The chemistry, laid over the machine. The transcriber's five steps — e ( p ( g·g ) p ) e — are an orbital cycle: an electron at rest, a photon written in across a gap, the excited state held as the stored bit, a photon read back out, the electron at rest again. And the design your simulation chose by trial — silicon to write, copper to read — is exactly what the band model predicts: a gap gates (rejects noise on write), no gap dumps (lossless on read). The engineering decision and the physics agree; here is why.
| e · rest | electron in the substrate — filled HOMO (Si valence) |
| p · write | absorb HOMO→LUMO; Si gap gates — noise below it is rejected |
| g·g · store | excited state held = the bit; the gap energy is the stored colour |
| p · read | emit LUMO→HOMO; Cu no gap — electron dumps lossless |
| e · rest | ground again — signal delivered to the substrate |