In our efforts to synthesise large quantities of the nucleoside analogue LNA,^1,2 we regularly purify quantities of up to 100 g of highly functionalised nucleoside and carbohydrate derivatives. Conventional silica gel column chromatography techniques including Flash Column Chromatography³ afford poor separation of amounts exceeding 10 g demanding purification of the reaction mixtures after division into several minor batches. This is unfeasible due to the large quantities of silica gel, solvents and time required.

In our search for a simple and efficient method for large scale separation of complex mixtures, we re-discovered and further developed a technique first described by L. M. Harwood called “Dry Column” Flash Chromatography.⁴ This technique has mainly been used for small to medium scale purifications (< 2 g) with separation power rivalling that of analytical TLC. We have now adapted this technique to large scale chromatography (over 10 g compound mixture) without losing the excellent resolving power observed with smaller amounts on smaller columns. We have prepared and eluted large columns (diameter 10–13 cm) with up to 100 g of complex mixtures in 1–2 h depending on the number and size of the fractions collected, the efficiency of the pump and the kind of silica type used. We have chosen to name this improved version of Harwood’s technique Dry Column Vacuum Chromatography (DCVC) since “Flash” is misleading in this context.⁵

The general experimental setup suggested by the inventor^4,6 and others^7,8 consists of glassware found in any organic chemistry laboratory. However, the setup is not practical as it is necessary to disassemble it for each fraction (Figure 1), which is tedious and time consuming.

The setup is significantly improved by using an ordinary separatory funnel,⁹ an adapter with a sidearm and a three-way stopcock attached on the sidearm allowing one to let air into the funnel without turning the pump off (Figure 2a). The only drawback with this setup is that with large sintered glass funnels (diameter over 10 cm) it tends to get very high and as a consequence impractical. We have solved this problem by designing a compact vacuum proof receiver (Figure 2b). Our adaption of the experimental setup has significantly reduced the time required for DCVC and provides a construction that is easy to handle for large as well as small scale chromatography.

The experimental procedure can be summarised as follows (see experimental section for detailed procedure):

1. A sintered glass funnel of the appropriate size¹⁰ is filled with approximately 6–7 cm of loose silica and tapped to give a level surface.
2. Vacuum is applied and the surface is pressed firmly to give a completely level, well compacted bed approximately 4.5–5.5 cm high.
3. The column is checked for voids and channels by pouring n-heptane onto the silica bed while applying vacuum. If the column is packed properly, the solvent will descend in a horizontal line.
4. The mixture to be separated is dissolved in an appropriate low boiling solvent (eg. ethyl acetate, methanol) and pre-adsorbed on silica.¹¹ The silica is added as a thin uniform layer on top of the column and step (2) is repeated.
5. The column is gradient eluted with a suitable solvent system and fractions are monitored by TLC.

Several problems can arise due to evaporation of solvents during chromatography. The most significant being difficulty in performing a controlled gradient elution and water condensing on the glassware. This can be prevented by avoiding volatile solvents such as diethyl ether, dichloromethane and pentane. Mixtures of n-heptane, ethyl acetate, and methanol are recommended as they are easy to handle and remove after separation has been accomplished.

Elution of compounds is usually accompanied by frothing on the underside of the sinter. As a rule of thumb, a compound will elute in the solvent mixture where the R_f value is ca. 0.5 on analytical TLC.

When chromatographing a mixture of compounds (approx. 20–50 g) with a ΔR_f ~0.05 (by TLC) on a sintered glass funnel with a diameter of 10 cm, we usually collect 100 mL fractions by gradient elution. Normally this yields compounds that are >95% pure by HPLC. If a better separation is desired a different solvent gradient and collection of smaller fractions usually improve the result.

In order to obtain good separation, the choice of the correct silica type is essential. Resolution is determined by particle size. As a rough guide, reduced particle size gives better resolution. On the other hand, too small particles will result in an extremely compact column that elutes very slowly. According to the inventor⁶ and others,⁷ TLC silica without the gypsum binder (5–25 µm) should be used for DCVC because it is cheaper than the silica used for regular chromatography (e.g. Flash Column Chromatography) and because “regular” silica is too free flowing to allow easy column packing. This is also true in our hands for Merck Silica Gel 60 (40–63 µm) used for Flash Column Chromatography because the normal packing procedure does not yield a sufficiently compact column. As a result, the solvent just pours through the columns affording very poor separation. If the column is packed using a high vacuum pump, compact columns with good separation can be prepared. However, this approach is undesirable for several reasons, most of all because of the long time required to prepare such a column.

If Merck Silica Gel 60 (15–40 µm) is used, the problems are avoided. We have experienced no problems when using Merck Silica Gel 60 (15–40 µm) instead of TLC silica (5–25 µm) and in our hands the same time is needed to prepare a column with either type of silica. Although it is difficult to reproduce the exact chromatography conditions, the data suggest that TLC silica (5–25 µm) without the gypsum binder offers a slightly better separation compared to Merck Silica Gel 60 (15–40 µm) as expected due to smaller particle size. TLC silica (5–25 µm) is about 10–15% cheaper than Merck Silica Gel 60 (15–40 µm). Furthermore, a column packed with TLC silica (5–25 µm) requires approximately 10–15% less silica by mass compared to Merck 60 (15–40 µm). However, a column packed with TLC silica becomes very compact and therefore significantly slower in eluting.

Our preferred silica is Merck Silica Gel 60 (15–40 µm) as it offers excellent resolution and gives considerably faster eluting columns than those packed with TLC silica.

A typical example of a compound mixture that was separated successfully is shown in Figure 3. An analytical TLC of the compound mixture in 10% n-heptane, 90% ethyl acetate (v/v) showed 4 compounds with R_f values: 0.45 (1), 0.25 (2), 0.20 (3) and 0.10 (4). The compounds were successfully separated and isolated on a 10 g scale in 1.5 hours including packing of the column, elution and evaporation of the solvent fractions containing the pure compounds. The column (diameter 10 cm, height 5 cm, Merck silica gel 60, 15–40 µm) was eluted with 0–100% ethyl acetate in n-heptane (v/v) with 5% increments in ethyl acetate concentration for each fraction collected (twenty-one 100 mL fractions). After isolation, the identity of the compounds depicted below was established by NMR. The desired compound 2 was isolated in 98+% purity by HPLC and ¹H-NMR.

In conclusion, DCVC provides a robust, fast and economical method for large as well as small scale separations. In addition, the technique is superior in resolution power to Flash Column Chromatography because of the reduced diffusion during chromatographic separation. This means that each compound elutes in fewer fractions and that less cross contamination of fractions is observed.

We have used Dry Column Vacuum Chromatography for purification of a wide range of complex mixtures comprising highly functionalised nucleosides and carbohydrates with excellent results. It is possible to perform DCVC using mixtures of nonchlorinated, nonvolatile solvents such as n-heptane, ethyl acetate and methanol. This and the greatly reduced amount of silica used compared to other techniques reduces the enviromental impact of the process. Separation of compounds with ΔR_f ~0.05 (by analytical TLC) should at least be expected, but with experience separations as efficient as those with analytical TLC can be achieved.

Chromatography equipment was assembled as depicted in Figure 2a or 2b. Analytical TLC was performed using Merck 5554 silica 60 aluminum sheets. The silica used for DCVC was Merck 15111 silica gel 60 (15–40 µm) and Merck 9385 silica gel 60 (40-63 µm) purchased from Merck Eurolab. TLC silica without binder (5-25 µm) was purchased from Sigma-Aldrich Denmark. Solvents were HPLC grade from LABSCAN. Vacuum was applied with a Vacuubrand MZ 2C diaphragm vacuum pump (1.7 m³/h, 10 mbar) instead of a water aspirator for enviromental reasons. Sintered glass funnels (porosity 3) with diameters up to 17 cm are commercially available.