Introduction

In a previous paper [1], the microstructure of quenched 4340 steelFootnote 1 tempered at 200 °C for 1 h was shown to consist of martensite with, predominantly, transition-iron-carbide precipitates. Analysis of crystallographic aspects of the precipitates showed that the data could be interpreted as from a hexagonal epsilon-carbide phase or an orthorhombic eta-carbide phase [1, 2]. Results from 4340 steels [3, 4] are, in many ways, similar to data collected from other alloys with martensitic microstructures [5,6,7,8,9,10,11,12,13,14,15,16,17,18,19]. Oftentimes, transition-iron-carbide precipitates appear to exhibit a rodlike or needlelike morphology [2, 6, 7, 10, 12, 15, 16, 18], but upon closer examination at high magnification, numerous transmission electron microscope (TEM) images have revealed that the “rods” were composed of individual, closely spaced, equiaxed precipitates of very small size (e.g., less than 10 nm) [2, 8,9,10,11,12, 15, 16, 18]. These nearly equiaxed precipitates oftentimes were aligned along <100> directions in martensite crystals [2,3,4,5,6, 12, 17,18,19]. Additionally, kinked segments were observed that were at least partially responsible for a “wavy” appearance of some precipitates or arrays of precipitates. These segments typically showed a 20°–30° angle with respect to the “rod” axis [1,2,3,4,5, 7, 8, 10, 12, 15,16,17,18,19].

In several studies [12,13,14,15, 20], steels with high alloy and/or high carbon contents were examined. These alloys were designed to possess martensite-start (Ms) temperatures below 20 °C so that the earliest stages of tempering or aging could be investigated. Results revealed interesting microstructural features that were not included in the seminal descriptions of stages I–III of martensite tempering [21]. Key observations were consistent with a spinodal decomposition mechanism in which a modulated structure with high- and low-carbon regions developed within martensite crystals. Taylor et al. [13, 20] examined Fe–Ni–C steels, and these authors showed that apparently planar modulations develop with {023} habits that eventually evolve into platelike transition-iron-carbide precipitates with {012} habits, where the indices refer to the tetragonal martensite phase.Footnote 2

Ohmori and Tamura [15] also reported a “tweedlike” structure, but suggested that it was not a result of planar modulations of carbon, but rather related to arrays of <121> dislocations introduced during the austenite-to-martensite transformation. They suggested that very fine epsilon-carbide precipitates formed as needles on these closely spaced line defects. The precipitates then grew into rods aligned along <100> martensite directions, followed by coalescence into plates or disks with {010} martensite habit planes.

Despite the observation of similar streaking events in electron diffraction patterns (EDPs) and similar tweedlike, modulated microstructures in the earliest stages of aging or tempering, two very different descriptions of microstructural evolution have been proposed [13, 15]. In one study, the aged microstructure evolved toward the formation of transition-iron-carbide precipitates in the form of plates [13], whereas in the other, the precipitates formed as needles aligned in <100> martensite directions [15]. Since the previous work showed precipitation phenomena in 4340 steel were similar to those reported for high-alloy steel [22], the purpose of the present work was to see if any of the hypotheses in these previous papers could provide a complete understanding of stage I tempering of a commercially processed 4340 steel.

Experimental Procedures

The material obtained was end-cut slices from Charpy V-notch specimens and measured 10 by 10 mm with a thickness near 1 mm. The steel manufacturer verified that the alloy adhered to the chemical composition requirements for 4340 steel, and an approximate composition (in mass pct.) is 0.4 C, 0.7 Mn, 0.3 Si, 1.7 Ni, 0.9 Cr, and 0.2 Mo. The thermal history of the steel was reported as austenitized, quenched, and then tempered for 1 h at 200 °C.

Three-millimeter-diameter disks were produced from the thin coupons described above. Specimens were then electrochemically jet-polished until perforation with a commercial twin-jet electropolishing device. After perforation, specimens were rinsed, dried, and inserted into a TEM.

The TEM used was operated at an accelerating potential of 120 kV. Conventional bright-field (BF) and centered-dark-field (CDF) images were generated. Selected-area-electron-diffraction (SAED) spot patterns were recorded on photographic film. The camera constant for each pattern was calibrated based on partial ring patterns from surface oxide (magnetite), and results were compared with a previous calibration that utilized a high-purity aluminum standard. Calibration of the rotation between images and diffraction patterns was conducted with crystals of molybdenum trioxide on an amorphous carbon substrate.

To improve the accuracy of the electron diffraction analyses, eccentricities of diffraction rings were measured, and a method of correcting for distortions was employed [23], followed by recalibration of the camera constant. This approach established upper limits of uncertainties at ± 0.005 nm for interplanar spacings and ± 4° for interplanar angles.

Results

Region Chosen for Detailed Analysis

Two different views of an exceptionally large lath of martensite are shown in Fig. 1. This lath is the same as that analyzed in a previous paper [1] and has proven to be comparatively easy to work with because of its size, orientation, and location within a thin-foil TEM specimen. In Fig. 1a, the lath of interest is located at the left side of the micrograph, and it shows clear differences in brightness and in contrast for its “upper” and “lower” portions. This feature is strongly reminiscent of martensite midribs that are ubiquitous in plate martensite [24, 25]. If in fact this apparent “discontinuity” is the location of the first-formed portion of the lath and if it “grew toward the top and bottom of this micrograph,” then the lath width is more than three micrometers wide, likely to be one of the thickest laths in the specimen [26]. Three arrows denote the positions of prior austenite grain boundaries that existed in the parent microstructure prior to quenching and the transformation of austenite to lath martensite. The size of this lath compared with adjacent ones suggests that it was one of the first-formed laths [27], which would likely make it susceptible to autotempering since the Ms temperature of 4340 steel is calculated to be about 300 °C [28]. In Fig. 1b, the differences between the “upper” and “lower” portions of this lath are less noticeable, consistent with the interpretation that it is, most likely, a single lath rather than two adjacent laths. Additionally, an elongated array of transition-iron-carbide precipitates is labeled at “A,” which is analyzed later in the paper.

Fig. 1
figure 1

Exceptionally large lath within the martensitic microstructure of 4340 steel. (a) Well-defined “upper” and “lower” portions and evidence of prior austenite grain boundaries (arrows). (b) Elongated linear aggregate of transition-iron-carbide precipitates (at A) in the same lath, but a different tilt condition. TEM BF images

Full size image

The goal of the work was to provide detailed analyses of complex fine features associated with precipitation phenomena in what is presumed a representative large martensite lath. To accomplish this work, several sessions were needed, and the specimen was removed after each session. Upon reinserting the specimen for the next session, no attempt was made to orient the specimen the same fashion, although it is believed that the same side was “up” within the microscope for all sessions. The images presented in this paper have been positioned such that the [010] martensite direction is either (i) within about 10°–20° or (ii) precisely at the horizontal position (within about a degree), and the difference between these two conditions will be made clear in the text of this paper for each of the remaining figures.

Transition-Iron-Carbide Precipitates in 4340 Steel

Rodlike or needlelike transition-iron-carbide precipitates frequently are observed aligned along <100> martensite directions [2,3,4,5,6, 12, 17,18,19]. However, in some cases, the precipitates are believed to be a hexagonal epsilon-carbide phase, and in others, the phase is reported to be an orthorhombic eta-carbide phase. The differences in crystallographic aspects of these two phases are extremely small [2], and throughout the remainder of this paper, the precipitates will be described simply as a transition-iron-carbide phase. Other research [2, 8,9,10,11,12, 15, 16, 18] has shown that small, transition-iron-carbide precipitates possess equiaxed shapes (possibly spherical) that, oftentimes, are aligned in a limited number of directions within crystals of martensite. This evidence comes predominantly from CDF images, and similar information gleaned from BF images is far less common.

Figure 2 includes a bright-field TEM image and its associated EDP, both at the same tilt condition. The diffraction pattern (Fig. 2c) shows that the [010] martensite direction is horizontally oriented, while the [\( \bar{1} \)00] martensite direction is oriented vertically. The image (Fig. 2a) has been rotated into the same configuration as the EDP. The multi-beam diffracting condition explains why the bright-field image exhibits such strong contrast from the martensite matrix. Although this strong contrast was not effective for examination of the details of fine-scale, transition-iron-carbide precipitates, it did reveal some extremely elongated precipitates or precipitate arrays that were not immediately obvious at other tilt conditions nor in adjacent, smaller laths of martensite. Figure 2b is the same image as Fig. 2a, but has an over-layer on which precipitates were highlighted with black, lenslike features.

Fig. 2
figure 2

Linear aggregates of transition-iron-carbide precipitates (parallel to arrows). (a) BF TEM image, (b) over-layer emphasizing individual vs. aggregates of precipitates, and (c) the associated EDP. Rotation between image and EDP has been accounted for in the placement of the figures. The horizontal direction coincides with the [010] martensite direction

Full size image

Transition-iron-carbide precipitates are present throughout Fig. 2a, and an elongated precipitate feature is shown immediately below the letter “D.” The feature at “D” shows a long axis that is extremely close to a [\( 010 \)] martensite direction, and appears to possess a lenslike or lenticular morphology.Footnote 3 Similar features are shown throughout the remainder of this lath. These precipitates or arrays of precipitates are approximately 100 nm in length. However, between two labels denoted “A” in Fig. 2a, there exists an elongated feature nearly 500 nm in length with kinked segments that produce an alignment of about 15° away from the [\( 010 \)] martensite direction. This and similar elongated features will be characterized later in this paper and appear not to have been documented previously. Figure 2b helps to emphasize the “difference” between the 100-nm-long precipitate features and the 500-nm-long linear aggregates of precipitate features.

In Fig. 2a, a “downward-sloping arrow” is nearly parallel to the feature at “A–A,” and a similar feature of nearly the same orientation is highlighted with the labels “B–B.” Alternatively, the “upward-sloping” arrow highlights a similar feature at “C–C.” Notice that these two arrows are near symmetrically positioned about the horizontal [010] martensite direction, both at angles of about 15°.

Since the 100-nm-long precipitate features have been observed previously and are the more common feature throughout this large lath, they will be analyzed first. Figure 3 shows a pair of bright-field and centered-dark-field transmission electron micrographs from a region located at the right of Fig. 2a, but at higher magnification and at a very different tilt condition than that of Fig. 2. This fact is also demonstrated by the associated EDP (Fig. 4) with a [\( \bar{1}39 \)] beam direction, about 19° away from a [\( 001 \)] orientation. The [010] martensite direction has an orientation that is not in the plane of Fig. 4; it points slightly upward from the image plane at an angle of about 18°. Its projection onto the plane common to Figs. 3a, b, and 4 is in the horizontal position (i.e., a degree or two below \( {\mathbf{g}}_{{03\bar{1}}} \)). This direction is indicated by the black arrow oriented horizontally in Fig. 3a. The feature labeled “D” in Fig. 2a is also labeled “D” in Fig. 3. As compared with Fig. 2a, Fig. 3 shows the rodlike nature rather than a lenticular morphology and also shows that the “rod” actually consists of several small, near-equiaxed precipitates aligned along [010] martensite. The feature at “D” in Fig. 2b is shown in more realistic fashion beneath its original depiction, where several small circles have been used to indicate the near-equiaxed shapes observed in Fig. 3.

Fig. 3
figure 3

Rodlike arrays of closely spaced transition-iron-carbide precipitates. (a) BF and (b) CDF TEM images. The horizontal and near-vertical directions coincide with projections of [010] and [100] martensite directions, respectively (see arrows)

Full size image
Fig. 4
figure 4

EDP associated with Fig. 3. Rotation between image and EDP has been accounted for in the placement of the figures. Beam direction is [\( \bar{1} \)39] martensite

Full size image

The transition-iron-carbide precipitates of (possibly) spherical morphology are approximately 5 nm in diameter, consistent with the previous work. A kink is also present at the left side of the array in Fig. 3. Elsewhere in these micrographs, other rodlike arrays with occasional kinks are oriented in the same direction. The remaining rodlike arrays are oriented nearly 90° to the horizontal ones and are nearly parallel to a projection of the [\( 100 \)] martensite direction (the arrow pointing downward in Fig. 3), again with some kinked segments. All of the “rodlike” arrays in this figure apparently consist of closely spaced precipitates that are nearly touching (about 10 nm apart) rather than a monolithic rodlike crystal.

Figure 5 shows the rodlike array at “D” in Figs. 2, 3a and b, but now at a higher magnification. Figure 5a is a conventional bright-field image, and Fig. 5b is a CDF image from a transition-iron-carbide diffraction spot. These images come from a thin-foil specimen estimated to be about 50 nm in thickness locally, and therefore, although the features appear to be touching, they could be at slightly different heights within the specimen. A good example consistent with this description is shown by the features labeled “1” and “2” in Fig. 5a and b. Typically, closely spaced second-phase precipitates should be associated with some type of microstructural defect or discontinuity in the matrix phase that stimulated precipitation, such as a dislocation or a twin/matrix interface.

Fig. 5
figure 5

Higher-magnification versions of Fig. 3 (a) is BF, (b) is CDF, (c) is a near-weak-beam image of the martensite matrix, and (d) is BF image at a different tilt condition

Full size image

Figure 5c is a near-weak-beam image from a martensite matrix diffraction spot and at a slightly different tilt condition than those shown in Fig. 5a and b. In Fig. 5c, segments of martensite matrix dislocations are labeled “1” and “2.” These features are discussed in more detail later in this paper where it is shown that these lines are consistent with <111> martensite directions. Figure 5d is an enlargement of “D” from Fig. 2a, and when compared, Fig. 5a and d shows that images of small transition-iron-carbide precipitates can exhibit very different appearances and correspondingly result in very different interpretations of morphological features depending on local diffracting conditions.

During a subsequent session with the same electron microscope, this region was once again located, and the kinked, elongated features measuring nearly 500 nm in length were examined at a number of different tilt conditions in an attempt to gain a better perspective regarding their morphology. Figures 6 and 7 show the resultant bright-field images and the associated EDPs, respectively. (These images are oriented at various angles (5°–15°) away from a horizontal [010] martensite direction). The most prominent feature in Fig. 6a is emphasized by two small “end-cap” arrows below the label “A.” Above this label, a large arrow is used to highlight the macroscopic direction of this kinked linear array; this direction is within a degree of the projection of the [\( \bar{1}\bar{4}0 \)] martensite direction onto the plane of the image.

Fig. 6
figure 6

Rodlike arrays of transition-iron-carbide precipitates measuring about 100 nm and linear aggregates of rodlike arrays (about 500 nm long) “attached” by kinked segments. The black arrows in the upper right corners of each image indicate projections of the [0\( \bar{1} \)0] (near horizontal) and [\( \bar{1}00 \)] (near vertical) martensite directions. (a) through (d) correspond with EDPs in Fig. 7

Full size image
Fig. 7
figure 7

EDPs associated with Fig. 6. The beam directions referenced to martensite are [023] ((a) in both figures), [\( \bar{1}57 \)] tending toward [\( \bar{1}69 \)] ((b) in both figures), [\( \bar{1}57 \)] tending toward [\( \bar{1}35 \)] ((c) in both figures), [\( \bar{3}35 \)] ((d) in both figures), and [\( \bar{1}23 \)] ((e) without a corresponding image in Fig. 6)

Full size image

Above the large arrow, a second linear aggregate of rodlike arrays “attached” by kinked segments is shown that possesses a slightly different macroscopic direction. The feature below “A” in Fig. 6a is similarly highlighted by long arrows in Fig. 6b and c, again indicating that the feature is aligned in the [\( \bar{1}\bar{4}0 \)] martensite direction. In Fig. 6c, the linear feature is rather difficult to discern from martensite matrix features, and a version at higher magnification is presented as an inset at the top of this image. The images and long arrows are consistent with a stereographic trace analysis of the 500-nm-long feature that revealed alignment in the [\( \bar{1}\bar{4}0 \)] direction.

In each part of Fig. 6, the black arrows in the upper right-hand corners backed with white rectangles denote the projected [\( 0\bar{1}0 \)] martensite direction (to the left and slightly downward most of the time) and the projection of the [\( \bar{1}00 \)] martensite direction (nearly straight up). None of the EDPs possess \( {\mathbf{g}}_{{0\bar{2}0}} \), and only one possesses \( {\mathbf{g}}_{{\bar{2}00}} \); therefore, the angles between the projected [0\( \bar{1} \)0] and \( [\bar{1}00] \) directions are not 90°, and these positions of these projections vary between each tilt condition. The image in Fig. 6d seemed to show some notable distortion, as is discussed further below.

The EDPs show that the first tilt condition (Figs. 6a and 7a) corresponded with a [023] martensite zone axis, the second tilt condition (Figs. 6b and 7b) corresponded to [\( \bar{1}57 \)] with a second zone near \( \left[ {\bar{1}69} \right] \), the third also corresponded with [\( \bar{1}57 \)] but with a second zone of spots tending toward \( \left[ {\bar{1}35} \right] \) (Figs. 6c and 7c), and the fourth was near a [\( \bar{3} \)35] zone axis (Figs. 6d and 7d). The tilt conditions according to readings from the goniometer were read as approximately − 10°, − 2.5°, + 2.5°, and + 25°. The plus/minus designations refer to the 0° condition, which occurs when the specimen is normal to the incident electron beam. A fifth tilt condition resulted in Fig. 7e, a pattern with a [\( \bar{1}23 \)] zone axis, and the tilt was + 10°; however, an image corresponding with this region in the large lath was not recorded. The angles show a consistent indexing scheme; therefore, a reliable analysis is anticipated. Table 1 provides a summary of this tilt-angle information.

Table 1 Tilt-angle data
Full size table

An important observation from Fig. 6a–c is that the 500-nm-long feature of interest does in fact approximate as what will be referred to as a linear aggregate of rodlike arrays “attached” by kinked segments. Its near-linear nature is supported by the observations of Fig. 6a–c. Unlike the 100-nm-long rodlike precipitate arrays that show strong alignment with <100> martensite directions with only an occasional kink (refer to Fig. 2b at “D”), the 500-nm-long features possess enough kinked segments that the macroscopic direction is well away from the <100> directions (refer to Fig. 2b). Thus, it is concluded that the rodlike arrays and linear aggregates of rodlike arrays have much in common and probably develop in a similar fashion. Thus, a distinction between the two arrays may be somewhat artificial, but will be retained to facilitate further discussion.

Returning to Fig. 6d, there are some similarities with respect to Fig. 6a–c, but some important differences that cannot be ignored. The feature to the left of “A” in Fig. 6a seems clearly to be the same as that labeled “A” in Fig. 6d. The linear array above the long line in Fig. 6a (labeled “E”) seems to correspond with that below the lower long arrow in Fig. 6d. These observations suggest some type of rotation between Fig. 6a and d, but as the specimen was not rotated, one might conclude that Fig. 6d suffers from distortion since it is tilted at a significant angle (25°) compared with the other images, 10° or a few degrees. However, despite this problem, the elongated features still appear to be roughly linear with kinked segments, and instead of being parallel to [\( \bar{1}\bar{4}0 \)], they are parallel to [\( 1\bar{4}0 \)], i.e., a symmetrical deviation about the [\( 0\bar{1}0 \)] martensite axis.

In many alloy systems, fine-scale second-phase precipitates form at dislocations because of a reduction in the energy barrier for nucleation. Figure 8 is broader view of the region shown by Fig. 5c. The dark feature (no image intensity) in the center of Fig. 5c is directly beneath the label “D” in Fig. 8, and this same feature is at “D” in Figs. 2a and 3. In Figs. 5c and 8 (same image but different magnifications and fields of view), the rodlike array is at a sharp angle to the linear white features in this figure, which are martensite matrix dislocations. Numerous similar features of this same orientation as well as other orientations are shown in Fig. 8. The EDP associated with Fig. 8 was nearly the same as that shown by Fig. 4 (within half a degree). Therefore, the latter figure was used to project four <111> directions onto Fig. 8 to represent the expected directions of screw dislocations in this crystal of body-centered-cubic tempered martensite. Assuming the matrix dislocations to be screw dislocations [29], the [\( 1\bar{1}1 \)] direction is nearly parallel to the feature at “a.” Likewise, the features at “b,” “c,” and “d” align with [\( 1\bar{1}\bar{1} \)], [\( 111 \)], and [\( 11\bar{1} \)] directions, respectively. Table 2 contains information regarding the projections of the four <111> directions onto the image plane of Fig. 8. While many dislocations aligned along [\( 11\bar{1} \)] and [\( 111 \)] are present, those along [\( 1\bar{1}\bar{1} \)] and [\( 1\bar{1}1 \)] seem rather low; in fact, the feature at “a” is believed to be a small plate of cementite. The frequency of observation might be related to the dislocation orientation with respect to the plane of the image. Specifically, the [\( 1\bar{1}1 \)] is nearly aligned with the electron beam, and therefore, the lengths of these dislocation line segments would be small and possibly obscured in this complex image. Elsewhere, there are numerous features that cannot be completely explained at this point. Nonetheless, the observations are broadly consistent with dislocation lines that seem to show little “affinity” to the rodlike precipitate arrays. Support for the claim that transition-iron-carbide precipitates do not seem to form on dislocations in martensite has been provided in other studies [5, 8, 13].

Fig. 8
figure 8

Near-weak-beam TEM image from martensite reflection. Matrix martensite dislocations, many aligned with <111> directions

Full size image
Table 2 Dislocation characteristics
Full size table

Discussion

A key observation from this work is that 500-nm-long linear aggregates of rodlike arrays attached by kinked segments have been observed, and reasons for their existence are not readily apparent. Additionally, these linear aggregates seem to have much in common with the transition-iron-carbide precipitates that form in rodlike arrays, which measure about 100 nm in length, and possess occasional kinked segments. The rodlike arrays are aligned along <100> martensite directions, whereas the linear aggregates of rodlike arrays lie along <140> martensite directions. This difference is possibly related to the occasional kink in rodlike arrays, but the linear aggregates seem to be connected to adjacent linear aggregates by more prominent kinked segments. However, the rodlike components of the linear aggregates are aligned along <100> martensite directions.

The prominent role of <100> martensite directions in precipitate alignment is difficult to explain since martensite dislocations are aligned, predominantly, along <111> directions. Additionally, the lath examined in this study showed no evidence of internal twins, ruling out any role they might play. While reasons for the alignment of 5-nm-diameter equiaxed precipitates are not readily apparent, the observation that they are separated by about 10 nm is very difficult to explain.

Tekin and Kelly [5] presented a wide range of observations regarding the tempering of steels, and one particularly interesting point was that “at no stage of tempering was precipitation on dislocations observed, though dislocations associated with the precipitates were sometimes found.”

Hirotsu and Nagakura [8] observed “dark feather-like images,” likely similar to or the same as the rodlike arrays with occasional kinks observed in the present work and stated that this was “due to dislocations, along which the carbide particles precipitated” [8]. They used dark-field imaging to observe transition-iron-carbide precipitates and concluded that carbide particles precipitated “periodically along dislocations” at 10 nm intervals [8].

In a subsequent paper on transition-iron-carbide precipitation in martensite, Hirotsu et al. [10] showed several features similar to those in the present work, although some differences between their work and the present work were also apparent. They reported that precipitates were not always formed along dislocations, and “a different mechanism must be considered to interpret the needlelike alignment of carbide particles” [10].

Although Taylor et al. [13] observed platelike rather than rodlike arrays of small equiaxed transition-iron-carbide precipitates, they too reported that dislocations in martensite did not “play a significant role in the nucleation of the transition carbides” [13]. While noting the difficulty in directly observing any possible association between dislocations and nuclei of second-phase precipitates, they stated that “no nucleation was directly observed on dislocations” [13].

As shown by Taylor et al. [13, 20] as well as by other investigators, high-alloy and/or high-carbon martensitic steels with low Ms temperatures show that planar modulations of high- and low-carbon regions develop prior to the first stage of tempering. The preferred planes for these modulations appear to be close to {012} and {023} martensite planes. If one considers the intersections between {012} planes in a tetragonal phase, they are <100> and <221> martensite directions. For {023} planes, the intersections are <100> and <332> directions. The formation of adjacent, small precipitates along <100> directions in body-centered-cubic iron (similar to tempered martensite) might be encouraged since this is an elastically soft direction. Additionally, if intersecting planar modulations of high-carbon regions develop, one might envision a further elevation of carbon atoms at these intersections.

If one accepts the concept of intersecting planar modulations as the source of the <100> alignment of precipitate arrays in this system, then the kinks might be envisioned as forming along other intersections, which are <221> directions for {012} planar modulations, and <332> directions for {023} planar modulations. However, the angles between <221> and <100> directions or between <332> and <100> directions are 48–50°, not very close to the 20°–30° angles that are observed. Additionally, the <332> directions are 10° away from the elastically hard <111> directions, and so in this case, the “choice” of kinked segments seems not to be explained. However, Fig. 6d provides some evidence of a symmetric crystallographic nature to the kinks. The commonly observed “kink angle” of 20°–30° in this region (and many other similar observations) also suggests that there might be a crystallographically related reason for the kink directions, but it is not apparent at this time the source of the governing or responsible principles.

Thus, although clearly in need of further experimental observations, it is hypothesized that the presence of intersecting planar modulations of high-carbon regions that result from spinodal decomposition provides the source for the formation of closely spaced (nearly touching) transition-iron-carbide precipitates along elastically soft <100> directions. The presence of kinked segments in transition-iron-carbide precipitates, oftentimes described as the “wavy” nature of these precipitates, has been shown not to be associated with “wavy” dislocations, although a second plausible explanation for these kinks has been offered. It is hypothesized that adjacent transition-iron-carbide precipitates are “connected” by cementite precipitates that either form during quenching (autotempering), or during tempering as the transition carbides are replaced by the thermodynamically more favorable cementite phase. Although cementite plates have been observed in this work and the previous work, there is little or no association between the two phases at kinked segments. Additionally, the kinked segments typically show about a 20°–30° angle with respect to the longer segments, whereas if they were cementite, they should appear to be aligned along <110> directions, i.e., at a 45° angle.

It is interesting to note some similarities between the aligned precipitate arrays in the present work and the tweedlike microstructures reported by others in various alloy systems. In these systems, precipitate phases belong to the tetragonal crystal system, while the parent phase is cubic. As stated in Ref. [30]: “It is generally accepted that the tweed structure arises from a natural tendency to minimize the elastic energy of the precipitate distribution.” It may be that some concepts from this field of study are relevant to the present system.

Results from the present paper combined with the previous work provide for detailed analyses of several aspects of the early stages of transition-iron-carbide precipitation during tempering of a commercially produced 4340 steel; however, the scope is exceedingly limited as one large lath has been analyzed from single heat-treatment condition. Naturally, a follow-up to this work should include, at least, a series of times at low tempering temperatures, or an extended time–temperature matrix. To date, the literature contains many such studies, and the present work seeks to fill some voids that remain in our understanding of the earliest stages of tempering of martensitic steels. Further work in this area is ongoing.

Finally, it should be noted that while different stages of microstructural evolution exist during aging and tempering, observations from various studies may reflect different mechanisms of microstructural evolution that are a function of alloy content and processing history. For example, the large lath chosen in this work did not show any evidence of twinning, and this was considered a major advantage during the interpretation of complex EDPs [1]. In regions of twinned martensite, one should expect differences in features of transition-iron-carbide precipitation as compared with untwined martensite. Another example of expected differences would be comparison of microstructural evolution in steels quenched with extremely high cooling rates (e.g., laboratory processing of thin coupons) versus slower cooling rates in large parts that are processed in industry. Still, another set of different conditions is realized in martensitic steels with Ms temperatures well above ambient (e.g., 300 °C in 4340 steel) versus Ms temperatures well below 20 °C (e.g., Fe–25Ni–0.4C and Fe–15Ni–1C model alloys [13]).

Summary

SAE 4340 steel was tempered at 200 °C for 1 h. The following points of summary are provided.

  1. 1.

    Small, nearly equiaxed transition-iron-carbide precipitates (less than 10 nm in diameter) were observed in rodlike arrays, measuring about 100 nm in length. The rodlike arrays are aligned predominantly along <100> martensite directions.

  2. 2.

    “Kinked segments” are occasionally observed in the rodlike arrays that deviate from <100> martensite directions by about 20°–30°.

  3. 3.

    A large lath presumed to have formed at comparatively high temperature also showed less frequent linear arrays measuring nearly 500 nm in length. To distinguish them from the more common rodlike arrays, they were referred to as “linear aggregates of rodlike arrays attached by kinked segments.”

  4. 4.

    Linear aggregates of rodlike arrays attached by kinked segments possess a macroscopic direction of about <140> as referenced to the tempered martensite matrix.

  5. 5.

    Transition-iron-carbide precipitates do not show a clear association with matrix dislocations of the martensite phase.

  6. 6.

    A hypothesis is offered that suggests a link between the linear aggregates and planar modulations of high- and low-carbon regions that may develop prior to the onset of stage I of tempering.