Sunteți pe pagina 1din 12

The Mechanism of Brittle Fracture in a Microalloyed Steel: Part I.

Inclusion-Induced Cleavage
D.P. FAIRCHILD, D.G. HOWDEN, and W.A.T. CLARK The cleavage resistance of two microalloyed steels (steels A and B) was studied using several tests, including the instrumented precracked Charpy and Charpy V-notch (CVN) techniques. Ductile-tobrittle transition temperatures were measured for the base-metal and simulated heat-affected zone (HAZ) microstructures. Steel B showed inferior cleavage resistance to steel A, and this could not be explained by differences in gross microstructure. Scanning electron fractography revealed that TiN inclusions were responsible for cleavage initiation in steel B. These inclusions were well bonded to the ferritic matrix. It is believed that a strong inclusion-matrix bond is a key factor in why TiN inclusions are potent cleavage initiators in steel. Strong bonding allows high stresses in a crack/notchtip plastic zone to act on the inclusions without debonding the interface. Once an inclusion cleaves, the strong bond allows for transfer of the TiN crack into the ferritic matrix. It was estimated that only 0.0016 wt pct Ti was tied up in the offending inclusions in steel B. This indicates that extended times at high temperatures during the casting of such steels could produce TiN-related toughness deterioration at even modest Ti contents.

I. INTRODUCTION

SEVERAL authors[14] in the 1950s demonstrated the detrimental effect of Fe3C (cementite) on the cleavage toughness of steel. A few authors[5,6] during this period speculated that cementite was directly responsible for fracture, but they reported such events as special cases and did not identify the exact mechanism of initiation. In 1965, McMahon and Cohen[7] convincingly demonstrated that the cracking of cementite particles located at ferrite grain boundaries represents a primary cleavage initiation mechanism in steel. This established that brittle fracture in steel often begins in or near some minute particle that is harder or more brittle than the surrounding ferrite matrix. Subsequently, this concept contributed to the formulation of the shear-cracking theory, whereby cleavage in steels was observed to initiate at cracking cementite lamellae in pearlite,[2,812] and of the theory that nonmetallic inclusions often act as cleavage initiators in steel weld metals[1316] and in pearlitic steels.[17] Over the past few decades, there has been a steady decrease in the carbon content and the impurity (P and S) level of structural steels, and as a result, typical cleavage initiators like cementite particles and nonmetallic inclusions have been reduced in number and size. This has contributed to improved brittle-fracture resistance. However, despite these advances, three facts virtually guarantee that brittle fracture in steel will always remain a concern. First, because of continuing improvements in structural steels, users have responded by selecting these materials for more-severe service conditions: applications that are less tolerant of oversights or unfortunate mishaps. Second, structural steels have a bcc iron matrix, and, with respect to cleavage resistance,
D.P. FAIRCHILD, formerly Graduate Fellow, The Ohio State University, is Research Specialist with Exxon Production Research Co., Houston, TX 77252. D.G. HOWDEN, Associate Professor, Department of Industrial, Welding, and Systems Engineering, and W.A.T. CLARK, Professor, Department of Materials Science and Engineering, and Associate Dean, Graduate School, are with The Ohio State University, Columbus, OH 43210. Manuscript submitted August 29, 1997.
METALLURGICAL AND MATERIALS TRANSACTIONS A

bcc iron is a borderline material.[18,19] Dislocation flow is essential for cleavage prevention, and, depending primarily on the microstructure, service temperature, and loading rate, dislocations in steel can move with relative ease or with much difficulty. Third, structural steels will always be fusion welded, and this guarantees the presence of microstructures (the weld metal and heat-affected zone (HAZ)) which are typically inferior to the highly processed parent material. By 1980, the study of cleavage in steels was perceived by many as a mature science, and, in general, the capabilities of modern structural steel were in high regard. Then, in 1982, local brittle zones (LBZs) were discovered.*[20] The
*The term discovered is used here to denote detection, definition, and labeling. Low-toughness HAZ regions in steel welds were not new as of 1982. The 1982 contribution was in describing the mechanism of LBZ formation, defining the LBZ microstructure by transmission electron microscopy (TEM), explaining the effect of LBZs on mechanical properties and structural performance, and naming the phenomenon.

LBZs were first identified as small regions of low cleavage resistance in the HAZs of some offshore platform steels.[21] The LBZ issue rekindled interest in cleavage studies, and LBZ-motivated research has led to many interesting results.[2028] One important realization is that certain cleavage-initiation details are still unknown. For example, concerning the martensite island mechanism,[2936] some have speculated that island debonding initiates cleavage,[37,38] while others claim that these islands are not responsible for cleavage at all.[39] The current study investigates the cleavage resistance of a modern, microalloyed steel. Despite a relatively lean chemistry and the use of thermomechanical control processing (TMCP), this steel displayed degraded toughness due to the presence of brittle particles. In Part I of this article, the degraded toughness is quantitatively demonstrated using several toughness tests, and scanning electron fractography is used to identify the cleavage initiation sites. Certain aspects of the offending particles, which appear to make them particularly potent cleavage initiators, are discussed.
VOLUME 31A, MARCH 2000641

Table I. Base Metal Chemistries (Weight Percent) and Strengths (MPa)


Steel
A B Steel A B

C
0.11 0.094 Al 0.031 0.028

Mn
1.53 1.40 V 0.076 0.005

P
0.011 0.008 Nb 0.028 0.001

S
0.001 0.001 Ti 0.011 0.048

Si
0.23 0.31 B 0.0002 0.0002

Cu
0.26 0.011 Ca 0.0045 0.0039

Ni
0.12 0.014 N 0.0041 0.0049

Cr
0.012 0.011 Yield 484 474

Mo
0.011 0.011 Ultimate 584 572

Table II. Gleeble Program Details


Gleeble Program
1 2 3

Cooling Time 800 C to 500 C


6.2 s 34 s 54 s

Cooling Rate at 800 C


86 C/s 15 C/s 9 C/s

In Part II of this article, additional fractographic details and the transmission electron microscopy (TEM) results are combined with aspects of existing dislocation-based cleavage models to develop a specific model of cleavage initiation in this steel. II. EXPERIMENTAL MATERIALS AND PROCEDURES A. Materials Two commercial X-65 linepipe steels (steels A and B) were investigated. Steel B is the primary focus of the study (toughness degradation is observed), while steel A is included for comparison purposes. Both steels were produced by the same manufacturer to the same specification, and, therefore, the chemistries and microstructures are similar. The chemistries and tensile properties are listed in Table I. Steels A and B were acquired as short pipe sections with wall thicknesses and diameters of 13 to 19 mm and 510 to 610 cm, respectively. Both steels were manufactured using TMCP. B. HAZ Simulations Three different HAZ microstructures were produced using a Gleeble thermal-cycle simulator. The thermal programs were derived from Reference 40. Thermal-cycle details are given in Table II and Figure 1. All cycles were coarse-grain HAZ (CGHAZ) simulations with a peak temperature of 1350 C. The cycles represent several welding techniques relevant to either pipe manufacture or pipeline construction. Gleeble program 1 simulates a girth weld produced by the shielded metal arc technique. Gleeble program 2 simulates a longitudinal seam weld produced by the double submerged arc technique. Gleeble program 3 simulates a flash-butt girth weld. The Gleeble programs are ordered by numbers 1 through 3, according to increasing heat input (decreasing cooling rate). The Gleeble specimens were machined with the long dimension parallel to the pipe axis. Thermal cycling was conducted with narrow jaw spacings to simulate actual HAZ widths (the spacing defines about two HAZ widths). Spacings of 9, 12, and 15 mm were used for thermal cycles 1
642VOLUME 31A, MARCH 2000

Fig. 1HAZ thermal cycles.

through 3, respectively. Gleeble simulations were conducted in air, and the oxide layer was not removed. This short-cut has been proven to not affect toughness.[38] C. Microstructural Characterization Samples of each base metal and of each steel/Gleeble combination were optically examined at magnifications ranging from 100 to 1000 times. Cross sections were oriented such that the line of sight was parallel to the pipe hoop direction (transverse to the steel rolling direction). Samples were prepared using standard metallographic techniques and etched with 2 pct Nital. Scanning electron microscopy (SEM) was used to investigate the second-phase carbide morphology within the base metals. This work was conducted on a Hitachi S510 scanning electron microscope, and the samples were gold sputtered. The prior-austenite grain size in the Gleeble microstructures was measured using the line-intercept method on large montage micrographs. Each montage was constructed by connecting six standard prints into a two-by-three matrix. Montages were generated at 200 and 400 times magnification. As is often the case for Gleeble-simulated CGHAZs, no special techniques beyond Nital etching were necessary to reveal the prior-austenite grain boundaries (this will be apparent in the optical micrographs shown later). The basemetal ferrite grain size was also measured using the lineintercept method; however, individual micrographic prints were used instead of montages. Vickers microhardness (200 g and 1 kg) traverses spanning the 10-mm specimen width were conducted on the optical microscopy samples. Hardened specimens were prepared for comparison to the microhardness data. A small
METALLURGICAL AND MATERIALS TRANSACTIONS A

piece (10 mm3) of each steel was austenitized at 940 C, held for about 25 minutes, and quenched in water. D. Quantitative Microscopy: Inclusions As will be shown subsequently, titanium nitride (TiN) inclusions were found to control the cleavage resistance of steel B. Therefore, the optical appearance of these inclusions was studied, and an attempt was made to quantify their number per unit volume in both steels. These inclusions were not readily observed in the etched condition, as the microstructure created visual background noise. The optical microscopy sections were repolished and examined unetched. The general problem of relating counts made on a random plane section to the number of particles in the three-dimensional volume has apparently not been solved;[41] however, methods for some special cases have been developed. If the particles are assumed to be same-sized spheres, Eq. [1] relates the number of particles per unit volume (NV) to the number of particles per unit area observed in the cross section (NA) and the number of particles intercepting a random lineal traverse (NL). The assumption concerning spherical shape is not precise for the cuboidally shaped TiN inclusions; however, a useful approximation can still be obtained. NV N2 A 4NL
[1]

Fig. 2Instrumented precracked Charpy specimen geometry.

The value of NA was estimated for steel B by examining approximately 450 fields of view at 400 times magnification. The number of inclusions per field was counted, averaged, and divided by the field-of-view area. For steel A, this same procedure was used, except that approximately 100 fields of view were examined. The lineal intercepts were determined using an image analyzer, whereby each field of view was transferred to the computer environment and a horizontal line grid was superimposed. The lines were spaced about 4.4 m apart. The number of inclusion intercepts was counted for approximately 100 fields of view. The counts were averaged and divided by the total line length. The average TiN inclusion size in steel B was estimated using the image analyzer. Numerous adjacent fields of view were enlarged on the computer monitor by about 1000 times. The lengths and widths of all inclusions within each field of view were measured using a mouse-operated, pointand-click measuring tool. Approximately 175 inclusions were sized. E. Fracture Testing For each steel in the as-received condition and in combination with each Gleeble program, 12 to 24 instrumented precracked Charpy (IPC) specimens were tested over a range of temperatures to produce ductile-to-brittle transition curves (eight curves total). The IPC specimens were prepared for precracking by placing a 2.5-mm-deep, 0.20-mm-wide surface notch at the location of interest by wire electrodischarge machining. For the Gleeble specimens, this location was coincident with the thermocouple spot weld. Fatigue precracking was conducted in three-point bending using a Fatigue Dynamics Inc. precracker with automatic shutdown control. This machine was programmed to shut off once the precrack propagated to about 50 pct of the specimen
METALLURGICAL AND MATERIALS TRANSACTIONS A

thickness. With respect to starting and finishing loads, precracking was conducted according to British Standard 5762 guidelines for crack-tip opening displacement testing. Figure 2 shows the final specimen geometry. The IPC testing was conducted using a Dynatup drop tower. The drop load was 227 kg, the drop height was 152 mm, and the tup velocity upon contacting the specimen was about 1.68 m/s. The specimens were cooled in a commercial, recirculating cooling bath for a minimum of 30 minutes prior to testing. The bath temperature was controlled to within 0.5 C using a thermocouple embedded inside of a dummy Charpy specimen that was placed in the bath adjacent to the test specimens. Each broken specimen was examined under a stereo microscope to determine the fatigue crack length and thus, the size of the fracture area. The crack length was measured at five equally spaced locations and averaged. A 95 pct confidence interval for the length of all fatigue cracks was 5.13 0.03 mm. The total fracture energy for each specimen was divided by the fracture area (normalized to give J/cm2) to minimize the effect of fatigue crack length variations. Although both initiation and propagation energies were determined for each specimen, the total energy will be presented as the primary measure of IPC toughness. Each set of transition-curve data was statistically curve fitted using a hyperbolic tangent model and the nonlinear regression routine of Statistical Analysis Software.[38,42,43] The ductile-to-brittle transition temperature (DBTT) was that temperature denoting the point on the curve halfway between the upper and lower shelves. The DBTT was used as the primary measure of cleavage fracture resistance. Standard Charpy V-notch (CVN) tests were conducted for both steels in combination with Gleeble programs 1 and 2. Approximately 15 specimens were tested for each steel/ Gleeble program combination to produce full transition curves. These data were statistically modeled using the same method as for the IPC data. Double-fatigue-crack, four-point bend tests were conducted for both steels in combination with Gleeble programs 1 and 2. For each case, approximately six specimens were tested at various temperatures ranging from 80 C to 10 C. The specimen geometry and loading configuration are shown schemetically in Figure 3. The Gleebled regions and the precracks were produced as described previously for the IPC specimens. Four-point loading to fracture was conducted using a screw-driven load frame, and the crosshead speed was 1.07 mm/min. The double-crack, four-point bend tests were conducted primarily to study the micromechanisms of
VOLUME 31A, MARCH 2000643

Fig. 3Double crack, four-point bend specimen geometry.

cleavage fracture in the HAZ microstructures, and this subject will be addressed in Part II of this article. For Part I, these double-crack specimens provide (1) fracture-toughness data for comparing the cleavage resistance of the steels and (2) fracture surfaces for SEM examination. F. Fractography Visual inspection of the broken toughness specimens was conducted to see if the fracture morphology differences between steels A and B could be distinguished by the naked eye. A machinists rule and a handheld 10 times magnifying glass were used. Scanning electron fractography was conducted for the purpose of locating and inspecting cleavage initiation sites, and the specimens examined were those tested at temperatures below the DBTT. The area within several hundred microns of the notch/crack tip was examined using the cleavage river patterns as a road map. This work was conducted using a PHILIPS* XL 30 scanning electron
*PHILIPS is a trademark of Philips Electronic Instruments Corp., Mahwah, NJ.

microscope accompanied by a DX-4 EDAX (Energy Dispersive Analytical X-Ray) system. Secondary electron imaging was the primary mode of operation and was conducted at 20 kV. The study of inclusion composition using energydispersive spectroscopy (EDS) was conducted at 10 kV to minimize interaction volume sampling of adjacent and underlying regions. III. EXPERIMENTAL RESULTS A. Microstructural Characterization The optical micrographs are shown in Figure 4. The base metals show alternating light and dark microbands, and the grains are elongated as a result of TMCP rolling. The light phase is ferrite, but resolution of the dark, carbide-containing areas required SEM examination. The carbide-containing areas were found to be pearlite (Figure 5). The HAZ microstructures created by Gleeble program 1 appeared to be about 75 pct bainite and 25 pct lath martensite. Although no attempts were made to quantify the exact proportions of upper and lower bainite, the presence of both morphologies was confirmed by TEM, and upper bainite was identified as the dominant morphology ( 80 to 95 pct). Proeutectoid ferrite was essentially absent in Gleeble 1 microstructures, which is likely a result of the Mn content[44] and the fast cooling rate of the shielded metal arc simulation. The microstructures created by Gleeble program 2 are notably coarser than those for program 1 (due to slower
644VOLUME 31A, MARCH 2000

Fig. 4Optical micrographs.

cooling and increased time at high temperatures), and no martensite is present. Steel B displayed no proeutectoid ferrite at the prior-austenite grain boundaries, while very small
METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 5Scanning electron micrograph of the base metal, second phase in steel B. Fig. 7Microhardness data.

Fig. 8Optical images of TiN inclusions.

B. Inclusion Quantification
Fig. 6Prior austenite grain sizes.

amounts were found in steel A. Steel A showed a small amount of acicular ferrite, but upper bainite was again the primary constituent in both steels. The microstructures created by Gleeble program 3 showed only a slight increase in the amount of grain-boundary ferrite over that created by program 2. This was somewhat surprising, considering the very slow cooling rate of Gleeble program 3. The gross microstructure in both steels showed a mix of upper bainite, Widmanstatten ferrite, and acicular ferrite. Overall, the HAZ microstructures in steels A and B for all Gleeble simulations appeared very similar. The base-metal ferrite grain sizes for steels A and B were 3.9 and 5.2 m, respectively. These small grain sizes were produced by a combination of (1) austenite grain-boundary pinning by microalloy-related precipitates, (2) the rolling sequences during TMCP, and (3) the accelerated cooling at the finish of TMCP. The prior-austenite grain sizes for the Gleeble simulations are shown in Figure 6. For all Gleeble programs, steel A had smaller grains than steel B. This difference is believed to be real, because the linear-intercept technique was applied until a statistically significant number of boundary intercepts were generated. The microhardness results are shown in Figure 7, and, generally, the hardness decreased with decreasing cooling rate. Steel A produced a higher microhardness than steel B in all conditions. The differences were most marked for the cases where fast-to-moderate cooling rates were applied; i.e., for the quenched condition and for Gleeble programs 1 and 2. This can be explained by noting that steel A has a slightly higher carbon and total alloy content.
METALLURGICAL AND MATERIALS TRANSACTIONS A

Although TiN inclusions existed in steel A, there were too few to conduct quantitative work, and Nv could not be calculated. The main problem was counting lineal intercepts. Only seven intercepts were obtained, for a total line length of 560 mm. Too many fields of view yielded zero intercepts. The value of NA , however, was determined and found to be 6.4/mm2. The values of NL , NA , and NV for steel B were determined to be 0.24/mm, 47.5/mm2, and 7263/mm3, respectively. The average inclusion size in steel B was 1.56 m, and a 95 pct confidence interval spanned from 1.42 to 1.71 m. Several large TiN inclusions from steel B are shown in Figure 8. The TiN inclusions appeared yellow-gold in color, and about 30 pct contained dark spots. The EDS results revealed that these spots are probably oxides, calcium aluminates, spinels, or sulfides. It can be surmised that many TiN inclusions nucleate on these slag particles in the liquid steel. C. Fracture Results An IPC transition curve, including the data points, is shown in Figure 9 to demonstrate the typical level of IPC data scatter experienced in this investigation. Most tests were conducted at temperatures in the transition regime to increase the accuracy of DBTT determination. The magnitudes of the upper and lower shelves were known due to prior IPC testing reported elsewhere.[38] To aid visual clarity, the remainder of the toughness transition curves will be presented without the data points. The base-metal and Gleeble simulation IPC transition curves are shown in Figure 10. The transition temperatures are compared in Figure 11. Steel B consistently displays higher transition temperatures, which is indicative of lower cleavage resistance.
VOLUME 31A, MARCH 2000645

Fig. 12CVN transition curves. Fig. 9Example IPC transition curve with data points.

was the case for the IPC results, steel B consistently displays a higher DBTT, which is indicative of lower cleavage resistance. During initial CVN testing, it was found that the upper-shelf specimens of both steels stopped the hammer, absorbing all 325 J of available energy. Subsequently, several specimens were reduced to two-thirds of their width and tested at upper-shelf temperatures. All subsize specimens were expelled from the anvil, but none broke into two pieces (several appeared to be dragged through the anvil). Although the subsize tests provide an estimate of upper-shelf energies, these values cannot be considered exact. From the double-crack, four-point bend tests, values of stress intensity (KI) were calculated using Eq. [2] and a value of f(a/W ) 1.5 for pure bending.[45] KI
Fig. 10IPC transition curves.
max

max

a)f (a/w)

[2]

maximum bending stress,

where a f(a/w) crack length, and function of geometry.

Fig. 11IPC transition temperatures.

The CVN transition curves are shown in Figure 12, and the transition temperatures are compared in Figure 13. As
646VOLUME 31A, MARCH 2000

Fig. 13CVN transition temperatures.

METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 15IPC fracture surface appearance (Gleeble program 2, test temperature 32 C).

Fig. 14Double crack, four-point bend test data.

Plane-strain validity was checked using Eq. [3]. The yield strengths used in Eq. [3] were measured on Gleebled tensile samples tested at 20 C.[38] t or a
where t
ys

2.5

KI
ys

[3]
Fig. 16Steel A and B CVN specimens showing a distinct difference in fracture behavior.

specimen thickness, and yield strength. E. SEM Fractography A number of IPC, CVN, and double-crack, four-point bend fracture surfaces were examined by SEM. Tracing cleavage river patterns to the fracture origins was arduous. On a microscale, fracture did not necessarily travel in a logical direction (away from the notch/crack tip or away from the initiation site). Cleavage appeared to wind its way through the microstructure, maneuvering around high-angle boundaries and first cleaving areas favorably oriented to the local stress state. This hypothesis was honed primarily by examination of the Gleeble specimens rather than the basemetal specimens. The HAZ microstructures produced coarse cleavage features that were easier to interpret. The Gleeble specimens often displayed large (30 to 80 m) facets that were assumed, for two reasons, to be either individual prioraustenite grains or individual bainitic packets. First, the largest facets were equal in size to the prior-austenite grains and the smaller facets were equal to either the smallest prioraustenite grains or the largest packets. Second, the facet boundaries were the most prominent ridge-like features on the fracture surfaces and, in CGHAZs, the most prominent high-angle boundaries in the microstructure are either the prior-austenite grain boundaries or, in the case of lath microstructures like martensite or bainite, the packet boundaries. From each cleavage origin, several fracture paths emanated radially. By comparing the cleavage patterns of several paths over a distance of a few hundred microns, it appeared that the ease of cleavage propagation varied among the paths. Some paths showed very flat surfaces with relatively minor ridges between facets. Others showed a more tortuous morphology, with many minute ledges within the facets and
VOLUME 31A, MARCH 2000647

The yield strengths for steel A, processed with Gleeble programs 1 and 2, were 883 and 690 MPa, respectively. The yield strengths for steel B, processed with Gleeble programs 1 and 2, were 910 and 669 MPa, respectively. Due to the small cross section of the double-crack specimens (10 10 mm), only one test met plane-strain validity. The one valid KIC test was recorded on a specimen of steel B, processed with Gleeble program 1, and tested at 69 C. The temperature range for the double-crack, four-point bend tests was insufficient to generate full transition curves. The KI data are shown in Figure 14. The four-point bends tests indicate that steel B has lower cleavage resistance than steel A. D. Visual Fractography Observations concerning the lower cleavage resistance of steel B vs steel A were possible by visual inspection. An example is shown in Figure 15, which is a schematic of the fracture surfaces of two IPC specimens processed using Gleeble program 2 and tested at 32 C. Steel A was within the upper transition, while steel B was within its lower transition. Steel A absorbed 131 J/cm2 of energy, while steel B absorbed 60 J/cm2. The fracture surface of steel A was dominated by microvoid coalescence, while the fracture surface of steel B was dominated by cleavage. Figure 16 shows a photograph of two CVN specimens that were processed with Gleeble program 1 and tested at 7 C. The specimen for steel A stopped the hammer, while the specimen for steel B failed by cleavage and absorbed 11 J of energy.
METALLURGICAL AND MATERIALS TRANSACTIONS A

prominent ridges between facets. It is believed that the fracture propagation rate varied along different paths and, therefore, the material located at points equidistant from the fracture origin did not cleave at precisely the same instant in time. Considering this time element was useful when examining new crack-front regions for initiation sites. The tortuous paths were often ignored, as the paths of easy propagation pointed more directly to the fracture origin. No distinct initiation features were found in steel A. Steel B, however, consistently showed broken, cuboidal, or angular-shaped inclusions at fracture origins. All steel B fracture faces displayed multiple inclusion-related initiation sites along the crack front. Typically, these sites were no further than 200 m from the crack/notch tip; most were about 100 m from the tip. When an inclusion-related initiation site was discovered in one fracture face, the analogous location in the mating face was traced and inspected. In all cases, both sites were found to contain mating inclusion halves. Using EDS analysis (presented subsequently), the inclusions were tentatively identified as TiN. The TiN hypothesis was confirmed using microdiffraction in the transmission electron microscope, and this work will be presented in Part II of this article. Figures 17(a) through (d) show a TiN-related initiation site from a double-crack, four-point bend specimen processed using Gleeble program 1 and tested at 0 C. In Figure 17(a), the bright, v-shaped feature near the top of the image is the fatigue-crack tip. In the center of the image, about 70 m below the crack tip, there is a diamond-shaped feature that is centrally located with respect to the cleavage river patterns. Figure 17(b) shows this feature at higher magnification. It is a fractured TiN inclusion, and Figure 18 shows the EDS spectrum from this particle. Figures 17(c) and (d) are high-magnification images of the mating TiN halves. It appears from Figure 17(d) that cleavage initiated within the inclusion along its lower-right edge (oriented at 45 deg). Some small facets can be seen in the TiN at this location. The crack then propagated toward the eleven oclock position, and, thus, the initial direction of travel was toward the fatigue crack. At the left-most corner of the TiN (in Figure 17(d)), the crack did not penetrate the matrix, but did so at the uppermost corner. After it had penetrated about 2 m into the matrix, the crack spread radially in all directions. Despite experiencing a very stressful event, this TiN inclusion remained well bonded to the matrix; no voids or separations are present along the inclusion-matrix interface. Figures 19(a) through (d) show a TiN-related initiation site from a base-metal, IPC specimen tested at 115 C. This initiation event involved two TiN inclusions (arrows in Figure 19(a)). The fatigue-crack tip (out of view) is positioned 112 m above these particles. Figure 19(b) shows the right-hand-side TiN, and the local river patterns in the ferrite matrix clearly emanate from this particle. Figures 19(c) and (d) are high-magnification images of the mating TiN halves, and several features are worth noting: (1) a number of sidecracks formed in the particle, (2) while some river patterns exist on the particle fracture surfaces, the bulk of the surfaces are smooth, (3) the particle did not debond from the matrix, and (4) an empty cavity exists in the particle center. The cavity is the former site of a spherically shaped, nonmetallic inclusion that was expelled during the fracture process. Figure 20 shows the EDS spectrum from the TiN
648VOLUME 31A, MARCH 2000

in Figure 19(b), and the Mg peak is likely related to remnants of the lost nucleus. Although TiN-related initiation sites were found in the base-metal specimens of steel B, they were less common than in the Gleeble specimens. Many base-metal initiation sites contained no distinct features. IV. DISCUSSION A. The Cleavage Mechanism in Steel B In the base-metal condition and for all Gleeble simulations, steel B displayed lower cleavage resistance than steel A. This trend was confirmed using several test methods and was even discernible with the naked eye through visual fractography. Early in this study, considerable effort was invested to determine if steel Bs lower cleavage resistance could be attributed to factors previously reported in the literature:[21,46] i.e., microalloy precipitation, upper-bainitic microstructure, high-carbon martensite islands (martensiteaustenite constituent), large prior-austenite grains, and free nitrogen. It was found that these factors could not explain steel Bs performance, as its chemistry, grain size, and microstructure were very similar to steel A. With respect to hardness, the results seemed opposite to the toughness trend. Normally, within the same class of steel, a lower hardness would be associated with higher toughness. However, in this study, steel B possessed both lower hardness and lower toughness. One clue concerning the controlling cleavage mechanism in steel B was found by noting that a wide variety of microstructures were tested. For both steels, the base metals were fine-grained ferrite pearlite, while the three HAZs contained mixtures of martensite, bainite, grain-boundary ferrite, Widmanstatten ferrite, and acicular ferrite. Although steels A and B displayed similar structures when similarly processed, it is important to note that a broad range of microstructures were generated by the processing conditions. One might ask, How can steel B consistently display lower cleavage resistance than steel A over such a broad microstructural range?; or, How can steel B display lower cleavage resistance than steel A when their microstructures are the same? The most likely answer is that the controlling cleavage mechanism is not related to the gross microstructure and that the culprit is a constant. The cleavage mechanism in steel B was determined using scanning electron fractography by tracing cleavage river patterns to their origins and finding broken TiN, inclusions. Attributing cleavage initiation in steel B to these inclusions is consistent with a mechanism that is a constant and not dependent on gross microstructure, because large TiN particles (for example, those larger than about 0.5 m) in steel form in the liquid during casting and are present in all conditions subsequent to casting. Once formed, these inclusions are thermally stable up to, or slightly beyond, the steels melting point and, thus, remain unaltered in all microstructural conditions.[4750] B. TiN Inclusions as Cleavage Initiators Although the fractographic results for only two initiation sites were shown previously, about 25 cases were examined in the SEM. The fact that broken inclusion halves were
METALLURGICAL AND MATERIALS TRANSACTIONS A

(a)

(b)

(c)

(d )

Fig. 17(a) Scanning electron fractograph of TiN initiation site in steel B. Note the diamond-shaped feature (the TiN) located near the image center. (b) Higher-magnification image of the cleavage origin from (a). (c) Higher-magnification image of the TiN inclusion shown in (b). This figure mates to (d). (d ) Mating fractograph to (c).

Fig. 18EDS spectrum from the inclusion in Fig. 17(b). Peaks are K unless otherwise noted.

found in all mating fracture faces indicates that, during the initiation event, the TiN inclusions break into two pieces as opposed to debonding from the ferrite. The inclusion-matrix bond in the TiN-ferrite system is apparently very strong. This observation rules out an initiation mechanism where, in the plastic zone of the crack/notch tip, the inclusion slowly debonds from the matrix, creating a microvoid that grows to become a critical Griffith flaw. Such debonding has been proposed to be a key element of the cleavage initiation mechanism from martensite islands in weld HAZs.[37,38] The strong inclusion-matrix bond creates two phenomena that are responsible for TiN inclusions being potent cleavage initiators in steel. First, because the inclusion-matrix interface remains intact under load, high stresses can be transferred from the matrix to the inclusion. In the plastic zone of a crack or notch tip, these stresses ultimately reach a critical level and the TiN cleaves. TiN has a B1 (NaCl)type structure and is inherently brittle, as are most refractory compounds.[51,52] This brittleness is demonstrated in the previous fractographs by the relatively flat, featureless fracture
VOLUME 31A, MARCH 2000649

METALLURGICAL AND MATERIALS TRANSACTIONS A

(a)

(b)

(c)

(d )

Fig. 19(a) Scanning electron fractograph of a TiN initiation site in a steel B, base metal IPC specimen. The arrows denote fractured TiN inclusions. (b) Higher-magnification image of the right-hand TiN cleavage origin from (a). (c) Higher-magnification image of the TiN inclusion shown in (b). This figure mates to (d). (d ) Mating fractograph to (c).

Fig. 20EDS spectrum from the inclusion in Fig. 19(b). Peaks are K unless otherwise noted.
650VOLUME 31A, MARCH 2000

surfaces within the TiN inclusions and by the occurrence of sidecracking. Second, the strong interface allows the TiN crack to transfer from the inclusion, across the inclusionmatrix interface, and into the ferrite. If the inclusion-matrix bond were weak, the crack might arrest at the interface and never enter the ferrite. Because steels A and B were manufactured to the same specification and had similar microstructures, it appears that the TiN inclusions in steel B were responsible for significant transition temperatures shifts (reduced cleavage resistance) relative to steel A. Specifically, the following occurred: (1) a DBTT shift of about 30 C for the base metal, as measured by the IPC test; (2) DBTT shifts of about 40 C to 70 C for simulated HAZs as measured by the IPC test; and (3) DBTT shift of about 50 C to 60 C for simulated HAZs, as measured by the Charpy test. The larger shifts for the HAZs as compared to the base metals are believed to be caused by the coarse HAZ microstructures. Coarseness creates longer and more intense dislocation pileups, which
METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 21A graph showing how the mount of Ti tied up as TiN varies with inclusion size.

are believed to be an important driving force in cracking the TiN inclusions (in analogy to pileups being a driving force in cracking cementite in steels[53] and nonmetallic inclusions in weld metals[1316]). While the base metals had ferrite grain sizes ranging from 4 to 5 m, the HAZ prioraustenite grains ranged from 52 to 95 m. This explains the sensitivity of weld HAZs to the presence of potent cleavage initiators like TiN inclusions. C. The Amount of Ti Tied Up as TiN Inclusions Because the TiN inclusions in steel B greatly affect its cleavage resistance, and because steel B contains an appreciable aount of Ti as compared to most microalloyed steels, an estimation was made of the amount of Ti in steel B tied up as TiN inclusions. This calculation was based on a hypothetical 1 mm3 volume of steel. Using the quantitative microscopy results, it was assumed that this volume contained 7263 TiN inclusions for each calculation, each being a 1.6 m cube. The following list describes the quantity produced step: the volume of a TiN unit cell (a 4.24 A), the volume of a 1.6 m TiN inclusion, the number of unit cells per inclusion, the number of Ti atoms per inclusion (4 atoms per unit cell), the number of Ti atoms per cubic millimeter of steel, the weight of these Ti atoms, the weight percent these atoms comprise in the steel, and the percent of Ti tied up in the 7263 inclusions. This calculation yielded a value of 0.0016 wt pct of Ti tied up in the inclusions, which represents 3.3 pct of the total Ti content of steel B (0.048 wt pct). The 3.3 pct estimate cannot be considered exact for several reasons. First, it was assumed that each inclusion was a pure TiN cube, but many inclusions contained slag-particle nuclei that occupied a significant volume. Second, the premise of a perfectly stacked array of unit cells ignores the inevitable presence of vacancies, dislocations, grain boundaries, impurities, etc. Third, the quantitative microscopy exercise was based on the assumption that all inclusions were same-size spheres, and Figure 8 shows that this is not the case. In light of these possible inaccuracies, Figure 21 was produced. This graph shows how the amount of Ti tied up in TiN inclusions varies with inclusion size. Three lines are plotted, one for
METALLURGICAL AND MATERIALS TRANSACTIONS A

the calculation just reviewed (7263 inclusions per mm3), and two for the cases of 40 pct from this inclusion level. The 40 pct value was an arbitrarily selected worst-case error. Figure 21 shows that, in the range of inclusion sizes consistent for steel B, even large errors are tolerated while still supporting the main conclusion that the amount of Ti in the TiN inclusions is small (2 to 6 pct of the total Ti contained in steel B). This implies that cleavage initiation by TiN inclusions is possible for relatively modest Ti contents. The degree to which these inclusions are present will depend on the time spent at very high temperatures (i.e., on the casting schedule) and on the Ti content of the steel, which influences the driving force for TiN formation. Steels that experience longer times in the liquid state and/or have higher Ti contents will be more likely to suffer reduced toughness as a result of TiN-initiated cleavage. The casting schedule for steels A and B was not available to the authors; however, because steel A contained so few TiN inclusions despite being Ti microalloyed, it is anticipated that steel A spent less time in the liquid state than did steel B. This processing difference is believed to be one of the primary factors in determining the toughness variation between the steels. One variable pertinent to the number of TiN inclusions that form is the nitrogen content. A calculation was made to approximate the amount of nitrogen in steel B tied up as TiN. This yielded 4.7 ppm of nitrogen tied up in the TiN inclusions. In other words, only 10 pct of the total nitrogen in steel B is estimated to be tied up in the TiN inclusions. The amount of nitrogen contained in steels A and B was essentially the same, 41 and 49 ppm, respectively. Although relatively low by structural steel standards, this nitrogen level was apparently more than ample to create TiN-related toughness degradation in steel B. Again, this indicates that the ingredients for toughness degradation are present in many Ti microalloyed steels; however, whether the ingredients act in a deleterious way depends on the steel processing route. It is concluded that two key factors contributed to the toughness degradation of steel B vs steel A: (1) steel B had a higher Ti content (0.048 vs 0.011 wt pct), and (2) it is believed that steel B spent more time in the liquid state during casting. V. CONCLUSIONS The cleavage-fracture resistance of two microalloyed TMCP steels (steels A and B) was investigated. Fracture tests were conducted using base-metal and simulated HAZ specimens. Visual and scanning electron fractography were used to study the cleavage initiation mechanisms. 1. In all microstructural conditions and for several different toughness test methods, steel B displayed inferior cleavage resistance to steel A. The DBTTs of steel B, as measured by the instrumented precracked Charpy and standard CVN tests, were 35 C to 70 C higher than those for steel A. 2. The chemistries and microstructures of steels A and B were similar. The factors which are commonly associated with reduced cleavage resistance in the CGHAZs of structural steels (large prior-austenite grains, upper bainite, microalloy precipitation, martensite islands, and free nitrogen) could not explain steel Bs inferior toughness.
VOLUME 31A, MARCH 2000651

3. Scanning electron fractography revealed no distinct features at the cleavage origins in steel A, but broken TiN inclusions were found at the cleavage initiation sites in steel B. It was concluded that these brittle particles were responsible for steel Bs inferior toughness. By quantitative microscopy, it was found that TiN inclusions were essentially absent in steel A, but abundant in steel B. 4. Fractographic details showed that the TiN inclusions were brittle and well bonded to the ferrite matrix. Within the plastic zone at a notch or crack tip, the strong inclusionmatrix bond allows matrix stresses to be transferred to the inclusion. Once a TiN inclusion cleaves, this continuous interface permits the TiN crack to transfer from the inclusion, across the interface, and into the ferrite. The strong inclusion-matrix bond is a key feature that makes TiN inclusions potent cleavage initiators in steel. 5. Steel B contained 0.048 wt pct Ti, but only 0.0016 wt pct (3.3 pct of the total) was estimated to be tied up in the offending inclusions. This implies that a casting schedule that dwells at high temperatures for long periods could generate these deleterious inclusions at even modest Ti contents (although higher Ti contents will make matters worse).

ACKNOWLEDGMENTS This article is based on a portion of DPFs Ph.D. Dissertation and was funded by an American Welding Society Fellowship. Edison Welding Institute is acknowledged for occasional use of their facilities. Conversations with Drs. Malcolm Gray, Claire Davis, Peter Harrison, and Shuji Aihara are sincerely appreciated. REFERENCES
1. C. Crussard, R. Borione, J. Plateau, Y. Morillon, and F. Maratray: J. Iron Steel Inst., 1956, June, pp. 146-77. 2. W.H. Bruckner, W.P. Rees, B.E. Hopkins, and H.R. Tipler: Weld. J., 1950, Sept., pp. 467s-476s. 3. N.P. Allen, W.P. Rees, B.E. Hopkins, and H.R. Tipler: J. Iron Steel Inst., 1953, June, pp. 108-20. 4. J.C. Danko and R.D. Stout: Trans. ASM, 1957, vol. 49, pp. 189-200. 5. N.J. Petch: J. Iron Steel Inst., 1953, May, pp. 25-28. 6. A.H. Cottrell: Trans. TMS, 1958, Apr., pp. 192-03. 7. C.J. McMahon and M. Cohen: Acta. Metall., 1965, vol. 13, pp. 591-04. 8. A.R. Rosenfield, E. Votava, and G.T. Hahn: Trans. ASM, 1968, vol. 61, pp. 807-15. 9. L.E. Miller and G.C. Smith: J. Iron Steel Inst., 1970, Nov., pp. 998-05. 10. Y.J. Park and I.M. Bernstein: Metall. Trans. A, 1979, vol. 19A, pp. 1653-64. 11. J.J. Lewandowski and A.W. Thompson: Metall. Trans. A, 1986, vol. 17A, pp. 1769-86. 12. J.J. Lewandowski and A.W. Thompson: Acta. Metall., 1987, vol. 35, pp. 1453-62. 13. J.H. Tweed and J.F. Knott: Met. Sci., 1983, vol. 17, pp. 45-54. 14. J.H. Tweed and J.F. Knott: Acta Metall., 1987, vol. 35, pp. 1401-14. 15. J.H. Tweed and J.F. Knott: Met. Constr., 1987, vol. 19, pp. 153-58. 16. P. Bowen and J.F. Knott: Proc. 7th Int. Conf. on Offshore Mechanics and Arctic Engineering, Houston, TX, ASME, Golden, CO, 1988, pp. 503-07. 17. D.J. Alexander and I.M. Bernstein: Metall. Trans. A, 1989, vol. 20A, pp. 2321-35. 18. A. Kelly, W.R. Tyson, and A.H. Cottrell: Phil. Mag., 1967, vol. 15, pp. 567-87. 19. E. Smith: Conf. on Mechanics and Physics of Fracture, Metals Society and Institute of Physics, Cambridge, United Kingdom, 1975, pp. 47-56.

20. D.P. Fairchild: Exxon Production Research Co., Houston, TX, J.Y. Koo: Exxon Corporate Research Laboratories, Clinton, NJ, private communication, 1982. 21. D.P. Fairchild: Welding Metallurgy of Structural Steels, Denver, CO, J.Y. Koo, ed., TMS, Warrendale, PA, 1987, pp. 303-18. 22. C.P. Royer: Welding Metallurgy of Structural Steels, Denver, CO, J.Y. Koo, ed., TMS, Warrendale, PA, 1987, pp. 255-62. 23. C. Thaulow, A.J. Paauw, and K. Guttormsen: Welding J., 1987, vol. 66 (9), pp. 266-79. 24. P.L. Harrison and P.H.M. Hart: Proc. Int. Conf. on Weld Failures, The Welding Institute, London, 1988, pp. 185-201. 25. P.H. Bateson, S.E. Webster, and E.F. Walker: Proc. 7th Int. Conf. on Offshore Mechanics and Arctic Engineering, Houston, TX, ASME, Golden, CO, 1988, pp. 257-65. 26. S.E. Webster and E.F. Walker: Proc. 7th Int. Conf. on Offshore Mechanics and Arctic Engineering, Houston, TX, ASME, Golden, CO, 1988, pp. 395-03. 27. T.M Scoonover: Proc. Int. Conf. on The Metallurgy, Welding, and Qualification of Microalloyed (HSLA) Steel Weldments, Houston, TX, Nov. 68, 1990, pp. 276-305. 28. P.L. Harrison, D.J. Abson, A.R. Jones, and D.J. Sparkes: in Quantitative Methods in Fractography, ASTM STP 1085, B.M. Strauss and S.K. Putatunda, eds., ASTM, Philadelphia, PA, 1990, pp. 102-22. 29. H. Ikawa, H. Oshige, and T. Tanoue: Trans. Jpn. Welding Soc., 1980, vol. 11 (2), pp. 87-96. 30. J.H. Chen, Y. Kikuta, T. Araki, M. Yoneda, and Y. Matsuda: Acta Metall., 1984, vol. 32 (10), pp. 1779-88. 31. J.Y. Koo and A. Ozekcin: Welding Metallurgy of Structural Steels, Denver, CO, J.Y. Koo, ed., TMS, Warrendale, PA, 1987, pp. 119-35. 32. S. Aihara and T. Haze: TMS Annual Meeting, Phoenix, AZ, Jan. 2528, 1988, paper no. A88-14. 33. S. Aihara and K. Okamoto: Proc. Int. Conf. on The Metallurgy, Welding, and Qualification of Microalloyed (HSLA) Steel Weldments, Houston, TX, Nov. 68, 1990, pp. 402-26. 34. B.C. Kim, S. Lee, N.J. Kim, and D.Y. Lee: Metall. Trans. A, 1991, vol. 22A, pp. 139-49. 35. S. Lee, B.C. Kim, and D. Kwon: Metall. Trans. A, 1992, vol. 23A, pp. 2803-16. 36. S. Lee, B.C. Kim, and D. Kwon: Metall. Trans. A, 1993, vol. 24A, pp. 1133-41. 37. C.L. Davis and J.E. King: Metall. Mater. Trans. A, 1994, vol. 25A, pp. 563-73. 38. D.P. Fairchild: Ph.D. Thesis, The Ohio State University, Columbus, OH, 1995. 39. K. Ohya, J. Kim, K. Yokoyama, and M. Nagumo: Metall. Mater. Trans. A, 1996, vol. 27A, pp. 2574-82. 40. D.P. Fairchild and D.G. Howden: The Ohio State Univ. Res. Rep., Sept. 1989. 41. R.T. DeHoff and F.N. Rhines: Quantitative Microscopy, McGraw-Hill Inc., New York, NY, 1968, ch. 5. 42. W. Oldfield: ASTM Standardization News, 1975, Nov., pp. 24-29. 43. W. Oldfield, W.L. Server, R.A. Wullaert, and K.E. Stahlkopf : Int. J. Pressure Vessel Piping, 1978, vol. 6, pp. 203-22. 44. T. Sakuma and R.W.K. Honeycombe: Mater. Sci. Technol., 1985, vol. 1, pp. 351-56. 45. H. Tada: The Stress Analysis of Cracks Handbook, 2nd ed., Paris Productions Inc., St. Louis, MO, 1985. 46. M. Suzuki, K. Tsukada, and I. Watanabe: Proc. Int. Conf. on Offshore Welded Structures, The Welding Institute, London, 1982, pp. P16 and 1-12. 47. L. Meyer, F. Heisterkamp, and D. Lauterborn: Processing and Properties of Low Carbon Steel, Ferrous Committee of TMS-AIME, Warrendale, PA, 1973, pp. 297-321. 48. D.C. Houghton, G.C. Weatherly, and J.D. Embury: Proc. Conf. on Thermomechanical Processing of Microalloyed Austenite, Pittsburgh, PA, Aug. 1719, 1981, TMS-AIME, Warrendale, PA, 1981, pp. 267-92 49. W. Roberts: Proc. Conf. HSLA Steels, Technology and Applications Philadelphia, PA, Oct. 36, 1983, ASM, Metals Park, OH, 1983, pp. 33-65. 50. Z. Chen, M.H. Loretto, and R.C. Cochrane: Mater. Sci. Technol., 1987, vol. 3, pp. 836-44. 51. H.H. Hausner and M.G. Bowman: Fundamentals of Refractory Compounds, Plenum Press, New York, NY, 1968. 52. L.E. Toth: Transition Metal Carbides and Nitrides, Academic Press Inc., New York, NY, 1971. 53. E. Smith: Int. J. Fract. Mech., 1968, vol. 4(2), pp. 131-45.

652VOLUME 31A, MARCH 2000

METALLURGICAL AND MATERIALS TRANSACTIONS A

S-ar putea să vă placă și