At the end of the day, a function like DES is a wave function, as in quantum mechanics. That is, the plaintext, the ciphertext and key are 3 wave functions. As with any wave function, each is just two bit-vectors:- one for the real component and one for the imaginary component and time t (where t is the next bit in the bit-vector..). Rather than think of bits as 0 and 1, use –1 and 1 (with zero crossings built in). One has the classical mental picture now a wave function oscillating away in two phases.
Of course, we know that any such wave function can be broken down into its Fourier components. Consider now that each round of a 16-round DES calculation is (conceptually) one of those Fourier components. This is not particularly far fetched, as we know that single-round characteristics propagate as multi-round characteristics (in DES). So its not only that the rounds “add” (to make the final DES ciphertext output), they are adding in the spectral domain too. This preserves the characteristic, which supports guessing.
Now turn to nuclear chemistry. The point of the the fission process in a controlled nuclear explosion is to detonate the lithium. Lets now assume that one can have a controlled detonation of the latter (rather than exploding a hydrogen bomb). The point is to create an environment allowing for the representation of huge number of combinatoric possibilities. One wants an actual space in which every possible combination of DES plaintext and ciphertext, for every difference pair, can simultaneously exist.
Since the number of atoms in the universe is relatively small, we have to go sub-nuclear to find such a space. We have to go into the wave functions of electrons themselves; which are “combinatoric” spaces to start with. In the smeared-out quantum state machine, we have all the possibilities, assuming we model each des difference bit (for plaintext or ciphertext) as a particular electron.
Now, just like colossus was tuned up the the signal/noise ratio of the tunny cipher and essentially flashed orange when the detector detected a case distinct from uniform (i.e. a snigificance test passed, suggesting a subkey), so we want a thermonuclear reaction that “flashes orange” at some point in the evolution of its quantum state machine – to tell us that the subkey was found. Of course, we want to set the initial conditions so that reaction A, vs reaction B, vs C…each correspond to a particular bit of the subkey.
Now one of the things NSA school of cipher making is good at is subkey schedule design. But its also quite good at the reverse – which is taking a implication of a known subkey and using that as a constraint in a second-level attack on the key. This means we want two classes of thermonuclear cryptanalysis: one for difference pairs (that constrain the stats), and one for the subkey constrains used the error-correcting decoder that deduces the key.
Now why do we want thermonuclear reaction vs. merely a fission reaction? Well we want the weight constraints of the cipher design to be built into the chemistry itself – so their constraints – that are designed to diffuse the probabilities when calculating in the digital space – become a way of controlling the evolution of the quantum machine represented by the instance of reaction. One wants the sequence of heat state of the reaction, as energy is released by the detonation of the lithum, to evolve through the weight states of the math, that is.
So if we now think about what kind of machines can practically exist one thinks of a magnetic resonance machine observing not your insides but a second reaction. It exists to image “slices” of the reaction, in much the same way that any measurement projects the quantum space onto the projective plane associated with one of the eigenvalues of the operator (the magres machine’s chosen slice). Since the reaction cannot be repeated or controlled much, one presumably wants several magres operators measuring simultaneously, sampling different eigenspaces. Of course, each must do this in a manner that is so correlated with the core reaction that it – by itself or in combination with other imagers – does not influence the quantum evolution.
Now we have seen folks build such machines – think of CERN particle colliders – with the view to so speeding the electron particle that, in better thinking, one has fully evolved the wave equation of its quantum lifecycle, allowing one to see within and see its own constituent makeup – having “flashed orange”.