Special Points For Brilloofn-Zone Integrations

PH YSICA L RE VIE% B VOLUME 16, NUMBER 4 15 AUGUST 1977
Special points for Brilloofn-zone integrations

D. J. Chadi
Xerox Palo Alto Research Center, Palo Alto, California 94304
(Received 9 March 1977)
The efficiency of two different methods for obtaining "special" points useful for Brillouin-zone integrations
of periodic functions is compared. We find that for some Bravais lattices (such as body~tered cubic and
hexagonal), the method suggested by Monkhorst and Pack leads to different and sometimes less efficient
point sets than those previously obtained by Chadi and Cohen. For a two-dimensional oblique lattice, special
points twice as efficient as those suggested by Cunningham are given.
Two different methods, one by Chadi and Cohen'

(4)
(CC) and the other by Monkhorst and Pack' (MP),
for generating "special" point sets for Brillouin-
zone integration of periodic functions of wave vec-
tor have been suggested so far. In their paper MP
state that their method gives sets of points identi- g=i
i
(x(A (R) = 0, m = 1, 2, . . . ,N (6)
cal to those of CC and additional intermediate sets. for N as large as possible. The special points
By applying their method to cubic Bravais lattices have been defined' to be those which result in the
which they consider, we find this statement to be
largest N for a given number of points.
true for the simple cubic and face-centered-cubic The method of CC is based on the idea that if
lattices but not for the body-centered-cubic lattice
A (k, ) andA (%,) are equal to zero for certain val-
or for other Bravais lattices such as a hexagonal
lattice. For the latter we find the CC point sets to }
ues of m denoted by (i}f, and (M, ), respectively,
then a new set of points can be generated' from%,
be more efficient than those obtained by applying
the method of MP. In the following we briefly re-
and k, such that (5) is satisfied for m in both ) (I,
view the major features of each method and dis-
and
4
{Mj. The new points are related to k, and Q,
cuss the reasons for the difference in the point sets
they generate. At the end of the paper we also give R, =k, + T(R„
the special points for a two-dimensional oblique where T, runs over the point-group operations of
lattice. These points are twice as efficient as
those given by Cunningham. ' the Bravais lattice. Successive applications of (6)
using the R,. 's instead of %, and a new point R, (in-
The special points method is related to obtaining
stead of%, ), which sets A (%,) equal to zero for
the average over the Brillouin zone of a periodic
values of m not covered by fc, and R„ lead to suc-
function
cessively larger point sets. The weighting factor
f$) f++ f fn= 1
A (k),
for each point is determined only by the symmetry
of that point, as discussed in Ref. 1.
MP' have suggested another interesting method
where
for generating special points. Their method re-
duces the three-dimensional problem to three one-
dimensional problems. To show this we write
~ik ' R ~i2+ni+a ei2m n2&84' i2~3~'y
with 0& c + c y By summing over the inequival- (7)
ent members of the star of k, a periodic function where, if (F„E„7g and( G„G„Gg are the prim-
f(k) can always be made to have the full symmetry itive translation vectors in real and reciprocal
of the Bravais lattice as in (1). The Brillouin- space, we have
f,
zone average off (k) is equal to ; we are, there-
R=n, t, +n, E, +n, ES, (8)
fore, interested in the points k, for which
f, = Q u(f (k() fr=kaG(+ k((G. 2+ k G3 ~ (9)

To eliminate exp(ik R) from the Fourier trans-
is a good approximation, with a, being the weight-
ing factor of ki To accomplish this we need
f
f»m of (k) by using a sum over points R„ it is pos-
sjble and sufficient to consider the three one-dim-
16 1746
16 SPECIAL POINTS FOR BRILLOUIN-ZONE INTEGRATIONS 1747
ensional problems associated with each exponential fact that in picking a "generating wave vector" CC
on the right-handside of (7). If we take choose points that set the sum of exp(t'k R} over
-—1)/2N, r=1, 2, . . . , N related lattice vectors equal to zero, whereas by
k =(2r N (10)
construction MP set each exponential term indi-
and similarly for ks( and k&(&), then the sum vidually equal to zero. The latter procedure may
exp(i% R) over the points not always lead to the best possible point sets.
For example, for the same accuracy a straight-
k„„=ku " G, + %&is) Ga+ k&i G3 i r = I, . . . , Nc
t&
forward application of MP's method for a hexagon-
s=1, . . . , N&, t=1, . . . , N„, (11) al lattice leads to larger point sets than those of
CC, although it may be possible to modify MP's
is zero for all R for which at least one of the fol- method to obtain the same results.
lowing conditions is satisfied: 0& (n, &N, or
0& (u, ( &Nsz or 0 & )n, ( &N„; [the relation between
(
%e now discuss special points for two-dimen-
sional Brillouin-zone integrations. Cunningham
the n's and R being given by (8)j. The size of the
has obtained the special point sets for two-dimen-
resulting point set is determined by the numbers
sional zones. However, the point sets he gives
N~, NS, and N~. for the oblique lattice are not the optimal ones.
For the simple cubic and face-centered-cubic
Defining a general oblique lattice by the primitive
Bravais lattices, the method of MP outlined above
translation vectors
gives special point sets identical to those obtained
by CC and also additional intermediate sets. For X, =a(1, 0), K, =a(5, P),
the body-centered-cubic (bcc) structure, however,
the special points in ascending order of accuracy
the point sets resulting from the two methods are
are for set 1
different. For example, CC find the two points
(units of 2s/a, where a is the lattice constant) %=(I, (1-8)/P) (units of w/2a), a= 1;
(4&4&4)yt (t2 (s&4&4}t (12) for set 2 in units of s/2a
' for each point,
with weighting factors a of —, k, =(1, (1 —8)/p), %2=(1, (1+ 8)/p),
whereas the method of MP gives
a, =a, =~;
', ', '),
%, =(-, —, —, %, =(-,', 0, 0),
(13)
3 for set 3 in units of s/4u
Q2= g ~
The first set of points is slightly more efficient %, =(l, (1-5)/P), R, =(1, —(1+5)/P),
for Brillouin-zone integrations than the second k, =(1, (3 —8)/P), %, =(1, —(3+8)/P),
set; they eliminate the first five nearest-neighbor
k, =(3, 3(1 —8)/P), R, =(3, —3(1+8)/P),
shells from the Fourier transform of (%} as com- f
pared to the first four shells for the second set of %, =(3, (1-35)/P), %, =(3, —(1+38)/P),
points. The larger point sets obtained from either ~ ~ l
Q ~
Q
method, although always different, appear to be
roughly equivalent' in their efficiency: The eight- The last two sets given above have the same ac-
point set of CC fails for R= (4, 0, 0)a; the six- curacy as those given by Cunningham but are a
point set' obtained by applying MP's method fails factor of 2 more efficient. The method of MP can
for R= (2, 2, 2)a. The difference in the special also be used to determine intermediate point sets
point sets obtained by CC and MP results from the for the two-dimensional case.
'D. J. Chadi and M. L. Cohen, Phys. Rev. B 8, 5747(1973). are identical, the bcc points of MP can also be used for
2H. J. Monkhorst and J. D. Pack, Phys. Rev. B 13, 5188 other cubic lattices. For example, for the fcc lattice
(1976). the points equivalent to thoseyiven in Eq. (13) would
~S. L. Cunningham, Phys. Rev. B ~,
4988 (1974). -
be kg—(g,1 1 I 1
g, g), k2=(g, 0, 0), k3=(1, ~, 0), with nf 1
The same set of points results if k &+ Ti Ki is taken in-
stead of (6).
f
u2 = o3= . For the bcc lattice k& aud R& are equivalent
points. For a function having the full symmetry of the
'For the bcc Bravais lattice the star of k points obtained Bravais lattice, the two-point scheme of CC gives
by applying the method of MP (Ref. 2) contains fewer more accurate results than the three points given above.
points than those of CC (Ref. 1). For functions f(k) 6For the bcc lattice the expression given by MP (Ref. 2)
which do not have the full symmetry of the lattice, it for the number of special points in a given set does not
may, therefore, be preferable to use the MP point actually apply to the sets derived by applying their
sets. Since in an extended-zone scheme the special method. It applies correctly, however, to the special.
point sets for all cubic lattices [whether simple cubic point sets that can be obtained by transforming the
or face-centered-cubic (fcc) or body-centered cubic] simple cubic special points into the bcc Brillouin zone.

Special Points For Brilloofn-Zone Integrations

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Special Points For Brilloofn-Zone Integrations

Uploaded by

Copyright:

Available Formats

PH YSICA L RE VIE% B VOLUME 16, NUMBER 4 15 AUGUST 1977

Special points for Brilloofn-zone integrations

Two different methods, one by Chadi and Cohen'

f, = Q u(f (k() fr=kaG(+ k((G. 2+ k G3 ~ (9)

You might also like