3. In the absence of interferents, the precision and detection limits of SWS and DWS are normally comparable. DWS may, however, be more precise than SWS where the latter measurement has substantial associated sample presentation error.In the presence of interferents, DWS precision is markedly reduced because of increase of independent and square root errors. In particular, dark current noise may become significant under these conditions. 5. The use of very small wavelength differences, as in differential spectroscopy, results in poor measurement precision.While the above conclusions are generally valid, it must be stressed that the different functional dependences of errors of different types on AI and D lead to a complex overall dependence which will vary with individual instruments and measurements. T o some extent, therefore, it is appropriate to consider each instrument and each type of measurement 4.as a special case.The timedependent supply and removal of sample atoms have been measured separately for two commercial graphite thermal atomizers. I n either type the release of the atoms from the graphlte wall Is determlned by the wall temperature and descrlbed by an Arrhenlus-type rate constant. With common heatlng rates, the equivalent tlme constant Is about 1 8. The removal of the atoms from the cell depends on the type of atomizer used. I n the Varian Techtron mlnlfurnace, dlffuslon Is the domlnatlng process and the equlvalent time constant is less than 70 ms. I n the larger Perkln-EIPner furnace operated under flow conditions, convection leads to equally small time constants. I f , however, the argon flow is slopped completely, dlffudon and, to some extent, expansion raise the time constant to about 1 s. Theoretical analysis and experimental measurement show that only In the latter case some 25% of the sample can be contained in the cell. Under common operating conditions, this efficlency is less than 10%.

No abstract