀ structure on solidification from the melt.

Prior art attempts to synthesise tetrataenite have started from melt compositions containing Fe, Ni and other additives such as Ti, Mo, Ir or Nb. The prior art melt compositions did not contain P, and did not contain C, and those prior art studies confirmed that the solid formed by solidifying the melt had a ccp structure, and not the L1 ₀ phase obtained by the present invention.

In preferred embodiments, the melt comprises P, and the melt may have the formula Fe _a XpPnZe. Preferably 0.01 < (h + Q) £ 0.25. Preferably 0.005 < h < 0.25, or particularly preferably 0.01 < h < 0.20. Preferably 0 < Q £ 0.25, or 0 < q £ 0.15, or 0 < q £ 0.10. The inventors have found that compositions having phosphorus contents in this range successfully form the desired L1 ₀ phase on solidification straight from a melt.

In particularly preferred embodiments where the additives comprise or consist of P and C, the melt has the formula Fe _a XpP _n Ce. Preferably 0.01 £ (h + Q) £ 0.25, or particularly preferably 0.05 £ (h + Q) £ 0.20. Preferably 0.005 < h < 0.25 and 0 < Q £ 0.25. Particularly preferably 0.01 £ h < 0.13 and 0 < Q £ 0.07. The inventors have found that melts having phosphorus and carbon contents in this range form the desired L1 ₀ phase on solidification.

In particular, the presence of phosphorus in the melt may be a key factor in the formation of the permanent-magnet L1 ₀ phase on solidification. In preferred embodiments, the melt therefore comprises P, optionally in addition to further additives Z. Preferably the melt comprises at least 0.5 at.% or 1 at.% phosphorus.

The melt is preferably prepared by combining the constituent elements in the desired atomic fraction and melting under vacuum or in an inert atmosphere. The melt is a liquid, or molten, mixture of the constituent elements of the melt composition, and may be formed by combining the constituent elements (or compounds of the constituent elements) in the correct proportions for the desired melt composition, and heating the mixture to a temperature at which all of the ingredients have mixed into a liquid form. The melt may be formed by conventional means, for example by arc-melting.

The Solid Material

In the solid material that is formed from the melt, there is a phase having an L1 ₀ structure. Once the solid material has cooled to a temperature of 25 °C, for example, the solid material contains a phase having an L1 ₀ structure.

The composition of the solid material, and the phases within the solid material, is determined by the solute partitioning process that takes place during solidification. A first solid phase is formed on solidification from the melt, and at an interface between the first solid phase and the melt, a portion of the one or more additives Z is rejected from the first solid phase into the melt by solute partitioning. As the solute partitioning process causes a portion of the one or more additives Z to be rejected from the first solid phase as it forms, however, the composition of the first solid phase is different from that of the melt. When cooled to room temperature, the first solid phase has an L1 ₀ structure.

While the composition of the melt may be expressed with the formula Fe _a NipZ _Y , the composition of the L1 ₀ phase in the solid material may be expressed as is FeeNi _£ Z _? . Preferably (d + e + z) = 1. The solute partitioning that takes place during solidification means that the L1 ₀ phase contains a lower additive content than the melt, such that z < y.

The Fe:Ni ratio in the solid L1 ₀ phase itself is also higher than the Fe:Ni ratio in the melt.

Where the melt has the formula Fe _a Ni _p Z _Y , the solid material comprises tetrataenite as the L1 ₀ phase.

Where the composition of the L1 ₀ phase in the solid material is FeeNi _£ Z _? preferably 0.001 < z £ 0.03, particularly preferably, 0.005 < z < 0.02. In other words, the content of additive which remains in the solidified L1 ₀ phase is likely to be between 0.1 at.% and 3 at.%, or between 0.5 at.% and 2 at.%. The inventors consider that some non-zero amount of additive Z must remain in the L1 ₀ phase in order for the L1 ₀ phase to successfully form from the melt, the solute rejection during solidification necessarily means that the additive-content of the solid L1 ₀ phase is significantly lower than that of the initial melt.

In preferred embodiments, the L1 ₀ phase may comprise 0.3 at.% to 1 at.% additive, particularly preferably 0.5 at.% to 1 at.% additive.

In particularly preferred embodiments, the composition of the L1 ₀ phase in the solid material is FeeNi _£ P , and 0.001 < z < 0.03, preferably 0.002 £ z £ 0.01 or 0.004 £ z £ 0.01 . Thus, the L1 ₀ phase may comprise 0.1 at.% to 3 at.% P, particularly preferably 0.2 at.% to 1 at.% P.

Where the composition of the crystalline phase having an L1 ₀ structure is FeeNi _£ Z _? , preferably the iron content Fee is 0.5 < d £ 0.65, or 0.5 < d £ 0.65, and particularly preferably 0.55 < d £ 0.65. Preferably the nickel content, Nί _e , is 0.35 < e £ 0.5, or particularly preferably 0.35 < e £ 0.45. Unlike prior art attempts which focused on 50-50 iron-nickel compositions, the iron content Fee of the L1 ₀ phase formed by the present invention is preferably greater than 50 at.%. Particularly preferably, the L1 ₀ phase formed by the present method may have an Fe content of 53 at.% to 64 at.%. The inventors have found that compositions within this range successfully demonstrate the ordered tetragonal L1 ₀ structure.

In the solid material formed from the melt, the proportion of the one or more additives Z in the L1 o phase is lower than the proportion of Z in the melt. In other words, the composition of the melt is Fe _a Ni _p Z _Y , and the composition of the L1 ₀ phase in the solid material is FeeNi _£ Z _? , and z < y. Where the composition of the melt is Fe _a XpPnZ _e , the composition of the L1 ₀ phase in the solid material may be FeeNi _£ P , in which z < (h + Q) and z < h.

For example, the inventors have successfully prepared crystalline L1 ₀ phases having eight compositions in the range from Fe ₅ 3Ni ₄ 6 to Fe ₆ 3Ni ₃ 6 using the present method. The quantity of additives Z remaining in the solid L1 ₀ phases was found to be as low as 0.4 at.%, and all of these compositions have demonstrated the expected tetragonality that confirms the presence of tetrataenite.

The solid material may comprise a second solid phase in contact with the L1 ₀ phase, in which the second phase contains a higher proportion of the additive Z than the L1 ₀ phase.

The solid material may comprise the L1 ₀ phase (which is inherently crystalline) dispersed in another metallic phase, which may be a crystalline or amorphous metallic matrix. The metallic matrix may comprise one or more metallic or metal-additive phases. For example, when C and P are used as additives in the melt, the metallic matrix may comprise an Fe-Ni-C phase and an Fe-Ni-P phase.

The L1 o crystalline magnetic phase may be dispersed as a plurality of dendrites or other crystalline forms in a polycrystalline metallic matrix.

Solidification Process

Prior art attempts to synthesise tetrataenite from melts have used melt spinning (US2014/0210581 A1 ) to solidify melts having desired compositions. Melt-spinning is a “rapid solidification” technique, in which the cooling rate is in the range of 10 ⁵ to 10 ⁶ K/s. However, the authors of US2014/0210581A1 used this cooling technique to arrive at alloys of the desired composition having a disordered ccp structure. In US2014/0210581 A1 and WO2016/036856A1 the authors have focused on the use of severe plastic deformation and annealing below the order-disorder temperature of 320 °C to force a phase transformation to occur, in order to eventually arrive at a material containing an L1 ₀ phase. In the present method, the inventors have found that crystalline material having an L1 ₀ structure can be solidified directly from a melt. Severe plastic deformation steps and annealing are advantageously not required in order to arrive at the L1 ₀ structure. Thus the present method preferably does not comprise plastic deformation. The present method preferably does not comprise annealing below a temperature of 320 °C.

The inventors have found that this is possible over a range of parameters, for example using macroscopic casting techniques (on a copper stage or in cylindrical copper moulds) with cooling rates broadly in the range 10 to 10,000 K-s ^_1 , such that the particular method used to solidify the melt does not appear to be a determining factor in and of itself. The inventors consider that the key requirement for direct formation of the L1 ₀ phase is the ability for solute partitioning to take place during solidification.

Another factor that the inventors consider to be key in the present invention is that the formation of the chemically ordered tetragonal L1 ₀ structure in Fe-Ni may occur above the generally accepted Fe-Ni order-disorder temperature T ₀ D of 320 °C, which is in stark contrast to what is generally believed and assumed by a person skilled in the art and which is not reflected in the accepted understanding of the literature.

The majority of prior art approaches have relied on forcing an ordering phase transformation from ccp to L1 ₀ to occur below the accepted T ₀ D of 320 °C. At this temperature, atomic mobility is extremely low, which is why, for example, WO2016/036856A1 annealed samples at 290 °C for 6 weeks.

The present inventors consider, however, that the generally accepted Fe-Ni order-disorder temperature is incorrect, and that the actual T ₀ D may be significantly higher than 320 °C. This value was stated in the prior art for the ideal composition Fe ₅ oNi ₅ o (at.%); it may be incorrect even for that composition, or T ₀ D may vary significantly with composition, even for small deviations from the ideal stoichiometry.

Although investigations are ongoing, the inventors estimate that the actual T ₀ D is around 550-600 °C (823-873 K), though the actual T ₀ D may be sensitive to the exact compositions being considered.

One theory on how an L1 ₀ phase has been obtained directly from solidifying melts is that the liquid solidifies directly to the L1 ₀ phase. The inventors consider that a more plausible theory is that on cooling, the melt first forms a ccp-structured solid phase, and then that ccp phase subsequently automatically undergoes an ordering transition to L1 ₀ during the same solidification step (i.e. without requiring further processing steps of deformation or annealing). For that ccp to L1 ₀ transition to occur during cooling, the order-disorder temperature must be sufficiently high for the transition to occur during the timescale of cooling the melt, as at lower temperatures there is not enough atomic mobility. The inventors consider that a T ₀ D of around 550 °C to 600 °C would explain the L1 ₀ ordering that has been observed to occur on cooling directly from the melt.

The step of forming the solid material from the melt may comprise the step of cooling the melt such that a portion of the melt solidifies into the first solid phase.

The method preferably comprises the step of solidifying the phase having an L1 ₀ structure at a temperature greater than 320 °C, and preferably greater than 450 °C, particularly preferably greater than 500 °C. The first solid phase may therefore be formed at a temperature greater than the generally accepted Fe-Ni order-disorder temperature of 320 °C. The teaching in the prior art would suggest that it is impossible to obtain an L1 ₀ phase at temperatures greater than 320 °C, as the L1 ₀ structure of Fe-Ni has long been considered to be thermodynamically unstable at such temperatures.

The inventors consider that it is highly likely that the solidification of the solid material from the melt occurs close to equilibrium (close to the phase diagram), such that solidification starts at a temperature above 1100 K.

The solid material, containing the phase having an L1 ₀ structure, may be formed or solidified from the melt by casting, additive manufacturing, or directional solidification. For example, casting at a variety of cooling rates has been tried and found to successfully produce the magnetic L1 ₀ phase.

The method may comprise solidifying the solid material, containing the L1 ₀ phase, in a batch process or a continuous process.

In the present method the solid material comprises a phase having an L1 ₀ structure before carrying out any optional further processing steps. Unlike in the prior art, no further processing steps are necessary after the solid material has been formed from the melt, as the phase having an L1 ₀ structure is already present in the solid material. Any further processing steps are therefore optional, and are not responsible for the formation of the phase having an L1 ₀ structure.

In the present method, the L1 ₀ phase is formed without the need for any annealing step. However, the L1 ₀ structure may optionally be optimized by annealing. For example, the present method may comprise the step of annealing the solid material at a temperature greater than 320 °C, and preferably lower than 550 °C. In the prior art, plastic deformation and annealing steps carried out on the ccp alloy have been intentionally restricted to temperatures below 320 °C. In the present invention, however, as the inventors have found the actual order-disorder temperature to exist around 550 °C to 600 °C, annealing the solid material at a temperature between 320 °C and 550 °C, or between 350 °C and 500 °C, may advantageously improve the degree of L1 ₀ ordering, while providing significantly higher atomic mobility than prior art annealing steps below 320 °C.

In the present method, even where an annealing step is used, the solid material comprises a phase having an L1 ₀ structure before carrying out any annealing step.

During the process of solidifying the solid material from the melt there must be a solid-liquid interface between the first crystalline solid phase and the melt, and the melt close to the interface contains a higher proportion of the one or more additives Z than a portion of the first solid phase close to the interface.

During solidification, the solid material close to the interface may have the same composition as the melt far from the interface. In the special case of well controlled directional solidification, for example, the temperature gradient is so steep that the solid-liquid interface is flat (i.e. no cells or dendrites). In that case, after the initial transient in composition as the first solid phase forms in the melt, the solid may grow with the same composition as the liquid and could be 100 % tetrataenite.

The melt may be solidified to form the L1 ₀ phase and a second solid phase in contact with the L1 o phase, in which at least a portion of the second solid phase adjacent to the L1 ₀ phase contains a higher proportion of the additive Z than the L1 ₀ phase. For example, as the melt is cooling and the solid L1 ₀ phase is being formed, solute rejection of the additives Z from the first solid phase into the melt will mean that the melt contains a higher and higher additive- content as the solidification progresses. After a certain point, the remaining melt having a high additive concentration may solidify to form a second solid phase which, thanks to the high additive-content, has a different crystal structure from the L1 ₀ phase.

The step of forming the solid material from the melt involves cooling the melt, so that at least a portion of the melt has a temperature below the freezing point of the first solid phase, and the first solid phase begins to form.

The method preferably comprises the step of cooling the melt at a cooling rate at which solute partitioning takes place at the solid-liquid interface between the first solid phase and the melt.

In preferred embodiments, the method may comprise the step of cooling the melt at a cooling rate of between 1 10 ^-2 K-s ^_1 and 2 10 ⁴ K-s ^_1 . Cooling rates of roughly 1 10 ^-2 K-s ^_1 are used in processes such as the continuous casting of steel, and are expected to be suitable for the present method as the slow cooling rate is likely to allow time for the solute partitioning to take place at temperatures below the real T ₀ D of the composition, but high enough that atomic mobility allows the L1 ₀ phase to form.

Preferably the melt may be cooled at a cooling rate of between 1 K-s ^_1 and 1 10 ³ K-s ^_1 , preferably between 10 K-s ^_1 and 1 10 ² K-s ^_1 .

The melt spinning technique suggested in some prior art documents typically has a cooling rate of around 10 ⁵ — 10 ⁶ K-s ^_1 . The present invention therefore differs from the prior art in the cooling rates used during solidification.

At very high cooling rates, solidification is too fast to permit solute partitioning. When partitioning is eliminated, or even just substantially reduced, the formation of solid dendrites from the melt is hindered, they lose their side-branches and become ‘cells’, and ultimately the solid-liquid interface is simply planar. Lower cooling rates may therefore be preferred in order to allow solute partitioning to occur, and to allow the atomic rearrangement that may be required to form the L1 ₀ phase directly from the melt.

The method may comprise the step of applying a magnetic field during solidification, in particular applying a magnetic field at the solid-liquid interface during solidification of the first solid phase. This may advantageously polarise the magnetic L1 ₀ phase as it is formed.

Steady-state directional solidification One particularly preferred method of forming the solid material from the melt is directional solidification.

Solid formation from a supercooled liquid typically takes the form of solid dendrites growing into the supercooled liquid. In the present case, when the melt is cooled to below its freezing point, dendrites of the desired L1 ₀ phase may grow into the melt. The dendrites are somewhat rich in Fe and Ni compared to the liquid alloy, which is rich in additive Z such as P and C. The dendritic growth form is a natural instability caused by the solute partitioning between solid and liquid. If the temperature gradient is high enough, however, then that hinders the dendritic instability. Indeed, if the gradient is high enough, the solid-liquid interface is forced to be planar (lying very close to the isotherm at the liquidus composition).

With suitable directional solidification apparatus, a high temperature gradient of this kind can be imposed on the sample and translated along the sample length. In the Bridgman- Stockbarger technique, for example, the sample is lowered at a carefully controlled rate through a furnace with a zone of the controlled temperature gradient, and in this way steady- state directional solidification can be achieved.

Making a magnet

According to a further aspect of the present invention there is provided a method of making a magnet, comprising the steps of solidifying a melt as defined in any preceding claim, isolating the L1 ₀ phase, and forming the magnet from the L1 ₀ phase.

The step of isolating the L1 ₀ phase may involve separating the L1 ₀ phase from any other non- Li o phase in the solid material.

The L1o phase is preferably formed without plastic deformation.

In the present method, the L1 ₀ phase is formed without the need for any annealing step. However, the L1 ₀ structure may optionally be annealed, which may lead to improvements in the L1o structure.

The method may comprise the steps of milling the L1 ₀ phase to form a powder, and/or compacting the L1 ₀ phase into a magnet. Magnetic Solid Material

According to a further aspect of the present invention there is provided a magnetic solid material comprising an L1 ₀ structure, the composition of the solid material being FeeNi _£ Z _? , in which Z is one or more additives.

The one or more additives Z preferably comprise P, or consist of P.

Instead of, or in addition to, P, the one or more additives Z comprise one or more of C, B, Si, S, Al, V, Cu, Mn and Ti. Alternatively, the one or more additives Z comprise one or more of C, B, Si, S, Al, V, Cu, Mn. In some embodiments Z comprises C in addition to P.

The composition of the L1 ₀ phase in the solid material is FeeNi _£ Z _? , and preferably 0.5 < d £ 0.65, preferably in which 0.5 < d £ 0.65, and particularly preferably in which 0.55 < d £ 0.65. In preferred embodiments, 0.001 < z < 0.03, preferably in which 0.005 < z < 0.01 . The contents of the composition add up to 100% - in other words (d + e + z) = 1.

Particularly preferably the composition of the L1 ₀ phase in the solid material is FeeNi _£ P , in which 0.001 < z < 0.03, preferably in which 0.004 £ z £ 0.01 .

The magnetic solid material is preferably a permanently magnetic solid material.

The magnetic solid material is preferably the solid material formed by the method of the first aspect of the invention. Features of the invention discussed in relation to the other aspects of the invention may apply equally to the magnetic solid material of the present aspect.

Magnet

According to a further aspect of the present invention there is provided a magnet comprising a solid material having an L1 ₀ structure, the composition of the L1 ₀ solid material being Rb _d Nί _e Z _z in which Z is one or more additives.

The L1 o solid material has the composition FeeNi _£ Z _? , which may be termed tetrataenite. The magnet may therefore be a magnet comprising tetrataenite. The one or more additives Z preferably comprise one or more non-metal or metalloid additives, such as one or more additives selected from the list: B, C, Si, P, S, Ge, As, Sb, Te. The one or more additives Z may comprise one or more of P, C, B, Si, S, Al, V, Cu, Mn and Ti. Preferably the one or more additives Z comprise P, and optionally one or more of the other additives Z. Particularly preferably the one or more additives Z comprise P and C or consist of P and C.

In some preferred embodiments the composition of the L1 ₀ solid material is Rb _d Nί _e Rz.

Preferably 0.001 < z < 0.03, preferably in which 0.004 £ z £ 0.02, or 0.01 < z < 0.015. That is, the additive Z content of the L1 ₀ solid material may be between 0.1 at.% and 3 at.%, preferably between 0.4 at.% and 2 at.% or between 1 at.% and 1 .5 at.%.

The iron content of the L1 ₀ solid material Fee may be 0.5 < d £ 0.65, preferably 0.5 < d £ 0.65, and particularly preferably 0.55 < d £ 0.65.

The solid material having an L1 ₀ structure preferably has an order-disorder transition temperature greater than 320 °C and preferably greater than 350 °C, or 400 °C, or 450 °C, or 500 °C.

Alternative Aspects

In a further aspect of the invention there is provided a method of making a magnetic material, comprising the steps of: preparing a melt containing Fe, Ni, and one or more additives Z; and cooling the melt at a cooling rate of between 1 c 10 ^-2 K-s ^_1 and 2 c 10 ⁴ K-s ^_1 , so that the melt solidifies into a solid material that at 25 °C comprises a crystalline magnetic phase with a L1 ₀ crystal structure.

In a further aspect of the invention there is provided a method of making a magnetic material, comprising the steps of: preparing a melt containing Fe, Ni, and one or more non-metal or metalloid additives , and forming a solid material from the melt, the solid material comprising a phase having an L1 ₀ structure.

The one or more additives may be selected from the list: B, C, Si, P, S, Ge, As, Sb, Te. The one or more additives may preferably comprise one or more of P, C and S. Particularly preferably the one or more additives may comprise P, and optionally C. The method may comprise the step of cooling the melt to obtain the solid material comprising a phase having an L1 ₀ structure directly from the melt. Alternatively, the method may comprise the steps of cooling the melt to obtain a solid material, and subsequently annealing the solid material at a temperature of between 320 °C and 550 °C, or between 350 °C and 500 °C, to obtain a solid phase having an L1 ₀ structure.

In a further aspect of the invention there is provided a method of making a magnetic material, comprising the steps of: preparing a melt containing Fe, Ni, and P, and forming a solid material from the melt, the solid material comprising a phase having an L1 ₀ structure.

The melt may additionally comprise C. The inventors have found that melts containing phosphorus alone, and also a combination of phosphorus and carbon reliably form the desired L1 ₀ phase on solidification.

Features described above in relation to any one aspect of the invention are equally applicable to every other aspect of the invention.

Brief Description of the Drawings

The invention will now be described, by way of example only, by reference to the following figures, in which:

Figure 1 shows the shapes of samples made by arc-melting and casting;

Figure 2a is a photograph showing an as-cast rod of Fe ₅ oNi ₃ oPi3C ₇ (at.%) alloy formed using the method of the present invention;

Figure 2b is a scanning electron microscopy (SEM) image that shows the microstructure in the rod in Figure 2a, showing dendrites of Fe-Ni monocrystals formed during solidification, with an inset showing the tetragonal L1 ₀ structure of Fe-Ni;

Figure 2c shows a high-angle annular dark-field (HAADF) STEM image and EDX elemental maps around one dendrite side-arm in the rod in Figure 2a;

Figure 2d shows the unit cell of the tetragonal L1 ₀ structure of Fe-Ni; Figure 3 shows X-ray diffraction data from Fe ₅ oNi ₃ oPi3C ₇ alloy both as-cast and after annealing at 1123 K for 15 min, with an inset showing peak splitting due to the tetragonal L1 ₀ structure of the as-cast dendritic monocrystals;

Figure 4a to 4d show transmission electron microscopy (TEM) measurements of the tetragonality of the Fe-Ni phase within as-cast Fe ₅ oNi ₃ oPi3C ₇ ;

Figure 4a shows selected-area electron diffraction (SAED) on the [110] zone axis;

Figure 4b shows intensity profiles along the solid and dashed lines in Figure 4a;

Figure 4c shows a high-resolution STEM image on the [110] zone axis;

Figure 4d shows intensity profiles along the solid and dashed lines in Figure 4c;

Figure 5a shows a bright-field TEM image of the Fe-Ni primary phase extracted from an as- cast Fe55Ni ₃₅ P _6. 5C ₃ .5 button;

Figure 5b and Figure 5c show differential phase contrast images of the magnetic domain structures in the primary phase in Figure 5a;

Figure 5d and Figure 5e, for comparison, show differential phase contrast images of the magnetic domain structure in a sample of ccp Fe ₆ oNi ₄ o;

Figure 6a, Figure 6b and Figure 6c show SEM images of the microstructures of as-cast Fe ₅ oNi ₃₀ Pi3C ₇ rods of 1 mm, 2 mm and 3 mm in diameter;

Figure 6d, Figure 6e and Figure 6f show SEM images of the microstructures of as-cast Fe ₅ 3Ni ₃₂ P9. ₇ 5C ₅ .25 rods of 1 mm, 2 mm and 3 mm in diameter;

Figure 7a is an SEM image showing the microstructure of an as-cast Fe ₅ 5Ni ₃₅ P6 _. 5C ₃ .5 button;

Figure 7b is an SEM image showing the microstructure of an as-cast Fe ₅ 6Ni ₃₆ P5 _. 2C2.8 button;

Figure 8a is an SEM image showing the microstructure of an as-cast Fe58 _. 5Ni38 _. 5P3 button;

Figure 8b is an SEM image showing the microstructure of an as-cast Fe59 _. 5Nis9 _. 5P1 button;

Figure 9a shows the temperature dependence of the equilibrium long-range order parameter in CU ₃ AU; Figure 9b shows the possible evolution of the long-range order parameter upon heating Cu ₃ Au that is initially partially disordered;

Figure 10 shows the ccp-to-L1 ₀ ordering upon heating disordered CuAu;

Figure 11a shows the order-disorder transition in L1 ₀ Fe-Ni through in-situ TEM measurements of the tetragonality (c/a) on heating and cooling past the order-disorder temperature;

Figure 11b shows the temperature-time profile for the in-situ heating and cooling;

Figure 12a shows the SAED pattern of the L1 ₀ Fe-Ni phase in an as-cast FessNhsPe.sCs.s button;

Figure 12b shows the SAED pattern, now indicating a ccp phase, from the sample in Figure 12a after it has been heated and cooled past the order-disorder temperature.

In order to explore the conditions under which the L1 ₀ phase may be formed using the present method, a variety of alloy compositions have been tested and cast in various sizes and at different cooling rates. The resulting solids have been characterized for example by scanning electron microscopy SEM, which confirms the size and proportion of dendrites in the solid material, and transmission electron microscopy TEM, to confirm the phase identification of Uotetrataenite, and to obtain some insight on magnetic properties.

The inventors have found that tetrataenite can be formed starting from a variety of initial melt compositions, and using a variety of cooling methods and cooling rates. By starting with different melt compositions, the inventors have also obtained a range of different compositions of the L1 ₀ crystalline phase. For example, tetragonal L1 ₀ phases have been formed with compositions in the range: Fe ₆ 4Ni ₃ 6 to Fe ₅₃ Ni ₄ 7; these all also have a low content (<1 at.%) of phosphorus.

In Example 1 set out in detail below, a melt with composition Fe ₅ oNi ₃ oPi ₃ C ₇ (atomic percent) was cast into rods with 1 mm diameter in a water-cooled copper mould (cooling rate >10,000 K-s ^_1 ). Tetrataenite was found to be present in the solid obtained directly from the melt, without further steps of plastic deformation and/or annealing. In further experimental examples, the inventors have tested melts containing different additive contents. Compositions of Fe-Ni-P-C containing 15 at.% (P and C), 10 at.% (P and C), 8 at.% (P and C), 3 at.% P and 1 at.% P have been tested and found to yield the L1 ₀ structure with similar degree of tetragonality as Fe ₅ oNi ₃ oPi C ₇ , as described in Example 3 below.

Two different conventional apparatuses were used to prepare the samples discussed in the following examples.

Samples in the form of rods 100 were prepared using the suction casting crucible shown schematically in Figure 1a, which is made up of an arc-melting stage 10 and a mould 20 that contains a cylindrical cavity positioned underneath the arc-melting stage 10. Both the stage 10 and the mould 20 are typically formed from copper. An arc-melting electrode 30 is positioned near the stage 10.

In use, raw materials are arc-melted on the arc-melting stage 10 to form a melt, followed by suction casting into the cylindrical mould cavity that (for different moulds) has a diameter of 1 mm, 2 mm or 3 mm. The mould 20 is cooled with water at 286 K, and the melt is cast in the form of rods having a diameter of 1 mm, 2 mm or 3 mm. The cooling rate during solidification of such a rod is estimated to be in the range 10,000 K-s ^_1 to 20,000 K-s ^_1 . (The basis for the estimation is explained in Example 3.)

Figure 1 b shows an alternative apparatus, in which a crucible 40 is formed by a recess in the surface of an arc-melting stage 50. In use, the raw materials are placed in the crucible, and arc-melted by the electrode 30. The arc-melted alloy solidifies to a roughly hemispherical “button” 200 with a diameter of roughly 2 cm. When arc-melting is stopped, the melt solidifies on the copper stage, into a solid “button” of alloy. The cooling rate during solidification of such a button is estimated to be in the range 10 K-s ^_1 to 100 K-s ^_1 .

Example 1. Formation of L1 ₀ Fe-Ni by casting without subsequent processing

Sample Preparation: Raw materials with overall atomic percentages of Fe ₅ oNi ₃ oPi C ₇ were selected. A melt of a master alloy was prepared directly from a mixture of elemental Fe (99.98 % purity), Ni (99.99 % purity), C (99.99 % purity) and the compound FeP (99 % purity) using an Edmund Buhler GmbH MAM-1 compact arc-melter. The melt was subsequently cast into rods 1 mm in diameter and at least 3 cm long in a water-cooled (286 K) copper mould attached to the arc-melter (as shown in Figure 1 a).

Figure 2a shows the Fe ₅ oNi ₃ oPi3C ₇ alloy cast as a rod of 1 mm diameter. XRD analysis confirms that the sample is fully crystallised, and microcrystals in the shape of dendrites are observed across the specimen, as shown in the SEM image of Figure 2b.

As deduced from transmission electron microscopy - energy-dispersive X-ray (TEM-EDX) elemental mapping (shown in Figure 2c), the dendrites are metallic Fe-Ni with very low additive (P, C) content. The composition of the dendrites is estimated to be 99.05 at.% Fe ₆ 4Ni ₃ 6 (at.% metals) + 0.95 at.% P. These EDX measurements are accurate to ±0.5 at.%.

The dendritic microcrystals are surrounded by a metallic matrix, containing a mixture of three phases. There are two metal-additive phases: iron-rich carbide (Fe 62±1 Ni 15±1 C 22±5 at.%) and iron-nickel phosphide (Fe 40±1 Ni 38±1 P 12±5 at.%). Additionally, a few particles in the polycrystalline matrix are metallic Fe-Ni with very low additive (P, C) content. Formally, these are considered to be the same phase as the dendrites.

This microstructure is typical for a “hypo-eutectic” alloy. The dendrites are the “primary phase” that forms first on cooling the liquid. There is solute partitioning between the dendrites and the remaining liquid. As the system is cooled and the dendrites grow, the liquid becomes more concentrated in solute elements partitioning out of the dendrites. Eventually, the liquid reaches a eutectic point (combination of temperature and composition) at which it freezes to a mixture of two or more phases, one of which is the same as the primary phase. In the present example, the remaining liquid freezes to a eutectic mixture of three phases.

Transmission electron microscopy shows the metal-additive matrix phases to be finely divided into complex shapes with shortest cross-sections on an order of 100 nm. Overall, the metallic matrix is polycrystalline as shown by selected-area electron diffraction (SAED). The pattern of the intermixture of phases in the matrix is characteristic of formation by eutectic freezing.

Figure 3 shows the XRD pattern recorded from a cross-section of the as-cast rod. All XRD peaks can be assigned to the mixture of three phases in agreement with the TEM study. The iron-rich carbide phase (Fe ₃ C structure) has the orthorhombic space group 62 Pnma with lattice parameters a = 0.5083 nm, b = 0.6733 nm and c = 0.4530 nm; the iron-nickel phosphide phase ((Feo _. sNio _. s^P structure) has the tetragonal space group 821-4 with lattice parameters a = 0.9005 nm and c = 0.4480 nm.

In Figure 2b, the straight rows of rounded light-contrast regions are cross-sections through the side-arms of a single dendrite spine. TEM imaging (bend contours) and SAED show that, as expected, all such regions in a row are in identical crystallographic orientation. Each dendrite is thus confirmed to be a monocrystal. As expected from the near-alignment of the spines of neighbouring dendrites in a region such as that in Figure 2b, the crystallographic axes of the dendrites show strong preferred orientation. The overall volume fraction of the dendritic phase is estimated to be 16±3 %.

Solid Fe-Ni alloys over a broad range of composition (including the Fe:Ni ratio in the primary phase in the present system) have the cubic close-packed crystal structure ccp (also known as face-centred cubic, fee, which is its conventional crystallographic lattice) with space group 225 Fm3m. In contrast, the XRD data for the present system (Figure 3) are best fitted by assuming that the Fe-Ni-rich dendrites have the tetragonal space group 123 P4/mmm, with lattice parameters a = 0.2547 nm and c = 0.3625 nm (ICSD 108455). The crystal structure is known as L1 ₀ (according to the Strukturbericht classification). The L1 ₀ structure is shown in Figure 2d; if the iron and nickel atoms were randomly assigned to the lattice sites (rather than being ordered, as shown) the structure would be ccp. As the ccp and L1 ₀ structures are so closely related, their X-ray diffraction patterns show essentially the same peaks, but some of the peaks that are single and of normal width in the ccp pattern show a distinct splitting in the L1 ₀ pattern. The inset in Figure 3 shows a close-up of a broad peak, interpreted to be the result of this splitting, and indicating the tetragonal L1 ₀ structure of the dendritic primary phase.

The identification of the L1 ₀ phase is confirmed by electron diffraction and high-resolution scanning TEM. In the diffraction pattern in Figure 4a, the spacing of the diffraction maxima along the crystallographic directions indicated by the solid and dashed lines would be identical in the ccp pattern, but in fact they are slightly different as shown by the scans in Figure 4b. Figure 4c is a high-resolution image with the same two crystallographic directions marked by solid and dashed lines. The profiles of the intensities along these two lines, in Figure 4d, clearly show different periodicities.

The L1 ₀ structure of Fe-Ni is illustrated in Figure 2d. The L1 ₀ structure is conventionally assigned a body-centred tetragonal (bet) lattice, but it is useful to consider it in terms of a face-centred (fct) lattice. The volume of the fct unit cell is twice that of the bet unit cell: a _ct = 2a _bct ; <¾ = Cb _ct - The key point is that the fct lattice has basis vectors parallel to those of the fee lattice of the ccp structure; as a result, the deviation of its da ratio from one directly describes the deviation from a cubic structure.

The average da ratio (fct) obtained from XRD is 1.007. This tetragonal distortion was confirmed by the analysis of the SAED patterns and high-resolution (S)TEM data in Figure 4.

These data show that the FeNi monocrystals in the cast samples have the tetragonal L1 ₀ structure otherwise known as tetrataenite.

A portion of the rod in Figure 2a was annealed under high vacuum for 15 minutes at 1123 K (i.e. just below the onset of melting). The X-ray diffractogram of the annealed sample (Figure 3) shows that the peaks assigned to the carbide and phosphide phases are essentially unchanged by the annealing. In contrast, the peaks assigned to the Fe-Ni phase are changed. Those peaks that show splitting in the diffractogram of the as-cast sample (e.g. the peak highlighted in the inset in Figure 3) are no longer split, indicating that the tetragonal distortion has disappeared. The XRD peaks from the annealed sample are consistent with cubic close-packed (ccp “gamma-phase”) Fe-Ni; this has an fee lattice with parameter a = 0.3593 nm.

EDX elemental mapping of a monocrystal-polycrystalline phase boundary of the kind shown at the perimeter of the rounded single-phase regions in Figure 2c, shows the composition of the monocrystalline region to be uniform. The composition of the sample changes sharply at the boundary of the monocrystal, with much higher concentrations of the P and C additives being contained in carbon-rich and phosphorus-rich phases in the polycrystalline matrix. This is direct evidence that the P and C additives that were part of the initial melt are rejected from the primary phase (ccp or L1 ₀ ) by solute partitioning during solidification.

The formation of the characteristic hypoeutectic microstructure, as seen in Figure 2b, suggests a normal solidification sequence, and specifically no special effects of rapid solidification. On cooling, the primary crystalline phase nucleates and grows. The solute partitioning between solid and liquid leads to composition gradients in the liquid immediately ahead of the advancing interface with the solid. These gradients lead to a well-known instability (driven by “constitutional supercooling”) that leads to dendritic growth shapes. As the sample is cooled, the primary-phase dendrites continue to grow until the eutectic point is reached and the remaining liquid freezes to a mixture of three phases.

The structure of the primary phase as studied at room temperature is L1 ₀ . The L1 ₀ structure can be formed by chemical ordering of Fe and Ni atoms on the atomic sites within the ccp structure. There are then two pathways to form the L1 ₀ structure by solidification: (i) the L1 ₀ structure could be formed immediately at the solid-liquid interface, facilitated by high atomic diffusivity in the liquid; or (ii) the phase forming at the solid-liquid interface could be ccp that later (i.e. at lower temperature) undergoes a disorder-to-order transformation within the crystalline state to L1 ₀ . The observations so far do not give any evidence supporting either pathway. In particular, the results of annealing, shown in Figure 3, do not lead to unambiguous conclusions on how the high-temperature annealing led to the appearance of the ccp phase. The annealing was carried out on a sample containing dendrites in a matrix of three phases. The possibility that the phase compositions change during the annealing cannot be ruled out. Indeed, in the inset in Figure 3, the split peak for the L1 ₀ structure not only merges into one for the ccp structure, but also shifts to higher diffraction angle. The most likely explanation of this shift, is an increase in the iron content of the dendrites during the annealing. Ideally, potential ordering or disordering transitions in Fe-Ni phase should be considered at fixed composition.

But whichever pathway is followed, it seems that the L1 ₀ phase can be formed by solidification from the melt, at conventional not ‘rapid’ cooling rates, and without any need for subsequent processing such as plastic deformation or annealing.

Example 1 demonstrates that a bulk solid material with uniformly dispersed L1 ₀ monocrystalline dendrites has been successfully synthesised directly by casting, without subsequent processing, of an Fe-Ni based alloy containing P and C as additives. The crystal structure and composition have been confirmed by XRD and systematic TEM studies. The crystallographic c/a ratio in the L1 ₀ phase is within the range of values reported in the literature for the L1 ₀ phase, including tetrataenite as found in meteorites. Together with the high Fe content (64 at.%), this makes this L1 ₀ phase a highly promising candidate for a rare- earth-free permanent-magnet material.

Analysis Methods - Example 1 Specimen preparation. As-cast rods were sliced longitudinally, i.e. parallel to their axis, mechanically polished from both sides and placed on SEM stubs for imaging. For TEM examination, lamellae were prepared by FIB milling.

X-ray diffraction. XRD analysis of the sliced rods was performed with CoKa radiation using a Bruker D2 PFIASER diffractometer combined with an energy-dispersive LYNXEYE XE detector ( A = 1.79026 A).

Scanning electron microscopy. SEM study was performed using a Helios Nanolab SEM/FIB. The microstructure of the samples was studied by Ga ⁺ ion-beam scanning (i.e. distinct from conventional secondary- or backscattered-electron scanning) to acquire “channelling contrast” which allows to map grains with different crystallographic orientation. When the incident beam becomes parallel to a set of crystallographic planes inside a grain, the intensity of this grain in the image decreases. This is the origin of the channelling contrast, in which each image shows a qualitative crystallographic orientation map of the region of interest.

Transmission electron microscopy. For TEM measurements, the rods, sliced longitudinally, were ground and polished with diamond paste to a thickness of 60 pm. Subsequently, the samples were exposed to double-sided ion milling (Gatan model 691) using argon ions with a primary beam energy of 4 keV at a rotation speed of 2.5 rpm. For the first 20 minutes, the gun angles were set to 6°, followed by 4° for 70 minutes. Additionally, a cross-sectional TEM lamella was prepared by focused ion-beam (FIB) milling using a Helios Nanolab FIB/SEM. The samples were polished with a series of diamond abrasive films with decreasing grit sizes down to 0.1 pm. A standard secondary-electron detector was used to image the Fe-Ni dendrites. The image acquisition and spectroscopic analysis were conducted using a FEI Tecnai Osiris TEM/STEM with field-emission gun operated at 200 keV, equipped with Super-X windowless EDX detector. High-resolution (S)TEM was carried out on a (S)TEM Titan G2 60-300 microscope (FEI, Netherlands) equipped with an image and probe aberration corrector. The microscope was operated at 300 kV acceleration voltage, giving a spatial resolution below 1 A.

Example 2. Magnetic properties of L1n Fe-Ni phase made bv casting To evaluate the magnetic properties of the L1 ₀ phase, the magnetic domain structure within monocrystals of the phase was investigated by virtual bright field differential phase contrast scanning transmission electron microscopy (VBF DPC STEM).

A button-shaped sample, roughly 2 cm in diameter, of FessNhsPe _. sCs _. s (at.%) was cast on the stage of the arc-melter (as shown in Figure 1b). A longitudinal slice was cut parallel to the rotation axis of the button. For TEM, a lamella was prepared by focused ion-beam (FIB) milling.

Figure 5a shows a TEM bright-field image from the lamella. The primary dendritic phase in this sample has the L1 ₀ structure, with a composition close to Fe ₆ 2Ni ₃ 7Pi (at.%). In VBF-DPC imaging, the light-dark contrast indicates the strength and sign of magnetic-field components, so the direction of the magnetization in each domain can be determined. For this lamella, Figure 5b and Figure 5c show two orthogonal components £? _x and By of the in-plane magnetic field B at remanence, measured by VBF-DPC. This is the remanent state under zero applied external field. The x and y axes are shown in Figure 5a. It is clear that the magnetic state is a single domain. The orientation of the magnetization vector is indicated by the arrow in Figure 5c.

A similar lamella was prepared from an as-cast 3-mm-diameter rod of Fe ₆ oNi ₄ o (i.e. with an Fe:Ni ratio designed to match that of the L1 ₀ phase just discussed). This rod formed single phase ccp Fe-Ni with the same composition. As discussed in relation to Example 3, this sample did not contain any additive, and did not form the L1 ₀ phase. Figure 5d and Figure 5e show the components Ek and B of the in-plane magnetic field at remanence for this ccp sample. In contrast with the L1 ₀ sample (Figure 5b and Figure 5c), the magnetic structure is multi-domain with alternating magnetization vectors in adjacent domains. In this case, the overall remanent magnetization is close to zero. Such a domain pattern is typical for thin samples with rectangular shape and negligible monocrystalline magnetic anisotropy. In these patterns (known as the Landau-Lifshitz structure), the triangular flux-closure domains minimize the magnetostatic energy of the sample.

Comparison of the domain structures in the L1 ₀ and ccp phases suggests that in L1 ₀ , the magnetic anisotropy is high enough to overcome the magnetostatic energy due to the sample geometry. As given by SAED, the c-axis of the L1 ₀ Fe-Ni crystal in Figures 5a-c is oriented parallel to the magnetization vector shown in Figure 5c but tilted by ~9° out of plane. Thus, the observed single-domain state is directly related to the uniaxial magnetocrystalline anisotropy of the tetragonal L1 ₀ Fe-Ni. This state gives a high remanent magnetization, as is needed for permanent-magnet applications.

Example 2 demonstrates that for a given Fe-Ni ratio, the L1 ₀ phase made by the methods in this invention has magnetic properties very different from the closely related ccp phase. The chemical ordering in the L1 ₀ phase leads to a tetragonal distortion of the crystalline lattice and to uniaxial magnetocrystalline anisotropy. This anisotropy is consistent with literature reports of the highly desirable magnetic properties of tetrataenite. The anisotropy leads to single-domain states in the crystalline grains. For a monocrystalline sample, or with preferred orientation of the c-axis in a polycrystalline sample, the remanent magnetization is high, near to magnetic saturation (all moments aligned). This demonstrates that samples prepared using the method of the present invention are highly applicable to permanent-magnet applications.

Analysis Methods - Example 2

Magnetic measurements in TEM. The magnetization of the samples was first saturated by application of an external magnetic field of 2 T. The samples were characterized in the remanent state after removal of this field. In addition to those methods used for Example 1 , virtual bright field differential phase contrast scanning transmission electron microscopy (VBF DPC STEM) was used to characterize magnetic domain structures.

Cross-sectional specimens with a final thickness of about 70 nm were prepared by FIB milling. Low-magnification (LM-STEM) images for the VBF-DPC were acquired on a Titan 60-300 electron microscope equipped with a high-brightness electron gun (x-FEG), and high- angle annular dark-field (HAADF) STEM detector. The microscope was operated at 300 keV acceleration voltage. The LM-STEM was intentionally tuned for the objective lens (OL) = 0 % condition to attain a zero-magnetic-field environment. For highest sensitivity of the VBF-DPC method, in the case of the semi-convergence angle a being bigger than or equal to the collection angle b, geometrical considerations suggest to apply an overlap y between the zero-field diffraction disc and the VBF detector such that y = b (i.e. the edge of the diffraction goes through the centre of the VBF detector). The convergence semi-angle of the STEM probe was measured to be a = 0.6 mrad. The VBF detector was used with a selected-area (SA) aperture of 40 pm diameter, resulting in a collection semi-angle b = 0.2 mrad. In this scheme, the VBF detector has the area of the standard HAADF detector limited by the SA aperture. As for the standard DPC technique, the VBF-DPC method was used to acquire two orthogonal in-plane components of the magnetic field in the sample, and thus to visualise its magnetic domain structure. Example 3. Different compositions and sizes of castings

In addition to the Fe ₅ oNi ₃ oPi3C7 and FessNhsPe.sCs.s (at.%) compositions described in relation to Example 1 and Example 2, the inventors have studied several other compositions, all made by arc-melting as already described. These compositions have been prepared from melts having a defined “overall composition”, and cast either as rods of different diameters, or as buttons (with diameters of 2 cm), as depicted in Figure 1 a and Figure 1 b. All of the solid samples were found to contain a primary dendritic phase. The volume fraction of the dendritic phase in the overall solid sample, the composition of the primary dendritic phase, and the tetragonality of the primary dendritic phase have been characterised.

Details of the samples that have been made are set out in Table 1 :

Table 1

^# lt is generally accepted that for casting as shown in Figure 1 a, the cooling rate should be inversely proportional to the square of the rod diameter [J.F. Loftier, A.A. KOndig, F.H. Dalla Torre, Rapid solidification and bulk metallic glasses — processing and properties, In: J.R. Groza, J.F. Shackelford, E.J. Lavernia, M.T. Powers (editors), Materials Processing Handbook (CRC Press 2007)]. Cooling rates in the solidification range have been measured for similar casting for rods with diameters of 3, 5 and 10 mm [D.V. Louzguine-Luzgin, G. Xie, Q. Zhang, C. Suryanarayana, A. Inoue, Formation, structure, and crystallization behaviour of Cu-based bulk glass-forming alloys, Metallurgical and Materials Transactions A 41 A (2010) 1664-1669]. The values given are as measured for the 3-mm- diameter rods, and are scaled (with d ^{~ 2} ) for the other diameters d.

^† These values are estimated from the area fraction of the primary dendritic phase seen in SEM micrographs (e.g. Figure 2b and Figures 6-8). The area fraction appears to vary somewhat across the rod samples in particular. These values do not include the fraction of the same phase included within the eutectic matrix. The uncertainty in these values of volume fraction is estimated as ±3 %.

^* These are samples cast on the stage of the arc-melter (the melt is not ejected into a mould). Under the action of surface tension and gravity, the molten alloy forms a button shape that is roughly hemispherical (Figure 1 b). The diameter of the button is roughly 2 cm.

*ln these samples, the tetragonality has not been quantified. The evidence nevertheless suggests that the structure is tetragonal, since all of these samples show similar SAED patterns on the (lOO)-type zone axis.

^Difficult to quantify, but very close to 100 %. Figure 6a, Figure 6b and Figure 6c show the microstructures of as-cast Fe ₅ oNi ₃ oPi3C ₇ in rods of 1 mm, 2 mm and 3 mm diameter respectively, for which the estimated cooling rates are (10-20) x 10 ³ , (2-6) x10 ³ , and (1-3)x10 ³ K-s ^_1 . The rods of larger diameter experience a lower cooling rate, and the microstructures are coarser: the dendrite spines are of larger diameter and the secondary side-arm spacing is greater. These effects are explained by slower cooling leading to the dendrites growing more slowly and with blunter tips. The microstructures in the larger-diameter rods also show a greater volume fraction of the primary dendritic phase. This effect is explained by slower cooling permitting greater equilibration of the compositions of the primary solid phase and the remaining liquid before the onset of eutectic solidification of the liquid.

Figure 6d, Figure 6e and Figure 6f show the microstructures of as-cast Fe53Ni32P9.75C5.25 in rods of 1 mm, 2 mm and 3 mm diameter respectively. These show the same effects of lower cooling rate on larger diameter rods (coarser dendrites of the primary dendritic phase, occupying a higher volume fraction) as in Figures 6a-c. In addition the lower additive (P+C) content (15 at.% compared to 20 at.% in the former case) leads, for each rod diameter, to a greater volume fraction of the primary dendritic phase. This shows that the volume fraction of Fe-Ni primary phase must tend to 100 % as the additive content of the system tends to zero.

The trends seen in Figure 6 are extended in Figure 7. The overall sample compositions have still lower (P+C) content, 10 at.% in Figure 7a, and 8 at.% in Figure 7b. Melts of these compositions were not ejected into cylindrical moulds, but left to freeze as button-shaped samples in the stage of the arc-melter. These buttons have diameters of roughly 2 cm, and so are cooled significantly more slowly (R ~ 10-100 K-s ^_1 ) than a 3-mm-diameter rod. The resulting primary dendritic phase is coarser and occupies a larger volume fraction of the overall solid, as is particularly clear in Figure 7b.

These trends are seen further in Figure 8. For the overall composition Fe _58.5 Ni _38.5 P ₃ (at.%) cast as a button, the dendritic growth mode of the primary phase can still be detected, but this phase occupies nearly 100 % of the sample volume (Figure 8a). For the overall composition Fe _59.5 Ni _39.5 P ₁ (at.%) cast as a button, the microstructure (Figure 8b) shows only equiaxed grains of a single phase. With such a low additive content, it appears that the liquid can solidify almost completely to a single-phase Fe-Ni solid solution; X-ray diffraction reveals only a trace presence of the phosphide phase. The key characteristics of the microstructures as a function of composition and cooling rate are summarized in Table 1 . The composition Fe ₆ oNi ₄ o with no additive content solidifies to a ccp phase, and not to the L1 ₀ phase, which indicates that some additive content is key to the formation of the L1 ₀ phase. In all the other cases listed in Table 1 , there is strong evidence for crystallographic tetragonality indicating that the primary phase has the L1 ₀ and not the ccp structure.

The data in Table 1 allow for comparison of the Fe:Ni ratio in the overall composition (i.e. in the liquid melt being cast) and in the primary-phase dendrites. In all cases, the Fe:Ni ratio is higher in the dendrites than in the overall alloy. This direction of partitioning is a typical result of solute partitioning during dendrite growth. The consistent behaviour across the ranges of both composition and cooling rate suggests that there is a wide range of conventional solidification processing conditions under which the L1 ₀ phase is obtained. The production of the L1 o phase is not reliant on precisely controlled composition, nor on extreme (e.g. rapid) or precisely controlled cooling.

The desired L1 ₀ primary phase is formed directly by casting, even with extremely low additive contents in the melt. For example, a melt containing as little as 1 at.% P has been shown to directly form the desired L1 ₀ primary phase. It appears that the presence of carbon is not necessary for the formation of the L1 ₀ phase. The formation of essentially single-phase material with overall composition Fe59 _. 5Ni39 _. 5P1 suggests that up to an order of 1 at.% P can be incorporated in the L1 ₀ phase. This is corroborated by EDX measurements of phosphorus content in the primary-phase dendrites. The inventors consider that these small phosphorus contents are critical in favouring formation of the L1 ₀ phase.

The desired L1 ₀ primary phase is formed directly by casting even for the cooling rates in copper-mould casting, associated with near-equilibrium partitioning of solutes during solidification.

With the range of possible compositions and cooling rates, as-cast samples can contain coarse dispersions of the L1 ₀ phase occupying high volume fractions. This is extremely different from anything previously documented in the prior art. Example 4. Experimental evidence for the order-disorder temperature ( 7OD)

In order to understand how the L1 ₀ phase could form directly on solidification from a melt, the inventors have considered the order-disorder temperature ( 7 ₀ D) of the Fe-Ni materials being studied.

The lamella studied in Figure 5a was subjected to in-situ heating in a transmission electron microscope (TEM). This enables real-time detection of the tetragonality (c/a). The degree of deviation from one (i.e. from cubic symmetry) indicates the degree of ordering. The isolation from other regions is important, as neighbouring regions of different composition might interdiffuse with the L1 ₀ phase, changing its composition. In that case, any ordering or disordering (occurring on heating) might be an effect of changing composition, and not a true order-disorder transition.

Figure 9a shows the temperature dependence of the long-range order parameter h for a Cu ₃ Au alloy in equilibrium. At high temperature, the Cu and Au atoms randomly occupy sites in the cubic close-packed (ccp) structure {h = 0). At low temperature, the site occupancy is ordered, with Cu atoms occupying the face-centre sites and Au atoms occupying the corner sites in the unit cell {h = 1 ). The ordering to form the Cu ₃ Au (L1 ₂ ) structure from ccp is similar to that for the formation of L1 ₀ from ccp, but is for the case when the two species are present in a ratio of 3:1 (rather than 1 :1 for L1 ₀ ). In Figure 9a, the solid line shows a computed order parameter h, and the dots are measured data.

On heating the fully ordered Cu ₃ Au structure, ^ decreases somewhat (this means that some Cu atoms occupy Au sites and vice versa), and then there is a sharp a “first-order” transition in which there is a discontinuous sharp drop to near-total disorder (h = 0). This drop indicates the order-disorder temperature T ₀ D. The inventors expect that the temperature dependence of the equilibrium long-range order parameter for the L1 ₀ -to-ccp transition would have the same general form as that shown in Figure 9a.

Figure 9b shows that if the starting phase were only partially ordered (here 77 ^» 0.45), then on heating (arrow 1 ) at first nothing happens because there is insufficient atomic mobility for the structure to change. On further heating, the mobility increases exponentially and the h rapidly reaches (arrow 2) its equilibrium value and then follows the equilibrium curve (arrows 3 to 5). Stoichiometric CuAu shows an order-disorder transition from L1 ₀ to ccp exactly analogous to that for FeNi. Figure 10 shows experimental results for heating a sample of CuAu from a starting condition of complete disorder (77 = 0); this profile is broadly similar to that schematically shown in Figure 9b: ordering is followed by progressive and then sharp disordering.

Figure 11 shows data (closed circles for heating, open circles for cooling) obtained for a primary phase extracted from a FessNhsPesCss (at.%) button that was cast on the stage of the arc-melter as shown in Figure 1b. The primary phase itself has the composition Fe62Ni37.05P0.95. The data in Figure 11 was obtained by heating and cooling an as-cast L1 ₀ monocrystal made using the method of the present invention. The heating and cooling are conducted in-situ in the TEM, permitting measurement of the tetragonality at different temperatures. The degree of tetragonality (deviation of c/a from one) is proportional to the long-range order parameter 77 conventionally used to describe the transition from ccp (77 = 0) to fully ordered, stoichiometric (i.e. Fe ₅ oNi ₅ o) L1 ₀ (77 = 1). The value of c/a that would correspond to complete ordering (77 = 1 ) is unknown.

On heating, the tetragonality shows ordering followed by sharp disordering, with some similarity to the schematic example in Figure 9b and to the data in Figure 10. The maximum tetragonality is -1.01 1 , and this can be regarded as a lower-bound estimate of the value representing full ordering (77 = 1). The initial tetragonality (1.005) matches the values found in meteorites. The heating rate is rather low -0.2 K-s ^_1 . The cooling rate is -0.1 K-s ^_1 . In contrast, the cooling rate in casting is of the order of 10 K-s ^_1 to 100 K-s ^_1 .

The data in Figure 11 suggest that:

(i) the as-cast L1 ₀ phase that was analysed is not fully ordered.

(ii) the partly ordered L1 ₀ can become more ordered on heating.

(iii) the increase in tetragonality is unambiguous evidence that the temperature in that regime (350 °C to 550 °C) is less than the actual T ₀ D.

(iv) the collapse in tetragonality between 550 °C and 600 °C (823 K to 873 K) shows the probable region of T ₀ D. For Fe ₅ oNi ₅ o and closely related compositions, there is no generally accepted value of T ₀ D. In the earliest work, T ₀ D was taken to be around 320 °C [L. Neel, J. Pauleve, R. Pauthenet, J. Laugier, D. Dautreppe, Magnetic properties of an iron-nickel single crystal ordered by neutron bombardment. Journal of Applied Physics 35 (1964) 873-876] and later work suggested a value between 400 °C and 450 °C [S. Goto, H. Kura, E. Watanabe, Y. Hayashi, H. Yanagihara, Y. Shimada, M. Mizuguchi, K Takanashi, E. Kita, Synthesis of single-phase /.10-FeNi magnet powder by nitrogen insertion and topotactic extraction. Scientific Reports 7 (2017) 13216]. The present inventors consider that the higher value of 550° to 600 °C in their work is the effect of additive addition.

(v) there is enough atomic mobility around T _O D for the ordering and disordering to occur in real time, with rather low heating rate.

(vi) if disordering can occur on heating, the reverse process of ordering should occur on cooling: this is seen for the CuAu example in Figure 10. In the present case, even though the cooling is slower than the heating, order does not re-develop on cooling. It is not clear why this is so. Flowever, just the same phenomenon is found for tetrataenite in meteorites: there is disordering on heating, but no restoration of ordering on cooling. The reason may be that on heating and after the disordering, a different type of ordering occurs, or that there is chemical interdiffusion with neighbouring regions. The continuing changes seen above 600 °C in the present work may indicate such effects.

Comparison of the SAED patterns from the sample before and after the in-situ heating provides particularly clear confirmation of L1 ₀ -type ordering in the as-cast sample. Figure 12 compares the diffraction patterns on the same [001] zone axis. In each case, the main diffraction spots form a square array. Indexing the reflections according to the fct and fee lattices (i.e. with the same crystallographic axes for each pattern), the first main spot to the right of the origin can be indexed as 200, and the first main spot above the origin can be indexed as 020. Taking the vectors from the origin to these two spots and combining them by integer addition, the entire grid of main spots can be constructed. Figure 12b shows that for the ccp structure with an fee lattice, these are the only observed reflections. In Figure 12a, in contrast, there are additional very weak reflections, one of which is circled. With the main spots indexed as above, the circled spot is 110. This reflection is “systematically absent” for the fee lattice, but is permitted for the fct lattice. That the 110 spot (with its symmetry- equivalents) is visible, is unambiguous evidence that the iron and nickel atoms are not randomly assigned to the structural sites, but show ordering of the kind shown in the inset in Figure 2b. Conclusions of TO investigation

The conclusions that can be drawn from this investigation of the order-disorder temperature are as follows:

1. For this particular composition of L1 ₀ Fe-Ni (Fe ₆ 2Ni ₃ 7Pi), T _O D is in the range 550 °C to 600 °C (823 K to 873 K).

2. With this T _O D value, and the normal cooling rates for casting, it may be that only partial ordering is achieved. This is consistent with the rates of ordering seen upon heating.

3. T _O D for this composition is high enough to permit a detectable extent of ordering to L1 ₀ to develop during normal casting.

4. With this T ₀ D, it should be possible to increase the degree of ordering to L1 ₀ by annealing (e.g. at 550 °C). A higher degree of ordering favours a greater magnetic anisotropy (which is desired).

5. That opens up the possibility that if casting did not give L1 ₀ with a degree of ordering as high as would be desired, the required ordering might be developed on subsequent annealing.

6. The development of magnetically aligned (i.e. polarized) material is likely to be useful. The ability to develop ordering on annealing is therefore highly attractive, as that annealing could be conducted under an externally imposed magnetic field.

Analysis Methods - Example 4 ln-situ heating in TEM. In-situ imaging was conducted using an in-situ heating holder (DENSsolutions) under the usual TEM vacuum (<10 ^-6 Pa). The resistance of the platinum coil in the chip is monitored in a four-point configuration, and the temperature is calculated using constants provided by the manufacturer, with some correction at higher temperature [Yu.P. Ivanov, C.M. Meylan, N.T. Panagiotopoulos, K. Georgarakis, A.L. Greer, In-situ TEM study of the crystallization sequence in a gold-based metallic glass, Acta Materialia 196 (2020) 52-60]. On the heating holder, the sample temperature can be controlled with temperature accuracy >95 %, temperature homogeneity >99.5 % and thermal stability 0.005 K. Due to the absence of thermal drift, data acquisition could be started immediately after each temperature increase.

Figure Captions

Figure 1. Casting of samples after arc-melting, (a) Arc-melting on the copper stage is followed by suction casting into the cylindrical cavity that (for different moulds) has a diameter of 1 , 2 or 3 mm. The mould is cooled with water at 286 K. (b) The arc-melted alloy forms a roughly hemispherical “button” 2 cm in diameter, and this solidifies on the copper stage. Figure 2. As-cast rod of Fe ₅ oNi oPi ₃ C ₇ (at.%) alloy, (a) an optical photograph of a 1-mm- diameter rod formed using the method of the present invention; (b) is a scanning electron microscopy (SEM) image of a polished lateral cross-section of the rod in (a), showing dendrites of Fe-Ni monocrystals formed during solidification; the inset shows the body- centred tetragonal (bet) L1 ₀ structure of Fe-Ni; (c) a high-angle annular dark-field (FIAADF) STEM image and EDX elemental maps around one dendrite side-arm in the rod in (a). Figure 3. X-ray diffraction analysis of as-cast and annealed Fe ₅ oNi3oPi3C ₇ (at.%) alloy. The measured traces (labelled XRD) are shown for the as-cast sample, and after annealing at 1123 K for 15 min; the radiation is CoKoc, and the diffractograms are acquired from a lateral cross-section of the rod in Figure 2a. The expected peaks for the phases ccp Fe-Ni, L1 ₀ Fe- Ni, (Fe ₅ oNi ₅ o)3P and Fe C are shown at the bottom. The patterns for (Fe ₅ oNi ₅ o)3P and Fe ₃ C are from tabulated data for the stoichiometric phases. The patterns for the ccp and L1 ₀ phases are obtained by adjusting the lattice parameters to match the measured diffractograms. The inset shows a close-up of a Bragg peak that is split in the as-cast sample and is single after annealing. The splitting indicates the tetragonality associated with L1 ₀ - type ordering of iron and nickel atoms on sites that without this ordering would constitute a cubic-close-packed (ccp) structure.

Figure 4. Transmission electron microscopy of the Fe-Ni phase within as-cast

Fe ₅ oNi ₃ oPi ₃ C ₇ . This is from the rod shown in Fig. 2a. (a) Selected-area electron diffraction (SAED) on the [110] zone axis; (b) intensity profiles along the solid and dashed lines in (a) (c) High-resolution STEM image on the same [110] zone axis; (d) intensity profiles along the solid and dashed lines in (c).

Figure 5. Differential phase contrast imaging of magnetic domain structures, (a) A bright-field TEM image of a lamella of the Fe-Ni primary phase extracted from a Fe ₅ 5Ni ₃₅ P6 _. 5C ₃ .5 (at.%) button that was cast on the stage of the arc-melter. The primary phase has the L1 ₀ structure with a composition of Fe ₆ 2Ni _37. o5Po.95. (b— d) Differential phase contrast (DPC) imaging, mapping the magnitude of two in-plane orthogonal components of magnetization B* and B at remanence: (b,c) for the sample in (a); (d,e) for a sample of ccp FeeoNUo. In (b,c), the components B* and B are referred to the axes in (a), and the magnetization vector, projected into the plane of the micrograph, is shown by the arrow in (c); in this case the magnetic structure is a single domain with a high remanent magnetization, parallel to the c-axis of the L1 ₀ structure. In (d,e), the magnetic structure is multi-domain with a remanent magnetization close to zero.

Figure 6. Microstructures of as-cast Fe ₅ oNi oPi ₃ C ₇ and Fe ₅₃ Ni ₃₂ P ₉ .7 ₅ C ₅ .2 ₅ rods. Scanning electron microscopy (SEM) images of polished longitudinal cross-sections of rods cast with different diameters (cooling rates). For both (a) Fe ₅ oNi ₃ oPi3C ₇ and (b) Fe53Ni32P9.75C5.25 compositions (at.%), the rods of larger diameter (cooled more slowly) show a higher volume fraction of coarser primary-phase monocrystalline dendrites of Fe-Ni. This volume fraction is also higher when, in (d-f) compared with (a-c), the overall P, C content of the original melt is lower.

Figure 7. Microstructures of as-cast FessNUsPe.sCs.s and Fe ₅₆ Ni ₆ P5.2C2.8 buttons.

Scanning electron microscopy (SEM) images of polished longitudinal sections of button shaped samples cast on the stage of the arc-melter. With these low P, C contents in the original melt, (a) 10 at.% in Fe ₅₅ Ni ₃₅ P _6.5 C _3.5 , (b) 8 at.% in FessNhsPs^.s, there are high volume fractions of primary-phase monocrystalline dendrites of Fe-Ni.

Figure 8. Microstructures of as-cast Fe _58.5 Ni _38.5 P ₃ and Fe _59.5 Ni _39.5 P ₁ buttons. Scanning electron microscopy (SEM) images of polished longitudinal sections of button-shaped samples cast on the stage of the arc-melter. With these lower P, C contents in the original melt, (a) in Fe _58.5 Ni _38.5 P ₃ , (b) in Fe _59.5 Ni _39.5 P ₁ , the volume fractions of primary-phase monocrystalline dendrites of Fe-Ni are even higher than in Figure 7.

Figure 9. The order-disorder transition in Cu ₃ Au. (a) This plot shows the temperature dependence of the equilibrium long-range order parameter h in Cu ₃ Au as calculated from atomistic simulations (solid line) and as measured (data points). The l_1 ₂ -ccp order-disorder transition temperature is 635 K (362°C). (b) The dashed curve and arrows superposed on the plot show the expected changes in h upon heating a sample of CU ₃ AU that is initially partially ordered ( h ^« 0.45). The plot in (a) is adapted and simplified from [F. Cleri, G. Mazzone, Vittorio Rosato, Order-disorder transition in Cu ₃ Au: A combined molecular- dynamics and cluster-variation-method approach. Physical Review B 47 (1993) 14,541 — 14,544]

Figure 10. The ccp-to-L1 ₀ ordering upon heating disordered CuAu. The sample is subjected to “isochronal heating”, i.e. holding successively for 20 min at each temperature (data points). The dashed line shows the behaviour upon subsequent isochronal cooling. The long-range order parameter is calculated from the electrical resistivity of the samples. Adapted and simplified from [M Spanl, W. Puschl, B. Sprusil, J. Sachl, V. Sima, W. Pfeiler, Change of microhardness in stoichiometric CuAu. Materials Transactions 43 (2002) 560- 565]

Figure 11. The order-disorder transition in L1 ₀ Fe-Ni. (a) Tetragonality c/a of the Fe-Ni primary phase extracted from a FessNhsPesCss (at.%) button that was cast on the stage of the arc-melter. The primary phase itself has the composition Fe ₆₂ Ni _37.05 P _0.95 . Lattice- parameter measurements were made in-situ in the TEM during heating (closed circles) and during subsequent cooling (open circles). The long-range order parameter is proportional to the amount that c/a deviates from one. Similarly to the behaviour shown schematically in Figure 9b and by the measurements in Figure 10, the degree of order first increases on heating and then falls sharply at the order-disorder transition temperature 7b D, which appears to be in the range 550°C to 600°C (823 K to 873 K). In contrast to the reversibility seen in Figure 10, ordering, i.e. (c/a) > 1 , does not return upon cooling (b) The temperature-time profile for the in-situ heating and cooling.

Figure 12. SAED revealing the L1o-ccp order-disorder transition in the Fe-Ni phase.

These SAED patterns on the [001] zone axis were obtained at room temperature from the lamella of Fe-Ni (extracted from a button of FessN sPe _. sCs _. s) used to obtain the results in Figure 5 and Figure 11. (a) is in the as-cast condition, and indicates the L1 ₀ phase; (b) is after the heating profile in Figure 11b, and indicates the ccp phase. The main reflections are the same in each case. In (a), four weak reflections, one of them circled for emphasis, can be seen in a square array around the central spot. These additional reflections can arise only when the iron and nickel atoms are distributed non-randomly on the sites of the ccp structure, thus forming the L1 ₀ structure as shown in the inset in Fig. 2b. Thus, for the same basic structure, (a) represents chemical order and (b) represents chemical disorder.

The work leading to this invention has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement No 340025.

The project leading to this application has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No 695487).