Architectures

The array simulations are a great way to produce a basic quantitative estimate of the behavior of the artificial muscles with reasonably accurate values for the dimensions. However, in the end, these structures need to be practical and manufacturable. As a result, specific microfluidic architectures needed to be developed, for the prototyping of the actual devices.

To illustrate this point, notice that the COMSOL simulations took the microcapacitor plates as disconnected structures essentially floating inside the bulk material. This is a good simplification for the simulations, but a practical working muscle requires electrical wiring to every plate, to function as designed. To achieve this, our basic approach planned on using microfluidic channels built by 3D printing, emptied, and then filled with conducting material, to form the electrical wiring. Hence, both 3D printing and microfluidics imposed their own requirements on the necessary structures.

Microfluidics requires that there is fluidic access to each plate. However, practicality requires that the number of input/output ports on the overall device is kept to a minimum. For example, if the artificial muscle is built with 10,000 microcapacitors, it is not practical to require 20,000 inputs to be able to apply voltage to every capacitor in the array. The simple solution then is to array the capacitors fluidically as well. 

To accommodate, we came up with the idea of dividing the artificial muscle into muscle fiber bundles, and then dividing each bundle into individual fiber. Each fiber would need two polarities connected in an alternating fashion to a column of stacked parallel capacitor plates. We realized then that all the plates of the same polarity can be part of the same microfluidic channel. Then just two channels would be sufficient to form each fiber. 

3D-printing imposed requirements as well. The printer cannot print over empty space, so it uses sacrificial material wherever a cavity is to be formed inside the resin, e.g. a microfluidic channel. After printing, the sacrificial material must be removed. That is done fluidically, which means that the architecture should facilitate that. For example, there can be no dead ends allowed in the channel matrix, because then it would not be possible to remove the sacrificial material from the dead end. This led to the idea that the entire channel forming one polarity within the same muscle fiber must by necessity be a single channel with no splits to dead ends. 

Combining this requirement with the need to attach to and clear rectangular plates led us to the idea of connecting to each plate along diagonally opposite corners of the square. This would ensure the best possible clearing as there will be no dead ends. Next, because there are two different alternating polarities, we would have to use the other set of diagonally opposite corners for the other polarity. 

These ideas were assembled into the basic pattern of the microfluidic architecture.
 

On the diagram, structure A shows the basic element of the array, include two positive and two negative plates. Notice that the red plates are always connected through their top right and bottom left corners, while the blue plates are always connected through their bottom right and top left corners. 

Structure B shows how the basic structure can be copied and pasted to put together a single muscle fiber. Note that in that fiber, there are only two inputs and two outputs for an unlimited number of microcapacitors. 

Fibers must be combined in muscle fiber bundles. Moreover, within a bundle, the fibers must act in the same direction to add force. Hence, they are positioned in parallel to the longitudinal axis of the muscle. Structure C shows two parallel fibers. Their inputs and outputs can also be combined by same polarity, which means that both fibers only require two common inputs and two common outputs. 

The same arraying can be done in both lateral directions. Structure D on the diagram shows 4 muscle fibers arrayed in a 2x2 formation and all coupled to just two inputs and two outputs. 

This approach to arraying can be generalized to produce a muscle fiber bundle of NxM fibers, so long as both N and M are powers of 2, i.e. N=2^n and M=2^m. 

This double-helix architecture was successfully patented.