Exceptional-point-like Sensing near Hermitian Critical Points

arXiv:2601.17266 · physics.optics · Submitted 2026-01-24 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "Exceptional-point-like Sensing near Hermitian Critical Points".

Mira: Here's the gist of the paper. A non-Hermitian system at an exceptional point (EP), a specific critical point (CP) associated with the parity-time symmetric phase transition, exhibits a sublinear response to perturbation and promise unprecedented sensitivity beyond the linear-response Hermitian sensors, so far operating at the diabolic points (DP) System Comparison and Response Characteristics The paper compares conventional Hermitian DP-based sensors, non-Hermitian sensors, and the proposed CP-based sensor The CP-based Hermitian sensor exhibits an EP-like response and enhanced sensitivity breaking the Petermann-factor limit of sensitivity in nonHermitian counterparts This is achieved because of the inherent eigenvector orthogonality, this CP-based Hermitian sensor exhibits an EP-like response and enhanced sensitivity The CP-based sensor measures the extremum separation of transmittance spectrum, ∆µ, see the right panel in Fig.

Kai: 1e.

Mira: Below a bias level that η(ϵb) ≤ ηCP, the transmission is a single peak.

Kai: At ηCP, the transmission shows a plateau profile.

Title and authors: Kai: Let's talk about who wrote this. The paper, "Exceptional-point-like Sensing near Hermitian Critical Points," is by Jiang-Shan Tang, Long-Qi Xiao, Hao-Dong Wu, Yuwei Jing, Han Zhang, Ya-Ping Ruan, Wuming Liu and Yan-Qing Lu Kai.

Mira: I've looked at the authors' work before. This group has strong expertise in condensed matter theory and how these critical phenomena manifest in different physical systems Mira.

Lev: From an engineering standpoint, we need to look closely at what they actually built and measured. How close are these theoretical predictions to what you can actually achieve on hardware? Lev.

Kai: Well, the paper does show experimental validation using transmission spectra for theta equal to pi over four in figures two c through e Kai.

Mira: And they compare the response of this CP-based sensor directly against the diabolic point based sensor shown in figure three a Mira.

Lev: What's striking is that the diabolic point sensor shows a linear response, fitting nicely with a line, while everything else—VP, AP and HP space—shows an abrupt change from zero to a square-root dependence on the magnetic field B Lev.

The paper's summary: Kai: To summarize what this paper is saying about the CP-based Hermitian sensor, it shows that below a certain bias level, where eta epsilon b is less than or equal to eta CP, the transmission just looks like a single peak Kai.

Mira: And at that critical bias level, eta CP, the transmission profile changes to a plateau shape Mira.

Lev: Then as you push the bias slightly above that point, they find this spectral extremum splitting increases rapidly with the perturbation. The paper clearly shows this transition from a linear dependence to a square root dependence on the perturbation right at eta CP Lev.

Kai: This square root behavior is what clearly manifests as a critical point, which they call CP Kai.

Mira: And what's really interesting is that this enhanced sensitivity breaks the Petermann factor limit that you see in non-Hermitian counterparts because of inherent eigenvector orthogonality in this CP-based Hermitian sensor Mira.

Lev: That lack of noise amplification due to orthogonal eigenvectors is a huge win for real hardware, because it completely avoids the strong gain induced noise that plagues active non-Hermitian sensors Lev.

The paper's improvements: Kai: The authors suggest a few ways this concept can be improved. They point out that the sensitivity can be further increased by three to four orders if you sweep the laser central frequency really fast using an acousto-optic modulator Kai.

Mira: And they also mention that combining this with extra stabilization to reduce technical noises could help pave the way for testing the standard model and searching for axions Mira.

Lev: If we take that suggestion about sweeping the frequency fast, we need to think about the hardware requirements. That level of sweep speed puts a lot of strain on our control systems and how much noise you can tolerate before it gets washed out Lev.

Kai: They also focus on using AI models to predict and optimize operating conditions near these critical points where that square root dependence transitions exhibit that square root behavior Kai.

Mira: And they are also looking at using AI frameworks to dynamically select between different measurement modes like VP, AP, and HP based on the calculated CP position gBCP derived from the polarization orientation theta Mira.

Lev: I'd want to see how robust those AI models are when the system is right near that exceptional point, because if they rely on precise calculations of gBCP, any small error in that calculation could lead to a huge error in performance Lev.

Conclusion: Kai: So to wrap up this paper on "Exceptional-point-like Sensing near Hermitian Critical Points," they demonstrate CP-enhanced sensing of refraction in a chiral and passive Hermitian optical system Kai.

Mira: The main point is that this CP-based sensor has an EP-like response and sensitivity that is well beyond what the diabolic point based sensor can do Mira.

Lev: It’s a simple sensor, but it’s ultrasensitive to small changes in the refractive index, and they claim it can achieve a refraction sensitivity dramatically superior to ten to the power of negative eleven over square root Hertz for the VP measurement Lev <ref:2601.17266#pg1>.

Kai: And they showed that even when intensity noise dominates other contributions, the noise amplification factors are close to unity in all three measurement scenarios, clearly confirming that the Petermann factor is one Kai.

Mira: So they confirm that by using this CP-based sensor, you avoid the Petermann factor limit because of the inherently orthogonal eigenvectors Mira.

Lev: From an engineering view, seeing noise amplification factors close to unity means we don't have to worry about massive noise amplification during operation, which makes this a very practical concept for building real sensors Lev.

Kai: This work paves the way for facilitating tests of the standard model and searching for axions by using these simple but ultrasensitive sensors Kai.

Mira: It’s a solid piece of work showing how critical points in Hermitian systems can unlock better sensing capabilities than we thought possible Mira.

College of Engineering and Applied Sciences, National Laboratory of Solid State Microstructures, Nanjing University · College of Advanced Interdisciplinary Studies, National University of Defense Technology Changsha · Xi’an Institute of Applied Optics Xi’an · School of Physics, Nanjing University · Laboratory of Solar and Space Instruments, Nanjing Institute of Astronomical Optics and Technology Chinese Academy of Sciences · Beijing National Laboratory for Condensed Matter Physics Institute of Physics Chinese Academy of Sciences Beijing National Laboratory for Condensed Matter Physics

physics.optics

Submitted: 2026-01-24

Updated: 2026-01-24

Comments: 23 pages, 10 figures

Journal ref: Chinese Phys. Lett. 43 (2026) 110402

DOI: 10.1088/0256-307X/43/11/110402

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 83/100

The gist: A non-Hermitian system at an exceptional point (EP), a specific critical point (CP) associated with the parity-time symmetric phase transition…

Key concepts

Exceptional Point (EP)
An EP is a specific critical point in non-Hermitian systems where two eigenvalues and their corresponding eigenvectors coalesce. This coalescence leads to unique physical phenomena, such as the divergence of the Petermann factor, which is key to understanding the system's extreme sensitivity.
Critical Point (CP)
A CP is a specific critical point associated with a parity-time symmetric phase transition in Hermitian systems. At this point, the spectral extremum splitting shows a square-root dependence on perturbation, signaling a unique sensing characteristic.
Petermann Factor (PF)
The PF quantifies non-orthogonality between the eigenmodes of a system. For Hermitian systems, PF is 1. However, near an EP in non-Hermitian systems, the PF diverges. This divergence indicates maximal non-orthogonality and noise amplification.
Sublinear Response
This refers to a sensing scheme where the response to a small change in the refractive index follows a sublinear relationship rather than a linear one. This behavior is achieved by exploiting critical points, allowing for enhanced sensitivity beyond traditional linear-response sensors.

Terminology

Summary

A non-Hermitian system at an exceptional point (EP), a specific critical point (CP) associated with the parity-time symmetric phase transition, exhibits a sublinear response to perturbation and promise unprecedented sensitivity beyond the linear-response Hermitian sensors, so far operating at the diabolic points (DP)

System Comparison and Response Characteristics

The paper compares conventional Hermitian DP-based sensors, non-Hermitian sensors, and the proposed CP-based sensor The CP-based Hermitian sensor exhibits an EP-like response and enhanced sensitivity breaking the Petermann-factor limit of sensitivity in nonHermitian counterparts This is achieved because of the inherent eigenvector orthogonality, this CP-based Hermitian sensor exhibits an EP-like response and enhanced sensitivity The CP-based sensor measures the extremum separation of transmittance spectrum, ∆µ, see the right panel in Fig. 1e. Below a bias level that η(ϵb) ≤ ηCP, the transmission is a single peak. p. 1

Where the paper says it

Unlike the DP-based sensor observing the eigenfrequency splitting, our CP-based sensor measures the extremum separation of transmittance spectrum, ∆µ, see the right panel in Fig. Below a bias level that η(ϵb) ≤ ηCP, the transmission is a single peak.

At ηCP, the transmission shows a plateau profile. p. 1
Where the paper says it

Below a bias level that η(ϵb) ≤ ηCP, the transmission is a single peak. At ηCP, the transmission shows a plateau profile.

Slightly biased above this point, the spectralextremum splitting (SES) increases rapidly with the perturbation. p. 1
Where the paper says it

At ηCP, the transmission shows a plateau profile. Slightly biased above this point, the spectralextremum splitting (SES) increases rapidly with the perturbation.

The SES displays a transition from a linear dependence to a square-root dependence on the perturbation at ηCP, clearly manifesting a CP. p. 1
Where the paper says it

The two peaks locate at opposite sides of a spectral dip. The SES, defined as ∆µ = µ+ − µ−, displays a transition from a linear dependence to a square-root dependence on the perturbation at ηCP, clearly manifesting a CP.

Noise Amplification and Noise Mitigation

The sensitivity is more critical than the response for a sensor The enhanced dynamic response have been successfully demonstrated in the gain-loss non-Hermitian sensors. Inactive non-Hermitian sensors exhibit an important advantage over its active counterpart, because it completely avoids the strong gain-induced noise. p. 1

Where the paper says it

The PF-amplified noise is thus partly suppressed and a threefold enhancement is achieved for the signal-to-noise ratio. A passive non-Hermitian sensor exhibits an important advantage over its active counterpart, because it completely avoids the strong gain-induced noise.

The CP-based sensor contains no gain medium and its eigenvectors are inherently orthogonal, completely avoiding the PF problem because PF remains unity. p. 5
Where the paper says it

Remarkably, our Hermitian sensor contains no gain medium. Its eigenvectors are inherently orthogonal, completely avoiding the PF problem because PF remains unity.

The total noise in a Hermitian sensor reduces to σ2DP/CP = σ2shot + σ2cl + R2DP/CPσ2signal + σ20. p. 6
Where the paper says it

sensor reduces to σ 2 DP/CP = σ 2 shot + σ 2 cl + R 2 DP/CPσ 2 signal + σ 2 0 . (4) In a non-Hermitian sensor, these two noise contributions are greatly amplified by the divergent PF near the EP.

In a non-Hermitian sensor, these two noise contributions are greatly amplified by the divergent PF near the EP. p. 6
Where the paper says it

sensor reduces to σ 2 DP/CP = σ 2 shot + σ 2 cl + R 2 DP/CPσ 2 signal + σ 2 0 . (4) In a non-Hermitian sensor, these two noise contributions are greatly amplified by the divergent PF near the EP.

In our CP-based sensor, the noise amplification, defined as Gσ = σCP/σDP, is eliminated because the eigenmodes are orthogonal. p. 6
Where the paper says it

(4) In a non-Hermitian sensor, these two noise contributions are greatly amplified by the divergent PF near the EP. In our CP-based sensor, the noise amplification, defined as Gσ = σCP/σDP, is eliminated because the eigenmodes are orthogonal.

Experimental Validation and Performance Metrics

The experimental transmission spectra for θ = π/4 shown in Figs. 2c-e confirm the depicted concept for the CP-based sensor in Fig. p. 6

Where the paper says it

Results The experimental transmission spectra for θ = π/4 shown in Figs. 2c-e confirm the depicted concept for the CP-based sensor in Fig.

1e. The responses of the DP-based and CP-based sensor are shown in Fig. p. 7
Where the paper says it

The CP is first observed in the VP spectrum, secondly in the AP and then in the HP, in good agreement with the theory (see the asterisks in Fig. The responses of the DP-based and CP-based sensor are shown in Fig.

3a. The DP-based sensor displays a linear response (purple), well fitted by a line. p. 7
Where the paper says it

The responses of the DP-based and CP-based sensor are shown in Fig. The DP-based sensor displays a linear response (purple), well fitted by a line.

In contrast, all three responses in the VP, AP and HP space exhibit an abrupt change from null to a square-root dependence on the magnetic field B. p. 7
Where the paper says it

The DP-based sensor displays a linear response (purple), well fitted by a line. In contrast, all three responses in the VP, AP and HP space exhibit an abrupt change from null to a square-root dependence on the magnetic field B, see the log-log plot in the inset.

Above BCP, the experimental responses are well fitted by the theoretical results given in Eq. p. 7
Where the paper says it

In contrast, all three responses in the VP, AP and HP space exhibit an abrupt change from null to a square-root dependence on the magnetic field B, see the log-log plot in the inset. Above BCP, the experimental responses are well fitted by the theoretical results given in Eq.

2. The response enhancement is defined as GR,m = RCP,m/RDP(m = VP, AP, HP). p. 7
Where the paper says it

This feature is not observed in the EP- and TPD-based sensors. In comparison with the DP-sensor, the response enhancement is clearly observed in the CP-based measurements and is evaluated as GR = RCP/RDP, see Figs.

Theoretically, these response enhancements approaches to divergent values near the CP. p. 6
Where the paper says it

In our CP-based sensor, the noise amplification, defined as Gσ = σCP/σDP, is eliminated because the eigenmodes are orthogonal. Clearly, when the signal-related contribution is negligible, leading to σCP ≈ σDP, the sensitivity enhancement, evaluated as GS = SDP/SCP = GR/Gσ, approaches to the response improvement GR.

Conclusion and Outlook

We have demonstrated CP-enhanced sensing of refraction in a chiral and passive Hermitian optical system This CP-based sensor exhibits an EP-like response and sensitivity well beyond the DP-based counterpart. Our sensor is very simple but ultrasensitive to a small change of the refractive index. p. 8

Where the paper says it

Such a general sensing method can be applied to diverse bipartite-coupled platforms and is expected to significantly advance the ongoing sensor revolution [47]. Our sensor is very simple but ultrasensitive to a small change of the refractive index.

p. 8
Where the paper says it

Such a general sensing method can be applied to diverse bipartite-coupled platforms and is expected to significantly advance the ongoing sensor revolution [47]. Our sensor is very simple but ultrasensitive to a small change of the refractive index.

The sensitivity can be further improved by three to four orders if the laser central frequency is fast swept with an AOM. p. 8
Where the paper says it

Our sensor is very simple but ultrasensitive to a small change of the refractive index. The sensitivity can be further improved by three to four orders if the laser central frequency is fast swept with an AOM.

p. 8
Where the paper says it

Our sensor is very simple but ultrasensitive to a small change of the refractive index. The sensitivity can be further improved by three to four orders if the laser central frequency is fast swept with an AOM.

In combination with extra stabilization reducing technical noises, this work paves the way for facilitating the test of the standard model and searching for axions. p. 8
Where the paper says it

The sensitivity can be further improved by three to four orders if the laser central frequency is fast swept with an AOM. In combination with extra stabilization reducing technical noises, this work paves the way for facilitating the test of the standard model and searching for axions [48, 49], which cause a very tiny change to the refraction index of vacuum.

p. 8
Where the paper says it

The sensitivity can be further improved by three to four orders if the laser central frequency is fast swept with an AOM. In combination with extra stabilization reducing technical noises, this work paves the way for facilitating the test of the standard model and searching for axions [48, 49], which cause a very tiny change to the refraction index of vacuum.

Methods and Theoretical Framework

The system is governed by a Hamiltonian H = ν0 − iκ igB − igB ν0 − iκ. The MO-induced interaction converts the eigenmodes from linear polarization to circularly polarization such that aˆ+ = (ˆaH +iaˆV)/√2 and aˆ− = (ˆaH − iaˆV)/√2. p. 3

Where the paper says it

It allows the DP- and CP-based measurement and thus facilitates comparison. The MO-induced interaction converts the eigenmodes from linear polarization to circularly polarization such that aˆ+ = (ˆaH +iaˆV)/ √ 2 and aˆ− = (ˆaH −iaˆV)/ √ 2.

The eigenvalue of the system can be easily obtained as ν± = ν0 − iκ ± p g2B2 + δ2. p. 9
Where the paper says it

We first consider the Hermitian case, where κH = κV = κ. The eigenvalue of the system can be easily obtained as ν± = ν0 − iκ ± p g 2B2 + δ 2 .

The CP position is determined by numerically solving µ+,VP − ν0 = µ−,VP − ν0 = 0, as shown in the orange solid curve in Fig. p. 1
Where the paper says it

1e and the inset in Fig. Therefore, the dynamic response (blue curve) is enhanced by a divergent factor ∝ 1/ √ ϵ compared to a conventional Hermitian sensor, as shown in Fig.

2b. The SES in the CP-based sensor is defined as ∆µm = µ+,m −µ−,m(m = VP, AP, HP). p. 1
Where the paper says it

The SES, defined as ∆µ = µ+ − µ−, displays a transition from a linear dependence to a square-root dependence on the perturbation at ηCP, clearly manifesting a CP. Comparing to the DP-based sensor, the dynamic response is enhanced by a factor of √ ϵ near the CP.

Petermann Factor Derivation

The Petermann factor PF = ⟨ψLψL⟩⟨ψRψR⟩, which for two dimensional matrices gives the closed expression PF = 1/2 (1 + Tr H†0H0 Tr(H20)). Clearly, we have PF = 1 for Hermitian systems. p. 1

Where the paper says it

The enhanced dynamic response have been successfully demonstrated in the gain-loss non-Hermitian sensors [8–11]. In these active systems, one mode is amplified to balance the loss of the other.

p. 1
Where the paper says it

The enhanced dynamic response have been successfully demonstrated in the gain-loss non-Hermitian sensors [8–11]. In these active systems, one mode is amplified to balance the loss of the other.

For non-Hermitian systems at the exceptional point (EP) where η = ∆κ, the two eigenmodes coalesce and the PF diverges, reflecting maximal non-orthogonality. p. 0
Where the paper says it

A system near a CP is very sensitive to a perturbation and can be utilized as an ultrasensitive sensor. By exploiting the parity-time (PT ) symmetric phase transition, a non-Hermitian sensor near the so-called exceptional point (EP), where its eigenvalues and eigenvectors simultaneously coalesce [1–4], has shown the essential advantages in response [5–21] and sensitivity [22–24] over a conventional Hermitian counterpart operating near the diabolic point (DP).

p. 0
Where the paper says it

A system near a CP is very sensitive to a perturbation and can be utilized as an ultrasensitive sensor. By exploiting the parity-time (PT ) symmetric phase transition, a non-Hermitian sensor near the so-called exceptional point (EP), where its eigenvalues and eigenvectors simultaneously coalesce [1–4], has shown the essential advantages in response [5–21] and sensitivity [22–24] over a conventional Hermitian counterpart operating near the diabolic point (DP).

The final expression demonstrates that both the fundamental quantum noise Θ and the technical drive amplitude noise Sα are amplified by a factor of PF. p. 5
Where the paper says it

For a real signal from environment, this contribution disappears. The final term σ0 reflects the background noise imposed by technical constraints including the imperfect quantum efficiency and dark electric currents of the photoelectric detectors.

Fitting of Experimental Results

The extracted coupling strength g is 0.90 MHz/mT for the DP sensor, and 0. p. 7

Where the paper says it

The fitting parameter g takes the values of 0. 90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively.

p. 7
Where the paper says it

The fitting parameter g takes the values of 0. 90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively.

89, 0.89, and 0.90 MHz/mT for the VP, AP and HP measurements of the CP sensor. p. 7
Where the paper says it

The fitting parameter g takes the values of 0. 90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively.

p. 7
Where the paper says it

The fitting parameter g takes the values of 0. 90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively.

The close values indicate a good agreement between the experimental observations and theoretical predictions. p. 7
Where the paper says it

90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively. These close values indicate a good agreement between the experimental observations and theoretical predictions.

p. 7
Where the paper says it

90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively. These close values indicate a good agreement between the experimental observations and theoretical predictions.

The extracted coupling coefficients g = 0. p. 1
Where the paper says it

the measured frequency separation as a system response χ in the δ − η parameter space (inset surfaces), where δ is half of the frequency detuning between the two modes, the perturbation-induced coupling strength η is proportional to the normalized perturbation ϵ (ϵ ≪ 1), the dynamic response R (blue curves), defined as the derivative of the response with respect to perturbation, i. R = ∂χ/∂ϵ.

p. 1
Where the paper says it

the measured frequency separation as a system response χ in the δ − η parameter space (inset surfaces), where δ is half of the frequency detuning between the two modes, the perturbation-induced coupling strength η is proportional to the normalized perturbation ϵ (ϵ ≪ 1), the dynamic response R (blue curves), defined as the derivative of the response with respect to perturbation, i. R = ∂χ/∂ϵ.

882, 0.874, 0.863 MHz mT−1 and the decay rates of κ = 2. p. 3
Where the paper says it

We use νH and νV to denote their resonance frequencies, respectively. The left and right mirrors causes external decay rates κ ex H and κ ex V to the HP and VP modes, respectively.

p. 3
Where the paper says it

We use νH and νV to denote their resonance frequencies, respectively. The left and right mirrors causes external decay rates κ ex H and κ ex V to the HP and VP modes, respectively.

403, 2.466, and 2.433 MHz. The very close coupling coefficients and the decay rates validate our experimental observations. p. 7
Where the paper says it

90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively. These close values indicate a good agreement between the experimental observations and theoretical predictions.

p. 7
Where the paper says it

90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively. These close values indicate a good agreement between the experimental observations and theoretical predictions.

Noise Amplification Analysis

The noise amplification is defined as Gσ = σCP/σDP = q [σ2other + (RCPγCPB)2]/[σ2other + (RDPγDPB)2]. Clearly, when σsignal is negligible, Gσ ≈ 1 (see Figs. p. 6

Where the paper says it

In our CP-based sensor, the noise amplification, defined as Gσ = σCP/σDP, is eliminated because the eigenmodes are orthogonal. Clearly, when the signal-related contribution is negligible, leading to σCP ≈ σDP, the sensitivity enhancement, evaluated as GS = SDP/SCP = GR/Gσ, approaches to the response improvement GR.

4b and 4c). The sensitivity enhancement is GS = GR/Gσ. p. 6
Where the paper says it

In our CP-based sensor, the noise amplification, defined as Gσ = σCP/σDP, is eliminated because the eigenmodes are orthogonal. Clearly, when the signal-related contribution is negligible, leading to σCP ≈ σDP, the sensitivity enhancement, evaluated as GS = SDP/SCP = GR/Gσ, approaches to the response improvement GR.

When the intensity noise dominates other contributions, the noise amplification factors are close to unity in all three measurement scenarios, clearly confirming that PF = 1. p. 7
Where the paper says it

As shown in Fig. 4b, when the intensity noise dominates other contributions, the noise amplification factors are close to unity in all three measurement scenarios, clearly confirming that PF = 1.

The sensitivity enhancement reaches larger values of 7. p. 7
Where the paper says it

4b, when the intensity noise dominates other contributions, the noise amplification factors are close to unity in all three measurement scenarios, clearly confirming that PF = 1. The sensitivity enhancement reaches larger values of 7.

51, 5.49 and 7.40 in the VP, AP and HP measurements, respectively. p. 7
Where the paper says it

The fitting parameter g takes the values of 0. 90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively.

Data Availability

The data that support the plots within this paper are available via figshare All other data used in this study are available from the corresponding authors upon reasonable request. The acknowledgements state that this work was supported by the Innovation Program for Quantum Science and Technology (Grants No. p. 0

Where the paper says it

Exceptional-point-like Sensing near Hermitian Critical Points Jiang-Shan Tang1,2,∗ , Long-Qi Xiao1,∗ , Hao-Dong Wu1,3 , Yuwei Jing1 , Han Zhang4 , Ya-Ping Ruan5 , Wuming Liu6,† , Yan-Qing Lu1,† , Keyu Xia1,7,8,† 1College of Engineering and Applied Sciences, National Laboratory of Solid State Microstructures,Nanjing University, Nanjing, 210023, China . 2College of Advanced Interdisciplinary Studies, National University of Defense Technology, Changsha, 410073, China.

2021ZD0301400), the National Natural Science Foundation of China (Grants No. p. 0
Where the paper says it

Exceptional-point-like Sensing near Hermitian Critical Points Jiang-Shan Tang1,2,∗ , Long-Qi Xiao1,∗ , Hao-Dong Wu1,3 , Yuwei Jing1 , Han Zhang4 , Ya-Ping Ruan5 , Wuming Liu6,† , Yan-Qing Lu1,† , Keyu Xia1,7,8,† 1College of Engineering and Applied Sciences, National Laboratory of Solid State Microstructures,Nanjing University, Nanjing, 210023, China . 2College of Advanced Interdisciplinary Studies, National University of Defense Technology, Changsha, 410073, China.

U25A6008, No. 92365107, No. 12305020, No. 12334012, and No. 62575225), the National Key R&D Program of China (Grants No. p. 0
Where the paper says it

Exceptional-point-like Sensing near Hermitian Critical Points Jiang-Shan Tang1,2,∗ , Long-Qi Xiao1,∗ , Hao-Dong Wu1,3 , Yuwei Jing1 , Han Zhang4 , Ya-Ping Ruan5 , Wuming Liu6,† , Yan-Qing Lu1,† , Keyu Xia1,7,8,† 1College of Engineering and Applied Sciences, National Laboratory of Solid State Microstructures,Nanjing University, Nanjing, 210023, China . 2College of Advanced Interdisciplinary Studies, National University of Defense Technology, Changsha, 410073, China.

2019YFA0308700 and No. 2019YFA0308704), the Program for Innovative Talents and Teams in Jiangsu (Grant No. JSSCTD202138), China Postdoctoral Science Foundation (Grant No. p. 0
Where the paper says it

Exceptional-point-like Sensing near Hermitian Critical Points Jiang-Shan Tang1,2,∗ , Long-Qi Xiao1,∗ , Hao-Dong Wu1,3 , Yuwei Jing1 , Han Zhang4 , Ya-Ping Ruan5 , Wuming Liu6,† , Yan-Qing Lu1,† , Keyu Xia1,7,8,† 1College of Engineering and Applied Sciences, National Laboratory of Solid State Microstructures,Nanjing University, Nanjing, 210023, China . 2College of Advanced Interdisciplinary Studies, National University of Defense Technology, Changsha, 410073, China.

2023M731613), and Jiangsu Funding Program for Excellent Postdoctoral Talent (Grant No. 2023ZB708). The author contributions detail the roles of each researcher in the project. p. 5
Where the paper says it

The more important sensitivity S is defined as the ratio of measurement uncertainty or noise σ to the dynamic response R with in the total measurement τ , namely S = σ/R√ τ [43]. In a gain-loss non-Hermitian sensor operating near the EP, the total noise basically includes five contributions arising from various sources.

The supplementary materials provide details on experimental setups and noise characterization. p. 5
Where the paper says it

Below, we experimentally examine the CP-enhanced response and sensitivity at these three CPs (see Supplementary Material for details in experimental setups). The more important sensitivity S is defined as the ratio of measurement uncertainty or noise σ to the dynamic response R with in the total measurement τ , namely S = σ/R√ τ [43].

Supplementary Material Details

The main component of our experiment is a high-quality magneto-optical (MO) Fabry-Pérot (FP) cavity, as shown in Figs. 5 and 6. A 20 mm-long (111)-cut cylindrical terbium gallium garnet (TGG) crystal is placed at the cavity center. p. 3

Where the paper says it

The effective length of the cavity is Lc = 80 mm. The MO medium is a (111)-cut cylindrical terbium gallium garnet (TGG) crystal, made by the company.

The probe beam is generated by a tunable external cavity diode laser (Toptica DLC pro), and its intensity is modulated by an acousto-optic modulator (AOM) driven by an applied voltage V. p. 3
Where the paper says it

To validate this concept, we examine the response and sensitivity enhancement by artificially adding intensity noise to the probe field and the output field. We artificially add a strong intensity noise with a zero mean value to the probe field by randomly modulating the probe field via an acousto-optic modulator (AOM).

The output beam is adjusted by two convex lenses to yield a waist diameter of approximately 225 µm at the cavity center. p. 3
Where the paper says it

Simulations of this type of noise is performed by adding a noise field randomly modulated in its amplitude into the output fields via a beam splitter [42] (see details in Supplementary Material). Theory The Fabry-Pérot cavity supports two optical modes: the HP and VP modes, denoted as aˆH and aˆV.

The noise distributions are presented on the right panels of the noises. p. 7
Where the paper says it

Figure 4 compares the noise, and the sensitivity of the CP- and DP-based sensors in three cases of no artificial noise (left panel), adding intensity noise (middle panel) and background noise (right panel). The lines and data for the DP-based sensor are presented for comparison.

The experimental data for the sensitivity, sensitivity enhancement and the noise amplification without artificial noise are fitted (see Methods for details). p. 8
Where the paper says it

Red, blue, and green data points correspond to the VP, AP, and HP measurements, respectively. Experimental data for the sensitivity, sensitivity enhancement and the noise amplification without artificial noise are fitted (see Methods for details).

The results show that in all three cases, the experimental responses are similar in all three cases, meaning that they are independent on the noise. p. 7
Where the paper says it

2 in each measurement scenario (see Methods). As indicated by the theoretical model, the experimentally observed responses are similar in all three cases, meaning that they are independent on the noise.

The CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor. p. 7
Where the paper says it

Figures 4g-i show the sensitivities. It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor.

p. 7
Where the paper says it

Figures 4g-i show the sensitivities. It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor.

p. 7
Where the paper says it

Figures 4g-i show the sensitivities. It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor.

p. 7
Where the paper says it

Figures 4g-i show the sensitivities. It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor.

p. 7
Where the paper says it

Figures 4g-i show the sensitivities. It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor.

p. 7
Where the paper says it

Figures 4g-i show the sensitivities. It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor.

The VP measurement achieves a refraction sensitivity dramatically superior to 10−11/√Hz. p. 7
Where the paper says it

It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor. Particularly, the VP measurement achieves a refraction sensitivity dramatically superior to 10−11/ √ Hz.

p. 7
Where the paper says it

It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor. Particularly, the VP measurement achieves a refraction sensitivity dramatically superior to 10−11/ √ Hz.

p. 7
Where the paper says it

It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor. Particularly, the VP measurement achieves a refraction sensitivity dramatically superior to 10−11/ √ Hz.

p. 7
Where the paper says it

It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor. Particularly, the VP measurement achieves a refraction sensitivity dramatically superior to 10−11/ √ Hz.

p. 7
Where the paper says it

It is clear that the CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor. Particularly, the VP measurement achieves a refraction sensitivity dramatically superior to 10−11/ √ Hz.

The noise amplification is shown to be close to unity in all three measurement scenarios, clearly confirming that PF = 1. p. 7
Where the paper says it

As shown in Fig. 4b, when the intensity noise dominates other contributions, the noise amplification factors are close to unity in all three measurement scenarios, clearly confirming that PF = 1.

p. 7
Where the paper says it

As shown in Fig. 4b, when the intensity noise dominates other contributions, the noise amplification factors are close to unity in all three measurement scenarios, clearly confirming that PF = 1.

p. 7
Where the paper says it

As shown in Fig. 4b, when the intensity noise dominates other contributions, the noise amplification factors are close to unity in all three measurement scenarios, clearly confirming that PF = 1.

The CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor. The VP measurement achieves a refraction sensitivity dramatically superior to 10−11/√Hz. The noise amplification is shown to be close to unity in all three measurement scenarios, clearly confirming that PF = 1. The CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor. The VP measurement achieves a refraction sensitivity dramatically superior to 10−11/√Hz.

Characterization of Noise Injection

To emulate laser intensity noise, a uniformly distributed random modulation is superimposed onto the AOM drive voltage. The additional fluctuations increase the variance to δV2 = 9 × 10−5 V2. p. 6

Where the paper says it

The insert shows the same data in a double-logarithmic fashion. b-d, Response enhancement, GR, of the CP sensor relative to the DP sensor without additional noise (b), with added laser intensity noise (c), and with background noise (d).

In this case, the transmitted laser field is subject to white-noise amplitude modulation, thereby providing a controlled setup for testing the PF amplification of these fluctuations. p. 7
Where the paper says it

In our experiment, both the DP- and CP-based sensors directly measures the intensity of transmitted fields. Thus, the intensity noise of the probe laser field has the same effect as the shot noise and thus can also simulate the shot noise.

The noise amplification displays almost the same level as in the case of adding intensity noise (Fig. p. 7
Where the paper says it

In this simulations, the artificial background noise is dominant. The noise amplifications display almost the same level as in the case of adding intensity noise (Fig.

4c). The sensitivity enhancements are close to the response improvement factors (Fig. p. 7
Where the paper says it

The noise amplifications display almost the same level as in the case of adding intensity noise (Fig. The sensitivity enhancements are close to the response improvement factors (Fig.

4f), reaching a maximum of about 9. p. 7
Where the paper says it

The sensitivity enhancements are close to the response improvement factors (Fig. 4f), reaching a maximum of about 9.

64 in the VP measurement. p. 1
Where the paper says it

The sensitivity is more critical than the response for a sensor. It is crucially dependent on both the dynamic response and the noise in measurement [40].

The CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor. The VP measurement achieves a refraction sensitivity dramatically superior to 10−11/√Hz.

Characterization of Orthogonal Cavity Modes

At zero field, the cavity eigenmodes are HP and VP. These two modes are expected to be degenerate in both resonance frequency and linewidth. p. 3

Where the paper says it

Considering the intrinsic losses κ in H and κ in V , we respectively denote the total losses of the HP and VP modes as κH = 2κ ex + κ in H and κV = 2κ ex + κ in V . This HP and VP eigenmodes are linearly polarized and degenerate in resonance frequency such that νH = νV = ν0 and in dissipation such that κH = κV = κ, leading to vanishing detuning δ = 0 and gain/loss contrast ∆κ = 0.

The frequency difference between the HP and VP modes is found to be 0. p. 1
Where the paper says it

the measured frequency separation as a system response χ in the δ − η parameter space (inset surfaces), where δ is half of the frequency detuning between the two modes, the perturbation-induced coupling strength η is proportional to the normalized perturbation ϵ (ϵ ≪ 1), the dynamic response R (blue curves), defined as the derivative of the response with respect to perturbation, i. R = ∂χ/∂ϵ.

0069 MHz, which is negligibly small compared to the linewidth. p. 7
Where the paper says it

The fitting parameter g takes the values of 0. 90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively.

These observations confirm that, at B = 0, the HP and VP eigenmodes are well degenerate in both the resonance frequency and the decay rate. p. 3
Where the paper says it

Considering the intrinsic losses κ in H and κ in V , we respectively denote the total losses of the HP and VP modes as κH = 2κ ex + κ in H and κV = 2κ ex + κ in V . This HP and VP eigenmodes are linearly polarized and degenerate in resonance frequency such that νH = νV = ν0 and in dissipation such that κH = κV = κ, leading to vanishing detuning δ = 0 and gain/loss contrast ∆κ = 0.

The small difference of 0. p. 6
Where the paper says it

2c-e confirm the depicted concept for the CP-based sensor in Fig. For B < BCP, corresponding to a small ∆n, the VP, AP and HP transmissions exhibit singlet-peak structures, implying a null SES.

06 MHz arises from unavoidable experimental imperfections but can be safely neglected. p. 7
Where the paper says it

90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively. These close values indicate a good agreement between the experimental observations and theoretical predictions.

Petermann Factor and Amplification of Noise

The effective Hamiltonian of the coupled system takes the form Hc = − (∆ − iκH) ˆa†H aˆH − (∆ - iκV) ˆa†V aˆV + iη a†H aˆV − a†V aˆH. The real part of λ corresponds to the mode frequency shift, while the imaginary part describes effective the gain or loss of the eigenmode. p. 1

Where the paper says it

Conventionally, a Hermitian two-mode sensor with a vanishing gain/loss contrast, ∆κ = 0, operates at the DP where δ = 0 and ϵb = 0. For any perturbation, its eigenfrequency splitting follows an ϵ-dependence, ∆ν = Re[ν+]−Re[ν−] ∝ ϵ, as shown by the real parts Re[ν±] of the two complex eigenfrequencies in Fig.

The non-orthogonality is quantified by the Petermann factor PF = ⟨ψLψL⟩⟨ψRψR⟩, which for two dimensional matrices gives the closed expression PF = 1/2 (1 + Tr H†0H0 Tr(H20)). p. 0
Where the paper says it

Here, we report the CP-resulting square-root response to the refractive index change and enhanced sensitivity in a simple chiral Hermitian cavity without phase transitions. Because of the inherent eigenvector orthogonality, this CP-based Hermitian sensor exhibits an EP-like response and enhanced sensitivity, breaking the Petermann-factor limit of sensitivity in nonHermitian counterparts.

Clearly, we have PF = 1 for Hermitian systems. For non-Hermitian systems at the exceptional point (EP) where η = ∆κ, the two eigenmodes coalesce and the PF diverges, reflecting maximal non-orthogonality. This final expression demonstrates that both the fundamental quantum noise Θ and the technical drive amplitude noise Sα are amplified by a factor of PF. p. 5
Where the paper says it

For a real signal from environment, this contribution disappears. The final term σ0 reflects the background noise imposed by technical constraints including the imperfect quantum efficiency and dark electric currents of the photoelectric detectors.

The CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor. The VP measurement achieves a refraction sensitivity dramatically superior to 10−11/√Hz.

In this section, we present the theoretical fittings to the experimental results in the CP sensing without artificial noise as an example. The extracted g is 0.90 MHz/mT for the DP sensor, and 0.89, 0.89, and 0.90 MHz/mT for the VP, AP and HP measurements of the CP sensor. The close values indicate a good agreement between the experimental observations and theoretical predictions. The extracted coupling coefficients g = 0.882, 0.874, 0.863 MHz mT−1 and the decay rates of κ = 2.403, 2.466, and 2.433 MHz. The very close coupling coefficients and the decay rates validate our experimental observations. The theoretical model is in good agreement with the experimental results. p. 7

Where the paper says it

90 MHz/mT for the DP sensor, the VP, AP and HP measurements of the CP sensor, respectively. These close values indicate a good agreement between the experimental observations and theoretical predictions.

The noise amplification is shown to be close to unity in all three measurement scenarios, clearly confirming that PF = 1. The CP-based sensor exhibits better sensitivities in all measurement scenarios than the DP-based sensor.

We have demonstrated CP-enhanced sensing of refraction in a chiral and passive Hermitian optical system This CP-based sensor exhibits an EP-like response and sensitivity well beyond the DP-based counterpart. Our sensor is very simple but ultrasensitive to a small change of the refractive index. The sensitivity can be further improved by three to four orders if the laser central frequency is fast swept with an AOM. In combination with extra stabilization reducing technical noises, this work paves the way for facilitating the test of the standard model and searching for axions. The data that support the plots within this paper are available via figshare All other data used in this study are available from the corresponding authors upon reasonable request.

The system devices and parameters have been introduced in Sec.. A linearly polarized laser field with 9 frequency excites the MO Fabry-Pérot cavity with a drive amplitude α. p. 9

Where the paper says it

frequency ν excites the MO Fabry-Pérot cavity with a drive amplitude α. Its polarization orientates at an angle θ to the horizontal plane.

The Langevin equations of motion are given by daˆH dt = i h H, ˆ aˆH i + L[ˆaH] = (i∆H − κH) ˆaH + gBaˆV + √2κexα cos θ. p. 9
Where the paper says it

(5), the first two terms correspond to the free Hamiltonian, the third term describes the photon interaction, and the last two terms account for the external driving. The Langevin equations of motion are given by daˆH dt = i h H, ˆ aˆH i + L[ˆaH] = (i∆H − κH) ˆaH + gBaˆV + √ 2κ exα cos θ , (6a) daˆV dt = i h H, ˆ aˆV i + L[ˆaV] = (i∆V − κV) ˆaV − gBaˆH + √ 2κ exα sin θ .

The output fields follow from the input-output relations, aˆH,out = √2κexaˆH and aˆV,out = √2κexaˆV. p. 9
Where the paper says it

(6) to zero yields the system’s steady-state solution: ⟨aˆH⟩ = − √ 2κ exα [(i∆ + iδ − κV) cos θ − gB sin θ] (i∆ − iδ − κH) (i∆ + iδ − κV) + (gB) 2 , (7a) ⟨aˆV⟩ = − √ 2κ exα [gB cos θ + (i∆ − iδ − κH) sin θ] (i∆ − iδ − κH) (i∆ + iδ − κV) + (gB) 2 . (7b) The output fields follow from the input-output relations, aˆH,out = √ 2κ exaˆH and aˆV,out = √ 2κ exaˆV.

The VP, AP, and HP transmissions are defined as TVP = ⟨aˆV,out⟩2 / α2. p. 9
Where the paper says it

(7b) The output fields follow from the input-output relations, aˆH,out = √ 2κ exaˆH and aˆV,out = √ 2κ exaˆV. The VP, AP, and HP transmissions are defined as TVP = |⟨aˆV,out⟩|2/|α| 2 , TAP = |⟨aˆH⟩eH + ⟨aˆV⟩eV| 2/|α| 2 , and THP = |⟨aˆH⟩|2/|α| 2 , where {eH, eV} represent the HP and VP unitary vectors satisfying ei · ej = δi,j with i, j = H,V.

Hamiltonian and Transmission Spectrum

The system Hamiltonian is Hˆ = − (∆ − δ) ˆa†H aˆH − (∆ + δ) ˆa†V aˆV + igB a†H aˆV - a†V aˆH + i√2κexα cos θ a†H - α squared sin θ squared V.

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

  1. Move beyond linear response limitations by designing sensing schemes that exploit critical points in Hermitian systems to achieve sublinear response and enhanced sensitivity, as described by CP-resulting square-root response to the refractive index change.

  2. Develop AI models capable of predicting and optimizing operating conditions near critical points where spectral extremum splitting transitions exhibit a square-root dependence on perturbation, which is the signature of a CP.

  3. Implement noise mitigation strategies within sensing algorithms that leverage the inherent orthogonality of eigenvectors in Hermitian systems to avoid the Petermann factor limit, as shown by the eigenvectors of this CP-based Hermitian sensor are intrinsically orthogonal.

  4. Create AI systems that can dynamically select between different measurement modes (VP, AP, HP) based on the calculated CP position gBCP derived from polarization orientation theta.

  5. Design AI frameworks to incorporate noise amplification factors like PF into sensitivity calculations, allowing for real-time estimation of the noise amplification G sigma = σCP/σDP to quantify sensor performance under varying noise conditions.

Abstract

A non-Hermitian system at an exceptional point (EP), a specific critical point (CP) associated with the parity-time symmetric phase transition, exhibits a sublinear response to perturbation and promise unprecedented sensitivity beyond the linear-response Hermitian sensors, so far operating at the diabolic points (DP). Despite great advancements, its sensitivity enhancement is fundamentally limited by the divergent Petermann factor, intrinsically rooted in the non-Hermitian eigenvector degeneracy, and practically by the system complexity. Here, we report the CP-resulting square-root response to the refractive index change and enhanced sensitivity in a simple chiral Hermitian cavity without phase transitions. Because of the inherent eigenvector orthogonality, this CP-based Hermitian sensor exhibits an EP-like response and enhanced sensitivity, breaking the Petermann-factor limit of sensitivity in non-Hermitian counterparts. This work paves the way towards exploring the Hermitian CPs for ultrasensitive sensing outperforming both the EP- and DP-based sensors.

Related papers