Beruflich Dokumente
Kultur Dokumente
3871-3886, 1997
m,C1997 Acta Metallureica Inc.
Published by Elsevier S%sm Ltd. All rights kserved
Printed in Great Britain
PII: S1359-6454(97)00027-X
1359.6454197 $17.00 + 0.00
Pergamon
Sciences,
and N. HANSEN2
Sandia
National
9 August
Laboratories,
Livermore,
CA 94550,
1996)
Abstract-Deformation
of metals from medium to high strains introduces
significant changes in the
microstructure
and the texture. The microstructure
evolves into a lamellar structure with boundaries
of
small to medium misorientation
angles mixed with high angle boundaries.
The latter category consists
of deformation
induced boundaries
plus the original grain boundaries.
The number of deformation
induced high angle boundaries is significantly larger than the number of original grain boundaries.
Mechanisms for the formation of the deformation induced boundaries are suggested based on grain
subdivision processes which can lead to formation of different texture components
within an original grain.
The distribution
of their misorientations
is estimated
based on these mechanisms.
This estimate is
compared
to experimental
findings for Al, Ni and Ta deformed to large strain by rolling or in torsion.
This estimate and the findings are discussed and good support is established for the basic assumption
that
grain subdivision accompanied
by a strong texture evolution can lead to a very significant increase in the
in a deformed metal. These findings provide the essential physical
fraction of high angle boundaries
background for the construction of theoretical models for the distributions. C 1997Acta Metallurgica Inc.
1. INTRODUCTION
Heavily cold worked metals are subdivided by grain
boundaries
and dislocation
boundaries
which are
arranged in a lamellar or subgrain structure [lb3].
The frequency and distribution
of these boundaries
determines
the properties
of the deformed
metal,
including
the flow stress, texture, recrystallization
behavior
and the formability.
A comprehensive
review of the general microstructure
and texture for
the large strain state of metals was last given in 1980
in Ref. [4].
The boundaries
in heavily cold worked metals
originate not only from grain boundaries
present in
the undeformed
metal, but also from dislocation
boundaries which form during plastic deformation
in
grains that are subdivided
into regions that are
smaller than the original grain size. Qualitatively
it
has been known that these boundaries
have a large
angular spread and that the misorientation
across
many boundaries is of the magnitude characteristic
of
ordinary high angle boundaries
[2]. A quantitative
analysis of this angular spread has, however, only
recently
been possible
by the introduction
of
semi-automatic
and automatic
microscopic
techniques for the determination
of local crystallographic
orientations
on the micrometer
and submicrometer
scale [e.g. 5-101.
As a result, a framework
for the formation
of
dislocation
boundaries
during plastic deformation
has been formulated
[e.g. 111. This framework
has
been supported
by many quantitative
experimental
observations
of a variety of metals [e.g. 51. Both this
2. PREVIOUS WORK
The development
of boundaries with a large spread
in misorientations
has been known qualitatively since
the discovery
of grain break-up
by Barrett and
Levenson [e.g. I]. Further, it has been shown that the
misorientation
across many of these boundaries is of
a magnitude characteristic
of high angle boundaries,
e.g. through X-ray, optical and electron microscopic
techniques [14]. High angle boundaries
are defined
as those with angles above 15-20 [12]. Because the
size scale of these boundary spacings is frequently less
it is difficult
to obtain
than one micrometer,
quantitative
microstructural
and crystallographic
data. Thus there are but a handful of reports that
have observed and measured individual high angle
deformation
boundaries
following medium to large
strain deformation.
These include the observations
in
wire drawn Fe-Si [2], an unstable Al single crystal
[ 131, friction
following
channel
die deformation
deformed Cu [14], torsion deformed Al [15], rolled
Al-Zr-Si
alloy containing
particles [ 161, aluminum
rolled to intermediate
strains [17] in heavily rolled
nickel [18], and heavily rolled aluminum [19, 201. A
statistical analysis of the expected grain boundaries
compared
to the observed
high angle boundaries
shows that most of these high angle boundaries
are
formed by the deformation
process at the places
where grains subdivide and are not original grain
boundaries
[ 181. These different observations
show
3871
HUGHES
3872
and HANSEN:
GRAIN
microstructures
andgrain subdivision
SUBDIVISION
MECHANISMS
large strains
small strains
HUGHES
and HANSEN:
GRAIN
(a)
n
SUBDIVISION
MECHANISMS
3873
(b)
I
1.0 I
I\
GNBs
-1
5% cr
?? 10% cr
A 30% cr
T 50%/a
fit
0
W
A
I
5%cr
10% cr
30% cr
50% cr
fit
e/e,,
e/e,
Fig. 2. Probability
density functions of the boundary
misorientation
angles, normalized
by the average
misorientation
angle, scale to the same function for 5-50% cr. or 6~ = 0.06-0.80. (a) Scaled probability
density for IDB misorientation
angles. The curve fit for all of the IDB data is given by equation (2) with
CI= 3. (b) Scaled probability
density for GNB misorientation
angles. The curve fit for all of the GNB data
is given by equation (2) with IX= 2.5 [36, 521.
P(Q,&I = we/e:)
(1)
.m) =
CP
r(a)
3874
HUGHES
and HANSEN:
the word
coined
to describe
a misorientation
calculated based on a minimum angle relationship
between crystallites
determined
by considering
all
[39].
symmetry operations
3.2.2. Texture mechanisms. Of equal importance
to the origin
of high
angle
boundaries
are
mechanisms
involving
grain subdivision
and the
evolution of a preferred texture during deformation,
including:
(i) the rotation of a
preferred
crystal
mation;
(ii) ambiguity of slip
orientations
that
within a grain.
(4
0.1
0.08
contribution from
islocation accumulation
-: 0.06
Q)
g
0.04
0.02
0
(b,
10 20 30
Disorientation
40 50
(deg)
60
texture development:
equal probability of each
ideal component as a
nearest neighbor
20
Disorientation
40
(deg.)
60
GRAIN SUBDIVISION
F.c.c. rolling
MECHANISMS
deformation
conditions
F.c.c. torsion
Crystal orientation
jhkl] (unv)?
Label
Cube (D)
Rotated cube (RC)
s (SI, s:. SI, Sd)
Copper (C, and C-1)
Brass (B and B?)
Goss (G)
a fiber
/r fiber
(I li)[2ii]
EXPERIMENT
Cube (D)
Rotated cube (RC)
Goss (G)
Taylor (TI)
Taylor (Tz)
Cross Taylor (CT,)
Cross Taylor (CT?)
jl00}<00l>
{ooi~(iio)
(O_llI;(lOO>
(lll)[211]
(l I l)[2iil
(iii)[oii]
(ii I)[OI I]
The properties
of the three materials
and the
various deformation
conditions used in this study are
shown in Table 2. Thin foils for transmission
electron
microscopy (TEM) were made to view the samples in
the longitudinal side plane for rolling or the Z-theta
plane for torsion.
(Z is the shear plane normal,
theta is the shear direction.)
Transmission
electron
microscopy and convergent beam diffraction analysis
of these samples was performed.
Orientations
of
individual
crystallites
were obtained
from
the
convergent
beam Kikuchi patterns.
Crystallites
as
small as 40 nm could
be measured
with this
technique. The Kikuchi patterns were analyzed using
a computer method based on the techniques in Refs
[IO, 421 to obtain
the orientation
matrices
for
individual crystallites. The minimum angle misorientation relationship
(disorientation)
between adjacent
crystallites separated by dislocation
boundaries
was
calculated by considering all 24 symmetry operations
for the orientation
matrices in a standard manner,
The axis/angle pairs for the disorientations
were also
calculated. The axis/angle pairs were quantitatively
spacing
following
deformation
direction.
respectively,
in rolling,
or the shear
plane and
and observation
or : dire&Ion
for torsion
Spacmg
Material and deformation
condltlonst
Nickel (99.99%) f.c.c.
80 flrn initial grain size:
Rolhng 70% cr. (<>M= 1.4)
Rollin 90% cr. ((,M = 2.7)
Rolling 98% cr. (fthl = 4.5)
Torsion (c.~ = 4.1)
Alummum (99.5%) f.c.c.
90 pm mitral grain sire:
Rolling 90% cr. (c,, = 2.7)
Aluminum (99.8%) f.c.c.
I50 x 300 pm mltial grain size:
Rolling 90% c.r. (c\~ = 2.7)
Tantalum (99.99%) b.c.c.
44 pm imtial grain size:
Rollmg 90% x.r: ((ih, = 2.7)
Original grain
boundaries$ (pm)
All dislocation
boundaries (pm)
High angle
boundaries. H > IS (pm)
Observation
length (pm)
32.0
8.0
1.6
11.5
0.40
0.22
0.16
0.17
0.72
0.53
0.37
0.34
13.7
12.2
7.7
9.6
9.0
0.50
2.00
38.X
15.0
0.35
2.10
42
4.4
0.25
1.80
shear
compared
to axis/angle
pairs for coincident
site
lattices. The disorientation
axes were also compared
to the sample axes of the adjacent crystallites.
A
negative or positive disorientation
angle was assigned
by considering whether the disorientation
axis is in a
left hand or right hand triangle, respectively.
Measurement
errors for these Kikuchi methods
include errors that arise from measuring the Kikuchi
pattern itself and cumulative errors in the measured
orientations
due to slight overall bending in the thin
foils. Errors in both the orientation
and misorientation between adjacent crystallites based on pattern
measurement
are 0.1 to I, as examined
in Refs
[lo. 421. Note that orientation
angle changes due to
overall foil bending are restricted by both the thick
supporting
sample rim and the geometry
of the
sample holder and microscope.
Furthermore,
these
errors were carefully examined using single crystal
thin foils measured over large distances of 500 to
1000 pm. Such careful measurement
showed that
cumulative
errors in the orientation
measurement
that would develop owing to slight foil bending are
only about 2 over 100 kirn of sample [43]. Additional
evidence
for the goodness
of TEM orientation
analyses is found in Ref. [44] in which orientations
in
AND RESULTS
4.1. E,xperirnent
Crystal orientation
{hk/J<uow)i
Label
Fiber linkmg
the sequence:
BI, Sz. CI, SI.
B:, Si, Cz, SI
tThe designation
jhk/)(uw)
refers either to the rolling
direction, respect&y
for torsion.
4.
B.c.c. rolling
Crystal orientation
(hk/)(ucw)1_
Label
3875
3876
3877
3878
HUGHES
and HANSEN:
GRAIN
SUBDIVISION
MECHANISMS
GRAIN
distributions
SUBDIVISION
MECHANISMS
3879
a disorientation
of less than 15- relative to the ideal
component.
An orientation
not within 15 of any
ideal component is taken to be a random orientation.
The ideal components
considered
throughout
the
paper are listed in Table 1. The complementary
variants of the same ideal component
were also
distinguished,
since these variants differ from each
other by high angle rotations (e.g. 60^ about (1 11)).
The most striking features in Fig. 6(a)-(e) are the
large number
of high angle disorientations
that
accompany a very fine scale pattern of local texture
components.
Note that most (>2/3) of the very high
angle changes, >30 , in Fig. 6(a))(e) separate two
different ideal texture components.
In contrast, only
l/4 separate an ideal component
from a random
component
and less than l/l5 separate two random
components.
Remarkably,
owing to the many high
angle changes,
nearly
all of the ideal texture
components
and their variants are frequently found
over a distance of just a few micrometers
in all these
various structures. For f.c.c. rolling, the orientations
can be seen to span a wide range of the x and /Ifibers
plus some random
components
[see especially
Fig. 6(a), (b) and (d)]. These orientations
are
frequently arranged in space so that they alternate
back and forth along the length of the c( and /r fibers.
Analogously,
for torsion, observed orientations
span
and alternate along the {l I ~)(uz~M~) and (Ml){ 110)
fibers [Fig. 6(c)]; for b.c.c. rolling all of the ideal
components
along the {OOl}(urn~) or (111 ](uw~)
fibers are found [Fig. 6(e)]. Strikingly, this trend can
already
be observed
at a 70% c.r. (E,~ = 1.4)
[Fig. 6(a)] and is maintained
with increasing strain.
At these large strains the random components were
scattered throughout
the microstructure.
Although
random components
could be found in regions with
either S-bands or localized glide bands, the frequency
of the random components
was not any higher than
elsewhere.
In fact the localized glide bands were
typified by having a normal deformation
texture. e.g.
copper or S.
It must be noted that texture components
of a
different color are not always separated by high angle
boundaries
in Fig. 6. This result occurs as a
consequence
of the texture grouping we use that is
based on the classification of an orientation
as ideal
if it is within I5 of the ideal, combined with the fact
that adjacent ideal orientations
along the b fiber are
only 19.4 apart.
The local measurements
also show that each of the
deformation
texture components
is composed
of a
very large number of small volumes with orientations
scattered around the ideal orientations.
An estimate
of the size of such volumes for rolling shows that their
average dimension in the ND direction may be l/3 to
l/5 of the reduced
grain size (Table 2). The
dimensions
in the RD and TD directions gives an
average aspect ratio for the pancake shaped volumes
of about 335. This average aspect ratio is somewhat
misleading because of the wide distribution of lengths
for the pancake shaped volumes. Within lamellar
3880
HUGHES
and HANSEN:
GRAIN
SUBDIVISION
In
.!2
-a
5
p
z
G
20
Disorientation
40
()
60
MECHANISMS
10
8
6
4
20
40
Disorientation
(deg.)
60
40
20
Disorientation
(deg.)
60
in
40
20
Disorientation
(deg.)
60
iii
20
Disorientation
40
(deg.)
iii
60
40
20
Disorientation ()
60
Fig. 5. Histograms
showing the distribution
of the magnitude
of disorientations
across dislocation
boundaries measured along either the normal direction for rolling or the z direction for torsion. Note that
the histograms
generally show two peaks in their distribution.
The disorientation
axes are plotted in
standard triangles. Axes for 0, ISI > 35; A, 18 < /0/ i 35; 0,]0\
< 18. (a) Nickel 70% cr.,(b) nickel
90% cr., (c) nickel 98% cr., (d) nickel deformed by torsion, (e) aluminum 90% c.r., smaller grain size
and (f) tantalum 90% x.r.
Cd)
_0
(b)
B,
4
6
8
10 12
Distance along ND (Fm)
mm~~mm
B2 C2 S1
S, S,
-60
__
14
.,00Dl04
5
10
15
Distance along
20
25
30
ND (Fm)
(e)*
m
-(001)[~10](001)<uvw>(111)[011] (lll)filZ]
random
W?
(lll)[i2l](lll)<uvw>
60
40
1
2
3
4
5
Distance along ND (pm)
60
20
B
Bg
i:
8
-20
-40
7
;
-60
u
mm mm%3
A, A 2* A, A2 B1 Bz C 110 fiber111 fiber random
CC)s
2345678
Distance
along
ND (Km)
60
40
a
z
.2
m
z
z
s:
a
20
0
-20
-40
-60 _
Distance
6
8
along 2 axis (pm)
10
The efects
crystal structure,
The
formation
of grain orientation,
material purity,
of
high
angle
boundaries
mode
and
sr.C-l.RD*
10
15
20
25
30
35
4u
Fig. 7. Disorientation
angles measured across dislocation
boundaries
within a single lamellar band in the rolling
direction.
The color shading shows that similar texture
components
are maintained
within the band.
3882
GRAIN
local orientations
may be affected
by several
parameters
including
grain orientation,
material
purity, crystal structure, strain level and deformation
mode.
4.5. I. Orientation effkcts. In general the high angle
boundaries
are distributed
widely throughout
the
distances measured. These high angle changes occur
both in the grain centers and near original grain
boundaries. At the same time, there are quiet regions
with only low to medium angle changes and regions
with clusters of high angle boundaries.
For the case
of f.c.c. rolling, these quiet regions were associated
with average grain orientations
along the c1 fiber,
including
Goss and brass components
[see, e.g.
Fig. 6(d)]. For b.c.c. rolling these regions were
primarily
associated
with the rotated cube orientations (see, e.g., Fig. 6(e) and Ref. [53]). In no case
did these regions with low to medium angle changes
extend across the whole grain width; rather they
occupy fractions of a grain, even for the rotated cube
orientation
in tantalum.
In contrast
to the quiet
areas, the clusters of high angle changes in f.c.c.
rolling are associated
with alternations
primarily
along the /jfiber including finely distributed
regions
with the variants
of S, copper
and brass plus
occasionally a Goss orientation.
No clear orientation
effects could be identified for the case of torsion,
although somewhat larger frequencies of high angle
changes are associated with orientations
along the
(111) fiber compared to the (110) fiber.
4.5.2. Material purity and type. Similar numbers
of high angle boundaries
and varieties
of local
orientations
are observed for a range of material
purity from 99.999% to 99.5% (Table 2 plus Ref.
[ 191). However,
significantly
more
high
angle
boundaries
were observed in nickel than in aluminum, consistent with a finer spacing of dislocation
boundaries
in nickel compared to aluminum.
4.5.3. Crystal structure. The different deformation
textures
that develop reflect the different
crystal
structures, f.c.c. or b.c.c. as expected. However, both
the b.c.c. and f.c.c. metals show similar trends with
respect to the number of high angle boundaries and
texture components.
However, the average number of
texture components
in b.c.c. tantalum, 24, is smaller
than that for f.c.c. 2-7, at the same strain. This
difference is partly related to the stability of the
rotated cube grains in the tantalum and the larger
number
of ideal texture
components
for f.c.c.
compared to b.c.c.
4.5.4. Grain size and shape effects. The grain sizes
in this study ranged from medium-small
(44 pm) to
medium-large
(150 p). Within this grain size range no
observable
effects of either the initial or deformed
grain size were observed.
For example, the same
average spacing of high angle boundaries is observed
for aluminum with two different initial grain sizes
(Table 2). Both aluminum samples showed the similar
varieties of high angle boundaries
and local texture
components.
All of these suggest a minor role of the
SUBDIVISION
MECHANISMS
Boundary formation
GRAIN
SUBDIVISION
MECHANISMS
3883
orientations
in creating
high angle boundaries
depends on the starting texture of the material. While
this mechanism
is a potent source of high angle
boundaries
in polycrystals
with strong
starting
textures such as cube, its potency diminishes with an
increasingly
random
starting
texture since fewer
regions are near these special orientations.
For the
weakly
textured
materials
in this study,
this
mechanism is only of small to moderate importance.
5.1.6. Etolving texture. A more potent source of
high angle boundaries may be found if one considers
the whole of texture evolution
and not just the
evolution
of special orientations.
Large crystal
rotations are required to bring the starting texture to
the final preferred texture.
The above hypothesis is supported by considering
crystal rotations
using a Taylor model. During
deformation,
an initially random grain is reoriented
along a path in orientation
space that is defined by
the rotation
or reorientation
velocity
field in
orientation
space [58]. Since both the magnitude and
direction
of this velocity depends
on location in
orientation space, each part of a subdivided grain will
have a somewhat different reorientation
velocity and
direction.
While this rotation
field is generally
smooth, there are regions in which a small change in
orientation
causes a reasonably large change in the
velocity.
Large
differences
in velocity
between
adjacent crystallites in a grain result in the evolution
of large differences in orientation and misorientation.
As texture evolves, a random subdivided grain will
approach
regions with large velocity changes
as
follows.
The characteristic
structure of the rotation field
causes crystals of any general or random orientation
to be funneled into orientation streams flowing to one
end orientation.
However, as the crystal volumes of
a subdivided grain are swept along these reorientation streams, the volumes rotate near one or more
surfaces in orientation
space that separate streams
with branching
paths to different end orientations.
Because a grain has been subdivided,
it contains a
range of crystal orientations,
each with its own
rotation rate and direction, since the latter depend on
the current orientation
with respect to the deformation axes. Consequently,
different parts of a grain
will approach these branching paths at different times
and locations in orientation
space. These differences
will cause some parts of a grain to follow one branch
while others will follow a different branch, thereby
creating some very high angle boundaries
when the
different end orientations
are reached. The initial
crystal orientation and deformation
mode will dictate
the degree of subdivision necessary to cause this large
scale splitting. For special orientations
already near
the surfaces
in orientation
space that separate
branching
paths,
grain subdivision
into regions
separated by only a couple of degrees of misorientation is required. In contrast, for orientations
that are
farthest away from these branches, grain subdivision
3884
of
The prediction
based on the addition of distributions in Fig. 3 based on both microstructure
and
texture provides good qualitative agreement with the
experimental
distributions
in Fig. 5. While the
simplified model represented
in Fig. 3 captures the
main features
observed,
it is not meant to be
quantitative.
For example, the distribution
peak at
the small angle range occurs at a somewhat larger low
angle in the predictions than in the experiments. This
moderate difference is indicative of the interactions
between the dislocation and the texture mechanisms
which we know exist, but do not account for in the
simple model. Also the contribution
to the distribution from the dislocation mechanisms
is based on
an extrapolation
of both the average angle and the
scaling observed at small to moderate strains. This
extrapolation
is acceptable for strains below c,~ = 2.7
(90% c.r.), although
it slightly over-predicts
the
average angle from the dislocation contribution.
From the experimental
results, we know that
dislocation
boundaries
form which subdivide
the
original grains and that a preferred texture evolves.
Thus both dislocation
processes and texture evolution occur together
in the experiments.
This is
reflected in the distribution
of misorientation
data in
Fig. 5, clearly showing a large peak in the distribution
at lower angles and a second population
at high
angles.
The dislocation
mechanisms
[Fig. 3(a)]
contribute most to the peak observed at small angles,
but only 1% of those boundaries
contribute
to the
angle range greater than either 30 or 45 at 70% and
90% c.r., respectively.
The experiments
show a
significantly larger percentage of lo-25% for the high
angles This high angle population is in a similar range
to the broad peak formed by preferred
texture
formation
in a subdivided
grain [Fig. 3(b)]. The
texture
mechanisms
contribute
to only a small
portion of the distribution
at the lowest angle range.
Combining
the two distributions
in the simplified
model of Fig. 3 to obtain a more quantitative
prediction
would require the incorporation
of the
grain subdivision
as described by Fig. 3(a) into a
crystal plasticity model. This incorporation
is easiest
to envision, for example, in a finite element based
plasticity model. As texture evolves within subdivided
grains in the polycrystal
model, a fraction of high
angle boundaries will grow at some of the boundaries
Material
Nickel:
Rolling
Rolling
Rolling
Rolling
Torsmn
Aluminum:
Rolling
(99.5%
Rolling
(99.8%
Tantalum:
Rolling
and deformation
conditions
Average spread
()
1.4)
2.7)
3.5)
4.5)
9.2
10.7
11.9
12.6
16.9
12.9
11.0
texture
13.1
Calculated deformation
textures frequently predict
much higher intensities at the peak intensity points in
the ODF as well as much narrower angular spreads
in intensity
about those peaks than is found in
experiments
[59]. Smoothing
factors used in recent
texture simulations to reduce the sharpness of those
calculated
textures to the approximate
levels observed in experiment, include 7.5 [.59] and 7-8 [60].
However, larger smoothing
factors from 9 to 17,
based on Table 3, may be more representative
for
textures developing at large strains. As an additional
application of the present data, the orientation spread
owing to the microstructure
can be introduced
into
polycrystal
models to predict the formation
of the
high angle boundaries.
5.4. Effect qf material
and deformation
parameters
3885
during recrystallization.
Specific orientation relationships may result in accelerated
growth,
but in
most cases the migration
of high angle grain
boundaries
will be slowed down or stopped
by
orientation
pinning. Orientation
pinning arises by
impingement
between
a growing
grain
and a
deformed
crystallite
of similar
orientation
that
leads to the replacement
of a mobile high angle
boundary
with a much less mobile
low angle
boundary [61].
6. CONCLUSIONS
Deformation
of metals from medium
to large
strain introduces
significant changes in microstructure, texture and local crystallography.
These changes
are illustrated by experimental
examples for different
metals deformed by different processes. The following conclusions
are drawn:
The microstructure
evolves
into a lamellar
structure of dislocation boundaries of small and
medium angles mixed with high angle boundaries. The number of the latter is significantly
larger (e.g. about 3-5 times in rolling, 27 times in
torsion)
than the number
of original
grain
boundaries.
?? The macro-texture
evolves into a typical deformation texture containing
a number of texture
components.
Local crystallographic
measurements show that the volume fraction
of the
individual
texture components
is composed
of
micrometer
and submicrometer
sized volumes
distributed throughout
the deformed microstructure.
?? The distribution
of misorientations
across lamellar boundaries
has been estimated
based on
assumptions
that those boundaries are formed by
evolving grain subdivision processes, which can
lead to different texture components
within an
original grain. The estimated distributions
are in
good agreement
with those determined
experimentally at large strain by automated
Kikuchi
diffraction techniques.
?? The
formation
of high angle grain boundaries
during plastic deformation
leads to a deformation induced reduction in grain size. This grain
refinement
will have an effect both on the
mechanical and on the thermal behavior of the
deformed metal.
??
3886
HUGHES
and HANSEN:
GRAIN
SUBDIVISION
36.
31.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
MECHANISMS