The mirror reflects all the energy of the photon whereas the photodiode absorbs the energy and its state is altered by it. How nature knows the difference is a deep question. Presumably the rules governing nature ensure this consistency. What that looks like is anyone's guess.
>I would have said that quantum erasure experiments make by far the most sense if you consider measurement to never happen
I agree, but then this interpretation comes with the exponentially growing cost of actual branching. In my view, this rationally must be the interpretation of last resort, for reasons stated in the other threads.
And as mentioned elsewhere, all QM theories do have to deal with exponentially increasing branching. A cost that apparently increases if we grind up photodiodes and turn them into mirrors. If wavefunction branching is a problem, how does QM get away with it when nobody is looking?
Deep questions usually have simple answers! The simplicity is what makes the answer difficult to accept.
Presumably some state is changed somewhere in the system to indicate a detection.
>If wavefunction branching is a problem, how does QM get away with it when nobody is looking?
Do we know that branching is unbounded when we're not looking? Is there a detectable difference in a large, presumably entangled system and its classical counterpart?
Yes, but "state change" happens all the time! When the photon bounces off the mirror, even if 100% of its energy is reflected, it imparts momentum to the mirror. When the photon propagates through free space, it "changes state". All state changes occur exactly in accordance with the time-evolution of the Schrodinger equation.
> Do we know that branching is unbounded when we're not looking?
This is a basic principle of QM, which would have pretty measurable macroscopic impact if it were invalid. "Branches" are not even finite in number, even for a single particle in free space; its position and momentum distributions are both continuous, with a minimum possible area guaranteed by the Heisenberg uncertainty principle, and even so much as precisely localizing its momentum already creates an infinity of equally-likely possible positions.
It depends on what you consider the system. For a "measuring apparatus" to signal the outcome of a measurement, its internal state must change in some structured way. That is, the configuration of its atoms must change in a discernible way as to indicate one outcome over another. Perhaps the change in momentum is too diffuse or too noisy to result in an informative state change with implications for collapse. Perhaps one's context is relevant to whether a state change results in collapse (e.g. Wigner's friend). I'm more inclined to deny a universal definite reality than to accept many-worlds.
> "Branches" are not even finite in number, even for a single particle in free space
This only has problematic metaphysical implications if we assume the definite reality involves particles rather than waves or some indeterminate in-between. Branching due to "measurement" appears to be different owing to its problematic metaphysical implications. The question I'm asking is do we have experimental evidence of "very large" macroscopic entanglement?
It certainly shouldn't! If wavefunction collapse happens for a photodiode but not for a mirror, that means that there should be some objective law of nature that says "if you arrange silicon atoms in a flat, polished plane, the wavefunction does not collapse, but if you dope them with phosphorus and nitrogen, and apply an electric field, it does collapse". Whether the wavefunction collapses or not shouldn't depend on what we consider the system to be.
This is passing the buck. What makes a change discernible? You know there's only one place the bucks stops: The photon frees an electron in the photodiode, which creates a small flow of current; that current is wired to an amplifier, which flips a MOSFET gate, triggering the flow of charge into a capacitor, which results in a voltage that exceeds a comparator threshold, which triggers a software interrupt in a microprocessor, which increments a digital counter, which alters the light output in an LED screen, which a human observes and scribbles in their lab book.
At which point did the system change become "discernible", and change from a quantum wavefunction to a classical-but-randomly-behaved outcome? MWI provides an answer: when the observer became entangled with the system it was observing.
Nature doesn't even 'know' there's a photodiode there, or a human observing the screen. It's just a sea of fermions and bosons all doing their thing, with various positions and velocities. If some of them are arranged in a way that we call a "photodiode", that's only our orderly ontology imposed on the messiness of the natural world. The photodiode itself is merely a bubbling sea of quantum particles, just like its surroundings.
> The question I'm asking is do we have experimental evidence of "very large" macroscopic entanglement?
How large is "very large"? Is the Cosmic Microwave Background big enough? [1][2]
According to MWI, the classical-physics behavior we observe in macroscopic objects, and the unpredictable decay of nuclear isotopes, are exactly what we would expect from macroscopic entanglement. So all you have to do to quantum-entangle a cat is to literally put it in a box with a geiger counter and a uranium puck. But in order to prove something is quantum entangled by the standards of Copenhagen, you need to be able to create an interference pattern that an instrument can record, which necessarily requires that the instrument cannot itself be entangled with the system it's measuring. Otherwise it won't measure an interference pattern, it will measure (randomly but decidedly) one result or the other.
[1] https://arxiv.org/pdf/2106.15100v1
[2] https://www.worldscientific.com/doi/abs/10.1142/S02182718110...
It's not about consideration in terms of choice of what constitutes a system; the systems are defined by sensitivity to informative states. It's like a quantum analog to the light cone, the information boundary defines the system. But this information boundary is much more elaborate and dependent on the dynamics of the structures involved. In this case, the mirror imparts a different information dynamic than the diode and so has different implications for the information boundaries.
What the rules are that govern the interactions that integrate the informative states of two systems isn't obvious. Funnily enough, I've thought a lot about this problem in an entirely different context. Somewhat surprising that it comes up naturally in quantum mechanics. I don't have any fully worked out details, but I do have some intuition. The issue is related to supervenience of macro structures on micro details. The macro structures can carry information (i.e. have mutual information) about external systems; this is how two systems are informationally integrated. A measurement is when the macrostate of one system becomes correlated with the microstate of another system. The correlation is between macro and micro because for state to be useful requires it to interact in a specific way to result in a macrostructure correlation which can in principle be computed with. It's the information analog of free energy. Not all interactions produce state that one can compute with. Information integration requires precisely this kind of state. I suspect the exact nature of this kind of interaction can be characterized in principle.
>What you apparently are inclined to deny is the notion that you could ever be put in a quantum superposition yourself, and that this would merely feel like observing one particular nondeterministic outcome out of a deterministic distribution of possible outcomes. Superpositions for thee, but not for me!
I don't deny that I could, I just don't seem to be. And the interpretation that coheres with unobserved superposition I reject for external reasons. I don't have any qualms about the implications for personal identity from MWI (assuming that's what you're getting at).
Let's say that I send a photon through a beam-splitter which will deflect its momentum either right or left. QM predicts that the photon at the output of the beamsplitter will be in a superposition of both trajectories, and we can prove this by adding mirrors that combine both possible output beams back together with another and generate an interference pattern. The interference pattern will be detected by a sensitive photodiode array, capable of detecting individual photons.
Would you say the interaction with the photodiode array counts as a measurement? And the interaction with the mirror does not? How does QM know whether a photon interaction with a bunch of atoms in a mirror coating, vs in a silicon diode, creates a constraint on the evolution of the system? Why is one interaction measurement-like in nature, but not the other? Is there some property of silicon diodes that make them immune to being put into quantum superpositions?
I would have said that quantum erasure experiments make by far the most sense if you consider measurement to never happen, and the result is merely a superposition of both outcomes. The photon is put into a superposition of both states when it reaches the beamsplitter, the detector screen is put into a superposition of both states when the superposition-photon reaches it, and so are you in turn when you observe the detector screen.