Generalize spectroscopy

[eggerdj]

Summary

This small extension allows the spectroscopy experiment to take a preparation circuit before running the spectroscopy pulse.

Details and comments

The preparation circuit allows us to e.g. prepare the first excited state and do spectroscopy on the 1<->2 transition. If the user gives a preparation circuit then the circuits to run will use this circuit followed the spec circuit with measurement.

For example

prep = QuantumCircuit(1)
prep.x(0)

spec12 = QubitSpectroscopy(qubit, np.linspace(...), unit="Hz", pre_circuit=prep)

will create circuits like

           ┌───┐┌───────────────────┐ ░ ┌─┐
      q_0: ┤ X ├┤ Spec(4759400508.) ├─░─┤M├
           └───┘└───────────────────┘ ░ └╥┘
measure: 1/══════════════════════════════╩═
                                         0 

[nkanazawa1989]

I agree this is efficient in terms of implementation but I don't think this is right direction.

I think 0-1 spectroscopy and 1-2 spectroscopy should be defined as dedicated class, in terms of the concept of qiskit experiment, i.e. every experiment is dedicated to specific type of experiment. This is why we have experiment_type in metadata. This implementation deprecates the experiment_type field.

If we want to keep this logic, we should make this a private super class and define 0-1 spect and 1-2 spect as its subclass.

Another issue is, for example, qutrit cal can be done with qubit discriminator with 0-1 inversion pulse attached to measurement. See Fig 3.1 of this article. I prefer this method because default measurement pulse is not tuned to qutrit and separation of 1 and 2 blobs are not quite good. Sometime we see unwanted mixing of these states.
https://qudev.phys.ethz.ch/static/content/science/Documents/semester/Andreas_Landig_semesterthesis_131020.pdf

[oliverdial]

Not part of this PR, but I notice some hand-wired unit conversion code in the spectroscopy classes. If we continue to allow non-SI units this seems likely to be part of almost every experiment class -- should we pick a unit conversion library? (Otherwise it seems likely we'll end up with each class accepting slightly different sets)

[eggerdj]

How bout like so de9daf0 ?

[eggerdj]

@nkanazawa1989 How about 3d8b6da ? My main concern with this is that we may end up having many different types of spectroscopy experiments lying around.

[nkanazawa1989]

Thanks Daniel, this looks almost good to me. I can approve this PR with couple of minor changes.