Use of the small punch test for the estimation of ductile-to-brittle transition temperature shift of irradiated steels

The small punch test is evaluated as a screening procedure for irradiation embrittlement of reactor pressure vessel steels. In particular, the correlation between ductile-to-brittle transition temperatures obtained from this small specimen test technique and from the standard Charpy impact test is investigated. Small punch tests and Charpy impact tests at different temperatures were performed on various steels including materials from original reactor pressure vessels in the unirradiated, neutron irradiated and annealed condition. It is demonstrated that the small punch test is a reliable and effective supportive means for the estimation of the irradiation-induced shift of the ductile-to-brittle transition temperature. It was found that a tanh-fit of the normalized small punch energy in dependence of temperature is preferable in comparison to the two-curve fit of the total small punch energy. paper aims at evaluating a data set of transition temperatures based on the SP test with respect to their correlation with Charpy-based transition temperatures. Two different procedures for the analysis of SP


Introduction
The small punch (SP) test has been established as a small specimen test technology to support development and monitoring of structural materials [1][2][3][4][5][6]. It yields estimations of mechanical properties with small amounts of material and is therefore an appropriate means for screening.
For nuclear components, the SP test provides the possibility for an evaluation of ageing mechanisms, namely radiation-induced hardening and embrittlement and thermal ageing. Major benefits arise from a lower total activity of a SP sample (approximately 1:220 as compared to a Charpy-size specimen) and from the possibility of re-using already tested standard specimens. Indeed, the Slovakian utility VUJE recently included the SP test into their monitoring concept of the nuclear power plant (NPP) Bohunice [39,40]. The shift of the DBTT is a key element of reactor pressure vessel (RPV) embrittlement surveillance [41,42]. It is therefore interesting to evaluate whether the DBTT shift can be reliably measured by means of the SP test.
As for the SP-test-based estimation of the DBTT, T SP , the following simple correlation has long been proposed [1,5]: where T CVN is the DBTT from the Charpy impact test using standard specimens 10 × 10 × 55 mm with a 2 mm V-notch (so-called CVN specimens). The factor of proportionality α depends on SP geometry, the material class, the fitting procedure for T SP , and the criterion for the definition of T CVN (e.g. T 41J ,T 47J , T 68J , T 50%US ). The factor of proportionality is significantly smaller than one, which is a consequence of the lower constraint in the SP test. Values of α = 0.32 … 0.45 are reported in the literature [4,5,9]. This paper aims at evaluating a data set of transition temperatures based on the SP test with respect to their correlation with Charpy-based transition temperatures. Two different procedures for the analysis of SP data are applied and compared with one another. Potential benefits for embrittlement monitoring of RPV steels are discussed.

Materials
A number of seven different steels were investigated. These materials include bainitic reactor pressure vessel (RPV) steels, base metals (BM) as well as weld metals (WM), and the ferritic-martensitic 9Cr steel P91. The latter material could be included as it is comparable to RPV steels in terms of tensile properties and toughness. Moreover, the experimental program includes different irradiation conditions, namely unirradiated (u), neutron irradiated (i) and neutron irradiated and annealed (ia). The detailed list of materials is given in Table 1 and the chemical compositions in Table 2.
The material 10KhMFT originates from the original RPV of the nuclear power plant Greifswald, unit 4, operated from 1979 to 1990. Within the post mortem investigation program, material was extracted from the vessel wall and extensively characterized [43,47,51]. The extracted trepans were cut into layers and allowed a through-thickness analysis of the microstructure and the material properties. The different layers of the trepans represent various irradiation conditions as the neutron fluence changes from 4.4 × 10 19 n/cm 2 (E > 0.5 MeV) at the inner side of the wall to 0.9 × 10 19 n/cm 2 at the outer side (irradiation temperature 270 • C) [52,53]. Moreover, as a consequence of the welding technology, there are significant through-thickness variations of the chemical composition and of the structure [47]. This means that the different layers can be considered as "different materials" for the purpose of this paper. A subset of the irradiated samples was annealed for 100 h at 475 • C (condition ia). This treatment was proven to induce a recovery of the mechanical properties comparable to the unirradiated condition [51,54]. Table 3 gives an overview of the different layers included in this investigation (e.g. 4-6-02 means unit 4, trepan no. 6, layer 02).
The material 15Kh2MFAA originates from the original RPV of the NPP Greifswald, unit 8, which was not put into operation, i.e. the material is unirradiated. The RPV was produced by SKODA steelworks. For decommissioning, the vessel was cut into large segments of 0.4 × 1 m [45]. The Chary-V and SPT samples were fabricated from the middle thickness layers of the forged ring 031.
The materials 22NiMoCr3.7 and SST38 were extracted from the original RPV of the NPP Biblis C, which was also never put into operation. Pieces of the RPV wall were delivered to HZDR by ENSI (the Swiss Federal Nuclear Safety Inspectorate) [47].
JFL and JRQ are so-called IAEA reference steels produced by Kawasaki Steel Corporation. These steels were used in numerous international projects and benchmarks. JFL is a low-Cu steel comparable to A508 Cl. 3, while JRQ is similar to A 533B Cl. 1, but with intentionally increased Cu content. In JRQ, the structure varies with thickness location due to the manufacturing process. Lower bainite und tempered martensite was found in the near surface regions, while granular bainite dominates in the centre [48]. Therefore, specimens from two respective layers are investigated (cf. Table 1).
P91 is a ferritic-martensitic Cr-steel for high temperature applications. It is widely used in thermal power plants and belongs to the candidate materials for Gen-IV and fusion reactor components. The P91 samples were manufactured from a hot-rolled and tempered pipe. There was no significant local variation of hardness and structure [50].

Charpy impact tests
Standard Charpy-V specimens (10 × 10 × 55 mm, V-notch 2 mm) were tested according to ISO 148-1. An instrumented TIRA WPM PSd300 machine was used (energy 300 J, impact velocity 5.5 m/s, hammer mass 20 kg). The control of test temperature was realized by a combination of liquid N 2 and electrical heating. The test temperature range was from − 25 … 225 • C for the irradiated steels and − 150 … 150 • C for the unirradiated steels. The number of tests per series was N ≥ 10 (irradiated steels) and N ≥ 15 (unirradiated steels). The WWER-440 materials (10KhMFT, 15Kh2MFAA) were tested in orientation T-S (according to the Russian standard PNAE G-7-008-86), the other materials in orientation T-L (Lrolling direction, Ttransverse direction, Sthickness direction). As RPV steels do not exhibit significant structural anisotropy, the orientation does not affect the DBTT [55].

Small punch tests
SP specimens of 10 × 10 × 0.5 mm were manufactured from tested or unused Charpy-sized SE(B) specimens. The orientation was T for all materials (normal direction of the SP disc). In case of the unirradiated materials (cf. Table 1), slices of 10 × 10 × 0.6 mm were cut by electrical discharge machining (EDM) and subsequently ground to final thickness of 0.5 ± 0.005 mm with grit 2500. In case of the irradiated materials, grinding was not possible, therefore the surface finish was realized by two EDM post cuts. The final thickness of irradiated SPT samples was 0.54 ± 0.005 mm.
The main parameters of the SP set-up are: punch diameter d = 2.5 mm, receiving hole diameter D = 4 mm, receiving hole edge radius R E = 0.5 mm (cf. Fig. 1). The edge size is larger than proposed in the upcoming standard [38]. While the effect of the edge size on the estimation of tensile properties (in particular the yield stress) is significant, it can be  [50] neglected for the estimation of the ductile-to-brittle transition temperature [9]. The punch displacement v was measured by an inductive sensor with an accuracy of ±1 μm and corrected for the device compliance. The punch force was measured by means of a load cell placed between the puncher and the cross head of the testing machine with an accuracy of ±5 N. At least a number of 25 samples were tested per material and condition. In total, >500 SP tests were performed. The test temperatures were in the range from − 172 … + 100 • C for the irradiated materials and − 194 … + 100 • C for the unirradiated materials. For each test the force-displacement curve F(v) was recorded and the small punch energy E SP was calculated as integral of F(v) up to maximum force F m . As an example, Fig. 2 shows F(v) curves of material 10KhMFT at different test temperatures.

Determination of ductile-to brittle transition temperatures
The DBTT of a Charpy impact test series is determined by means of a tanh-fit of the impact energy as function of test temperature T.
A, B, C and T 0 are the fitting parameters, E US and E LS are the upper and lower shelf energies. The energy E(T) can either be the absorbed impact energy KV according to ISO 148− 1 or the total impact energy W t * average over the filling layers beyond the welding root.
A non-linear iterative least square procedure is used to identify the parameters A, B, C and T 0 . A statistical error estimation for the fit parameters is included [56]. The procedure is implemented in a software package developed at HZDR. Different constraint options are available for the fit, two of them were applied to our data: • no constraints • pre-defined lower shelf energy The standard option is the fit without constraints. The second option can be useful if (i) the standard fit provides a negative lower shelf energy (E LS < 0), which is physically meaningless, or if (ii) lower shelf data are missing.
Two fitting methods are available for the determination of the SPtest-based DBTT. The first one is also based on Equation (2). The shift parameter T 0 is directly taken as SP-DBTT (T SP = T 0 ). The normalized energy E n = E SP /F m is used for fitting. i.e. E(T) corresponds to E n [57]. E SP is the area under the force-displacement curve up to the displacement v m at maximum force F m [37]. From the definition of E n and from the shape of the force-displacement curve, it can be concluded that v m ≤ 2E n holds. Note that the engineering unit of E n is mJ/N = mm. The above mentioned fitting constraint (pre-defined lower shelf energy) is especially useful for SP tests of very tough materials where T SP is close to liquid nitrogen temperature or even lower [9].
The second approach for the determination of SP-based DBTT relies on a two-curve exponential fitting of E SP (T) [2,9]: where the index "b" refers to the brittle region and the index "d" to the ductile region. A b , B b , C b , A d , B d , and C d are the fitting parameters (A b , . The intersection of the two curves E SP,b (T) and E SP,d (T) marks the maximum of the fitted E SP (T) dependency, E max , while the minimum energy is given by E min = E SP (0 K). The SP transition temperature T SP2 is defined as temperature where E SP,b (T SP2 ) = 0.5× (E min + E max ) holds. The index "2" was added to discriminate T SP2 (twocurve-fit, Eq. (4)) from T SP (tanh-fit, Eq. (2)).
As an example, Fig. 3 shows the tanh-fit of E n (T) without constraints for material 10KhMFT, layer 4-6-02. The fitting parameters are: A = 0.634 mm, B = 0.366 mm, C = 40.7 K, T SP =− 129.5 • C. Fig. 4 shows the two-curve-fit of E SP (T) based on the same data set. The data points were assigned to either E SP,b or E SP,d according to the brittle or ductile fracture appearance of the associated F(v) curve [9] cf. Fig. 2. In Fig. 4, the blue points are associated with E SP,b and the red points with E SP,d (fitting parameters: A b = 0.399 mJ, B b = 121.3 mJ, C b = 0.0157 K − 1 , A d = 1395 mJ, B d = 5594 mJ, C d = − 0.00808 K − 1 , temperature in K). Fig. 4 shows that the small punch energy in the ductile range (E SP , d ) decreases with increasing temperature. This is a consequence of the decreasing yield stress of the material. The normalization leads to an elimination or significant reduction of the decreasing behaviour of the upper shelf since the maximum force is also decreasing with temperature. This makes the tanh-fit applicable to E n (T).

Results
The SP-based and Charpy-based DBTTs are listed in Table 4. For material 10KhMFT, the layers from trepans 4-6 and 4-4 which have a similar thickness position (cf. Table 3) are combined in one row.
The statistical error analysis yielded uncertainties of (standard deviation). The DBTTs from Table 4 Table 5. The tanh-fits of E n (T) were done without constraints (cf. Section 2.4). If the resulting lower shelf energy was outside the range 0.1 mm ≤ E LS ≤ 0.3 mm, the fit was repeated with a pre-defined lower shelf E LS = 0.25 mm. Within the regression analysis it was also checked whether the forced intercept of the T SP − T 47J correlation at 0 K according to Eq. (1) is justified. The regression with 2 parameters yielded T SP = 0.41⋅T 47J + 3.5K, hence the absolute offset is negligible.
The obtained correlations were used to evaluate the DBTT of the core weld of the RPV Greifswald unit 4. Fig. 7 shows T 47J as a function of the

Discussion
The correlations between T SP and T 47J (T 41J , T 0 ) work surprisingly well (Fig. 5) while the correlations between T SP2 and T 47J (T 41J , T 0 ) exhibit a larger scatter (Fig. 6). This can also be seen in the corresponding R 2 values. If we consider the CVN based DBTTs as the "true" values and the transformed SP-based values as their predictions, we can define an error as follows: The errors e SP,41J , e SP,0 , e SP2,47J , e SP2,41J , e SP2,0 are defined analogously. The corresponding α values are listed in Table 5. The maximum and average values of these errors are given in Table 6. The maximum value for the T SP − T 47J correlation is e SP,47J = 16 K. In Fig. 5, this corresponds to the horizontal distance of the green coloured data point from the regression line (material 15Kh2MFAA, Greifswald 8). This maximum error is small as compared to the usually expected irradiationinduced DBTT shift in RPV steels. The errors for the correlations based on the two-curve fit (T SP2 ) are significantly higher. In particular the materials with low DBTTs produce this scatter. Obviously, in case of missing lower shelf data, the twocurve-fit, Eq. (4), is less robust in comparison to the tanh-fit, Eq. (2).
The values obtained for the factor of proportionality α range from 0.39 up to 0.44 (cf. Table 5). This is in accordance with values reported in the literature [4,5,9]. Two main factors affect the α-values: the type of Charpy-DBTT (T 41J , T 47J , T 0 ) to which the SP-DBTT is correlated, and the fitting method (T SP vs. T SP2 ).
The application of the T SP − T 47J correlation to the core weld of the RPV of NPP Greifswald 4 demonstrates the usability of the SP test for the evaluation of irradiation induced embrittlement. For the as-irradiated condition, Fig. 7 indicates a characteristic through-thickness  dependence of T 47J observed in the Charpy impact test, which exhibits a pronounced minimum close to the welding root. This is nicely reproduced with the SP-based T SP (Fig. 7) for the as-irradiated condition. A similar through-thickness dependence was reported for the fracture toughness (expressed by the reference temperature of the master curve concept) [43,44]. An explanation will be provided below. The recovery of mechanical properties due to the annealing treatment is clearly indicated by significantly lower DBTTs (Fig. 7). Again, the Charpy and SP based values of T 47J give a coherent picture. Moreover, the SP test provides additional information for the layers for which Charpy tests are not available since the material from the trepans was used up.
In the annealed condition, there is no significant difference between the DBTTs for the welding root (around 35 mm distance from the inner surface, see Fig. 7) and the filling layers (distances beyond 50 mm). This is in contrast to the as-irradiated condition. As the annealed condition can be considered as an equivalent for the unirradiated condition, we can conclude that the particular through-thickness dependence for the as-irradiated DBTTs is caused by irradiation. In particular, the irradiation-induced DBTT shift in the welding root (~52 K) is significantly lower as compared to the filling layers (~106 K). At a first glance, this behaviour seems to be unexpected as the neutron fluence is decreasing from the inner surface to the outer surface (Section 2.1). Nevertheless, this finding is in line with the through-thickness variation of the Cu content. Cu is known to be the most important impurity element in RPV steels for their susceptibility to irradiation embrittlement. Indeed, the Cu content was reported to be lowest in the welding root [43].
It should be emphasized that this conclusion was possible through the re-use of tested Charpy-sized specimens by means of SP testing. There was no more material left for a complete Charpy impact testing of all layers in the annealed condition. This aspect of the re-use of tested standard samples could also be of general importance for extended RPV surveillance programmes under long term operation. The use of small specimen technologies is a promising option to overcome the lack of surveillance materials [39,40,58,59].

Conclusions
It was demonstrated that the SP test can be used as a supportive technique to monitor irradiation embrittlement of RPV steels. The DBTT shift can reliably be estimated based on the correlation between SPbased and Charpy-based transition temperatures (Eq. (1)). The tanh-fit of the normalized SP energy E n (T) provides a better correlation as compared to the two-curve fit of the total SP energy E SP (T). The SP test is an option for the re-use of tested irradiated standard specimens.
CRediT authorship contribution statement E. Altstadt: Conceptualization, Methodology, Investigation, Table 4 SP and Charpy based transition temperatures.   Correlation between E SP based T SP2 (two-curve-fit) and T 47J .

Table 5
Factor of proportionality α and coefficient of determination for the correlation between Charpy-based and SP-based DBTTs Eq. (1).
T SP tanh-fit of E n (T) T SP2 two-curve-fit of E SP (T)  1)) as a function of position in the core welding seam of the RPV of Greifswald unit 4 (material 10KhMFT); as irradiated (irr) and annealed (ia) conditions; the background picture shows the structure of the weld (at the left end the austenitic overlay cladding is visible).

Table 6
Maximum and average prediction errors (K) over all data points used for the correlation analysis (cf.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.