The emission spectra of Cr

4+

in 57.5%AlO

1.5

-35%CaO-7.5%BaO glass were measured at different temperature from 15K to 300K. The energy levels of Cr

4+

in this host fall into the weak field regime on the Tanabe Sugano diagram

the lowest excited state is

3

T

2

and the emission spectrum is a broad band (Fig.2). The spectrum at 15K (Fig. 1) can be approximated by a Pekarian

[5]

g

(

E

zp

-

p

tω)=

e

-S

S

P

/p!

By fitting the spectrum with this function we have got

E

zp

=8400±13cm

-1

ω=320±5cm

-1

S= 3.58±0.08. Strictly speaking

the spectrum should be the convolution of the phonon sideband and an inhomogeneous lineshape

which is usually expressed by a Gaussian. By calculating the convolution and comparing with Fig.1

we estimated that the inhomogeneous line width is within 300~400cm

-1

(FWHM). The emission band widths obtained from Fig.2 are shown in Fig.3 as solid cycles. According to the single frequency approximation

its temperature dependence would be

[5]

D

=

D

0

[1+

e

-tω/

kT

/1-

e

-tω/

kT

]

1/2

=

D

0

coth

1/2

[tω/

kT

]

where

D

0

is the band width at 0K. With the parameters obtained above

D

should be as shown by the solid line in Fig.3 (labeled by ωV/ωU=1). Evidently

it does not agree with the experimental result. The temperature dependence of the width of the phonon sideband originates from the thermal excitation of higher vibronic states. Transition from these states covers larger energy range. If the high vibrational states coupled to the excited electronic state are less populated

the temperature dependence of the bandshape would be weaker. This will happen if the phonon coupled to the excited electronic state possesses energy larger than that coupled to the ground state. Thus it makes us apply the model where the ground state and the excited state couple unequal frequency phonons

[6]

. Keeping

S

E

zp

and ηω

U

obtained from the single frequency approximation fitting and taking ω

V

/ω

U

as a new parameter

we calculated the temperature dependence of the emission bandwidths. The calculated results are shown in Fig.3

ω

V

/ω

U

=1.15 results in the best fitting.