TetrahedronLetters No. 21, pp. 22-26, [Link] Press Ltd.

Printed in
Great Rritain.

ON THE WRCRANISM OF THE FORMOSE FBACTION

Ronald Rreslow

Dept. of Chemistry,Columbia University,.NewYork, N. Y.

(Reoeived25 November 1959)

IN 1861 Butlerowl first noted the formation of sugar-likesubstanceswhen

formaldehydeis treated with mild base. Since that time numerous investi-

gations of this reaction have been undertaken,2and many mechenismihave

been suggested for the process, but to date no mechanism appears to have

been formulatedwhiah does not employ "active formaldebydewor some other

ill-definedspecies. It is the purpose of this communicationto aall

attention to the considerablebody of data available on this reaction

to point out a mechaniem which is consistentwith such data and which iS

also consistentwith chemical experience.

Although a heterogeneousmixture of products ia usually obtained on

treatment of formaldehydesolutionswith base, conditionshave been realized

in which as much as 5& of the original formaldehydehas been oonverted to

glyoolaldehyde,3 Furthermore,trioeea, tetroses,eto., aould easily arise

1
A. Butleroa, C. R. Aced. Soi., Paris 22, 145 (1861).
2
For a short review with leading referenoes,cf. W. Langenbeck,
Tetrahedron2, 185 (1958).
3 W, Langenbeck,Angen. Chem, 66, 151 (1954).

22
No.21 Cn the mechanism of the formose reaction 23

from aldol condensationsinvolvingglycolaldehydeas the active methylene

componentsnd formaldehyde,glycolaldehyde,etc., as the csrbonyl component.

For this reason it has long been obvious that the formation of glycolal-

dehyde from formaldehydeis the process to be explained,and we will not

further consider the subsequentcondensationreactions in this paper.

The reaction of two formaldehydemolecules by an "aldo condensation8V

mechanism is of course very unlikely, and the finding that there is a long

inductionperiod in the formose reaction4 suggests that if such a direct

process does occur it must be very slow. However, kinetic studies have

shown that after this induction period a rapid reaction ensues.4 This

suggests sn autocatalyticprocess, and this has been confirmedby the

finding that the addition of glyoolaldehyde,glyceraldehyde,or dihydroq-

acetone causes the inductionperiod to disappear, the reaction starting

immediatelybut at the ssme rate as that observed in the fast part of the

normal formcse reaction. This has led to the suggestionthat there are

two processes by which the formaldehydecondensationoccurs: (1) a slow

direct condensation,possibly involving a formyl anion, but forming

glyoolaldebydeat an indetectablerate. (2) a fast reaction, catalysedby

glycolaldehyde,.inwhich formaldehydecondenses to a second molecule of

glycolaldehyde.

The first reaotion is certainly interesting,but it is relatively

inacessiblefor study if it proceeds at an indetectablerate, and if the

rate is slow the reaction is not inconceivableas formulated,being related

4 Careful kinetic studies, with references to the earlier work, are

to other oases in whioh some moderate aoidity of hydrogens on a double bond

has been observed.5 The high rate of the second process strongly indioates,

however, that some more usual types of reaction are involved in its

mechanism. We wish to point out that, although minor modificationsin

details are possible, the following series of simple oondensationsand

tautomerizationssuffioe to rationalizethe process:

(1) 'XI20+ HOCH2-CHO _3 HOCH2-CHOH-CHO

(2) HOCH2-CHOH-CHO ,- HOCH2-CO-CH20H

(3) CH20 + HOCH2-CO-CH20H + HOCH2-CHOH-CO-CH2OR

(4) HOCH2-CHOH-CO-CH20H \d HOCH2-CHOH-CHOH-CHO

(5) HOCH2-CHOH-CHOH-CHO __+ HOCH2-CHO + HOCH2-CHO

The first aldol oondensationis unexoeptionable,and the tautcmerizationof

glyoeraldehydeto dihydroxyacetonethrough the common enol is also well

hOWIl. This osn of course undergo a second aldol condensationto form the

keto tetrose. The crucial step is the next one, in which this keto l&rose

tautomerizesto the aldotetrose. This is simply the aldol oondensation

, and by a simple reverse aldol reaction oan

regenerate the original glyoolaldehydeand liberate the new one, which has

thus been synthesizedfrom two formaldehydemoleoules.

This mechanism is oertainlyoonsistentwith the operation of small

amounts of glycolaldehyde,glyoeraldehyde,or dihydroqacetone in

eliminatingthe induction period but not affecting the rate of the fast

process, assuming that a steady state oonoentrationof these intermediates

would be establishedbecause of the competing further condensations.

5 R. Rreslow, J. Amer. Chem. Soo, SO, 3719 (1958).

-thermore, this mechanism explains why benzoylcarbinolcan give both an

immediatereaction and a faster reaction when it is added as catalyst to

the formaldehydereaction mixtures3 for it oan function in place of

glycolaldehyde,thus eliminatingthe induction period, and condensation

reactiona involving benzoylcarbinolmight zell be even faster than those

involving glycolaldehyde,a presumptionwhich is supported by the obeer-

vation that m- and p-cyanobenzoyloarbinol

the p-methoxy compound ia somewhat worae.

Many other catalysts are known for this reaction,2and moat of them,

e.g. benzoin, could not function in place of glycolaldehydein our

mechanism,a factor which has undoubtedly contributedto the overlookingof

this mechanism previously. However, in all these cases the effect is to

shorten the induction period only, and the eventual reaction rate is not

accelerated,but is identicalwith that in the fast part of the normal

reaction. Thus we suggest that many of these other catalysts facilitate

the formation of the first molecule of glyoolaldehyde,but that the catalysts

are not involved in the fast section of the reaction. It is possible to

suggest reasonablemechanism for all of these cases by which the first

glycolaldehydemolecule might be produced more rapidly, but these will not

be outlined here since our prime concern is with the second fast part of

the reaction.

It has also been found that the reaction ia very dependent on the

metal ion involved in the baae,5 being rapid with thallium or lead hydroxides,

relatively fast with calcium hydroxide, and of negligible importancewith

sodium or potaeaium hydroxide. It haa even been found that in the presence

of an appropriatemetal ion the reaction can be acid catalyzed. All of this

is explicable in terms of the well known stabilizingeffect of chelating

ions on enediols,nhich are postulatedintermediatesin the tautomerisation

and condensationsteps of our mechanism. Finally, the uniqueness of

aldehydes can oondense to an acyloin with an active mathylene group, snd

therefore,it alone can lead to an autooatalyticreaction.

The function of glycolaldehydein this mechanism is clearly analogous

to that of cysnide ion in the benzoin condensation. A oarbonionis

involved (in this case the enolate ion of glyoolaldehyde)whose carbon also

has aarbonyl-likecharaoter (in this case by tautomerisationto a true

carbonyl group). The prediationof other catalysesof this type and some

other prediotionsbaaed on our suggestedmechanism are currentlyunder

investigation.