Valence force field simulations utilizing large supercells are used to inve
stigate the bond lengths in wurtzite and zinc-blende InxGa1-xN and AlxGa1-x
N random alloys. We find that (i) while the first-neighbor cation-anion she
ll is split into two distinct values in both wurtzite and zinc-blende alloy
s (RGa-N1 not equal RIn-N1), the second-neighbor cation-anion bonds are equ
al (RGa-N2 = RIn-N2). The second-neighbor cation-anion bonds exhibit a cruc
ial difference between wurtzite and zinc-blende binary structures: in wurtz
ite we find two bond distances which differ in length by 13% while in the z
inc-blende structure there is only one bond length. This splitting is prese
rved in the alloy, and acts as a fingerprint, distinguishing the wurtzite f
rom the zinc-blende structure. (iii) The small splitting of the first-neigh
bor cation-anion bonds in the wurtzite structure due to nonideal c/a ratio
is preserved in the alloy, but is obscured by the bond length broadening. (
iv) The cation-cation bond lengths exhibit three distinct values in the all
oy (Ga-Ga, Ga-In, and In-In), while the anion-anion bonds are split into tw
o values corresponding to N-Ga-N and N-In-N. (v) The cation-related splitti
ng of the bonds and alloy broadening are considerably larger in InGaN alloy
than in AlGaN alloy due to larger mismatch between the binary compounds. (
vi) The calculated first-neighbor cation-anion and cation-cation bond lengt
hs in InxGa1-xN alloy are in good agreement with the available experimental
data. The remaining bond lengths are provided as predictions. In particula
r, the predicted splitting for the second-neighbor cation-anion bonds in th
e wurtzite structure awaits experimental testing. (C) 1999 American Institu
te of Physics. [S0021-8979(99)09601-2].