3.1 Characterization of Substrate and Feedstock
Light optical microscopy (LOM) images of the as-received microstructure in the AISI 316L feedstock are shown in Figs. 3a and 3b. The microstructure consists of coarse austenitic grains, typically 50 – 100 mm, that contain twins and shear bands associated with prior cold work equivalent to approximately 10% reduction [33]. LOM and scanning electron microscope (SEM) images of the as-received microstructure in the AISI 4340 substrate are provided in Figs. 3c and 3d, respectively. The observed microstructure consists of tempered martensite, and is consistent with the nominal hardness of 42 HRC.
Table 4 provides the measured yield strength, tensile strength, and hardness in the substrate and filler feedstock, along with specified values related to tempering of the 4340 or cold work in the 316L. The measured hardness of 426 HV0.2 in the 4340 substrate, which corresponds to 43 HRC per ASTM E140-02 [34], is again consistent with the specified minimum of 42 HRC. The measured hardness is also consistent with the measured yield and tensile strengths of 1286 and 1368 MPa, respectively, and corresponds to a tempering temperature of approximately 500 oC for two hours [35]. In the 316L feedstock, the measured tensile strength of 724 MPa approximately correspond to a 1/8-hard condition per ASTM A666-15 [36].
Table 4: Results of tensile and hardness testing of the as-received substrate and feedstock alloys, along with published values for different temper/cold-work conditions.
Material
|
Condition
|
Yield
Strength,
MPa
|
Tensile
Strength,
MPa
|
HRB/C
|
HV
|
Source
|
4340
|
Substrate
|
1286
|
1368
|
43C, avg.
|
426
|
Measured
|
Temper 205 oC
|
1860
|
1980
|
53C, max.
|
560
|
Ref. 30
|
Temper 315 oC
|
1620
|
1760
|
50C, max.
|
513
|
Temper 425 oC
|
1365
|
1500
|
46C, max.
|
458
|
Temper 540 oC
|
1160
|
1240
|
39C, max.
|
382
|
Temper 650 oC
|
860
|
1020
|
31C, max.
|
310
|
Temper 705 oC
|
740
|
860
|
24C, max.
|
260
|
316L
|
Feedstock
|
586
|
724
|
24C, avg.
|
257
|
Measured
|
Annealed
|
205
|
515
|
95B
|
-
|
Ref. 31
|
1/8 hard
|
380
|
690
|
-
|
-
|
1/4 hard
|
515
|
860
|
-
|
-
|
1/2 hard
|
760
|
1035
|
-
|
-
|
3.2 AFSD Groove-Filling
Fig. 4 shows LOM images of metallurgical cross-sections from the AFSD groove-filled locations G1 – G4, indicated in Fig. 2a. The sections have been etched with nital, which reveals the microstructure in the 4340 substrate but has no effect on the 316L filler. The images show the deformation of the groove geometry to be asymmetrical, with more deformation on the advancing side of the tool (corresponding to the left side of the images). This is consistent with prior reports of asymmetry in the interfacial mixing during deposition of AA2024 on AA6061, where the asymmetry was attributed to the interaction between the in-plane material expansion (driven by compression of the feedstock) and the material flow (driven by rotation of the tool head) [12]. In addition, the deformation is more pronounced in the smaller-radius grooves 1 and 2 (Figs. 4a and 4b) than in larger-radius grooves 3 and 4 (Figs. 4c and 4d). This suggests that the lateral constraint formed by the smaller radius transfers the shear forces to the substrate more effectively. Both of the deeper grooves, Figs. 4b and 4d, experienced cracking or lack of fusion at the interfaces, indicated by the white arrows in Fig. 4. This is shown in more detail in Fig. 5, where the images reveal different behavior in the two groove geometries. The smaller diameter groove 2 in Fig. 5a exhibits lack of fusion in the deeper parts of the groove (lower right) and well-bonded interface with localized cracking of the substrate in the shallower parts (upper left). In contrast, the larger diameter groove 4 in Fig. 5b exhibits lack of fusion along most of the substrate-filler interface. The feature indicated by the dark arrow in Fig. 5a is a trail of carbide inclusions the AFSD 316L introduced during the stir process.
Beneath the filled grooves, the microstructure shows two distinct layers of HAZ that vary in depth and relative thickness for the different geometries. The microstructures shown in Fig. 6 correspond to the four numbered locations in Fig. 4a: 1) 316L AFSD filler, 2) HAZ1, 3) HAZ2, and 4) substrate. Compared to the as-received 316L microstructure (Figs. 3a and 3b), the microstructure in the AFSD 316L (Fig. 6a) is significantly refined, with grain size <25 µm and minimal twinning. The HAZ1 microstructure in Fig. 6b is predominantly martensitic, indicating that the region austenized above ~727 oC, then transformed to martensite during rapid cooling []. The presence of martensite suggests decreased toughness near the interface. At the bottom of HAZ1 there is a gradual transition to the HAZ2 microstructure (Fig. 6c), which is comprised of coarse carbides consistent with upper bainite. This suggests that the HAZ2 region austenized, but did not cool rapidly enough to form martensite. In addition, HAZ2 has a significant gradient in the grain size or coarseness of the structure, with the bottom of HAZ2 being finer than the initial tempered martensitic structure found in the substrate. This further supports the idea that HAZ2 austenized during the process, with higher maximum temperatures resulting in larger grains closer to the surface. Finally, the microstructure in the substrate below HAZ2 (Fig. 6d) is consistent with the as-received microstructure in Figs. 3c and 3d. SEM images of the microstructures at the bottom of HAZ2 and the top of the underlying substrate are shown at higher magnification in Figs. 7a and 7b, respectively. The images confirm the bainitic structure in the HAZ2 to have a finer overall morphology than the substrate, implying a low austenizing temperature, with islands of hypoeutectoidal ferrite.
Fig. 8 shows the results of microhardness depth profiling at the locations indicated by the dashed lines in Figs. 4a and 4b. The data show the hardness of the AFSD 316L to be approximately 88% of the original feedstock, 225 HV0.2 versus 257 HV0.2 from Table 4. This decrease in hardness is associated with dynamic recrystallization that occurs during the AFSD process, leading to the competing effects of relaxation of the strain hardening and decrease in grain size. In HAZ1 the hardness is substantially elevated, exceeding 600 HV0.2 and approaching the maximum quenched hardness of ~670 HV for 0.38% carbon steel []. This hardness is consistent with the largely martensitic microstructure observed in Fig. 6b. The HAZ2 hardness ranges from ~347 HV0.2 near the boundary with HAZ1 to ~405 HV0.2 at the boundary with the substrate microstructure, with the gradient being attributed to the observed trend in grain size discussed above. Finally, the hardness in the substrate below HAZ2 ranges from ~340 HV0.2 near HAZ2 to the baseline value of ~426 HV0.2 reported in Table 4. This gradient near the boundary with HAZ1 does not correlate with any visible microstructural changes, and is attributed to partial stress relieving and tempering that occurs at temperatures between 595 oC and 675 oC [37]. The hardness data are consistent across the different locations in the two samples, and are summarized in Table 5.
Table 5: Vickers microhardness results from the depth profiles in Fig. 8, but segregated by test location. ‘n’ refers to the number of measurements.
|
Grooves (Fig. 8)
|
Overlap Clad (Fig. 13)
|
n
|
Average
|
St. Dev.
|
n
|
Average
|
St. Dev.
|
316
|
AFSD
|
41
|
225
|
30
|
16
|
249
|
47
|
4340
|
HAZ1
|
9
|
637
|
30
|
20
|
605
|
49
|
4340
|
HAZ2
|
38
|
386
|
29
|
16
|
392
|
17
|
4340
|
Substrate
|
92
|
424
|
16
|
98
|
417
|
11
|
The observed HAZ’s are the result of a complex thermal profile resulting from the rapid heating, short term exposure at the maximum temperature, and air cooling during each of the multiple AFSD passes. The two zones are differentiated by the maximum temperature attained, which was higher near the surface, and by the cooling rate, which was faster near the surface. These factors interact to affect the extent of austenization, the prior austenitic grain size, and the formation of bainite versus martensite. The presence of martensite in HAZ1 suggests that the material transformed to austenite, then cooled sufficiently rapidly to inhibit the formation of bainite. Similarly, the gradient in the grain size in HAZ2, as well as the refined grain size relative to the original microstructure, suggest that the alloy reached a lower transformation temperature, and cooled less rapidly than HAZ1, with the net effect of converting the original tempered martensite primarily to upper bainite.
The difference in hardness between the AISI 4340 substrate and the AFSD filler metal creates certain implications with respect to mixing at the interface. For example, comparison of the deposition patterns in Fig. 4 with prior published work relating to more evenly matched, and softer, alloys suggests unusually distinct interfaces in the current samples [12,13]. In fact, prior work on AFSD and FSW has defined several characteristic microstructural zones associated with solid state friction processes: stir zone (SZ), thermomechanical affected zone (TMAZ), heat affected zone (HAZ), and base metal (BM) [14,16]. The TMAZ refers to regions where the substrate material has been plastically deformed, with evidence of material flow, breakup of inclusions, and distortion of interfaces or boundaries. This region is generally absent from the observed microstructures for 316L AFSD on AISI 4340. This is illustrated in Fig. 4, where the visible deformation of the substrate is limited to localized distortion of the grooves, mostly on the advancing side of the first pass (the left side of the images). As a result of the apparent lack of shear deformation, the extent of mixing of the substrate and filler alloys at the interface was evaluated further.
Fig. 9 shows LOM images from the left edge of groove 1 in section G1, and from the right edge of groove 3 in section G3. As with the prior images, the reagent that was used to etch the substrate has no effect on the 316L filler alloy, thus providing a strong contrast between the two materials. The resultant images show evidence of localized mixing, with visible trails of substrate alloy entrained into the filler metal. The observed mixing is typically short-range at the interface, <50 mm, with some evidence of longer-range entrainment of substrate alloy into the AFSD 316L visible in the upper-half of Fig. 9b. To further reveal this aspect of the mixing, the AFSD 316L filler was electrolytically etched in oxalic acid for 90 s at 1.0 A/cm2. This etching condition is designed to reveal possible sensitization due to depletion of Cr from the grain boundaries [39]. The resultant microstructures are shown in Fig. 10 for the area from Fig. 9a. The thickening of the grain boundaries in the filler alloy indicated by ‘1’ in Fig. 10b suggests the early stages of potential sensitization, which can be associated with the formation of carbides due to the ingress of C from the graphite lubricant, or to the formation of sigma phase that has been reported to occur during friction stir processing of AISI 316L [40]. The effect is generally localized to the interfaces between the substrate and filler, indicated by ‘1’ in Fig. 10b and highlighted in Fig. 10c, and between subsequent AFSD passes, indicated by ‘2’ in Fig. 10a. Away from these regions, for example at ‘3’ in Fig. 10b, the etching does not significantly attack the filler alloy, except for a few areas with the appearance of shear-driven flow, location ‘4’ in Fig. 10a. Note that while the localized thickening of the grain boundaries does indicate some degree of decreased resistance to corrosion relative to the unaffected parts of the filler, the oxalic acid etch is a sensitive screening test that may detect conditions which are not necessarily detrimental from a performance perspective [39]. As such, further investigation is required to develop a fuller understanding of the corrosion performance of AFSD 316L, and the potential role of lubricants used during the process.
Tensile testing was performed on sub-sized specimens extracted from grooves 1 and 3, and oriented perpendicular to the AFSD/groove axis. The thickness of the specimens from groove 1, T1 – T3, was targeted to match the depth of the groove in order to isolate the contribution of the AFSD filler and the filler-substrate interfaces. In contrast, the thickness of the specimens from groove 3, T4 –T6, targeted approximately equal amounts of filler and substrate. The results tabulated in Table 6 show the average yield and tensile strengths for the groove 1 specimens to be 516 and 725 MPa, respectively, while those from groove 3 are 645 and 1391 MPa, respectively.
Table 6: Results of tensile testing AFSD locations perpendicular grooves 1 and 3, with additional data parallel to a thick multi-pass cladding.
Location
|
Orientation
|
Sample
ID**
|
Yield
Strength,
MPa
|
Tensile
Strength,
MPa
|
Groove 1
|
Transverse
|
T1
|
513.7*
|
717.3
|
T2
|
489.5
|
709.6
|
T3
|
542.6
|
747.4
|
Groove 3
|
Transverse
|
T4
|
674.3
|
1345.1
|
T5
|
621.2
|
1404.9
|
T6
|
640.5
|
1424.3
|
Cladding
316L only+
|
Axial
|
N/A
|
591.6
|
634.2
|
N/A
|
405.4
|
695.2
|
N/A
|
561.3
|
618.4
|
* Estimated from cross-head displacement due to malfunction of extensometer.
** From Fig. 2
+ Not shown
Fig. 11 shows the fracture cross-sections for samples T1 from groove 1 and T5 from groove 3. The images show ductile necking in the AFSD filler (top), with minimal deformation in the harder substrate (bottom). The T1 sample is shown in Fig. 11a, to be predominantly filler alloy (~80%), while the T5 sample is approximately one-half substrate (~50%), which accounts for the elevated strength values. Assuming a linear rule of mixtures, a simple linear regression can be used to estimate the tensile strength of the 4340 substrate and 316L filler as 1599 and 524 MPa, respectively. For the 316L filler, this corresponds to a well-annealed condition per Table 4, which is not surprising given the microstructure observed in Fig. 6. For the 4340 substrate, the interpretation is more challenging due to the non-uniform microstructure discussed above. However, the data indicate that immediately below the grooves, the substrate material has an elevated tensile strength, which is consistent with the elevated hardness and the observed martensitic microstructure. In both samples, the fracture propagated through the filler alloy, rather than along the interface with the substrate, and occurred at approximately the minimum thickness of the underlying substrate material. This indicates that, for this test geometry, the interfaces are at least as strong as the filler alloy. For comparison, an additional set of tensile tests was performed on samples extracted from a thick single-width cladding, and oriented axially along the AFSD deposition direction. Those samples, which were comprised solely of AFSD alloy with no attached substrate material, had average yield and tensile strengths of 519 and 649 MPa, respectively.
3.3 AFSD Cladding
Surface cladding by AFSD was also explored as a potential method suitable for material replacement over larger areas. Fig. 2b showed a region of cladding that was deposited with approximately 20% overlap between passes. Four metallurgical samples prepared from the locations identified as O1 - O4 in Fig. 2b are shown in Fig. 12. The cross-sections show a similar HAZ pattern comprising two layers with microstructures comparable to those previously shown in Fig. 6. Hardness depth profiles performed at the locations indicated by the dashed lines in Figs. 12c and 12d are presented in Fig. 13. The depth profiles are comparable to those from the groove-filling geometry presented in Fig. 8, with a slight decrease in the hardness of the filler material relative to the as-received condition, HAZ1 hardness again in excess of 600 HV0.2, HAZ2 hardness averaging 392 HV0.2, and BM hardness increasing from a minimum near HAZ2 to the baseline measured in the as-received plate. In general, HAZ1 is thicker than in the groove-filling examples from Fig. 8, while HAZ2 is less pronounced and lacks the measureable gradient between the boundaries with HAZ1 and BM. These differences are attributed to the different thermal histories associated with the two substrate geometries and AFSD process parameters. The hardness data are again summarized in Table 5.
The sharp appearance of the interface between the substrate and cladding shown in Fig. 12, and the absence of a significant TMAZ below the deposited material, are similar to the characteristics observed in the groove-filling geometry, Figs. 4 and 6. Closer inspection of the interface reveals comparable localized mixing (Figs. 14a and 14b) and evidence of sensitization of the associated grain boundaries (Figs. 14c and 14d). Additional etching is visible in the interface between consecutive passes and along apparent flow lines (Fig. 14b). As above, the etch-pattern suggests that the shear mixing that bonds the interface has a potentially deleterious effect on the localized corrosion resistance of the filler alloy, possibly due to the formation of carbides enabled by incorporation of C from the graphite lubricant. Again, it is worth noting the oxalic acid etch is a sensitive screening test that may detect conditions which are not necessarily detrimental from a performance perspective [39]. As such, additional investigation of the corrosion resistance of AFSD 316L repairs represents a significant opportunity for future work.
The subsurface hardness of the AFSD 316L cladding was shown in Table 5 and Fig. 13 to be ~249 HV0.2, which is consistent with the normal surface hardness of 96 HRB determined by Rockwell testing. In contrast, the measured subsurface hardness of the as-received substrate provided in Table 4 is 426 HV0.2, while the normal surface hardness determined by Rockwell testing is 41 HRC. The difference in hardness between the as-received AISI 4340 substrate and the surface of the AFSD 316L filler material may raise some concern about loss of wear resistance resulting from the resurfacing. This was evaluated by performing two types of wear testing at the locations indicated W1 – W4 in Fig. 2b: two-body wear by abrasive loop contact, and erosion by solid particle impingement [31,32]. The results are provided in Table 7, from which it can be seen that the volume loss in the AFSD 316L is less than 70% of that in the as-received 4340 substrate during the two-body abrasion, regardless of the difference in hardness. The results also show that the mass loss during the erosion test was comparable for the two materials, differing by less than 3%.
Table 7: Wear rates during abrasive loop contact or erosion by particle impingement. Test locations are defined in Fig. 1.
Material
|
Test Location*
|
ASTM
Method
|
Number of
Tests
|
Average Volume Loss,
mm3
|
Average
Mass Loss,
mg
|
4340
|
W1
|
G174†
|
4
|
1.582
|
N/A
|
316L AFSD
|
W2
|
G174†
|
4
|
1.086
|
N/A
|
316L AFSD
|
W3
|
G174†
|
4
|
1.069
|
N/A
|
4340
|
W1
|
G76‡
|
3
|
N/A
|
9.4
|
316L AFSD
|
W4
|
G76‡
|
3
|
N/A
|
9.2
|
* From Fig. 2
† Abrasion Resistance by Abrasive Loop Contact
‡ Erosion by Solid Particle Impingement Using Gas Jets
Studies have shown a linear relation between hardness and abrasion resistance for steels with similar metallurgical phase compositions; however, recent work has also shown a strong dependence between wear resistance and phase composition during two-body abrasion of steels with the same initial hardness [41,42]. Narayanaswamy, et al., investigated four steels with different microstructures, but all with hardness in the range of 325-360 HV, and reported wear rates in the following order: pearlite < bainite < martensite < tempered martensite [41]. In that work, for an abrasive (SiC) particle size of 58 microns, the relative two-body wear rates of the four phases were found to be 1.00, 1.36, 2.59 and 2.62, respectively. The differences in wear rate were attributed, in part, to the capacity of the two-phase microstructures (pearlite and bainite) to accommodate realignment of the harder carbide phase to the abrasion sliding direction by ductile deformation of the softer component. In addition, the lack of plastic deformation in the martensite and tempered martensite makes those phases more vulnerable to Hertzian fracture on the abrading surface, potentially leading to higher wear rates [43].
It is well established that austenitic stainless steels undergo strain-induced martensitic transformation in a near-surface layer during abrasive wear [44,45,46]. This creates a multi-phase microstructure that allows realignment of the harder martensite towards the abrasive sliding direction, and favors ploughing and wedge formation with narrow/deep surface tracks. In contrast, the tempered martensite of the AISI 4340 resists plastic deformation, thus favoring cutting (material loss) with wide/shallow surface tracks. Fig. 15 shows SEM images of the wear surface from the as-received AISI 4340 and the AFSD 316L, locations W1 and W2 in Fig. 2b. The images show deeper grooves with more plastic deformation and obvious ploughing in the AFSD material, Fig. 15b. In contrast, the surface of the AISI 4340 in Fig. 15a exhibits shallower grooves with less evidence of plastic deformation, which is consistent with cutting as the dominant abrasion mechanism. In this context, it is noteworthy that the mechanisms involving plastic deformation during abrasive wear, ploughing and wedge formation, only deform/translate material, whereas cutting results in material removal. The current results for the two-body abrasive loop test, Table 7, are thus consistent with prior reports of elevated wear resistance in steel microstructures comprised of two-phases, hard and soft (brittle and ductile), relative to microstructures comprised of a single hard phase.
Material loss during erosion by particle impingement is generally thought to occur when the indent formed by an impinging particle creates a protruding rim of plastically deformed material that is later removed as wear debris [47]. In this case the formation of the rim is dependent upon the flow stress and ductility of the target material, and the removal of the rim as debris depends upon the work-hardening of the displaced material. Thus, it is thought that the elongation to failure plays a significant role in the resistance to erosion. In addition, sub-surface investigations on polished 316L have revealed the formation of a nanograined surface layer approximately1 mm in depth, and have suggested that damage accumulation from particle impacts results in the formation of a strain-induced martensitic phase transformation [47]. As a result, the comparable performance of the as-received AISI 4340 and the AFSD 316L in the solid particle impingement test is not surprising.
Three-point bend testing was performed on the multi-layer cladding designated S1 in Fig. 2a, and for comparison on bare 4340 substrate. Fig. 16a shows the test configuration, where the span L is approximately 90 mm and the deflection D is such that the AFSD 316L is on the tensile side of the sample. Fig. 16b shows plots of the flexural stress σf versus flexural strain εf for the bare and clad substrates, where
$${\sigma }_{f}=\frac{3FL}{2b{d}^{2}}$$
4
$${ϵ}_{f}= \frac{6Dd}{{L}^{2}}$$
5
In Eqs. (4) and (5), F, b, and d represent the applied load, sample width, and sample thickness, respectively. The results, which are summarized in Table 8, show that the yield and ultimate strengths of the clad sample are 60% and 73% of the unclad substrate, respectively. This is due to the lower tensile strength of the AFSD 316L relative to the 4340 substrate. The clad bar also shows improvement in the strain to failure of 47%, which is again attributable to the performance of the 316L cladding on the tensile side of the bar. Note from Table 8 that the samples were configured to be similar in thickness, 11.3 vs. 14.1 mm, but with approximately one half (52%) of the clad sample comprising cladding. Thus, the interface between the cladding and substrate is approximately at the neutral axis, neglecting the slight difference in moduli, with the cladding in tension and the substrate in compression. Fig. 16c shows the clad sample after the bend test. Near the bottom of the image, it can be seen that the final AFSD layer has fractured and delaminated. This indicates poor interlayer bonding at this thickness of cladding (7.4 mm).
Table 8: Results of three-point bend testing.
|
4340
Substrate
|
316 AFSD
/ 4340
|
Change,
%
|
Yield Strength, MPa
|
2025
|
1224
|
-39.6
|
Ultimate Strength, MPa
|
2621
|
1912
|
-27.1
|
Strain at failure, %
|
7.6
|
11.2
|
47.4
|
Thickness 4340, mm
|
11.3
|
6.7
|
N/A
|
Thickness 316, mm
|
0
|
7.4
|
N/A
|
Total Thickness, mm
|
11.3
|
14.1
|
N/A
|