SCIENTIFIC ABSTRACT We propose to use ISOPHOT to observe a sample of 31 ultraluminous galaxies. These objects have been selected from the QMW IRAS Galaxy Catalogue (QIGC, MNRAS 253, 485) fulfilling the criteria to have IR luminosities greater than 1.E+12 L<sun> and to be brighter than 3 Jy at 60 micron. In addition we will observe the most luminous IR galaxy known, F10214+4724 with 3.E+14 L<sun> (Nature 351). The ultraluminous galaxies are central to the debate about the nature of the sources that generate this enormous power. Using ten ISOPHOT passbands, we will define the spectral energy distributions of our sample in the range 10-200 micron. The increased knowledge of the spectral energy distributions provided by these observations will permit an examination for active nuclei (hidden quasars), which, based on lower-luminosity archetypes, could produce an emission component peaking in the 10-30 micron region. Hidden populations of young stars should be evident at longer wavelengths. These data will provide the basis for detailed radiative transfer calculations which will describe the environments of the power sources, and they will also permit estimates to be made of the total dust mass contributing to various parts of the spectral energy distribution, particlularly the virtually unexplored region between 120 an 240 micron. OBSERVATION SUMMARY The ten passbands are P_10, P_11.5, P_16, P_25, C_60, C_90, C_120, C_135, C_180 and C_200. The ratio I(P_10)/I(P_11.5) will tell about the potential existence of a deep 9.7 micron absorption feature. All passbands from 11.5 to 200 micron will be used to assess the IR energy distribution. For the P band measurements the 52 arcsec aperture will be used which is best matched to the total size of most objects in order to measure their integral flux. The P band measurements will be chopped using rectangular chopper mode due to the low contrast of source and background. The steep increase of the source flux towards longer wavelengths makes PHT-C measurements in chopped mode unfeasible. Instead they are performed in sparse map mode with one pointing on the target and a second pointing on an off-position in order to obtain the background contribution. Two sparse maps have to be performed per target, one each for the C100 and C200 filter passbands. The epected source fluxes between 60 and 200 micron are well above the confusion limits (except for F10214+4724, see below) so that no special measurement provisions are necessary. In order to measure the complete spectrum of each source a CONCATENATED sequence of the AOTs P03-P37-P39-P37-P39 is chosen. Exposure times will be 64 sec (32 sec on-source) for the three shortest wavelengths and 32 sec (16 sec on-source) for the remaining ones. At 10 and 11.5 micron fluxes of 40 mJy can be measured with S/N = 3-5. For the longer wavelengths calculations of the photon noise give always sufficiently high S/N ratios of 10-1000, however longer exposure times have been selected to account for detector drift effects. Because F10214+4724 is significantly fainter than all the other objects (190 mJy at 60 um) the PHT-C array measurements will be done in a different manner in order to suppress a possible cirrus confusion. Using AOT P32 a linear scan on a 4 times 1 raster will be performed in each filter. Because multi-filter maps can only be performed with the same detector, a combination of C100 and C200 filters yields a concatenated sequence of P03-P32-P32 for this object. The integration time per raster point has been taken as 65 sec for C100 (13 chopper steps) and 49 sec for C200 (7 chopper steps).