L. Hedin et Jd. Lee, External losses in photoemission from strongly correlated quasi-two-dimensional solids - art. no. 115109, PHYS REV B, 6411(11), 2001, pp. 5109
Expressions are derived for photoemission, which allow experimental electro
n energy loss data to be used for estimating losses in photoemission. The d
erivation builds on new results for dielectric response and mean free paths
of strongly correlated systems of two-dimensional layers. Numerical evalua
tions are made for Bi2Sr2CaCu2O8 (Bi2212) by using a parametrized loss func
tion. The mean free path for Bi2212 is calculated and found to be substanti
ally larger than obtained by Norman et al. [Phys. Rev. B 59, 11 191 (1999)]
in a recent paper. The photocurrent is expressed as the convolution of the
intrinsic approximation for the current from a specific two-dimensional la
yer with an effective loss function. This effective loss function is the sa
me as the photocurrent from a core level stripped of the dipole matrix elem
ents. The observed current is the sum of such currents from the first few l
ayers. The correlation within one layer is considered as a purely two-dimen
sional (2D) problem separate from the embedding three-dimensional (3D) envi
ronment. When the contribution to the dielectric response from electrons mo
ving in 3D is taken as diagonal in q space, its effect is just to replace b
are Coulomb potentials in the (3D) coupling between the 2D layers with dyna
mically screened ones. The photoelectron from a specific CuO layer is found
to excite low-energy acoustic plasmon modes due to the coupling between th
e CuO layers. These modes give rise to an asymmetric power-law broadening o
f the photocurrent an isolated two-dimensional layer would have given. We d
efine an asymmetry index where a contribution from a Luttinger line shape i
s additive to the contribution from our broadening function. Already the lo
ss effect considered here gives broadening comparable to what is observed e
xperimentally. Our theory is not related to the loss mechanism recently dis
cussed by Joynt [R. Joynt, Science 284, 777 (1999); R. Haslinger and R. Joy
nt, J. Electron Spectrosc. Relat. Phenom. 117-118, 31 (2001)] which adds ad
ditional broadening beyond what we calculate. A superconductor with a gappe
d loss function is predicted to have a peak-dip-hump line shape similar to
what has been observed, and with the same qualitative behavior as predicted
in the recent work by Campuzano et al. [Phys. Rev. Lett. 83, 3709 (1999)].