All-optical processing with molecular switches

Abstract

A gradual transition from electrical to optical networks is accompanying the rapid progress of telecommunication technology. The urge for enhanced transmission capacity and speed is dictating this trend. In fact, large volumes of data encoded on optical signals can be transported rapidly over long distances. Their propagation along specific routes across a communication network is ensured by a combination of optical fibers and optoelectronic switches. It is becoming apparent, however, that the interplay between the routing electrical stimulations and the traveling optical signals will not be able to support the terabit-per-second capacities that will be needed in the near future. Electrical inputs cannot handle the immense parallelism potentially possible with optical signals. Operating principles to control optical signals with optical signals must be developed. Molecular and supramolecular switches are promising candidates for the realization of innovative materials for information technology. Binary digits can be encoded in their chemical, electrical, or optical inputs and outputs to execute specific logic functions. We have developed a simple strategy to gate optical signals with optical signals by using a photoactive molecular switch. We have demonstrated that NAND, NOR, and NOT operations can be implemented exclusively with optical inputs and optical outputs coupling from one to three switching elements. Our remarkably simple approach to all-optical switching might lead to the development of a new generation of devices for digital processing and communication technology.

Optical communication networks transport hundreds of gigabits per second over hundreds of kilometers (1, 2). The efficient transmission of data relies on the integration of optoelectronic devices and optical fibers. They are responsible for the generation, amplification, conduction, routing, and detection of optical signals. At each step, the continuous interplay between optical and electrical stimulations controls data processing. In particular, electronically controlled switches guide the traveling optical signals along programmed routes across the communication networks. Signal routing requires multiple optoelectronic and electroopitc conversion steps. From input fibers, the propagating optical signals reach the routing devices, where they are converted into electrical signals, processed in this form, reconverted into optical signals and, finally, directed to specific output fibers. These inevitable optoelectronic and electroopitc conversions are a major “bottleneck” to the development of optical networks (3, 4). Besides the obvious loss in signal intensity associated with each conversion step, these switching devices are relatively slow and can process only few signals simultaneously. Only a tiny fraction of the potential transmission capacity of optical fibers is used in present communication networks. These limitations are stimulating considerable research efforts to develop innovative operating principles for optical switching (5–8).

A variety of clever and efficient strategies to completely eliminate undesired optoelectronic and electroopitc conversion steps have already been identified (5–8). It is now possible to maintain the propagating signals exclusively at the optical level. However, these methods rely heavily on electronics and, presumably, they will not be able to support the terabit-per-second capacities that will be needed in the near future. Their major limitation is that the switching operations are still controlled by electrical stimulations. The traveling optical signals are routed in response to electrical signals, which cannot handle the immense parallelism offered by optical fibers. Strategies to control optical signals in response to optical signals are potentially much more attractive. However, the identification of reliable operating principles for all-optical switches is a challenging goal.

Molecular switches (9–12) are emerging as alternative materials for digital processing (13–15). Simple logic functions (16) have been implemented at the molecular level, operating individual molecular switches (17–32) or ensembles of communicating molecules (33–35). These chemical systems elaborate chemical, electrical, and/or optical signals with logic algorithms that are dictated by the design of the molecular components. In this article, we demonstrate that arrays of independent photoactive molecular switches can execute NAND, NOR, and NOT operations (16) that rely exclusively on optical inputs and optical outputs.

Methods

The spiropyran derivative SP (see Fig. 1) was synthesized by following a published procedure (32). MeCN was purchased from EM Science and distilled over CaH₂ under N₂. The optical networks shown in Figs. 2–4 were assembled in the sample compartment of an absorption spectrometer (Varian Cary Bio 100). Each cell was filled with a solution of SP (in MeCN, 10⁻⁴ M) and was irradiated for 5 min at 254 nm (Mineralight UVGL-25 lamp; Ultraviolet Products). The output intensity was measured immediately after, rather than during, the application of the optical inputs.

Figure 1.

UV light, visible light, and H⁺ induce the interconversion between the three states SP, ME, and MEH. Absorption spectra of SP (in MeCN, 10⁻⁴ M, 25°C) recorded after (a in Lower) and before (b in Lower) irradiation at 254 nm for 5 min.

Figure 2.

An optical signal (yellow arrows) travels from a light source to a detector after passing through a quartz cell containing a solution of SP (MeCN, 10⁻⁴ M). The intensity of the optical output (O) switches between low and high values as the optical input (I) is turned on and off. This signal transduction behavior is equivalent to a NOT operation.

Figure 4.

An optical signal (yellow arrows) travels from a light source to a detector after passing through three quartz cells containing a solution of SP (MeCN, 10⁻⁴ M). The intensity of the optical output (O) switches between low and high values as the three optical inputs (I1, I2, and I3) are turned on and off. This signal transduction behavior is equivalent to a three-input NOR operation.

Results and Discussion

Recently, we have developed a three-state molecular switch (Fig. 1) that responds to chemical and optical inputs to produce optical outputs (32). The colorless spiropyran state SP switches to the purple merocyanine form ME on irradiation with UV light. Alternatively, SP switches to the yellow-green protonated merocyanine MEH on acidification. The colored forms ME and MEH return to the colorless state SP when they are irradiated with visible light. Similarly, ME reisomerizes to SP when stored in the dark. The addition of acid to ME produces MEH, and the treatment of MEH with base restores ME.

The direct interconversion between the two states SP and ME relies on the photoisomerization of a spiropyran chromophore (36–41). This switching process is controlled solely by optical inputs, and the two isomers have distinct absorption properties in the visible region (32). In MeCN, the purple-state ME has a strong absorption band at 563 nm (a in Fig. 1 Lower) and can block an incident optical signal having this particular wavelength. The colorless state SP, instead, does not absorb at 563 nm (b in Fig. 1 Lower) and can release the same signal. The photoinduced transformation of SP into ME is extremely fast and involves the formation of colored intermediates (42). Time-resolved laser spectroscopy has demonstrated that, in the case of the parent 6-nitrospiropyran, colored species can be detected within 10 ps from irradiation. Thus, an optical input addressing SP can be transduced, at least in principle, into an optical output on a picosecond time scale. It follows that this ultrafast all-optical molecular switch is particularly attractive for digital processing at the molecular level.

In the optical network illustrated in Fig. 2, a monochromatic optical signal (yellow arrows) travels from the visible light source to the detector passing through a quartz cell. The wavelength of this signal is 563 nm and corresponds to the absorption maximum of the purple-state ME (a in Fig. 1 Lower). The cell contains a solution (MeCN, 10⁻⁴ M) of the molecular switch. The irradiation of this solution with an UV light input (red arrow) induces the interconversion of SP into ME. The isomerization of ME back to SP occurs thermally with a first-order rate constant of (41 ± 1)⋅10⁻⁴ s⁻¹, if the input stimulation is turned off (32). When the molecular switch is in the “nonabsorbing” state SP, the intensity of the optical output reaching the detector is 100%. However, it fades to 3% when the molecular switch is in the “absorbing” state ME. Thus, the optical output (O) switches between low and high values as the optical input (I) is turned on and off (lower left table in Fig. 2). It is worth noting the analogy between this all-optical switch and conventional field-effect transistors (43). In a basic complementary metal oxide semiconductor (CMOS) field effect transistor inverter, for example, a voltage output switches between low and high values as a voltage input is turned on and off. Of course, the switch in Fig. 2 processes optical signals, whereas a CMOS inverter elaborates electrical signals.

The optical network illustrated in Fig. 3 incorporates an additional switching element. In this instance, the traveling optical signal (yellow arrows) has to pass through two quartz cells before reaching the detector. Both cells contain a solution (MeCN, 10⁻⁴ M) of the molecular switch and are addressed by independent optical inputs. When the molecular switches in both cells are in the nonabsorbing state SP, the intensity of the optical output reaching the detector is 100%. However, it fades to 3–4% when the molecular switch in one of the two cells is in the absorbing state ME. The output intensity drops to 0% when both switching elements are in the absorbing state ME. Thus, the optical output (O) is high only when both optical inputs (I1 and I2) are turned off (in Fig. 3 Lower Left).

Figure 3.

An optical signal (yellow arrows) travels from a light source to a detector after passing through two quartz cells containing a solution of SP (MeCN, 10⁻⁴ M). The intensity of the optical output (O) switches between low and high values as the two optical inputs (I1 and I2) are turned on and off. This signal transduction behavior is equivalent to a two-input NOR operation.

Following a similar approach, all-optical networks with n input terminals can be realized introducing n independent switching elements between the visible light source and the detector. The resulting array of switches transduces 2ⁿ strings of n optical inputs (I1–In) into a single optical output (O). However, it is sufficient to address only one of the n switches to block the traveling optical signal. The intensity of the optical output O is high only when all of the input stimulations I1–In are turned off. As an example, a simple array with three switching elements is shown in Fig. 4. The output intensity is 100% only when all three switching elements are in the nonabsorbing state SP. If only one of the three switching elements is converted into the absorbing state ME, the output intensity fades to 3–4%. When two or all switching elements are in the absorbing state ME, the output intensity drops to 0%. Thus, the optical output (O) is high only when all of the optical inputs (I1, I2, and I3) are turned off (Fig. 4 Lower Left).

The tables at the bottom left of Figs. 2–4 summarize the signal transduction behavior of the three all-optical networks. In all instances, the input signals can be either off or on. Binary digits can be encoded on all inputs, applying a positive logic convention (off = 0, on = 1). Similarly, a logic threshold can be set arbitrarily at 5% for the output intensity. The values above and below can be defined high and low, respectively. Now, binary digits can be encoded on all outputs, applying, once again, a positive logic convention (low = 0, high = 1). Following these assumptions and conventions, the signal transduction of the all-optical switch in Fig. 2 translates into the truth table of a NOT gate (16). The output O is 1 when the input I is 0 and vice versa. The signal transductions of the all-optical networks in Figs. 3 and 4, instead, translate into the truth tables of a two-input NOR and a three-input NOR (16), respectively. The output O is 1 only when all of the inputs are 0. All of the other combinations of two- and three-input strings are converted into a 0.

Comparison of the combinational logic circuits illustrated in Figs. 2–4 reveals that the addition of one switching element along the path of the traveling optical signal (yellow arrows) corresponds to the addition of one OR gate in the circuit. In fact, the transition form the one-cell system (Fig. 2) to the two-cell system (Fig. 3) corresponds to the integration of an OR gate at the input terminal of the NOT gate. Similarly, the transition from the two-cell system (Fig. 3) to the three-cell system (Fig. 4) corresponds to the connection of a second OR operator to one of the two inputs of the first OR gate. It follows that a sequence of n cells along the path of the master signal (yellow arrows) is equivalent to a combinational logic circuit incorporating a sequence of n − 1 OR gates terminated by one NOT operator. The truth table of this hypothetical logic circuit is that of a n-input NOR gate.

In the examples of Figs. 3 and 4, two and three switching elements are operated sequentially. The master optical signal (yellow arrows) has to pass through all switches before reaching the detector. In principle, however, alternative configurations can be considered. For example, two switching elements can be operated in parallel to implement a NAND function (16). Two independent visible light sources can send two optical signals to the same detector. The optical output is the sum of the intensities of the two signals. The path of each signal can be intercepted by one quartz cell containing the molecular switch. Thus, the two merging optical signals are switched independently, operating two switching elements in parallel. The signal transduction of this parallel array can be predicted on the basis of the behavior observed for the one-cell system in Fig. 2. A low output intensity can reach the common detector only when the input stimulations of both switching elements are on. The two merging optical signals are blocked by the absorbing states of the two switching elements. In the other three cases, the output intensity is high. One or both merging signals can reach the detector unaffected. Following the same logic assumptions and conventions applied to the three systems in Figs. 2–4, the signal transduction of this hypothetical two-cell system translates into the truth table of a NAND gate (16). The output O is 0 only when both inputs are 1. It is 1 for the other three input strings.

Conclusions

The photoinduced isomerization of a spiropyran chromophore can be exploited to block and release a monochromatic optical signal falling in the visible absorption range of the photogenerated isomer. This simple operating principle can be used to gate optical signals in response to optical signals. We have demonstrated this hypothesis, implementing logic operations on arrays of switching elements that rely on the interplay of optical inputs and optical outputs. Our chemical approach to all-optical processing is extremely versatile. Independent all-optical switches can be configured sequentially and/or in parallel to satisfy a variety of digital designs. In addition, the ability to reproduce optically the universal functions NAND and NOR might have fundamental implications in optical computing. Virtually any logic algorithm can be implemented integrating exclusively NAND or exclusively NOR gates (16). In fact, these two universal operators are the dominant logic components of digital electronics (43). It is worth noting, however, that practical applications can emerge only after a substantial amount of further fundamental studies. Crucial parameters, including the switching speeds and potential photodegradation of the molecular components, must be carefully assessed. In any case, our experiments certainly demonstrate that all-optical processing can be implemented with simple molecular switches.