Primary Carbide Formation in Tool Steels: Potential of Selected Laboratory Methods and Potential of Partial Premelting for the Generation of Thermodynamic Data

To predict the solidification and product properties of tool steels with complex chemical compositions, an understanding of the transformation behavior is crucial. Therefore, the quaternary Fe–C system with 10 wt% Cr and 3 wt% W (a subsystem of cold work steels, with M7C3 and M23C6 carbides) and the Fe–C system with 6 wt% W and 5 wt% Mo (simplified high‐speed steel, with M6C and MC carbides) are selected. The motivation for this study is to develop a methodology for the safe and fast production of model alloys and the close to equilibrium performance of differential scanning calorimetry (DSC) measurements. Regular diffusion annealing of as‐cast carbidic steels is time‐consuming, but with an additional heat treatment during the DSC measurement in the semisolid zone (30–50% liquid phase fraction), a status close to equilibrium can be achieved within minutes due to the high diffusion. To prove the potential of the equilibration by partial premelting in the DSC, additional equilibration and quenching experiments are performed in a Tammann furnace and investigated using a scanning electron microscope and X‐ray diffraction analysis. By combining these methods, carbide types and the transformation temperatures can be verified to evaluate and construct complete phase diagrams.


Introduction and Motivation
To predict the solidification, perform solidification calculations, and control the product properties of complex alloyed tool steels, a deep understanding of the transformation behavior during and after solidification, depending on the chemical composition, is a key competence. The knowledge of primary carbide formation plays an essential role in higher alloy tool steels, where many different carbide types (e.g., M 7 C 3 , M 23 C 6 , M 6 C, and MC) can form. Table 1 gives an overview of typical carbide types and their properties. [1][2][3][4][5][6] The metastable M 2 C carbide only forms during fast cooling, and after annealing treatment at temperatures between 900 and 1500°C, decomposes into the stable forms MC and M 6 C. [7] Primary and eutectic carbides form directly from the melt and have an average size of 1-100 μm. Secondary carbides are formed during solid-phase transformations and have an average diameter of 0.1-1 μm. Especially with higher alloyed carbidic steels, thermodynamic calculations often show significant differences, visualized, e.g., in Figure 1. To make precise thermodynamic statements about new alloys, reliable investigation methods are essential. This work provides the results of comprehensive method development for a basic understanding of the sample preparation, characterization, and generation of phase diagrams with carbides from the high-alloy Fe-C-Cr-W-Mo system.
As multicomponent systems with more than five alloying elements include a wide variety of complex phases, two simplified subsystems were selected: the quaternary Fe-C model system with 10 wt% Cr and 3 wt% W (simplified cold work steel %X210CrWMoV1-2 without molybdenum and vanadium, including M 7 C 3 and M 23 C 6 carbides) and the Fe-C system with 6 wt% W and 5 wt% Mo (simplified high-speed steel %HS6-5-2 without chromium and vanadium, including M 6 C and MC carbides). In addition, all other alloying elements (e.g., Si, Mn, Al) and all trace and microalloying elements (S, P, Ti, Nb) were deliberately not taken into account. [8] Figure 1 shows the phase diagrams of the Fe-6W-5Mo-C var (a,b) and Fe-10Cr-3W-C var (c,d) systems, calculated with the thermodynamic databases TCFE10.1 (ThermoCalc) and FSstel2020 (FactSage). According to the database description, both phase diagrams calculated with TCFE10.1 DOI: 10.1002/srin.202200503 To predict the solidification and product properties of tool steels with complex chemical compositions, an understanding of the transformation behavior is crucial. Therefore, the quaternary Fe-C system with 10 wt% Cr and 3 wt% W (a subsystem of cold work steels, with M 7 C 3 and M 23 C 6 carbides) and the Fe-C system with 6 wt% W and 5 wt% Mo (simplified high-speed steel, with M 6 C and MC carbides) are selected. The motivation for this study is to develop a methodology for the safe and fast production of model alloys and the close to equilibrium performance of differential scanning calorimetry (DSC) measurements. Regular diffusion annealing of as-cast carbidic steels is time-consuming, but with an additional heat treatment during the DSC measurement in the semisolid zone (30-50% liquid phase fraction), a status close to equilibrium can be achieved within minutes due to the high diffusion. To prove the potential of the equilibration by partial premelting in the DSC, additional equilibration and quenching experiments are performed in a Tammann furnace and investigated using a scanning electron microscope and X-ray diffraction analysis. By combining these methods, carbide types and the transformation temperatures can be verified to evaluate and construct complete phase diagrams.
and FSstel2020 refer to the same literature sources (Gustafson [9,10] ) and show the same carbide types but different phase transformation temperatures due to individual and unpublished database developments.
The most significant difference in the Fe-6W-5Mo-C system between the databases is the calculation of the solidus line (L þ γ! L þ γ þ M 6 C) between 0.5 and 1.5 wt% carbon, as visualized in Figure 1a,b. The database FSstel2020 calculates the solidus temperature approx. 50°C lower than TCFE10.1.
For the Fe-10Cr-3W-C var system, visualized in Figure 1c,d, the peritectic region-where liquid, FCC, and BCC coexist-is very similar. After that, the phase diagram shows significant differences in the stable phase areas. FSStel2020 calculates the three-phase region consisting of M 7 C 3 , M 23 C 6 , and FCC at around 100°C lower than TCFE 10.1.
The underlying raw data [11][12][13][14][15][16] of the thermodynamic descriptions by Gustafson [9,10] were published between 1950 and 1977 and did not provide directly comparable alloys to those considered in the phase diagram in Figure 1. As trustworthy investigation methods are vital to generate reliable thermodynamic data, a literature review of the most common methods was carried out to achieve an overview of the current state of research. Table 2 summarizes the results of a comprehensive literature review of the Fe-C-Cr-W-Mo system. The compilation clearly shows that the differential thermal analysis (DTA)/differential scanning calorimetry (DSC) methods (both methods based on measuring the change in enthalpy [17] ) were used to determine the phase transition temperatures. Only referring to thermodynamic calculations without experimental verification has to be considered critically with respect to still uncertain phase diagrams. The analysis of the carbide types was almost always performed by means of metallographic examinations using a scanning electron microscope with back-scattered electrons (SEM-BSE). In order to confirm the respective carbide phases, the X-ray diffraction (XRD), electron backscatter diffraction (EBSD), or transmission electron microscopy (TEM) methods were used. The methodology, determination of phase transformations by DTA/DSC, identification of carbide types by SEM, and confirmation by XRD, EBSD, or TEM are based on well-known methods. [18] These are all excellent methods for investigating the present sample condition.

Literature Review
For the creation of phase diagrams, however, the samples must be in a state of thermodynamic equilibrium to provide reproducible and reliable measurement results. Therefore, the steel production, solidification, and formation of primary and secondary carbides, heat treatment, and achievement of an equilibrium state play an important role.
The experimental work in the publications listed in Table 2 is based on mostly different approaches in this regard. For example, Fischmeister [19] generated the samples by remelting in a regular 10 kg laboratory induction furnace and Zhu [24] by melting and alloying in an induction furnace and subsequent electro slag remelting of a 500 kg ingot. Wang, [20] on the other hand, produced the samples by remelting 50 g samples 10 times in an arc furnace. The wide variety of melting practices, specimen size, solidification conditions, and heat treatment of the samples results in limited comparability of the results. As there is no standard for sample production, heat treatment, and the measurement parameters for determining the phase transformations and the carbide type-but rather a large number of different approaches-the following chapter points out an own method development to clarify a valid workflow.

Experimental Program, Sample Preparation, and Chemical Analysis
The experimental program and actual chemical analyses are summarized in Table 3. All samples were produced in highfrequency remelting (HFR) furnaces from LINN High Therm GmbH ("Lifumat-Met-3.3-Vac" for 60 g and "PlatiCast-600-Vac" for 450 g samples), visualized in Figure 2, which is a proven method for the preparation of laboratory samples. [26,27] 60 g HFR samples are cost-effective and sufficient for performing chemical analysis and DSC measurements. If there is a need for more sample material for further investigations, e.g., annealing treatments and XRD measurements, large 450 g HFR samples are used. Both HFR sample sizes show the same microstructure and sample quality.
Commercial-purity iron cylinders (99.9 wt% Fe), pure chromium (99.5 wt% Cr), tungsten powder (99.9%, 44 μm, Alpha Aesar LOT: 61800610), molybdenum powder (99.95%, 149 μm Alpha Aesar LOT: Q24F052), aluminum wire, and a previously prepared Fe-4.4 wt% C alloy were used as input material. The melting process was always carried out in a new alumina crucible under argon 5.0 overpressure atmosphere. After the melt was homogenized for 3 min by inductive steering, to reduce temperature and chemical gradients and dissolve the Mo and W powders, the finished melt was spin-cast into a copper mold. The final chemical analysis of the Cr/W sample was determined by optical emission spectroscopy (OES, type OBLF QSG 750) and for the Mo/W-alloys by X-ray fluorescence spectrometer (XRF, Thermo Fisher XRF ARL 9900).
The use of fine, high-purity powders of the refractory metals molybdenum and tungsten placed in an iron cylinder, as shown in Figure 2, resulted from an extensive in-house Table 1. Typical carbide types and their properties. [3] Carbide type

Microhardness
[HV] Lattice type Carbide-forming elements M 3 C 800 Orthorhombic Fe (cementite), Cr, Mn M 7 C 3 1100-1400 Hexagonal and orthorhombic [4,5] Fe, Cr There is an increased risk of oxygen input when using metallic powders due to the large surface area of the fine powder. To eliminate the risk of oxygen, all samples were deoxidized with 0.03 wt% aluminum wire, which is common practice. The use of conventional ferroalloys, such as ferrotungsten and ferromolybdenum, is not possible due to their high content of undesirable trace elements like C, Si, Mn, P, S, and N; especially the high phosphorus content has a negative effect. Using highpurity lumpy tungsten (T Melting ¼ 3422°C) and molybdenum (T Melting ¼ 2623°C) is impossible because of their high melting points. In order to evaluate the sample quality, the HFR samples were examined metallographically, polished without etching by SEM-BSE (field-emitting scanning electron microscope, JEOL 7200F; JEOL Germany GmbH, Freising, Germany) with backscattered electron image mode (BSE) and different magnifications, visualized in Figure 5    a uniform solidification structure and the examination shows that all powder particles were completely dissolved, i.e., no Mo or W powder particles were found, which would clearly light up in BSE mode.

DSC Setup
The well-established DSC method provides a way to record all phase transitions associated with an exo-or endothermic effect (¼enthalpy change). Detailed information on the DSC technique to characterize phase transformations can be found in the NIST recommended practice guide by Boettinger. [17] The DSC analysis was carried out in a DSC 404F1 Pegasus (NETZSCH Gerätebau GmbH, Selb, Germany) with an Rh furnace and a platinum DSC sensor instrumented with type S thermocouples. Al 2 O 3 crucibles with a volume of 85 μL and lids were used for all experiments; the reference was an empty crucible in each trial. All DSC measurements start with the following "start sequence" consisting of three full furnace evacuations and purge cycles, followed by 45 min of intensive purging with 150 mL min À1 Ar 5.0 (purity 99.999 pct) at room temperature. During the measurement, the protective tube of Table 2. Overview of currently examined systems and their analysis methods. [8] Source System Carbide types Analysis methods a) Phase transf. b) Carbide type Fischmeister [19] Fe Rahimi [5] Fe-17Cr-9Ni-6Mn-4Al-0.42C Wang [20] Fe-10Cr-0.15C M 3 C, M 23 C 6 , M 7 C 3 a) DSC b) SEM, in situ TEM Maurizi-Enrici [21] HSS steel M 2 C, MC, M 6 C, M 23 C 6 a) DTA b) SEM, EBSD Bombac [22] Fe-C-Cr M 23 C 6 , M 7 C 3 a) DTA b) SEM, XRD Guo [23] Fe-10Cr-3Mo-3V-1W MC, M 2 C, M 7 C 3 a) DSC b) SEM, XRD Zhu [24] 8Cr13MoV M 7 C 3 a) ThermoCalc b) TEM, XRD Wieczerzak [25] Fe-25Cr-5Mo-0.82C M 23 C 6 , M 7 C 3 , M 6 C a) ThermoCalc b) SEM, EBSD, XRD Table 3. Test alloys, chemical analysis, and analysis performed. [8] Sample label and sample weight Chemical analysis [wt%] Analysis performed the Rh furnace was continuously purged with 70 mL min À1 Ar, and a thermally active Zr getter was placed below the DSC sensor to avoid oxidation of the sample at temperatures above 350°C. The experimental setup was calibrated by measuring the melting points and melting enthalpies of NETZSCH's standards of pure metals In, Bi, Al, Ag, Au, Ni, and Co. In all DSC experiments, a heating rate of 10°C min À1 was applied, and samples of 50 mg (2.1 mm Â 2.1 mm Â 1.5 mm, ground, not polished) were used. A relatively low sample mass of 50 mg was selected for all experiments to guarantee near-equilibrium conditions without any temperature gradients inside the sample. Melting of Fe-based alloys takes place under a considerable change in heat (ΔH > 200 J g À1 ). Within the DSC analysis of strongly endothermic phase transformations, special care must be taken to determine the equilibrium temperatures accurately. An effective way to exclude the experimental setup influences on the DSC signal is provided by NETZSCH's Tau-R software (NETZSCH Gerätebau GmbH, Selb, Germany). [28,29] The Tau-R method enables the determination of the equilibrium data from only a single DSC experiment. At this point, the authors refer to their previous work [26,27,30] regarding more detailed information on the DSC setup, the Tau-R method, and its successful application to characterize melting equilibria in steel using DSC. In the following sections, an onset in the DSC signal corresponds to the first deviation from the baseline that can be assigned to the beginning of the phase transformation, and a peak defines the end of the phase transformation.

Annealing and Quenching Technique and XRD Setup
All heat treatments in this work were performed using a Tammann furnace (HRTK 32 Sond., Ruhstrat, Germany). This electric resistance tube furnace with an Al 2 O 3 protective tube inside was modified with a gas-tight lock at the top and bottom to quickly take out the samples for the quenching experiments to freeze the high-temperature structure. The samples were placed in the center of the furnace, hanging on a Mo wire, next to the Pt10Rh-Pt thermocouple for the furnace control, to have a good temperature control. The sample size was %30 mm Â 30 mm Â 10 mm, and the samples were fixed either directly with a high-temperature-resistant Mo wire or in a suitable Al 2 O 3 crucible for solid-liquid experiments.
The furnace was continuously purged with argon and due to the carbon heating tube and its reaction with the oxygen; the final oxygen content was extremely low (0.001 ppm) to ensure that no oxidation of the samples occurred. For quenching, a tin bucket with fresh ice water was placed under the furnace, the lower sluice was opened, and the Mo wire was cut off at the top, making the sample fall into the ice water within 0.1 s. In addition, several samples can be placed in the furnace simultaneously and quenched after different times, see Figure 3b.
Selected samples, listed in Table 3, were annealed and quenched and subsequently examined using XRD measurements to determine the present phases. The XRD samples were cut, ground, and polished from the quenched samples with the following dimensions: 25 mm Â 25 mm Â 3 mm. XRD measurements were performed on a D8 Discover diffractometer (by Bruker AXS, Karlsruhe, Germany) using a molybdenum X-ray tube at 40 kV, 40 mA (Mo Kα1 radiation, λ ¼ 0.7093 Å). The irradiated sample diameter was approximately 4 mm; the diffraction angle range of 12-58°2θ was scanned with a step size of 0.03°2θ and a corresponding measurement time of 6°s step À1 . The Diffrac.EVA 6.0.0.7 software (Bruker AXS) was used for the qualitative analysis of the crystalline phases. The quantitative phase analysis was conducted by means of the Rietveld refinement technique using the Topas V5.0 software (also by Bruker AXS).

Method Development for the Characterization of Fe-6W-5Mo-C alloys
The first DSC measurement discussed is of alloy Fe-6W-5Mo-0.8C, illustrated in Figure 4, showing a) the high-temperature phase transformations under equilibrium conditions (Tau-R calculation [28] ) and b) and c) the performed time-temperature program. The alloy was measured with the following conditions: 1) signal "green" (sample a) with a continuous heating rate of 10 K min À1 from 450 to 1470°C (T max ¼ approx. 20°C above the expected liquidus temp.) and cooling to room temperature with À30 K min À1 ; 2) signal "blue" (sample b, to test the reproducibility) with a targeted time-temperature program consisting of fast heating with 30 K min À1 to 1170°C (T 1 ¼ 100°C below the carbide peak, determined from the first measurement) and an isothermal annealing phase of 15 min followed by continuous  heating with 10 K min À1 again to 1470°C; 3) subsequently, the cooling down with À30 K min À1 is stopped at 1050°C (T 2 ¼ well below the carbide formation during cooling, determined from the first measurement) and the solidified sample isothermally annealed for 15 min; and 4) signal "purple" (¼sample b) shows the reheating with 10 K min À1 of the same sample again to 1470°C and at last cooling to room temperature with À30 K min À1 .
The DSC measurements of both samples (a and b) show a very high agreement despite the different time-temperature programs, which is simply a "pretest" to check if the measurements and samples are balanced.
The second melting, within the time-temperature program visualized in Figure 4c, of sample b (purple line) also shows (despite slight decarburization that probably took place) an identical transformation behavior for the carbide and liquidus peaks (only the peritectic peak is slightly delayed to higher temperatures due to nucleation phenomena). The cooling down signal, visualized in Figure 4a, shows that strong nucleation phenomena result in remarkable supercooling of the phase transformation and will not be used to determine transformation temperatures.
For a deeper understanding of the DSC measurements, the DSC samples were examined metallographically after the trials, as exhibited in Figure 5. The primary carbide present from the solidification structure (DSC with À30 K min À1 cooling) could be identified as M 6 C based on the herringbone-like morphology, which has a similar configuration to that reported in the literature by Chaus. [31] Furthermore, the investigation shows that the carbide formation of the as-cast HFR sample and the sample melted and solidified in the DSC appears to be comparable, but the M 6 C carbide formation in the HFR sample is significantly finer due to the high cooling rate from the spin casting process in a copper mold.
To confirm the suspected M 6 C phase, a sample was annealed at 1180°C for 4 h, quenched in ice water and analyzed by XRD measurement according to the technique described in Section 3.3. Figure 6a shows the microstructure after annealing at 1180°C for 4 h, where the M 6 C carbides are already formed; see Figure 5 with the original herringbone-like morphology after solidification for comparison. The XRD analysis presented in Figure 6b clearly confirms the presence of primary M 6 C carbide in the high-temperature range (note that the phase fractions should only be considered as a guide). Despite the high quenching speed, the austenite phases cannot be "frozen" and martensite is formed in large amounts. Furthermore, very small amounts of MC phase near or below the detection limit of the XRD method could be present; however, no individual diffraction peaks of such a phase could be observed (overlapping effects for the strongest maxima are clearly present). Future TEM investigations would be useful here. According to thermodynamic calculations using TCFE 10.1 and the FSstel2020 database, MC should form below 800°C. Obviously, as the quenching speed is apparently not high enough, the formation cannot be fully suppressed. A faster cooling rate, e.g., by the use of smaller samples, would be an improvement in this regard.
The joint investigations of the alloy Fe-6W-5Mo-0.8C consisted of: 1) DSC measurements visualized in Figure 4a including first melting and remelting; 2) SEM-BSE studies in Figure 5 of the as-cast and solidified DSC samples and in Figure 6a of samples annealed at 1180°C for 4 h and quenched; 3) XRD measurement of a sample annealed at 1180°C for 4 h and quenched in Figure 6b.
The results confirm that the primary carbide is an M 6 C type. The phase sequence in the considered temperature range from 1050 to 1500°C is as follows: The other two Fe-6W-5Mo-C alloys show comparable behavior: 1) samples a, b and the remelting were identical in each case; 2) with the lower carbon content of 0.5 wt% C, the carbide peak was also smaller; 3) with the high carbon content of 1.2 wt% C, the carbide peak was significantly larger; and 4) the alloy with 1.2 wt% C remains in the γ phase region and no peritectic phase transformation occurs anymore because the higher carbon content stabilizes the austenitic region and δ-delta ferrite is no longer thermodynamically stable.
The determined phase transformation temperatures of the alloys are summarized in Table 4. For the Fe-C-6W-5Mo system, DSC measurements of as-cast HFR samples represent a safe and proven method for determining the phase transformations in the high temperature range. Even if the use of the triedand-tested DSC method appears to be safe and uncomplicated, Figure 5. SEM-BSE image of the alloy Fe-6W-5Mo-0.8C: a) before and b) after the DSC measurement, with different magnifications. Reproduced with permission. [8] Copyright 2022, ASMET. c) Typical M 6 C carbide from literature. Reproduced with permission. [31] Copyright 2016, Springer Nature. this must be questioned anew and critically for each system, as the subsequent study shows.

Method Development for the Characterization of Fe-10Cr-3W-C Alloys
The second DSC measurement discussed is of alloy Fe-10Cr-3W-1.5C, visualized in Figure 7a, and was carried out using the same methodology as the Fe-6W-5Mo-0.8C alloy before. Only the time-temperature program, outlined in Figure 7b,c, was adjusted to the actual transformation temperatures.
The DSC measurements of samples a and b visualized in Figure 7a show a good agreement. However, the DSC signal for the remelting of sample b (purple line) shows a new separate peak at %1200°C that has not been previously detected and should be studied in detail. The subsequent DSC signal from the second melt is relatively similar. Nevertheless, the onset of the large carbide peak is difficult to evaluate due to peak overlap with the small peak. Due to a slight decarburization of approx. 0.02 wt% C between the first and second melting cycles, the liquidus temperature rises and the liquidus peak is shifted slightly to higher temperatures. Using only the second melting alone is therefore not satisfactory. It should be noted that it is impossible to make any statements about the present phases and carbide types based only on this DSC measurement.
Metallographic investigations were carried out the same way as for the alloy before, presented in Figure 8a,b. Dark lamellar carbides appear in the material, similar to M 7 C 3 in morphology, [32] as shown in Figure 8c. Furthermore, some isolated "brightly shining" areas appear in the SEM-BSE image, indicating a high concentration of elements with a higher mole mass than Fe. But these bright areas are only very small (%1 μm), indistinct, and cannot be identified yet. The examination of the solidified sample from the DSC shows the dark lamellar-like M 7 C 3 carbides again, but now also clearly pronounced, a bright carbide type with a comparable configuration to M 23 C 6 carbides, reported in the literature [33] and shown in Figure 8d.
The observation of a second carbide type in the micrograph of the solidified DSC sample, visualized in Figure 8b, is likely the explanation for the new small carbide peak during the second DSC measurement with remelting. Primary M 7 C 3 and secondary M 23 C 6 are typical carbides in higher alloy chromium and tungsten steels but are difficult to distinguish optically. The SEM-BSE investigations in Figure 8 clearly show the dark Cr-containing M 7 C 3 carbides and the obviously bright M 23 C 6 carbides in the BSE mode, due to their high content of the heavy element tungsten. Figure 8a shows that no relevant amount of M 23 C 6 carbide was formed in the as-cast HFR sample because of the rapid solidification and cooling. The microstructure shows a fine and homogeneous structure, but still without secondary carbide (i.e., the corresponding alloying elements are still forcibly dissolved in the matrix). The M 23 C 6 carbide forms when the material melts again and solidifies slowly during the DSC measurement, as visualized in Figure 8b.
For a better understanding of the sample quality and the formation of the M 23 C 6 carbide, annealing tests were performed on as-cast HFR samples in the high-temperature region of the γ þ M 23 C 6 þ M 7 C 3 phase region, immediately before the M 23 C 6 dissolution. For this purpose, the samples were annealed for 2 and 4 h at 1180°C and quenched in ice water according to the annealing and quenching technique described in Section 3.3. These investigations should also show whether it is possible to pretreat the as-cast HFR samples before the DSC measurement to reach an equilibrium state. The micrographs of the annealed samples showed that the amount of the bright M 23 C 6 carbide Table 4. Compilation of the experimentally determined data. [8] Fe-6W-5Mo-C system DSC run: first melting of the virgin material Sample label/temperatures in°C determined by DSC Symbol in Figure 12 Fe-6W-5Mo-0. Fe-10Cr-3W-C system DSC run: first melting, after equilibration cycle (partial premelting) Symbol in Figure 12 Fe-10Cr-3W-   increases significantly after 2 h, due to the precipitation of the dissolved elements from the matrix of the as-cast HFR sample. However, even after an annealing time of 4 h, the structure is still not comparable with the remelted DSC sample from Figure 8b. This means that a classical 4 h diffusion annealing of the HFR samples is insufficient to reach an equilibrium state for the DSC samples. Diffusion annealing with longer holding times was not performed, as this treatment is an expensive and slow method. At this point, it must also be noted that a detailed study of the M 7 C 3 -M 23 C 6 transition was not conducted because it would require further investigations, as described in the following sources. [34,35] In order to quickly reach a near-equilibrium state in the γ þ M 23 C 6 þ M 7 C 3 phase field without complete remelting and solidification in a second DSC cycle, with the risk of decarburization and losing alloying elements in the sample, the following promising approach was tested. Internal DSC test series with different variants of time-temperature programs using the alloy Fe-10Cr-3W-1.5C showed the following findings: 1) an annealing treatment in the solid state of the sample is not sufficient to reach an equilibrium state, as also shown by the Tammann furnace experiments with 2 and 4 h of annealing in Figure 9; 2) however, partial semisolid melting, in this case 5 min isothermal at 1300°C and slow cooling again approx. 50-100°C below the expected phase transformation, allows the rapid attainment of an equilibrium state and the formation of the M 23 C 6 carbide peak, visualized in Figure 10a; 3) a partial melting up to a liquid content of approx. 30-50%, based on the evaluation of the first DSC measurement of a new sample, proves to be very efficient; and 4) this partial premelting cycle in the semisolid area and subsequent slow cooling in the DSC time-temperature program enables a reliable determination of the phase transformations without long annealing times, the risk of decarburization, and the high thermal stress over 1400°C during complete remelting in the DSC. DSC measurements in the high-temperature range lead to ongoing aging of the heating elements of the furnace and the platinum DSC sensor. In particular, the aging of the thermocouples (shorter lifetime and more calibration measurements are necessary), the thermomechanical deformation of the DSC sensor and the risk of sticking of the Al 2 O 3 crucibles are associated with exposure times at higher temperatures.
This semisolid holding during the DSC measurement is also advantageous over full melting in the sense that possible adverse interactions between the liquid phase, the furnace atmosphere, and the Al 2 O 3 crucible material, e.g., decarburization, evaporation, oxidation, contamination, and reactions with the crucible, are restricted. Furthermore, the typical balling of specimens during melting, which reduces the contact area with the bottom of the DSC crucible and weakens the DSC signal, would not occur during a semisolid treatment. Despite these advantages, the danger of new segregations forming during solidification (depending on the respective alloy) must be pointed out, which can occur with both classic melting through and partial melting.
To evaluate the microstructure after the partial premelting (¼situation before the actual DSC measurement), the first step of the time-temperature program of the DSC cycle, visualized in Figure 10b, was simulated in the Tammann furnace according to the annealing and quenching technique described in Section 3.3. This extra step in the Tammann furnace was necessary as the closed DSC plant does not allow the extraction of a hot sample out of the system. The investigation in the SEM-BSE mode clearly shows an almost identical structure to the fully melted and solidified DSC sample from Figure 8 and the semisolid premelted and quenched microstructure in Figure 10.
This confirms that an equivalent structure can also be achieved without complete melting. Additional XRD measurements could not be performed in this case due to the too-small sample.
Comparable investigations with a semisolid treatment were also carried out by Omar [36] on M2 high-speed steel, by Mohammed [37] on D2 cold-work tool steel, and by Wang [38] on a 9Cr18 steel with the aim of dissolving all carbides and homogenizing the material in a short process time. The typical reported holding time during the partial melting was 5 min, identical to the own time-temperature program used in this work and visualized in Figure 10b. The motivation of these investigations was to characterize and optimize the alloy behavior for the semisolid forming (SSF, e.g., thixoforming) process.
The great advantage of heat treatment in the semisolid area is that the diffusion takes place very quickly. This means that the carbides and alloying elements in the matrix dissolve very rapidly and the austenite grains are retaining and no noticeable segregation effects occur during subsequent cooling. [36] This additional "equilibration step" with partial premelting can be easily integrated into the DSC time-temperature program visualized in Figure 10b and increases the measuring time by only %40 min (depending on the required temperature levels), Figure 9. SEM-BSE image of the alloy Fe-10Cr-3W-1.5C from different annealed states at 1180°C. Reproduced with permission. [8] Copyright 2022, ASMET.
www.advancedsciencenews.com l www.steel-research.de steel research int. 2023, 94, 2200503 Figure 10. a) DSC measurement after the partial premelting of the alloy Fe-10Cr-3W-1.5C, b) the performed time-temperature program, and c) SEM-BSE image of the situation after the partial premelting. Reproduced with permission. [8] Copyright 2022, ASMET. which is acceptable in terms of the system load in the hightemperature range and the measurement costs.
Despite the previous DSC and SEM-BSE investigations, the two suspected carbide types M 23 C 6 and M 7 C 3 need to be confirmed according to the technique described in Section 3.3. 1100°C was chosen as the annealing temperature to investigate whether the thermodynamic calculation with the TCFE10.1 (phase area with γ þ M 23 C 6 þ M 7 C 3 ) or the FSstel2020 (γ þ M 7 C 3 ) database fits better as visualized in Figure 1c,d. Figure 11a shows the microstructure of a sample annealed at 1100°C for 4 h and quenched, where the two different carbide types can be identified in the BSE mode, comparable with the previous investigations shown in Figure 8 and 7. The XRD analysis presented in Figure 11b clearly confirms the presence of the M 23 C 6 and M 7 C 3 carbides. The diffraction pattern of the hexagonal M 7 C 3 phase was used for the Rietveld refinement. No further XRD peaks fitting to the corresponding pattern of the orthorhombic phase with the same composition could be clearly found in the actual measurements. According to the previous investigations of samples annealed at 1100°C for 2 and 4 h, respectively, as well as of quenched material, as depicted in Figure 9, there is still a Figure 12. Calculated phase diagrams and experimentally determined data: a) Fe-6W-5Mo-C var with the TCFE10.1, b) Fe-6W-5Mo-C var with the FSstel2020 database, c) Fe-10Cr-3W-C var with the TCFE10.1, and d) Fe-10Cr-3W-C var with the FSstel2020 database. risk that diffusion annealing for 4 h is insufficient to reach a steady state for the M 7 C 3 -M 23 C 6 transition. Therefore, the phase fractions given in Figure 11 should again be considered only as a guide.
The joint investigations of the alloy Fe-10Cr-3W-1.5C consisted of: 1) DSC measurements visualized in Figure 7a including first melting and the remelting and visualized in Figure 10a the first melting after partial premelting; 2) SEM-BSE studies in Figure 8 of the as-cast and solidified DSC samples, in Figure 10c of samples after partial premelting at 1300°C and quenched and in Figure 11a of samples annealed at 1100°C for 4 h and quenched; and 3) XRD measurement of a sample annealed at 1100°C for 4 h and quenched in Figure 11b.
The results confirm that the primary carbide is an M 7 C 3 and the secondary M 23 C 6 . The phase sequence in the considered temperature range from 1050 to 1500°C is as follows: The other two Fe-10Cr-3W-C alloys show similar behavior: 1) samples a and b were identical in each case and the DSC measurements c1 and c2 after the partial premelting show the separate secondary carbide peak; 2) with the lower carbon content of 1.2 wt% C, both carbide peaks (primary and secondary) were also smaller; and 3) with the high carbon content of 2.0 wt% C, both carbide peaks were significantly larger. The determined phase transformation sequence and the corresponding equilibrium transformation temperatures of the alloys are summarized in Table 4.

Compilation of the Results
The phase transformations determined for the Fe-6W-5Mo-C and Fe-10Cr-3W-C alloys are compiled in Table 4. For the system Fe-6W-5Mo-C, the temperatures are the average of two measurements (samples a and b), melting the first time of an as-cast sample in each case, with the time-temperature programs of Figure 7a,b. For the system Fe-10Cr-3W-C, the temperatures are the average of two measurements (samples c1 and c2), melting the first time of a partial premelted sample in each case, with the time-temperature programs of Figure 10b. A comparison with the calculated phase diagrams in Figure 1 and the experimentally determined data from Table 4 is presented in detail in Figure 12. Figure 12a,b shows that the liquidus temperature of all Fe-6W-5Mo-(0.5/0.8/1.2)C alloys is in good agreement (AE3°C) with both thermodynamic databases (TCFE10.1 and FSstel2020). On the other hand, the solidus temperature (Tγ þ M6C!γ þ M6C þ L) is calculated well by the TCFE10.1 database (AE10°C) visualized in Figure 12a, but the FSstel2020 databases calculate this transformation significantly too low (>80°C), compiled in Figure 12b.
For the Fe-10Cr-3W-(1.2/1.5/2.0)C system, the liquidus temperature is again well described (ΔT < 10°C) with both databases, visualized in detail in Figure 12c,d. The predicted solidus temperatures for the FSstel2020 database are about 40°C lower and for the TCFE10.1 40°C higher than the measured values. The most significant difference is in the area of the three-phase region (γ þ M 7 C 3 þ M 23 C 6 ): the calculation of TCFE 10.1 shows a much closer agreement with the measured values.
The calculated two-phase area (γ þ M 7 C 3 ) from the FSstel2020 could be definitively refuted with the structural investigations of Figure 9 (annealing treatment þ SEM-BSE at 1180°C with an M 23 C 6 þ M 7 C 3 microstructure) and a supplementary XRD measurement at 1100°C with both carbide types presented in Figure 11b. The measurement results indicate that thermodynamic databases should not be used without careful evaluation.

Conclusion
A promising experimental procedure has been developed to measure the equilibrium transformation temperatures of primary carbides in tool steels by DSC measurements. Regular diffusion annealing of as-cast carbidic steels is very time-consuming, but with an additional heat treatment step in the semisolid zone with a 30-50% liquid phase fraction, a high level of equilibration can be achieved within a few minutes due to the high diffusion rates. This additional heat treatment step with partial premelting of the as-cast sample can be easily integrated into the DSC time-temperature program and increases the measuring time by only %40 min (depending on the required temperature levels). Furthermore, this semisolid holding during the DSC measurement is also advantageous over full melting due to the following points: 1) reduced risk of reactions between: atmosphere ⇔ sample ⇔ crucible; 2) no negative balling of specimens and the risk of reduced contact area; and 3) reduced thermal load on the DSC system in the high-temperature range.
This described methodology with partial premelting is an essential improvement to regular DSC measurements, where often only the first heating is considered (caution-secondary carbides often do not appear yet) or DSC measurements where only the second time heating is considered, after a complete premelt (with the risks previously described).
To prove the potential of this equilibration step by partial premelting in the DSC, additional equilibration and quenching experiments were performed at selected temperatures in a Tammann-type furnace and investigated using SEM-BSE and XRD analysis. By combining these methods, carbide types and the transformation temperatures can be verified to evaluate and construct complete phase diagrams in the high temperature range. It must be noted, however, that no detailed investigations of the solid-solid transformations, e.g., the M 7 C 3 -M 23 C 6 transition and possible formation of new segregation during the partial premelting, were applied.
Finally, it can be stated that the laboratory sample production from high-purity alloying elements using an HFR furnace with spin-casting and DSC measurements with regular heating and supplemented by partial premelting in the DSC timetemperature program is a proven method for creating phase diagrams. Especially by using small 60 g HFR samples for alloy variations, which can be produced quickly and cheaply, and automated DSC measurements, the influence of individual alloying elements can be investigated very efficiently. The comparison of the measured phase transformation temperatures with results from the thermodynamic databases FSstel2020 and TCFE10.1 shows good agreement of the liquid temperatures but significant discrepancies between the measured and predicted solidus temperatures.