Class separations

In contrast to trace impurity removal, the use of adsorption for bulk separation in the liquid phase on a commercial scale is a relatively recent development. The first commercial operation occurred in 1964 with the advent of the UOP Molex process for recovery of high purity / -paraffins (6—8). Since that time, bulk adsorptive separation of liquids has been used to solve a broad range of problems, including individual isomer separations and class separations. The commercial availability of synthetic molecular sieves and ion-exchange resins and the development of novel process concepts have been the two significant factors in the success of these processes. This article is devoted mainly to the theory and operation of these Hquid-phase bulk adsorptive separation processes. 

Fig. 6. Fiber length distribution for (a) a long sample (group 4) and (b) a short sample (group 7) of chrysotile successive length classes separated by 50 p.m.

A first distinction which is often made is that between methods focusing on discrimination and those that are directed towards modelling classes. Most methods explicitly or implicitly try to find a boundary between classes. Some methods such as linear discriminant analysis (LDA, Sections 33.2.2 and 33.2.3) are designed to find explicit boundaries between classes while the k-nearest neighbours (A -NN, Section 33.2.4) method does this implicitly. Methods such as SIMCA (Section 33.2.7) put the emphasis more on similarity within a class than on discrimination between classes. Such methods are sometimes called disjoint class modelling methods. While the discrimination oriented methods build models based on all the classes concerned in the discrimination, the disjoint class modelling methods model each class separately. 

UNEQ is applied only when the number of variables is relatively low. For more variables, one does not work with the original variables, but rather with latent variables. A latent variable model is built for each class separately. The best known such method is SIMCA. 

As explained in Section 33.2.1, one can prefer to consider each class separately and to perform outlier tests to decide whether a new object belongs to a certain class or not. The earliest approaches, introduced in chemometrics, were called SIMCA (soft independent modelling of class analogy) [27] and UNEQ [28]. 

Figure 8.14 Senipreparative class separation of a diesel engine exhaust sample. Column 25 cm x 7.9 mm, lO micrometer Porasil. Solvent gradient hexane to 5% methylene chloride over 5 min., linear gradient to 100% methylene chloride over 25 min., isocratic for 10 min., linear gradient to 100% acetonitrile over 10 min., Isocratic for 5 min., step change to tetrahydrofuran for 10 min.,...

This technique was used by Delmas et al. [404] to separate lipid extracts in seawater into various classes. Lipid classes that have been eluted away from the point of application may be burnt off the rod in a partial scan, allowing those lipids remaining near the origin to be developed into the place that has just been simultaneously scanned and reactivated. By analysis of complex mixtures of neutral lipids in this stepwise manner it is possible to be more selective about lipid class separations as well as to be more confident about assigning identities to peaks obtained from a seawater sample. In addition, this approach also reduces the possibility of peak contamination by impurities which would normally coelute with marine lipid classes (e.g., phthalate esters [403]). 

The SMB process was invented by Broughton in 1961 and developed by Universal Oil Products under the general name Sorbex . Initially used for separating n-paraffins in bulk, it is now used for a variety of individual-isomer separations and class separations, and is currently attracting considerable interest for separating pharmaceutical enantiomers. The SMB process is described in Section 17.9.4 and in a growing literaturel-21 22>11 - 74),... 

SEC can be used to accomplish a class separation in which one component of the sample elutes in either excluded volume or permeation volume we term this application as group fractionation. Alternatively, SEC can be used to resolve two or more species within the included volume (e.g., between V0 and VP). We term this application simply as fractionation. 

Although the development of a SIMCA model can be rather cumbersome, because it involves the development and optimization of J PCA models, the SIMCA method has several distinct advantages over other classification methods. First, it can be more robust in cases where the different classes involve discretely different analytical responses, or where the class responses are not linearly separable. Second, the treatment of each class separately allows SIMCA to better handle cases where the within-class variance structure is... 

Fig. 2. A plot of the two largest principal components developed from all of the features in the dataset does not show class separation. When principal components are developed from the features that contain information about the classes, sample clustering on the basis of class is evident in a principal component plot of the data.

Since dioxiranes are electrophilic oxidants, heteroatom functionalities with lone pair electrons are among the most reactive substrates towards oxidation. Among such nucleophilic heteroatom-type substrates, those that contain a nitrogen, sulfur or phosphorus atom, or a C=X functionality (where X is N or S), have been most extensively employed, mainly in view of the usefulness of the resulting oxidation products. Some less studied heteroatoms include oxygen, selenium, halogen and the metal centers in organometallic compounds. These transformations are summarized in Scheme 10. We shall present the substrate classes separately, since the heteroatom oxidation is quite substrate-dependent. 

Quadratic discriminant analysis (QDA) is a probabilistic parametric classification technique which represents an evolution of EDA for nonlinear class separations. Also QDA, like EDA, is based on the hypothesis that the probability density distributions are multivariate normal but, in this case, the dispersion is not the same for all of the categories. It follows that the categories differ for the position of their centroid and also for the variance-covariance matrix (different location and dispersion), as it is represented in Fig. 2.16A. Consequently, the ellipses of different categories differ not only for their position in the plane but also for eccentricity and axis orientation (Geisser, 1964). By coimecting the intersection points of each couple of corresponding ellipses (at the same Mahalanobis distance from the respective centroids), a parabolic delimiter is identified (see Fig. 2.16B). The name quadratic discriminant analysis is derived from this feature. 

We will consider five chemical classes separately ... 

There are many factors to consider in selecting a column for a specific separation. Table 3.3 lists a variety of applications along with columns that have been found suitable for each analysis. While each of these columns has been used successfully for the class of compounds shown, they will not separate all compounds of that class. Separation of isomers can be difficult and special columns are often required. However, the columns listed are useful for the first step in the analytical problem. 

Ligands such as aniline (an), 1,2-diaminobenzene (dab, o-phenylenediamine) and 2,2 -dia-minobiphenyl886 887 are classed separately, not because their ability to bind to a central metal is any less than the ligands discussed previously, but because of their potential non-innocent behavior888 with respect to internal redox reactions. Indeed, the dark blue complex isolated from the air oxidation of Con/dab in aqueous ethanol (a conventional route to yellow Co(diamine)3+ systems) has been shown to have structure (117) with five-coordinate Co11.888 Related diimine complexes have been reported for Ni11889 as well as the conventional Ni(dab)(+,890 Co(dab)3+ 891 and Pt(dab) + 892 systems. 

Table D1.6.1 Solvent Systems for Lipid Class Separation on an Iatroscan TLC-FID...

Although use of a constant humidity chamber significantly improves the reproducibility of lipid class separation by Chromarods, minor... 

Most HPLC is based on the use of so-called normal-phase columns (useful for class separations), reverse-phase columns (useful for homolog separations), and polar columns (used in either the normal- or reverse-phase mode). Since reverse-phase HPLC columns are generally easier to work with, almost all authors use high-performance reverse-phase liquid chromatography with octade-cyl chemically bonded silica as the stationary phase and nonaqueous solvents as mobile phases (so-called NARP, or nonaqueous reverse-phase chromatography). 

Table 4 Recent HPLC Phospholipid Class Separation Methods Using an Acetonitrile-Based Mobile Phase...

WS Letter. A rapid method for phospholipid class separation by HPLC using an evaporative light scattering detector. J Liq Chrom 15 253-266, 1992. 

AA Karlsson, P Michelsen, A Larsen, G Odham. Normal-phase liquid chromatography class separation and species determination of phospholipids utilizing electrospray mass spectrometry/tandem mass spectrometry. Rapid Communications Mass Spectrom 10 775-780, 1996. 

McKay and Latham (1981) have also determined compound class distributions in the high-boiling distillates and the residua for four crude oils. As shown in Table VIII, the content of heteroatom compounds increases with increasing boiling point. The 675°C+ residuum may have nearly 10 times the acids, bases, or neutral Lewis (pyrrolic and amides unreactive to column resins) bases compared to the VGO portion (370-535°C). Grizzle et al. (1981) have also employed compound class separations and have observed similar trends. McKay and Latham (1981) calculated that each acid, base, or neutral nitrogen molecule in the <675°C residuum contains three to five heteroatoms. 

However, there are numerous theoretical models available in the literature that attempt to derive meaningful values from thermal data. Solution phase reactions are generally easier to model than solid-state or heterogeneous reactions and the discussion that follows will consider the two classes separately. In all cases, it is necessary to know the time at which the reaction was initiated, t0, in order to analyse the data correctly. Note that this does not mean the reaction must be initiated directly in the instrument—this is difficult for ampoules prepared on a bench-top—it means that the time axis on the resulting power-time curve must be corrected for the delay caused by loading. No fitting model requires the reaction to run to completion in order to return the correct reaction parameters—if this were the case it would take up to 10,000 years to model some reactions based on the sensitivity of the instrument ... 

---