This journal is c The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2011 New J. Chem., 2011, 35, 1787–1794 1787

Cite this: New J. Chem., 2011, 35, 1787–1794

Hydrothermal formose reactionw

Daniel Kopetzki* and Markus Antonietti

Received (in Montpellier, France) 1st March 2011, Accepted 11th May 2011

DOI: 10.1039/c1nj20191c

The self-condensation of formaldehyde is a one pot reaction resulting in a complex mixture of

carbohydrates. Based on a simple chemical, the reaction was previously considered as a prebiotic

source for sugar generation. Usually, a high pH and the presence of catalytically active species are

required. Here, the formose reaction was performed under hydrothermal temperatures up to 200 1C,

and carbohydrates were obtained under even simpler conditions. We found no pronounced

catalytic influence of active cations, and a slightly alkaline pH was sufficient to induce the

reaction. Maximum yield was reached in very short times, partly less than 1 minute. No selectivity

for a particular carbohydrate, although searched for, was found. Contrary to reactions performed

at lower temperatures, hexoses were only formed in negligible yields, whereas the shorter

carbohydrates accounted for the major fraction. Among the pentoses, ribose and the ketoses with

corresponding stereochemistry were formed in higher yields compared to the reaction at lower

temperature. Furthermore, we identified 2-deoxyribose in the product mix and found strong

indications for the presence of other deoxy compounds. Hence, the hydrothermal formose

reaction shows some remarkable differences compared to the conventional reaction at moderate

temperatures.

1 Introduction

As nearly 3.5 billion year old microfossils and typical isotope

patterns in sediments indicate, life arose quite early on earth,1

just about some hundred million years after the moon/earth

collision. Thus, the primary synthesis of biomolecules took

place at environmental conditions very different to the current

ones. The prebiotic chemistry community tries to resolve

the question, how complex organic molecules can be formed

from simple precursors, and to elucidate plausible mechanisms

proceeding in prebiotic environments.2 This is hampered by the

lack of knowledge about the true conditions on the early earth,

which includes the composition of the atmosphere, whether it

was neutral or reducing,3 or the temperature of the ocean.4

In the famous Miller experiment,5,6 the generation of amino

acids was proven in a simple reducing atmosphere subjected to

spark discharges. Recent work has mainly focused on the

formation mechanisms towards peptides7 and nucleic acids,8,9

while the focus of this paper lies more in the formation and

chemistry of carbohydrates. In this respect, we will investigate

whether high temperature is feasible as an energy source.

Habitats with superheated water are available in submarine

areas with volcanic activity and such hydrothermal environ-

ments with temperatures around 200 1C at a pressure of 20 bar

or higher were also present at ocean sites in the hadean period.10

The presented synthesis of carbohydrates is based on

formaldehyde, which can be clearly formed under hadean

conditions and is generally considered as a prebiotic molecule

having contributed to the local chemistry.11 It can be synthe-

sized by photoreduction of CO2, but is also available via

electric discharges.12 This simple compound can be converted

to a mixture of different carbohydrates in a one pot reaction,

called the formose reaction. Under alkaline conditions and

with certain catalysts, formaldehyde polymerises to form

sugars.13 Due to the ease with which complex carbohydrates

are synthesised from a very simple precursor, the formose

reaction has been considered to have contributed to the origin

of life. However, due to the fast degradation of sugars and the

missing selectivity, this is doubtful.14–16

The kinetics of the formose reaction is quite complex, due to

its autocatalytic nature.17 Formaldehyde usually does not

react with itself establishing a carbon–carbon bond, so that

the simplest sugar glycolaldehyde is formed slowly (see Scheme 1).

However, as soon as some condensation product is present, a

cascade of reactions is initiated, ultimately leading to the

formation of various straight-chain and branched carbo-

hydrates,18,19 plus their decomposition products. Elongation

of the carbohydrate backbone occurs via base-catalyzed aldol

condensation with formaldehyde, accompanied by isomerisations.

Max-Planck-Institute of Colloids and Interfaces, Research CampusGolm, D-14424 Potsdam, Germany.E-mail: daniel.kopetzki@mpikg.mpg.de; Fax: +49 331 567 9502;Tel: +49 331 567 9538w Electronic supplementary information (ESI) available: Moderatetemperature experiments, NMR and GC data. See DOI: 10.1039/c1nj20191c

NJC Dynamic Article Links

www.rsc.org/njc PAPER

Dow

nloa

ded

by U

nive

rsity

of

Oxf

ord

on 1

0 N

ovem

ber

2011

Publ

ishe

d on

16

June

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

1NJ2

0191

CView Online / Journal Homepage / Table of Contents for this issue

1788 New J. Chem., 2011, 35, 1787–1794 This journal is c The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2011

In alkaline medium retro aldol reactions take place as well,

whereupon both resulting fragments can act as initiators

again, thus leading to an autocatalytic system. Due to their

relative stability pentoses and hexoses are the main products.

For prolonged reaction times however they decompose, recogni-

sable by the yellowing of the solution.

Often, Ca(OH)2 is used as base because of its high catalytic

activity. Ca2+ can coordinate the enediol form of carbo-

hydrates and thus stabilises the deprotonated species.20

Performing the reaction in NaOH in the absence of other

catalytically active cations does not yield any sugars. Apart

from Ca(OH)2 many other catalysts of the formose reaction

have been identified.21 Naturally occurring minerals and clays,

quite abundant in nature, can also evoke the formation of

carbohydrates when refluxing a formaldehyde solution.22–24

It is even possible to induce the formose reaction photo-

chemically, resulting in the formation of highly branched sugar

alcohols.25 Apart from formaldehyde other small molecules can

also be employed to build up carbohydrates. Using short

sugars, like glycolaldehyde or glyceraldehyde, other catalysts

and less harsh conditions are sufficient. Examples include zinc

prolate,26 silicate27 or dipeptides,28 but in none of these cases a

successful formose reaction using solely formaldehyde could

be performed.

The formose reaction is usually performed at moderate

temperatures or occasionally at around 100 1C.29 In recent

work, we have presented a continuous flow reactor which

allows us to perform organic reactions in water at high

temperatures and pressures with high control and precision.30

It was for instance shown that formic acid acts under such

conditions as an effective hydrogenation agent, while simple

salts can take an unexpected role of being a catalyst. In the

present attempt, we will apply this set-up to the formose

reaction. Using simple hydrothermal reaction conditions and

formaldehyde without additional initiators in the presence of

only simple salts, reaction sequences are analysed. The moti-

vation to study the formose reaction under such conditions is

based on the lack of data on the hydrothermal behaviour of

formaldehyde yielding complex molecules,31,32 but also on the

fact that early earth conditions might have included various

aqueous environments under similar conditions.

2 Results and discussion

Effects of added salt

To establish hydrothermal conditions with high precision and

control, reactions were conducted in a continuous flow reactor.

A 0.5 M formaldehyde solution was heated to 200 1C under a

pressure of 100 bar. It should be noted that such a high

concentration is not along with prebiotic conditions. To vary

the pH and to probe for the potential catalysis of simple ions,

salts were added. Despite the fact that certain cations are

necessary at moderate temperatures, we first used inactive

sodium or potassium salts. Control experiments were conducted

at 60 1C in 0.05 M Ca(OH)2 or 0.1 M NaOH, respectively

(ESIw, Fig. S1–S3).The conversion of formaldehyde in different salt solutions is

shown in Fig. 1. In pure water and even under acidic conditions

(such as diluted acetic acid), formaldehyde is consumed relatively

fast within a timescale of minutes. An induction period, as

described for the formose reaction at moderate temperatures,

is not observed. With increasing basicity of the added salts, the

conversion is accelerated. Even the barely basic sodium sulfate

shows some effect. This trend is continued following the series

acetate, hydrogen carbonate and hydrogen phosphate. In a

carbonate buffer (50 mM NaHCO3, 50 mM Na2CO3), the

formaldehyde is consumed in less than one minute.

Of course, the fact that formaldehyde vanishes does not

prove the formation of carbohydrates. In fact, the NMR

spectra of the reactions performed in pure water, in acetic

acid and also when sodium acetate was added, just show the

Cannizzaro products formic acid and methanol (ESIw, Fig. S4).These solutions remained clear and colourless even for

prolonged reaction times and did not show the typical yellow-

ing point, resulting from decomposition products when

carbohydrates were formed. This colour change was however

observed for all more basic salts, indicating a successful

Scheme 1 Simplified mechanism of the formose reaction.

Fig. 1 Conversion of formaldehyde in the presence of various salts at

200 1C and 100 bar.

Dow

nloa

ded

by U

nive

rsity

of

Oxf

ord

on 1

0 N

ovem

ber

2011

Publ

ishe

d on

16

June

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

1NJ2

0191

C

View Online

This journal is c The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2011 New J. Chem., 2011, 35, 1787–1794 1789

formose reaction despite the absence of catalytically active

cations. In a control experiment with microwave heating, this

yellowing occurred in glass vessels as well, suggesting that it is

not the reactor material which causes the observed effects.

Influence of an initiator and calcium ions

We also tested for the influence of both the cation variation

and the presence of an initiator. A 0.2 M sodium acetate or,

respectively, a 0.1 M calcium acetate solution were used to

investigate the effect of a potential cation catalysis. Both

solutions contain the same amount of acetate and thus exhibit

similar basicity. Fig. 2 shows that up to two minutes reaction

time the consumption of formaldehyde is identical in both

cases. Subsequently, the reaction accelerates in the calcium

acetate solution. An interpretation could be that at this point

small carbohydrates such as glycolaldehyde must have formed,

which can specifically interact with the Ca2+-ions and accel-

erate the formose reaction. Testing the reaction solutions from

the first two minutes, indeed predominantly the Cannizzaro

products were found in both solutions. In case of the calcium

salt, the solution turns yellow at later stages, and carbo-

hydrates are formed. The maximum yield of carbohydrates

is already reached for a formaldehyde conversion of 55%. This

is to be compared with the formose reaction at moderate

temperatures where the yield is at a maximum shortly after all

formaldehyde is consumed. According to the data, decomposition

of carbohydrates seems to be more accelerated at higher

temperature, which is expected. The overall yield of sugars is

rather small. Integrating over the signals of all linear sugars

with two to six carbon atoms, we obtain only 3.4%. The

composition and distribution of sugars will be discussed later.

To compare this salt catalysis, reactions were carried out

with the addition of glycolaldehyde as a promoter (1 mol%

with respect to formaldehyde). This simplest sugar is a power-

ful initiator of the formose reaction. Indeed, the conversion of

formaldehyde is much faster (Fig. 2). Here, the sodium acetate

solution also turns yellow, and carbohydrates are formed. In

case of Na(CH3COO) the total yield is 5.1%, whereas for

Ca(CH3COO)2 it reaches 12.0%. Again, the maximum of

carbohydrate formation is reached well before complete

formaldehyde consumption.

At moderate temperatures, the presence of Ca2+ is crucial

for a successful formose reaction. Even with initiator no

carbohydrates are formed in the absence of catalysts. The

initiator only reduces the induction period but does not

improve the yield when the formose reaction is performed in

Ca(OH)2 solution. In our experiments the total yield of

(straight-chain) carbohydrates could reach 26.5%.

Under hydrothermal conditions however, an initiator and a

catalyzing counterion simply step into the competition of

formose and Cannizzaro reactions. In the presence of glycol-

aldehyde the carbohydrate formation starts immediately

whereas in the absence a substantial amount of formaldehyde

is lost before by self-disproportionation. To keep the experi-

ments limited to simple chemicals, the use of an initiator was

omitted for further studies. The more basic salts NaHCO3 and

K2HPO4 or a sodium carbonate buffer also resulted in a

yellowing of the solution after relatively short reaction times

(Fig. 1) and coupled carbohydrate formation. Not only

the speed of formaldehyde consumption, but also the yield

of carbohydrates was improved compared to the addition

of more neutral salts, such as Ca(CH3COO)2. This again

points to the fact that a slightly more basic pH is of

greater relevance than the specific presence of Ca2+ under

hydrothermal conditions.

Typical kinetics on the example of K2HPO4 addition

For the model case of 0.1 M K2HPO4 addition, the kinetics of

formation and decomposition of carbohydrates was analysed

in detail. Fig. 3 shows the concentration of formaldehyde and

the products as added masses of carbohydrates with the same

number of carbon atoms. The products were analysed by gas

chromatography as their alditol acetates. By reduction with

NaBH4 the number of species is decreased, which facilitates

analysis. Aldoses yield the same products as ketoses with

corresponding stereochemistry. Only the linear carbohydrates

were quantified. It is known that branched sugars are also

formed in formose reactions,18 and the GC chromatograms

showed some weak peaks that could be attributed to them. A

control experiment, in which the reduction step was omitted,

Fig. 2 Conversion of formaldehyde with addition of sodium or

calcium acetate in the presence and absence of the initiator glycol-

aldehyde at 200 1C and 100 bar.

Fig. 3 Kinetics of formaldehyde consumption and product forma-

tion of a 0.5 M formaldehyde solution in 0.1 MK2HPO4 at 200 1C and

100 bar.

Dow

nloa

ded

by U

nive

rsity

of

Oxf

ord

on 1

0 N

ovem

ber

2011

Publ

ishe

d on

16

June

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

1NJ2

0191

C

View Online

1790 New J. Chem., 2011, 35, 1787–1794 This journal is c The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2011

indicated that negligible amounts of sugar alcohols are

present after reaction (ESIw, Fig. S5). This proves that there

is only little cross-disproportionation of carbohydrates with

formaldehyde during reaction. The induction period in 0.1 M

K2HPO4 is only around 25 s. During this time span about 20%

of formaldehyde is lost via disproportionation. Afterwards the

formation of carbohydrates is very rapid. It takes less than

another 30 s to reach maximum yield.

Contrary to the reaction at ambient temperatures, where

hexoses are the main products, here the shorter carbohydrates

are preferentially formed. Despite the high stability that is in

principle expected for hexoses, their amount is insignificant at

200 1C. With the exception of glycolaldehyde, all different

carbohydrates peak around the same reaction time. At 60 1C

and under catalysis by Ca(OH)2 the different sugars form

consecutively, with the maximum yield of shorter carbo-

hydrates being reached earlier than that of the longer ones.

This reflects a chain-like growth of the carbohydrates by

successive additions of formaldehyde.

We speculate that under hydrothermal conditions we have a

very fast equilibration between all the different compounds.

Using glycolaldehyde or dihydroxyacetone as initiator did not

alter product distribution, as compared to the reaction without

initiator. The ineffectiveness of different initiators to influence

selectivity is however well described for the formose reaction.

Not only the formation, but also the decomposition is very

rapid. One intermediate seems to be glycolaldehyde itself,

because its concentration still increases while all other sugars

diminish. The nature of the decomposition products was

however not further investigated. It is known that some

polymeric species are formed. At long reaction times the

solution turned dark brown and ultimately turbid and exhibited

a strong caramel like odour.

Gas chromatographic analysis revealed the presence of

other compounds besides straight-chain carbohydrates. The

chromatogram of the reaction at maximum yield is shown in

Fig. 4. Three different ions were scanned. The signal atm/z=43

originates from the acetoxy cation. Thus all compounds with

hydroxyl groups show a peak in this trace. The ion m/z = 115

is characteristic for alditols. It originates from a fragment with

3 carbon atoms after loss of acetic acid and ketene.33 The ion

m/z = 129 is formed the same way when an additional methyl

group is present. Consequently, deoxysugars show this signal.

In this GC trace also some extra peaks are present. Most

importantly, we were able to quantify 2-deoxyribitol. Its

identity was validated by comparing the complete mass spectrum

and the retention time with the reference compound. The

maximum yield was however only 0.18 mg per 100 mg

formaldehyde. Still, it is worth noting that not only the true

condensation products of formaldehyde are found. As the

various peaks with the m/z= 129 ion fragment indicate, many

other deoxy compounds are probably formed as well. Further

indications for the presence of deoxy species are gained from

NMR spectra that show peaks in the region of 1–2 ppm, which

cannot be attributed to ordinary carbohydrates (ESIw, Fig. S6and S7). In the control experiment at 60 1C in Ca(OH)2 we also

detected 2-deoxyribitol after reduction with NaBH4. Here, the

maximum yield was lower (0.11 mg per 100 mg formaldehyde).

The decomposition of 2-deoxyribitol is slower than that of the

other carbohydrates. This is a general observation for all

compounds showing a fragment with m/z = 129.

To evaluate the amount of self-disproportionation of

formaldehyde during the reaction, pH measurements were

conducted after the synthesis for the model reaction in

0.1 M K2HPO4. The variation of pH (after threefold dilution)

is shown in Fig. 5. As expected the pH decreases throughout

the whole reaction process. Although no sugars were formed

in the first 30 s, the pH falls rapidly. Under the assumption

that the change in pH is only caused by the formation of

formic acid, its concentration can be calculated with the law of

mass action, taking into account the self-disproportionation

of water and the equilibria of the different phosphate species

and formic acid (with the dissociation constants Kw = 10�14,

pK1 = 2.15, pK2 = 7.20 and pK3 = 12.35 for phosphoric

acid34 and pK = 3.76 for formic acid).35 The calculated

concentration is also plotted in Fig. 5. Large amounts of acid

are created. The maximum concentration of formic acid

reaches 100 mM, which is one-fifth of the formaldehyde

initially present. The Cannizzaro reaction thus accounts for

40% of the formaldehyde consumption. This loss is much

Fig. 4 GC traces of the reduced and acetylated formose products for

different ions (each scaled with different factors), for the reaction of

0.5 M formaldehyde in 0.1 M K2HPO4 at 200 1C and 100 bar.

Fig. 5 Variation of pH (squares) and calculated amount of formic

acid (circles) for the reaction of 0.5 M formaldehyde (triangles) in

0.1 M K2HPO4 at 200 1C and 100 bar.

Dow

nloa

ded

by U

nive

rsity

of

Oxf

ord

on 1

0 N

ovem

ber

2011

Publ

ishe

d on

16

June

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

1NJ2

0191

C

View Online

This journal is c The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2011 New J. Chem., 2011, 35, 1787–1794 1791

higher than in the conventional formose reaction at moderate

temperatures. Due to the low salt concentration, the buffer

capacity is exhausted quite fast. When the formation of

carbohydrates finally takes place, the solution is neutral or

even slightly acidic.

Influence of buffer capacity

The change in basicity might indeed stop the formose reaction.

To conduct the reaction under stable pH conditions, the salt

concentration was raised to 0.5 M. A sodium carbonate buffer,

consisting of equal proportions of Na2CO3 and NaHCO3, was

employed. With these modified conditions the pH remained in

the alkaline range also for prolonged reaction times. Perform-

ing the reaction at 200 1C resulted in a dark brown solution

already after less than half a minute. Contrary to the reactions

with low buffer capacity, the solution at high pH did not

become turbid, even for prolonged reaction times. It appears

that no carbonisation occurs in alkaline media.

Due to the fast reaction, the temperature was lowered to 150 1C.

Here, kinetics is comparable to the system with 0.1 M

K2HPO4 at 200 1C. However, the induction period and the

subsequent formation of carbohydrates are clearly more dis-

tinct (see Fig. 6). The formation of carbohydrates occurs quite

abruptly. It takes only 6 s from the beginning of their forma-

tion to maximum yield. Again, the shorter carbohydrates

account for the major part, though the relative fraction of

hexoses is more pronounced than at higher temperatures. The

product distribution is slightly altered. In the strong alkaline

medium, pentoses are formed preferentially. When the buffer

concentration was low and the final pH slightly acidic, pen-

toses, tetroses and trioses were formed in equal amounts.

Besides the very fast formation of carbohydrates, their decom-

position is highly accelerated in alkaline media. This is prob-

ably the reason why the overall yield was not improved by a

higher buffer capacity. Under those conditions it reaches only

7.6%. It turns out that the lower buffer capacity might even

have been advantageous. At the end of the reaction the pH

turned neutral, and sugars are more stable under these condi-

tions than at high pH.

Carbohydrate selectivity

It is an interesting question whether the nature of the added

salt or catalyst had some effect on the product selectivity.

Table 1 shows the maximum yield with different salts. Salts are

ordered according to increasing basicity. The total yield nicely

indicates that the starting pH and not any ion catalysis is the

decisive factor. When working under hydrothermal conditions

the yield is lower compared to moderate temperatures, but less

harsh conditions, e.g. a significantly lower alkaline pH, are

necessary to induce the formose reaction. Among the hydro-

thermal reactions, using a simple 0.1 M carbonate buffer

resulted in the highest relative yield. Concentration and type

of buffer are optimal, as the solution is initially alkaline,

allowing the formose reaction to proceed. When formaldehyde

was consumed and the maximum of carbohydrate yield

reached, the solution became neutral. Here, sugars exhibit

the highest stability.

Taking a look at various reactions performed at 200 1C no

difference in selectivity towards a special sugar was apparent.

The relative amounts of carbohydrates with different carbon

numbers are roughly identical. As discussed earlier, the most

striking difference is the virtually complete absence of hexoses at

200 1C and the preferential formation of shorter carbohydrates.

Two different alditols, erythritol and threitol, originate from

the post reaction reduction of the tetroses. Both are formed in

approximately equal amounts at 200 1C as well as at 60 1C.

This is different regarding pentoses. Here, three sugar alcohols

are identified. Under the assumption that all possible stereo-

isomers of pentoses are synthesised with the same probability,

one should obtain 50% arabinitol and 25% of ribitol and

xylitol each. Indeed, sugars yielding arabinitol account for

roughly half of the (linear) pentoses. However, we found a

higher concentration of ribitol compared to xylitol under

hydrothermal conditions. This selectivity is reversed at 60 1C.

We can speculate that this is an effect of different reactivities

of the pentoses. Fig. 7 shows their fraction against reaction

time for the hydrothermal reaction in 0.1 M K2HPO4 and the

reaction at 60 1C in 0.05 M Ca(OH)2. In the early stages ribitol

dominates over xylitol in both cases. Under Ca(OH)2 catalysis,

Fig. 6 Kinetics of formaldehyde consumption and product forma-

tion of a 0.5 M formaldehyde solution in 0.5 M carbonate buffer at

200 1C and 100 bar.

Table 1 Maximum yield of different carbohydrates at hydrothermaland moderate temperatures of a 0.5 M formaldehyde solution indifferent salts, concentration of salts was 0.1 M except for Ca(OH)2with 0.05 M

Yield at maximum [%]

200 1C, 100 bar 60 1C

Ca(OAc)2 NaHCO3 K2HPO4

NaHCO3/Na2CO3 Ca(OH)2

Glycolaldehyde 0.95 0.80 1.41 1.20 0.55Trioses 0.56 1.41 2.40 2.87 1.73Tetroses 0.87 1.67 2.99 3.19 3.15As erythritol 0.51 0.82 1.54 1.53 1.54As threitol 0.37 0.85 1.45 1.66 1.62

Pentoses 0.90 1.46 2.48 2.94 9.17As ribitol 0.25 0.41 0.73 0.82 1.65As arabinitol 0.49 0.69 1.22 1.38 4.27As xylitol 0.16 0.35 0.52 0.74 3.25

Hexoses 0.16 0.16 0.30 0.29 11.79Total 3.44 5.50 9.58 10.48 26.50

Dow

nloa

ded

by U

nive

rsity

of

Oxf

ord

on 1

0 N

ovem

ber

2011

Publ

ishe

d on

16

June

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

1NJ2

0191

C

View Online

1792 New J. Chem., 2011, 35, 1787–1794 This journal is c The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2011

this is reversed very rapidly. Under hydrothermal conditions

however, this only occurs for very prolonged reaction times,

when the overall yield of sugars is already very low. In a

neutral solution at 100 1C, ribose decomposes faster than

xylose, correlating with the percentage of free aldehyde in

solution.14 Although ribose seems to be formed preferentially,

its reactivity is higher. At 60 1C a possible sink for ribose is the

continued reaction to hexoses, which is negligible under

hydrothermal conditions. Therefore, the fraction of ribitol

does not decrease until the maximum yield of carbohydrates

has been reached.

The effect of temperature was investigated in the range of

125 1C to 200 1C in 0.1 M K2HPO4 (Fig. 8). Here, 1 mol%

glycolaldehyde was added to reduce the reaction time, as it is

difficult to handle extended reaction times with the employed

flow setup. In particular we checked whether more hexoses

would be synthesised at lower temperatures. Surprisingly this

was not the case, at least not in significant terms and in the

analysed temperature range. Chromatograms at lower tempera-

tures revealed that less side products are formed. For instance,

the detectable amount of 2-deoxyribitol decreased. However,

total yields were not significantly affected.

Carbohydrate stabilisation

of K2HPO4 to the formaldehyde solution, were equally valid

for other salts. Only the overall yield is influenced by the

nature of the additive, but not the relative product distribution.

However, when the additive contains reactive sites, other

reaction pathways are possible, which could push the formose

reaction towards a certain product. Since the fast decomposition

is the main cause for the low yield, the reaction can only be

improved by either stabilising or trapping the products. We

performed the reaction in the presence of 0.1 M adenine (its

solubilisation was achieved by heating to 100 1C) and observed

formaldehyde consumption as well as a colour change. However,

no free carbohydrates could be detected and NMR spectra did

not show any carbohydrate signals in the region of 3.5 to 4.5 ppm

(ESIw, Fig. S8). Clearly, the presence of amines blocks the

formose reaction and directs it towards other pathways. When

zinc prolate was used as additive in the formose reaction, a

very rapid consumption of formaldehyde took place (see Fig. 1).

This salt was shown to be an efficient catalyst for aqueous

aldol reactions.26 At the high temperatures used in our experi-

ments, however, conversion levels of proline were similar to

those of formaldehyde, indicating an incorporation of the

amino acid into the products. Again, no carbohydrates could

be detected.

Another means of stabilisation, complexation of sugars with

borate, also was not successful. Borate minerals were shown to

stabilise ribose and other carbohydrates.36 When we performed

the reaction in 0.125 M Na[B(OH)4] or in a 0.125 M borate

buffer prepared with borax, we observed a fast conversion of

formaldehyde (ESIw, Fig. S9), but NMR spectra revealed that

only the Cannizzaro reaction took place. Even when 1 mol%

glycolaldehyde was added as initiator, only marginal amounts

of carbohydrates were formed which furthermore decomposed

quite rapidly. Also increasing borate concentration to 0.5 M

and accordingly lowering temperature to 130 1C did not result

in significant carbohydrate formation. Obviously borate is

an efficient catalyst of formaldehyde disproportionation

under hydrothermal conditions and is unable to stabilise

carbohydrates at elevated temperatures. Recently, silicate

was reported as another compound complexing and stabilising

carbohydrates.27 Its use was however not probed by us, as

possibly occurring silica precipitation was suspected, which

would damage the employed flow reactor.

3 Conclusions

We have analysed the feasibility of the formose reaction under

hydrothermal conditions. In contrast to the counterpart at

ambient temperature, less demanding conditions are required.

Catalytically active species are not essential and only have

minor effects regarding yield. A slightly alkaline solution is

sufficient to induce the reaction. In fact, depending on buffer

capacity, the pH can even drop to slightly acidic at the end of

the reaction. A high pH throughout the whole reaction is in

fact disadvantageous, as product decomposition is favoured

above a pH too high. No selectivity towards a particular

product could be detected. However, considerable differences

in the outcome of the reaction with respect to moderate

temperatures were found. Hexoses, being the main product

at low temperature, are only formed in negligible amounts.

Regarding the distribution of pentoses, we found a reversed

selectivity towards carbohydrates yielding ribitol and xylitol.

Fig. 7 Fraction of pentoses for the hydrothermal reaction in 0.1 M

K2HPO4 at 200 1C and 100 bar (a) and the reaction at 60 1C in 0.05 M

Ca(OH)2 (b) showing arabinitol (circles), ribitol (squares) and xylitol

(triangles).

Fig. 8 Consumption of 0.5 M formaldehyde in 0.1 M K2HPO4 with

1 mol% glycolaldehyde at 200 1C (squares), 175 1C (circles), 150 1C

(upward triangles) and 125 1C (downward triangles) under a pressure

of 100 bar.

Dow

nloa

ded

by U

nive

rsity

of

Oxf

ord

on 1

0 N

ovem

ber

2011

Publ

ishe

d on

16

June

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

1NJ2

0191

C

View Online

This journal is c The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2011 New J. Chem., 2011, 35, 1787–1794 1793

Apart from the true condensation products of formaldehyde,

also deoxysugars were detected. Hydrothermal conditions

seem to facilitate alternative reaction pathways. The low yield

caused by rapid decomposition is a key issue. Attempts to

stabilise or trap the products under hydrothermal conditions

have been unsuccessful so far. These findings do not provide

convincing evidence that the formose reaction played a role in

the prebiotic formation of carbohydrates. Still, they nicely

demonstrate the extent hydrothermal reactions can differ from

moderate temperature chemistry. The aforementioned char-

acteristics of the formose reaction under hydrothermal and

moderate temperatures are summarised in Table 2.

4 Experimental

The experiments were performed in the X-Cube Flash

continuous flow reactor from Thales Nano, which was coupled

to a Gilson GX-271 autosampler. A 4 mL reactor made out of

Hastelloy C-22, an alloy based on nickel, chromium and

molybdenum, was employed. Residence time was calculated

taking into account the density ratio of water under reaction

and ambient conditions. Because of the relatively low solute

concentration, the density of pure water was used.37

Formaldehyde solution was prepared by refluxing an aqueous

suspension of paraformaldehyde, which was purchased from

Sigma-Aldrich. Salts and carbohydrate standards were purchased

from different commercial suppliers. N-Methyl imidazole,

sodium borohydride, acetic anhydride and chromotropic acid

disodium salt dihydrate were from Sigma-Aldrich.

NMR spectra were recorded on a Bruker DPX 400 NMR.

10%Deuterium oxide was added to aqueous samples containing

volatile compounds. Otherwise pure deuterium oxide was

employed as solvent.

Carbohydrate analysis was carried out with an Agilent

6890N GC with Agilent 5975 mass spectrometer (EI ionisation)

after conversion of the sugars to their corresponding alditol

acetates.38 For calibration curves, standards were prepared

from the corresponding aldoses, except for erythritol, threitol

and iditol, which were directly employed as such.Myo-inositol

was added as internal standard immediately after reaction.

Afterwards carbohydrates were reduced to alditols using

NaBH4. Excess hydride was destroyed by addition of 500 mLacetic acid. Subsequently to reconcentration to about 1 mL,

200 mL were taken and acetylated with 2 mL acetic anhydride

and 200 mLN-methyl imidazole for 10 min at room temperature.

The solution was quenched with 5 mL water and after

decomposition of excess acetic anhydride the alditol acetates

were extracted with 1 mL dichloromethane, which then was

dried over anhydrous Na2SO4. Separation was achieved on a

DB-225ms column. The ions m/z = 43, 115 and 129 were

scanned, each with a dwell time of 50 ms. The ion at m/z = 43

was used for the quantification of glycolaldehyde, m/z = 129

for 2-deoxyribitol and m/z = 115 for all other alditols and the

internal standard.

Amino acids were converted to volatile products by deriva-

tisation with ethyl chloroformate.39 400 mL of a 4 : 1 mixture

of ethanol : pyridine were added to 600 mL of aqueous sample.

50 mL ethyl chloroformate was then added and the vial briefly

shaken. The resultingN-ethoxycarbonyl ethyl esters were extracted

with 1 mL chloroform containing 1% ethyl chloroformate and

separated on an HP-5ms column by gas chromatography.

The conversion of formaldehyde was monitored photo-

metrically with the chromotropic acid method19 using a Perkin

Elmer Lambda 2 spectrometer. This assay is characterised

by very high sensitivity and selectivity. However, there were

slight interferences with formose products, which causes the

absorbance not to reach the baseline even for prolonged

reaction times. The chromotropic acid reagent was prepared

by dissolving 0.5 g of chromotropic acid disodium salt dihydrate

in 3.75 mL of water and subsequent addition of 100 mL

concentrated sulfuric acid. To determine the concentration of

formaldehyde the sample was diluted with water to a maximum

of 3 mM HCHO. 100 mL were reacted with 1 mL of reagent

solution at 100 1C for 15 min. After cooling to room temperature

and dilution with 5 mL water the absorbance was read off at

578 nm against a blank solution, prepared by the same protocol

with pure water.

References

1 J. W. Schopf, Science, 1993, 260, 640–646.2 A. Eschenmoser, Tetrahedron, 2007, 63, 12821–12843.3 G. H. Shaw, Chem. Erde: Geochem., 2008, 68, 235–264.4 S. A. Wilde, J. W. Valley, W. H. Peck and C. M. Graham, Nature,2001, 409, 175–178.

5 S. L. Miller, Science, 1953, 117, 528–529.6 S. L. Miller, J. Am. Chem. Soc., 1955, 77, 2351–2361.7 B. M. Rode, Peptides, 1999, 20, 773–786.8 R. Shapiro, Orig. Life Evol. Biosph., 1995, 25, 83–98.9 L. E. Orgel, Crit. Rev. Biochem. Mol. Biol., 2004, 39, 99–123.10 G. Macleod, C. Mckeown, A. J. Hall and M. J. Russell, Orig. Life

Evol. Biosph., 1994, 24, 19–41.11 I. Cleaves and H. James, Precambrian Res., 2008, 164, 111–118.12 G. Schlesinger and S. L. Miller, J. Mol. Evol., 1983, 19, 383–390.13 A. Butlerow, Justus Liebigs Ann. Chem., 1861, 120, 295–298.14 R. Larralde, M. P. Robertson and S. L. Miller, Proc. Natl. Acad.

Sci. U. S. A., 1995, 92, 8158–8160.15 C. Reid and L. E. Orgel, Nature, 1967, 216, 455.16 R. Shapiro, Orig. Life Evol. Biosph., 1988, 18, 71–85.17 R. Breslow, Tetrahedron Lett., 1959, 1, 22–26.

Table 2 Comparison of the formose reaction under hydrothermal and moderate temperatures

High temperature and pressure Moderate temperature

Only slightly basic initial conditions necessary Strongly alkaline conditionsCatalytically active ions have minor effects Catalytically active ions necessaryInitiator improves yield Initiator does not influence yieldInduction period even without initiator quite short Initiator drastically reduces induction periodNature of initiator does not influence outcome of reaction Nature of initiator does not influence outcome of reactionShorter carbohydrates are formed, rarely hexoses Main products are hexoses and pentosesSelectivity pentoses: arabinitol > ribitol > xylitol Selectivity pentoses: arabinitol > xylitol > ribitolLower yield Higher yield

Dow

nloa

ded

by U

nive

rsity

of

Oxf

ord

on 1

0 N

ovem

ber

2011

Publ

ishe

d on

16

June

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

1NJ2

0191

C

View Online

1794 New J. Chem., 2011, 35, 1787–1794 This journal is c The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2011

18 R. D. Partridge, A. H. Weiss and D. Todd, Carbohydr. Res., 1972,24, 29–44.

19 P. Decker, H. Schweer and R. Pohlmann, J. Chromatogr., A, 1982,244, 281–291.

20 A. N. Simonov, O. P. Pestunova, L. G. Matvienko andV. N. Parmon, Kinet. Catal., 2007, 48, 245–254.

21 T. Mizuno and A. H. Weiss, Adv. Carbohydr. Chem. Biochem.,1974, 29, 173–227.

22 A. G. Cairns-Smith, G. L. Walker and P. Ingram, J. Theor. Biol.,1972, 35, 601–604.

23 A. W. Schwartz and R. M. Degraaf, J. Mol. Evol., 1993, 36, 101–106.24 N. W. Gabel and C. Ponnamperuma, Nature, 1967, 216, 453–455.25 Y. Shigemasa, Y. Matsuda, C. Sakazawa and T. Matsuura, Bull.

Chem. Soc. Jpn., 1977, 50, 222–226.26 J. Kofoed, J. L. Reymond and T. Darbre, Org. Biomol. Chem.,

2005, 3, 1850–1855.27 J. B. Lambert, S. A. Gurusamy-Thangavelu and K. B. A. Ma,

Science, 2010, 327, 984–986.28 S. Pizzarello and A. L.Weber,Orig. Life Evol. Biosph., 2010, 40, 3–10.29 V. A. Likholobov, A. H. Weiss and M. M. Sakharov, React. Kinet.

Catal. Lett., 1978, 8, 155–166.

30 D. Kopetzki and M. Antonietti, Green Chem., 2010, 12, 656–660.31 M. Osada, M. Watanabe, K. Sue, T. Adschiri and K. Arai,

J. Supercrit. Fluids, 2004, 28, 219–224.32 S. Morooka, C. Wakai, N. Matubayasi and M. Nakahara, J. Phys.

Chem. A, 2005, 109, 6610–6619.33 Y. Nygren, S. A. Fredriksson and B. Nilsson, J. Mass Spectrom.,

1996, 31, 267–274.34 R. N. Goldberg, N. Kishore and R. M. Lennen, J. Phys. Chem.

Ref. Data, 2002, 31, 231–370.35 J. L. S. Bell, D. J. Wesolowski and D. A. Palmer, J. Solution

Chem., 1993, 22, 125–136.36 A. Ricardo, M. A. Carrigan, A. N. Olcott and S. A. Benner,

Science, 2004, 303, 196–196.37 E. W. Lemmon, M. O. McLinden and D. G. Friend, Thermo-

physical Properties of Fluid System, in NIST Chemistry WebBook,NIST Standard Reference Database Number 69, NationalInstitute of Standards and Technology, Gaithersburg, MD,20899, http://webbook.nist.gov, retrieved October 15, 2010.

38 A. B. Blakeney, P. J. Harris, R. J. Henry and B. A. Stone,Carbohydr. Res., 1983, 113, 291–299.

39 P. Husek, J. Chromatogr., A, 1991, 552, 289–299.

Dow

nloa

ded

by U

nive

rsity

of

Oxf

ord

on 1

0 N

ovem

ber

2011

Publ

ishe

d on

16

June

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

1NJ2

0191

C

View Online