No abstract

No abstract

Recent developments in the field of gas-liquid chromatography have made it possible to measure thermodynamic properties of solution with a high degree of reliability.Nowadays it is possible to evaluate and eliminate1 the contribution of interfacial adsorption to the net retention volume, enabling one to isolate the role of the solute-solvent interactions in retention without interference from concentration effects at the liquid-solid or liquid-vapor interfaces. Information contained in the resulting corrected partition coefficient promised to be extremely useful in the study of intermolecular interactions.The importance of these developments is not that such a partition coefficient has not been previously available, but that the glc approach allows its measurement with an 'Alternatively, the interfacial effects may be made the object of study, see, e.g., Karger, et al. (9).

No abstract

Propagation of uncertainty from a set of measurements to a derived quantity is an important topic whose fundamental ideas have been competently addressed in this Journal (1-4). However, there is a pitfall associated with the uncritical use of the standard formulas promulgated in many texts that has not been emphasized to date. It is readily illustrated with the data collected in a common undergraduate physical chemistry experiment, and is the subject of this note.We use spreadsheets nowadays in much of our data workup. Students learn early on how to use the data regression tool therein, and we try to instill in them the significance and importance of the estimates of uncertainty that are provided-for example, for the slope, m, and intercept, b, of the common y = mx + b correlation. These should be examined for consistency with the students' own estimates of uncertainties in the measured variables, and they of course provide essential information for the propagation of uncertainty to derived quantities. A familiar example of the latter is the enthalpy of vaporization derived from the slope of a plot of the logarithm of vapor pressure versus inverse temperature.One of the standard experiments, in fact, in our first physical chemistry lab course is the measurement of the vapor pressure of water as a function of temperature (see, e.g., ref 5). Students adjust the pressure above the boiling liquid and wait for the temperature to stabilize. They then measure the pressure with a mercury meter-stick manometer and the temperature with a thermistor capable of 0.02 K precision. Ten to 12 data pairs are collected between 170 and 750 torr.Pollnow (6) has done a beautiful job of illustrating the importance, in proper optimization of parameters, of the weighting of pressures when the dependence of vapor pressure on temperature is linearized. However, this is a subtlety unjustified by the quality of the data in question here, and our students are instructed to do a simple least squares regression of ln P vs. 1/T:The spreadsheet supplies them with estimates of the "best fit" parameters m and b as well as their uncertainties. They use the information for m to estimate the enthalpy of vaporization at the midpoint of the temperature range of the data, and its uncertainty. This calculation is almost always satisfactory.Students are told also to use this equation to predict the normal boiling point of water by substituting 760 torr for P and solving for T, and to estimate the uncertainty in T by propagation of the uncertainties in m and b. They obtain very respectable values for the boiling point, but their estimates of uncertainty are unreasonably large (1 to 2 K) if they use the familiar "propagation of error" formula σ T 2 = (∂T/ ∂m) 2 σ m 2 + (∂T/ ∂b) 2 σ b 2 where σ represents standard deviation.How can it be that in a series of measurements characterized by roughly 0.05 K uncertainty in temperature (the uncertainty in P contributes less than 0.04 K to the uncertainty in boiling point in our experiment), the estimated uncertainty...