Professional Documents
Culture Documents
In this section, we use Kane’s model24 to find the band- FIG. 1. A wurtzite crystal consists of two interpenetrating hex-
edge energies and define important physical parameters in- agonal closely packed sublattices, displaced by 5/8 of the c~0001!
cluding the optical momentum-matrix elements. axis. The dashed lines show the boundary of a unit cell.
A. The Hamiltonian matrix elements Here, V(r) is the periodic potential, H so accounts for the
The Schrödinger equation for the periodic part u nk(r) of spin-orbit interaction, and s i ,i5x,y,z are the Pauli spin ma-
trices, which are defined in Appendix A.
the Bloch function and the energy E n (k) near the band edge
In the set of basis functions near the zone center,
can be written as24,26
Hu nk~ r! . H 0 1 S
\ 2k 2
1
2m 0 m 0
\
k•p1H so u nk~ r! D u iS↑ & , u 15 2 U ~ X1iY !
A2
↑ ,L u 25 U ~ X2iY !
A2 L
↑ ,
5E n ~ k! u nk~ r! , ~1!
where u 3 5 u Z↑ & ;
U L U L
~4!
p2 ~ X2iY ! ~ X1iY !
H 05 1V ~ r! , ~2! u iS↓ & , u 45 ↓ , u 55 2 ↓ ,
2m 0
A2 A2
\
H so5 ¹V3p• s5H sx s x 1H sy s y 1H sz s z . ~3!
4m 20 c 2 u 6 5 u Z↓ & .
~5!
54 k–p METHOD FOR STRAINED WURTZITE SEMICONDUCTORS . . . 2493
where the x-y plane has been used. For example, (x,y) mapping to
a 60° rotation around the z axis (5c axis! leaves the Hamil-
k 6 5k x 6ik y ~6! tonian invariant, and the wave functions u X & and u Y & trans-
and we have used the following definitions for the energies: form to
^ S u H 0 u S & 5E c , 1 A3
u X 8& 5 u X & 1 uY &,
2 2
^ X u H 0 u X & 5 ^ Y u H 0 u Y & 5E v 1D 1 ,
^ Z u H 0 u Z & 5E v , ~7! A3 1
u Y 8 & 52 uX&1 uY &. ~9!
2 2
^ X u H sz u Y & 52iD 2 ,
Therefore,
^ Y u H sx u Z & 5 ^ Z u H sy u X & 52iD 3 ,
3 1
and two Kane’s parameters, P 1 and P 2 , which are related to ^ Y u H 0u Y & 5 ^ Y 8u H 0u Y 8& 5 ^ X u H 0u X & 1 ^ Y u H 0u Y &
4 4
the interband momentum-matrix elements, ~10!
K U UL iS
\ ]
i ]z
Z 5
m0
\ 1
P ,
and we find ^ Y u H 0 u Y & 5 ^ X u H 0 u X & . We also obtain
^ X u H 0 u Y & 50, and ^ S u H 0 u X & 50. The properties
^ S u ] / ] x u X & 5 ^ S u ] / ] y u Y & and ^ Y u H sx u Z & 5 ^ Z u H sy u X & can be
K U UL K U UL
iS
\ ]
i ]x
X 5 iS
\ ]
i ]y
Y 5
m0
\ 2
P . ~8!
derived in a similar manner.
E 2 5E v 1
D 1 2D 2
2
1 AS D 1 2D 2
2 D 2
12D 23 , au 2 1bu 6 , bu 3 1au 5 , ~11!
E 3 5E v 1
D 1 2D 2
2
2 AS D 1 2D 2
2 D 2
12D 23 , bu 2 2au 6 , 2au 3 1bu 5 ,
\ 2k 2 ~ E g 1D 1 1D 2 !~ E g 1D 2 ! 2D 23 \
E ~ k! 5E c 1 1 P5p1 s3¹V. ~20!
2m 0 E g @~ E g 1D 1 1D 2 !~ E g 12D 2 ! 22D 23 # 4m 0 c 2
3 P 22 ~ k 2x 1k 2y ! It is noted that the last term in P is usually neglected for
simplicity.8,26 The band-edge wave function can be written
~ E g 12D 2 ! as
1 P 21 k 2z , ~16!
~ E g 1D 1 1D 2 !~ E g 12D 2 ! 22D 23
A B
which can also be written in terms of the transverse and u k~ r! 5 (j a j 8~ k! u j 80~ r! 1 (g a g~ k! u g 0~ r! ~21!
longitudinal effective masses. 8
54 k–p METHOD FOR STRAINED WURTZITE SEMICONDUCTORS . . . 2495
and we choose the six valence-band wave functions, Using Löwdin’s method,28 the six-by-six Hamiltonian
$ u 1 ,u 2 , . . . ,u 6 % in Eq. ~4!, as the bases for the states of matrix for the valence bands can be written as the sum of a
interest ~called class A), and all other states of no interest are band-edge contribution and a k-dependent contribution.
called class B.
21 H 636,j j 8 ~ k! 5H 636,j j 8 ~ k50 ! 1D j j 8 ~23!
u u 1& 5 u ~ X1iY ! ↑ & ,
A2 and the k-dependent matrix is
1
u u 2& 5 u ~ X2iY ! ↑ & ,
A2 D j j 85 ( ab
D j j8k ak b ,
a,b
u u 3 & 5 u Z↑ & ,
u u 4& 5
1
A2
u ~ X2iY ! ↓ & , ~22! ab
D j j85
\2
F
d d 1
2m 0 j j 8 ab (g
B b
m 0 ~ E 0 2E g !
a
p aj g p g j 8 1 p bj g p g j 8
G , ~24!
21
u u 5& 5 u ~ X1iY ! ↓ & , where the indices j, j 8 51,2, . . . , 6PA, g PB, and
A2 a , b 5x,y,z. The band-edge Hamiltonian matrix has been
u u 6 & 5 u Z↓ & . obtained from Kane’s model,
3 4
E v 1D 1 1D 2 0 0 0 0 0
0 E v 1D 1 2D 2 0 0 0 A2D 3
0 0 Ev 0 A2D 3 0
H 636 ~ k50 ! 5 . ~25!
0 0 0 E v 1D 1 1D 2 0 0
0 0 A2D 3 0 E v 1D 1 2D 2 0
0 A2D 3 0 0 0 Ev
D 333 5 F L 1 k 2x 1M 1 k 2y 1M 2 k 2z
N 1k xk y
N 2k xk z
N 1k xk y
M 1 k 2x 1L 1 k 2y 1M 2 k 2z
N 2k yk z
N 2k xk z
N 2k yk z
M 3 ~ k x 1k 2y ! 1L 2 k 2z
2
G uX&
uY &,
uZ&
~26!
2m 0 S
11
B
2p Xx g p gx X
(g m 0~ E 0 2E g ! D 5
\2
2m 0 S
11
B
2 p xY g p xg Y
(g m 0~ E 0 2E g ! D ,
5
\2
2m 0
11 S B
2p y p y
(g m 0~ EY0g2Eg Yg ! D ,
M 25
\2
2m 0
11 S B
2pz pz
(g m 0~ EX0g2Eg Xg ! D
L 25
\2
2m 0 S
11
B
2p Zz g p gz Z
(g m 0~ E 0 2E g ! D , 5
\2
2m 0 S
11
B
2 p zY g p zg Y
(g m 0~ E 0 2E g ! D , ~27!
2496 S. L. CHUANG AND C. S. CHANG 54
M 35
\2
2m 0
11 S B
2p x p x
(g m 0~ EZ0g2Eg Zg ! D L 1 2M 1 5N 1 .
S D
B written as
\2 2p y p y
5 11 (g m 0~ EZ0g2Eg Zg ! ,
3 4
2m 0 F 2K * 2H * 0 0 0 u u 1&
2K G H 0 0 D u u 2&
\2
B
p Xx g p gy Y 1p Xy g p xg Y 2H H* l 0 D 0 u u 3&
N 15
m 20 (g ~ E 0 2E g !
, H5
0 0 0 F 2K H u u 4&
,
0 0 D 2K * G 2H * u u 5 &
B B 0 D 0 H* 2H l u u 6&
\2 p Xx g p gz Z 1p Xz g p gx Z \2 p Yy g p gz Z 1p zY g p gy Z
N 25
m 20 (g ~ E 0 2E g !
5
m 20 (g ~ E 0 2E g !
, ~31!
where
3 4
D 11 21 2D *
D* \2
23
K5 A ~ k 1ik y ! 2 ,
D 21 D 11 D 23 0 2m 0 5 x
2D 23 D *
23 D ZZ \2
D 636 5 , H5 A ~ k 1ik y ! k z ,
D 11 D 21 D 23 2m 0 6 x
* 2D 23
*
D5 A2D 3 .
0 D 21 D 11
23 2D 23
D* D ZZ
We obtain the relations between the band-structure param-
~28!
eters derived using the k•p method and the more commonly
used A i parameters in the Pikus-Bir model,
where the whole matrix can be expressed by using only four
distinctive matrix elements, \2 \2 \2
A 5L 2 , A 5M 3 , A 5M 2 2L 2 ,
2m 0 1 2m 0 2 2m 0 3
D 115 S L 1 1M 1
2 D
~ k 2x 1k 2y ! 1M 2 k 2z , \2
A 5
2m 0 4
L 1 1M 1
2
2M 3 ,
\2 N1
A 5 ,
2m 0 5 2
\2
A 5
~33!
N2
2m 0 6 A2
.
with the deformation potentials, D 1 ,D 2 , . . . ,D 6 , at the cor- stants. The general strain-stress relation ~Hooke’s Law! for
responding positions of A 1 ,A 2 , . . . ,A 6 . the hexagonal crystal can be found in Ref. 29.
F5D 1 1D 2 1l1 u ,
IV. THE INVARIANT METHOD
G5D 1 2D 2 1l1 u , FOR THE EFFECTIVE-MASS HAMILTONIAN
l e 5D 1 e zz 1D 2 ~ e xx 1 e y y ! , H5D 1 J 2z 1D 2 J z s z 1D ~ J 1 s 2 1J 2 s 1 !
\2
\2 1 @~ A 1 1A 3 J 2z ! k 2z 1 ~ A 2 1A 4 J 2z !~ k 2x 1k 2y !
u5 @ A k 2 1A 4 ~ k 2x 1k 2y !# 1 u e , 2m 0
2m 0 3 z
~34! 2A 5 ~ J 21 k 22 1J 22 k 21 ! 22A 6 k z ~@ J z J 1 # k 2 1 @ J z J 2 # k 1 !#
u e 5D 3 e zz 1D 4 ~ e xx 1 e y y ! , 1 ~ D 1 1D 3 J 2z ! e zz 1 ~ D 2 1D 4 J 2z !~ e xx 1 e y y !
\2 2D 5 ~ J 21 e 2 1J 22 e 1 ! 22D 6 ~@ J z J 1 # e z2 1 @ J z J 2 # e z1 ! ,
K5 A ~ k 1ik y ! 2 1D 5 e 1 ,
2m 0 5 x ~38!
\2 where
H5 A k ~ k 1ik y ! 1D 6 e z1 ,
2m 0 6 z x
1
D5 A2D 3 , J 65 ~ J x 6iJ y ! ,
A2
where
e 6 5 e xx 62i e xy 2 e y y , 1
s 6 5 ~ s x 6i s y ! , ~39!
2
e z6 5 e zx 6i e yz . ~35!
It should be pointed out that under the cubic 2 @ J z J 6 # 5J z J 6 1J 6 J z ,
approximation,8,14 the following relations hold for the pa-
rameters A i s and D i s: and we have specified the bases, $ Y 11 , Y 10 , and Y 121 % , with
the operator matrices, J x , J y , J z , and the Pauli spin matrices
A 1 2A 2 52A 3 52A 4 , A 3 14A 5 5 A2A 6 , D 2 5D 3 , explicitly defined in Appendix A. It should be noted that
~36! because of the changes in the basis functions involving
D 1 2D 2 52D 3 52D 4 , D 3 14D 5 5 A2D 6 .
changes of signs and phase factors in early and recent papers,
Therefore, only five band-structure parameters, such as A 1 , confusion occurred in deriving the Hamiltonian matrix.
A 2 , A 5 , D 1 , and D 2 , and three deformation potentials are Since the previous derivations for the wurtzite Hamiltonian
necessary for the calculation of the valence-band structures. were based on the invariant method, the sign and the phase
The cubic approximation idea is based on the similarity be- factor in front of the A 5 and A 6 ~or the H and K terms! were
tween the wurtzite structure and the cubic crystal. The above arbitrarily chosen as long as the Hamiltonian stays Hermit-
relations are derived8 if we map the c ~0001! axis to the z 8 ian. A consistent approach is to derive the Hamiltonian using
axis along the (111) direction and set the x 8 and y 8 axes the k•p method, as shown in Sec. III.
along the @ 112 # and @ 110# directions, respectively, in the In writing the Hamiltonian, we have kept in mind that the
coordinate system. It has also been found8,14 that a seventh band-structure parameters A i ,i51,2,3, . . . ,6, are all real
coefficient A 7 for the linear k terms vanishes; therefore, we numbers, following our definitions using the k•p method.
discard it at the beginning and keep only the quadratic terms The Hamiltonian ~38! can also be simplified to the following
of k, in addition to the band-edge energy terms. form:
For a strained-layer wurtzite crystal pseudomorphically
grown along the ~0001! (z axis! direction, the strain tensor H5 ~ D 1 1 u ! J 2z 1D 2 J z s z 1l
e% has only the following nonvanishing diagonal elements:
1D ~ J 1 s 2 1J 2 s 1 ! 2 ~ K * J 21 1KJ 22 !
a 0 2a
e xx 5 e y y 5 , 2 ~ H * 2 @ J z J 1 # 1H2 @ J z J 2 # ! , ~40!
a
3 4
F 2H * 2K * 0 0 0 u Y 11↑ & ~ 5 u u 1 & )
2H l H* D 0 0 u Y 10↑ & ~ 5 u u 3 & )
2K H G 0 D 0 u Y 121 ↑ & ~ 5 u u 2 & )
H5 . ~41!
0 D 0 G 2H * 2K * u Y 11↓ & ~ 5 u u 5 & )
0 0 D 2H l H* u Y 10↓ & ~ 5 u u 6 & )
0 0 0 2K H F u Y 121 ↓ & ~ 5 u u 4 & )
This result, when written in the same order of the bases where k x 1ik y 5k t exp(i f ) has been used.
$ u 1 ,u 2 , . . . ,u 6 % , gives an identical result as that in Eq. ~31!, First, it is easy to see that we can remove the f
which was derived from the second-order k•p method taking dependence in the matrix elements by introducing
the symmetry of wurtzite crystals into consideration. the bases, $exp(2i3 f /2) u Y 11↑ & , exp(i f /2) u Y 121 ↑ & ,
It should be pointed out that some sign changes in the exp(2i f /2) u Y 10↑ & , exp(i3 f /2) u Y 121 ↓ & , exp(2i f /
choice of bases are possible, such that the original definitions 2) u Y 11↓ & , exp(i f /2) u Y 10↓ & % . By pairing the bases $ u 1 ,u 4 % ,
of the K and H terms of Pikus and Bir can still be used. For $ u 2 ,u 5 % , and $ u 3 ,u 6 % , we form the bases u 1 & , u 2 & , u 3 & ,
example, using $ 2u 1 ,u 2 ,iu 3 ,u 4 ,2u 5 ,iu 6 % , we find a sign u 4 & , u 5 & , and u 6 & by using the basis transformation, u i &
change in front of K and an extra factor of i in both H and 5ST i j u j , where the T matrix is defined as
D. In this set of bases, the complete Hamiltonian matrix
differs from the original form in Ref. 8, only by an extra i
3 4
a* 0 0 a 0 0
factor in front of D. We believe that our formulation gives
consistent results based on both the Luttinger-Kohn model 0 b 0 0 b* 0
and the invariant method. 0 0 b* 0 0 b
T5 , ~43!
V. BLOCK-DIAGONALIZATION
a* 0 0 2a 0 0
AND THE VALENCE-BAND DISPERSIONS 0 b 0 0 2b* 0
The full six-by-six Hamiltonian matrix can be block di- 0 0 2b* 0 0 b
agonalized following a similar procedure26,30 to that of the
where
zinc-blende structure. We note that the off-diagonal terms
such as K and H contain the f dependence and write
1 1
\2 a5 e i ~ 3 p /413 f /2! , b5 e i ~ p /41 f /2! . ~44!
K5K t e 2 f , K t 5 A k2 ; A2 A2
2m 0 5 t
\2
H5H t e i f , H t5 A kk , ~42! The Hamiltonian matrix is then block diagonalized,
2m 0 6 t z H 8 5UHU † 5T * HT t ,
3 4
F Kt 2iH t 0 0 0
Kt G D2iH t 0 0 0
iH t D1iH t l 0 0 0
H 85 , ~45!
0 0 0 F Kt iH t
0 0 0 Kt G D1iH t
0 0 0 2iH t D2iH t l
where the superscript † means taking both transpose ~super- VI. ANALYTICAL EXPRESSIONS
script t) and complex conjugate (*). The transformation FOR THE EFFECTIVE MASSES
matrix26 for the components of the wave functions is AND OPTICAL MOMENTUM-MATRIX ELEMENTS
U5T * . By writing the upper-left Hamiltonian as H U and the The effective masses along the transverse and longitudi-
lower-right Hamiltonian as H L , we find the relation nal directions can be derived as follows:
H U 5(H L ) * 5(H L ) t . ~i! Near the zone center (k approaches zero!, we keep
54 k–p METHOD FOR STRAINED WURTZITE SEMICONDUCTORS . . . 2499
TABLE I. Normalized inverse effective masses of the heavy-hole ~HH!, light-hole ~LH!, and crystal-field
split hole ~CH! bands along the c axis ~5 z axis! and transverse (t) to the c axis.
Valence energy m 0 /m z m 0 /m t
E 3 ~CH band!
F S DG
2 A11
E032le
E032E02
A3 F S DG
2 A21
E032le
E032E02
A4
only terms of up to the second order in k and find ~for a finite the strain effects in Tables I and II are simple and useful,
D i ,i51,2,3) where E 01 , E 02 , and E 03 , are the band-edge energies, E 1 ,
E 2 , E 3 in Eq. ~46!, evaluated at k50,
E 1 5F,
E 01 5D 1 1D 2 1 u e 1l e ,
E 25
G1l
2
1 AS G2l
2 D 2
1D 2 , ~46! E 02 5
D 1 2D 2 1 u e
2
1l e 1 AS D 1 2D 2 1 u e
2 D 2
12D 23 ,
~47!
E 35
G1l
2
2 AS G2l
2 D 2
1D 2 .
E 03 5
D 1 2D 2 1 u e
2
1l e 2 AS D 1 2D 2 1 u e
2 D 2
12D 23 .
The band-edge effective masses along the longitudinal We also redefine Eq. ~12! to include the strain effects,
(z5c axis! and the transverse (x or y) directions can then be
obtained and are tabulated in Table I. Using the band-edge E 02 2l e
wave functions ~11!, we also obtain the momentum-matrix a5 ,
elements in Table II for the optical transition from the con-
A~ E 02 2l e ! 2 12D 23
duction band edge to the three valence-subband edges ~C-
HH, C-LH, and C-CH transitions labeled as E 1 , E 2 , and A
2D 3
b5 , ~48!
E 3 , respectively!, when the optical polarization is parallel or A ~ E 2 2l e ! 2 12D 23
0
perpendicular to the c axis. Our analytical results including
and it is straightforward to show that
TABLE II. Strain-dependent interband momentum-matrix ele-
ments z^ S u ê•pu v i & z2 for ê polarizations along the c axis ~5z) and E 02 2l e E 03 2l e
a 25 0, b 25 ,
perpendicular to the c axis. The band-edge wave functions v i are E 02 2E 3 E 03 2E 02
listed in Eq. ~11!. Here, the energy parameters for the interband
transition elements are related to Kane’s parameters, P 1 and P 2 in which are used in Tables I and II.
Eq. ~18!, by E pz 5(2m 0 /\ 2 ) P 21 and E px 5(2m 0 /\ 2 ) P 22 . Note that The conduction-band edge has a hydrostatic energy shift,
a 2 1b 2 51; a and b are defined in Eq. ~48!. P ce ,
S D S D
4 px
E 2 ~LH band! m0 m 0
b2 E a2 E For a GaN layer sandwiched between two thick
2 pz 4 px Al x Ga 12x N layers, the strain is compressive in nature, since
E 3 ~CH band!
a2 S D
m0
E
2 pz
b2 S D
m0
E
4 px
the lattice constant a of the GaN layer before deformation is
larger than the lattice constant a 0 of Al x Ga 12x N layers.
Therefore, e xx is negative, and the band-edge shifts will be
Sum m0 m0
E E positive for the conduction band and negative for the top
2 pz 2 px valence bands. The net band-gap shift is determined by
2500 S. L. CHUANG AND C. S. CHANG 54
E c 2E 01 5E g 1 P c e 2 ~ u e 1l e ! , ~50!
and it can be compared with the lowest band-edge transition
energy ~A line! of the photoluminescence spectrum, with an
exciton correction energy of about 28 meV. Similarly,
E c 2E 02 and E c 2E 03 with the same exciton correction can be
used to compare with the transition energies of the B and
C lines in the photoluminescence spectrum.
~ii! Away from the zone center, as k approaches a big
number, we can ignore D 2 and D 3 ; the rest of the Hamil-
tonian has analytical solutions:
E 1 5F 8 2K t ,
1
E 2 5 @ F 8 1K t 1l1 A~ F 8 1K t 2l ! 2 18H 2t # , ~51!
2
1
E 3 5 @ F 8 1K t 1l2 A~ F 8 1K t 2l ! 2 18H 2t # ,
2 FIG. 3. Interband energies calculated for the transitions between
the conduction band E c and the HH, LH, and CH bands, or called
where F 8 5D 1 1l1 u . Using the above expressions, the A, B, and C lines. The symbols are experimental data from GaN
valence-band effective masses away from the zone center samples at temperatures below 10 K collected in Ref. 23. The solid
can be obtained analytically and are tabulated in Table I. lines are calculated using the parameters in Table III in expressions
Note that in the special case of D 2 5D 3 50, these analytical in Eq. ~47! for E 01 , E 02 , and E 03 , and E c in Eq. ~49! with an exciton
effective masses are the exact solutions of the three-by-three energy (E ex'0.028 eV! correction @ E g ~10 K!2E ex'3.471 eV#.
Hamiltonians H U and H L . The above analysis also explains The dashed lines are calculated using the same equations with pa-
the difference between the expressions for the effective rameters taken from Ref. 23.
masses used in Refs. 8 and 14.
e xx 52 @ C 111C 1222(C 213/C 33) # 21 T and T is the magnitude
of the equivalent in-plane compressive stress in kbar. It is
VII. NUMERICAL EXAMPLES
noted that the fit is not unique, since the roots for D cr and
For a special case, if we ignore the anisotropic property of D so are symmetric, as can be seen from Eqs. ~14!. It is usu-
the wurtzite crystals, we can assume that a cz 5a ct 5a c . ally believed that D cr>D so53D 2 ; and D cr522 meV,
Since the experimental data only provide the total band-gap D so511 meV, have been used. These values affect the mag-
shift as a function of an externally applied pressure, only the nitudes of E 01 , E 02 , E 03 , and the polarization selection rules
total value for the interband deformation potential, a, is re- for optical transitions between the conduction band and the
ported. Theoretically, the hydrostatic deformation potentials LH (E 02 ) and CH (E 03 ) valence bands. We have varied the
a5 ] E g / ] (lnV)5ac2av ~interband!, a c ~conduction band!, parameters using D 1 516 meV, D 2 5D 3 54 meV and found
a v ~valence band!, and the shear deformation potential, b, that our fit with the data in Fig. 3 is within 2 meV. Our
have been used for zinc-blende structures. Possible rules for parameters will not reverse the polarization selection rule
the partition ratio, u a c /a u and u a v /a u , are used based on compared with previous report,27 while the D i parameters of
different theories. Note that the common convention that a Ref. 23 will have different polarization selection rules for the
and a c are negative and a v is positive is used. For wurtzite B and C lines of the photoluminescence spectrum from those
structure, the parameters, D 1 and D 2 , play similar roles23 as in Ref. 27. More experimental data on the polarization selec-
the hydrostatic deformation potential, a v ; while D 3 and D 4 tion rules, especially near the B and C lines of the interband
play similar roles as the shear deformation potential, b, of transitions, are required to resolve this issue. We then
cubic crystals. Various values of a for GaN, such as13,22 changed the deformation potentials: a c 50.5a524.08 eV,
27.8, 29.2, and 211.8 eV have been reported. A set of D 1 50.7 eV, D 2 52.1 eV, D 3 51.4 eV, and D 4 520.7 eV,
parameters, using a528.16 eV, a c 50, D 1 5D 2 58.16 eV, and obtain the solid lines in Fig. 3. These parameters satisfy
D 3 5D 4 523.71 eV, D 1 510 meV, D 2 56.2 meV, and the cubic approximation, D 1 2D 2 52D 3 52D 4 . There are
D 3 55.5 meV have been used23 to fit experimental data col- few reports on the experimental values of a c and D i s, since
lected from a few samples for temperatures lower than 10 K. these parameters cannot be independently measured so far.
Note that these parameters were obtained by a linearization The above values are of the same magnitudes as the hydro-
of the valence-band-edge energies with respect to strain. We static and shear deformation potentials of the other III-V
have used the same parameters to calculate the transition zinc-blende crystals.
energies ~with exciton corrections! for the A-line, B-line, and In Figs. 4~a! and ~b!, we plot the band-edge energy shifts
C-line interband transitions in the photoluminescence spec- for the conduction band and the three valence bands as a
trum using the exact expressions, ~47! and ~49!, instead of function of the in-plane compressive strain up to a strain
the linearization formulas. The results are shown as the value of a magnitude 0.01 ~or one percent!. Note that we use
dashed lines in Fig. 3, while the data points collected in the convention e xx to be negative for a biaxial compression
Ref. 23 are shown as symbols. We have used case. One percent of compressive strain is the amount of the
54 k–p METHOD FOR STRAINED WURTZITE SEMICONDUCTORS . . . 2501
band along the k t direction, and they have similar effective ization!. The analytical expressions are tabulated in Table II.
masses in the k z direction. Since D 2 and D 3 are small, the We also calculate the energy parameters for the interband
band structures appear to be dominated by the features de- optical-transition oscillator strengths and obtain E px 515.7
termined by the effective masses away from the zone center. eV, and E pz 5 13.9 eV. Notice the sum rules for each polar-
This suggests that for a GaN quantum-well structure, the ization in Table II. All the matrix elements are normalized to
heavy-hole and light-hole subband energies will be very the total value for the TE polarization. The sum of all three
close and the density of states for the heavy-hole subbands transitions, C-HH, C-LH, and C-CH, for the TE polarization
will be much larger than that for the light-hole subbands. is unity, and the sum of all three matrix elements for the TM
The normalized matrix elements for the optical transitions polarization is E pz /E px 50.89.
from the conduction band to the HH, LH, and CH bands are
VIII. CONCLUSIONS
plotted as a function of the in-plane strain, as shown in Fig.
7 for the optical polarization (ê) parallel to the c (z) axis In conclusion, we have derived the effective Hamiltonian
~TM polarization! and perpendicular to the c axis ~TE polar- for the wurtzite crystals using both the k•p method and the
54 k–p METHOD FOR STRAINED WURTZITE SEMICONDUCTORS . . . 2503
ACKNOWLEDGMENTS
s x5 F G
0
1
1
0
, s y5 F G 0
i
2i
0
, s z5 F 1
0
0
21
G , ~A1!
F G F G
fundamental band-structure parameters, the interband
momentum-matrix elements, and the energy parameters. A 0 1 0 0 2i 0
unitary transformation is also found to block diagonalize the 1 1
J x5 1 0 1 , J y5 i 0 2i ,
six-by-six matrix into two three-by-three matrices and it sim- A2 A2
plifies the derivations for the valence-band dispersions and 0 1 0 0 i 0
F G
effective masses. Analytical expressions for the effective
1 0 0
masses near and far away from the band edges and the band-
edge wave functions are derived within the framework of the J z5 0 0 0 . ~A3!
k•p method. Analytical expressions and numerical results 0 0 21
for the effective masses and optical interband transition ma-
trix elements taking into account the strain effects have also The basis functions are in the following order:
been presented. This Hamiltonian will be very useful for the
calculation of both electronic and optical properties of 21 1
Y 115 u X1iY & , Y 105 u Z & , Y 121 5 u X2iY & .
strained bulk and quantum-well structures, using GaN/ A2 A2
Al x Ga 12x N or In y Ga 12y N/Al x Ga 12x N materials, which ~A4!
have a great potential for applications to blue-green laser
We also use
F G
diodes and electronic devices.
0 1 0
1
J 15 ~ J x 1iJ y ! 5 0 0 1 ,
A2
0 0 0
J 25
A2
1
~ J x 2iJ y ! 5 1
0
F G
0 0
0
1
0
0 .
0
~A5!
Ĝ 8 5Ĝ, 4 6
L 1 2M 1 5 ~ L 1 2M 1 ! 1 ~ N 1 1N *
1 !. ~B6!
16 16
1 A3 A3 1
X 8 5 X1 Y, Y 8 52 X1 Y , ~B4! Therefore,
2 2 2 2
L 1 2M 1 5N 1 , ~B7!
x8
1 A3 y y8
A3
1
p 5 p x1 p , p 52 p 1 p y,x
since N 1 is real. The relation ~B7! is important, since it re-
2 2 2 2
lates the Hamiltonian ~31! derived using the k•p method to
in that derived in the invariant method.
1
S. Nakamura, T. Mukai, and M. Senoh, Jpn. J. Appl. Phys. 30, 17
S. L. Chuang and C. S. Chang, Appl. Phys. Lett. 68, 1657 ~1996!.
L1998 ~1991!. 18
A. Rubio, J. L. Corkill, and M. L. Cohen, Phys. Rev. B 49, 1952
2
S. Nakamura, T. Mukai, and M. Senoh, Appl. Phys. Lett. 64, ~1994!.
1687 ~1994!. 19
J. Neugebauer and C. G. Van de Walle, Phys. Rev. B 51, 10 568
3
X. H. Yang, T. J. Schmidt, W. Shan, J. J. Song, and B. Golden- ~1995!.
berg, Appl. Phys. Lett. 66, 1 ~1995!. 20
A. Rubio and M. L. Cohen, Phys. Rev. B 51, 4343 ~1995!.
4
S. T. Kim, H. Amano, I. Akasaki, and N. Kiode, Appl. Phys. Lett. 21
S.-H. Wei and A. Zunger, Phys. Rev. Lett. 76, 664 ~1996!.
64, 1535 ~1994!. 22
W. Shan, T. J. Schmidt, X. H. Yang, S. J. Hwang, J. J. Song, and
5
H. Amano, T. Tanaka, Y. Kunii, K. Kato, S. T. Kim, and I. B. Goldenberg, Appl. Phys. Lett. 66, 985 ~1995!; W. Shan et al.,
Akasaki, Appl. Phys. Lett. 64, 1377 ~1994!. ibid. 66, 3492 ~1995!.
6
W. Fang and S. L. Chuang, Appl. Phys. Lett. 67, 751 ~1995!. 23
B. Gil, O. Briot, and R.-L. Aulombard, Phys. Rev. B 52, R17 028
7
G. E. Pikus, Zh. Éksp. Teor. Fiz. 14, 1075 ~1962! @Sov. Phys. ~1995!.
JETP 14, 898 ~1962!#. 24
E. O. Kane, J. Phys. Chem. Solids 1, 249 ~1957!.
8
G. L. Bir and G. E. Pikus, Symmetry and Strain-Induced Effects 25
J. M. Luttinger and W. Kohn, Phys. Rev. 97, 869 ~1955!.
in Semiconductor ~Wiley, New York, 1974!. 26
S. L. Chuang, Physics of Optoelectronic Devices ~Wiley, New
9
W. Y. Ching and B. N. Harmon, Phys. Rev. B 34, 5305 ~1986!. York, 1995!.
10 27
C. Y. Yeh, Z. W. Lu, S. Froyen, and A. Zunger, Phys. Rev. B 46, R. Dingle, D. D. Sell, S. E. Stokowski, and M. Ilegems, Phys.
10 086 ~1992!. Rev. B 4, 1211 ~1971!.
11
B. J. Min, C. T. Chan, and K. M. Ho, Phys. Rev. B 45, 1159 28
P. Löwdin, J. Phys. Chem. 19, 1396 ~1951!.
~1992!. 29
A. E. H. Love, A Treatise on the Mathematical Theory of Elas-
12
A. Rubio, J. L. Corkill, M. L. Cohen, E. L. Shirley, and S. G. ticity ~Dover, New York, 1944!, pp. 151–160.
Louie, Phys. Rev. B 48, 11 810 ~1993!. 30
C. Y.-P. Chao and S. L. Chuang, Phys. Rev. B 46, 4110 ~1992!.
13
M. E. Christensen and I. Gorczyca, Phys. Rev. B 50, 4397 ~1994!. 31
Physics of Group IV Elements and III-V Compounds, edited by K.
14
M. Suzuki, T. Uenoyama, and A. Yanase, Phys. Rev. B 52, 8132 H. Hellwege and O. Madelung, Landolt-Börnstein New Series,
~1995!. Group III, Vol. 17, Pt. a ~Springer-Verlag, Berlin, 1982!.
15
S. Kamiyama, K. Ohnaka, M. Suzuki, and T. Uenoyama, Jpn. J. 32
Properties of Group III Nitrides, edited by J. H. Edgar ~INSPEC,
Appl. Phys. 34, L821 ~1995!. IEE, London, 1994!.
16
T. Uenoyama and M. Suzuki, Appl. Phys. Lett. 67, 2527 ~1995!. 33
S. Strite and H. Morkoç, J. Vac. Sci. Technol. B 10, 1237 ~1992!.