Here are my presentation boards for my final outcome. For an overview of the entire process, click here. For a brief explanation, follow along.

In a nutshell, my project worked with a process called reaction diffusion. A process which is generated by mixing Bromium and an acid. In doing so, the chemicals interact to form incredibly mesmerising patterns. and these patterns formed the basis for generating an architectural form.

The resulting architecture was something which exhibited a growth-like aesthetic, forming a coral-like looking structure. The benefit to using reaction diffusion for this process was the amount of variation that could be achieved by ever so slightly adjusting certain parameters in the initial set up. The final output, perhaps lacked architectural definition, but there is certainly a wealth of other possibilities that could be afforded to us by further exploring the possibilities of this phenomenon.

I’ve finally been able to garner more control over the reaction diffusion process, I can now control the seed number, chemical origin, speed of the reaction, scale of the reaction, the Flow and kill rate, the size of the mesh output, and the iteration steps between successive frames. In the model below, I worked with a simple polar array to create the form. But having done this, the possibility for more complex arrays is certainly achievable, and this takes me one step closer to realising architectural form.

I went back to an earlier method of reaction diffusion whereby the model is created in 3 Dimensions through a method of time based extrusion, which creates a model layer upon layer of the current reaction diffusion frame. I did this 3D print at a very fine scale, so the result came out super clean! What was also great about this 3D print is that it required no structure whatsoever in order to print, so it produced a very very clean model!

Last week I got to work on creating a catalog of hundreds of possibilities of extracting architectural quality from a reaction diffusion model. The catalyst for this was a good find in the u-skate world. I found a lower subset of the reaction which in my mind seemed to have a lot more architectural possibility in it due to the large spaces that opened up inside of it. (Found in the Grey Scott Model at F = 0.0140, k = 0.0450)

Here is the catalog I produced for the presentation. In it, I’ve tried to rationalise specific sectional cuts of the input models into the different architectonics they could possibly represent:

  • passages
  • archways
  • windows
  • levels
  • spaces
  • cantilevers

The hope in doing this was that it would give me a clearer grasp on the architectural possibilities available inside this container of ideas, and that I could make attempts to composite some of these variations into an architectural solution. The only problem I have with doing this is that it takes away from the purely logical process which we’re working with in my current design paper, and becomes a more manual process.

Similar to my last post, I did another animation to inspect a different reaction diffusion mesh result (see below). I think the last one certainly had stronger moments that can be turned into architecture, and this one tended to provide more channels and pathways as opposed to the one previously which tended to open up more spaces.

I’ve realised as of late that the parametric process, while being a wonderful exploration and visualisation tool, is hardly ever going to produce architecture directly as an output, and this is likely because architecture is determined not by a deductive method, but rather an inductive method of reasoning. Architecture relies on abstractions – the designer. For the most part, at some point the designer needs to either enter the realm of the building, or form needs to exit the speculative realm and become something rational. That’s not to say that it can’t be done, but nine times out of ten, in order to turn a process into something tangible, one has to deviate from the logical process. To borrow a line out of The Terminator, “What is it that makes us human? It’s not something you can program. You can’t put it into a chip… The difference between us and machines.” Machines do not have the ability to make assumptions, that’s an inherently human quality, and that’s where the divide lies between process and architecture.

Hence, at this point, I need to step in and begin to manipulate the form on my own, over the next few days I will be cataloging my own forays into what can be done with the form, to begin doing manual iterations and architectural abstractions. The way I imagine the process going at the moment is that my architectural programme for the building is largely going to be developed as a result of the form, where I begin to populate the spaces I see as I see fit.

In this step I went back to the origins of how these models are created. They sequentially stack up iterations of reaction diffusion, and then a mesh is calculated which wraps around a given threshold value in order to produce such a mesh.

I did a similar thing earlier on when I tried to take sections through the 3D reaction diffusion models, but I’m now doing it with the earlier models, and here is the animated result.

I’ve hit a bit of a roadblock with the modeling side of reaction diffusion temporarily. Over the last week I spent a bit of time trying to understand SideFX Houdini, in order to use a definition which simulated a much more powerful reaction diffusion model. Unfortunately, tight time-frames probably mean I won’t be able to use it and need to search for alternative means.

This post is going to look into a little bit more of the theoretical side of the design (building on my previous discussion). I’ve been reading a book lately entitled How the Leopard Changed its Spots by Brian Goodwin. In a later chapter, Goodwin enters into a discussion about life at the edge of chaos. “For complex non-linear dynamic systems with rich networks of interacting elements, there is an attractor that lies between a region of chaotic behaviour and one that is ‘frozen’ in the ordered regime, with little spontaneous activity. Then any such system, be it a developing organism, a brain, an insect colony, or an ecosystem will tend to settle at the edge of chaos. If it moves into the chaotic regime it will come out again of its own accord; and if it strays too far into the ordered regime it will tend to ‘melt’ back into dynamic fluidity where there is a rich but labile order, one that is inherently unstable and open to change.”

This passage from the book I believe is the essence of the process I’ve been looking into. It speaks to the essence of emergent order. This also begins to tie in with Wolfram’s theories of Classes of complexity (1986). There are four classes:

  • Class 1 – patterns evolve into a stable state – all randomness is purged
  • Class 2 – Most of the patterns evolve into a stable or oscillating state – most of the randomness is filtered out, but some remains
  • Class 3 – patterns begin to evolve in a seemingly random manner, localised noise filters out much of the initial randomness
  • Class 4 – patterns reach extreme complexity, forming localised clusters of order, but all the while remaining chaotic for long periods of time

It is thought that reaction diffusion exhibits class 4 complexity, coined by Chris Langton (Chris Langton was one of the pioneers in the field of complex interactions and activity) as, “Life at the edge of complexity.” But it is this state of complexity that reaction diffusion exists in, this system is dynamic and changeable, it has a rich pattern of activity.

This is important for several reasons. As I’ve tried to reinforce before, the reaction diffusion process does not enter an equilibrium state. Essentially the four classes can be further grouped into 2 states, a) ordered systems which settle into an unchanging state, and b) where patterns produce extremely complex activities, separated by rules that result in partial ordering. State b also refers back to the u-skate boundary, where I initially discussed this idea. Furthermore, this also reinforces the idea of dynamic stability. While dynamic stability may seem counter-intuitive at first, this rather aptly rationalises the idea. The idea that elements are free to move from an ordered to an unordered state on their own accord without exerting themselves too far in either direction perfectly describes the behaviour exhibited by the reaction diffusion systems.

At this point, I need to modify my research question a little. What I realised is that my previous question implies a rather strict dependency on reaction diffusion, which I sense I may be moving away from in the coming weeks. Hence,

How can processes like reaction-diffusion inform architectural design?

My concept for my design paper is based on an subsidiary of the Belousov-Zhabotinsky reaction, namely – reaction diffusion. If you mix certain substances together and then leave them to rest, patterns spontaneously begin to emerge.

image from flickr.com

These patterns were first noticed in the 1950’s by two Russian scientists after which the reaction is named, Belousov and Zhabotinsky. The pattern forms rings which start from an origin and make their way to the extents of the petri dish, and continue to form new rings at regular intervals. What made these reactions rather interesting is that they do not form interference patterns as would be expected from moving wave-fronts, and the chemical does not reach an equilibrium position (ie. the reaction continues indefinitely).

What is key about this process is the idea of emergence and emergent patterns. Emergence refers to how behaviours at a finer scale can influence details at a larger scale. “There is no plan, no blueprint, no instructions about the pattern that emerges. What exists in the field is a set of relationships between the components of the system such that the dynamically stable state into which it goes naturally has spatial and temporal pattern.” Ultimately the goal of emergence is to create some sort of order from an initial (chaotic) system.

So my question is,

How can architecture emerge through processes which mimic reaction diffusion systems?

And this is where things begin to get tricky. How do the dynamic properties of excitable media begin to inform a design logic. How do we represent dynamic stability in the built environment? In looking for an answer, I stumbled across a research paper entitled ‘Adaptive Morphologies‘ which began to look at this idea. One potential approach that is suggested early on is to simply grow the building out of the site, but this translates into a rather loose relationship between emergence and form articulation on site.

The paper goes on to talk about dual evolutionary strategies as being the likely way forward, ie, there needs to be a constant negotiation between the design intent and the material response. So what if we began to look at the dual evolutionary strategy in our design context? The two parties here become the form on the site and the site itself. If we look at these two parties as the mediums for our Belousov-Zhabotinsky reaction, we could likely begin to understand these as two colliding substances. So what is the intention of the form, how does the site react to its intention, and then how does the form respond to the site? This dialogue between the site and the design could be the catalyst for the emergence of the form.

Another alternative proposed by Adaptive Morphologies suggests that it is the designers job to mediate between intention and emergence. The designer could be free to tune a shape by manipulating the state space of the cells, all the while iterating through the emergent process. Through this, the designer takes on a whole new relationship with the program, and has a true balancing act on his or her hands. Of course this method does not entirely lend itself to the design process as we know it, and at this stage is something better suited to conceptualisation of shape rather than as a finished state.

The other alternative is that we could look at the emergent patterns which form on the site not directly as a model, but perhaps as a representation of movement through the building to be. Hence, rather than generating a form based on where the agents move in 3D space, we begin to look at the movement of the agents in 3D space as the mapping strategy to help define the programme of the building. The benefit of doing this is that the aesthetic of the reaction diffusion process is no longer the main driver for system. This was one of the limitations I came across (outlined in an earlier post) whereby we can’t exactly create new forms from reaction diffusion processes, all the possible forms that can be created out of this system lie on a predefined boundary.

I leave you with a quote from Lance Reddick, in the TV show fringe, “At the risk of sounding sentimental, I’ve always felt there are people who can leave an indelible mark on your soul, an imprint that can never be erased.” Can we view the relationship between form and material assemblage, or design intent and emergence as an imprint that can never be erased?

Today I realised my understanding of reaction diffusion was slightly inaccurate. I finally understood a few things about the reaction diffusion model (see previous work here). For starters, Reaction Diffusion is a chemical process in which substances react with each other and diffuse at the same time. This website contains the image below which defines an area called the u-skate world, as well as a being a good resource for understanding reaction diffusion.

(n.b. Flow rate is the rate at which chemicals are added into the system and kill rate is the rate at which particles precipitate into their ‘final’ state)

This image graphs the range of possibilities of reaction diffusion outcomes. The horizontal axis, k (kill rate) ranges from 0.05 to 0.07 units, and the vertical axis ranges from 0.04 to 0.08 units. Each box in this graph shows what happens when the coinciding kill and Flow rates meet. Beforehand, I had always plugged arbitrary numbers into each parameter, but I now realise the process is completely deterministic. I still find it very interesting that the system does not reach an equilibrium point however, as I’ve left several simulations running for days on end only to find the form continues to evolve.

So looking at the u-skate boundary, we can pull out particular Flow and kill rates as specified in the graph above in order to generate particular patterns.For instance,

  • at the upper extents of u-bounds (u because the graph is u shaped), we get a lot of long ‘worm’ formations (where F = 0.0740 and k = 0.0610)
  • as we move down through the graph, we still get a lot of worm formations, but much shorter (where F = 0.0340 and k = 0.0610)
  • on the blue fringes we get the formation of solitons (where F = 0.0340 and k = 0.0570)
  • and on the red fringes of the pattern we get a mitosis formation (where F = 0.0220 and k = 0.0610)
Reaction Diffusion Equation large
Equation of the Gray Scott Reaction Diffusion Model

The equation essentially models two substances (u and v) in a fixed volume, and these two equations return the rate of change of each substance and hence form the basis for how the form is generated. For purposes of simplicity, we can assume that all other terms apart from Flow and kill rate in the equation remain as constants.

Going back to some of the models that I’d created, and according to Andrew Adamatzky’s book, ‘Reaction-Diffusion Computers’, the ending representation is not of the two chemicals in their current ‘resting position’, rather the darker regions define where a precipitate has formed, and the white regions are a precipitate free area. As the reaction continues, the chemicals continue to diffuse through the container, forming larger areas of precipitation. However, they can also follow a phenomenon called The Liesegang Phenomenon which is defined as an oscillatory precipitation in the wake of a moving diffuse front. This means that the precipitation of the chemical is not a finite state, the chemicals can oscillate through a state of precipitate to precipitate free.

Something interesting which I came across in Andrew Adamatzky’s book is that reaction diffusion is actually being developed as a means of unconventional computing. It is agreed that naturally at some point we are going to reach the limits of what our current computers do, because currently they use serial computing methods, whereas the future could well be biological and chemical processing, as they perform predominantly via parallel mechanisms. Tomasso Toffoli once said, “A computing scheme that today is viewed as unconventional may well be because its time hasn’t come yet – or is already gone.” So perhaps reaction diffusion could form the basis for computation of the future. Till then, reaction diffusion seems to mainly be used as an art form, as I will be hoping to use it.

Here is the latest step in the reaction diffusion process (full process here). I now can produce 3D models of the system and can extract sequential sections, I just need to figure out how they fit together. At the moment it’s still not quite architectural, but I did this in order to help me inspect what’s going on at each level.

The model was made in processing, and the animation itself done in grasshopper and then composited in after effects.