Additive manufacturing (AM) represents a cutting-edge technology that potentially offers significant reductions in both manufacturing time and cost. Due to completely different principles of manufacturing, both the AM feed stock material and build process must be qualified, especially for application in the nuclear industry. The Procedure for the Acquisition of New Material Data (PTAN) of the RCC-MRx code for design and construction rules (AFCEN, 2017) and Section 3, Division 5 of ASME Boiler and Pressure Vessel Code (BPVC) (ASME, 2013) are two examples of procedures for the qualification of new materials. The qualification process includes defining a material property file, which, among others, includes physical and material properties from a wide range of tests and laboratories to establish material confidence limits in view of variation of properties. These can be due to material taken from a thick or thin plate, heat-to-heat variability, different cooling rates from AM melt, and later reheating by following passes of the heat source, etc.

The EU-funded research project NUclear COmponents Based on Additive Manufacturing (NUCOBAM) recently ended (2020–2024) (NUCOBAM, 2024). It examined the qualification of AM 316L stainless steel to demonstrate the compatibility of AM 316L in a light water reactor (LWR)-irradiated environment (Konstantinović et al., 2025). A more comprehensive program is currently being considered in ASME (Messner, 2023). In this work, laser powder bed fusion (LPBF) was used because no excessive ferrite phase builds up during rapid cooling.

The nuclear community (scientists, reactor designs, and regulators) is currently debating whether traditional material qualification tests, which assume that material properties derived from standard tests reduce uncertainty through extensive testing, are sufficient or if component-level tests are necessary (Messner, 2023; Torres and Gordon, 2021; Lalé and Viguier, 2024; ASM Handbook, 2017). In NUCOBAM, the process involved qualification of 316L feedstock powder first, followed by qualification of the AM build process and platforms, and qualification of the AM build material through a number of mechanical tests. Finally, a valve body and debris filter were manufactured for component testing. In this work, we focus on the recently standardized (EN, ASME, and RCC-MRx) small punch test (CEN EN, 2024; ASTM, 2024; RCC-MRx, 2022). This test uses small disks for test pieces (8 mm diameter, 0.5 ± 0.005 mm thickness), which is a crucial advantage when only a small amount of test material is available (Figure 1). Specifically, the small punch creep (SPC) tests were carried out to estimate creep properties, followed by validating these properties by performing a small number of standard uniaxial creep (UC) tests.

The article is structured as follows. We first describe the feedstock material, followed by a description of the methods used. Next, we discuss the specificities of the analysis of the SPC test results of AM 316L material and the transferability of these results to UC. Conclusions are given at the end.

A 316L stainless steel feedstock powder was used to manufacture bars for material tests using the laser powder bed fusion (LPBF) additive manufacturing process in accordance with the ASTM F3184 (ASTM F3184-16 Standard Specification for Additive Manufacturing Stainless Steel Alloy, 2024). The basis for the feedstock powder chemical composition requirements was taken from ASTM F3184, with additional requirements imposed for the weight % of N, O, Cu, Co, Ta, B, and Cu. The chemical composition of the feedstock powder is comparable to the 316L(N) forged plate but lower in the C (0.02 vs. 0.024) and Mn (0.8 vs 1.794) content and higher in Si (1 vs. 0.689). Table 1 provides the chemical composition, while the measured particle size distribution is given in Table 2.

The powder feedstock was then used to manufacture bars on a build platform (Figure 2). In this article, we use bars manufactured using a Renishaw AM250 single laser machine at the Advanced Manufacturing Centre (AMRC), University of Sheffield, and Commissariat à l'énergie atomique et aux énergies alternatives (CEA), using an SLM Solutions SLM280HL single laser, both using argon as an inert gas. The bars used in this work are in the as-built z-direction, protruding out of the build platform, as shown in Figure 2.

A heat treatment consisting of soaking at 1,066 ± 14 °C with a holding time of 1 h 15 min ±15 min was applied to the AMRC build platform, while for the CEA, the bars were cut from the build platform prior to application of the same thermal treatment. The holding time counter initiated when a lower temperature limit, 1052 °C = 1066 °C–14 °C, was attained. An argon atmosphere was used, with the prescribed heating rate of 4 °C/min and a cooling rate of at least 10 °C/min. We refer to this heat treatment as HT2 to distinguish it from other heat treatments applied within NUCOBAM.

In addition to HT2, some bars were aged for 10,000 h at 450 °C. The following combinations of materials were used in the current article: a) AMRC, HT2; b) AMRC, HT2 aged; and c) CEA, HT2, aged. To simplify labeling, these will be referred to as AMRC NA (not aged), AMRC Aged, and CEA Aged. These materials were then used to manufacture all small punch and uniaxial creep test pieces.

Four bars from the internal AMRC qualification process were used for three UC tests (at initial stress of 250 MPa, 270 MPa, and 290 MPa) due to an insufficient amount of originally shipped material. The heat treatment of these bars was performed at the JRC, following the same HT2 procedure.

Figure 6 shows the SPC initial deflections, u₀, defined as a deflection 5 s after the application of the full force load. These deflections are the result of an elastic–plastic response of a test piece. Because the calculation of the equivalent stress σ e q , Equation 1, depends strongly on the deflection, it is important to accurately capture these deflections. The initial deflections follow a linear trend relative to the load, with some scatter even within the same material. AMRC NA tests were performed first and have the largest scatter. Lessons learned in applying the test procedure reduced the scatter later on during AMRC Aged and CEA Aged tests. All the test pieces conformed to the very strict EN 10371:2021 thickness tolerances (0.5 ± 0.005 mm), and both the test piece temperature and the force load were strictly controlled. Some scatter in results within the same material is probably due to differences in the microstructure of test pieces. Because LPBF is used, the test pieces are basically welds, and local variations in material properties can be expected. Furthermore, surface roughness of the test pieces and the silicon nitride Si3N4 bearing ball, resulting in different friction effects between the tests, could be a contributing factor. The minor influencing factor could be the small variations in the silicon nitride Si₃N₄ bearing ball diameter (grade 5, ±0.0013 mm diameter tolerance) and load application rate, which we are not able to strictly control on our test machines.

Figures 7–14 show measured deflections during SPC tests at load rates from 500 N to 240 N (Jorge and Myriam, 2025). The symbols indicate the time at which the minimum deflection rate was estimated, using the procedure described above. In general, the aged material exhibits shorter creep life. In addition, the minimum deflection rate occurs earlier for the aged material; see Table 5. The exception is the AMRC NA FK-13 test at 260 N, where the minimum deflection rate occurs earlier than in both CEA Aged tests (GB-08, GB-09). In general, the CEA Aged creep lives are somewhat shorter than AMRC Aged lives at loads from 500 N to 300 N. At 260 N and 240 N loads, there is no significant difference between the two.

In Figure 9, we can see that all the deflection rate curves at 400 N force load exhibit two minima, with the first one having a significantly lower deflection rate than the second one. The two minima are even more apparent at decreased force load, 350 N, but are now almost evenly matched in terms of deflection rate. Calculation of the equivalent initial stress σ e q , Equation 1, according to the EN 10371:2021, depends on the identification of the minimum deflection rate. However, the standard assumes that only one minimum deflection rate is observed. At 350 N, it is unclear if we should select the first or the second minimum, especially because they are evenly matched in terms of deflection rate, and the estimation of the deflection rate itself is subject to the approach for dealing with the noise in the measured deflection signal and the derivative calculation method.

Another question that emerges is what causes the first minimum. In this work, we conjecture that a significant crack develops and manifests in the first minimum. It is well known that the region of maximal stress/strain in small punch tests continuously moves radially outwards during the test. We therefore conjecture that if the material is ductile enough, the crack that results in the first minimum in the deflection rate curve arrests as the zone of maximal stress/strain redistributes radially outwards. At a certain point, a second crack initializes and starts to propagate, resulting in the second minimum, shortly before the rupture of the test piece. At the 240 N load, one can see a significant local increase in the deflection at approximately 0.9 mm for the FK-10 AMRC NA and FV-08 AMRC tests (the orange and green curves in Figure 14). We assume these sudden deflection increases are due to a final/secondary crack initialization. To evaluate these conjectures, we performed an interrupted SPC test with AMRC NA material. Because no more test pieces conformant with the strict 0.5 ± 0.005 mm requirement of the EN 10371:2021 were available, we used a test piece with a thickness of 0.4661 mm, labeled FI-07. The EN 10371:2021 standard provides a formula for estimating the ultimate tensile stress, R m , from the small punch tensile test (Equation 2). In the formula, the ultimate tensile stress is linearly related to the initial thickness of the test piece, h 0 . β R m stands for the correlation factor, F m is the maximum force, and u m is the deflection at maximum force. Assuming that u m is approximately the same, we can calculate the force for the thinner test piece that would be equivalent to the 0.5-mm test piece (Equation 3). Recognizing that this is a very crude assumption, we calculate the force to be applied as 326 N = 350 N × 0.4661 mm/0.5 mm. R m = β R m · F m / h 0 · u m (2) F m   t h i n = F m · h 0   t h i n h 0 (3)

Based on the SPC tests at 350 N, we decided to interrupt the SPC test for the first time 12 h after application of the force load of 326 N. We defined the interruption as stopping the furnace heating by setting the target furnace temperature to 22 °C, triggering the cool down. During the cool-down, the force load remained at 326 N. Following the cool-down, we removed the dead weights, removed the test piece from the test machine, and performed a scanning electron microscope (SEM) analysis of the test piece. Two crack fronts are visible, as shown in Figures 15, 16. The first crack front with a diameter of 1.28 mm is fully developed after 12 h. The second front, with a diameter of 1.46 mm, has only started developing.

Next, we put the FI-07 test piece back into the test machine, added the same dead weights to reapply a 326 N force load, and heated the test piece back to 650 °C. After reaching the target temperature, we waited for 16 h and 48 min (cumulative time 12 h + 16 h + 48 min) and then initiated the second interruption, using the same procedure as for the first interruption. A second SEM analysis was performed (see the lower part of Figure 16). The analysis showed the arrest of the first crack front and full development of the second, much larger crack front that only started forming after the first interruption. We did not continue the SPC tests beyond this point, 16 h and 48 min (cumulative time 12 h + 16 h + 48 min). The deflection versus time curve and the calculated deflection rates indeed show two minima (Figure 17), although these minima are less pronounced than the ones at 350 N (Figure 10). Nevertheless, this confirms our conjecture that cracks formed during the SPC test. More interrupted tests are required to confirm that crack initiation results in a local minimum in the deflection rate.

Figure 18 shows the time to rupture versus the applied force for the SPC test data for the three materials and the regression lines. This confirms again that the aged materials have consistently reduced creep life compared to the non-aged material.

Nuclear design codes, and in particular evaluation of material properties, are primarily based on standard uniaxial tests. Uniaxial tests were conducted to assess the SPC test data and to assess the transferability. To this end, we calculated the equivalent stress σ e q , Equation 1, for all the SPC tests, and compared σ e q versus rupture time to the applied initial stress of UC tests σ i n i t versus rupture time, Figure 19. One UC test was performed for each material at 150 MPa, 170 MPa, and 190 MPa initial stress σ i n i t . For the AMRC NA, we also performed UC tests at 210 MPa, 250 MPa, 270 MPa, and 290 MPa initial stress σ i n i t . Table 6 gives an overview of the UC test results.

We can observe that UC tests for the three materials overlap perfectly and are distributed along the regression line with an R² = 0.9938, Figure 19. In contrast to SPC tests, there is practically no difference between the three materials in UC tests. Furthermore, the scatter in the UC results between three different materials is much smaller than in the SPC results. We can also see that the transferability of SPC results to UC, by evaluating σ e q from the SPC tests, is relatively good. Equation 1 from the EN 10371:2021 standard works quite well. One can therefore estimate the equivalent UC initial stress load condition from the SCP tests. However, the evaluated σ e q slightly overestimates the stress below t_r = 50 h and underpredicts it above that value due to different slopes of the SPC and UC trendlines.

Several explanations are possible for the SPC trendline having a different slope than the UC trendline. First, Equation 1 is based on one minimum in the SPC deflection rate, whereas here, we can have two minima, as shown in Figure 10. In this work, we always select the one with the lowest deflection rate value to use in Equation 1. However, as shown in Figures 15–17, the first minimum is probably the one where the initial crack has already formed. If the second minimum is used, the measured deflection is not solely due to creep but is significantly affected by the crack propagation. Second, the orientation of the SPC test pieces in relation to the uniaxial test piece is the one shown in red in Figure 1. This was driven by the limited amount of material, as more disc slices for SPC test pieces can be extracted from a bar in the red orientation than in the gray SPC orientation. In the SPC test, most of the stress acts in the radial direction of the SPC test piece. With the red SPC test piece orientation, the red SPC test piece has stress perpendicular to the stress in the uniaxial test piece. We are therefore testing a different material direction in the SPC tests. In fact, we should have used the gray SPC test piece orientation, but as mentioned, this was not possible due to the limited amount of available material. The SPC test piece orientation in relation to the test build can have a significant impact on the time to rupture in LPBF. For Inconel 718, Zhang et al. (2025) report significantly shorter rupture times for the SPC test piece in the build direction (red color in Figure 1) compared to the one perpendicular to the build direction (gray color in Figure 1). In this case, only aged heat treatment was applied. However, when they applied the solution and aged heat treatments, the differences in rupture times decreased significantly. This behavior is at least in part driven by a different microstructure (Zhang et al., 2025) of the two directions. Third, and probably the most important, Equation 1 has been developed based on 97 UC tests and 159 SPC tests with both new and service-exposed materials, mainly low-alloy steels and 9Cr steels, such as 14MoV63, X20CrMoV121, P91, P92, Eurofer97, and stainless steel 316L (CEN EN, 2024). However, this data set had a large share of P92 steel. Equation 1 coefficients 1.916 and 0.6579 are probably not the best for AM 316L material.

In this article, we provide a significant set of small punch creep (SPC) and uniaxial creep (UC) data of additively manufactured (AM) 316L stainless steel at 650 °C. We show that SPC tests of this material result in double minima for deflection rates at relatively short rupture times (tens of hours). For a specific case, we show that these two minima are related to crack initiation and propagation. The measured SPC deflection is then not only due to the creep of the test piece but also due to its cracking. This could potentially be an issue for SPC testing of non-ductile materials. Further work is necessary to confirm additionally that the multiple crack initiations in SPC tests result in multiple deflection rate minima, to determine which minimum should be used for estimating σ e q , and/or whether a completely different method should be used. We demonstrate the transferability of SPC results to UC results by calculating that the σ e q works quite well for the AM 316L in this work. However, the σ e q underestimates the stress for the shorter rupture times and overestimates the stress for longer rupture times. This is probably due to the EN 10371:2021 standard assuming a single minimum in the SPC deflection rate, while we show here that two minima can occur. It is clear, however, that the EN 10371:2021 standard must be re-evaluated to accommodate this complexity.