Additive Manufacturing of Tool Steels

The public funded project “AddSteel” aims to develop functionally adapted steel materials for additive manufacturing (AM). Based on the AM process laser powder bed fusion (LPBF), the holistic process chain, including alloy design, powder atomization, AM, and postheat treatment, is considered to achieve this objective. Tool steels are usually characterized by higher carbon content and limited weldability, leading to limited processability for LPBF. To extend these limitations, different approaches for tool steels are investigated: for high‐carbon tool steels, the effects of lower martensite start temperature are investigated using the alloy 1.2842 as an example. A low martensite start temperature seems to be advantageous for crack‐free processing with LPBF. In order to avoid a high hardness level after rapid cooling, the use of a hot work steel with a carbon content of 0.2 wt% is investigated. Due to the chemical composition of the material, a moderate preheating temperature <300 °C is required. In addition, very high scanning speeds are possible with an improved shielding gas flow. Finally, the experience along the process chain with the standard steels is used for a modification of the alloy 1.2344. The effects of this modification on AM and heat treatment are investigated.


Introduction
In recent years, additive manufacturing (AM) developed from a prototype manufacturing method to an approved manufacturing process for complex functional parts capable for scaling on industrial level. [1] In the field of AM for metallic components, laser powder bed fusion (LPBF) is regarded as the AM process most suitable to achieve this state. [2,3] With LPBF, complex parts with high detail resolution and a high degree of freedom in design can be produced from a powder bed through layer-wise selective melting of metal powder using laser radiation. Due to the high hardness and the wide range of adjustable material properties, the processing of tool steels by means of LPBF shows great potential, e.g., for the manufacturing of tools with integrated cooling channels or lightweight-optimized structural components. [4,5] However, conventional tool steel alloys are usually characterized by higher carbon content and limited weldability, leading to limited processability for LPBF. [6][7][8][9][10][11] The local heat input, high heating and cooling rates, and recurring remelting during the LPBF process lead to large internal stresses which often result in distortion and crack formation in parts made of tool steels. [12][13][14][15][16][17] In this article, results of the public funded project "AddSteel" are presented, in which the effects of different martensite start temperatures on the processability of tool steels using LPBF are investigated. Preliminary studies showed that tool steels with a lower martensite start temperature (M S ) could be processed crack-free for a wider range of LPBF parameter combination. [18][19][20][21] At first, this effect is validated using two standard tool steels, one cold work steel with high carbon content (Cryodur 2842) and one hot work steel (Thermodur 2322). Modified versions of these standard steels with an adapted chemical composition are subsequently investigated with the aim of achieving an increased productivity along the LPBF process chain by maintaining a high part density without crack formation. Finally, the chemistry for the hot work tool steel Thermodur 2344 is adapted to decrease the specific martensite start temperature and its general processability with LPBF. In addition, the processability for parameter sets with higher productivity is investigated.

Materials and Investigations
In this work, five different tool steels were investigated: the cold work tool steel Cryodur 2842, [22] the hot work tool steel Thermodur 2322, [23] and their corresponding modified versions as well as modified Thermodur 2344. The chemical composition of all steels in wt% is given in Table 1. The nominal composition of standard Thermodur 2344 according to ref. [24] is also represented as a reference. The adjusted chemical composition for the modifications is highlighted.
The public funded project "AddSteel" aims to develop functionally adapted steel materials for additive manufacturing (AM). Based on the AM process laser powder bed fusion (LPBF), the holistic process chain, including alloy design, powder atomization, AM, and postheat treatment, is considered to achieve this objective. Tool steels are usually characterized by higher carbon content and limited weldability, leading to limited processability for LPBF. To extend these limitations, different approaches for tool steels are investigated: for high-carbon tool steels, the effects of lower martensite start temperature are investigated using the alloy 1.2842 as an example. A low martensite start temperature seems to be advantageous for crack-free processing with LPBF. In order to avoid a high hardness level after rapid cooling, the use of a hot work steel with a carbon content of 0.2 wt% is investigated. Due to the chemical composition of the material, a moderate preheating temperature <300°C is required. In addition, very high scanning speeds are possible with an improved shielding gas flow. Finally, the experience along the process chain with the standard steels is used for a modification of the alloy 1.2344. The effects of this modification on AM and heat treatment are investigated. Powder material used in LPBF requires mainly spherical particles in a specific particle size distribution (PSD). Gas atomization processes like electrode induction melting inert gas atomization (EIGA) are state of the art for producing such powder material. Sulfur (S) was added to decrease the surface tension of Cryodur 2843 and Thermodur 2322 and thus to increase the powder output of EIGA processes in the desired PSD ( Figure 1) while maintaining a round particle shape with no internal porosity.
This leads to an increased productivity along the LPBF process chain. The increased content of manganese (Mn) in Thermodur 2344 shifts M S to lower temperatures, close to M S of Cryodur 2842. The M S for all materials was simulated using the thermodynamic calculation software ThermoCalc. For the standard materials, as well as for their modifications, a powder analysis was performed regarding particle shape, using SEM and PSD. All particles were identified as spherical shaped with a size of 10-45 μm in diameter and with no internal porosity. An exemplary SEM image from Cryodur 2842 mod. particles and the resulting quantile distribution are shown in Figure 2.
All presented investigations were carried out on a MIDI LPBF system from Aconity3D using a single-mode fiber laser with the maximum laser power of P L ¼ 400 W and a Gaussian intensity distribution. The beam was focused to 80 μm in diameter. Using argon as shielding gas, an oxygen concentration O 2 < 100 ppm was achieved. Additionally, the machine is equipped with a resistive base plate preheating system achieving a maximum temperature of 800°C. The preheating system reduces the size of the build volume to a diameter of 100 mm. To compare the results for each material, a standard test layout was defined, containing 24 cubic test samples (Figure 3), where each of the 10 Â 10 Â 10 mm 3 cubes can be assigned to a different process parameter combination. The scan sequence of the samples is against the flow direction of the shielding gas. Influence of the part position on the base plate is taken into account with five marked reference test  samples which are manufactured with the same process parameter set. In case the variance in density of these samples would be significant, an influence of part position could be assumed.
The density of all samples was measured with a computerassisted image analysis on a cross section using images with 100x magnification. The cross sections were performed parallel to the build direction ( Figure 3, plane A) and then grounded and polished with 1 μm diamond suspension. As baseline of this work, all materials were manufactured with a layer thickness of D S ¼ 30 μm. To compare different process parameter combinations, the volume energy density E V and the theoretical build rate V th are used ( Figure 3). For Cryodur 2842, Thermodur 2322 and their modifications, a volume energy density between E V ¼ 35-130 J mm À3 was used to manufacture the samples at room temperature. The reference test samples were manufactured with E V ¼ 83.33 J mm À3 . Due to the higher carbon content of the cold work steels Cryodur 2842 and Cryodur 2842 mod., an additional investigation with a preheating temperature of 200°C was performed. Compared to the other steels investigated in this work, Thermodur 2344 mod. has a higher content of carbide forming elements Cr and Mo and a medium carbon content. In expectation of a more limited processability with LPBF, the volume energy was adapted to E V ¼ 47-167 J mm À3 and manufacturing was performed by preheating temperatures of 200, 300, and 500°C. [24] In a second step, the layer thickness D S and the hatch distance dy S for all modified materials were increased to achieve an increased productivity. Based on the results from the baseline setup, laser power P L , scan speed v S , and preheating temperature T P were chosen such that a density >99.5% can be achieved. For Cryodur 2842 mod. and Thermodur 2322 mod., layer thicknesses of D S2 ¼ 90 μm and D S3 ¼ 120 μm with corresponding hatch distances of dy S2 ¼ 100 μm and dy S3 ¼ 120 μm were investigated. For Thermodur 2344 mod., the layer thickness was increased to 60 μm. A summary of all process parameters is shown in Table 2.

Standard Process Parameter Set (D S ¼ 30 μm)
To confirm that tool steels with a lower M S can be processed crack-free with a density >99.5% for a wider range of LPBF parameter combinations, a standard parameter set (D S ¼ 30 μm) was examined for all materials. The results of the density measurement of the test samples made of Cryodur 2842 and the modified version is shown in Figure 4.
For nearly all investigated parameter combinations, a density >99.5% can be achieved and no crack formation is detected. Preheating to 200°C shows almost no effect on the standard Cryodur 2842. However, for the modified versions the density can be increased significantly above the target density at low E V when using preheating. Independent of the material or the preheating temperature, a similar density is achieved for the reference samples. The small variance of these samples indicates a stable process regardless of the position on the building platform. Although Thermodur 2322 has a much lower carbon content compared to Cryodur 2842, the range of parameter combinations in which a density >99.5% can be achieved is much smaller ( Figure 5).
For Thermodur 2322 and its modification, the density to be achieved can be reached for E V > 75 J mm À3 . Although Thermodur 2322 transforms predominantly bainitic, a theoretical M S of %400°C can be determined for very rapid cooling (<12 s). [23] This transformation may account for the smaller parameter range in which Thermodur 2322 can be processed with a density >99.5% compared to Cryodur 2841. However,     the range of process parameters in which Thermodur 2322 can be processed with such a density is much wider than for other nonbainitic tool steels, like, for example, Thermodur 2344. As before, no crack formation is detected for all investigated parameter combinations. The modification of Thermodur 2322 with an increased sulfur content can also be processed but in an even smaller range of parameter combinations to achieve density >99.5%. As already with Cryodur 2842, the reference samples of both types of Thermodur 2322 show a small variance indicating a stable process all over the build plate. For the alloy Thermodur 2344 mod., Mn content was increased resulting in a martensite start temperature close to Cryodur 2842 ( Figure 1). However, the higher amount of carbide forming elements limits the processability of this alloy with LPBF, as the density analysis shows ( Figure 6). The target density can only be achieved in isolated cases, but the overall density seems to be higher for samples with a lower E V . There is also a tendency for greater density to be achieved with a lower preheating temperature. A detailed analysis of the detected defects shows that an increased temperature is accompanied by an increased microcrack formation (Figure 7).
In average, these cracks are <200 μm in length, mostly oriented in build direction and scattered across the investigated area. Especially for preheating temperatures >300°C, the incidence for crack formation increases significantly. The measured density for the reference samples varies strongly compared to Cryodur 2842 and Thermodur 2322, which indicates an influence of the sample position on the build plate. Especially the reference samples close to the shielding gas inlet show a decreasing density with increasing preheating temperature. This effect matches the observation of an increased amount of process by-products (spatters and particularly vaporized material) when processing Thermodur 2344 mod. at different temperatures compared to the other investigated alloys in this work. A reason for the increased process by-products may be the increased Mn content. Compared to other alloy elements, Mn has a low evaporation temperature, leading to more vaporized material when manufactured at higher temperatures. Combined with the scan sequence where the samples near the shielding gas inlet are scanned toward the end of a layer and the higher speed of the shielding gas flow at the inlet, an increased interaction between melt pool and vaporized material could lead to the decreasing density.

Increased Productivity Process Parameter Set (D S > 30 μm)
Based on the results of the standard parameter set, a set with increased productivity was investigated for all modified materials. As the parameter set for Cryodur 2842 mod. to achieve a density >99.5% is considerably larger compared to for Thermodur 2322 mod., only the results for Thermodur 2322 mod. (Table 2) are shown in Figure 8 as an example. The targeted density can be achieved for each layer thickness with several parameter sets independent of the layer thickness applied. The minimum E V for which this density is reached decreases with increasing layer thickness. For D S2 ¼ 90 μm, a density >99.5% is achieved for E V > 40 J mm À3 ; for D S3 ¼ 120 μm, the critical E V decreases to >34 J mm À3 . The increasingly steep course of the curves for higher layer thicknesses indicates a limit for a dense processing of Thermodur 2322. For a constant laser power, the theoretical build rate V th increases with decreasing volume energy density (Figure 3). At this point it should be noted that the maximum laser power (400 W) of the LPBF system was used for processing some of these samples. For E V ¼ 34.7 J mm À3 , which used P L ¼ 400 W, the highest build rate of V th ¼ 11.52 mm 3 s À1 can be achieved.
In contrast to the expected behavior derived from Thermodur 2322, Thermodur 2344 mod. shows that with a greater layer thickness almost all process parameter sets investigated are above the density to be achieved. As an example, this is shown in Figure 9 for the samples that were manufactured with a preheating temperature of 300°C.
A detailed analysis of the cross sections for the samples build with D S2 ¼ 60 μm reveals that no crack formation was detected. Furthermore, the density variance of the corresponding reference sample is significantly smaller. While the amount of process by-products remains quite large, also compared to the manufacturing of Thermodur 2322 mod. at higher layer thicknesses, the proportion of spatter and vaporized material seems to shift. For D S ¼ 60 μm, more spatters and a decreased amount of vaporized material can be observed. By remelting a larger amount of powder, less energy is available to vaporize already solidified material. In addition, the LPBF system's shielding gas flow seems to be designed to remove spatter from the processing zone. The flow velocity required for this often leads to turbulence when vaporized material is discharged, so that the vaporized material can interact increasingly with the melt pool, leading to the decreased density for samples with D S ¼ 30 μm.

Conclusion
The results in this work confirm that tool steels with a low martensite start temperature can be manufactured using LPBF for a wide range of process parameter sets. A density >99.5% can be achieved even for sets with a maximum layer thickness of D S ¼ 120 μm aiming for productivity on an industrial scale. Chemical modifications of standard tool steel alloys can be used furthermore to increase the productivity along the LPBF process chain. An increased sulfur content for the cold and hot work steels investigated improves the initial particle size distribution after the EIGA process without changing the shape or the porosity of the particles generated. In addition, no influence on the density could be detected during the processing of these modifications. Finally, the positive influence of a lower martensite start temperature could also be transferred to a hot work steel with limited weldability through a chemical modification. However, when adjusting elements with a low evaporation temperature compared to the other alloy elements, an adjustment of the LPBF process parameters must be considered, too.