Bonding of Fe and Fe-C, Fe-Cr, and Fe-Ni alloys;

Boride-layer growth kinetics

C. M. Brakman, A. W. J. Gommers, and E. J. Mittemeijer
Laboratory of Metallurgy, Delft University of Technology, Rotterdamseweg 137, 2628 AL Delft,
The Netherlands
(Received 23 February 1989; accepted 22 June 1989)
Specimens of pure Fe and of Fe-0.8 mass % C, Fe-0.5 mass % Cr, Fe-4.0 mass % Cr,
Fe-4.0 mass% Ni, and Fe-10.0 mass% Ni alloys were borided in boriding powder. A
boron-compound layer developed consisting of a surface-adjacent "FeB" sublayer on top of an
"Fe2B" sublayer. Layer-growth kinetics were analyzed by measuring the extent of penetration
of the "FeB" and "Fe 2 B" sublayers as a function of boriding time and temperature in the
range 1025-1275 K. Layer growth is dominated by B diffusion through "FeB/Fe 2 B". This
diffusion process is of strongly anisotropic nature. Consequently, ragged interfaces occur
between the substrate and the boride layers. The depths of the tips of the most deeply
penetrated "FeB" and "Fe 2 B" needles have been taken as measures for diffusion in the easy
[001] diffusion directions. Assuming unidirectional B diffusion and parabolic growth, a
simple model of layer growth has been given. It accounts for the specific volume difference
between "FeB" and "Fe2B". In contrast with earlier work, the model provides values for the
kinetic parameters for growth of each of the phases in the boron-compound layer.

I. INTRODUCTION strate underneath the compound layer can be ignored usu-

Bonding's most prominent technological advantage is
the production of a very hard, wear-resistant surface layer
on metallic substrates.1'2 Industrial boriding is mainly ap-
plied to steel and ferrous alloys. For process control in auto-
mated installations, knowledge of kinetic parameters of the
boriding process is essential. The separate and combined
influences on boriding kinetic parameters of the many alloy-
ing elements in steels are not known.
This paper deals with a kinetic analysis of the boriding
of pure iron and iron-carbon, iron-chromium, and iron-
nickel alloys. In earlier work devoted to the boriding of iron
and carbon steels3'4 no distinction was made between layer-
growth rates of the (Fe,M)B and (Fe,M)2B sublayers. In
the present analysis, use is made of boron-penetration depth
obtained from needle-length measurements of both (Fe, M)B
and (Fe, M)2B phases. A simple model for bilayer growth is
applied providing separate determination of the kinetic pa-
rameters for boron diffusion in each of both phases. The set
of kinetic parameters obtained in this work for the boriding
of iron and the iron-based alloys indicated allows steering of
the constitution of the boron-compound layer.
II. BORIDE-LAYER GROWTH KINETICS
On boriding iron and ferrous alloys a boron-compound
layer can develop on the surface of the specimen. In gen-
eral the compound layer is composed of two sublayers: an
(Fe, M)B sublayer on top of an (Fe, M)2B sublayer. An al-
loying element M originally present in the matrix can be in-
corporated in the borides. In the sequel the borides will be
denoted by FeB and Fe2B. The diffusion zone in the sub-
ally; the solubility of boron in Fe is very small.5
It is assumed that (i) layer growth is rate-controlled
by boron diffusion in the FeB and Fe2B sublayers and
(ii) boride-layer growth occurs as a consequence of B diffu-
sion perpendicular to the surface of the specimen. A linear
B-concentration profile is assumed within each sublayer
(cf. the small homogeneity ranges of about 1 at. % for both
borides reported in Ref. 6). Further assuming that the boron
surface and interface concentrations do not change during
boriding, the kinetic model given in Appendix A can be
adopted. Using c0 = 0 and cb — cc = 0 (Appendix A;
Fig. A-l), it follows from Eqs. (A-6) and (A-7)
C FeB = DFs,B(cs - ca) = 2m(mca + ncc) (la)
C Fe2B = DFe2B(cb - cc) = 2n(am + n)cc (lb)
where m and n are the "parabolic" growth constants for the
FeB and Fe2B sublayers, respectively. The concentrations
cs ~ ca and cb ~ cc and the iron atom volume ratio a
[Eq. (A-l)] can be assessed for boriding of pure Fe from
lattice parameter data7: cs = 100.5 X 103 mol m"3, cb =
59.8 X 103 mol m" 3 , and a = 0.84. [The orthorhombic
unit cell of FeB has parameters a = 0.5506 nm, b =
0.4061 nm, c = 0.2952 nm; the tetragonal unit cell of
Fe2B has parameters a = 0.5109 nm, c = 0.4249 nm.
Both unit cells contain 4 molecules. In fact, all (more re-
cent) other lattice-parameter values reported for FeB in the
literature are taken from Ref. 7 (mostly, but not always,
after conversion of kX into A or nm), but the order of a, b,
and c varies from source to source (Table I). The present
choice of c in any case ensures compatibility with Ref. 3
for indication of the overall diffusion direction and the fiber-

1354 J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989 1989 Materials Research Society
 C M . Brakman, A.W. J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics

TABLE I. FeB lattice-parameter data from the literature.

Ref. Year a (kX or nm) b (kX or nm) c (kX or nm)

7 1929 5.495 kXa 4.053 kXa 2.946 kX"

8 1930 4.053 kXa 5.495 kXa 2.946 kXa
9 1933 4.053 kXa 5.495 kXa 2.946 kXa
10 1950 5.506 x 10 ' nm 2.952 x 10"' nm 4.061 x 10~' nm
3 1967 4.053 kXb 5.495 kX" 2.946 kXb
11 1967 5.495 kX 2.946 kX 4.053 kX
12 1985 5.506 x 10"' nm 2.952 x 10"1 nm 4.061 x 10"' nm
Present paper 1989 5.506 x 10"' nm 4.061 x 10"' nm 2.952 x 10"' nm

"These data have been given in A, but, by international agreement of 1947, have to be taken as given in kX units (1 kX = 0.1002056 nm).15
'These data have been incorrectly denoted as given in A units.

texture axis. The lattice-parameter values for Fe2B from
Ref. 7 agree (after conversion from kX into nm) with those
from more recent work1314 within experimental accuracy.]
It is assumed that cs — ca and cb — cc do not depend
significantly on temperature in the temperature range ap-
plied. Then, the dependencies of CFeB and CFe2B on tem-
perature are governed by the diffusion coefficients Z)FeB and
£>Fe2B. Hence:
CFeB = fcFeB exp(-0 F e B /Rr)
CFe2B = £Fe2B exp(-0 F e 2 B /Rr)
where £FeB and £Fe2B are constants. Consequently, the acti-
vation energies Q and the pre-exponential factors k for B
diffusion in FeB and Fe2B can be derived from the slopes
and intercepts at the ordinates of the straight lines obtained
by plotting In CFeB and In CFe2B as a function of the re-
ciprocal boriding temperature.
Note that the growth rates of both sublayers inter-
depend [cf. Eqs. (la) and (lb)]. Hence, for the determination
of the kinetic parameters relevant to only FeB- or only
Fe2B-sublayer growth, sets of corresponding data for both
sublayers are required. This is exemplified by the proce-
dure given above for activation-energy determination. Con-
versely, calculation of the growth rate of one sublayer from
kinetic parameters is possible only if the latter are known
for both sublayers. The corresponding not-straightforward
procedure is described in Appendix C.
Neither FeB nor Fe2B has cubic crystal symmetry.
Consequently, diffusion in these materials is of anisotropic
nature and grains may exhibit needle-like growth. For both
compounds [001] appears as the easiest diffusion direc-
tion.3 Substitutional B diffusion can occur in both com-
pounds as a consequence of small homogeneity ranges of
about 1 at. %6 in the Fe-B phase diagram or unequilibrium
amounts of vacancies at the boron sublattice.
Both sublayers have been shown to exhibit a (001)
texture.3'16> 17 Further, it has been found that the Fe2B (001)
texture became increasingly sharp for increasing penetra-
tion depth. 31617 A similar FeB behavior is probable but
difficult to prove (thin layer).
The following interpretation is employed. Initially,
the direction of the largest B concentration gradient is per-
pendicular to the specimen's surface where B atoms are
supplied. Then, due to anisotropic diffusion, those crystal
nuclei with orientations allowing the largest B transport
rate perpendicular to the surface grow out preferentially.
Subsequently, the easy diffusion direction keeps favoring
preferential growth of these crystals perpendicular to the
surface, although lateral B concentration gradients can
(2a) occur. Growth selection leads to the (001) fiber texture
(2b) becoming increasingly sharp for increasing penetration
depth. In the present experiments it is expected that the
needles most deeply penetrating into the substrate ex-
hibit orientations close to or with [001] perpendicular to
the surface.
Hence, an unequivocal interpretation in terms of dif-
fusion of boron in FeB or Fe2B is obtained only if the
respective needles penetrating most deeply are selected
for indication of layer thickness. [One also excludes then
needles with the correct [001] orientations but (i) not sec-
tioned through their tips by the metallographic preparation
technique, (ii) nucleated at a later stage in the process.] The
kinetic parameters thus obtained correspond to diffusion in
[001] directions for FeB and Fe2B. This discussion implies
that averaging over needles of various lengths to define an
"average thickness" obstructs interpretation in terms of
fundamental mechanisms.
III. EXPERIMENTAL
A. Specimen preparation and boriding
Specimen codes and chemical compositions of the
materials used are listed in Table II.
Specimen geometry and treatment of the materials
prior to boriding are indicated in Table III. Before borid-
ing, all specimens were metallographically prepared by
hand-grinding on 240, 320, 400, and 600 paper and subse-
quently polishing with 3 /xm and 1 jum diamond paste.
The boriding times and temperatures are given in
Table IV. Boriding was performed in a closed, square-
shaped case (material: DIN 1.4841) using a fresh batch of

J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989 1355

 C M . Brakman, A. W.J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics

TABLE II. Composition (mass %) of alloys used.

Fe C Cr Mn Si P s Ni

Fe balance 0.005 0.0013 0.004

Fe-0.8C balance 0.81 0.00 0.00 0.30 0.000 0.005 0.00
Fe-0.5Cr balance 0.013 0.53 0.014 0.027 0.001 0.007 0.057
Fe-4Cr balance 0.014 3.81 0.024 0.029 0.002 0.003 0.011
Fe-4Ni balance 0.010 0.00 0.00 0.040 0.000 3.99
Fe-lONi balance 0.007 0.00 0.00 0.059 0.000 9.99

TABLE III. Specimen treatment before bonding.

Condition of raw material Heat treatment Specimen shape

Fe Chill-cast rod, 25 mm diam None Disks, 2.3-2.5 mm thickness, sectioned

into 4 equal parts
Fe-0.8C Rolled sheet, 2 .3-2.5 mm thickness 30 min 850 °C + cooling in quiet air Square samples 10 mm x 10 mm
Fe-0.5Cr Rolled sheet, 2 .3-2.5 mm thickness 28 h homogenizing at 1100 °C in quartz Square samples 10 mm x 10 mm
tube + cooling in quiet air
Fe-4Cr Rolled sheet, 2 .3-2.5 mm thickness 28 h homogenizing at 1100 °C in quartz Square samples 10 mm x 10 mm
tube + cooling in quiet air
Fe-4Ni Chill-cast rod, 15 mm diam, turned to None Disks, 2.3-2.5 mm thickness, sectioned
10 mm diam into 2 equal parts
Fe-lONi Chill-cast rod, 15 mm diam, turned to None Disks, 2.3—2.5 mm thickness, sectioned
10 mm diam into 2 equal parts

TABLE IV. Boriding times and temperatures.

Boriding temperatures (K)

1023 1073 1123 1173 1273

Alloy Boriding times (h)

Fe 4,6,8 1,41/4,81/2 11/4,21/3,4 1,4,81/2 i/4,l,2

Fe-0.8C 1,4,8 1,4%, 8 Vi, l!/3,2
Fe-0.5Cr 4,8 1/2,1,4 1/2, 1,4
Fe-4Cr 1,4,8 1,4,8 l,5,9!/2 l,2!/2,4
Fe-4Ni Vi, 1,4 i/2,l,4 1/2, 1,4
Fe-lONi 1,4,8 1,4,8 4,8 1/4,1/2
(1223 K)

700 grams of Durborid 2 powder (tradename of Degussa
AG, GB Durferrit und Industrieofenbau, Postfach 1345,
D-6450 Hanau 1, FRG) for each experiment. The case's
dimensions were such that the powder filled it up to the
lid. The specimen was located in the approximate center of
the case. Boriding followed after placing the case in a
normal chamber furnace without use of inert gases. Prior
to experiments, ample time was allotted for establishment
of thermal equilibrium in the furnace. As a result of pre-
liminary experiments it was estimated that boriding started
half an hour after transferring the specimen to the furnace;
after that, the so-called incubation period (see Sec. IV.A)
sets in.
B. Specimen treatment after boriding
It was found that a cleaning treatment, before mount-
ing of the specimen, removed part or nearly all of the FeB
sublayer. This is probably caused by the extreme brittle-
ness of the FeB sublayer. Thus the electrolytical deposition
of a supporting Ni surface layer requiring a clean surface
before deposition had to be abandoned. Therefore, sub-
sequent to cooling of the specimen after boriding, the
specimen was carefully removed from the case and resin-
embedded without any further treatment.
Using 400 and 600 grinding papers, a 1-mm layer of
material was removed [Fig. l(a)] in order to eliminate

1356 J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989

 C M . Brakman, A.W. J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics
face used foi
penetration depth
measurements
(a) (b)
FIG. 1. Metallographic preparation before (a) and after (b) boriding.
Arrow indicates grinding and polishing direction after boriding.
"corner effects" (i.e., layer growth by inward diffusion of
boron from more than one side). Care was taken that the
abrasive action of both grinding and polishing always
occurred in only one and the same direction [Fig. l(b)].
After polishing using 3 jum and 1 /xm diamond paste,
etching took place in nital (2% nitric acid in ethanol). With
the exception of very short boriding times and/or very low
boriding temperatures (not considered in this paper), an
FeB layer was observed for all materials displayed in
Table II. It was also found that the FeB layer thickness
was larger on the "A-side" of the specimen [Fig. l(b)] as
compared to the "B-side," indicating the extreme brittle-
ness of the FeB phase. (This distinction between A and B orig-
inates from the grinding/polishing direction employed.)
The conclusion seems to be justified that one always
tends to wnJer-estimate FeB-layer thickness. An apparent
absence of FeB top layers after boriding may be due to in-
appropriate specimen handling.
C. Penetration-depth measurements
Penetration depth was measured using a Leitz Durimet
device. For both FeB and Fe2B sublayers the penetration
depth was determined by selecting those needles reaching
most deeply into the substrate (see discussion in Sec. II).
Values for the penetration depths u and v (see Fig. A-l)
were obtained as averages over 10 measurements (needle
tip-surface distances) taken along a distance of 5 mm par-
allel to the surface of the specimen.
IV. RESULTS AND DISCUSSION
A. Parabolic growth constants
Using measured data for the thickness parameters u
and v [see Fig. A-l and Eq. (A-5)], plotting of u2 and v2
versus treatment time t yielded straight lines in general
(but see below), indicating that parabolic layer growth
occurred.
Usually these straight lines did not pass through the
origin. The intercept of the time axis can result from inac-
curate estimation of the time needed to heat up the speci-
men (see Sec. III. A) as well as from an (actual) incubation
time for boride development. No indication was obtained
for significant difference between intercepts of the time
axis for u\t) and v2(t) (apart from those differences due to
experimental problems described below).
J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989
As discussed in Sec. III.B, part of the FeB sublayer
can scale off easily during cooling after boriding and/or
subsequent specimen preparation for examination, leading
to too small values for u and v at variable times. Further,
the boriding potential of the atmosphere may decrease as a
result of exhaustion of the boriding powder, leading to too
small values for u and v at long times and at high tempera-
tures (Table V). Because of the systematic nature of both
errors, the following procedure was established:
(i) In the v2 vs t plot two points are selected yielding
the steepest straight line, but with an intercept of the ab-
scissa (taken as the effective incubation time) smaller than
the shortest boriding time employed.
^ii) The incubation time yields an additional point for
the u2 vs t graph. Next, the_(FeB) straight line is drawn
through this point and the u2 data point leading to the
steepest straight line.
(iii) After determination of the slopes of the straight
lines for both v2 and u2 vs t, their square roots yield
2(n + m) and 2m, respectively, according to Eq. (A-5).
Graphs of u2 and v2 vs t for boriding at 1123 K for all
alloys investigated are given in Figs. 2 to 7. The m and
n values for the various materials and boriding tempera-
tures are given in Tables_VI-a and Vl-b.
All values used for u2 and v2 are averages of 10 mea-
surements. Errors for the slopes were assessed from maxi-
mal and minimal estimates for u2 and v2 according to
{(w2)1'2 ± <ruf and {(v2)"2 ± crv}2 where au and o\, denote
the standard deviations of u and v, respectively.
B. Activation energy and pre-exponential
constant for layer growth
Activation-energy values were determined on the basis
of Eqs. (1) and (2) using the parabolic sublayer-growth
constants m and n gathered in Tables Vl-a and Vl-b. For
each material, plots of In CFeB and In CFe2B vs reciprocal
boriding temperature were made (see Figs. 8 to 13). Errors
in CFeB and CFe2B [cf. Eqs. (la) and (lb)] as a consequence
of the growth-constant errors (see above) are indicated in
Figs. 8 to 13. Straight lines were fitted to these data points
employing a least-squares procedure. Activation-energy
values were derived from the slopes of the straight lines
TABLE V. Examples of affected layer-thickness determinations due to
scaling off and exhaustion of boriding powder (see data marked*).
Material and boriding Boriding FeB layer FeB + Fe2B layer
temperature time (h) thickness (fim) thickness (pun)
Fe-4Cr 1 27.3 100.7
(1273 K) 2!/2* nil 137.5
4* nil 185.0
Fe 1/4 16.2 71.3
(1273 K) 1 25.0 133.5
2* 24.1 189.6
1357
 C M . Brakman, A.W.J. Gommers, and E. J. Mittemeijer: Boride-layer growth kinetics

15000
Fe
Tbor1123K
r>- * v2(FeB+Fe2B)
• u2(FeB)

10000

f. 5000

A i' r " : 't \

(a) (b)
FIG. 2. (a) Cross section of borided Fe specimen (4 h, 1123 K). (b) Averaged squares of layer thicknesses u and v (cf. Fig. A-l) vs boriding time t at
1123 K for Fe specimen.

15000

- 3000
" 10000 -
'- 2000

5000 -
- 1000

FIG. 3. (a) Cross section of borided Fe-0.8C specimen (4 h, 1123 K). (b) Averaged squares of layer thicknesses u and v (cf. Fig. A-l) vs boriding
time t at 1123 K for Fe-0.8C specimen.

1358 J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989

 C M . Brakman, A.W.J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics

Fe-0.5Cr
T bor 1123K
10000

5000

(b)
FIG. 4. (a) Cross section of borided Fe-0.5Cr specimen (4 h, 1123 K). (b) Averaged squares of layer thicknesses u and v (cf. Fig. A-l) vs boriding
time t at 1123 K for Fe-0.5Cr specimen.

10000
Fe-4 Cr
Tbo_r1123K
A V 2(FeB+Fe 2 B)

5000

(a) (b)
FIG. 5. (a) Cross section of borided Fe-4Cr specimen (4 h, 1123 K). (b) Averaged squares of layer thicknesses u and v (cf. Fig. A-l) vs boriding time t
at 1123 K for Fe-4Cr specimen.

J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989 1359

 C M . Brakman, A. W.J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics

20000
Fe-IONi

rat Tbor1123K
* v2(FeB+Fe 2 B)
15000 3000

10000 2000

5000 1000

(a) (b)
FIG. 6. (a) Cross section of borided Fe-4Ni specimen (4 h, 1123 K). (b) Averaged squares of layer thicknesses u and v (cf. Fig. A-l) vs bonding time t
at 1123 K for Fe-4Ni specimen.

10000

1500

5000 1000

500

(a) (b)
(J)
FIG. 7. (a) Cross section of borided Fe-IONi specimen (4 h, 1123 K). (b) Averaged squares of layer thicknesses u and v (cf. Fig. A-l) vs bonding
time t at 1123 K for Fe-IONi specimen.

1360 J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989

 C M . Brakman, A. W.J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics

TABLE Vl-a. Parabolic FeB-layer growth constants m (growth in [001] direction).

Bonding temperatures (K)

1023 1073 1123 1173 1273

FeB layer growth constants m (10 6 ms "2)

n ns7 0066 ft 1 47 0,l66

Fe 0 029 U - 1 4 / 0.l27
U.U3/ 0.048
(0.034) (0.089) (0.136) (0 .288)
(0.056)
0 jynO-525
Fe-0.8C
(0.057) (0.095) (0.150) (0.229) (0.481)
Fe-0.5Cr 0.064°;™ 10124
J O 103
(0.041) (0.069) (0.110) (0.169) (0 .362)
A ftc70.066
Fe-4Cr U.U3/0047 0.260°;™
(0.023) (0.126) (0.260) (0 .899)
Fe-4Ni (0.057) n til0158 0 289° 31°
U.141O 125
(0.017) (0.304) (1 .238)
(0.132)
Fe-lONi (0.050)
O83J
(1223 K)
(0.014) (0.039) (0.098) (0.216) (0 .437)

Super- and subscripts indicate error range as obtained from the experiments. Values between parentheses: m back-calculated from kinetic data /t and Q
(Tables VII and VIII) according to Appendix C.

TABLE Vl-b. Parabolic Fe2B-layer growth constants n (growth in [001] direction).

Boriding temperatures (K)

1023 1073 1123 1173 1273

Fe2B layer growth constants n (10 ms "2)

Fe O.154o O.315o'ilo ".O3.J0.827

(0.144) (0.220) (0.323) (0.460) (0.853)
Fe-0.8C 0.206S 0.538Q 537

(0.105) (0.158) (0.228) (0.319) (0.576)

Fe-0.5Cr O 0 235 0.488g;^
(0.154) (0.233) (0.340) (0.480) (0.879)
Fe-4Cr (1 2 4 2 0 2 4 9
U.Z4Z 02 37
(0.057) (0.096) (0.153) (0.462)
(0.231)
ft 171 0.377
Fe-4Ni 0.186 0185 0.306g;^ U.J/IQ.364
(0.124) (0.193) (0.282) (0.388) (0.628)
1*L10.270 II ,i«->0.488
Fe-lONi 0.199":2™ 0 .262 0 2 5 6 O.4s2 0 4 7 7
(1223 K)
(0.130) (0.193) (0.270) (0.357) (0.457)

Super- and subscripts indicate error range as obtained from the experiments. Values between parentheses: n back-calculated from kinetic data k and Q
(Tables VII and VIII) according to Appendix C.

(Table VII). Extrapolation toward \/T = 0 yielded the
temperature independent pre-exponential constants k
according to Eqs. (2a) and (2b). Results are shown in
Table VIII. The kinetic data are interpreted here as repre-
senting B diffusion along [001] directions in FeB and
Fe2B, respectively (see discussion in Sec. II).
1. Iron
Substitutional boron diffusion perpendicular to (001)
appears possible in Fe2B [there is a chain of B atoms along
[001]; see Figs. 14(d, e)]; in the case of FeB a zigzag path
(corresponding to a zigzag chain of B atoms; see Fig. 3 of
Ref. 9) seems likely for net transport along [001].
Using crystallographic data (see aside in Sec. II), the
shortest distance between an Fe atom and a B atom experi-
enced by a B atom during diffusion in Fe2B along its path
perpendicular to (001) equals about 0.191 nm [at that mo-
ment the B atom is surrounded by 4 Fe atoms at this dis-
tance; see Fig. 14(e)]. For the zigzag path diffusion of B in
FeB each atomic jump (corresponding to a "zig" or a "zag")

J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989 1361

 C M . Brakman, A.W. J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics

~<n -16 Fe
*
E
o • FeB
E • Fe2B
J-17

-18
\

\
-19
N\
\
\
-20

\
-21
\ -21 -

-22
7.8 8.0 8.5 9.0 9.5 10.0
1 FIG. 10. In CFeB and In CFe2B vs \/T for Fe-0.5Cr specimen [cf. Eqs. (1)
T
(1CHK-1)
and (2)].
FIG. 8. In CFl,B and In C¥t2B vs \/T for pure Fe specimen [cf. Eqs. (1)
and (2)]. is along a direction such as [hid] with h = (1 - 4u)/a2, k -
l/b2mdl = (-16 uv + 4M + 4v - 2)1 c2 where u = 0.125
and v = 0.18. [Symbols u and v used here should not be
confused with those used for the sublayer thicknesses in
the rest of this paper. Note that a and b crystal axes have
been interchanged in this paper as compared to Ref. 9 (see
aside in Sec. II and Table I). Hence u from Ref. 9 is asso-
ciated with b in this paper and v with a]. This atomic jump
occurs perpendicular (to within about 2°) to a plane such as
(hkl) with/i = 2/(4v - 1), k = h/(l - 4u)andk + I = 2.
The shortest distance between an Fe atom and a B atom
during diffusion in FeB occurs in the jump indicated and
equals about 0.197 nm [at that moment the B atom is
surrounded by 4 Fe atoms in the plane described; see
Fig. 14(c)].
The unconstrained shortest distance between an Fe
and a B atom is about 0.215 nm.10'19 Then, consideration
of the boron misfits in the jumps indicated above would
suggest about the same value for the activation energies of
boron diffusion in FeB and Fe2B.
The small difference between the values for the ac-
tivation energy of boron diffusion in FeB and Fe2B (cf.
-21 -
Table VII) may be explained as a consequence of a dif-
ference in bonding of the B atom. Because the shortest
Fe-B distance is somewhat smaller in FeB than in Fe2B
(1CHK-1) (0.215 nm vs 0.218 nm10), the covalent nature and thus
FIG. 9. In CFeB and In CFe2B vs \/T for Fe-0.8C specimen [cf. Eqs. (1) the strength of the Fe-B bond may be largest for FeB
and (2)]. (cf. Ref. 20). Further, the shortest B-B distance in FeB

1362 J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989

 C M . Brakman, A.W. J. Gommers, and E. J. Mittemeijer: Boride-layer growth kinetics

Fe-£Ni

• FeB
• Fe2B
u
c

-16 -

-17 -

-18 -

\
-19 - \
IK

-20 --
-21 -

FIG. 11. In CFeB and In CFe2B vs \/T for Fe-4Cr specimen [cf. Eqs. (1)
and (2)].
is about twice the covalent radius of the B atom whereas
the corresponding value in Fe2B is significantly larger
(0.180 nm vs 0.212 nm10). It may thus be suggested that
disruption of the bond of the B atom with a nearest B
neighbor is more difficult in FeB than in Fe2B.
Activation-energy values were reported earlier for this
case of boriding Fe; data are shown in Table IX. However,
a comparison with the present results is difficult, because
(i) the FeB and Fe2B sublayers were treated as one in
Refs. 3 and 4 and the values reported for the activation
energy reflect some average; (ii) it appears likely that the
values for the activation energies reported for FeB and
Fe2B in Ref. 21 were derived just from the temperature
dependence of m alone and n alone, respectively, there-
by ignoring the interdependence of m and n for bilayer
growth as incorporated in Eqs. (la) and (lb); and (iii) spe-
cial precautions, to avoid artifacts due to scaling off, and
thickness measurements on the basis of the needles most
deeply penetrated, were not made for the analyses reported
in the literature.
The diffusion coefficient is proportional to the square
of the length of the atomic jump and the jump frequency
(the latter factor contains the exponential temperature de-
pendence as characterized by the activation energy). One
atomic jump of B in Fe2B occurs along [001] and covers a
1 t 1 \
-21
78 8.0 8.5 9.0 95 10.0
FIG. 12. In CFeB and In CFe2B vs 1/71 for Fe-4Ni specimen [cf. Eqs. (1)
and (2)].
distance of c/2 = 0.212 nm, whereas one atomic jump of
B in FeB along the zigzag path (not along [001]) covers a
projected distance along [001] of c/2 = 0.148 nm. There-
fore, considering only jumps along the chain (either zigzag
or straight) of boron atoms, the ratio of the pre-exponential
factors of the boron-diffusion coefficients valid for net
diffusion along [001] in FeB and Fe2B is estimated by
(0.148)2/(0.212)2 = 0.5. No experimental data are avail-
able to verify this prediction. Using the experimental value
for kFeB/kFe.lB (= 2.25; cf. Table X) and the above value
for the ratio of pre-exponential factors of the diffusion
coefficients, it follows that the boron-concentration differ-
ence across the FeB sublayer is about 4.5 times as large as
for the Fe2B sublayer [see Eqs. (1) and (2) and Fig. A-l].
2. Iron-chromium and iron-nickel
Considerable amounts of Cr and Ni can be dissolved
in FeB and Fe2B.22
The atomic radii of Cr and Ni are about the same and
they are larger than that of Fe. Adopting the geometrical
model for B diffusion, discussed above for boriding Fe,
it would be expected that Cr and Ni dissolved on the Fe
sublattice of the borides will generally lead to activation

J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989 1363

 C M . Brakman, A.W.J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics

needs not be governed only by diffusion of B, but, for ex-

ample, the necessary transport of M and Fe by diffusion on
the "Fe sublattice" may (also) be rate determining. Indeed,
especially the activation energy for (Fe,M)B-sublayer
growth is affected on alloying Fe with Cr or Ni. Further,
diffusion of Cr on an (a or y) Fe (sub)lattice is more rapid
than that of Ni.24 This might be related to the observation
of an activation energy for (Fe, M)B-sublayer growth larger
for FeNi than for FeCr (cf. data for about the same amounts
of alloying elements in Table VII: Fe-4Cr vs Fe-4Ni).

-20 -
-21 -
7.8 8.0
FIG. 13. In CFeB and In CFe2B vs 1/r for pure Fe-lONi specimen [cf.
Eqs. (1) and (2)].
energies for B diffusion larger than those for Fe (cf. data
for Fe, FeCr, and FeNi in Table VII). On the other hand,
composition-depth profile data for FeCr (8 mass % Cr) and
FeNi (8 mass % Ni) borided at 1173 K indicate that on borid-
ing a redistribution of the alloying element, M(=Cr, Ni),
occurs: as compared to the bulk compositions the (Fe, M)B
sublayers appear to be impoverished in Ni and enriched in
Cr, respectively, while the (Fe,M) 2 B sublayers exhibit
about the bulk content of M. 22 ' 23 Then the apparent ac-
tivation energy for, in particular, (Fe, M)B-sublayer growth
3. Iron-carbon
Carbon does not dissolve significantly in FeB and
Fe 2 B. 3 On bonding carbon is driven ahead of the boride
layer and, together with boron, it forms borocementite,
Fe3(B, C) as a separate layer between Fe2B and the matrix
[Fig. 3(a)]. [From lattice-parameter measurements it was
found that the borocementite in the present case contained
about 4 mass % B corresponding to Fe3(B067C0 33).17] Thus,
part of the boron supplied is used for the formation of
borocementite.
Consequently, a bilayer-growth model for bonding
Fe-C alloys does not hold in principle. However, a descrip-
tion based on an effective bilayer-growth model may be used
yet, if a number of simplifying assumptions are made:
(i) the kinetics for the three-layer system [FeB +
Fe2B + Fe3(B, C)] are considered to be controlled only by
B diffusion in FeB and Fe2B, and
(ii) the fraction of all B atoms arriving at the Fe 2 B/
Fe3(B, C) interface and used for Fe3(B, C) formation does
not depend on time and temperature.
The model pertaining to the above conditions has been de-
scribed and discussed in Appendix B where it has also been
shown for QFe2B and made likely for gFeB that they can be
obtained using the formalism of Eqs. (la) and (lb).
Minor changes of the Q values are observed as com-
pared to the case of boriding Fe (Table VII). This supports
applicability of the above model. Earlier work on boriding
carbon steels also hinted at small effects of carbon on
boriding kinetics.4

TABLE VII. Activation-energy values obtained using bilayer growth model (Appendix A).

Correlation coefficient
least-squares fit
GFB (X R
(kj mor 1 ) (U mol"1) FeB Fe2B

Fe 175 157 1.11 0.995 0.996

Fe-0.8C 176 154 1.14 0.984 0.981
Fe-0.5Cr 177 155 1.14 0.999 0.997
Fe-4Cr 289 210 1.38 1.000 0.997
Fe-4Ni 311 178 1.75 0.995 0.986
Fe-lONi 286 157 1.82 0.983 0.999

1364 J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989

 C M . Brakman, A. W.J. Gommers, and E. J. Mittemeijer: Boride-layer growth kinetics

Jo
• Feinf
o Fe i n ^
• Binf
• B in^c

(a)

O Fe atoms
• B atoms

0.293

(hkl);see
caption
(b) (c)

FIG. 14. (a) Projection of Fe and B atomic-center points onto (001) plane for FeB (according to Fig. 1 of Ref. 9). (b) Projection of Fe atoms onto the
plane perpendicular to which an atomic jump of B takes place in FeB (for hkl indices see text in Sec. IV.B). The center points of the hatched atoms 1—4
are in the plane of projection; the center points of the cross-hatched atoms 5 and 6 are above this plane [the numbering 1-6 corresponds with the one of
(a)]. The atomic radius of Fe has been taken equal to half the shortest Fe-Fe distance in the structure: 0.132 nm (cf. atom pairs 1-3 and 2-4). B atoms
have not been drawn, but projections of B atoms coincide with the crosses indicated, (c) Fe atom configuration around the B atom for FeB at the moment
of closest approach to the surrounding Fe atoms during an atomic jump of B; measures in nm. The B atomic jump occurs practically perpendicularly to
the plane through the Fe atoms 1-4 [cf. (b) and Sec. IV.B]. The shown crystallographic section of the iron sublattice corresponds with the front face of
the trigonal prism shown in Fig. 2 of Ref. 9. (d) Projection of Fe atoms onto the (001) plane perpendicular to which an atomic jump of B takes place in
Fe2B. The center points of the hatched atoms are in a plane at a distance of 3 c/4 above the plane of projection; the center points of the darkened atoms
are in a plane at a distance c/4 above the plane of projection. The atomic radius of Fe has been taken equal to half the shortest Fe-Fe distance in the
structure: 0.120 nm (cf. atom pair 4-5). B atoms have not been drawn, but projections of B atoms coincide with the crosses indicated, (e) Fe atom con-
figuration around the B atom for Fe2B at the moment of closest approach to the surrounding Fe atoms during an atomic jump of B perpendicular to
(001); measures in nm [see also Fig. l(b) in Ref. 18].

J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989 1365

 C M . Brakman, A.W. J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics

0.269

(001)

(d) (e)
FIG. 14 (continued)

C. Effect of kinetic parameters on the relative Values for kFtB/kFeiB, (QFeB - QFe2B), and m/n for
thicknesses of the (Fe, M)B and (Fe, M)2B 1173 K and 1273 K are presented in Table X. Very large
sublayers values of kFeB/kF<.lB occur for Fe-4Cr, Fe-4Ni, and Fe-
The ratio m/n is equal to the ratio of the thicknesses lONi, as compared to Fe, Fe-0.8C, and Fe-0.5Cr. This
of the (Fe,M)B and (Fe,M)2B sublayers [cf. Eq. (A-5)]. promotes large values for m/n. However, the correspond-
Using Eqs. (1) and (2) it follows ing (gFeB - 2Fe2B) values are also large, thereby counter-
acting the &FeB/^Fe2B effect to a considerable extent, but not
entirely (cf. Table X).
For practical applications a monophase (Fe,M) 2 B
CFe B
2 my1 surface layer is favored amongst others to reduce the
— +a brittleness.1 An (Fe,M)B top layer can be avoided if the
nl
bonding potential is sufficiently low. This may not al-
exp{-(2 FeB - G Fe2B )/Rr} (3) ways be possible, for example if a bonding powder as in
F C Fe 2 B this work is employed. Then, if the model used in this
Realizing that m and n are positive, it is immediately seen paper prevails, the thickness of the (Fe,M)B sublayer
that m/n increases with increasing CFeB/CFe2B. can be (made) small for [see Eq. (3)] a small kFtB/kFt2B
ratio, a large <2FeB - gFe2B difference, and a low bonding
TABLE VIII. Logarithms of the temperature-independent constants temperature.
*FeB = DfB(cs - ca) and kFc2B = />oe2B(c» - cc) where the Do denote the
pre-exponential factors of the corresponding diffusion coefficients
[cf. Eqs. (A-8) and (A-9)]. V. CONCLUSIONS
In k¥cB (mol m ' s ') In £Fe2B (mol m ' s ') (1) A simple model for diffusion-controlled bilayer-
growth kinetics has been given taking into account specific
Fe - 0.359 -1.17 volume differences. If the interface concentrations are
Fe-0.8C + 0.284 -1.96 unknown but approximately independent of temperature,
Fe-0.5Cr + 0.167 -1.25
Fe-4Cr + 3.33
as in the present case of boriding, the kinetic parameters
+ 11.9
Fe-4Ni + 14.7 +0.925 for each of both phases can be calculated yet from the
Fe-lONi + 11.5 -1.48 temperature dependencies of both parabolic layer-growth
constants.

1366 J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989

 C M . Brakman, A.W. J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics

TABLE IX. Activation-energy values for boriding of iron as reported in the literature.

Ref. Activation energy reported (kJ mol ') Details

[3] 176; both layers treated as one amorphous B powder boriding; B concentration measurements; armco specimens
[4] 132; both layers treated as one; very little FeB Ekabor 2 powder boriding; layer thickness measurements; armco specimens
[21] FeB: 82.5, Fe 2 B: 105.5 (eFeB/eFe2B = 0.78) gas boriding; separate layer thickness measurements; armco specimens

(2) Because of the anisotropic nature of boron diffu- found are close to those determined for the case of borid-
sion, the (Fe,M)B and (Fe, M)2B sublayers produced on ing iron.
boriding iron-based alloys exhibit pronounced (001) fiber
texture and needle-type morphology. Analysis of well-
ACKNOWLEDGMENTS
defined growth kinetics should be performed on the basis
of penetration-depth measurements of those needles ex- We are indebted to Ir. J. J. Smit for enthusiastic co-
tending most deeply into the substrate. Activation ener- operation in an early stage of the project. We express our
gies obtained are related to boron diffusion along [001] in gratitude toward Ir. M.A. J. Somers and Professor B.M.
both compounds. Korevaar for stimulating discussions and toward Mr. P. F.
(3) Activation energies and pre-exponential constants Colijn for skillful metallographic analysis. We are grateful
for growth of (Fe,M)B and (Fe, M)2B sublayers have to Dr. R . B . Helmholdt (Netherlands Energy Research
been determined from parabolic growth constants. In all Foundation, ECN, Petten, The Netherlands) for his pa-
cases the activation energies are larger for (Fe, M)B than tience and help in preparing computer drawings of the FeB
for(Fe,M) 2 B. and Fe2B crystal-structure projections.
(4) On boriding of pure iron, values for the activation
energy of (net) boron diffusion along [001] in FeB and REFERENCES
Fe2B are about the same. As compared to Fe2B, the some- 'A. Graf von Matuschka, Boronising (Carl Hanser Verlag, Munich,
what higher FeB activation energy is ascribed to stronger FRG, 1980).
2P. A. Dearnley and T. Bell, Surf. Eng. 1, 203-217 (1985).
boron-boron atom bonding in the chain along which substi-
3 H. Kunst and O. Schaaber, Harterei-Tech. Mitt. 22, 1-25 (1967).
tutional boron diffusion takes place.
4M-J. Lu, Harterei-Tech. Mitt. 38, 156-169 (1983).
(5) Activation energies increase on alloying with 5 O. Kubaschewski, Iron-Binary Phase Diagrams (Springer Verlag,
chromium and nickel — in particular for (Fe, M)B-sublayer Berlin, FRG, 1982).
growth. ^Binary Phase Diagrams, edited by T. B. Massalski et al. (ASM, Metals
(6) Boriding of iron-carbon (0.8 mass % C) leads to a Park, OH, 1986), Vol. 1, p. 356.
7T. Bjurstrom and H. Arnfelt, Z. Phys. Chem. (B) 4, 469-474 (1929).
surface layer composed of three compounds: FeB, Fe2B, 8 B. Hendricks and P. R. Kosting, Z. Kristallog. 74, 511-545 (1930).
and Fe3(BO67CO33). If severe constraints are obeyed, boron 9T. Bjurstrom, Ark. Kemi, Mineralogi och Geologi 11A, no. 5, 1-12
diffusion in FeB and Fe2B sublayers may still be analyzed (1933).
using the bilayer-growth model. The activation energies IOR. Kiessling, Acta Chem. Scand. 4, 209-227 (1950).
"W. B. Pearson, Handbook of Lattice Spacings and Structures of Metals
(Pergamon Press, Oxford, 1967), Vol. 1, p. 899.
12W. B. Pearson, Handbook of Lattice Spacings and Structures of Metals
TABLE X. Illustration of the effect of the kinetic parameters k and Q on (Pergamon Press, Oxford, 1985), Vol. 2, p. 1297.
the ratio m/n of the thicknesses of the boride sublayers [see Eq. (3)]. 13E. E. Havinga, H. Damsma, and P. Hokkeling, J. Less-Common Metals
27, 169-193 (1972).
m/n 14 B. Aronsson, T. Lundstrom, and I. Engstrom, Proc. Int. Symp. on
Anisotropy in Single-Crystal Refractory Compounds, edited by F. W.
Vahldiek and S.A. Mersol, 1967 (Plenum Press, New York, 1968),
(kJ mol"1) 1173 K 1273 K Vol. 1, pp. 3-22.
^International Tables for X-ray Crystallography (The Kynoch Press,
Fe 2.25 18 0.3 0.3 Birmingham, 1962), Vol. 3, p. 42.
Fe-0.8C 9.43 22 0.7 0.8 16M. Chochatbwski, K. Przybylbwicz, J. Szymariski, and J. Zelechowski,
Fe-0.5Cr 4.12 22 0.4 0.4 Metalurgia i Odlewnictwo 9, 411-421 (1983).
Fe-4Cr 5270 79 1.1 2.0 17J. J. Smit, M.Sc. Thesis, Delft University of Technology, Laboratory of
Fe-4Ni 960000 133 0.8 2.0 Metallurgy, 1984.
Fe-lONi 434000 129 0.6 1.0 18G. Hagg, Z. Phys. Chem. (B) 11, 152-162 (1930).
(1223 K) 19L. Pauling, The Nature of the Chemical Bond, 3rd ed. (Cornell Univer-
sity Press, New York, I960), p. 435.
Values for m/n have been calculated from the data for k and Q given 20W. B. Pearson, The Crystal Chemistry and Physics of Metals and Alloys
in Tables VII and VIII. Note that for boriding at 1273 K of Fe-4Cr and (Wiley, New York, 1972), p. 135.
Fe-4Ni an (Fe,M)B layer thickness is predicted twice as large as for 21 H. Planitz, G. Treffer, H. Konig, and G. Marx, Neue Hiitte 27,
(Fe,M)2B. 228-230 (1982).

J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989 1367

 C M . Brakman, A. W. J. Gommers, and E.J. Mittemeijer: Boride-layer growth kinetics
22C. Badini, C. Gianoglio, and G. Pradelli, J. Mater. Sci. 21, 1721-1729
(1986).
23 Z-S. Jiang, L-X. Zhang, L-G. Li, X-R. Pei, and T-F. Li, J. Heat
Treating 2, 337-343 (1982).
24 J. Fridberg, L-E. Torndahl, and M. Hillert, Jernkont. Ann. 153,
264-276 (1969).
25W. H. Press, B.P. Flannery, S.A. Teukolsky, and W. T. Vetterling,
Numerical Recipes (Cambridge University Press, Cambridge, 1986),
pp. 269-272.
APPENDIX A: BILAYER GROWTH EQUATIONS
The following case is analyzed. A substrate of solvent
element A has been saturated with solute element B
(Fig. A-l). The chemical potential of B in the surrounding
atmosphere is such that two layers, I and II, form on the
substrate. Each layer corresponds to a certain compound
of A and B. Layer/compound I is more rich in B than
layer/compound II. It is assumed that:
(i) layer-growth rates are controlled by solid-state
diffusion;
(ii) inward diffusion of B is unidirectional (perpen-
dicular to the surface);
(iii) differences in specific volume per solvent atom
for the phases concerned (I, II, and the substrate) are ac-
commodated fully in the diffusion direction;
(iv) linear B -concentration profiles occur in both
layers. [In principle the treatment should be given in terms
of activities. In view of the minute composition variations
which are possible within the sublayers (cf. Sec. II), it is
anticipated that the activity coefficient is practically con-
stant within a sublayer. Then the treatment presented is not
affected in principle.]
(v) (inter)diffusion coefficients, Dl and Du for layers I
and II, respectively, are constant. The same holds for the
interface concentrations cs, ca, cb, cc, and c0.
A. Mass balance for growth of layer I
An increase du of the thickness of layer I in time dt
occurs under simultaneous "consumption" of a layer of
thickness adu of layer II. The ratio of the specific volumes
per solvent atom, Vn and Vl for compounds II and I, respec-
tively, is given by a = Vn/V{. For instance, if compounds I
and II conform to AB and A 2 B, respectively, it follows:
a = VmA2j2V"AB (A-l)
where Vm denotes molar volume (in fact, a is equivalent to
a so-called Pilling-Bedworth ratio).
The B mass-balance for the growth of layer I thus reads:
du
(ca - <xcb)— = Di(cs ~ ca)/u
- Dn(cb - cc)/{v - u) (A-2)
where both u and v are distances to the (outer) surface.
B. Mass-balance for growth of layer II
Even if no reaction would occur at the layer II/substrate
interface, a shift of the interface would yet occur as a result
1368 J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989
of the reaction taking place at the I/II interface (if a ¥" 1).
As follows from the above discussion, this interface shift
amounts to (1 — a)du. Therefore, in the analysis of intrin-
sic growth of layer II, a net shift dvnel has to be considered:
dvmt = dv - (1 - a)du (A-3)
Then, proceeding analogously as for layer I, the B-
mass balance for growth of layer II reads:
(A-4)
where /3 denotes the ratio of the specific volumes per sol-
vent atom of substrate and compound II, respectively.
Note that the B amounts necessary to maintain the con-
centration gradients in the respective layers have been ne-
glected in the left-hand members of Eqs. (A-2) and (A-4).
The following general solutions satisfy Eqs. (A-2) and
(A-4):
u = 2mt112 and v - u = 2nV (A-5)
After rearranging it follows from Eqs. (A-2), (A-4),
and (A-5):
C, = Dics - ca) = 2m{mca + ncc - am(cb - cc)
- (3co(am + «)} (A-6)
and:
Cn = Dn(cb - cc) = 2n(am + n) (cc - f3c0) (A-7)
If the interface concentrations are not known, the diffu-
sion coefficients cannot be determined from measurements
of the "parabolic" growth constants m and n. However,
the kinetic parameters for diffusion in layers I and II can
still be obtained from measurements of m and n provided
the interface concentrations and in particular the differ-
ences (cs — ca) and (cb — cc) [see left-hand members of
Eqs. (A-6) and (A-7)] are approximately constant in the
temperature range considered. Then adopting Arrhenius-type
temperature dependencies for the diffusion coefficients:
C, = *, (A-8)
Cn = *n exp(-G n /Rr) (A-9)
where kt and kn are the pre-exponential constants, Ql and
(2n are the activation energies, and R and T denote the gas
constant and the absolute temperature, respectively.
Values for Ql and Qxl are determined from the slopes
of the straight lines obtained on plotting In Ct and In C n vs
1/r. Note that for the determination of Q:l knowledge of
(cc — f3c0) is not required, since the temperature depen-
dence of In Cn is fully determined by n(am + n) [see right-
hand member of Eq. (A-7)]. For the determination of Ql in
principle ca, cc, (cb — cc), and c0 should be known. How-
ever, precise values are not required; thus, if only small
concentration variations are possible within compounds I
and II, the stoichiometrical values can be adopted for ca
and cc, and (cb - cc) ~ 0 (see also Sec. II).
 C M . Brakman, A. W.J. Gommers, and E. J. Mittemeijer: Boride-layer growth kinetics

APPENDIX B: BILAYER GROWTH IN THE CASE (cc ~

OF "B LEAKAGE"
The treatment is based on that given in Appendix A.
Suppose that part of the amount of B arriving at the
layer Il/substrate interface at x = v (see Fig. A-l) is not
used for net growth of layer II but is absorbed by the sub-
strate. It will be assumed that the kinetics for the growth of
layers I and II are (still) controlled only by diffusion of B
in layers I and II. Then, considering a bucket with a hole
in the bottom to be filled with water, the present problem
may be described as "B leakage" (in this sense the case of
"a bucket without a hole" has been treated in Appendix A:
B-saturated substrate).
Assuming that the fraction, /, of the amount of B ar-
riving at the layer Il/substrate interface that is absorbed by
the substrate does not depend on time, the mass balance
for growth of layer II now reads [cf. Eq. (A-4)]
dvnet . (ch — cc)
The formula for the mass balance for growth of layer I [cf.
eq. (A-2)] does not change. Defining the now occurring
parabolic growth constants as m' and n' [cf. Eq. (A-5)],
the pendants of Eqs. (A-6) and (A-7) become
C, - DY(cs - ca)
r r
= 2m'\m'c,. + n'
(1-/)
Cr
- am \ch -
(1-/)
jam' + n')
(B-2)
C° (1 - / )
- cc) = 2n'(am' + n')
(1 - / )
(B-3)
A case covered by the above description is that deal-
ing with a "semi-infinite" substrate with "infinitely fast"
diffusion of B; c 0 = 0 for all t.
"B leakage" can also be realized such that the B ab-
sorbed by the substrate is incorporated in a (third) com-
pound layer adjacent to layer II (Fig. B-l). Under the above
constraints (B diffusion in layers I and II is rate controlling
and B diffusion in layer III is "infinitely fast") the mass
balance for growth of layer III reads
*W _ m (cb ~ cc)
(B-4)
11(v - u)
with
dwna = dw - (1 - a)du - (1 - 0)dvna (B-5)
where x = w indicates the layer Ill/substrate interface,
cd and c 0 denote the constant B concentrations in layer III
and the substrate, respectively, and y is the ratio of the
specific volumes per solvent atom of substrate and com-
pound III, respectively. For "parabolic" growth of layer III
with growth constant l(w — v = 2 • I tm), it follows from
Eqs. (A-3), (B-l), (B-4), and (B-5)
I = (am' + n') (B-6)
-f)(cd-yc0)
If the interface concentrations and the fraction / are
not known, the diffusion coefficients in layers I and II can-
not be determined from measurements of the "parabolic"

B concentration B concentration

x (distance from surface) x (distance from surface)

(1-a)du (1-a)du \ \ ^
adu \
1 •
1 1
in
1=1

1 1 ! n l substrate I j j substr
1 [
.— u —J
— u _J V Ydw net
(3dvne t w
V

I n substrate
m substrate
u + du v+dv u + du v+dv w+dw
FIG. A-l. Solute-element concentration profile assumed in bilayer FIG. B-l. Solute-element concentration profile assumed in three-layer
growth model. growth model.

J. Mater. Res.. Vol. 4. No. 6, Nov/Dec 1989 1369

 C M . Brakman, A.W.J. Gommers, and E.J. Mittemeijer: Boride-layergrowth kinetics

growth constants m' and n'. However, the kinetic parame- f(x,y) = ax2 + bxy + 1 = 0 (C-3)
ters for diffusion in layers I and II can still be obtained
g(x,y) = <fy2 + exy + 1 = 0 (C-4)
from measurements of m' and n', provided the interface
concentrations and the fraction / , and in particular the dif- where a = 201 X 10"9 (mol m" 1 s~l)/p
ferences (c, — ca) and (cb — cc) [see left-hand members b = 119.6 x 10"9 (mol m" 1 s"1)//?
of Eqs. (B-2) and (B-3)], are approximately constant in the d = 119.6 x 10"9 (mol m" 1 s^)/q
temperature range considered. This, again, leads to Eqs. e = 0.8407 x 119.6 X 10"9 (mol nT 1 s~l)/q
(A-8) and (A-9). using data for ca, cc, and a given in Sec. II. Values forp
Then, it immediately follows that the activation energy and q were usually of the order 100 x 10"9 (mol nT 1 s"1).
<2n in fact can be obtained by using the formalism of The roots of Eqs. (C-3) and (C-4) can now be ob-
Eq. (lb) or Eq. (A-7), since the temperature dependence of tained applying a two-dimensional version of the Newton-
In C n is fully determined by n'(am' + n') [see Eq. (B-3)]. Raphson method.25
If cb «= cc and / «= /3cd/cc the term between braces in Consider the functions/(*, y) and g(x,y), as defined
the right-hand member of Eq. (B-2) reduces to m'ca + above, with x and y as independent variables; the roots of
n'cc, and the activation energy Q1 can in fact be obtained Eqs. (C-3) and (C-4) are designated by xr and yr. Sup-
by using the formalism of Eq. (la). Even if/would signifi- pose xn and yn are (the n*) approximations for xr and yr.
cantly deviate from the above value, it can be shown that Then, Taylor series expansions for f(x,y) and g(x, y) around
in many cases only small effects on the temperature depen- (xn,yn) yield
dence of In Q occur [note that the temperature dependence
f(xr,yr) =fixn,yn) + hjx(xn,yn) + jjy{xn,yn)
of In (2m') is of comparable importance for the tempera-
ture dependence of In CY as the temperature dependence of
(C-5)
the logarithm of the term between braces; see Eq. (B-2)].
{xr,yr) = g(xn,yn) + hngx{xn,yn)
APPENDIX C: CALCULATION OF PARABOLIC (C-6)
LAYER-GROWTH CONSTANTS FOR BILAYER
where
GROWTH FROM THE KINETIC PARAMETERS
Q ANDfr ; f, = | = bx ;
In Tables VII and VIII the kinetic data <2FeB, <2Fe2B and
&FeB> ^Fe2B a r e given for the six alloys studied in this paper.
8* Tx ey> Sy
Once such kinetic data are available the parabolic layer-
growth constants for any desired temperature can in prin- and xr = xn + hn and yr = yn + j n .
ciple be calculated. For the calculation of m and n Eqs. (1) Since xH, yn, f(xr,yr)(=0), g(xr,yr)(=0), fx(xn,yn),
and (2) may be rewritten as follows: fy(x«,yn), gx(xn,yn), and gy(xn,yn) are known, estimates
2cam2 + 2ccmn + p = 0 (C-l) for hn and/ n , indicated by /?„' andj n ', can be calculated
from Eqs. (C-5) and (C-6) if the series expansions are
2ccn2 + 2accmn + q = 0 (C-2) truncated after the first derivatives. Hence a next [(« + l)th]
where approximation for xr and yr is obtained from
P= eXp(-g F e B /Rr) = xn h'n (C-7)
yn+1 = y n + i l (c-8)
This is a system of two nonlinear equations in m and n. In leading to new estimates h'n+l a n d ^ + 1 , etc.
order to find the roots an iterative procedure can be used. Evidently, the starting values x0 = 0 and y0 = 0 should
Because p and q are smaller than ca and cc by many not be used. Both parabolic layer-growth constants should
orders of magnitude, the constants in the equations have to be non-negative. Use of two small positive numbers (say
be adapted in order to obtain numerical stability. For the 0.01) as starting values provided an efficient approach. The
present problem the following approach has been proven iteration can be pursued until f(xn+l,yn+1) and g(xn+i,yn+l)
to suffice. both differ by a small amount (depending on the experi-
After division of the left-hand and right-hand members mental accuracy; cf. Tables VI-a, Vl-b, VII, and VIII)
of Eqs. (C-l) and (C-2) by p and q, respectively, and from zero [cf. Eqs. (C-3) and (C-4)]. Rapid convergence
defining x and y as dimensionless quantities as follows was achieved; only a few steps were necessary.
x = m times 1.10"6 s"2 meter" 1 ; Obviously, coincidence of the thus calculated and ex-
perimental data for m and n is achieved only if the corre-
y = n times 1.10"6 s1/2 meter" 1 , sponding experimental data points in the plots of In C vs
Eqs. (C-l) and (C-2) are transformed into equations with 1/7 fall exactly on the fitted straight line (see Figs. 8-13
coefficients of comparable order of magnitude: and Tables Vl-a and Vl-b).

1370 J. Mater. Res., Vol. 4, No. 6, Nov/Dec 1989