A Little History

1.1 Introduction

The physicist and percussionist Feynman is widely credited with envisioning atom assembly as a way of building ultra-small devices. While this vision can be realized by atom manipulation with scanning probe microscope tips, it is clear that chemists have a treasure chest of methods to arrange atoms in intricate arrays to produce molecules on a kilogram scale. Once synthesized, these molecules can be investigated one at a time – at the single molecule level – if necessary. More commonly, these molecules will be studied in large populations of many billions.

1.2 Early Proposals for Molecular Logic

Perhaps the best known early proposal for molecular logic and computation was announced in 1988 by Aviram and is outlined in Figure 1.1. Given his affiliation with IBM, it was natural that the system 1 used electric voltages for both the inputs and the outputs. The vision was as follows. A π-system1, an oligothiophene of ca. 25 nm length would be fixed via thiol terminals to two gold electrodes (electrode1 and electrode2). In this electrically neutral state, the oligothiophene would be non-conducting. The middle of the oligothiophene contains a spiro linkage so that another π-system can be positioned orthogonally. The latter π-system2 (different from or identical to the first oligothiophene) would be held in its radical cation form which is electrically conducting. The spiro linkage would hinder electron transfer between the two π-systems so that the first oligothiophene remains electrically neutral. However if a strong electric field is applied to the spiro linkage via two orthogonal switching electrodes aimed at it, theoretical calculations suggested that electron transfer would occur between the two p-systems. The first oligothiophene would then pass into its radical cation state which is an electron conductor, as mentioned earlier. Thus a voltage applied at one gold electrode1 would appear at the other (output) gold electrode2 in response to a voltage input applied across two switching electrodes aimed at the centre of 1, which are not shown in Figure 1.1. In other words, an electric field-induced insulator-to-conductor transition at the molecular scale would serve as the switching mechanism in this device. The above argument does not need electrode and electrode but they are required for building NOT logic gates, which are essential for general computing. Wiring of multiple copies of these switches was then expected to yield various logic functions.

As we go to press 24 years later, Aviram's proposal remains a proposal. Much effort has been expended by Tour, for example, to synthesize close relatives of 1 and many other derivatives in resourceful ways. Tour and his collaborators Allara, Weiss and Reed also performed pioneering two-electrode experiments on molecules simpler than 1, some of these results turning out to be controversial. Nevertheless, he concluded that it was too difficult to focus more than two metal electrodes selectively onto a small molecule such as 1 in order to perform the crucial test. Fabricating three-electrode devices based on single molecules remains a difficult art. Logic gates based on graphene and (bundled or single) carbon nanotubes have also been constructed.

Though not proposed for molecular logic-based computation as such, Aviram and Ratner's suggestion (made in 1974) for a molecular diode/rectifier reached a happier conclusion. An electron donor π-system linked via a s-bridge to an electron acceptor π-system such as 2 was Aviram's and Ratner's original suggestion. After a long odyssey and many controversies concerning the nature of the metal electrode–organic molecule interface in determining current output–voltage input profiles, the rectification behaviour of monolayers of, for example, 3 was demonstrated.

Aviram's baton has been picked up by several others to develop different approaches to molecular electronic logic and computation. These have focused on molecules different from 1, e.g. rotaxanes, catenanes, carbon nanotubes and graphene.

This electronic approach to molecular logic and computation was and is very popular, perhaps because of its direct connection with the burgeoning semiconductor computer industry and also because it engaged the concepts and techniques of chemists, physicists and engineers in equal measure. Naturally, there have been many good discussions in the literature. Well-resourced national programmes were launched in several countries. The USA operation occurred in two waves, with the first centred in the early 1980s and the second starting in the late 1990s and continuing to this day. The entire field of molecular logic-based computation has benefited as a result, but also (undeservedly) shared in the trauma of exposed hype at various times.

Hole-burning spectroscopy was another early approach which involved discussions on molecular computation. In the 1980s, Wild's laboratory demonstrated computing functions such as elementary addition on molecular substrates. This was a photochemical method. A laser picked out a few molecules in a specific characteristic microenvironment in a rigid medium on the basis of their characteristic absorption lines, to the accuracy of 0.0001 cm-1. These molecules were converted after photoselection to their tautomeric forms, which would persist. However, the logic functions were not intrinsic to the molecules but arose from optical images which were impressed upon them, for instance. The molecules served as an information storage medium, which is a highly sensitive one at appropriately low temperatures. So this approach does not come within the scope of this book, even though we need to pay homage to this very elegant science (Figure 1.2).

Chemists were developing a separate strand of thought after reflecting on biology. This was mentioned in the Pimentel report on science presented to the US House of Congress, which observed that research in molecular computation should be possible given that the brain is a pinnacle of such activity. However we must not forget that each cell possesses vital computational skills of its own.

As far as we are aware, the first claim of an intrinsically molecular-level logic experiment appeared in the conference literature. However, these reports concerning porphyrin molecules have never crossed over into the refereed primary literature in sufficient detail to allow corroboration via the usual scientific process. Additionally, we have been unable to locate any work which followed up the original claims. Birge's NAND gate molecules 4 and 5 were designed to contain two input chromophores to which light can be directed. A retinal Schiff base derivative was used as the output chromophore, whose absorption band was expected to shift upon electric charging of the porphyrin serving as a charge integrator. The input chromophores were stated to be selectively excitable although their structures do not differ significantly in 4. The electric charging upon excitation of the input chromophores was considered to arise from photoinduced electron transfers (PET, see section 4.5) in 4 and from internal charge transfer (ICT, see section 4.2) in the excited state of 5. Proper mechanistic evaluation is difficult in the absence of detailed information. Birge's later statements to the press do not give us much hope concerning these specific cases. However, the subsequent studies by Wasielewski, Gust and Andréasson are distinctly different and employ clear mechanistic designs. These represent progress along this general avenue.

1.3 Photochemical Approach to Molecular Logic and Computation

It was in 1993 that we published the first general and practical approach to intrinsically molecular logic. This approach, based on ion-driven luminescent signalling systems, grew steadily and without fuss as an activity among chemists and molecular physicists initially. This philosophy forms a substantial fraction of the molecular logic gates currently known. Its use of high-school chemistry fundamentals in combination with basic computer science has appealed to the younger generations of scientists from various backgrounds. The past 19 years have shown how molecules can begin to perform some of the computational functions achieved in semiconductor technology. and begin to apply these functions in biological contexts. Notably, the recognition that two-way communication with molecular logic devices can be achieved with chemical species and light signals launched progress down this avenue. It is important to note that Fromherz had also seen a connection between pH-dependent fluorescence phenomena and electronic devices.

Small molecules, with their large diversity in structure and properties, have yielded more and more logic functions including small- and medium-scale serial and parallel integration. Arithmetic systems such as half/full adders/subtractors are appearing steadily. We note that different logic functions are being achieved by different molecular structures, rather than by differently connecting copies of a parent structure, which is the semiconductor computing paradigm. Criticism of the photochemical approach, usually from a semiconductor electronic computing viewpoint, has eased in recent years as the biological paradigm – at cellular and organism levels – has become appreciated as an equally valid computing scheme. The photochemical approach has also demonstrated strategies seen only in modern semiconductor computing, e.g. concerning reversibly or irreversibly reconfigurable logic systems. Some of these are simultaneously and multiply reconfigurable. The latter cases are superposed logic systems, which are unheard of in conventional semiconductor-based computing and are close to what is dreamt of by the quantum computing community, who are suffering their own traumas concerning previous hype.

This basic photochemical approach, though conceived for small molecules, is now being applied to large biomolecules with much success and thus drawing in the molecular biology community. The information handling capability of DNA has never been in doubt, but now it is clear that the humblest molecules, synthetic or natural, are achieving similar information handling all around us and even inside.

Adleman's oligonucleotide approach to molecular computation, which does not directly involve Boolean logic, caused quite a stir in 1994 owing to the promise of solving problems considered too hard for solution by current semiconductor computers. The natural celebrity of DNA and the opportunity for molecular biologists to get involved in information science also aided its popularity. While this topic lies outside the scope of this book, it is important to note that some subsequent examples of this approach have involved Boolean logical arguments. Currently the progress appears to have been tempered following Ellington's critique of this approach. This progress is chronicled in a series of conference proceedings. It is telling that the title of the conferences has been modified since 2009 from 'International Meeting on DNA Computing' to 'International Meeting on DNA Computing and Molecular Programming', reflecting the growing influence of molecules other than DNA in computing. Another important trend in recent years is the application of the concepts of small-molecule Boolean logic to DNA and RNA, so that there is plenty of momentum but not along the original Adleman avenue. The structural capability of DNA has also permitted a tiling approach to molecular logic and computation.

The discussion above has shown that, as small as they are, molecules have many logic and computing tricks up their sleeves. A world map showing the laboratories involved in this enterprise (Figure 1.3) brings us almost up to the state of the present day.

Chemistry and Computation

2.1 Introduction

First of all, this book is not about computational chemistry. That is the subject that depends on computer programs based on quantum theory running on semiconductor-based hardware to provide information about atoms, molecules, materials and reactions. The more recent availability of density functional methods has made such calculations more accessible and useful to experimental chemists. Computational chemistry also includes methods which only require hand calculation to produce powerful insights.

On the other hand, this book is about molecules and chemical systems which possess innate ability to compute, at least in a rudimentary way, like machines based on semiconductor transistors (or magnetic relays or mechanical abacuses in the past) or like people. Although the field is only 19 years old, it already has applications (see Chapter 14) which are uniquely useful.

Chemists have grown up with the slogan 'Chemistry is the central science'. If chemistry meshes more with the spectrum of disciplines, the more central a science it will be. Meshing means, at a minimum, to engage with the concepts of that discipline and, more hopefully, to solve problems in that field. Given the atomic and molecular basis of living matter, it is not surprising that biology and medicine have received large contributions from chemistry. Even civil, mechanical and electronic engineering benefit from chemistry-based materials development. On the other hand, mathematics and computer science have not benefited from chemistry as much. Of course, chemistry has contributed to the material aspects of computer science, e.g. the production of pure semiconductors followed by controlled doping. An exception is the chemical manifestation of topology. Molecular logic and computation can help fill this gap.

It is clear that the modern information technology revolution has permeated society. However, there is a much older and more vital information revolution. This is the one operating within our genes, cells, nerves and brains. Chemical processes start and drive it, even though its complexities take us into the realms of biology. As the semiconductor-based revolution gradually deals with ever-smaller features, molecules become ever more attractive as information processors. Indeed, small molecules easily go to small vital spaces where semiconductor devices fear to tread. It therefore becomes a responsibility for chemists to explore the information-processing capabilities of molecules.

This should not be difficult, because chemists have been exposed to molecular information processing since high school. Many chemistry experiments involve the exposure of a compound to a reagent and/or heat. Similar operations, perhaps with less quantitation, occur in kitchens around the world several times each day. Physical organic chemistsconsider the key compound to be the substrate. The progress of the reaction is seen by the change of some visible property to that of the product, such as the colour. This response of the compound to the stimulus can be appreciated in a different way (Figure 2.1). To borrow a computer scientist's language, an input is applied to a molecular device so that an output will result.

Of course, the outputs of many simple semiconductor logic devices are continually controlled by the inputs. Therefore, reversible chemical reactions are the best suited to be interpreted in a computational manner. Otherwise such a molecular device would have the output frozen in one state even though the inputs undergo subsequent changes. However, there is another class of simple semiconductor logic devices which can be locked in a given output state. Some of these have been addressed by chemists, and we will discuss these in Chapter 11.