Materials Transactions, Vol. 43, No. 7 (2002) pp.

1758 to 1766

2002
c The Japan Institute of Metals
EXPRESS REGULAR ARTICLE

Microstructural Analyses of Grain Boundary Carbides of Tempered

Martensite in Medium-Carbon Steel by Atomic Force Microscopy ∗
Masao Hayakawa, Saburo Matsuoka and Kaneaki Tsuzaki
National Institute for Materials Science, Tsukuba 305–0047, Japan

The microstructures of JIS-SCM440 steel (containing 0.4%C, 1%Cr and 0.2%Mo; mass%) that had been tempered at 723 K were observed
by atomic force microscopy (AFM). In an AFM image of an electropolished surface, carbides and blocks were clearly distinguished, since the
rate of electropolishing depends on the crystal phase and crystal orientation on the surface. However, the prior austenite (γ ) grain boundaries
could not be clearly recognized. Therefore, an AFM image of the surface that had been etched with picric acid was taken at the same location
on the surface. Since picric acid selectively etches prior γ grain boundaries, the boundaries were clearly visible in the image. The two AFM
images were then superimposed to reveal prior γ grain boundaries, carbides, and blocks. Finally, the number and sizes of the carbides on the
prior γ grain boundaries were examined.

(Received April 26, 2002; Accepted June 3, 2002)

Keywords: tempered martensite, medium-carbon steel, atomic force microscope, prior austenite grain boundary, carbide

1. Introduction objectively and quantitatively.

Tempered martensite forms an important microstructure in
medium or high-carbon low-alloy steels for machine compo-
nents. It is necessary to understand the microstructures of
the tempered martensite to develop high strength steel with a
tensile strength of over 1500 MPa and superior resistances to
delayed fracture and fatigue. The authors have characterized
prior austenite (γ ) grains, packets, blocks, laths, precipitates
and dislocations by observing the structures of the marten-
site1, 2) using an atomic force microscope (AFM) and a trans-
mission electron microscope (TEM). In a previous paper,1)
we observed the tempered lath martensite of JIS-SCM440,
a medium-carbon low-alloy steel (containing 0.4%C, 1%Cr
and 0.2%Mo; mass%), using a light microscope (LM) and
a scanning electron microscope (SEM) as well as AFM and
TEM. The blocks of a medium-carbon steel are too small for
the conventional LM and SEM observations. A previous pa-
per1) showed that the blocks of the JIS-SCM440 steel were
successfully observed on an electropolished surface by AFM,
and the size distributions of the blocks were determined. The
rate of electropolishing depends on the crystal orientation of
the specimen surface, and this allows precise measurements
of nanometer-scale differences in height among the blocks.
However, the AFM image was not sufficient to identify the
prior γ grains. In the previous paper,1) a prior γ grain mea-
surement was taken using the LM images of a surface etched
with picric acid solution, which selectively etches the prior γ
grain boundaries.
Prior γ grain boundaries tend to precipitate numerous car-
bide particles, and the relationship between the grain bound-
ary carbides and the mechanical properties of the steel, such
as delayed fractures and temper brittleness, has been under
discussion.3–6) To gain an understanding of the mechanisms
of delayed fracture and to identify the microstructures that
are highly resistant to the delayed fracture, it is necessary to
establish methods for evaluating the grain boundary carbides
∗ ThisPaper was Originally Published in J. Japan Inst. Matals 65 (2001)
734-741.
This paper describes an AFM observation method that we
developed by making improvements on the previous method1)
which enables us to observe carbides, blocks, and prior γ
grain boundaries in the same area. The method uses the tem-
pered lath martensite microstructure of the JIS-SCM440 steel
and takes and superimposes two AFM images of the same
area: one of an electropolished surface that shows carbide
distribution and the other of a picric acid-etched surface at
which prior γ grain boundaries are selectively corroded. This
method proves effective in determining the number and the
size of the carbides on the prior γ grain boundary.
2. Experiment
2.1 Specimen
A JIS-SCM440 steel containing the same chemical com-
ponents as those described in the previous paper1) (0.40%C,
0.24%Si, 0.81%Mn, 0.02%P, 0.007%S, 1.03%Cr, and
0.16%Mo: mass%) was used for the specimen. The condi-
tions of the thermal treatment were also the same as those
in the previous paper.1) The specimen, which was a 80 mm
length bar with a diameter of 12 mm, was austenized at
1153 K for 2.7 ks and then oil-quenched to form the lath
martensite microstructure. To acquire a sufficient amount of
cementite precipitation (Fe3 C), the specimen was tempered
at 723 K for 5.4 ks and cooled in water. These tempering
conditions produced cementite precipitates but no alloy car-
bides such as Mo2 C. The tensile strength of the specimen
was 1400 MPa, and its Vickers hardness was 430.
2.2 AFM observation of the microstructure
AFM observation was conducted in the atmosphere. The
maximum scanning range was 90 µm for both the length and
the width and 5 µm for the height. The observation was con-
ducted under the vibration mode (the tapping mode) where
the silicon cantilever with a 20 N/m spring constant resonated
at 300 kHz. The scanning speed was 0.2 Hz for the electropol-
ished surface and 0.5 Hz for the etched surface. The can-
 Microstructural Analyses of Grain Boundary Carbides of Tempered Martensite
tilever, which had a probe tip of with a radius of 5–10 nm at
the curvature, was monolithically formed by etching a silicon
wafer.
A flowchart for the AFM observation is shown in Fig. 1.
The outline of the observation process is: 1) taking an AFM
image of the electropolished surface; 2) taking an AFM image
of the same location after etching the surface with picric acid
and finding that position using the Vickers indent as a marker;
3) superimposing the two images; and 4) observing the car-
bides and the prior γ grain boundaries in the superimposed
image.
The detail of the process is:
1) A 7 square mm plate with a thickness of 2 mm was cut
from the specimen bar such that the observation surface was
along the lengthwise (rolling) direction of the bar.
2) The observation surface was polished with 5 µm diamond
particles until the surface became mirror-finished.
3) A Vickers indent was formed on the surface as a position-
finding marker. A large indent was first formed by applying
a load of 490 N using a Vickers hardness tester. Then a load
of 9.8 N was applied using a micro Vickers tester, forming a
small indent at a distance three times the diagonal length of
the large indent along its diagonal direction. Electropolish-
ing and etching treatments made it difficult to distinguish the
small, shallow indents, but it was possible to locate them and
the target position easily by using the large indent as a marker.
4) To observe the carbides and blocks, electropolishing was
conducted using a solution that consisted of 8 vol% perchloric
acid, 10 vol% butoxyl ethanol, 70 vol% ethanol, and 12 vol%
distilled water. The temperature of the solution was 273 K,
and an electric voltage of 40 V was applied for 10 s. The spec-
imen was washed ultrasonically with ethanol immediately af-
ter electropolishing.
5) and 6) The electropolished surface was observed by
Cutting out the specimen
Mechanical polishing
Marking by a Vickers and a micro Vickers indent
Electropolishing
Location setting for the AFM observation around the micro Vickers indent
AFM observation of carbides and blocks
Picric acid etching
Location setting for the AFM observation in the same area as
AFM observation of prior grain boundaries
Superimposing of the AFM micrographs taken on and
Size measurement of the carbides on prior grain boundaries
Fig. 1 Flowchart of AFM observations.
1759
AFM after determining the point of observation near the
micro Vickers indent.
7) To observe the prior γ grain boundaries, the specimen
was removed from the AFM and immersed in a solution con-
sisting of 100 cm3 of saturated picric acid solution and 5 cm3
of interfacial active agent (Solution I) for 30 s. The speci-
men was then transferred into a solution that contained 50 cm3
of the first solution and 0.5 g of ferric chloride anhydrous
(Solution II)and was incubated for 60 s to allow selective cor-
rosion of prior γ grain boundaries. Then the specimen was
washed with ethanol and examined to determine whether it
was possible to identify the prior γ grain boundaries by the
Nomarski differential-interference light microscopy. The LM
photographs were taken to determine the grain sizes.
8) and 9) We found the point of observation using the micro
Vickers indent, adjusted the field of AFM observation to ob-
serve the same area of the picric acid-etched surface as the
area that we had observed after electropolishing the surface,
and carried out the AFM observation.
10) AFM images of the electropolished and the picric
acid-etched surfaces were superimposed using an image-
processing program.
11) We measured the size of the carbides separately over
the entire field, on the prior γ grain boundaries, on the packet
boundaries, and on the block boundaries using an AFM im-
age that showed all the prior γ grain boundaries, blocks, and
carbide microstructures, using the particle analysis program.
Figure 2 shows LM photographs of a micro Vickers indent
and an AFM cantilever on (a) a mirror finished surface, (b) an
electropolished surface, and (c) a picric acid-etched surface.
Since the diagonal lines of the indent were 65 µm, 40 µm, and
35 µm on the mirror finished, electropolished and picric acid-
etched surfaces, respectively, and the angle between the oppo-
site faces of the micro Vickers indenter was 136◦ , the depths
of electropolishing and etching should be 5 µm and 1 µm, re-
spectively. It was easier to find the target position by observ-
ing the electropolished surface first, followed by the etched
surface, since etching caused less abrasion than electropol-
ishing.
In addition to the aforementioned series of observations, a
nital-etched surface after the electropolishing was observed
under an LM.
3. Results and Discussion
3.1 LM observation of microstructures
Figure 3 shows the LM images of the surface etched with
(a) picric acid solution and (b) nital. The surface that had
been etched with picric acid showed selective corrosion at
the prior γ grain boundaries. The nominal grain size that
was determined from the microscope image in the view of
230 µm × 180 µm was 17 µm by the cutting method.7) The
micrograph of the nital-etched surface showed packets that
contained martensite lath in parallel alignment. However, the
prior γ grain boundaries were not distinguishable.
3.2 AFM observation of the microstructures on elec-
tropolished and etched surfaces
Figure 4 shows an AFM micrograph of the electropolished
surface. The black and white contrast in the image shows
1760 M. Hayakawa, S. Matsuoka and K. Tsuzaki

Fig. 3 Light micrographs of the JIS-SCM440 steel tempered at 723 K. (a)

and (b) were observed on the picric acid- and the 3% nital-etched surfaces,
respectively. (a) shows prior γ grain boundaries, and (b) shows martensite
laths.

Fig. 2 Light micrographs showing the micro Vickers indent and the AFM
cantilever taken on (a) the mechanically polished surface, (b) the elec-
tropolished surface and (c) the picric acid-etched surface.

the differences in the level on the surface. The differences in

the level are attributable to the difference in the rate of elec- Fig. 4 AFM image of the electropolished surface of the JIS-SCM440 steel
tropolishing, which depends on the crystal phase and orien- tempered at 723 K. Martensite blocks, represented by black and white
contrast, are clearly identifiable.
tation on the surface. Therefore, the black and white stripes
are the blocks. Both black and white contrasted regions and
uniformly colored regions are present. The uniformly colored
etched surface. It has almost the same surface area as that
regions are blocks consisting of laths of uniform crystal ori-
of Fig. 4. The selective corrosion at the prior γ grain bound-
entation. The differences in the level among blocks were less
aries was shallow, and there were boundaries that were not
than 10–20 nm. Packets are shown as black and white stripes
easy to identify. However, when the degree of corrosion was
in parallel alignment. The carbides form peaks on the BCC
increased, more prior γ grains eluted, making it difficult to
matrix surface and appear as white spots under these elec-
compare the image with the AFM image of the electropol-
tropolishing conditions.
ished surface.
Figure 5 shows an AFM micrograph of the picric acid-
 Microstructural Analyses of Grain Boundary Carbides of Tempered Martensite
Fig. 5 AFM image of the picric acid-etched surface of the JIS-SCM440
steel tempered at 723 K. Referring to the micro Vickers indent, Fig. 5 was
taken in the same area as Fig. 4. Prior γ grain boundaries, represented by
black lines, can be identified.
3.3 Superimposition of AFM images of the electropol-
ished and etched surfaces
We identified several prior γ grain boundaries on the elec-
tropolished surface in Fig. 4, but it was difficult to identify all
of them. This problem was solved by superimposing the AFM
image of the electropolished surface (Fig. 4) on the AFM im-
age of the etched surface (Fig. 5). Figure 6 shows the results
of superimposing these images. In Fig. 6(a), the AFM im-
age of the electropolished surface was not treated, whereas
the AFM image of the etched surface was binary-processed
to eliminate the transgranular data. Figure 6(b) shows the
boundaries with red lines on the image of the electropolished
surface to assist in the identification of the positions of the
prior γ grain boundaries.
3.4 Carbide in microstructure
3.4.1 Identification of carbide particle
The properties of carbides on prior γ grain boundaries and
the mechanisms of the intergranular cracks such as delayed
fracture3–6) have been thought to have a close correlation. The
identification of prior γ grain boundaries enabled us to iden-
tify the grain boundary carbides on an electropolished sur-
face. Figures 7 and 8 show AFM images after the image anal-
ysis of the carbides on the electropolished surface. In Figs.
7(a) and 8(a), all the carbides are marked in red. Figures 7(b)
and 8(b) show the carbide distributions on the three types of
boundaries: the prior γ grain boundaries that are identified
using the superimposition method, the packet boundaries that
are identified as the boundaries between the regions in which
black and white strips have parallel alignment, and the block
boundaries that are identified as the boundaries between the
blocks that are aligned in the same direction. The carbides on
these boundaries are marked in green.
Since the carbides project out from the electropolished
surface, the particle analysis program recognizes the sharp
changes in the levels as the outline of the particles (see Fig.
11(b)), and therefore can identify all the particles regardless
of their size. To distinguish the carbide particles from the
1761
(a)
5 m
(b)
5 m
Fig. 6 AFM images showing prior γ grain boundaries, blocks and carbides
for the JIS-SCM440 steel tempered at 723 K. (a) is the image created by
superimposing Fig. 5 on Fig. 4. (b) is the image showing prior γ grain
boundaries, identified in Fig. 4 with red lines.
simple surface roughness, the program was set to ignore the
differences in the level of 5 nm or smaller. In addition, fewer
carbide particles may have been counted, since small carbide
particles are likely to exfoliate during the process of elec-
tropolishing.
The square root of the plane area occupied by the carbide
particles was assumed to be the particle diameter. The his-
tograms of particle diameter that correspond to Figs. 7 and 8
are shown in Figs. 9 and 10, respectively. Besides these two
images, the carbides in two other images were analyzed. The
results are summarized in Table 1. Table 1(a) is for all the
carbides in a 5 µm × 5 µm field, Table 1(b) is for the carbides
on the prior γ grain boundaries, Table 1(c) is for the carbides
on the packet boundaries, and Table 1(d) is for the carbides
on the block boundaries.
Table 1(a) shows that the number (N ) of carbides in the
5 µm × 5 µm field was almost constant at 1950 to 2350 for
all four images. Table 1(b) shows that the number (Nγ ) of
carbides on an L γ 3.6 to 5.5 µm-long section of the prior γ
grain boundary was 44 to 77 (10 to 15 particles per µm). The
1762 M. Hayakawa, S. Matsuoka and K. Tsuzaki

Table 1 Summary of carbide analyses of four AFM images 5 µm × 5 µm in size.

(a) All carbide particles
AFM image Fig. 7(a) Fig. 8(a) — —
The total number of carbide particles, N 2344 1957 2292 2314
The apparent area fraction of all carbide particles, f app (%) 10 11 9.8 12
The apparent mean carbide particle diameter, Dapp /nm 30 34 30 33
The corrected mean carbide particle diameter, Dest /nm 23 25 24 23
The corrected mean carbide particle diameter derived
from the tip radius, dest /nm 20 24 20 23
(b) Carbide particles on prior γ grain boundaries
AFM image Fig. 7(b) Fig. 8(b) — —
The number of carbide particles on
a prior γ grain boundary, Nγ 67 45 77 44
The length of a prior γ grain boundary, L γ /µm 5.5 3.6 5.3 4.4
The corrected mean carbide particle diameter on
a prior γ grain boundary, Dest,γ /nm 31 29 27 29
The corrected occupied fraction of the carbide particles
on a prior γ grain boundary, f γ (%) 53 50 52 31
(c) Carbide particles on packet boundaries
AFM image Fig. 7(b) Fig. 8(b) — —
The number of carbide particles on a packet boundary, Np 26 50 71 18
The length of a packet boundary, L p /µm 2.3 2.7 4.3 1.6
The corrected mean carbide particle diameter on
a packet boundary, Dest,p /nm 27 25 28 29
The corrected occupied fraction of the carbide particles on
a packet boundary, f p (%) 35 52 55 38
(d) Carbide particles on block boundaries
AFM image Fig. 7(b) Fig. 8(b) — —
The number of carbide particles on a block boundary, Nb 15 17 20 9
The length of a block boundary, L b /µm 1.1 1.0 1.3 1.1
The corrected mean carbide particle diameter on
a block boundary, Dest,b /nm 21 27 26 32
The corrected occupied fraction of the carbide particles
on a block boundary, f b (%) 39 45 40 44
(e) Carbide particles on 5 section lines dividing an AFM image into 6 equal parts in the horizontal direction
AFM image Fig. 7(a) Fig. 8(a) — —
The total number of carbide particles on 5 section
lines of total length 25 µm, Nsec 77 74 78 94
The corrected mean occupied fraction of the carbide
particles on given lines, f sec (%) 7.8 7.1 8.0 8.9

program counted the carbide particles separately if there was
a difference in the level on their surfaces, even when the par-
ticles on the γ grain boundaries appeared to be continuous.
To investigate the size of the carbides, the monitored values
were corrected to remove the effects of the shape of an AFM
cantilever probe.
3.4.2 Correction of the size of carbides
The shape of the AFM cantilever probe is known have a
strong effect on the particle size estimation in AFM images,
if the particles are smaller than the size of the probe tip (5–
10 nm).8) Therefore, the methods for correcting the apparent
sizes of the carbides and removing the effects of the probe’s
shape were investigated.
(1) Correction using the area percentage of cementite
In this specimen, the carbide that precipitated was cemen-
tite. If all the carbon is converted into cementite, the volume
percentage of cementite f (vol%) can be expressed as a sim-
ple approximate expression:9)
f = 15.3c (1)
 Microstructural Analyses of Grain Boundary Carbides of Tempered Martensite
(a)
1 m
(b)
Block
boundary
Packet
boundary
1 m
Prior grain boundary
Fig. 7 AFM images after the image analysis of the carbides on the elec-
tropolished surface of the JIS-SCM440 steel tempered at 723 K. In (a), all
the carbides are shown in red; and in (b), the carbides on the prior γ grain
boundary, the typical packet and block boundaries are shown in green.
where c (mass%) is the carbon content. Since the carbon
content c is 0.4 mass%, its volume percentage should be
6.12 vol%. If the field of observation is much larger than
the size of the cementite particle and if the distribution is
isotropic, the area percentage of the plane should also be
6.12%.
Assuming that both the quantity of carbides that exfoliate
and the oxide films that are formed during electropolishing
can be ignored and that the field of the AFM image is much
larger than the carbide particles so that the theoretical area
percentage is effective, the expansion e of the carbide area,
which is caused by the AFM probe, can then be expressed as
e = f app /6.12 (2)
Where f app (%) is the area percentage of carbides
(=cementites) determined from the AFM image.
When the apparent size of the carbides in the AFM image
is designated as Dapp , the corrected size Dest of the particle
1763
(a)
1 m
(b)
Prior grain
boundary
1 m
Packet Block
boundary boundary
Fig. 8 AFM images after the image analysis of the carbides on the elec-
tropolished surface for the JIS-SCM440 steel tempered at 723 K. In (a), all
the carbides are shown in red; and in (b) the carbides on the prior γ grain
boundary, the typical packet and block boundaries are shown in green. The
observed area is different from that in Fig. 7.
after the effect of the AFM probe is removed is expressed as
√
Dest = Dapp / e (3)
The apparent carbide area percentage f app in each field,
which we derived from the AFM images (5 µm × 5 µm =
25 µm2 ) listed in Table 1(a), showed no significant difference
among the four fields and was 10 to 12%. Since the apparent
mean particle diameter Dapp was 30 to 34 nm, we obtained a
corrected particle diameter Dest of 23 to 25 nm from eqs. (2)
and (3).
(2) Correction using the radius of curvature of the probe
tip
This section describes the method for correcting the appar-
ent sizes of the carbides using the shape of the probe tip. As
schematically shown in Fig. 11(a), the tip of the probe moves
along the dashed line. The angle θ of opening is almost zero at
the tip of the probe. Therefore, when the height of the carbide
1764 M. Hayakawa, S. Matsuoka and K. Tsuzaki

(a) Estimated (carbide area), Dest/nm (a) Estimated (carbide area), Dest/nm
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
0.4 0.4

0.3 All carbides in Fig.7 0.3 All carbides in Fig.8

Frequency

Frequency
0.2 0.2

0.1 0.1

0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
(carbide area), Dapp / nm (carbide area), Dapp/nm

(b) Estimated (carbide area), Dest, /nm (b) Estimated (carbide area), Dest, /nm
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
0.4 0.4
Carbide particles along the
0.3 prior grain boundary 0.3 Carbides along the
Frequency

Frequency
prior grain boundary
0.2 0.2

0.1 0.1

0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
(carbide area), Dapp, /nm (carbide area), Dapp, Y /nm
(c) Estimated (carbide area) , Dest,p/nm (c) Estimated (carbide area), Dest,p/nm
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
0.5 0.4
0.4 Carbides along
the packet boundary 0.3 Carbides along
Frequency

Frequency
0.3 the packet boundary
0.2
0.2
0.1 0.1
0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
(carbide area), Dapp,p/nm (carbide area), Dapp,p/nm
(d) Estimated (carbide area), Dest,b/nm (d) Estimated (carbide area), Dest,b/nm
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
0.5 0.4
0.4 Carbides along
the block boundary 0.3 Carbides along
Frequency

Frequency

0.3 the block boundary

0.2
0.2
0.1 0.1
0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
(carbide area), Dapp,b/nm (carbide area), Dapp,b/nm

Fig. 9 Distributions of the square root of the carbide-occupied area mea- Fig. 10 Distributions of the square root of the carbide-occupied area mea-
sured in Fig. 7 for the JIS-SCM440 steel tempered at 723 K. All the car- sured in Fig. 8 for the JIS-SCM440 steel tempered at 723 K. All the car-
bides are analyzed in (a); and the carbides on the prior γ grain, the packet bides are analyzed in (a); the carbides on the prior γ grain, the packet and
and block boundaries are analyzed in (b), (c) and (d), respectively. block boundaries are analyzed in (b), (c) and (d), respectively.

from the surface is not too high, the diameter dest of the car-
bide, from which the effect of the probe shape is eliminated,
is given by
dest = Dapp − 2r
Where r and Dapp are the radius of curvature of the probe tip
and the apparent diameter of the carbide particle.
A cross section profile of an AFM image is shown in Fig.
11(b). Since the values along the vertical direction are not
affected by the probe tip, the heights of the carbides are 10
to 20 nm. On the other hand, the r of the probe tip is deter-
mined to be 5–10 nm. Assuming that r is 5 nm, eq. (4) shows
that the diameters dest of the carbides, from which the effect
of the probe shape was eliminated, are 20 to 24 nm from the
Dapp values of 30 to 34 nm, which are shown in Table 1(a).
These dest values are similar to the corrected Dest values (23
to 25 nm) that were determined using the area percentage in
the previous section. Since the apparent percentage of the
carbide-occupied area f app was almost the same (10 to 12%)
in all four fields in Table 1(a), the shape of the probe tip are
likely to be similar.
The previous paper1) reported that the diameter of the car-
(4) bide, which was estimated from a TEM image of the same
specimen, was 20 nm. The above corrected values and this
value are close, supporting the effectiveness of the correc-
tion methods for the AFM images. Therefore, the method
described in Section (1) is effective for correcting the sizes of
the carbide particles and quantitatively analyzing the distribu-
tion of carbides.
3.4.3 Distributions of carbides on the boundaries
Prior γ grain boundaries, packet boundaries and block
boundaries with high angle boundaries tend to be preferen-
tial sites for carbide precipitates. In particular, prior γ grain
boundaries contain a number of large carbide films, which are
likely to cause deterioration of intergranular cracks.10) To un-
derstand the mechanisms of intergranular cracks being related
to delayed fracture, it is useful to estimate the percentage of
grain boundary area covered by carbides and other informa-
 Microstructural Analyses of Grain Boundary Carbides of Tempered Martensite
(a)
Probe tip
r
Sample
surface Dapp
Carbide
(b)
Carbide height
25
Vertical Distance, Lv/nm
0 Fig. 12 Schematic drawing of grain boundary carbides and transgranular
25
0 1
Horizontal Distance, L H / m the carbides on the clearly visible packet and block bound-
Fig. 11 Schematic drawing of a vertical cross-section of the sample. The
broken line represents the trace of the movement of the AFM probe tip
in (a). Carbide heights and intervals between carbide precipitates were
measured from the real profile of the vertical cross section from an AFM
image in (b).
tion from the AFM images quantitatively. Table 1(b) shows
the corrected mean diameters of the carbides on the prior γ
grain boundaries Dest,γ . Since Dest,γ values are 27 to 31 nm,
they are larger than those for all the AFM images, Dest of 23 to
25 nm in Table 1(a). The histograms of the carbide diameter
in Figs. 9(b) and 10(b) show that many of the carbide particles
on the prior γ grain boundaries were larger than 50 nm. Since
these histograms use the square roots of the areas occupied by
the carbides, the shapes of the carbides are not reflected in the
histograms. The carbide particles that are larger than 50 nm
in Figs. 9(b) and 10(b) correspond to the carbide films that are
longer than 100 nm.
To represent the distribution of carbide films along the
grain boundaries, the percentage of the length covered by
the carbides to the total length (L γ ) of the grain boundary
( f γ = (Σlestγ )/L γ × 100%) was measured after correct-
ing the sizes of carbides, as shown in Fig. 12, where lestγ
is the corrected length of the carbide along the long sec-
tion of the grain boundary. On the other hand, the mean
percentage of the transgranular area covered by the carbides
( f sec = (Σlest )/L × 100%) was estimated. lest and L are the
correcting length of a carbide section which crosses the lines
dividing an AFM image and 25 µm in total with five 5 µm-
long lines dividing the AFM image, which is 5 µm × 5 µm
in size, into six (see Fig. 12). Table 1(b) shows that the f γ
values were 31 to 53%, and over 50% of their boundary was
covered by carbides in three out of the four AFM images. On
the other hand, the mean percentage of the transgranular area
covered by the carbides, f sec , was 7.1 to 8.9% for each of the
four AFM images.
Figure 6 shows that blocks can be identified by the black
and white contrasting shapes on the electropolished surface.
Packets were also easily identified, since the AFM images
1765
est
Carbide est
5 m
est
est
Prior
G.B.
5 m
carbides in an AFM image.
2 clearly showed the prior γ grain boundaries. For each image,
aries were analyzed. Figures 7(b) and 8(b) show the mea-
surements of the carbides for these two kinds of boundaries,
respectively. Figures 9(c), 9(d), 10(c), and 10(d) show the
histograms. As shown in Tables 1(c) and (d), the mean sizes
of the carbides on the packet and block boundaries Dest, p and
Dest,b were 25 to 29 nm and 21 to 32 nm, respectively, and
were larger than the mean Dest of 23 to 25 nm. The fractions
of the packet and block boundaries occupied by the carbides
( f p and f b ) were 35 to 55% and 39 to 45%, respectively, and
were larger than the transgranular percentage ( f sec ) of 7.1 to
8.9%, although the difference was smaller than that for the
prior γ grain boundary f γ .
As described above, we analyzed the distribution of car-
bides, which vary in size from several nanometers to over
100 nm and are present at rates of several thousand per AFM
image quantitatively by applying the particle analysis pro-
gram. The identification of prior γ grain boundaries was es-
pecially significant. Although we may need to obtain more
data by increasing the number of images, our methods gen-
erate a method for investigating the relationship between the
grain boundary carbides and the properties of steel, such as
the resistance to delayed fracture.
4. Summary
This study, which describes AFM microstructure observa-
tion methods, enables us to identify the carbides and prior
γ grain boundaries on the tempered lath martensite using
the same field of observation. After experimenting with a
medium-carbon low-alloy steel (JIS-SCM440) specimen that
had been austenized at 1153 K and tempered at 723 K, our
conclusions are as follows:
(1) We could observe the same location on a specimen
surface after it had been electropolished and after it had been
further etched with a picric acid solution by AFM using a
micro Vickers indent. Carbides and blocks were clearly rec-
ognizable on the electropolished surface and prior γ grain
boundaries were identified on the picric acid-etched surface.
(2) Superimposing the two AFM images permits the
1766 M. Hayakawa, S. Matsuoka and K. Tsuzaki
identification of prior γ grain boundaries on the electropol-
ished surface.
(3) Carbides on the prior γ grain boundaries could be
distinguished from transgranular carbides on the electropol-
ished surface, and the sizes of carbides were measured using
a particle analysis program. We calculated the increase in the
carbide area due to the radius of the curvature of the AFM
probe tip and developed methods for correcting the readings
for sizes of the carbides.
(4) The sizes and the distribution of the carbide particles
on the prior γ grain boundaries were shown to be larger than
the mean of all the carbides quantitatively.
Acknowledgements
The authors would like to thank Dr. Toshihiko Takahashi
and Dr. Toru Hara of the National Institute for Materials Sci-
ence for providing valuable advice.
REFERENCES
1) M. Hayakawa, T. Hara, S. Matsuoka and K. Tsuzaki: J. Japan Inst.
Metals 64 (2000) 460–466.
2) T. Hara, S. Yusa and K. Tsuzaki: Proc. of Intern. Conf. on Solid-
Solid Phase Transformations ’99, eds. by M. Koiwa, K. Otsuka and
T. Miyazaki (The Japan Institute of Metals, 1999) pp. 1549–1552.
3) C. J. McMahon, Jr.: Hydrogen Effects in Metals, eds. by I. M. Bernstein
and A. W. Thompson (TMS-AIME, 1981) pp. 219–234.
4) F. Nakasato and F. Terasaki: TETSU-To-HAGANE 61 (1975) 856–868.
5) R. M. Horn and R. O. Ritchie: Metall. Trans. A 9A (1978) 1039–1053.
6) S. Yusa, T. Hara and K. Tsuzaki: J. Japan Inst. Metals 64 (2000) 1230–
1238.
7) ASTM Designation E112-82, 1983 Annual Book of ASTM Standards,
03.03 (1983) pp. 121–155.
8) H. Masuda: Ferrum 3 (1998) 884–890.
9) S. Takaki: 141 • 142th Nishiyama Memorial Technology Lecture, ed.
ISIJ, Tokyo, (1992).
10) G. Krauss and C. J. McMahon Jr.: Martensite, eds. by G. B. Olson and
W. S. Owen (ASM International, New York, 1992) pp. 295–321.