- Group contribution method
A group contribution method is a technique to estimate and predict thermodynamic and other properties from molecular structures.
In today's chemical processes many hundreds of thousands components are used either as raw material, as intermediates, and as products. The
Chemical Abstracts Serviceregistry currently lists 33 million substances but many of these components are only of pure scientific interest.
For an optimal process layout and control it is necessary to know at least some basic chemical properties of the pure components and their mixtures. Experimental work is expensive and for mixtures containing many components it is almost impossible to get a complete overview over properties at different compositions because of the enourmous amount of needed measurements. But even for pure components it is in many cases impossible to get reliable data.
Predictive methods and especially methods based on group contributions can replace measurements if they are giving precise and reliable estimations. The estimated properties will always be not as precise as well-made measurements but for many purposes during process simulation, process synthesis, and process development the quality of estimated properties is sufficient. Since even measurements can give wrong results predictive methods can be used to check the results of experimental work.
A group contribution method uses the principle that some simple aspects of the structures of chemical components are always the same in many different molecules. The smallest common constituents are the atoms and the bonds. All organic component for example are built of
carbon, hydrogen, oxygen, nitrogen, halogens, and maybe sulfur. Together with a single, a double, and a triple bond there are only ten atom types and three bond types to build thousands, of components. The next slightly more complex building blocks of components are functional groups which are themselves built of few atoms and bonds.
A group contribution method is used to predict properties of pure components and mixtures by using group or atom properties. This reduces the number of needed data dramatically. Instead of needing to know the properties of thousands or millions components only data for a few dozens or few hundreds groups have to be known.
Additive group contribution method
The simplest form of a group contribution method is the determination of a component property by summing up the group contribution.
This simple form assumes that the property (normal boiling point in the example) is strictly linear dependent from the number of groups and additionally no interaction between groups and molecules are assumed. This simple approach is used for example in the
Joback methodfor some properties and it works well in a limited range of components and property ranges but leads to quite large errors if used outside the applicable ranges.
Additive group contributions and correlations
This technique uses the pure additive group contributions to correlate the wanted property with an easy accessible property. This is often done for the
critical temperature, where the Guldberg ruleimplies that Tc is 3/2 of the normal boiling point and the group contributions are used to give a more precise value than 3/2.
This approach often gives better results than pure additive equations because the relation with a known property introduces some knowledge about the molecule. Commonly used additional properties are the molecular weight, the number of atoms, chain length, and ring sizes and counts.
For the prediction of mixture properties it is in most cases not sufficient to use a purely additive method. Instead the property is determined from group interaction parameters.
where P stands for property and Gij for group interaction value.
A typical group contribution method using group interaction values is the
UNIFACmethod which estimates activity coefficients. A big disadvantage of the group interaction model is the need for many more model parameters. Where a simple additive model only needs ten parameters for ten groups a group interaction model needs already needs 45 parameters. Therefore a group interaction model has normally not parameter for all possible combinations.
Group contributions of higher orders
Some newer methods [Constantinou L., Gani R., "New Group Contribution Method for Estimating Properties of Pure Compounds", AIChE J., 40(10), 1697-1710, 1994] introduce second-order groups. The second-order order groups can be super-groups containing several first-order (standard) groups. This allows the introduction of new parameters for the position of groups. Another possibility is to modify first-order group contributions if specific other groups are also present [Nannoolal Y., Rarey J., Ramjugernath J., "Estimation of pure component properties Part 2. Estimation of critical property data by group contribution", Fluid Phase Equilib., 252(1-2), 1-27, 2007] .
Determination of group contributions
Group contributions are obtained from known experimental data of well defined pure components and mixtures. Common sources are thermophysical data banks like the
Dortmund Data Bank, Beilstein database, or the DIPPR data bank (from AIChE). The given pure component and mixture properties are then assigned to the groups by statistical correlations like e. g. (multi-)linear regression.
Important steps during the development of a new method are the
# evaluation of the quality of available experimental data, elimination of wrong data, finding of outliers
# construction of groups
# searching additional simple and easily accessible properties that can be used to correlate the sum of group contributions with the examined property
# finding a good but simple mathematical equation for the relation of the group contribution sum with the wanted property. The critical pressures, for example, is often determined as Pc=f(ΣGi2)
# fitting the group contribution
The reliability of a method mainly relies on a comprehensive data bank where sufficient source data have been available for all groups. A small data base may lead to a precise reproduction of the used data but will lead to significant errors when the model is used for the prediction of other systems.
Wikimedia Foundation. 2010.