Thin–layer Chromatography (TLC)

Parameter	High–performance TLC	TLC
Plate dimensions (cm2)	10 × 10	20 × 20
Layer thickness (mm)	0.1 or 0.2	0.1–0.25
Starting spot diameter (mm)	1–2	3–6
Diameter of separated spots (mm)	2–6	6–15
Solvent front migration distance (cm)	3–6	10–15
Time for development (capillary flow) (min)	3–20	20–200
Detection limitsa Absorption (ng)	0.1–0.5	1–5
Detection limits Fluorescence (pg)	5–10	50–100
Nominal particle size range (μm)	3–7	5–20
Apparent particle size (μm)b	5–7	8–10
Minimum plate height (μm)	22–25	35–45
Optimum velocity (mm/s)	0.3–0.5	0.2–0.5
Porosity Total	0.65–0.70	0.65–0.75
Porosity Interparticle	0.35–0.45	0.35–0.45
Porosity Intraparticle	0.28	0.28

In the basic TLC experiment, the sample is applied to the layer as a spot or band near to the bottom edge of the layer. The separation is carried out in a closed chamber by either contacting the bottom edge of the layer with the mobile phase, which advances through the layer by capillary forces, or the mobile phase is forced to move through the layer at a controlled velocity by an external pressure source or centrifugal force. A separation of the sample results from the different rates of migration of the sample components in the direction travelled by the mobile phase. After development and evaporation of the mobile phase, the sample components are separated in space, their position and quantity being determined by visual evaluation or in situ scanning densitometry aided by the formation of easily detected derivatives by post–chromatographic chemical reactions, as required.

Separations by column liquid chromatography (HPLC) and TLC occur by essentially the same physical process. The two techniques are frequently considered as competitors, when it would be more realistic to consider them as complementary. The attributes of TLC that provide for its co–existence as a complementary technique to HPLC are summarised in Table 2. Based on these attributes, TLC methods are most effective for the low–cost analysis of a large number of samples (e.g. drug screening in biological fluids and tissues, determination of the botanical origin and potency of traditional herbal medicines, stability testing and content uniformity testing), for the rapid analysis of samples that require minimum sample clean up or where TLC allows a reduction in the number of sample preparation steps (e.g. analysis of samples containing components that remain sorbed to the stationary phase or contain suspended microparticles). TLC is also preferred for the analysis of substances with poor detection characteristics that require post–chromatographic chemical treatment for detection. In other cases, HPLC methods are generally preferred, particularly if a large number of theoretical plates are necessary for a separation, for separations by size–exclusion and ion–exchange chromatography, and for trace analysis using selective detectors unavailable for TLC.

Attribute	Application
Separation of samples in parallel	Low–cost analysis and high–throughput screening of samples requiring minimal sample preparation
Disposable stationary phase	Analysis of crude samples (minimising sample preparation requirements)
Analysis of a single or small number of samples when their composition and/or matrix properties are unknown
Analysis of samples containing components that remain sorbed to the separation medium or contain suspended microparticles
Static detection	Samples that require post–chromatographic treatment for detection
Samples that require sequential detection techniques (free of time constraints) for identification or confirmation
Storage device	Separations can be archived
Separations can be evaluated in different locations or at different times
Convenient fraction collection for coupled column–layer chromatography
Sample integrity	Total sample occupies the chromatogram, not just that portion of the sample that elutes from the column

Conventional TLC plates can be prepared in the laboratory by standardised methods, but reproducible layer preparation is easier to achieve in a manufacturing setting and few laboratories prepare their own plates today. Precoated plates for high performance, conventional and preparative TLC are available in a range of sizes and different layer thickness, supported on glass, aluminium or plastic backing sheets. To impart the desired mechanical stability and abrasion resistance to the layer a binder, such as poly(vinyl alcohol), poly(vinyl pyrrolidone), gypsum or starch in amounts from 0.1 to 10% (w/w) is incorporated into the layer. An ultraviolet (UV)-indicator, such as manganese–activated zinc silicate of a similar particle size to the sorbent, may be added to the layer to visualise separated samples by fluorescence quenching. TLC plates with a narrow preadsorbent zone located along one edge of the layer are available to aid manual sample application.

Silica gel is the most important stationary phase for TLC, with other inorganic oxide adsorbents, such as alumina, kieselguhr (a silica gel of low surface area) and Florisil (a synthetic magnesium silicate), of minor importance. Most silica gel sorbents have an average pore size of 6 nm and are designed for the separation of small molecules (relative molecular mass < 700). The chromatographic properties of the inorganic oxide adsorbents depend on their surface chemistry and specific surface area. For silica gel, silanol groups are the dominant adsorption sites. The complementary sample properties that govern retention are the number and type of functional groups and their spatial location (Fig 3). The influence of functional group properties on selectivity is illustrated in Fig 2 for the separation of ethynyl steroids. The steroids with phenolic groups are the most strongly retained, followed by hydroxyl groups, and then ketone and ester groups. Subtle separation differences through steric hindrance at a functional group and differences in ring conformations are also seen, which allow the separation of steroids with very similar chemical properties.

Chemically bonded layers are prepared from silica gel by reaction with various organosilane reagents to form siloxane bonds, with some of the silanol groups present on the silica surface (Table 3). Reversed–phase alkylsiloxane–bonded layers with a high level of surface bonding cannot be used with mobile phases that contain a significant amount of water (>30% v/v) because of the inadequate mobile–phase velocity generated by capillary forces. Water compatibility for alkylsiloxane–bonded layers is achieved by increasing the particle size, using a reproducible although lower degree of silanisation, and by using modified binders. These layers are referred to as water wettable and are used for all types of reversed–phase separations, while layers with a high degree of silanisation are used predominantly with non–aqueous mobile phases. Alkylsiloxane bonded phases are used primarily (but not exclusively) for the separation of water–soluble polar drugs and weak acids and bases after ion suppression (buffered mobile phase) or ion–pair formation. Water compatibility is not a problem for polar chemically bonded phases, which can be used for both normal- and reversed–phase separations. For separations that cannot be achieved on silica gel, the polar chemically bonded phases are the most widely used stationary phases. The 3–aminopropylsiloxane–bonded layers can function as a weak anion exchanger for the separation of polyanions with a buffered mobile phase. Cellulose layers provide only weak retention of common drug substances and are used primarily to separate very polar compounds in biochemistry.

Type of modification	Functional group	Applicationa
Alkylsiloxane	Si–CH3	For reversed phase separations generally, but not exclusively
	Si–C2H5	Separation of water–soluble polar organic compounds (RPC)
	Si–C8H17	Weak acids and bases after ion suppression (RPC)
	Si–C18H37	Strong acids and bases by ion–pair mechanism (RPC)
Phenylsiloxane	Si–C6H5	Of limited use for drug analysis
Cyanopropylsiloxane	Si–(CH2)3CN	Useful for both RPC and NPC
		In NPC it exhibits properties similar to a low–capacity silica gel.
		In RPC it exhibits properties similar to short–chain alkylsiloxane–bonded layers (it has no selectivity for dipole–type interactions)
Aminopropylsiloxane	Si–(CH2)3NH2	Used mainly in NPC and IEC; limited retention in RPC
		Selectively retains compounds by hydrogen–bond interactions in NPC; separation order generally different to that in silica gel
		Functions as a weak anion exchanger in acidic mobile phases (IEC)
Spacer bonded propanediol	Si–(CH2)3OCH2 Si-CH(OH)CH2OH	Used in NPC and RPC, but more useful for NPC because of low retention in RPC
		Polar drugs selectively retained by hydrogen bond and dipole–type interactions in NPC; more hydrogen–bond acidic and less hydrogen–bond basic than aminopropylsiloxane–bonded layers in NPC; more retentive than aminopropylsiloxane–bonded layers in RPC
		Similar retention to short–chain alkylsiloxane–bonded layers, but different selectivity for hydrogen–bonding drugs

Table 3. Retention properties of silica based chemically bonded layers

TLC has found limited use for the separation of enantiomers. The most widely used approach employs ligand–exchange chromatography on reversed–phase layers impregnated with a solution of copper acetate and (2S,4R,2RS)-N-(2–hydroxydodecyl)-4–hydroxyproline. Separations result from stability differences in diastereomeric complexes formed between the drug, copper and the proline selector. Suitable drugs for this application require an amino acid or α-hydroxycarboxylic acid group for complex formation. A more versatile approach to the separation of enantiomeric drug substances by reversed–phase TLC is the use of chiral selectors, such as cyclodextrins or bovine serum albumin, as mobile–phase additives.

The technique of TLC involves a number of separate steps, namely preparing the layer, applying the sample, developing the plate and detecting the separated zones. These steps are described below.

Prior to chromatography it is common practice to prepare the layers for use by any or all of the following steps: washing, conditioning and equilibration. Layers may also be cut to preferred sizes using scissors for plastic- or aluminium–backed plates and diamond or carbide glass–cutting tools for glass–backed plates. Newly consigned precoated layers are invariably contaminated, or quickly become so, because of residual contaminants from the manufacturing process, contact with packaging materials and adsorption of materials from the atmosphere. To remove contaminants, single or double immersion in a polar solvent, such as methanol or propan–2–ol, for about 5 min is generally superior to predevelopment with the mobile phase. For trace analysis, sequential immersion and predevelopment may be required to obtain the best results.

For inorganic oxide adsorbents the absolute R_F (see later) value and the reproducibility of R_F values depend on the layer activity. The latter is controlled by the adsorption of reagents, most notably water, through the gas phase. Physically adsorbed water can be removed from silica gel layers by heating at about 120° for 30 min. Afterwards, the plates are stored in a grease–free desiccator over blue silica gel. Heat activation is not normally required for chemically bonded layers. Equilibration of activated layers by exposure to the atmosphere is extremely rapid and layer activation is at times an unnecessary step. In modern air–conditioned laboratories, layers achieve a consistent level of activity that should provide sufficient reproducibility for most separations. Inorganic oxide layers can be adjusted to a defined activity by exposure to a defined gas phase in an enclosed chamber. Since manipulation in the atmosphere almost certainly readjusts this activity, it is best performed after application of the sample zones in a developing chamber that allows both layer conditioning and development in the same chamber (e.g. a twin–trough chamber), or in a separate conditioning chamber immediately before development. Atmospheres of different constant relative humidity can be obtained by using solutions of concentrated sulfuric acid or saturated solutions of various salts. Acid or base deactivation can be carried out in a similar manner by exposure to, for example, ammonia or hydrochloric acid fumes.

Drugs are applied to TLC plates as spots or bands of minimum size with a homogeneous distribution of material within the starting zone. For high–performance layers, with desirable starting spot diameters of about 1.0 to 2.0 mm, this corresponds to a sample volume of 100 to 200 nL if applied by a dosimeter (micropipette). For conventional TLC plates, sample volumes five- to ten–fold greater are acceptable. Desirable properties of the sample solution are summarised in Table 4. If scanning densitometry is used for detection, manual sample application with hand–held devices is inadequate. For densitometry, the starting position of each spot must be known accurately, which is achieved easily with mechanical devices that operate to a precise grid mechanism. Also, the sample must be applied to the layer without disturbing the surface, something that is nearly impossible to achieve using manual application.

Sample application devices for TLC encompass a wide range of sophistication and automation. The most popular devices for quantitative TLC use the spray–on technique. A controlled nitrogen–atomiser sprays the sample from a syringe or capillary, to form narrow, homogeneous bands on the plate surface. The plate is moved back and forth under the atomiser on a translational stage to apply bands of any length between zero (spots) and the maximum transit length of the spray head. Bands are typically 0.5 or 1.0 cm in length, with the longer bands used primarily for preparative–scale separations. The rate of sample deposition is also adjustable to accommodate sample solutions of different volatility and viscosity. An advantage of spray–on devices is that different volumes of a single standard solution can be applied for calibration purposes and the standard addition method of quantification is carried out easily by overspraying the sample already applied to the layer with a solution of the standard. Fully automated sample applicators can be programmed to select samples from a rack of vials and deposit fixed volumes of the sample, at a controlled rate, to selected positions on the plate. The applicator automatically rinses itself between sample applications and can spot or band a whole plate with different samples and standards without operator intervention.

Glass microcapillaries for conventional TLC and fixed–volume dosimeters (which consist of a 100 or 200 nL platinum–iridium capillary sealed into a glass–support capillary) for high–performance TLC are also commonly used for sample application and require less sophisticated instrumentation. The capillary tip is brought into contact with the plate surface using a mechanical device to discharge its volume. A click–stop grid mechanism is used to provide an even spacing of the samples on the layer and a frame of reference for sample location during scanning densitometry.

Layers with a preadsorbent zone (a narrow zone prepared from a silica gel of low surface area with weak retention) simplify some aspects of sample application. This allows relatively large sample volumes or dirty samples to be applied to the preadsorption zone and their reconcentration to a narrow band at the interface between the preadsorbent and separation zones by a short development prior to chromatography. However, since the distribution of the sample may not be even within the band, the quantitative accuracy of densitometric measurements may be lowered using this approach.

The principal development techniques in TLC are linear, circular and anticircular, with the velocity of the mobile phase controlled by capillary forces or forced–flow conditions. In any of these modes continuous or multiple development can be used to extend the application range. Radial development is used rarely for drug analysis and is not considered further. Forced–flow development requires sophisticated equipment not commonly found in analytical laboratories, and is not described here.

For linear (or normal) development, samples are applied along one edge of the plate and the separation developed for a fixed distance in the direction of the opposite edge. Viewed in the direction of development, the chromatogram consists of a series of compact symmetrical spots of increasing diameter or, if samples are applied as bands, in rectangular zones of increasing width.

In continuous development the mobile phase is allowed to traverse the layer under the influence of capillary forces until it reaches some predetermined position on the plate, at which point it is evaporated continuously. Evaporation of the mobile phase usually occurs at the plate atmospheric boundary using either natural or forced evaporation. Continuous development is used primarily to separate simple mixtures with a short development length and a weaker (more selective solvent) than employed for normal development.

In unidimensional multiple development, the TLC plate is developed for some selected distance, then either the layer or the mobile phase is withdrawn from the developing chamber, and adsorbed solvent evaporated from the layer before repeating the development process. The principal methods of unidimensional multiple development are summarised in Table 5. Multiple development provides a very versatile strategy for separating complex mixtures, since the primary experimental variables of development distance and composition of the mobile phase can be changed at any development step, and the number of steps varied to obtain the desired separation. Multiple development provides a higher resolution of complex mixtures than does normal or continuous development, can easily handle samples of a wide polarity range (stepwise gradient development) and, because the separated zones are usually more compact, leads to lower detection limits. Equipment for automated multiple development is commercially available.

For drug mixtures that span a wide retention range, some form of gradient development is required to separate all the components in either a single chromatogram or in separate chromatograms for successive developments. Continuous solvent–composition gradients, as commonly employed in HPLC, are used rarely in TLC. These require experimental conditions that are less convenient than those for step mobile–phase gradients. In addition, step gradients can be constructed easily to mimic a continuous linear gradient, with the added advantage that the zone refocusing effect can be employed to minimise zone broadening. Gradients of increasing solvent strength are used to fractionate complex mixtures by separating just a few components in each step. Individual drugs are usually identified and quantified at the intermediate steps at which the drugs of interest are separated. In this way, the zone capacity can be made much larger than predicted for a complete separation recorded as a single chromatogram. However, this approach is tedious when many components are of interest and it is difficult to automate. Alternatively, if incremental multiple development is used, the sample can be separated for the shortest distance in the strongest mobile phase, with each subsequent, longer development using mobile phases of decreasing solvent strength. This strategy is most useful when the final separation is to be recorded as a single chromatogram, but it is limited in zone capacity because all the components must be located between the sample origin and the final solvent front. The two approaches for exploiting solvent–strength gradients are thus complementary and selection is made based on the properties of the sample. The decreasing solvent–strength gradient approach is the operating basis of automated multiple–development chambers.

In two–dimensional TLC the sample is spotted at the corner of the layer and developed along one edge of the plate. The solvent is then evaporated, the plate rotated through 90° and redeveloped in the orthogonal direction. If the same solvent is used for both developments, the sample is redistributed along a line from the corner at which the plate was spotted to the corner diagonally opposite. In this case, only a small increase in resolution can be anticipated. The realisation of a more efficient separation system implies that the resolved sample should be distributed over the entire plate surface. This can be achieved only if the selectivity of the separation mechanism is complementary in the orthogonal directions. Using two solvent systems with complementary selectivity is the simplest approach to implement in practice, but it is often only partially successful. In many cases the two solvent systems differ only in their intensity for a given set of properties and are not truly orthogonal. Chemically bonded layers can be used in the reversed–phase and normal–phase modes, and they enable the use of additives and buffers as a further way to adjust selectivity. The acceptance of two–dimensional TLC for quantitative analysis, though, will depend on providing a convenient method for in situ detection and data analysis. It seems unlikely that two–dimensional development will be more widely used in TLC, except for qualitative analysis, until the problems of detection are solved.

The development process in TLC can be carried out in a variety of vessels that differ significantly in design and sophistication. For convenience these are often categorised under the headings of normal (N-chamber) and sandwich (S-chamber), and further subdivided according to whether the internal atmosphere is saturated (N_S or S_S) or unsaturated (N_U or S_U). Sandwich chambers have a depth of gas phase in front of the layer of less than 3 mm, with other chamber designs indicated as normal chambers. Saturation of the vapour phase is achieved by using solvent–saturated pads or filter papers as a chamber lining.

The twin–trough chamber is the most popular of the simplest TLC developing chambers. It consists of a standard rectangular developing tank with a raised, wedge–shaped bottom. The wedged bottom divides the tank into two compartments, so that it is possible to either develop two plates simultaneously or to use one compartment to condition the layer prior to development. The horizontal developing chamber (Fig 4) can be used in either the normal or sandwich configuration for either conventional edge–to–edge or simultaneous edge–to–centre development. Starting the development simultaneously from opposite edges allows the number of samples separated to be doubled in the same time. The sandwich configuration of the horizontal developing chamber is not suitable for mobile phases that contain volatile acids, bases or large amounts of volatile polar solvents, such as methanol or acetonitrile, because of the restricted access of the saturated vapour phase to the dry portion of the separation layer.

The automated developing chamber increases laboratory productivity and improves the reproducibility of separations by providing precise control of layer conditioning, mobile–phase composition, solvent–front migration distance and drying conditions. This chamber can be used in the normal or sandwich configuration with all the operational features preselected on a microprocessor–based control unit and monitored by sensor technology.

The automated multiple–development chamber (Fig 5) provides the necessary conditions and control for automated separations by incremental multiple development with a decreasing solvent–strength gradient. The operating parameters of layer conditioning, solvent–front migration distance, mobile–phase composition and drying time for each development, and the total number of developments for the separation, are entered into the computer–based control unit. The complete separation sequence is carried out without further intervention. Each development is typically 3 to 5 mm longer than the previous one and, depending on the complexity of the desired mobile phase gradient, a total of 10 to 30 developments are used, which requires 1.5 to 4.5 h for completion.

About 1 to 10 μg of coloured substances with a quantitative reproducibility rarely better than 10–30% can be detected by visual inspection of a TLC plate. This may be adequate for qualitative methods, but for reliable quantification in situ spectrophotometric methods are preferred, as they are more accurate and far less tedious and time consuming than excising zones from the layer for determination by conventional solution spectrophotometry. The fluorescence–quenching technique enables visualisation of UV-absorbing drugs on TLC plates that incorporate a fluorescent indicator. The zones of UV-absorbing substance appear dark against the brightly fluorescing background of a lighter colour when the plate is exposed to UV light of short wavelength. The method is not universal, since it requires overlap between the absorption bands of the indicator (γ_max ≈ 280 nm with virtually no absorption below 240 nm) and the drug, but in favourable cases it is a valuable and non–destructive method for zone location.

All optical methods for the quantitative in situ evaluation of TLC chromatograms are based upon measuring the difference in optical response between a sample–free region of the layer and regions of the layer in which separated substances are present. Reflectance measurements can be made at any wavelength from the UV to the near infrared (185 to 2500 nm). The relationship between signal and sample amount in the absorption mode is non–linear, and does not conform to any simple equation. The principal method of quantification in TLC is by calibration using a series of standards that span the concentration range of the drug to be determined. The calibration curve is usually based on a second–order polynomial fit for the calibration standards, with individual samples quantified by interpolation only.

The determination of drugs that fluoresce on TLC plates is fundamentally different to absorption measurements. At low sample concentrations the fluorescence signal F is described adequately by F = φI₀εbC, where φ is the quantum yield, I₀ the intensity of the excitation source, ε the molar absorption coefficient, b the thickness of the TLC layer and C the sample amount. With the exception of the sample amount all terms in this expression are constant, or fixed by the experiment, and therefore the fluorescence emission is linearly dependent on the sample amount over two or three orders of magnitude.

The development technique is selected based on the number of detectable components in the mixture and their polarity range (Table 7). A single development with capillary–controlled flow may be too difficult or impossible for mixtures that contain more than eight to ten components of interest. In addition, if the range of polarities is too wide, multiple development techniques using mobile phase gradients are necessary. It is only necessary to separate the components of major interest from each other and from the less important components, which need not be separated individually. Method development is easier if standards for the relevant compounds are available. Standards simplify zone tracking and enable detection characteristics and the possibility of spectroscopic resolution of incompletely separated zones to be established. Standards are also required for calibration, if quantification is required, and to construct spectral libraries for identification purposes. The expected concentration range of relevant compounds may indicate the need for derivatisation to obtain the required detection limits and to increase zone separation of neighbouring compounds if one compound is a minor component with similar migration properties to a major component.

Development method	Dimensions	Zone capacity
Predictions from theory
Capillary–controlled flow	1	<25
Forced flow	1	<80 (up to 150 depending on pressure limit)
Capillary–controlled flow	2	<400
Forced flow	2	Several thousand
Based on experimental observations
Capillary–controlled flow	1	12–14
Forced flow	1	30–40
Capillary–controlled flow (AMD)	1	30–40
Capillary–controlled flow	2	About 100

Table 7. Zone capacity calculated or predicted for different conditions in TLC

A general guide to the selection of stationary phases for TLC separations is summarised in Fig 6. Silica gel is generally the first choice to separate drugs of low molecular weight that are soluble in moderately polar organic solvents. Reversed–phase chromatography on chemically bonded layers is generally used to separate drugs that are difficult to separate on silica gel because of inadequate retention, inadequate selectivity or zone asymmetry. Ionic compounds and easily ionised compounds are separated frequently by reversed–phase chromatography using buffered mobile phases (weak acids and bases) or ion–pair reagents (strong acids and bases). Only a limited number of stationary phases are available for ion–exchange chromatography, which is not a widely used separation mechanism in TLC.

Since the solvent used for the separation is evaporated prior to detection, a wider range of UV-absorbing solvents are commonly used in TLC than is the case for HPLC. Solvents must be of high purity, since involatile impurities and stabilisers remain sorbed to the layer that causes problems in the detection step. Multicomponent mobile phases can produce a mobile–phase gradient in the direction of development through demixing. If demixing is complete, zones with sharp boundaries are formed, which separate the chromatogram into sections of different solvent composition and, therefore, selectivity. Demixing effects are less apparent when saturated developing chambers are used. These considerations hinder optimisation strategies based on the composition of the mobile phases as popularised in HPLC.

The selection of a mobile phase to separate simple mixtures need not be difficult and can be arrived at quickly by guided trial and error methods. A solvent of the correct strength for a unidimensional development migrates the sample components into the R_F range 0.2 to 0.8, or thereabouts and, if of the correct selectivity, distributes the sample components evenly throughout this range. Solvent systems can be screened in parallel using several development chambers, as prescribed in the PRISMA model. To select suitable mobile phases, the first experiments are carried out on TLC plates in unsaturated chambers with ten solvents, chosen from the different selectivity groups, indicated by bold type in Table 8. After these screening experiments with single solvents, the solvent strength is either reduced or increased so that the substance zones are distributed in the R_F range 0.2 to 0.8. If the substances migrate into the upper third of the plate, the solvent strength is reduced by dilution with hexane (the strength–adjusting solvent). If the substances remain in the lower third of the plate with the single solvents, the solvent strength is increased by the addition of a strong solvent, such as water or acetic acid. A similar procedure is followed in the reversed–phase mode, except that solvent selection is limited to water–miscible solvents and water is used as the strength–adjusting solvent. From these trial experiments, those solvents that show the best separation are selected for further optimisation in the second part of the model.

Selectivity group	Solvent
I	n-Butyl ether, diisopropyl ether, methyl t-butyl ether, diethyl ether
II	n-Butanol, propan–2–ol, propanol, ethanol, methanol
III	Tetrahydrofuran, pyridine, methoxyethanol, dimethylformamide
IV	Acetic acid, formamide
V	Dichloromethane, 1,1–dichloroethane
VI	Ethyl acetate, methyl ethyl ketone, dioxane, acetone, acetonitrile
VII	Toluene, benzene, nitrobenzene
VIII	Chloroform, dodecafluoroheptanol, water

Table 8. Solvent–strength parameters and selectivity groups for solvents used for separations on silica gel

Between two and five solvents can be selected to construct the PRISMA model for solvent optimisation. Modifiers required to maintain an acceptable zone shape, such as acids and ion–pair reagents, can be added in a low and constant concentration, so that their influence on solvent strength can be neglected. The PRISMA model (Fig 7) is a three–dimensional geometrical design that correlates solvent strength with selectivity of the mobile phase. The model consists of three parts: the base or platform (represents the modifier), the regular part of the prism with congruent base and top surfaces, and the irregular truncated top prism (frustum). The lengths of the edges of the prism (S_A, S_B, S_C) correspond to the solvent strengths of the neat solvents (A, B, C). Since the selected solvents usually have different solvent strengths, the lengths of the edges of the prism are generally unequal and the top plane of the prism is not parallel and congruous with its base. If the prism is cut parallel to its base at the height of the lowest edge (determined by the solvent strength of the weakest solvent, solvent C in Fig 7), the lower part gives a regular prism, and the top and any planes, which represent weaker solvents diluted with a strength–adjusting solvent, are parallel equilateral triangles. The upper frustum of the model is used for mobile–phase optimisation of polar drugs in normal phase TLC, while the regular part is used to separate moderately polar drugs in normal–phase TLC and all separations by reversed–phase TLC.

For polar compounds, optimisation is always started on the top irregular triangle of the model, either within the triangle, when three solvents are selected, or along one side, for binary mobile phases. Any solvent composition on the face of the triangle can be represented by a three–co–ordinate selectivity point (P_S), each co–ordinate corresponding to the volume fraction of the solvent at that position on the triangle (Fig 7). Optimisation is commenced by selecting solvent combinations that correspond to the centre point P_S = 333 and three other points close to the apexes of the triangle P_S = 811, 181 and 118. If the separation obtained is insufficient, other selectivity points are tested around the solvent combination that gave the best separation. On changing the selectivity points on the top triangle the solvent strength changes as well, especially when the solvent strengths of the solvents used to construct the prism are significantly different. The solvent strength should be adjusted with the strength–adjusting solvent as required to maintain the separation in the optimum R_F range. Failure to obtain the beginning of a separation requires that a new prism be constructed, using a different solvent for at least one of the edges.

For reversed–phase TLC, the solvation–parameter model provides a convenient computer–aided approach to method development. Suitable water–miscible solvents with a range of selectivity include methanol, propan–2–ol, 2,2,2–trifluoroethanol, acetonitrile (or dioxane), acetone (or tetrahydrofuran) and dimethylformamide (or pyridine). For optimisation of systems (stationary phases and binary mobile phases), preliminary results in the form of system maps (a continuous plot of the system constants against mobile–phase composition) are required. System maps are a permanent record of the system properties used in all calculations and are available for most common layers and indicated solvents for selectivity optimisation. For each computer–simulated separation a retention map is calculated from the system map and displays the computed R_F values as a continuous function of the binary mobile–phase composition. A typical retention map for the computer–predicted separation of analgesics on an octadecylsiloxane–bonded layer with 2,2,2–trifluoroethanol–water mixtures as the mobile phase is shown in Fig 8. Solvent compositions that result in an acceptable zone separation are identified easily by visual inspection. Computer simulation of retention maps allows those systems (defined as a combination of stationary and mobile phase) likely to provide an acceptable separation to be identified before experimental work commences. The agreement between model predicted and experimental R_F values is generally good, typically better than 0.05 R_F units. A mixture–design approach is used to extend this method to ternary solvent mixtures.

For drug mixtures of a wide polarity range, stepwise changes in solvent composition are required to achieve a satisfactory TLC separation. Models to calculate migration distances using incremental multiple development with increasing and decreasing solvent–strength gradients have been described, but are complicated and not widely used. Optimised gradients for automated multiple development are usually arrived at by more pragmatic means. Methods based on a universal gradient commence with methanol, end with hexane and use either dichloromethane or methyl t-butyl ether as the intermediate solvent for separations on silica gel. By scaling and superimposing the chromatogram of the separation above the theoretical gradient profile, those regions of the chromatogram that affect the separation are identified easily. The solvent composition for the initial and final development steps is adjusted to eliminate those portions of the gradient that do not contribute to the separation. The gradient shape is modified to enhance resolution in those regions of the chromatogram that are separated poorly or to minimise regions devoid of sample zones. For moderately complex mixtures this approach is often satisfactory. If, after the above adjustments, the separation is inadequate it is necessary to identify a more selective solvent for problem regions in the gradient. The PRISMA model can be used at this point to identify more selective solvents to incorporate into the gradient as a replacement for the initial, terminal or base solvent.

Preparative TLC is used mainly to purify drugs or to isolate drug metabolites and impurities in amounts of about 1 to 100 mg for subsequent use as reference materials, structural elucidation, biological activity evaluation and other purposes. Scale up from analytical TLC is achieved by increasing the thickness of the layer (loading capacity increases with the square root of the layer thickness) and by increasing the plate length used for sample application. Precoated TLC plates for preparative chromatography vary in size from 20 cm × 20 cm to 20 cm × 40 cm and are coated with 0.5 to 10 mm thick layers, with the most popular thickness being 1.0 to 2.0 mm. As the average particle size (≈ 25 μm) and size distribution (5 to 40 μm) are larger for preparative layers, and as sample overload conditions are used commonly in preparative chromatography, invariably inferior separations in a longer time (≈ 1 to 2 h) are obtained compared with analytical separations. Resolution can be increased significantly by using wedge–shaped, gradient–thickness layers. These layers have a uniform increase in thickness from 0.3 mm at the bottom to 1.7 mm at the top. Sample bands are focused during migration by the negative mobile–phase velocity gradient created by the layer geometry.

Sample application is a critical step in preparative TLC, and if performed improperly can destroy all or part of the separation. The sample, usually as a 5 to 10 % (w/v) solution in a volatile solvent, is applied as a band along one edge of the layer to give a maximum sample load of about 5 mg/cm for each millimetre of layer thickness. Sample loads are usually lower for difficult separations and for cellulose and chemically bonded layers. Any of the automated band applicators for analytical TLC are suitable for sample application in preparative TLC. Manual sample application by syringe or glass pipette must be performed carefully to avoid damaging the layer and producing irregularly shaped migrating zones. A short predevelopment, of about 1 cm with a strong solvent, is often useful to refocus manually applied bands. Preparative layers with a preadsorbent zone are useful for manual sample application, since the focusing mechanism can be used to correct for poor sample–application technique. In all cases, it is important that the sample solvent is evaporated fully from the layer prior to the start of the separation to avoid the formation of distorted separation zones. It is usual to leave a blank margin of 2 to 3 cm at each vertical edge of the layer to avoid uneven development.

Most of the changes in preparative TLC over the past decade have occurred in the method of development. Conventionally, ascending development in large–volume tanks that hold a number of preparative layers in a rack is used commonly. In laboratories that perform preparative TLC on a regular basis, higher resolution and shorter development times are achieved by using forced–flow development or rotation planar chromatography (accelerated development using centrifugal force). These methods allow conventional development and elution with on–line detection and automated fraction collection to be used.

After development, physical methods of zone detection are used to identify the sample bands of interest. Layers that contain a UV indicator for fluorescence quenching or the adsorption of iodine vapours are useful for this purpose. If a reactive spray reagent is used for visualisation, it should be sprayed on a small strip of the chromatogram only, so as not to contaminate the remainder of the material. Once the bands of interest are located, the zones are scraped off the plate carefully with a spatula or similar tool. A number of devices based on the vacuum–suction principle for removing the marked zones from the plate are available also. Soxhlet extraction, liquid extraction or solvent elution with a polar solvent is used to recover drugs from the sorbent. For solvent extraction, water is often added to dampen the silica gel prior to extraction with a water–immiscible organic solvent. Chloroform and ethanol (methanol is less suitable because of its higher silica solubility) are widely used for solvent elution. Colloidal silica can be removed by membrane filtration prior to vacuum stripping of the solvent.

The retardation factor, or R_F value, is the fundamental parameter used to characterise the position of a sample zone in a TLC chromatogram. For linear development it represents the ratio of the distance migrated by the sample compared to the distance travelled by the solvent front:

R _F = Z_X/(Z_f − Z₀)

where Z_X is the distance travelled by the sample from its origin, (Z_f − Z₀) the distance travelled by the mobile phase from the sample origin, Z_f the distance travelled by the mobile phase measured from the mobile phase level at the start of the separation, and Z₀ the distance from the sample origin to the mobile phase level at the start of the separation. The boundary conditions for R_F values are 1 ≥ R_F ≥ 0. The R_F value is generally calculated to two decimal places. Some authors prefer to tabulate values as whole numbers, as hR_F values equivalent to 100R_F.

Drug identification

Database searches are used in systematic toxicological analysis to identify suspect substances in biological fluids and post–mortem tissue samples. Extracted samples are separated in one or more standard TLC systems. The corrected R_F values, often combined with the results of sequential post–chromatographic colour reactions, are then entered into the search program. The input data are automatically compared against a database of reference drugs, common metabolites, natural contaminants, etc., for identification. A number of chemometric procedures can be used for data analysis, but the most common approach is based on the mean list method.

It is assumed that the errors in individual measurements are random and can be described by a standard deviation. The precision of the separation system can then be described as the mean of the standard deviation of all substances separated in the system, called the system mean standard deviation. This allows a confidence interval or window to be assigned to the system as some multiple (typically three) of the system mean standard deviation. Each R_F value in the system database that appears in the window could be confused with the original substance. The number of substances identified as above is called the list length. Repeating the process for all R_F values in that system and averaging the individual list lengths provides the mean list length for that system. The mean list length indicates, on average, the number of substances in the database that qualify as candidates for the identification of a single drug. The shorter the mean list length, the greater the information potential of the system. Combining the results from additional retention parameters in complementary standard separation systems, colour reactions, spectroscopic data, etc., minimises the mean list length to the point that only a small number of candidate compounds for the unknown are indicated. More specific tests can then be used to identify the unknown from among the small number of indicated possibilities. For systematic drug identification in forensic toxicology, commonly two or more complementary TLC systems combined with the results from several in situ sequential colour reactions are used. For drugs of toxicological interest, a mean list length from two to ten is possible.

Systematic toxicological analysis takes advantage of the separation of an unknown substance in standard TLC systems (or other chromatographic systems) to establish the probable identity of the substance by reference to a database of candidate compounds using a statistical comparison approach, such as the mean list method. Suitable chromatographic techniques for systematic toxicological analysis must meet the following criteria:

TLC systems that meet these requirements are described below.

Since pH-dependent extractions are customarily used in drug extraction and work–up procedures, generally different TLC systems are used to separate acidic and basic drugs, with neutral drugs likely to occur in both fractions. The Committee for Systematic Toxicological Analysis of the International Association of Forensic Toxicologists (TIAFT) recommended 11 separation systems for drug identification (Table 9). Four systems (1 to 4a) are to separate neutral and acidic drugs and seven systems (4b to 10) are to separate neutral and basic drugs. Reference data are presented for about 1600 toxicologically relevant substances. For general drug screens, the use of two separation systems with a low correlation is recommended: systems 2 and 4(a) for neutral and acidic drugs and systems 5 and 8 for neutral and basic drugs (systems 7 and 8 are nearly as good). Combining colour reactions with the TLC data improves the certainty of identification significantly. Four colour reactions are carried out on the same plate in sequence. After each step the colour is noted and encoded by means of a colour chart (1, yellow; 2, orange; 3, brown; 4, red; 5, purple; 6, black; 7, blue; 8, green; 0, no spot observed). The sequence consists of formaldehyde vapour and Mandelin’s reagent, water, fluorescence under 366 nm irradiation and modified Dragendorff’s reagent. Other sequential colour reactions can be encoded and utilised in the same way. The Merck Tox Screening System (MTSS) contains the TIAFTTLC database and several other useful tools for searches using other chromatographic and spectroscopic databases and user–created databases.

Table 9. TLC systems recommended by TIAFT for systematic toxicological analysis (drug database, de Zeeuw et al. 1992)

Romano et al. (1994) presented data for 443 drugs in four TLC systems using high–performance silica gel TLC plates (Table 10).

TLC system
No	Mobile phase	Chamber type	Stationary phase	Reference compounds	hRcF*
(1)	Ethyl acetate–methanol– 30% ammonia (17:2:3)	Saturated	Silica gel	Morphine Strychnine Aminopyrine Cocaine	25 44 70 85
(2)	Cyclohexane– toluene– diethylamine (13:5:2)	Saturated	Silica gel	Clobazam Aminopyrine Mebeverine Amitriptyline	15 29 47 60
(3)	Ethyl acetate– chloroform (1:1)	Saturated	Silica gel	Caffeine Ketamine Flunitrazepam Prazepam	09 24 44 61
(4)	Acetone	Saturated	Silica gel impregnated with 0.1 mol/L KOH in methanol and dried	Imipramine Pericyazine Aminopyrine Lidocaine	20 37 62 78

Table 10. The separation systems recommended by Romano et al. (1994) for systematic toxicological analysis by TLC (drug database: Romano et al. 1994)

These systems use slight modifications of the mobile–phase compositions recommended by TIAFT. The UniTox system uses three TLC systems (Table 11). System 1 is designed to separate neutral and acidic drugs and systems 2 and 3 to separate basic, amphoteric and quaternary drugs. Two of the separation systems are based on reversed–phase separations designed to complement the more familiar silica gel separations. The database contains over 375 drugs of general toxicological interest, including a large number of amphetamines.

The Toxi-Lab system is a TLC kit for toxicological drug screening; it contains equipment for extraction, development, detection and identification. Separations are performed on unsupported, particle–embedded glass–fibre sheets with holes punched in them to receive samples and standards as extraction or reference disks. A combination of silica gel and reversed–phase separations together with sequential colour reactions is used for identification and confirmation purposes. The database is designed for computer searches with results entered in a standard format.

Pesticides are a further class of toxic substances of interest to systematic toxicological analysis because of their general availability, toxicity and potential confusion with drugs. Erdmann et al. (1990) developed a database for 170 commonly used pesticides separated in three standardised TLC systems (Table 27.12). The systems in Table 27.12 supplement those in Table 27.9, in which many common pesticides migrate with the solvent front. Systems 1 and 2 are recommended for general screening and system 3 for the identification of special compounds not distinguished in the first two systems.

TLC system
No	Mobile phase	Chamber type	Stationary phase	Reference compounds	hRcF
(1)	Hexane–acetone (4:1)	Saturated	Silica gel	Triazophos	21
Parathion–methyl	30
Pirimiphos–methyl	49
Quintozene	84
(2)	Toluene–acetone (19:1)	Saturated	Silica gel	Carbofuran	20
Azinophos–methyl	42
Methidathion	56
Parathion–ethyl	85
(3)	Chloroform–acetone (1:1)	Saturated	Silica gel	Nicotine	11
Ioxynil	39
PCP	60
Methabenzthiazuron	85

Thousands of general and validated methods are available for the determination of drugs as pharmaceutical products and in biological fluids. Since the zone capacity of TLC systems is small, there are no general methods for drugs as a class, but there are a large number of methods for individual drugs defined by therapeutic or chemical categories. These still represent substantial diversity driven by the need to optimise selectivity for each group of substances taken for analysis. This information can provide a useful starting point for system selection, but is no general substitute for systematic method development. For these reasons universal methods for general drug analysis do not exist and earlier attempts at systematised approaches for different drug categories have failed to keep pace with the growth in number of drugs in those categories. In addition, systems recommended for the separation of individual drug categories rarely prove optimal for the separation of individual drugs and their impurities or metabolites.

The TLC systems given below are general screening methods for nitrogenous bases (Systems TA, TB, TC, TL, TAE and TAF), for acids and neutral compounds (Systems TD, TE, TF and TAD) with a further three general screening methods (Systems TAJ, TAK and TAL). Furthermore, seven systems specific for pesticides (Systems TW, TX, TY, TZ, TAA, TAB and TAC) and another 19 systems covering specific groups of drugs are also listed. The drugs are divided into chemical or pharmacological groups, but some other drugs are included with certain groups if they are chemically similar and would be extracted with that group.

There may not be one best system for a particular separation and a number of systems can be applied from those suggested. However, each of the systems described has been selected because it gives a good spread of Rf values, has high reproducibility, and has low correlation with the other systems selected for that group of drugs. These systems have proved useful for a large number of groups of drugs over the years and are robust and dependable. At least three systems are given for each group, where possible. The Rf values of the reference compounds suggested for the general screening systems have been derived using solutions of approximately 2 mg/mL of each substance.

Fluorescent plates should always be used and the absorption or fluorescence of the drug under ultraviolet light (both 254 and 350 nm) should be used as a location procedure. The suggested locating agents include general ones to visualise any drug that might be present as well as more specific ones to pick out individual classes of drugs.

Molecule	TA	TB	TC	TE	TAE
Buclosamide	90	2	67	–	90
Chlorphenesin	82	–	–	62	–
Clotrimazole	–	–	–	76	80
Diamthazole	52	30	30	–	–
Econazole	80	9	61	75	78
Fenticlor	–	–	–	–	91
Fluconazole	–	–	–	35	67
Flucytosine	–	–	–	9	57
Griseofulvin	–	–	–	69	78
Hydroxystilbamidine	1	–	–	–	–
Itraconazole	–	1	–	79	87
Ketoconazole	–	–	–	50	68
Miconazole	73	11	67	80	77
Thioacetazone	78	–	–	–	–

Antibacterials

The tabulated systems, previously described, may be used together with the associated location reagents.

Antineoplastics

The tabulated systems, previously described, may be used together with the associated location reagents.