Relation between the Fatigue and Fracture Ductile-Brittle Transition in S500 Welded Steel Joints

: The formation and propagation of cracks occur through irreversible dislocation movements at notches, material defects, and grain boundaries. Since this process is partly thermally controlled, the resistance to dislocation movements at low temperatures increases. This slows both fatigue initiation and fatigue crack propagation. From recent experimental data, it can be seen that fatigue crack growth is accelerated below the fatigue transition temperature (FTT) that correlates with the ductile-brittle transition temperature (DBTT) found by well-known fracture mechanics tests, i.e., Charpy impact, fracture toughness, and CTOD. Hence, this study investigates the relation between FTT and DBTT in S500 high-strength steel base material and welded joints at low temperatures using fatigue crack growth, fracture toughness tests as well as scanning electron microscopy. From the tests, an almost constant decrease in fatigue crack propagation rate is determined with decreasing test temperature even below the DBTT. At − 100 ◦ C, the fatigue crack propagation rate is about half of the rate observed at room temperature for both base material and weld metal.


Introduction
Low temperatures cause challenging conditions for the structural integrity of steel structures in Arctic regions [1]. In order to ensure the safety of these structures against brittle failure, the structural design has to account for the static and dynamic structural responses that are often more severe than for other engineering structures [2][3][4][5][6]. In particular, the combination of high ice-related loads and low temperatures is difficult to account for in design [7][8][9][10][11][12][13][14][15][16][17][18]. This problem is amplified by the lack of comprehensive guidance for lowtemperature fatigue strength assessment of steel structures. A major knowledge gap in this regard is the relation between the fatigue and fracture ductile-brittle transition in welded steel joints.
With decreasing temperature, ferritic structural steels may undergo a ductile-to-brittle transition (DBT) behavior, which is characteristic of steels with a body-centered cubic (bcc) crystal structure. At temperatures below the ductile-to-brittle transition temperatures (DBTT), the mechanism of stable crack extension behavior changes from plastic blunting and tearing to cleavage-controlled brittle fracture. Temperatures below the fatigue transition temperatures (FTT) may result in the fatigue behavior showing an acceleration of crack propagation rates; however, the relation between static and fatigue transition temperature is currently purely empirical, see [19][20][21][22]. Thus, a better understanding of the effect of 2 of 19 temperature on fatigue damage mechanisms (fatigue crack initiation and propagation) is required to avoid detrimental fatigue performance at low temperatures and to use the beneficial effect of improved fatigue strength above the FTT, see Braun and Ehlers [23].
Alvaro et al. [24] provide a first extensive literature review on the effect of low temperature on the different aspects of the fatigue life of steels and their weldments; however, the discussion is mainly limited to fatigue crack initiation and propagation in base materials, due to the absence of literature on welded joints.
In two recent studies [19,25], Alvaro et al. extended their findings to thermally weld simulated materials (Gleeble machine tests); however, we are unaware of any comprehensive assessment of the relation between the fatigue and fracture ductile-brittle transition in actual welded steel joints. Hence, the current study aims at describing the relation of FTT and DBTT by means of (i) fatigue crack growth (FCG) rate tests, and (ii) Charpy impact and fracture toughness tests of a high-strength S500 structural steel and welded joints. To this goal, FCG rate tests were performed using single test specimens under constant crack tip loading conditions by successively reducing the test temperature. The underlying idea is that this allows minimizing specimen and setup-related uncertainties (due to weldingrelated deformations, defects, and microstructure). In particular, the microstructure is known to highly affect fatigue and fracture transition behavior, see [26][27][28][29][30][31]. For reference, FCG curves were determined for both the base material and the welded joints. Finally, the results are assessed by fracture surface investigations using scanning electron microscopy.
Knowledge about the actual material behavior under static fracture and cyclic fatigue behavior of welded joints at sub-zero temperatures is of major importance to ensure the structural integrity of engineering structures operating in Arctic regions but also for other applications such as aerospace or general cryogenic environments. Hence, this study seeks to help to gain knowledge on the relation between fatigue and fracture transition behavior for actual welded joints.

Specimen's Design
In the experiments, middle tension M(t) specimens were used to determine FCG rates for the base material and welded joints made of S500 high-strength steel. For the first type of specimen, the crack starter notch was placed in the homogeneous base material. The crack starter notch is placed in the middle of a welded joint for the second type, as shown in Figure 1. The specimens were manufactured from S500G1+M steel, see Braun [32] for basic material properties, chemical composition, and welding parameters. The butt-welded joints were fabricated by flux-cored arc welding. This steel type is known to be susceptible to cold cracking in underwater welding [33], but otherwise easily weldable. The chemical composition and the mechanical properties of the steel are taken from Braun et al. [34] and listed in Tables 1 and 2. The filler material was EN ISO 17632-A: T 50 6 1Ni P M21 1 H5.  Table 2. Mechanical properties of the S500G1+M steel [34]. The basic dimensions of the specimens are shown in Figure 1. In the gauge section, the samples were tapered by wire erosion to a width of 58 mm. The crack starter notch was placed in the middle of the specimens. The notch was also wire eroded and manufactured as a 15 mm long slot with runouts on both sides with an angle of 30°. In addition, the specimens made from the welded joint were milled in the gauge section to a constant thickness of 9.5 mm to remove the weld overfill and heat-treated to release welding residual stresses.

Preparation of the Test Specimens
First, to determine the initial crack length, crack detection gauges were glued to the specimen surface two millimeters from the tip of the notch. Second, copper-coated bolts were welded to the surface of the specimen to attach cables for current input and measurements. The arrangement was planned according to the standards of ASTM 647-15 [35]. The two bolts farthest away from the notch were used for current input (Y1), and another two bolts (YR.2, YR2.2) were placed on both sides of the notch to detect fluctuations in the power supply (VR). The two last bolts (Y0) were placed on opposite positions of the notch to determine the crack length by measuring the current potential drop over the growing crack (VM). The bolts were numbered from 1 to 6. The positions of the bolts are shown in Figure 2 and Table 3. In addition, a photograph of the specimen surface of a welded joint specimen is presented in Figure 3. The basic dimensions of the specimens are shown in Figure 1. In the gauge section, the samples were tapered by wire erosion to a width of 58 mm. The crack starter notch was placed in the middle of the specimens. The notch was also wire eroded and manufactured as a 15 mm long slot with runouts on both sides with an angle of 30 • . In addition, the specimens made from the welded joint were milled in the gauge section to a constant thickness of 9.5 mm to remove the weld overfill and heat-treated to release welding residual stresses.

Preparation of the Test Specimens
First, to determine the initial crack length, crack detection gauges were glued to the specimen surface two millimeters from the tip of the notch. Second, copper-coated bolts were welded to the surface of the specimen to attach cables for current input and measurements. The arrangement was planned according to the standards of ASTM 647-15 [35]. The two bolts farthest away from the notch were used for current input (Y 1 ), and another two bolts (Y R.2 , Y R2.2 ) were placed on both sides of the notch to detect fluctuations in the power supply (V R ). The two last bolts (Y 0 ) were placed on opposite positions of the notch to determine the crack length by measuring the current potential drop over the growing crack (V M ). The bolts were numbered from 1 to 6. The positions of the bolts are shown in Figure 2 and Table 3. In addition, a photograph of the specimen surface of a welded joint specimen is presented in Figure 3.   Figure 2. Positions of the welded bolts for current input and measurements.  Figure 2. Positions of the welded bolts for current input and measurements.  3. Photograph of one of the welded specimens with a machined surface, heat tint from heat treatment, the two crack detection gauges, and the copper-coated bolts for current mea ments.

Experimental Setup
The experiments were carried out in a fatigue test rig (Carl Schenck AG, Darms Germany) with a climate chamber to regulate the temperature. The specimens w clamped inside the resonant testing machine. For electrical insulation, resin plates w placed between the sample and hydraulic clamps, and the chamber was cooled by va izing nitrogen. An Omron temperature controller (Omron K.K.; Kyoto, Japan) regul the valve and ensured the selected temperature in the chamber. The temperature in the chamber, on the specimens, and at the load cell were measured with PT100 sen The values were constantly recorded and controlled. The diameter of the cables for cu input and measurements were chosen according to ASTM 647-15 [35] and insulated protection against temperature fluctuations. As a power source, a stabilized labora power supply unit (EA Elektro Automatik; Viersen, Germany) was used. Figure 3. Photograph of one of the welded specimens with a machined surface, heat tint from the heat treatment, the two crack detection gauges, and the copper-coated bolts for current measurements.

Experimental Setup
The experiments were carried out in a fatigue test rig (Carl Schenck AG, Darmstadt, Germany) with a climate chamber to regulate the temperature. The specimens were clamped inside the resonant testing machine. For electrical insulation, resin plates were placed between the sample and hydraulic clamps, and the chamber was cooled by vaporizing nitrogen. An Omron temperature controller (Omron K.K.; Kyoto, Japan) regulated the valve and ensured the selected temperature in the chamber. The temperature inside the chamber, on the specimens, and at the load cell were measured with PT100 sensors. The values were constantly recorded and controlled. The diameter of the cables for current input and measurements were chosen according to ASTM 647-15 [35] and insulated for protection against temperature fluctuations. As a power source, a stabilized laboratory power supply unit (EA Elektro Automatik; Viersen, Germany) was used.
A Peekel Autolog 3000 measurement amplifier (Peekel Instruments GmbH; Bochum Germany) with a maximum measurement frequency of 1 kHz was used for measurements. To verify the crack length on the surface, crack detection gauges with 0.1 mm spacing were used.

Johnson's Equation
The crack length change during the experiment was determined using Johnson's equation [36]: Herein, the crack length a depends on the specimen width W, the actual measured current V and the measured current V 0 at the reference crack length a 0 . The distance of the measuring pin is represented by Y 0 .

Stress Intensity Factor Range
The stress intensity factor range ∆K in the M(t) specimens is calculated according to ASTM 647-15 [35]: with force range ∆P, and form factor α and thickness B. ∆P is defined by the maximum and minimum force per load cycle, see Equation (3). The definition of α is presented in Equation (4). According to ASTM 647-15 [35], the error of Equation (2) is less than 1% for 2a/W ≤ 0.8.

Calibration and Pre-Tests
Before the actual experiments, the precision of the direct current potential drop method was tested. For this reason, calibration curves were created at room temperature, −20 • C, −50 • C, and −80 • C for both materials (base and weld material) with reference specimens in increasing ∆K tests. For correlation of the results, the crack length of the Johnson equation was compared with beach marks, which were created during the test, see Figure 4. In addition, the crack detection gauges-used for the start crack length of the Johnson equation-were calibrated based on the information gained from these experiments. The correction thus accounts for crack front curvature. The influence of the temperature on the load cell was also corrected during the experiments.

Reference Decreasing ΔK Experiments
Decreasing ΔK experiments were performed at room temperature with one specimen per material as a reference for the tests with decreasing test temperature according to ASTM 647-15 [35]. First, the experiments were started from a specific ΔK. Second, after reaching a defined crack length, the experiments were stopped, and the load ranges were reduced before the experiments were continued. Third, this procedure was repeated until no crack growth was observed. This value is then assumed to represent the threshold stress intensity factor range ΔKth.

Experiments with Decreasing Test Temperature
The experiments with decreasing temperatures were carried out with one specimen for each material (base and weld material). The advantage of this method is that just one specimen is required, and specimen-related influences are mitigated. In order to find a representative SIF range for comparison, a value slightly above the threshold SIF was chosen.
Both specimens were then tested at a constant SIF range and at a stress ratio R = 0.5 while gradually reducing the temperature. Due to the changing crack length, the applied loading was constantly adjusted to yield a constant stress intensity factor range of 8 MPa m. For the determination of an SIF range for the experiments in this study, decreasing ∆K tests according to ASTM 647-15 [35] were performed at room temperature. A ∆K range of 8 MPa √ m a little above the threshold value ∆K th was then chosen for the experiments with decreasing temperature. The threshold value is the ∆K for which technically no crack growth occurs anymore.

Reference Decreasing ∆K Experiments
Decreasing ∆K experiments were performed at room temperature with one specimen per material as a reference for the tests with decreasing test temperature according to ASTM 647-15 [35]. First, the experiments were started from a specific ∆K. Second, after reaching a defined crack length, the experiments were stopped, and the load ranges were reduced before the experiments were continued. Third, this procedure was repeated until no crack growth was observed. This value is then assumed to represent the threshold stress intensity factor range ∆K th .

Experiments with Decreasing Test Temperature
The experiments with decreasing temperatures were carried out with one specimen for each material (base and weld material). The advantage of this method is that just one specimen is required, and specimen-related influences are mitigated. In order to find a representative SIF range for comparison, a value slightly above the threshold SIF was chosen.
Both specimens were then tested at a constant SIF range and at a stress ratio R = 0.5 while gradually reducing the temperature. Due to the changing crack length, the applied loading was constantly adjusted to yield a constant stress intensity factor range of 8 MPa √ m. The procedure of the experiment can be summarized in seven steps: The specimen was clamped into the resonant testing machine.

2.
To avoid the notch shape influence, the resonant testing machine ran until the crack grew to an initial depth of 2 mm on each side of the notch. Upon reaching the crack detection gauges, the machine was stopped. 3.
The current crack length was taken as the reference crack length a 0 for Johnson's equation. The upper and lower applied load was adjusted to equal an SIF to 8 MPa √ m at a stress ratio of 0.5.

4.
The machine was then restarted and stopped once the crack had grown by 0.5 mm as proposed in ASTM 647-15 [35]. 5.
The stress ratio was changed to R = 0.75 to create the beach marks. 6.
The temperature was then reduced by 20 • C and the test resumed once the specimen's temperature had stabilized. The temperature does not deviate more than 0.5 degrees from the mean value of the respective temperature. 7.
Steps 3-5 were then repeated until the nitrogen tank was empty.
As the 600 L nitrogen tank did not provide enough nitrogen for the lowest test temperatures, the last two test temperatures for both materials were performed separately. To this goal, the chamber was stepwise cooled down to a temperature of −100 • C for the welded specimen (WS) and −110 • C for the base material (BM). As nitrogen displaced air, no ice formation was observed in the chamber during the experiments.
Both specimens were still intact after finishing the tests at the lowest temperatures. Thus, the specimens were opened to evaluate the beach marks. To assess the position of the beach marks, photographs of the fracture surfaces were taken. The length of the crack associated with the beach mark was determined as the mean value of three points on the curved shape of the beach mark with equal distance between each other and to the surface. The initial crack length estimated by the crack detection gauges was corrected by the information of the beach mark shape (crack front curvature).

Results of Fatigue Crack Growth Rate Tests
This section presents the results of the different experiments. The first subsection focuses on the results of calibration of the Johnson equation. Then the results of the decreasing ∆K tests with the goal to find a reasonable ∆K for the decreasing temperature tests are presented. Finally, the results of the experiments with the constant ∆K and decreasing temperatures are presented.

Calibration Test Results
In general, a good agreement was observed between calibration curves and Johnson's equation. For lower temperatures, the accuracy of Johnson's equation shows increasing but still insignificant deviations. At room temperature, the maximum deviation is lower than 2% for both types of specimens. At −80 • C the maximum deviation is still below 5% and the mean below 3%. The influence of the temperature on the accuracy is larger than the influence of the crack length. More detailed results are presented in Sallaba [37].

Decreasing ∆K Test Results
The threshold region was estimated according to ASTM 647-15 [35]. The values of crack growth rates between 10 −10 m/cycle and 10 −9 m/cycle were fitted and a value corresponding to 10 −10 m/cycle was used to define ∆K th . The decreasing ∆K experiments at room temperature yield a ∆K th of 3.85 MPa √ m and 2.35 MPa √ m for the S500 base material and for the welded specimen, respectively. The results of the base material experiments are presented in Figures 5 and 6. The results of the cracks in welded material are presented in Figures 7 and 8. The results of the decreasing ∆K experiments are presented in Figures 6 and 8  Interestingly, the crack growth rate is lower at 20 °C than at 0 °C. Below ature, the crack growth rate decreases constantly with decreasing temperatu °C. At −110 °C, the rate is slightly above the rate at −100 °C. This could be FTT; however, the change is small and still within the scatter band propose and Rolfe [38]. Figure 6 presents the results of the constant ΔK experiment together w of the decreasing ΔK experiment. As can be seen from this figure, the res temperature of both kinds of experiments are close together. Furthermore, constant ΔK experiments lay close to the target value of 8 MPa√m. The di the target value is caused by correcting the actual crack length after the ex addition, the trend of decreasing crack growth rate with decreasing test te clearly visible. Interestingly, the crack growth rate is lower at 20 °C than at 0 °C. Below this temperature, the crack growth rate decreases constantly with decreasing temperature until −100 °C. At −110 °C, the rate is slightly above the rate at −100 °C. This could be related to the FTT; however, the change is small and still within the scatter band proposed by Barsom and Rolfe [38]. Figure 6 presents the results of the constant ΔK experiment together with the values of the decreasing ΔK experiment. As can be seen from this figure, the results for room temperature of both kinds of experiments are close together. Furthermore, the ΔK of the constant ΔK experiments lay close to the target value of 8 MPa√m. The difference from the target value is caused by correcting the actual crack length after the experiments. In addition, the trend of decreasing crack growth rate with decreasing test temperature is clearly visible.

Results of the Decreasing Temperature Tests for the Weld Material
The results of the constant ΔK experiments of the specimen cracked in the weld material are displayed in Figure 7. Again, the crack growth rates increase slightly from room temperature to 0 °C, staying constant for 0 °C and −20 °C and decreasing for lower tem- the lowest temperature that was possible to achieve with the test setup without exc nitrogen consumption. The results of the decreasing ΔK and constant ΔK experiments of the welded men are presented together in Figure 8. The agreement is good but not as good as base material specimen. Many factors influence the crack growth in welded joints as welding deformation and material defects. In particular, the influence of seco bending stresses caused by specimen misalignment cannot be accounted for in su periments. The angular misalignment of both welded specimens was quite smal about 0.5°; nevertheless, misalignment effects on fatigue crack growth cannot be ruled out. The post-processed data uses the actual initial crack lengths evaluated fro beach marks. In summary, a constant decrease of fatigue crack growth rates was determin  The results of the decreasing ΔK and constant ΔK experiments of the welded specimen are presented together in Figure 8. The agreement is good but not as good as for the base material specimen. Many factors influence the crack growth in welded joints, such as welding deformation and material defects. In particular, the influence of secondary bending stresses caused by specimen misalignment cannot be accounted for in such experiments. The angular misalignment of both welded specimens was quite small with about 0.5°; nevertheless, misalignment effects on fatigue crack growth cannot be fully ruled out. The post-processed data uses the actual initial crack lengths evaluated from the beach marks. In summary, a constant decrease of fatigue crack growth rates was determined between −20 °C and −100 °C for the weld material, which is similar to the base material. Interestingly, the fatigue crack growth rates are similar for both materials throughout the tested temperature range. This could be related to the release of welding residual stresses

Results of the Decreasing Temperature Tests for the Base Material
The evaluated data of the experiments of the base material and constant ∆K is presented in Figure 5. The crack growth rate is plotted over the corresponding temperature. Barsom and Rolfe [38] developed a scatter band for mild steel depending on the crack growth rate. The scatter band is plotted as a vertical error bar for each crack growth rate to put the results in perspective.
Interestingly, the crack growth rate is lower at 20 • C than at 0 • C. Below this temperature, the crack growth rate decreases constantly with decreasing temperature until −100 • C. At −110 • C, the rate is slightly above the rate at −100 • C. This could be related to the FTT; however, the change is small and still within the scatter band proposed by Barsom and Rolfe [38]. Figure 6 presents the results of the constant ∆K experiment together with the values of the decreasing ∆K experiment. As can be seen from this figure, the results for room temperature of both kinds of experiments are close together. Furthermore, the ∆K of the constant ∆K experiments lay close to the target value of 8 MPa √ m. The difference from the target value is caused by correcting the actual crack length after the experiments. In addition, the trend of decreasing crack growth rate with decreasing test temperature is clearly visible.

Results of the Decreasing Temperature Tests for the Weld Material
The results of the constant ∆K experiments of the specimen cracked in the weld material are displayed in Figure 7. Again, the crack growth rates increase slightly from room temperature to 0 • C, staying constant for 0 • C and −20 • C and decreasing for lower temperatures. For −100 • C, the crack growth rate corresponds to 1.8 × 10 −9 m/cycle. This was the lowest temperature that was possible to achieve with the test setup without excessive nitrogen consumption.
The results of the decreasing ∆K and constant ∆K experiments of the welded specimen are presented together in Figure 8. The agreement is good but not as good as for the base material specimen. Many factors influence the crack growth in welded joints, such as welding deformation and material defects. In particular, the influence of secondary bending stresses caused by specimen misalignment cannot be accounted for in such experiments. The angular misalignment of both welded specimens was quite small with about 0.5 • ; nevertheless, misalignment effects on fatigue crack growth cannot be fully ruled out. The post-processed data uses the actual initial crack lengths evaluated from the beach marks.
In summary, a constant decrease of fatigue crack growth rates was determined between −20 • C and −100 • C for the weld material, which is similar to the base material. Interestingly, the fatigue crack growth rates are similar for both materials throughout the tested temperature range. This could be related to the release of welding residual stresses in the welded specimen by heat treatment. The only exception is the rate determined for the base material at 0 • C.

Fracture and Charpy Impact Toughness Properties
To compare the results of the previous section, Charpy impact and fracture toughness testing have been performed for different zones of the welded joints, i.e., the base material, the middle of the weld metal (WM), and the heat-affected zone (HAZ). For this purpose, specimens were machined according to ASTM E23-16b [39] and ASTM E399-20a [40] with notches in the different zones as presented in Figure 9. Calibrated thermoelements were attached to the Charpy and fracture toughness tests close to the notch or crack tip to perform the tests at intended test temperatures. in the welded specimen by heat treatment. The only exception is the rate determined for the base material at 0 °C.

Fracture and Charpy Impact Toughness Properties
To compare the results of the previous section, Charpy impact and fracture toughness testing have been performed for different zones of the welded joints, i.e., the base material, the middle of the weld metal (WM), and the heat-affected zone (HAZ). For this purpose, specimens were machined according to ASTM E23-16b [39] and ASTM E399-20a [40] with notches in the different zones as presented in Figure 9. Calibrated thermoelements were attached to the Charpy and fracture toughness tests close to the notch or crack tip to perform the tests at intended test temperatures.

Charpy Impact Toughness Tests
The test results of the Charpy impact testing are presented in Figure 10 including sigmoidal fits to assess the mean DBTT (e.g., T50 J) [41,42].

Charpy Impact Toughness Tests
The test results of the Charpy impact testing are presented in Figure 10 including sigmoidal fits to assess the mean DBTT (e.g., T 50 J ) [41,42].

Fracture Toughness Tests
The fracture toughness tests have been performed according to ASTM E399-20 [40] and the fracture toughness DBTT (T0) was determined based on the master curve theory by Wallin [48]. To this goal, the procedure outline in ASTM E1921-20 [49] was applied and standard 10mm × 20 mm single-edge bend specimens were tested. Herein, the DBTT is associated with the temperature (T0) at which the critical stress intensity factor is 100 MPa√m for an equivalent 25 mm thick specimen.
Fracture toughness tests are typically determined in a quasi-static manner; however, the goal of the here presented tests is to determine a DBTT based on a loading rate similar to the FCG rate tests. Thus, a higher loading rate of 1 mm/s was chosen. This rate approximately matches the stress intensity factor rate of the FCG rate tests ( ≈ 8 MPa√m ). This rate is about two times the rate at (T0) ( = 100 MPa√m s ); however, higher loading rates would have reduced the measuring accuracy, and more importantly, the DBTT is related to the logarithm of strain rate [50]. Thus, a difference in the loading rate by a factor of two does not have a significant effect on (T0).
The results of the tests are presented in Figure 11. As expected, the lowest DBTT is observed for the base material; however, in contrast to the Charpy impact tests, a lower DBTT is observed for the weld metal than for the heat-affected zone. Furthermore, a higher DBTT is obtained for BM and HAZ from the fracture toughness tests with a loading rate of 1 mm/s than from the Charpy tests. Different curve trends are observed for the different zones of the welded joint. First of all, the base material has DBTT (T 50 J = −119 • C) much lower than the minimum requirement of 50 J according to DNVGL-OS-B101 [43]. Furthermore, the heat-affected zone shows a very low DBTT of T 50 J = −85 • C. In contrast, a much lower toughness and higher transition temperature is observed (T 50 J = −39 • C) for the weld metal.
Testing of the middle of the weld metal is not required by EN 10225-1:2019 [44] but by DNVGL-OS-C401 [45]. It is typically assumed that the microstructure of the weld metal is inferior to the base metal, and the lowest toughness is usually associated with local brittle zones in the coarse-grain heat-affected zone [29]; however, in this study, the lowest toughness is found in the WM. This is thought to be related to the flux-cored arc welding (FCAW) process. A number of studies [28,46,47] observed much lower toughness in FCAW joints compared to other welding processes.

Fracture Toughness Tests
The fracture toughness tests have been performed according to ASTM E399-20 [40] and the fracture toughness DBTT (T 0 ) was determined based on the master curve theory by Wallin [48]. To this goal, the procedure outline in ASTM E1921-20 [49] was applied and standard 10mm × 20 mm single-edge bend specimens were tested. Herein, the DBTT is associated with the temperature (T 0 ) at which the critical stress intensity factor is 100 MPa √ m for an equivalent 25 mm thick specimen. Fracture toughness tests are typically determined in a quasi-static manner; however, the goal of the here presented tests is to determine a DBTT based on a loading rate similar to the FCG rate tests. Thus, a higher loading rate of 1 mm/s was chosen. This rate approximately matches the stress intensity factor rate of the FCG rate tests ( ). This rate is about two times the rate at (T 0 ) ( . K = 100 MPa √ m s ); however, higher loading rates would have reduced the measuring accuracy, and more importantly, the DBTT is related to the logarithm of strain rate [50]. Thus, a difference in the loading rate by a factor of two does not have a significant effect on (T 0 ).
The results of the tests are presented in Figure 11. As expected, the lowest DBTT is observed for the base material; however, in contrast to the Charpy impact tests, a lower DBTT is observed for the weld metal than for the heat-affected zone. Furthermore, a higher DBTT is obtained for BM and HAZ from the fracture toughness tests with a loading rate of 1 mm/s than from the Charpy tests.

Fracture Surface Investigations by Means of Scanning Electron Microscopy
To explain changes in fatigue and fracture mechanics properties, fracture surface in vestigations by means of scanning electron microscopy were performed. First, the fractur surfaces of the fracture toughness test specimens are assessed to create a reference to as sess the fracture surfaces of the fatigue test specimens. For comparison, two locations clos to the pre-fatigued crack tip in the middle of the fracture surface were magnified.
The comparison of the fracture surfaces for two specimens tested at −40 °C and −8 °C shows clear evidence of different fracture mechanisms. While the magnification of th surface of the specimen tested at −40 °C (Figure 12c) shows a large number of dimple (typical for ductile failure), the fracture surface of the specimen tested at −85 °C (Figur 12d) is dominated by transgranular cleavage planes.

Fracture Surface Investigations by Means of Scanning Electron Microscopy
To explain changes in fatigue and fracture mechanics properties, fracture surface investigations by means of scanning electron microscopy were performed. First, the fracture surfaces of the fracture toughness test specimens are assessed to create a reference to assess the fracture surfaces of the fatigue test specimens. Compared to the apparent difference in fracture surface morphology of the fracture toughness test specimen, almost no difference is observed when assessing the surfaces of the fatigue crack growth test specimens. Again, locations in the middle of specimen thickness were magnified for comparison. To avoid the influences of previous measurements, a location corresponding roughly to the middle of a load block was chosen. This was enabled by beach marking the test specimens. The fracture surfaces of the base material and welded joints specimen are compared at −100 °C, as this was the lowest test temperature of the welded test specimen, see Figure 13. This test temperature also led to the lowest fatigue crack growth rate measured at approximately ΔK = 8 MPa√m. Comparing both test specimens at room temperature and −100 °C, no change in fatigue crack growth mechanism can be observed with decreasing test temperature. This is not surprising for the base material specimens, as very low DBTTs were observed from both Charpy V-notch Compared to the apparent difference in fracture surface morphology of the fracture toughness test specimen, almost no difference is observed when assessing the surfaces of the fatigue crack growth test specimens. Again, locations in the middle of specimen thickness were magnified for comparison. To avoid the influences of previous measurements, a location corresponding roughly to the middle of a load block was chosen. This was enabled by beach marking the test specimens. The fracture surfaces of the base material and welded joints specimen are compared at −100 • C, as this was the lowest test temperature of the welded test specimen, see Figure 13. This test temperature also led to the lowest fatigue crack growth rate measured at approximately ∆K = 8 MPa √ m. Comparing both test specimens at room temperature and −100 • C, no change in fatigue crack growth mechanism can be observed with decreasing test temperature. This is not surprising for the base material specimens, as very low DBTTs were observed from both Charpy V-notch impact and fracture toughness testing; however, based on the DBTTs of the weld metal, a change in fatigue crack growth behavior was expected for the fatigue test specimens with cracks located in the weld metal. In fact, all four fracture surfaces are characterized by typical fatigue crack striations. impact and fracture toughness testing; however, based on the DBTTs of the weld metal, a change in fatigue crack growth behavior was expected for the fatigue test specimens with cracks located in the weld metal. In fact, all four fracture surfaces are characterized by typical fatigue crack striations.
(a) (b) (c) (d) Figure 13. Fracture surfaces of fatigue crack growth rate test specimens with crack located in the base material at (a) room temperature, and (b) −100 °C, and with crack located in the weld metal at (c) room temperature, and (d) −100 °C.

Knowledge Gaps with Respect to the FTT
To avoid brittle fracture of materials at low temperatures, international standards require that the DBTT remains below the anticipated service temperatures at all times; however, low service temperatures also influence other limit states. For fatigue limit state design, it is generally assumed that lower temperatures have no detrimental effect on the fatigue properties of steels as long as the service temperature remains above the DBTT, see ISO/TS 35105:2018 [21] or Alvaro et al. [24]. This assumption is backed by fatigue crack growth rate testing of base materials for a wide variety of steel grades.
In steel structures without welds, fatigue life can be divided into two parts, fatigue crack initiation and fatigue crack growth. Depending on the geometry, stress state, and presence of material defects, one might dominate the other. The mechanism of fatigue damage is known to be the result of complex dislocation arrangement governed by local irreversible plastic flow [24]. Since the static strength increases with lower temperatures, the resistance against dislocation movement is increased and thus both the fatigue crack

Knowledge Gaps with Respect to the FTT
To avoid brittle fracture of materials at low temperatures, international standards require that the DBTT remains below the anticipated service temperatures at all times; however, low service temperatures also influence other limit states. For fatigue limit state design, it is generally assumed that lower temperatures have no detrimental effect on the fatigue properties of steels as long as the service temperature remains above the DBTT, see ISO/TS 35105:2018 [21] or Alvaro et al. [24]. This assumption is backed by fatigue crack growth rate testing of base materials for a wide variety of steel grades.
In steel structures without welds, fatigue life can be divided into two parts, fatigue crack initiation and fatigue crack growth. Depending on the geometry, stress state, and presence of material defects, one might dominate the other. The mechanism of fatigue damage is known to be the result of complex dislocation arrangement governed by local irreversible plastic flow [24]. Since the static strength increases with lower temperatures, the resistance against dislocation movement is increased and thus both the fatigue crack initiation and the crack propagation are slowed [24]; yet, for metals that feature a DBTT, the fatigue crack growth rate can increase again for temperatures below the FTT.
From base materials tests, it is known that both the slope and the intersect of the FCG curves change significantly, see Moody and Gerberich [22]; however, there is very limited data available for welded joints, see Braun and Ehlers [23]. More importantly, the relation between the fatigue and fracture ductile-brittle transition has so far only been assessed empirically. In addition, the only known investigation of this relation for welded materials has been performed for thermally weld simulated material (Gleeble machine tests), see Alvaro et al. [19]. To the authors' knowledge, there have been limited investigations for actual welded joints, see [51][52][53]. This could be related to the complexity of such tests and additional influencing factors that are difficult to account for, e.g., welding-related specimen misalignment, weld defects, microstructural variations, and residual stresses. To overcome this problem, the current study performed tests with a single specimen by successively reducing the test temperature. Therefore, it is not possible to measure a full FCG curve but to perform a comparison of FCG rates at a constant ∆K. For this study, a ∆K of 8 MPa √ m was chosen, which is within the Paris regime of the FCG curve (linear part). The results of the experiments have shown decreasing crack growth rates for temperatures below 0 • C for crack locations both in the base and the weld material. The S500 base material was tested to a temperature of −110 • C and the welded specimen to −100 • C. Down to this temperature, no distinct fatigue ductile-to-brittle transition occurred. Only at −110 • C, a slight increase in FCG rate was observed for the base material.

Relation between the FTT and the DBTT
Comparing the results with the results of the Charpy V-notch impact and fracture toughness tests, a higher FTT would have been expected. Alvaro et al. [19] and Walters et al. [20], observed an FTT 15 • C and 12 • C higher than the Charpy DBTT for the T 40 J and T 27 J values, respectively. Kawasaki et al. [54] observed that the FTT occurred at the same temperature as the fracture toughness that they referred to as the "fatigue fracture toughness". For the tested S500 high-strength steel, DBTT T 50 J of −39 • C and −119 • C were obtained for the weld and base material, respectively. In contrast, fracture toughness DBTT of −71 • C and −104 • C were obtained for the same zones by tests with increased loading rates that match the stress intensity factor rate of the FCG rate tests. Those two zones were chosen for FCG rate testing based upon the lower Charpy impact toughness for the weld metal and because a notch location in the HAZ would intersect several other zones of the welded joint. Attributing a change in FCG rate behavior at low temperatures to a particular zone of the welded joint would therefore be difficult.
The FCG rate tests performed in this study show that there is no sudden increase in FCG rates down to the Charpy and fracture toughness DBTTs for S500 base material and even below the two DBTTs for the weld material for a stress range of ∆K = 8 MPa √ m and an R-ratio of 0.5. This seems contradictory to the results of Walters et al. [20]. At FCG test temperatures below the DBTT, no sudden brittle fracture would have been expected due to the low ∆K of 8 MPa √ m; yet, an increase in FCG rate was expected based on the results presented by Alvaro et al. [19] and Walters et al. [20]. In contrast, the specimens showed no sign of accelerated crack growth. The results seem to match better with the more subtle changes in FCG rate reported by Kawaski et al. [54]. This was also verified by scanning electron microscopy. The reason for this might be related to the choice of a low, constant ∆K for the FCG tests. This rate is in the lower part of the Paris regime of FCG curves and therefore representative for short cracks with slow crack propagation. Such a rate would, for example, characterize the majority of small fatigue cracks initiating at weld transitions. Alvaro et al. [19] observed cleavage burst-that they related to the FTT-at much higher ∆K of 17.2 MPa √ m. The effect of the FTT might therefore be strongly related to the magnitude of the stress intensity factor range. This would explain why no distinct FTT was observed even at temperatures below the DBTTs obtained from Charpy impact and fracture toughness tests; however, Walters et al. [20] pointed out the high scatter of crack growth tests in the transition region, which could also cover possible changes in FCG behavior.
Another explanation is that a number of authors have reported that the FCG curves at low temperatures tend to have higher slopes than the FCG curves at high temperatures, resulting in an intersection between the low-and high-temperature FCG curves somewhere in the Paris regime. These curves then have lower FCG values at low temperatures and low values of ∆K and higher FCG values at low temperatures and high values of ∆K, e.g., [22,51,55]. The intersection of FCG curves for room and low temperatures is presented in Figure 14.
growth tests in the transition region, which could also cover possible changes in FCG be havior.
Another explanation is that a number of authors have reported that the FCG curve at low temperatures tend to have higher slopes than the FCG curves at high temperatures resulting in an intersection between the low-and high-temperature FCG curves some where in the Paris regime. These curves then have lower FCG values at low temperature and low values of ΔK and higher FCG values at low temperatures and high values of ΔK e.g., [22,51,55]. The intersection of FCG curves for room and low temperatures is presented in Figure 14. Figure 14. A fatigue crack growth rate (da/dN) versus stress intensity factor range (∆K) curve for a low-carbon steel above and below the FTT, based on Yarema et al. [56] and taken from Braun [32].
With testing of multiple temperatures at only one value of ΔK, it is impossible to know if the ΔK value is below, at, or above the intersection. Hypothetically, materials tha were tested at the intersection should show little influence of temperature.
In addition, a number of other factors are known to affect the FTT and DBTT. Some of them are related to the test conditions, i.e., specimen geometry and load ratio [29]. Oth ers are related to the defects and inhomogeneities created during welding. The process o welding itself and the reheating and cooling of several weld layers has an impact on mi crostructure and mechanical properties [28], as do the welding consumables [30].
Typically, the highest ductile-brittle transition temperature is found in the coarse grained and intercritically reheated coarse-grained heat-affected zones [6,19,29,31]. Thi zone is, however, fairly small, particularly in thin-plated structures with few weld layers such as in ships. This might be the reason why the lowest Charpy notch impact toughnes was observed in weld metal of the test specimens used for this study. As the specimen were rather thin with a thickness of about 10 mm, different results might have been ob served in thick-plated structures with more weld layers. Another important aspect o welded structures is welding residual stresses. Typically, high tensile residual stresses are created in large welded structures. To account for this effect in a simplified manner, the weld material specimens of this study were first stress relieved by heat treatment and then tested at a high-stress ratio R = 0.5 to simulate high tensile residual stresses. In actual large scale structures, residual stresses are typically not constant. This effect will therefore alte  A fatigue crack growth rate (da/dN) versus stress intensity factor range (∆K) curve for a low-carbon steel above and below the FTT, based on Yarema et al. [56] and taken from Braun [32].
With testing of multiple temperatures at only one value of ∆K, it is impossible to know if the ∆K value is below, at, or above the intersection. Hypothetically, materials that were tested at the intersection should show little influence of temperature.
In addition, a number of other factors are known to affect the FTT and DBTT. Some of them are related to the test conditions, i.e., specimen geometry and load ratio [29]. Others are related to the defects and inhomogeneities created during welding. The process of welding itself and the reheating and cooling of several weld layers has an impact on microstructure and mechanical properties [28], as do the welding consumables [30].
Typically, the highest ductile-brittle transition temperature is found in the coarsegrained and intercritically reheated coarse-grained heat-affected zones [6,19,29,31]. This zone is, however, fairly small, particularly in thin-plated structures with few weld layers, such as in ships. This might be the reason why the lowest Charpy notch impact toughness was observed in weld metal of the test specimens used for this study. As the specimens were rather thin with a thickness of about 10 mm, different results might have been observed in thick-plated structures with more weld layers. Another important aspect of welded structures is welding residual stresses. Typically, high tensile residual stresses are created in large welded structures. To account for this effect in a simplified manner, the weld material specimens of this study were first stress relieved by heat treatment and then tested at a high-stress ratio R = 0.5 to simulate high tensile residual stresses. In actual large-scale structures, residual stresses are typically not constant. This effect will therefore alter the effect stress ratio at critical locations.

Aspects to Consider in Future Work
In summary, the chosen approach by testing only one specimen at different temperatures presented clear evidence of decreasing FCG rates at low temperatures; however, no distinct FTT was observed even below the DBTT obtained from Charpy and fracture toughness tests for the investigated high-strength S500 steel. Based on the assumption that brittle fracture occurs before the changing fatigue behavior significantly affects the lifetime of a structure, current regulations for material qualifications are aimed at enforcing sufficient toughness values at the design temperature, e.g., ISO 19906:2019 [57]. The main objective of many research projects is therefore to provide a reliable basis for estimating the transition to brittle fracture of base materials and welded structures; yet, above the transition temperatures, the fatigue strength is significantly increased. This effect is currently not considered in any design standard or relevant specifications, e.g., ISO/TS 35105:2018 [21]. In fact, compared to the well-investigated fracture ductile-to-brittle transition, the effect of temperature on fatigue properties of welded structures is scarcely investigated.

Conclusions
This study investigated the relation between the fatigue transition temperature and the ductile-brittle transition temperature by means of fatigue crack propagation tests at constant stress intensity factor range and fracture mechanics tests (i.e., Charpy impact, fracture toughness), respectively. For this purpose, the base material and welded joints of S500 high-strength steel were investigated. In addition, scanning electron microscopy photographs of the fracture surfaces were made to assess changes in the failure mechanism. From the experiments, the following conclusions are obtained:

•
From the tests, an almost constant decrease in fatigue crack propagation rate is determined with decreasing test temperature even below the ductile-brittle transition temperature of the weld metal. • At −100 • C, the fatigue crack propagation rate is about half of the rate observed at room temperature for both base material and weld metal of the S500 steel. • Compared to the apparent difference in fracture surface morphology of the fracture toughness test specimen, almost no difference is observed when assessing the surfaces of the fatigue crack growth test specimens of both base material and weld metal at −100 • C and thereby well below the ductile-brittle transition temperature.

•
The relationship between fatigue crack growth rate and temperature is thought to be related to the magnitude of the stress intensity factor range. This could explain why no distinct fatigue transition temperature was observed even at temperatures below the ductile-brittle transition temperatures obtained from Charpy impact and fracture toughness tests.
In recent years, there has been an increased interest in low-temperature effects on fatigue and the relation between fatigue and fracture ductile-to-brittle transition; nevertheless, there is still a lot of work to do on this subject to fully understand the fatigue and fracture transition behavior and to correctly account for temperature effects during design.