3 ms·
I've tried to read the paper (note IANP, I'm not a physicist) and some additional descriptions, remembering what I've read in Feynman lectures. The principles r
by acqq 13y ago
I've tried to read the paper (note IANP, I'm not a physicist) and some additional descriptions, remembering what I've read in Feynman lectures. The principles really sound like reading about the DRAM (with the pulses to write, to read and to "refresh" the content of memory) but with photons in the input and output instead of the electrons. Inside of the crystal, it's the electrons that get to move to higher levels when the photons hit the atom they orbit and then to the lower level emitting the photons out. That's how a plain glass window works. The difference is that the researchers add carefully controlled electromagnetic (light is also electromagnetic emission) signals (additional controlling inputs and outputs) which gives them the possibility to "write" at one point in time and "read" sometime later the same "image." The new achievement is that they reach 40 seconds or more between the write and read, which is a new record. There's even some application of genetic algorithms described in the paper used for preparing the conditions to make such write and read possible:
The loop starts with a random set (‘‘generation’’)
of preparation pulses (‘‘genetic individuals’’).
Each individual is described by a temporal array of intensity
and frequency values (‘‘genes’’). The self-learning
loop applies the pulses for EIT and determines the individuals
with the highest fitness, i.e., the best light storage
efficiency. The next generation is built by imitating concepts
of evolutionary biology: The best individuals are
copied into the next generation (cloning). Other good
individuals are modified by variations of their genes
(mutation) or combination with other fit individuals (inheritance).
The loop goes through several hundred generations, until the gene sequences (i.e., pulse shapes)
converge toward an optimum. Figure 3 shows the progress
of the self-learning loop, i.e., the increase of signal pulse
energy after light storage vs number of completed generations.
As expected, the fitness increases monotonically
with the generations.
I'd appreciate comments from anybody who knows and understands more.