Some protoplanetary disks show evidence of inner dust cavities .
Recent observations of gas and dust of these so-called transition disks support the hypothesis that these cavities originate from particle trapping in pressure bumps .
We present new Atacama Large Millimeter/submillimeter Array ( ALMA ) continuum observations at 336 GHz of two transition disks , SR21 and HD 135344B .
In combination with previous ALMA observations from Cycle 0 at 689 GHz , we compare the visibility profiles at the two frequencies and calculate the spectral index ( \alpha _ { mm } ) .
The observations of SR 21 show a clear shift in the visibility nulls , indicating radial variations of the inner edge of the cavity at the two wavelengths .
Notable radial variations of the spectral index are also detected for SR 21 with values of \alpha _ { mm } { \sim } 3.8 - 4.2 in the inner region ( r \lesssim 35 AU ) and \alpha _ { mm } { \sim } 2.6 - 3.0 outside .
An axisymmetric ring ( which we call the ring model ) or a ring with the addition of an azimuthal Gaussian profile , for mimicking a vortex structure ( which we call the vortex model ) , is assumed for fitting the disk morphology .
For SR 21 , the ring model better fits the emission at 336 GHz , conversely the vortex model better fits the 689 GHz emission .
For HD 135344B , neither a significant shift in the null of the visibilities nor radial variations of \alpha _ { mm } are detected .
Furthermore , for HD 135344B , the vortex model fits both frequencies better than the ring model .
However , the azimuthal extent of the vortex increases with wavelength , contrary to model predictions for particle trapping by anticyclonic vortices .
For both disks , the azimuthal variations of \alpha _ { mm } remain uncertain to confirm azimuthal trapping .
The comparison of the current data with a generic model of dust evolution that includes planet-disk interaction suggests that particles in the outer disk of SR 21 have grown to millimetre sizes and have accumulated in a radial pressure bump , whereas with the current resolution there is not clear evidence of radial trapping in HD 135344B , although it can not be excluded either .