Say a word about liquid metal. Thoughts on the hardware and software implementation of the T-1000

At temperatures close to room temperature, several metals are in the liquid state: cesium, melting point = 28.5 ◦C, francium = 27 ◦C, rubidium = 39.3 ◦C, mercury = -38.8 ◦C and gallium 29.8 ◦C. At the same time, mercury is very toxic, cesium and rubidium are too reactive, and francium, moreover, is radioactive and occurs in trace amounts. Compared to all these substances, the toxicity of gallium is minimal, in addition, its alloys with indium and tin are chemically stable. The special properties of gallium alloys, along with those mentioned above, are photothermal and photodynamic characteristics, as well as response to external stimuli and catalytic properties. Therefore, a hardware analog of a neuron can potentially be made from a gallium alloy. Also, such machines can be used in microfluidics, tomography, detection of cancer cells, elimination of vascular embolism.

But let us return to the fact that the controllability of gallium (and also its alloys) increases in narrow tubes. In such confined spaces, the alloy remains in a liquid state, and also reacts to magnetic and electrical influences, and even to light. That is why gallium alloys are promising for the production of micromachines. Currently, one of the main difficulties in the design of such devices is to ensure that they move autonomously in narrow channels to their destination and, upon arrival, perform relatively complex tasks, at least the delivery of an active substance. In such channels, Galinstan micromachines would move much faster than their solid counterparts and could even accelerate and change their direction of movement under the influence of a magnetic field. The narrower the channel, the faster the galinstanovka machine can move in it; establishedthat such a phenomenon is due to electroosmosis. As the forces that ensure the movement of a liquid micromachine in a narrow channel, there are known, for example, acceleration using hydrogen bubbles, pressure, ion gradient, ultrasound, ion and magnetic field. It has been proven that in an alkaline solution (NaOH), liquid metal gallium machines under the action of an electric field move to the cathode. They can be accelerated by expanding the channels along which they move, and guiding, deforming these channels as needed.

Nevertheless, such a movement is not entirely complete, since it requires constant external influence and is feasible only in laboratory conditions. The situation is complicated by the fact that nanoscale machines are forced to overcome the surface tension of the liquid, which, at their scale, significantly restricts movement. Therefore, the next generation of liquid nanomachines should not only independently extract energy for movement, but also be overgrown with a protective layer that will allow them to function longer in solutions with a changing acid-base balance.

Self-propelled micromachines

Synthetic self-powered motors, capable of spontaneously converting chemical energy into mechanical activity, thereby providing autonomous locomotion, would be perfect for creating miniature robots with sensor or detector functions. Based on galinstan designed micromotors of millimeter and centimeter sizes. Such machines float in a round Petri dish or in narrow channels with a different structure, developing a speed of up to several centimeters per second, moreover, they remain operational for up to 1 hour without an external source of energy. The metal is easily deformed and restores its shape, but, in addition, the engine exhibits “biomimetic” properties, bringing it closer to a mollusk. Just as a mollusk absorbs silicon by overgrowing a shell, gallium is amalgamated with aluminum. The activity of this process depends on several factors, including the volume of the engine and the aluminum content in the solution (solutions of sodium chloride or sodium carbonate are used for such fouling). In an alkaline solution (for example, sodium hydroxide), the aluminum layer erodes, hydrogen bubbles are released, which also ensure the movement of the micromachine. Nevertheless, in the currently available gallium micromachines, such a movement remains similar to Brownian, that is, uncontrollable. To give the desired vector to such a movement, the micromachines still need to be guided from the outside – for example, with the help of a laser. Naturally, for a machine to react to a laser, it must have light-sensitive elements. The combination of gallium alloys with light-sensitive compounds such as titanium dioxide, brings us to the next interesting aspect: it turns out that a liquid metal surface can exhibit the features of a “hardware neuron”.

Tactile Liquid Metal Components and Muscle for Robots

Based on liquid metal, the robot can be equipped with light-sensitive and tactile functions. Thus, it has been shown that it is possible to embed a network of tubules filled with a liquid alloy into an extensible silicone carrier – and to achieve that, when heated, this material changes color. A similar color change occurs in response to mechanical pressure. This primitive logic is like the one by which the octopus changes colorby responding to external stimuli. The skin of an octopus is riddled with a lot of nerves, and for him discoloration is a camouflage; a soft robot, in turn, can change color depending on the action being performed. It has been proven that skin discoloration in an octopus is not regulated by the brain; it is precisely the reaction of neurons to the incoming signal. The materials from which soft robots are made do not have electrical conductivity, but liquid metal drops, on the contrary, conduct both electricity and heat. The gallium-containing filling can react to both the gripping force and the shape of the object gripped by the robot. It is possible to mix gallium-indium alloy into the polymer already at the stage of manufacturing the parts for the robot. Initially, it is concentrated in the form of droplets, but in response to mechanical action, the droplets line up in a grid, like neurons. If cracks or holes appear in a polymer material, then the “neural network” spontaneously regroups, and the material retains its electrical conductivity. Moreover, the liquid metal elastomer can be used to produce muscle-like structures that not only change and maintain the shape required for work, but also return to their original state when heated. If you act on the gallium component of such a material with electricity, then it changes shape as required by the operator.

Sense of quorum

Finally, we return to the remark that liquid metal machines are almost Roy; they can work in concert if they have sensors for this purpose. Multi-agent systems of this kind can collectively perform complex tasks, in particular, build or search for something. Direct and indirect methods of coordination allow robots to exchange information, dynamically adjusting to changing situations. This behavior has a well-known (micro) biological analogue, the so-called “quorum sense” in bacterial films. Once in a nutrient medium or surrounding a specific cell, bacteria exchange chemical signals, thanks to which the entire colony or biofilm solves a common problem. This mechanism of intercellular communication allows each bacterium to estimate the size of the population (how many of us are there) and act in accordance with this information. Nanoscale robots with similar swarm intelligence could reproduce similar behavior in precision manufacturing or medicine. By the way, bacteria, united by a sense of quorum, often pose an additional danger, so microbiology carefully studies just suppression this mechanism (quorum quenching). Let us consider how to transfer this mechanism to a swarm of robots, in particular, how to implement in hardware an analog of signal molecules (autoinducers).

Apparently, the chemical communication of bacteria, acting only over short distances, in a swarm of robots could be realized using near field communication (NFC), that is, using radio signals. But already in 2006 it was supposedthat nanorobots operating in a liquid medium could also rely on (electro) chemical interactions if each agent carried a signaling molecule that serves as its beacon. When diffused in the environment, such robots could both concentrate and disperse, dynamically changing the density of the swarm and seeping through obstacles. If at the same time the swarm learned from previous experience on the basis of an evolutionary algorithm, then the robots could “vote” for this or that decision, as well as “vote” to decide whether the necessary concentration has been achieved to perform a particular operation, or if additional forces need to be tightened … Also, the sense of quorum allows you to take into account the frequency of incoming messages, and on the other hand, increase or decrease the activity of messages. Finally, robots in a swarm could, at the level of a sense of quorum, assess the energy state of the entire swarm and exchange charges if some agents begin to experience a lack of energy. On the other hand, the same algorithms could also implement joint suppression of the sense of quorum, so as not to block each other, or to prevent some robots from being cut off from the main part of the swarm.


Here I will not dare to fantasize about what size the smallest drop of the T-1000 could be, possessing all the properties of its polysplav and, accordingly, being a full-fledged robot. This can probably be due to the minimum possible size of the transistor (this is described in article, the translation of which may appear on the @Sivchenko_translate blog). In any case, this small excursion into the physics of liquid metal narrows well the range of hypotheses that explain many of the properties of the T-1000, in particular, its thermal and chemical weakness. It would be interesting to assume that this model could not only be alloyed with scandium or molybdenum to acquire sufficient refractoriness and sharpness of the cutting edges. The main difference between most of the described samples from the T-1000 is that a carrier medium is needed for their functioning, and the energy supply of a liquid metal robot still leaves much to be desired (the robot requires regular or constant feeding). Now I believe that using the T-1000 as an example, we see a hardware implementation of a complex neural network and a nanoscale swarm of robots at the same time, which once again makes us wonder where our technologies can lead us.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *