An original technique for the determination of the fraction of unsuspended solid particles at agitation speeds smaller than NJ, , is proposed. The technique is based on the "twin-systems" concept and is completed by a suitable tracing and detection technique for the solid phase. Results obtained with a fully baffled tank stirred by a Rushton turbine impeller show that the fraction of solids never leaving the tank bottom can be simply correlated to the ratio between agitation speed and Zwietering's complete suspension speed. It is shown that at agitator speeds of about 80% of Zwietering's complete suspension speed, practically all particles but a few, are already suspended. This finding implies that large energy and installation cost savings can be achieved, with respect to current design practices, without significantly affecting the solid-liquid mixing performance.On propose une technique originale pour la determination de Keywords: solids suspension, stirred tanks, partial suspension regime, twin systems.olids suspension in stirred vessels is an operation fre-S quently encountered in the process industry. Relevant examples include multiphase catalytic reactions, adsorption, crystallisation, dissolution, leaching and precipitation. As a consequence solids suspension in stirred vessel has been the subject of a number of investigations, most of which concern experimental methods and predictive correlations for the minimum agitation speed, Njs, required to "just suspend" all the particles, on the basis that this condition corresponds to the lowest specific power input at which all particles surface area is made available for effective processing.Several experimental methods were devised, that were claimed to allow the determination of NjS (Bourne and Sharma, 1974a; Buurman et al., 1986; Rao et al., 1988;Rieger and Ditl, 1994), but the most widely adopted method is still the visual method introduced by Zwietering (1958), by which the just complete suspension speed is defined as the agitator speed at which no solid particle rests on the bottom of the vessel for more than I s. This method clearly requires a transparent bottomed vessel, a feature easily available at laboratory scale but impracticable for large scale installations. The main advantage of this method is its intrinsic simplicity; but it must be pointed out that it is affected by subjective evaluations, so that only with a careful and skilled observation is it possible to get about +5% reproducibility at best in the case of dilute suspensions. No data can be reliably reproduced when solids concentration is above 8% by weight (Oldshue and Sharma, 1993).Zwietering (1958) also proposed a well known heuristic correlation for the estimation of Njs, on the basis of his own data and with the help of dimensional considerations. Similar *Author to whom correspondence should be addressed. E-mail address: abrucato@dicpm.unipa.it correlations were proposed by other authors (Nienow 1968) which were based in some cases on phenomenological models of the suspension phen...