Information Theory 2: What If Memory Could Travel as a Ripple?

One electron disturbs the quiet

Imagine an enclosure with two charges sitting at opposite ends. I give the first charge a gentle up-and-down motion. As it accelerates, it disturbs the electromagnetic field around it. The change spreads out, reaching the second charge a little later. That charge starts moving too. Its motion then produces another disturbance, and some of that disturbance travels back toward the first one.

For now I'm imagining a perfectly enclosed, lossless world. Nothing leaks out, nothing heats up, and nothing gets absorbed. I want to picture this exchange going on and on: one charge moves, the other feels it, then the first charge feels the response. It's a useful mental playground, even though a real setup would need its energy carefully accounted for.

Watch the little arrows carefully. Those closest to the moving charge change direction first. The arrows farther away respond later. Nothing needs to pick up the first charge and carry it over to the other side. The pattern of motion travels through space while each charge stays near its own position.

There is still the question of energy. Vibrating a charge takes energy, and radiation can carry that energy away. A carefully designed cavity can hold electromagnetic oscillations for a very long time. Two charges by themselves would need a mechanism to keep this exact back-and-forth motion going. Here I've simply assumed an ideal driver and enclosure so we can concentrate on how the disturbance travels.

What if space had a turnstile?

Now let me put a thin artificial layer between the charges. Imagine it lets waves pass from the left to the right, while stopping waves that come back from the right. The first charge moves and the second charge responds. When the second charge creates its own wave, the returning portion meets the layer and fades or gets redirected.

To actually make such a layer, I'd be looking at the physics of nonreciprocal isolators. Normal polarization filters select which way the electric field points. They don't normally decide which direction a wave is allowed to travel. Some specially designed components can favor one direction through magnetic bias or carefully timed modulation. The rejected energy has to go somewhere, usually into reflection, another path, or absorption.

I like this layer because it lets me imagine separate lanes for information moving in different directions. Of course, I'd still need some way to address individual memory cells and prevent several messages from overlapping. I'll come back to that when I reach the computer.

Could a vibration say “I love you Trishna”?

Now imagine I can control the vibration precisely. I could choose one pattern to mean 0 and another to mean 1. For a simple example, I'll use binary phase-shift keying (BPSK). Both bits use the same repeating wave. For a 1, I flip its phase halfway around the cycle. A receiver with the same clock can measure the difference and reconstruct the bits.

My first instinct was to say an upward-moving electron means 1 and a downward-moving one means 0. That can work as a sampling rule when both sides agree exactly when to look. Since any vibrating charge keeps changing direction, the clock matters a lot. Phase gives me a convenient way of putting that timing rule into the wave.

I'll try something personal: I love you Trishna (Trishna is my dear wife). Each character gets an eight-bit binary code, including the spaces. That gives me 18 bytes, or 144 bits. The blue trace below shows the wave pattern for the message, and the field arrows show the signal spreading from the sender toward the receiver. I'm sending the whole bitstream repeatedly so you can watch it.

Now I can picture a message riding on a vibration. To decode it reliably I'd need a good oscillator, shared timing, enough signal at the receiver, and ways to catch errors when noise gets in. All of that takes real circuitry and energy. Still, it's a starting point for making my thought concrete.

Distance becomes delay

If the signal leaves A and travels along a direct path in free space, the ideal propagation delay is distance divided by the speed of light. A centimeter takes about 33 picoseconds; ten centimeters about 0.33 nanoseconds; a meter about 3.34 nanoseconds. These are propagation times only, not RAM access times or end-to-end transfer latencies.

Here I want to watch what happens at the receiving end. I give the left charge one brief push and wait for the disturbance to reach the right charge. Once it starts wobbling, I draw that movement as a sine wave and mark the crests as 1 and the troughs as 0. An ordinary sine wave would produce an alternating sequence; sending a chosen message still needs the controlled encoding I explored in Part III.

My first reaction was that sending the signal straight through space should make it much faster than running it along a wire or optical fiber. The interesting part is the path and the hardware around it. Electrical signals in a circuit already propagate at a large fraction of the speed of light, and a fiber can take a nearly straight route too. A free-space connection might let me route signals differently, although its antennas and decoding circuits would add their own costs.

I'd also have to deal with the time the RAM takes to find and read a cell, the number of simultaneous requests, and how much information fits in a second. Those delays remain even if the wave itself travels through free space. So I'd want to measure the whole read operation, from memory address to the final byte arriving at the CPU.

RAM to CPU, without the usual conversation

Here's the computer I'm imagining. RAM holds the bits in ordinary cells to begin with. The CPU asks for an address. A small controller reads the selected cells and passes the bits to a set of tiny transmitters. They generate a controlled electromagnetic pattern that spreads toward a receiver near the CPU. The receiver measures that pattern, recovers the binary digits and hands the bytes to the processor.

For a first prototype, I'd keep the intuitive rule: upward motion can stand for one bit and downward motion for the other. I'd define a clock edge so everyone checks at the same moment. Later I would probably move to phase encoding, which is already well understood in radio communication. What appeals to me is the idea that the movement of a charge creates the message that another charge can sense.

The harder question is whether we can do more than exchange memory bytes. Can a carefully engineered electromagnetic mode store a state, and can a query excite a response that performs some computation as the wave propagates? Resonators can retain fields for finite times, and wave interference can implement useful analog operations. Persistent, dense, randomly addressable memory with reliable reads and writes is a very different challenge. Nothing in this essay demonstrates that conventional RAM can be replaced.

What survives the thought experiment?

There are already wonderful pieces of this picture scattered across engineering: radio links, photonics, wave-based processors, resonant cavities and directional microwave components. I'd like to bring some of those pieces together and ask a very specific question: could a tiny, addressable electromagnetic link between RAM and CPU outperform an electrical bus on latency, bandwidth, energy per bit, or reliability?

To test it, I would first build a single-channel tabletop demonstrator using a known bitstream, synchronized source and receiver, and a shielded or otherwise well-controlled channel. Next I would add a directional component and measure what happens to the return signal. Then I would compare against a wired reference using bit-error rate, bandwidth, end-to-end delay, and energy per delivered bit. Only then would multi-channel addressing, parallel reads, and memory semantics enter the discussion.

For now I'm calling this notebook entry Information Theory 2, because that's what popped into my mind when I thought of it. Shannon's theory already gives us powerful mathematics for communication with electromagnetic waves. Maybe what I'm really chasing is an architecture that uses fields more directly. I'd like to see how far I can push the idea before I decide what to call it.

Maybe the next computer will not move information between separate neighborhoods. Maybe it will learn to make the space between them part of the calculation.

Illustrations: five independent JavaScript Canvas renderings of an illustrative delayed dipole-field model. Each charge is given one finite oscillation when its incoming pulse arrives. The one-way barrier is an idealized directional mask. Arrows show a transverse field disturbance and its propagation delay across a 2D slice. They are educational schematics rather than Maxwell-equation simulations. I avoid presenting the gauge-dependent vector potential as a directly measurable ripple.

Further reading: Shannon (1948), A Mathematical Theory of Communication; Maxwell (1865), A Dynamical Theory of the Electromagnetic Field; standard treatments of antenna radiation, binary phase-shift keying, nonreciprocity, and computer memory hierarchies.