Thin Film Effects
I was recently contracted to do a journal cover for an accepted paper. Not long after finishing it I had an offer for a second. Both touch on a subject I'm particularly fond of - thin film devices. Given that the articles in question aren't out yet I won't be posting any of the graphics here. Instead I wanted to discuss the design of the second cover which addresses different deposition methods for organic solar cells.
Specifically, it focuses on blade coating, though spin coating, screen printing, drop casting, and all kinds of other techniques would be equally relevant here. A short description of these processes is that they are used to deposit some kind of "ink" - a specific molecule or mix of molecules dissolved in a solvent or mix of solvents - on a surface (usually called a substrate). With organic solar cells the layers are usually less than 300 nm thick, which is approximately 250 times smaller than the width of a human hair (a terrible metric to use since human hairs vary enormously in thickness).
The main goal of any coating process are very simple. Deposit the desired film thickness as uniformly as possible on the chosen substrate (glass, plastic, etc). Ideally the combination of molecules, solvents, and deposition conditions will give the best possible performance with the least amount of material wasted. For certain techniques there are also considerations of how fast you can deposit it (X meters or film per second). However, not every technique is created equally. Below is a bullet point comparison of spin coating and blade coating, the two techniques compared in the study I'm currently drafting a cover for.
Spin coating:
Gives uniform films with variable thickness (down to a few nm)
Gives control of drying effects through spin speed and duration
Suitable for small substrates
Modest instrument cost
Smaller instrument fits in most labs
Very widely used in academic research
Wasteful (most material is 'thrown off' the slide during the process
Blade coating;
Useful for thin films (I would say 50+ nm)
Less wasteful than spin coating
A precursor to true "Roll-to-roll" coating (think newspaper printing)
Better suited to larger area films
More expensive and larger instrument
Among the more widely used "large scale" techniques in academic research
Not as 'tuned in' for making high efficiency organic solar cells as spin coating
Now, on to the graphics side of things. One thing that is shared by both these techniques is that you're making a film on a surface. In spin coating the liquid falls on the spinning substrate. In blade coating a 'blade' drags the liquid across the substrate to create the film. And this is where we hit our graphics challenge. Both of these involve a liquid on a surface interacting with an object. Either it's a liquid interacting with a spinning surface or a liquid interacting with a moving blade.
At a very simple level you could make this figure, or even an animation through trickery. I've done both and they're showcased below. Both of these rely on animating layer thicknesses to "grow in" while the deposition process is occurring. Both are great for animations. However, as still figures they are… well, still. These are fine for in-paper figures (I actually prefer the simplicity in that setting) but there is an undeniable static lifelessness to them. So how can we add some pizzazz for a journal cover?
The most obvious way is fluid simulation. It's easily the coolest way as well. A big splash of liquid captured on a surface. It adds in that dynamic element, even in a still frame. Have the liquid whipping out to the sides for spin coating or have it getting dragged along for blade coating. Perfect really. There's just one issue. It relies on fluid simulation, which is slow and computationally intensive.
I generally don't recommend effects to my audience if I don't think they could be easily or quickly done on more modest hardware. For my own professional work, I may end up using a fluid simulation to get an initial approximation and then sculpting to get the final effect. I may just use a cloth simulation and then add thickness. In either case, it’s not very modular and it’s not necessarily the greatest thing to teach. Setting up simulation parameters, choosing effectors, making sure the dimensions are all exact so the physics engine doesn't behave unusually. There's a lot that can go wrong.
Alternatively, we have procedurally geometry, a somewhat more workable option. With geometry nodes in Blender we can easily accomplish this type of look. Want to iterate through designs by adjusting the rotation and scale of a point cloud? Want to update in essentially real time without having to re-bake a simulation? Want to add in effectors to locally change the scene or build in custom randomness? Geometry nodes are a good solution here. Combine instancing with metaballs or with a mesh to volume - volume to mesh combination and we're in good shape. Procedural materials could also work, but that's a bit less intuitive in my opinion.
As I work more on this cover I'll provide further updates here but this is my current thought process for making more 'dynamic' thin film images. If the procedural side wins out it will likely make it into a full showcase on the channel. If the fluid simulation side wins out it will end up getting a full dev blog post breaking it down.
