Quantum Inspired Microwave Phase Super-Resolution at Room Temperature
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.
Kai: We're looking at the paper: "Quantum Inspired Microwave Phase Super-Resolution at Room Temperature".
Mira: Here's the gist of the paper. Quantum metrology has been shown to surpass classical limits of correlation, resolution, and sensitivity <ref:2401.05026#pg2> Even quantum-inspired detection of classical signals may be advantageous in specific use cases <ref:2401.05026#pg2>.
Kai: The Core Concept The paper realizes room-temperature microwave frequency super-resolved phase measurements with trillions of photons, saturating the Cramer-Rao bound of sensitivity <ref:2401.05026#pg2>.
Mira: This is achieved by experimentally estimating the interferometric phase using the expectation value of the Parity operator by two methods <ref:2401.05026#pg2>.
Kai: The authors demonstrate a parity measurement of an interferometric phase between electromagnetic waves in the RF domain, where the thermal background is at least thousands of photons, and single-photon detection at room temperature is impossible <ref:2401.05026#pg2>.
Title and authors: Kai: So, we're looking at this paper, "Quantum Inspired Microwave Phase Super-Resolution at Room Temperature." It’s about using quantum ideas to get better phase measurements than what classical physics allows, especially when you're dealing with microwave frequencies and room temperature.
Mira: Exactly. The authors are showing that even without those super cold temperatures, you can achieve super-resolution and sensitivity by looking at the parity operator in a specific way. It sounds like they're tackling the idea of using quantum states to measure something classically limited, which is pretty interesting for practical applications.
Lev: From an engineering standpoint, room temperature microwave measurements are tough because of all that thermal background noise, but if this works without deep cooling, that’s a big deal for hardware development.
Kai: Right. So what they actually built was an experimental setup that functions like an optical Mach-Zehnder interferometer and uses a signal frequency of four point nine six gigahertz over a twenty five nanosecond duration. They used two methods to estimate the interferometric phase, which is the core idea here, and they achieved something pretty concrete with fifty six decibels of signal-to-noise ratio.
Mira: That's what they did: they used the expectation value of the parity operator to get that phase measurement. But before we get into that result, it’s important to understand what the underlying theory is saying about how resolution and sensitivity scale with each other for these measurements.
Lev: They define resolution as the full width at half maximum of some function M of phi, and sensitivity depends on both the precision of that measurement and its slope with respect to the phase. It’s a standard way to look at it in metrology, but they are using this framework to see how their quantum approach stacks up against classical bounds.
Kai: And they are showing that for coherent states, the expectation value of this parity operator gives them super-resolution for phase estimation and actually saturates the shot noise limit when there's no thermal noise present, which is a key point because it’s beating what we classically expect.
Mira: That saturation is what makes it powerful, but we have to be careful about the assumptions they are making. The theoretical framework relies on defining that parity operator Π as something related to the photon number operator N hat, and its expectation value depends on the probability of measuring a certain number of photons, which is tied back to the Wigner representation.
Lev: If you’re looking at a coherent state, they give you a specific formula for that expectation value involving N hat th and Nc multiplied by the sine squared of half phi divided by N hat plus one over two. That formula tells us how the measurement outcome changes based on the actual phase, which is what we need to know for any real-world implementation.
Title and authors: Kai: And this leads directly to their conclusion about the Cramer-Rao bound, which sets the maximum possible precision for measuring that phase under high signal-to-noise conditions. They show that this bound scales with a term involving the photon number operator N hat plus one over two times Nc, and it drops down to just shot noise when n th is zero.
Mira: That shot noise limit is what we expect from standard classical measurements, so showing they saturate it when thermal noise is absent confirms the quantum advantage in that specific scenario. But they also mention that for a coherent state, the CR bound itself has a classical limit related to half of the wavelength.
Lev: So they are comparing this new quantum method to two benchmarks: how fast your measurement precision improves with signal strength, and what the inherent noise floor is when you’re operating at room temperature.
Kai: The experimental methodology is where it gets really tangible for me—they set up something that mimics an optical Mach-Zehnder interferometer, but using microwave components. They sample the output directly with an oscilloscope and then do a Fourier analysis to get that phase space distribution.
Mira: They used two different ways to estimate the key quantity, W of zero over zero, first by fitting the phase space distribution with a maximum likelihood fit, and second by using a threshold method based on the signal distribution. It shows they have two distinct approaches to extracting this information from the data.
Lev: And their performance analysis is where things get specific about how good it actually is in practice. They show that the width of this parity measurement, which corresponds to their resolution, scales with one over the square root of the signal-to-noise ratio, and they report a highest resolution at fifty six decibels that is one thousand two hundred times smaller than the wavelength.
Kai: That factor of one thousand two hundred is what really jumps out to me when we think about practical applications. So, they found that both their resolution and their sensitivity scale with this same one over the square root of the signal-to-noise ratio, which is a consistent scaling relationship.
Mira: And they also discuss the trade-off between resolution and sensitivity being governed by a single parameter called a over sigma. This parameter tells you how you balance those two properties when choosing your measurement settings.
Lev: When we look at the appendix, they provide some more detail on how these things relate to general multi-mode Gaussian states, showing that the expectation value of the parity operator simplifies to something like pi times W of zero over zero, which is proportional to sigma squared times an exponential term involving mu squared divided by two sigma squared.
Title and authors: Kai: And from that they derive a minimum sensitivity for this method, and it’s given by vuut divided by SNR squared plus one. That formula tells us the lowest precision we can hope to get with this specific measurement technique under certain conditions.
Mira: That minimum sensitivity is what they compare to the classical shot noise limit when n th is zero, which we discussed earlier as a benchmark. They find that for both methods, resolution and sensitivity are of the same order of magnitude, which contrasts with classical methods where resolution doesn't change with signal strength at all.
Lev: The minimum achievable sensitivity using this technique only saturates the Cramer-Rao bound up to a factor of one point two six, which they calculated numerically. It suggests that while it gets very close to the best possible limit, there’s still a gap.
Kai: So, when we look at controlling noise, they note that the full width at half maximum of this parity measurement follows one over the square root of SNR scaling across several integration times with constant power.
Mira: And they flag a practical limitation here: increasing your integration time is limited by the setup’s extinction ratio, which in their case was ninety decibels. So even if you integrate longer, you can't always get better results because of that noise floor in the system itself.
Lev: When considering the energy within a certain integration time, they say that the photon number n th increases linearly with integration time but n th doesn't actually change over those times. The overall effect on the SNR is that it changes in exactly the same way in both pictures, so it doesn't fundamentally alter the scaling behavior of resolution and sensitivity.
Kai: And then they offer a direct measurement of this parity operator if you can measure photon number, which gives an error term that depends on n th plus one and e to four Nc times sine squared of half phi divided by two n th plus one minus one Nc squared times sine of phi.
Mira: That direct measurement sensitivity is also bounded by the same minimum value we talked about earlier, vuut divided by SNR squared plus one. It seems like both estimation methods point to that same fundamental limit for how precise we can be in this system.
Lev: The maximum likelihood estimation is another way they described it, where you measure the field quadrature distribution in phase space and fit it to a Gaussian distribution to estimate that W of zero over zero value, and the error in estimating that value at the origin scales with r divided by one plus mu squared over sigma squared.
Kai: And the sensitivity for this maximum likelihood estimation is given by s times cotangent squared of half phi plus n th plus one over two Nc divided by four sine of phi. This shows how you calculate your uncertainty when you use that fitting approach instead of just a direct measurement.
Title and authors: Mira: And the minimum sensitivity derived from that fitting method is also vuut divided by SNR squared plus one, which means both ways of estimating this quantity yield the same fundamental limit on precision for this system.
Lev: The threshold analysis involves looking at the probability of being inside some phase space area defined by a threshold radius a from the origin, and they show that the choice of that threshold affects both resolution and sensitivity.
Kai: They give you an approximation for this probability, W of zero over zero, which depends on integrating something over all angles and radii involving a function r W of r theta. The choice of your threshold radius a directly impacts how good your resolution and sensitivity turn out to be.
Mira: And they show that the minimum possible resolution with this method matches what you get from the maximum likelihood estimation, meaning they are equally good at setting the best achievable resolution for this specific measurement setup.
Lev: The minimum possible sensitivity using this method doesn't quite saturate the Cramer-Rao bound, but it asymptotically reaches it up to a factor of one point two six when calculated numerically.
Kai: So, to wrap up on all of this with the paper "Quantum Inspired Microwave Phase Super-Resolution at Room Temperature," we have two robust methods—the maximum likelihood fit and the threshold method for estimating the parity operator. Both methods give you a measure of how certain your phase estimate is, which is your sensitivity.
Mira: And both these approaches show that resolution and sensitivity scale with one over the square root of SNR, which means increasing your signal strength improves both properties simultaneously in these cases.
Lev: This work opens up possibilities for using this technique to refine range and direction measurements in free space, which is a useful extension since it moves it beyond just this specific lab setup.
Kai: And while they point to future work, they mention that reducing the inherent thermal noise of the microwave system by cooling parts of the setup or using squeezed states could lead to even better results.
Mira: That’s true, and squeezing states are another area where quantum metrology can give you super-sensitivity measurements, which is another direction we can explore within this Gaussian state framework.
Lev: And that’s where our hardware research comes in—we need to figure out how to actually implement those cooling strategies on real hardware and see what the overhead looks like for achieving those super-sensitive measurements.
Kai: So that's it for this paper, "Quantum Inspired Microwave Phase Super-Resolution at Room Temperature." It’s a lot of work showing that you can do high-fidelity measurements in challenging room temperature environments using quantum inspiration.
The paper's summary: Kai: So, to wrap up on that paper, it boils down to this: they realized you can get super-resolved phase measurements in the microwave range right at room temperature and they managed to hit the theoretical limit for sensitivity.
Mira: It really does sound like a lot of work because they are doing this without needing those incredibly cold setups that quantum hardware usually requires. They're using two ways to estimate that core quantity, the parity operator, and both methods lead them to the same fundamental limit on precision.
Lev: The main thing I see from an engineering standpoint is that they've shown how you can get a measurement resolution scaled by one over the square root of the signal-to-noise ratio across different integration times. That’s a consistent scaling relationship, which is good for planning hardware noise budgets.
Kai: Right, and they show this scaling holds even when you look at the maximum likelihood estimation and the threshold method together. It means you have a solid way to choose your settings based on that parameter 'a over sigma'.
Mira: That parameter tells you how to balance getting high resolution versus getting high sensitivity in your measurement setup. And they found that for similar performance between those two methods, that sweet spot is around one point seven for 'a over sigma'.
Lev: So if you're designing the actual hardware, knowing that the best balance is around one point seven helps you define where to aim your calibration efforts. It moves the discussion from just theory to what actually runs on a processor.
Kai: And they also pointed out that this technique isn't just for lab experiments; it can be used to refine measurements in free space for things like range and direction finding. That's a significant extension beyond just sitting in a controlled microwave environment.
Mira: Exactly, and looking ahead, the authors mention that the next steps involve pushing this further by cooling parts of the system down or using squeezed states to get even better super-sensitivity measurements. It's a clear path for improving performance.
Lev: Cooling is always the biggest hurdle for me; getting that level of noise reduction on real microwave hardware is tough, but if you can do that, those squeezed states would really put this technique into a new category for precision sensing.
The paper's improvements: Tom: So, to recap where we are: they showed you can get super-resolution in microwave measurements at room temperature by using parity operator estimation, and both their methods lead to the same fundamental sensitivity limit. Now we're going to look at what they suggest next for improvement.
Kai: The paper points out that they need to tackle the inherent thermal noise of the microwave system itself because even though it works at room temperature, you still have limitations from those background photons.
Mira: They suggest two big ways forward: first, significantly reducing that thermal noise by physically cooling parts of the setup down, and second, using squeezed states to get a super-sensitivity measurement on top of what they already have.
Lev: From my point of view, cooling is the most practical engineering hurdle right now. If you can bring down the temperature enough in this microwave system without losing your quantum coherence entirely, that would be a huge step toward real-world application.
Kai: Right, and using squeezed states is another area where quantum metrology gets an edge, so that's a clear direction for future work within the Gaussian state framework they are using.
Mira: It’s important to remember that both of these improvements are still framed within the same theoretical framework they developed for Gaussian states, which keeps things consistent.
Lev: I just wonder about the overhead; implementing those cooling strategies on actual hardware means you're dealing with complex integration and control systems, so we'll need to see how much extra complexity that adds to your error correction budget.
Conclusion: Kai: So we've covered how they built this experiment to get super-resolved phase measurements in the microwave range at room temperature, using those parity operator methods.
Mira: And we saw that both the maximum likelihood fit and the threshold method point to a shared fundamental limit on precision for this system. It’s a consistent result across different ways of extracting information from the data.
Lev: I just want to reiterate how much work that is for real hardware; they're showing you how these theoretical bounds translate into actual noise budgets you have to manage when you try to run this on a processor.
Kai: It really opens up possibilities for using this technique in free space, which means we can use it to refine measurements of range and direction without needing that super controlled environment.
Mira: And looking ahead, the next steps are clearly focused on improving that performance by either cooling down the system or using squeezed states to push beyond what they achieved here.
Lev: If you can tackle the thermal noise issue and get those squeezing measurements working reliably, that would be a major step for any real quantum sensor development.
Kai: So that’s where we head next, looking at how other condensed matter systems are handling these kinds of many-body problems.
Leonid Vidro, Liran Shirizly, Naftali Kirsh, Nadav Katz, Hagai S. Eisenberg
Racah Institute of Physics, Hebrew University of Jerusalem
quant-ph
Submitted: 2024-01-10
Updated: 2024-01-10
Journal ref: Physical Review Applied 22 (3), 034008 (2024)
DOI: 10.1103/PhysRevApplied.22.034008
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 75/100
The gist: Quantum metrology has been shown to surpass classical limits of correlation, resolution, and sensitivity <ref:2401.05026#pg2> Even quantum-inspired detection of classical signals may be advantageous…
Key concepts
- Parity Operator Expectation Value
- This operator is used to experimentally estimate the unknown interferometric phase of microwave signals. Its expectation value provides information about the phase, which is crucial for measuring how much a signal has shifted in its interference pattern.
- Cramer-Rao Bound (CR Bound)
- The CR bound sets the theoretical minimum limit on how precisely a measurement can estimate an unknown parameter, like the phase. The paper shows that their quantum-inspired method can reach this fundamental limit even with high signal quality.
- Resolution and Sensitivity Scaling
- Resolution describes how finely you can distinguish between different phase settings, while sensitivity describes the precision of estimating the actual phase. Both are shown to scale inversely with the square root of the signal-to-noise ratio (1/√SNR).
- Wigner Representation
- This is a mathematical tool used to describe quantum states, like those of microwave photons. In this context, it helps relate the measurement outcomes to classical phase space distributions, which is essential for using methods like maximum likelihood estimation.
Terminology
Summary
Quantum metrology has been shown to surpass classical limits of correlation, resolution, and sensitivity <ref:2401. p. 0 Quantum Inspired Microwave Phase Super-Resolution at Room Temperature
Leonid Vidro,∗ Liran Shirizly, Naftali Kirsh, Nadav Katz, and Hagai S. Eisenberg
Racah Institute of Physics, Hebrew University of Jerusalem, Jerusalem 91904, Israel
(Dated: January 11, 2024)
Quantum metrology has been shown to surpass classical limits of correlation, resolution, and
sensitivity. It has been introduced to interferometric Radar schemes, with intriguing preliminary
results. Even quantum-inspired detection of classical signals may be advantageous in specific use
cases. It has been introduced to interferometric Radar schemes, with intriguing preliminary
results. Even quantum-inspired detection of classical signals may be advantageous in specific use
cases.Where the paper says it
Where the paper says it
Where the paper says it
The Core Concept
The paper realizes room-temperature microwave frequency super-resolved phase measurements with trillions of photons, saturating the Cramer-Rao bound of sensitivity <ref:2401.05026#pg2>. This is achieved by experimentally estimating the interferometric phase using the expectation value of the Parity operator by two methods <ref:2401. p. 0 Following ideas demonstrated so far only in the optical domain, where practically no thermal
background photons exist, we realize room-temperature microwave frequency super-resolved phase
measurements with trillions of photons, while saturating the Cramer-Rao bound of sensitivity. We
experimentally estimate the interferometric phase using the expectation value of the Parity operator
by two methods. Following ideas demonstrated so far only in the optical domain, where practically no thermal
background photons exist, we realize room-temperature microwave frequency super-resolved phase
measurements with trillions of photons, while saturating the Cramer-Rao bound of sensitivity. We
experimentally estimate the interferometric phase using the expectation value of the Parity operator
by two methods. This
value was estimated by either a homodyne detection [10]
or by a photon-number resolving measurement [9]. In
this work, we demonstrate a parity measurement of an
interferometric phase between electromagnetic waves in
the RF domain, where the thermal background is at least
thousands of photons, and the detection of single photons
at room temperature is impossible. This
value was estimated by either a homodyne detection [10]
or by a photon-number resolving measurement [9]. In
this work, we demonstrate a parity measurement of an
interferometric phase between electromagnetic waves in
the RF domain, where the thermal background is at least
thousands of photons, and the detection of single photons
at room temperature is impossible.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Theoretical Framework
The resolution of a measurement is defined as Res(ϕ) = FWHM(M(ϕ)) <ref:2401.05026#pg7>. The sensitivity of the phase measurement is derived from the precision of the measurement ∆M and the slope ∆ϕ = ∆M ∂M/∂ϕ <ref:2401. p. 0 The width of the parameter’s function with respect to the phase is the resolution of the measurement
Res(ϕ) = FWHM(M(ϕ)). (1)
The sensitivity of the phase measurement is derived
from the precision of the measurement ∆M and the slope
arXiv:2401. The Parity operator is
defined as
Π = ( ˆ −1)Nˆ
, (3)
where Nˆ is the photon number operator. Its expectation value depends on the probability Pn of measuring n
photons, as well as it is proportional to the value at the
origin point of the Wigner representation of the quantum
state of the electromagnetic field [17] as function of the
field quadratures
⟨Πˆ⟩ =
X∞
n=0
(−1)n
· Pn = π · W(0, 0). For an
interferometric phase measurement of a coherent state
with µ
2
0/2 total number of photons and an interferometric
phase ϕ, at the output of one interferometer arm µ =
µ0 · sin(ϕ/2), and the number of photons is:
µ
2
/2 = Nc · sin2
(ϕ/2) = µ
2
0/2 · sin2
(ϕ/2) (A8)
and the phase
θ = ϕ/2. (A9)
The expectation value of the Parity operator thus follows:
⟨Πˆ ⟩ = π · W(0, 0) = 1
2Nth + 1
e
−
Nc·sin2(ϕ/2)
Nth+1/2
. (6)
For an interferometric phase measurement of a coherent
state with Nc photons on average and a phase ϕ, the
expectation value of the Parity operator is
⟨Πˆ ⟩ =
1
2Nth + 1
e
−
Nc·sin2(ϕ/2)
Nth+1/2
. (7)
The CR bound on measuring an interferometric phase ϕ
for high SNR is given by:
∆ϕmin =
r
2
SNR = 2
σ
µ
=
s
2
Nth + 1/2
Nc
, (8)
which is reduced to the shot noise limit when Nth = 0.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Experimental Methodology
The experimental setup is equivalent to an optical Mach-Zehnder interferometer <ref:2401.05026#pg10>. The output is sampled directly by an oscilloscope, then Fourier analyzed, effectively conducting Homodyne detection. p. 2 At one of the
arms, an IQ mixer controls the interferometric phase ϕ. The output is sampled directly by an oscilloscope, then
Fourier analyzed, effectively conducting Homodyne detection. At one of the
arms, an IQ mixer controls the interferometric phase ϕ. The output is sampled directly by an oscilloscope, then
Fourier analyzed, effectively conducting Homodyne detection. Several sequences
of the signal construct the phase space distribution from
which ⟨Πˆ ⟩ can be estimated. The signal frequency is 4. Several sequences
of the signal construct the phase space distribution from
which ⟨Πˆ ⟩ can be estimated. The signal frequency is 4. The signal frequency is adjusted to a minimum point of
the system spectral noise. The sample duration of 25ns
is such that the signal frequency would be exactly on the
grid of the discrete Fourier spectrum. The signal frequency is adjusted to a minimum point of
the system spectral noise. The sample duration of 25ns
is such that the signal frequency would be exactly on the
grid of the discrete Fourier spectrum. See Appendix B for more details. The W(0, 0) value is estimated by the two methods
described in Sec II; by a maximum likelihood fit to
the phase space distribution and by using the threshold method, where W˜ (0, 0) is evaluated by counting the
number of phase space samples within a threshold radius
a.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Performance Analysis
The maximum width of the parity measurement corresponds to the measurement resolution. The scaling of the FWHM is following 1/√SNR, and the highest resolution demonstrated at 56. p. 3 maximum (FWHM) of the parity measurement corresponds to the measurement resolution. The scaling of
the FWHM is following 1/
√
SNR and the highest resolution demonstrated at 56. maximum (FWHM) of the parity measurement corresponds to the measurement resolution. The scaling of
the FWHM is following 1/
√
SNR and the highest resolution demonstrated at 56. The scaling of
the FWHM is following 1/
√
SNR and the highest resolution demonstrated at 56. 1dB is 1200 times smaller than
the wavelength. The scaling of
the FWHM is following 1/
√
SNR and the highest resolution demonstrated at 56. 1dB is 1200 times smaller than
the wavelength. For both methods, the resolution and the sensitivity
are of the same order of magnitude, in contrast to the
classical method, where the resolution is unaffected by
the SNR. Both scale as 1/
√
SNR. For both methods, the resolution and the sensitivity
are of the same order of magnitude, in contrast to the
classical method, where the resolution is unaffected by
the SNR. Both scale as 1/
√
SNR. Both scale as 1/
√
SNR. The sensitivity of the
estimation by the maximum likelihood fit method saturates the CR bound and the threshold method reaches
the bound up to a factor of ∼ 1. Both scale as 1/
√
SNR. The sensitivity of the
estimation by the maximum likelihood fit method saturates the CR bound and the threshold method reaches
the bound up to a factor of ∼ 1. The freedom of choice of the threshold
parameter allows choosing a combination of the superresolution and the sensitivity, where the quality of one
comes at the expanse of the other. This trade-off is defined by a single parameter a/σ.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Future Extensions
The work extends to any domain and temperature between the quantum to classical divide because the Wigner representation reduces to the IQ distribution in this case of Gaussian states and considerable thermal noise. Further research includes significantly reducing the microwave system’s inherent thermal noise by cooling parts of the setup. p. 4 This work can be extended to free space
with phase estimation used for refining both range and
direction measurements. Further research includes significantly reducing the microwave system’s inherent thermal noise by cooling parts of the setup. This work can be extended to free space
with phase estimation used for refining both range and
direction measurements. Further research includes significantly reducing the microwave system’s inherent thermal noise by cooling parts of the setup. Further research includes significantly reducing the microwave system’s inherent thermal noise by cooling parts of the setup. Using squeezed
states, will result in a super-sensitivity measurement [1]. Further research includes significantly reducing the microwave system’s inherent thermal noise by cooling parts of the setup. Using squeezed
states, will result in a super-sensitivity measurement [1]. Using squeezed
states, will result in a super-sensitivity measurement [1]. All of these extensions are also in the framework of Gaussian states. Using squeezed
states, will result in a super-sensitivity measurement [1]. All of these extensions are also in the framework of Gaussian states.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Appendix Details
The derivation for a general multi-mode Gaussian state is straightforward. The expectation value of the Parity operator thus follows ⟨Πˆ⟩ = π · W(0, 0) = 1/2σ squared e−µ 2/2σ squared. p. 1 For white noise, such as the noise in
our system, the SNR scales down linearly with increased
integration time. The expectation value of the Parity
operator of such a Gaussian state is
⟨Πˆ ⟩ = π · W(0, 0) = 1
2σ
2
e
−
µ2
2σ2 . For white noise, such as the noise in
our system, the SNR scales down linearly with increased
integration time. The expectation value of the Parity
operator of such a Gaussian state is
⟨Πˆ ⟩ = π · W(0, 0) = 1
2σ
2
e
−
µ2
2σ2 . (A13)
Using this, the error of estimating the value at the origin,
normalized to a single phase space measurement is:
π · ∆W(0, 0) = ∆⟨Πˆ ⟩ =
1
2σ
2
e
−
µ2
2σ2 ·
r
1 +
µ2
σ
2
. (A14)
The sensitivity of this estimation is given by:
∆ϕ =
s
(cot2
(ϕ/2) +
Nth + 1/2
Nc
2
2
sin(ϕ)
, (A15)
the minimum value of this sensitivity is given by:
∆ϕmin =
vuut
2
SNR2 + 2s
1
SNR2
1
SNR2 + 1
−−−−−−→
SNR→∞ r
2
SNR,
which reduces to the CR bound for large values of SNR. (A13)
Using this, the error of estimating the value at the origin,
normalized to a single phase space measurement is:
π · ∆W(0, 0) = ∆⟨Πˆ ⟩ =
1
2σ
2
e
−
µ2
2σ2 ·
r
1 +
µ2
σ
2
. (A14)
The sensitivity of this estimation is given by:
∆ϕ =
s
(cot2
(ϕ/2) +
Nth + 1/2
Nc
2
2
sin(ϕ)
, (A15)
the minimum value of this sensitivity is given by:
∆ϕmin =
vuut
2
SNR2 + 2s
1
SNR2
1
SNR2 + 1
−−−−−−→
SNR→∞ r
2
SNR,
which reduces to the CR bound for large values of SNR. (A13)
Using this, the error of estimating the value at the origin,
normalized to a single phase space measurement is:
π · ∆W(0, 0) = ∆⟨Πˆ ⟩ =
1
2σ
2
e
−
µ2
2σ2 ·
r
1 +
µ2
σ
2
. (A14)
The sensitivity of this estimation is given by:
∆ϕ =
s
(cot2
(ϕ/2) +
Nth + 1/2
Nc
2
2
sin(ϕ)
, (A15)
the minimum value of this sensitivity is given by:
∆ϕmin =
vuut
2
SNR2 + 2s
1
SNR2
1
SNR2 + 1
−−−−−−→
SNR→∞ r
2
SNR,
which reduces to the CR bound for large values of SNR. (A13)
Using this, the error of estimating the value at the origin,
normalized to a single phase space measurement is:
π · ∆W(0, 0) = ∆⟨Πˆ ⟩ =
1
2σ
2
e
−
µ2
2σ2 ·
r
1 +
µ2
σ
2
. (A14)
The sensitivity of this estimation is given by:
∆ϕ =
s
(cot2
(ϕ/2) +
Nth + 1/2
Nc
2
2
sin(ϕ)
, (A15)
the minimum value of this sensitivity is given by:
∆ϕmin =
vuut
2
SNR2 + 2s
1
SNR2
1
SNR2 + 1
−−−−−−→
SNR→∞ r
2
SNR,
which reduces to the CR bound for large values of SNR. The solid red
line is at the Cramer-Rao bound. For both methods, the resolution and the sensitivity
are of the same order of magnitude, in contrast to the
classical method, where the resolution is unaffected by
the SNR. This trade-off is defined by a single parameter a/σ. The minimum achievable sensitivity with this method saturates the CR bound
only up to a factor of ∼ 1. The best possible sensitivity with this method does not saturate the CR bound
but asymptotically approaches it up to a factor of 1. 26
calculated numerically, see A 3. The best possible sensitivity with this method does not saturate the CR bound
but asymptotically approaches it up to a factor of 1. 26
calculated numerically, see A 3. 26 which was calculated numerically. The balance
point of similar performance of both resolution and sensitivity is at a/σ ≈ 1. 2 shows parametric
ROC curves for various values of locking precision ∆ϕl
parametrized with the detection threshold a. The locking precision is stated in terms of the CR bound. For example,
in Radar applications, it is common to plot a Receiver
Operating Characteristic (ROC) curve, which is the true
positive alerts versus false positive alerts probabilities. In our case, a true positive occurrence is when ϕ < ∆ϕl
and we received a sample within the threshold a.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Controlling Noise
The FWHM of ⟨Πˆ⟩ follows 1/√SNR scaling. Increasing the integration time is limited in practice by the setup’s extinction ratio, which was 90dB in our setup. p. 8 The
overall effect, however, does not change since the SNR
changes in the same way in both pictures. Increasing the integration time is limited in practice
by the setup’s extinction ratio, which was 90dB in our
setup. Higher SNR and higher resolution can also
be obtained by a longer integration time with constant
signal power. When considering the energy within a certain integration time, Nc increases linearly with integration time, however Nth does not change. The quadrature noise in this case scales as √
Hz. The
overall effect, however, does not change since the SNR
changes in the same way in both pictures. Such a high extinction ratio was achieved by controlling the losses in both arms of the interferometer and
minimizing the dark port output power. The resolution
feature of ⟨Πˆ ⟩ for several integration times with constant
power is shown in Fig.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Direct Measurement
A direct measurement of Πˆ is possible if a photon number measurement is available. The error of such a measurement is the standard deviation of the Parity operator and the calculated sensitivity of the measurement is ∆ϕ = r / sqrt(Nth+1 e 4Nc sin 2(ϕ/2) / (2Nth+1 − 1Nc squared sin(phi))). p. 6 In this case, its mean value could be calculated simply
by ensemble averaging. The error of such a measurement
is the standard deviation of the Parity operator and the
calculated sensitivity of the measurement is:
∆ϕ =
r
1
2Nth+1 e
4Nc sin2(ϕ/2)
2Nth+1 − 1
|
Nc
2Nth+1 sin(ϕ)|
. In this case, its mean value could be calculated simply
by ensemble averaging. The error of such a measurement
is the standard deviation of the Parity operator and the
calculated sensitivity of the measurement is:
∆ϕ =
r
1
2Nth+1 e
4Nc sin2(ϕ/2)
2Nth+1 − 1
|
Nc
2Nth+1 sin(ϕ)|
.Where the paper says it
Where the paper says it
Maximum Likelihood Estimation
By measuring the field quadrature distribution in phase space, the W(0, 0) value can be estimated using a maximum likelihood fit. The error in estimating the value at the origin, normalized to a single phase space measurement is ∆⟨Πˆ⟩ = 1/2σ squared e−µ 2/2σ squared · r / (1 + µ 2/σ 2). p. 6 The Fisher information for estimating the mean and variance of a 2-D
symmetric Gaussian distribution by a maximum likelihood fit is given by:
I(µ, σ2
) = 1
σ2 0
0
1
σ4
. (A13)
Using this, the error of estimating the value at the origin,
normalized to a single phase space measurement is:
π · ∆W(0, 0) = ∆⟨Πˆ ⟩ =
1
2σ
2
e
−
µ2
2σ2 ·
r
1 +
µ2
σ
2
. The Fisher information for estimating the mean and variance of a 2-D
symmetric Gaussian distribution by a maximum likelihood fit is given by:
I(µ, σ2
) = 1
σ2 0
0
1
σ4
. (A13)
Using this, the error of estimating the value at the origin,
normalized to a single phase space measurement is:
π · ∆W(0, 0) = ∆⟨Πˆ ⟩ =
1
2σ
2
e
−
µ2
2σ2 ·
r
1 +
µ2
σ
2
. (A13)
Using this, the error of estimating the value at the origin,
normalized to a single phase space measurement is:
π · ∆W(0, 0) = ∆⟨Πˆ ⟩ =
1
2σ
2
e
−
µ2
2σ2 ·
r
1 +
µ2
σ
2
. (A14)
The sensitivity of this estimation is given by:
∆ϕ =
s
(cot2
(ϕ/2) +
Nth + 1/2
Nc
2
2
sin(ϕ)
, (A15)
the minimum value of this sensitivity is given by:
∆ϕmin =
vuut
2
SNR2 + 2s
1
SNR2
1
SNR2 + 1
−−−−−−→
SNR→∞ r
2
SNR,
which reduces to the CR bound for large values of SNR. (A13)
Using this, the error of estimating the value at the origin,
normalized to a single phase space measurement is:
π · ∆W(0, 0) = ∆⟨Πˆ ⟩ =
1
2σ
2
e
−
µ2
2σ2 ·
r
1 +
µ2
σ
2
. (A14)
The sensitivity of this estimation is given by:
∆ϕ =
s
(cot2
(ϕ/2) +
Nth + 1/2
Nc
2
2
sin(ϕ)
, (A15)
the minimum value of this sensitivity is given by:
∆ϕmin =
vuut
2
SNR2 + 2s
1
SNR2
1
SNR2 + 1
−−−−−−→
SNR→∞ r
2
SNR,
which reduces to the CR bound for large values of SNR.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Threshold Analysis
The probability of the measurement being inside some phase space area defined by a threshold radius a from the origin is given by W˜ (0, 0) ≈ 1/πa squared ∫ Z 2π 0 dθ Za 0 dr rW(r, θ). The choice of a affects both the resolution and the sensitivity. p. 1 For a symmetric Gaussian distribution,
the marginal radial distribution is Rician. The choice of a affects both the resolution and the sensitivity of the measurement and one can choose a working
point that suits the specific requirements of the system. For a symmetric Gaussian distribution,
the marginal radial distribution is Rician. The choice of a affects both the resolution and the sensitivity of the measurement and one can choose a working
point that suits the specific requirements of the system. (A17)
Using this equation we can calculate the resolution for
a given phase space distribution. The error in calculating the Bernoulli probability p is just ∆p =
√p per
phase space sample point (and is reduced as the square
of the number of sample points) and so the sensitivity
can be calculated. With this normalization, the trade-off between resolution and sensitivity could be better appreciated. The
minimum possible resolution with this method matches
that of the maximum likelihood method. With this normalization, the trade-off between resolution and sensitivity could be better appreciated. The
minimum possible resolution with this method matches
that of the maximum likelihood method. The
minimum possible resolution with this method matches
that of the maximum likelihood method. However, The
minimum possible sensitivity does not saturate the CR
bound but asymptotically reaches it up to a factor of
∼ 1.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Summary
The paper demonstrates two methods for estimating the expectation value of the Parity operator, both based on the reconstruction of the Wigner representation by an effective Homodyne detection. The first method is based on a maximum likelihood fit of the phase space measurements to a Gaussian distribution. p. 4 Thus, as the
quantum and thermal fluctuations are scaled on a common footing, our work immediately extends to any domain and temperature between the quantum to classical
divide. The first method is based on a maximum likelihood fit
of the phase space measurements to a Gaussian distribution. It seems
that for this method, at least two measurements are required to estimate the maximum likelihood fit, but since
the noise in the system should remain constant, it can
be calibrated, and a single measurement would suffice to
estimate the current mean value of the distribution and
to complete the maximum likelihood fit. The threshold method is more straightforward to calculate, though
choosing the optimal threshold always requires calibration of the noise. The minimum achievable sensitivity with this method saturates the CR bound
only up to a factor of ∼ 1. Both methods can have
an enhanced resolution by increasing the SNR, e. Both methods can have
an enhanced resolution by increasing the SNR, e. , by
increasing the integration time. , by
increasing the integration time. The sensitivity of a measurement describes the certainty in which the phase is estimated. It is influenced
by the inherent noise in the physical state that is used
for sensing, and by the dependence between the measured quantity and the estimated parameter as in Eq. The resolution property is important when it is desirable to differentiate between two settings of the measured phase. Using the methods presented here, the resolution scales with the SNR the same
way as the sensitivity does, giving the resolution its intuitive interpretation in resolving and differentiating between settings. This work can be extended to free space
with phase estimation used for refining both range and
direction measurements.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
--- Page 1 ---
Quantum metrology has been shown to surpass classical limits of correlation, resolution, and sensitivity <ref:2401.05026#pg2> Even quantum-inspired detection of classical signals may be advantageous in specific use cases <ref:2401.05026#pg2>.
The paper realizes room-temperature microwave frequency super-resolved phase measurements with trillions of photons, while saturating the Cramer-Rao bound of sensitivity <ref:2401.05026#pg2>. This is achieved by experimentally estimating the interferometric phase using the expectation value of the Parity operator by two methods <ref:2401.05026#pg2>. The authors demonstrate a parity measurement of an interferometric phase between electromagnetic waves in the RF domain, where the thermal background is at least thousands of photons, and single-photon detection at room temperature is impossible <ref:2401.05026#pg2>.
The resolution of a measurement is defined as Res(ϕ) = FWHM(M(ϕ)). The sensitivity of the phase measurement is derived from the precision of the measurement ∆M and the slope ∆ϕ = ∆M ∂M/∂ϕ. p. 0 The width of the parameter’s function with respect to the phase is the resolution of the measurement
Res(ϕ) = FWHM(M(ϕ)). (1)
The sensitivity of the phase measurement is derived
from the precision of the measurement ∆M and the slope
arXiv:2401. The Parity operator is
defined as
Π = ( ˆ −1)Nˆ
, (3)
where Nˆ is the photon number operator. Its expectation value depends on the probability Pn of measuring n
photons, as well as it is proportional to the value at the
origin point of the Wigner representation of the quantum
state of the electromagnetic field [17] as function of the
field quadratures
⟨Πˆ⟩ =
X∞
n=0
(−1)n
· Pn = π · W(0, 0). For an
interferometric phase measurement of a coherent state
with µ
2
0/2 total number of photons and an interferometric
phase ϕ, at the output of one interferometer arm µ =
µ0 · sin(ϕ/2), and the number of photons is:
µ
2
/2 = Nc · sin2
(ϕ/2) = µ
2
0/2 · sin2
(ϕ/2) (A8)
and the phase
θ = ϕ/2. (A9)
The expectation value of the Parity operator thus follows:
⟨Πˆ ⟩ = π · W(0, 0) = 1
2Nth + 1
e
−
Nc·sin2(ϕ/2)
Nth+1/2
. (6)
For an interferometric phase measurement of a coherent
state with Nc photons on average and a phase ϕ, the
expectation value of the Parity operator is
⟨Πˆ ⟩ =
1
2Nth + 1
e
−
Nc·sin2(ϕ/2)
Nth+1/2
. (7)
The CR bound on measuring an interferometric phase ϕ
for high SNR is given by:
∆ϕmin =
r
2
SNR = 2
σ
µ
=
s
2
Nth + 1/2
Nc
, (8)
which is reduced to the shot noise limit when Nth = 0.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The experimental setup is equivalent to an optical Mach-Zehnder interferometer. The output is sampled directly by an oscilloscope, then Fourier analyzed, effectively conducting Homodyne detection. The signal frequency is 4.96GHz, and the sample duration of 25ns is such that the signal frequency would be exactly on the grid of the discrete Fourier spectrum.
The maximum width of the parity measurement corresponds to the measurement resolution. The scaling of the FWHM is following 1/√SNR, and the highest resolution demonstrated at 56.1dB is 1200 times smaller than the wavelength. Both methods show that the resolution and sensitivity scale as 1/√SNR.
The derivation for a general multi-mode Gaussian state is straightforward. The expectation value of the Parity operator thus follows ⟨Πˆ⟩ = π · W(0, 0) = 1/2σ squared e−µ 2/2σ squared. The minimum value of this sensitivity is given by ∆ϕmin = vuut / (SNR squared + 1).
A direct measurement of Πˆ is possible if a photon number measurement is available. The error of such a measurement is the standard deviation of the Parity operator and the calculated sensitivity of the measurement is ∆ϕ = r / sqrt(Nth+1 e 4Nc sin 2(ϕ/2) / (2Nth+1 − 1Nc squared sin(phi))).
By measuring the field quadrature distribution in phase space, the W(0, 0) value can be estimated using a maximum likelihood fit. The error in estimating the value at the origin, normalized to a single phase space measurement is ∆⟨Πˆ⟩ = 1/2σ squared e−µ 2/2σ squared · r / (1 + µ 2/σ 2). The sensitivity of this estimation is given by ∆ϕ = s(cot 2(ϕ/2) + Nth+1/2 Nc / 4 sin(phi)).
The probability of the measurement being inside some phase space area defined by a threshold radius a from the origin is given by W˜ (0, 0) ≈ 1/πa squared ∫ Z 2π 0 dθ Za 0 dr rW(r, θ). The choice of a affects both the resolution and the sensitivity. The minimum possible resolution with this method matches that of the maximum likelihood method.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved AI system can perform room-temperature microwave frequency super-resolved phase measurements with trillions of photons, saturating the Cramer-Rao bound of sensitivity while achieving a resolution up to 1200 times better than wavelength.
-
The system can utilize two methods for estimating the expectation value of the Parity operator: a maximum likelihood fit to phase space measurements and a threshold method, allowing for a trade-off between resolution and sensitivity defined by the parameter a over sigma.
-
AI systems can be calibrated to choose an optimal working point based on the trade-off between resolution and sensitivity, as shown in Figure A.1 where the balance point is at a over sigma approximately equal to 1.7 for similar performance of both methods.
-
The system can be adapted for phase locking applications by defining two binary thresholds, one for actual state deviation and one for estimation threshold, enabling calculation of true positive and false positive alerts using Receiver Operating Characteristic (ROC) curves.
Abstract
Quantum metrology has been shown to surpass classical limits of correlation, resolution, and sensitivity. It has been introduced to interferometric Radar schemes, with intriguing preliminary results. Even quantum-inspired detection of classical signals may be advantageous in specific use cases. Following ideas demonstrated so far only in the optical domain, where practically no thermal background photons exist, we realize room-temperature microwave frequency super-resolved phase measurements with trillions of photons, while saturating the Cramer-Rao bound of sensitivity. We experimentally estimate the interferometric phase using the expectation value of the Parity operator by two methods. We achieve super-resolution up to 1200 times better than the wavelength with 25ns integration time and 56dB SNR.
Related papers
- Fulqrum: a generalized framework for quantum subspace diagonalization
- Flux-Activated Resonant Control of a Bosonic Quantum Memory
- Machine learning the effects of many quantum measurements
- Barriers to dequantization of near-term molecular electronic structure methods
- BRIDGE: Lazy-Move Compilation for Dual-Species Neutral-Atom Quantum Computers
- Non-Hermiticity of an anomalous superradiant phase