Following an earlier study of the uncertainties for defining thicknesses by angle-resolved XPS, one of the major instrumental uncertainties has been evaluated that limits both precision and accuracy. For analysis of the thicknesses of SiO 2 on Si, certain angles of emission have been recommended, but an error of 1• in these angles leads directly to an error in the thickness of ∼1% from this contribution alone. This is significant since the total uncertainty required in the (International technology roadmap for semiconductors) ITRS is only 1%. In many instruments, to reduce sample-stage vibration and for other reasons, the angle and other setting adjustments are engineered with backlash. This, combined with the manufacturer's tolerances, can lead to angular errors above 1 • . We report here a device using a laser and reflectors, fixed to the sample mount, that allow the angle of emission to be set to a precision better than 0.1• and, furthermore, a method to set the zero angle of emission to 0.1 • . Using this geometrical device as well as by measurements of intensities in XPS, it is deduced that the data in our earlier report for the CCQM (Consultative Committee for Amount of Substance) intercomparison were for angles 1.89 • ± 0.15• too high. Consequently, by a re-analysis of all of that data, we find that our recommended attenuation length data for the Si 2p photoelectrons in thermal SiO 2 using Mg or Al Ka X-rays should be increased to 2.996 nm and 3.485 nm, respectively, an increase of 1.2% on the originally calculated values. These values now have standard uncertainties of 0.54% instead of the 20% of the TPP-2M calculations. This leads to an improved accuracy in the measurement of ultrathin thermal oxides on silicon by XPS, of better than 1% for thicknesses greater than 1.5 nm but less than 8 nm. Crown