{
"schema_version": 1,
"id": "PHYS-2026-08-09-01",
"canonical_url": "https://rimoooliii.github.io/physicsday/physics/PHYS-2026-08-09-01/",
"source_markdown_url": "https://rimoooliii.github.io/physicsday/physics/PHYS-2026-08-09-01.md",
"metadata": {
"schema_version": 1,
"id": "PHYS-2026-08-09-01",
"date": "2026-08-09",
"updated_at": "2026-08-09",
"title": "Entanglement alone does not make a quantum sensor robust",
"summary": "Use the Lindblad span, accessible Fisher information, and spatial-mode geometry to decide whether error correction or a decoherence-free sensor network can reject noise without erasing the signal.",
"language": "en",
"entry_kind": "daily",
"status": "published",
"level": "research",
"user_difficulty": "unrated",
"domains": [
"quantum-information",
"quantum-theory",
"mathematical-physics"
],
"estimated_minutes": 85
},
"content_markdown": "\n## The fracture\n\n**Entanglement can enlarge the response of a quantum sensor. It does not decide whether that response survives the dominant noise.**\n\nA GHZ state gives a phase proportional to the number of probes in an ideal unitary model. The same collective coherence can decay faster than a product-state coherence when each probe is noisy.\n\nError correction creates a sharper conflict. A code removes an operator direction by making it act identically on all logical states. If the signal occupies that direction, the code removes the signal as well.\n\nThe first design question is therefore geometric:\n\n$$\n\\boxed{\n\\text{Which component of the signal generator is distinguishable from the noise algebra?}\n}\n$$\n\nEntanglement becomes useful after that component has been identified and preserved.\n\n## 1. Sensitivity means accessible information\n\nLet a parameter $\\theta$ be encoded in a state $\\rho_\\theta$. A chosen measurement with outcomes $x$ produces probabilities $p(x\\mid\\theta)$ and classical Fisher information\n\n$$\nF_{\\rm C}(\\theta)\n=\\sum_x\\frac{[\\partial_\\theta p(x\\mid\\theta)]^2}{p(x\\mid\\theta)}.\n$$\n\nFor $\\nu$ independent repetitions and a locally unbiased estimator,\n\n$$\n\\operatorname{Var}(\\hat\\theta)\n\\geq \\frac{1}{\\nu F_{\\rm C}}\n\\geq \\frac{1}{\\nu F_{\\rm Q}}.\n$$\n\n$F_{\\rm Q}$ is the quantum Fisher information. It optimizes over measurements, so it is a ceiling on information available in the final quantum state. A laboratory measurement may attain less.\n\nFor a pure state evolving as $\\ket{\\psi_\\theta}=e^{-\\ii\\theta tG}\\ket\\psi$,\n\n$$\nF_{\\rm Q}=4t^2\\operatorname{Var}_\\psi(G).\n$$\n\nEntanglement can make $\\operatorname{Var}(G)$ scale as $N^2$. Noise can destroy that variance, prevent the optimal measurement, or introduce preparation and correction overhead that the bare QFI does not count.\n\nThis distinction prevents a common shortcut:\n\n$$\nF_{\\rm Q}\\sim N^2\n\\quad\\text{in an ideal model}\n\\not\\Rightarrow\n\\text{a robust }N^{-1}\\text{ experimental uncertainty}.\n$$\n\nThe standard review by [Pezzè et al.](https://journals.aps.org/rmp/abstract/10.1103/RevModPhys.90.035005) develops the relation among entanglement, QFI, measurements, and metrological scaling.\n\n## 2. The exact Markovian criterion\n\nConsider a finite-dimensional probe with parameter-independent Markovian noise:\n\n$$\n\\dot\\rho\n=-\\ii[\\theta G,\\rho]\n+\\sum_k\\left(\nL_k\\rho L_k^\\dagger\n-\\frac12\\{L_k^\\dagger L_k,\\rho\\}\n\\right).\n$$\n\nDefine the Hermitian Lindblad span\n\n$$\n\\begin{aligned}\n\\mathcal S=\\operatorname{span}_{\\mathbb R}\\{&I,\nL_k+L_k^\\dagger,\n\\ii(L_k-L_k^\\dagger),\\\\\n&L_k^\\dagger L_j+L_j^\\dagger L_k,\n\\ii(L_k^\\dagger L_j-L_j^\\dagger L_k)\\}_{j,k}.\n\\end{aligned}\n$$\n\nNumerical factors do not change the span. Including both Hermitian parts of $L_k^\\dagger L_j$ matters; leaving out the last term can give the wrong operator space.\n\nUnder noiseless ancillas and arbitrarily fast, accurate control, the result of Zhou, Zhang, Preskill, and Jiang is\n\n$$\n\\boxed{G\\notin\\mathcal S}\n$$\n\nif and only if error-corrected Heisenberg scaling in total sensing time is attainable. In this theorem,\n\n$$\nF_{\\rm Q}(T)=O(T^2)\n$$\n\nis the Heisenberg scaling, while failure of HNLS restricts the optimized QFI to at most $O(T)$ asymptotically.\n\nThis is the Hamiltonian-not-in-Lindblad-span, or HNLS, condition. The [original theorem](https://www.nature.com/articles/s41467-017-02510-3) also gives a semidefinite program for optimizing the code.\n\n> [!margin: Scope of the theorem]\n> HNLS is exact for the stated finite-dimensional, time-homogeneous Markovian model with suitable fast control. It is not an unconditional theorem about finite-rate hardware, colored noise, unknown dissipative parameters, or noisy ancillas.\n\nThe criterion is attached to the infinitesimal channel, not to a preferred notation for its jump operators. Unitary mixing of jump operators or the usual Lindblad gauge changes do not create a physical signal direction.\n\n## 3. Why correcting the noise can erase the signal\n\nLet $P$ project onto a two-dimensional sensing code. Correcting the infinitesimal error set requires conditions of the form\n\n$$\nPL_kP=\\lambda_kP,\n\\qquad\nPL_k^\\dagger L_jP=\\mu_{kj}P.\n$$\n\nInside the code, every correctable noise direction acts as a scalar. It cannot reveal which logical branch the probe occupies.\n\nSensing requires the opposite behavior from $G$:\n\n$$\nPGP\\neq gP.\n$$\n\nThe logical states must acquire different phases. If $G\\in\\mathcal S$, the correction conditions force $PGP$ to be scalar, so no logical phase remains.\n\nThis implication explains HNLS without reducing it to a commutator test.\n\nTake a qubit with\n\n$$\nG=\\frac12Z,\n\\qquad\nL=\\sqrt\\gamma Z.\n$$\n\nSignal and dephasing noise commute. They also lie along the same operator direction. A code that makes $Z$ invisible inside the logical space makes the frequency shift invisible there too.\n\nBy contrast, $[G,L]\\neq0$ does not by itself prove that $G$ lies outside $\\mathcal S$. Products $L_k^\\dagger L_j$ can generate directions absent from the list of individual $L_k$.\n\nCommutation answers whether two operations share a dynamical algebraic relation. HNLS answers whether the signal has any component that the entire infinitesimal noise channel cannot synthesize.\n\n## 4. The useful signal is a quotient direction\n\nWith the Hilbert-Schmidt inner product, decompose\n\n$$\nG=G_{\\parallel}+G_\\perp,\n\\qquad\nG_{\\parallel}\\in\\mathcal S,\n\\qquad\nG_\\perp\\perp\\mathcal S.\n$$\n\nHNLS says $G_\\perp\\neq0$. The correctable sensor uses this surviving component to generate a logical energy splitting.\n\nThe decomposition is a diagnostic. The optimal QFI prefactor also depends on spectral range, code constraints, ancillas, and control resources. A small $G_\\perp$ can satisfy HNLS while yielding a modest finite-time advantage.\n\nThe theorem separates two questions:\n\n1. **Scaling:** does any signal direction survive the Markovian noise algebra?\n2. **Performance:** how large and experimentally accessible is the logical generator built from that direction?\n\nThis is why “more entanglement” and “larger code distance” are incomplete objectives. Neither quantity identifies $G_\\perp$.\n\nYe and Zoller argue that programmable AMO platforms should combine entanglement, logical encoding, collective measurement, and networks for fundamental-physics searches.\n\nTheir [2024 Essay](https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.132.190001) supplies the program. HNLS supplies one exact design test within that program's ideal Markovian sector.\n\n## 5. Spatial filtering is the same geometry in a simpler algebra\n\nNow let $N$ sensors sample a field. Sensor $i$ has a diagonal generator $S_i$, and the field is expanded into one signal profile $\\mathbf s$ and nuisance profiles $\\mathbf n_a$:\n\n$$\nH(t)=\\gamma\\sum_{i=1}^N\n\\left[\n\\theta s_i+\\sum_{a=1}^r\\xi_a(t)n_{ai}\n\\right]S_i.\n$$\n\nChoose two product eigenstates labelled by opposite sensitivity vectors,\n\n$$\n\\mathbf w=(w_1,\\ldots,w_N),\n$$\n\nand prepare\n\n$$\n\\ket{\\psi_{\\mathbf w}}\n=\\frac{\\ket{\\mathbf w}+\\ket{-\\mathbf w}}{\\sqrt2}.\n$$\n\nThe nuisance phases cancel at every time if\n\n$$\n\\mathbf w\\cdot\\mathbf n_a=0\n\\quad\\text{for every }a.\n$$\n\nThe signal survives if\n\n$$\n\\mathbf w\\cdot\\mathbf s\\neq0.\n$$\n\nIf the columns of $N_{\\rm noise}$ are the nuisance profiles, the continuous optimum points along\n\n$$\n\\mathbf s_\\perp\n=\\left[I-N_{\\rm noise}\n(N_{\\rm noise}^{\\mathsf T}N_{\\rm noise})^+\nN_{\\rm noise}^{\\mathsf T}\\right]\\mathbf s,\n$$\n\nsubject to the sensitivity values allowed by the hardware. Here $+$ denotes the Moore-Penrose pseudoinverse.\n\nThus a protected spatial mode exists only when the signal has a component outside the nuisance-profile span. Discrete atomic levels can make the best allowed $\\mathbf w$ differ from the continuous projection.\n\nThis construction resembles HNLS, but the physical models differ. The spatial scheme cancels commuting phase profiles. General Lindblad noise includes quantum jumps and the product algebra $L_k^\\dagger L_j$.\n\n## 6. The three-ion realization\n\nSet the dimensionless positions to\n\n$$\n\\mathbf x=(-1,0,1)\n$$\n\nand expand\n\n$$\nB(x)=b_0+b_1x+\\frac{b_2}{2}x^2.\n$$\n\nThe constant, linear, and quadratic profiles are\n\n$$\n\\mathbf n_0=(1,1,1),\n\\quad\n\\mathbf n_1=(-1,0,1),\n\\quad\n\\mathbf s=\\left(\\frac12,0,\\frac12\\right).\n$$\n\nThe smallest nonzero integer solution of\n\n$$\n\\mathbf w\\cdot\\mathbf n_0\n=\\mathbf w\\cdot\\mathbf n_1=0\n$$\n\nis\n\n$$\n\\mathbf w=(1,-2,1)\n$$\n\nup to scale and sign. It retains $\\mathbf w\\cdot\\mathbf s=1$.\n\nA 2025 trapped-ion experiment encoded this pattern in different Zeeman sensitivities within the metastable manifold of three $^{40}\\mathrm{Ca}^+$ ions.\n\nThe entangled branch pair stayed inside a decoherence-free subspace for applied common-mode and gradient noise while sensing an effective quadratic field produced by AC-Stark shifts.\n\nThe implemented entangled protocol reduced normalized RMSE by a factor $2.6(1)$ relative to the implemented separable protocol. It also reduced RMSE by a factor $1.49(6)$ relative to the ideal comparator in the paper's restricted separable class.\n\nThese are finite, three-sensor results under a specified comparison. They do not demonstrate asymptotic Heisenberg scaling in network size. See [Bate et al.](https://arxiv.org/html/2501.08940v2).\n\n> [!margin: Why multilevel sensors matter]\n> The label $-2$ is not an eigenvalue of an ordinary Pauli $Z$. The experiment used multilevel ions with selectable magnetic sensitivities. A three-qubit derivation must engineer unequal coupling strengths instead.\n\n## 7. Where the analogy stops\n\nSpatial cancellation can be exact for arbitrarily strong fluctuating amplitudes $\\xi_a(t)$ when the spatial profiles stay fixed and all generators commute.\n\nIt fails when the nuisance field leaves the assumed profile span, sensor positions drift, local sensitivities are miscalibrated, or the noise couples through noncommuting operators.\n\nHNLS has different failure boundaries. Its necessity-and-sufficiency statement can fail outside a parameter-independent Markovian model because environmental memory carries information not represented by probe-only jump operators.\n\nA 2025 study models memory by enlarging the system to a probe plus an inaccessible hidden Markov environment. It derives generalized correction conditions and several sufficient routes to Heisenberg scaling.\n\nIt does not replace the Markovian HNLS theorem with one universal probe-only test. See [Mann et al., *PRX Quantum* 6, 030321](https://arxiv.org/abs/2503.07745).\n\nFinite correction rate creates another gap between theorem and apparatus. An error can act for a nonzero time before correction, leaving residual decoherence and shifting the inferred sensing frequency.\n\nThat shift is a systematic estimation bias, not extra statistical variance. [Rojkov et al.](https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.128.140503) show how it arises and how calibration can remove it.\n\n## 8. What the current frontier adds\n\n### Autonomous correction\n\nFast measurement and feedback are not the only route. Engineered dissipation can autonomously return a sensor to its code space.\n\nA 2026 result gives a sufficient construction that approximately restores $T^2$ scaling over a chosen finite interval at a finite ratio $R$ of correction to noise rates.\n\nThe additive error scales as $O(\\kappa T/R^c)$ for codes of order $c$ under the paper's conditions. The result narrows the gap between infinitely fast control and hardware, while retaining explicit rate and model assumptions.\n\nSee [Kwon et al., *npj Quantum Information* 12, 103 (2026)](https://www.nature.com/articles/s41534-026-01268-1).\n\n### Asymmetric codes\n\nMemory codes are rewarded for protecting every local direction. Sensors need a chosen local sum to remain visible.\n\nA 2026 preprint quantifies tradeoffs between code distance and QFI for several symmetric code families. It constructs asymmetric codes that relax protection along the signal direction while retaining growing protection in complementary directions.\n\nThis is evidence for a design principle, not yet a general experimental result: code geometry should follow signal geometry. See [Chen et al., arXiv:2512.20426v2](https://arxiv.org/abs/2512.20426).\n\n### Robustness as a model set\n\nExact orthogonality to one calibrated noise span can be brittle. A practical design should optimize over uncertainty in jump rates, spatial modes, correction latency, readout, and signal calibration.\n\nThe objective may be a worst-case classical FI after a realizable measurement, not the ideal QFI at the nominal model. This turns robust sensing into a minimax control and code-design problem.\n\nThe order of work should be explicit:\n\n$$\n\\text{identify nuisance modes}\n\\to\n\\text{find the surviving signal}\n\\to\n\\text{choose a code and state}\n\\to\n\\text{choose a measurement}\n\\to\n\\text{audit total resources and bias}.\n$$\n\n## The deliberately false claim\n\n**If the signal Hamiltonian commutes with the noise Hamiltonian, quantum error correction can always remove the noise without reducing signal sensitivity.**\n\nThe example $G\\propto L\\propto Z$ defeats it. The two operators commute perfectly and remain physically indistinguishable as generator directions.\n\nThe corrected statement is:\n\n> Commutation may simplify control. Protection preserves sensitivity only when the code makes the nuisance action trivial while leaving a nontrivial logical signal generator.\n\n## Collision: five statements with different strength\n\n1. **Ideal-state statement:** an entangled state can have $F_{\\rm Q}\\propto N^2$ under unitary encoding.\n2. **Markovian theorem:** ideal QEC restores $F_{\\rm Q}\\propto T^2$ exactly when HNLS holds under the theorem's assumptions.\n3. **Spatial statement:** a DFS cancels fixed nuisance profiles when an allowed sensitivity vector lies in their nullspace and overlaps the signal.\n4. **Experimental statement:** three multilevel ions demonstrated a protected finite-system advantage for constant and gradient noise.\n5. **Engineering statement:** finite-rate control, model error, preparation, readout, and estimator bias determine whether the advantage survives end to end.\n\nMoving from one line to the next requires new assumptions or new data.\n\n## Do this now\n\nThree multilevel sensors sit at\n\n$$\nx=(-1,0,1).\n$$\n\nTheir coupling is\n\n$$\nH=\\gamma\\sum_{i=1}^3B(x_i)S_i,\n\\qquad\nB(x)=b_0+b_1x+\\frac{b_2}{2}x^2.\n$$\n\nAssume $S_i\\ket{w_i}=w_i\\ket{w_i}$ and that the local levels needed below exist. Consider\n\n$$\n\\ket\\psi\n=\\frac{\\ket{\\mathbf w}+\\ket{-\\mathbf w}}{\\sqrt2}.\n$$\n\n### Target A\n\nFind the nonzero integer vector $\\mathbf w$ with the smallest maximum coefficient magnitude that cancels $b_0$ and $b_1$ while retaining $b_2$.\n\n### Target B\n\nCalculate the relative phase after time $t$.\n\n### Target C\n\nCalculate the pure-state QFI for estimating $b_2$. State how the answer changes for physical positions $(-d,0,d)$.\n\n### Target D\n\nSuppose an unmodelled quadratic nuisance is added. Can any state cancel it while remaining sensitive to $b_2$ through the same generator?\n\n### Oral check 1\n\nWhy does entanglement fail to help when the signal profile belongs to the nuisance-profile span?\n\n### Oral check 2\n\nWhat information does HNLS contain that $[G,L_k]$ does not?\n\n\nHints
\n\n- Impose $\\mathbf w\\cdot(1,1,1)=0$ and $\\mathbf w\\cdot(-1,0,1)=0$.\n- For $e^{-\\ii b_2K}\\ket\\psi$, use $F_{\\rm Q}=4\\operatorname{Var}_\\psi K$.\n- Two unknown coefficients multiplying the same operator direction cannot be identified from one accumulated phase.\n\n \n\n\nSolution outline
\n\nThe linear-noise constraint gives\n\n$$\n-w_1+w_3=0,\n$$\n\nso $w_3=w_1$. The common-mode constraint gives\n\n$$\nw_1+w_2+w_3=0,\n$$\n\nhence $w_2=-2w_1$. The primitive integer vector is\n\n$$\n\\boxed{\\mathbf w=(1,-2,1)}.\n$$\n\nIts quadratic overlap is\n\n$$\n\\mathbf w\\cdot\n\\left(\\frac12,0,\\frac12\\right)=1.\n$$\n\nThe branch energies are\n\n$$\nE_+=\\gamma b_2,\n\\qquad\nE_-=-\\gamma b_2.\n$$\n\nThe accumulated relative phase has magnitude\n\n$$\n\\boxed{\\Delta\\phi=2\\gamma b_2t}.\n$$\n\nIts sign depends on which branch is used as the phase reference.\n\nWrite\n\n$$\nK=\\frac{\\gamma t}{2}\n\\sum_i x_i^2S_i.\n$$\n\nThe two branches have $K$ eigenvalues $+\\gamma t$ and $-\\gamma t$. Therefore\n\n$$\n\\operatorname{Var}_\\psi K=(\\gamma t)^2\n$$\n\nand\n\n$$\n\\boxed{F_{\\rm Q}(b_2)=4\\gamma^2t^2}.\n$$\n\nFor $x=(-d,0,d)$, the eigenvalues are $\\pm\\gamma td^2$, so\n\n$$\nF_{\\rm Q}(b_2)=4\\gamma^2t^2d^4.\n$$\n\nIf an unknown nuisance multiplies the same quadratic profile and the same $S_i$, it is statistically indistinguishable from $b_2$. Cancelling that direction also cancels the target.\n\nOne must add another channel that changes their response, such as time modulation, a second species, a different transition, or independent prior information.\n\n \n\n## Exit ticket\n\n$G$ has a large norm but lies inside the Lindblad span. A weaker $G'$ has a nonzero component outside it.\n\nWhich generator is the candidate for error-corrected $T^2$ scaling, and what must still be calculated before claiming a practical advantage?\n\nA strong answer should mention the surviving logical spectral gap, control rate, total-cycle overhead, attainable measurement FI, and robustness to errors in the noise model.\n\n## Research checks worth doing next\n\n1. **Noise-span tomography.** Infer operator or spatial nuisance modes with uncertainty bars instead of assuming a diagonal noise model.\n2. **Projection stability.** Track how $G_\\perp$ changes under calibration drift and omitted weak channels.\n3. **Finite-cycle simulation.** Include correction latency, faulty ancillas, leakage, and control noise.\n4. **Measurement audit.** Compare QFI with classical FI for the readout that the platform can implement.\n5. **Bias audit.** Simulate the estimator mean as well as its variance after finite-rate correction.\n6. **Non-Markovian test.** Vary cycle time and look for memory-dependent performance that a time-local model cannot fit.\n7. **Network scaling.** Increase sensor number while holding state-preparation fidelity, nuisance rank, and total experimental time under explicit control.\n\n## Further reading\n\n- [Achieving the Heisenberg limit in quantum metrology using quantum error correction](https://www.nature.com/articles/s41467-017-02510-3) presents the HNLS theorem and code optimization.\n- [Quantum Sensing with Atomic, Molecular, and Optical Platforms for Fundamental Physics](https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.132.190001) frames the AMO program of entanglement, networks, and logical sensing.\n- [Experimental distributed quantum sensing in a noisy environment](https://arxiv.org/html/2501.08940v2) reports the three-ion spatially protected protocol and its comparison class.\n- [Practical limits of error correction for quantum metrology](https://iopscience.iop.org/article/10.1088/1367-2630/abf533) analyzes finite repetition rates and realistic resource limits.\n- [Bias in error-corrected quantum sensing](https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.128.140503) separates residual noise from systematic frequency bias.\n- [Quantum Error Corrected Non-Markovian Metrology](https://arxiv.org/abs/2503.07745) gives generalized conditions in a hidden Markov environment model.\n- [Restoring Heisenberg scaling in time via autonomous quantum error correction](https://www.nature.com/articles/s41534-026-01268-1) studies finite-rate autonomous protection.\n- [Bypassing the protection-sensitivity incompatibility via asymmetric codes](https://arxiv.org/abs/2512.20426) is a 2026 preprint on scalable, direction-dependent protection.\n\n## What to retain\n\n- QFI is an optimized information bound; the implemented measurement supplies classical FI.\n- Entanglement amplifies a preserved generator. It does not identify which generator survives noise.\n- HNLS uses the full Lindblad span, including $L_k^\\dagger L_j$ products.\n- Correctable noise must act trivially in the code, while the signal must act nontrivially.\n- Spatial DFS sensing is a nullspace problem with hardware constraints on allowed sensitivities.\n- The three-ion $(1,-2,1)$ state relies on multilevel atomic structure.\n- HNLS is exact inside an ideal Markovian control model.\n- Finite-rate correction can leave decoherence and estimator bias.\n- Non-Markovian memory requires an enlarged dynamical description.\n- Asymmetric protection is a promising design direction, currently supported by theory rather than a general experiment.\n\nNext: derive HNLS from the short-time Kraus expansion and show why the products $L_k^\\dagger L_j$ enter the channel tangent space.\n",
"content_sha256": "9cddc30601ae8f1e0ecde7d424f58523f0784678dcad0ef5649316f09ed5a7e8",
"formula_count": 115,
"formulas": [
{
"index": 1,
"display": true,
"tex": "\\boxed{\n\\text{Which component of the signal generator is distinguishable from the noise algebra?}\n}",
"line": 30,
"sha256": "f63441ebccf593da2670c68c1bfc6edd1b2d90e0a41981fbd392c138cada8a16"
},
{
"index": 2,
"display": false,
"tex": "\\theta",
"line": 40,
"sha256": "2b4af44cb17349bb9452292a447fe70d48c9ce08e68dd2f74e9628cf8ac625b8"
},
{
"index": 3,
"display": false,
"tex": "\\rho_\\theta",
"line": 40,
"sha256": "5673bc1347db3a1b8001f33911a937fdc6b86898a0211f2426324b5644451d9f"
},
{
"index": 4,
"display": false,
"tex": "x",
"line": 40,
"sha256": "2d711642b726b04401627ca9fbac32f5c8530fb1903cc4db02258717921a4881"
},
{
"index": 5,
"display": false,
"tex": "p(x\\mid\\theta)",
"line": 40,
"sha256": "10aa8c171857bd62163607cf3cd015ee54bf45b1d7d52c422b4d556859ff9789"
},
{
"index": 6,
"display": true,
"tex": "F_{\\rm C}(\\theta)\n=\\sum_x\\frac{[\\partial_\\theta p(x\\mid\\theta)]^2}{p(x\\mid\\theta)}.",
"line": 42,
"sha256": "a57bd16704ad1912e2e82e4ba60f6cade908e000098d3b0bf03ffd0fa26a1f0a"
},
{
"index": 7,
"display": false,
"tex": "\\nu",
"line": 47,
"sha256": "84eb2b123d1102b8f7c00b34116df32107cffa82d46efbac7fd9543fa7459c38"
},
{
"index": 8,
"display": true,
"tex": "\\operatorname{Var}(\\hat\\theta)\n\\geq \\frac{1}{\\nu F_{\\rm C}}\n\\geq \\frac{1}{\\nu F_{\\rm Q}}.",
"line": 49,
"sha256": "0086e7e74ce40d3be0340ce5f7ca92972a2b50b6bcb79cd11364c97b1b8231b5"
},
{
"index": 9,
"display": false,
"tex": "F_{\\rm Q}",
"line": 55,
"sha256": "0c25c03959b11dce031c4cfa31d8587f65648e3770e775f61749291f1fc6a7ab"
},
{
"index": 10,
"display": false,
"tex": "\\ket{\\psi_\\theta}=e^{-\\ii\\theta tG}\\ket\\psi",
"line": 57,
"sha256": "31b3ecf1b38cc63dff018a4b318993a4410ac8b75aa36c7433888acdc98917f6"
},
{
"index": 11,
"display": true,
"tex": "F_{\\rm Q}=4t^2\\operatorname{Var}_\\psi(G).",
"line": 59,
"sha256": "00d0e6151f2b238b47e71413330400d6a3ee38271ac909a57ec33a8eae4a12db"
},
{
"index": 12,
"display": false,
"tex": "\\operatorname{Var}(G)",
"line": 63,
"sha256": "da7fa084d4b8917af67e0ae26b731d53d2837b4e744922f85f5059d16161fe40"
},
{
"index": 13,
"display": false,
"tex": "N^2",
"line": 63,
"sha256": "20cba99e4019fcab21799c2071d97944ee3ceb1568b0cf59a15435049003203e"
},
{
"index": 14,
"display": true,
"tex": "F_{\\rm Q}\\sim N^2\n\\quad\\text{in an ideal model}\n\\not\\Rightarrow\n\\text{a robust }N^{-1}\\text{ experimental uncertainty}.",
"line": 67,
"sha256": "2c08fc1e667311ccaeff1fbe57dbd59474905c1af6f0019d4848c2d063f8c8d9"
},
{
"index": 15,
"display": true,
"tex": "\\dot\\rho\n=-\\ii[\\theta G,\\rho]\n+\\sum_k\\left(\nL_k\\rho L_k^\\dagger\n-\\frac12\\{L_k^\\dagger L_k,\\rho\\}\n\\right).",
"line": 80,
"sha256": "ce392201c8e712ee8482943131d50355fc21c5b0420f3269cf906dca1c3d47e4"
},
{
"index": 16,
"display": true,
"tex": "\\begin{aligned}\n\\mathcal S=\\operatorname{span}_{\\mathbb R}\\{&I,\nL_k+L_k^\\dagger,\n\\ii(L_k-L_k^\\dagger),\\\\\n&L_k^\\dagger L_j+L_j^\\dagger L_k,\n\\ii(L_k^\\dagger L_j-L_j^\\dagger L_k)\\}_{j,k}.\n\\end{aligned}",
"line": 91,
"sha256": "43f6846ddf83eb5e44af802bdc73075d3e5f05dbee046bb9f35fbdce0837fe8c"
},
{
"index": 17,
"display": false,
"tex": "L_k^\\dagger L_j",
"line": 101,
"sha256": "d9236c26dbb604df38ed22d86ff953d2b6dbab4a3048229e18b8e327fcfb537d"
},
{
"index": 18,
"display": true,
"tex": "\\boxed{G\\notin\\mathcal S}",
"line": 105,
"sha256": "ab05815f590a2c413b163daa3d063abc0366136a3f03d1a7f9fa4aa1490118a3"
},
{
"index": 19,
"display": true,
"tex": "F_{\\rm Q}(T)=O(T^2)",
"line": 111,
"sha256": "67f76b0f521ed56ddce6b9b12f437e49437ae7d3cd1f27a2c1b82088c864fc0d"
},
{
"index": 20,
"display": false,
"tex": "O(T)",
"line": 115,
"sha256": "7088196ac8eeb47ffbaf05c8468099a504cb2a0fb634ca34a3fab9272a20d84d"
},
{
"index": 21,
"display": false,
"tex": "P",
"line": 126,
"sha256": "5c62e091b8c0565f1bafad0dad5934276143ae2ccef7a5381e8ada5b1a8d26d2"
},
{
"index": 22,
"display": true,
"tex": "PL_kP=\\lambda_kP,\n\\qquad\nPL_k^\\dagger L_jP=\\mu_{kj}P.",
"line": 128,
"sha256": "b20541ff69d70c5264066ddd6b50f8f0bade54446b3ae629eea5cc730602e0f5"
},
{
"index": 23,
"display": false,
"tex": "G",
"line": 136,
"sha256": "333e0a1e27815d0ceee55c473fe3dc93d56c63e3bee2b3b4aee8eed6d70191a3"
},
{
"index": 24,
"display": true,
"tex": "PGP\\neq gP.",
"line": 138,
"sha256": "4612a39ec28bf87d3fe74660d0a6632001864b92c862dc5de3c6f2d735e6dba5"
},
{
"index": 25,
"display": false,
"tex": "G\\in\\mathcal S",
"line": 142,
"sha256": "3d14af93d1f882c9f8d6962a5989199216549f47c96ba8925d08600c7a4c7b63"
},
{
"index": 26,
"display": false,
"tex": "PGP",
"line": 142,
"sha256": "6cc6036cfafd30f58f0d4caf3bfee97303cb9f3241aa9c3493367c2845bac5c2"
},
{
"index": 27,
"display": true,
"tex": "G=\\frac12Z,\n\\qquad\nL=\\sqrt\\gamma Z.",
"line": 148,
"sha256": "b99857f2580d3e46ebf92ce3b68d0d934b24ef2e35eb3528c8f9ef1b638acec4"
},
{
"index": 28,
"display": false,
"tex": "Z",
"line": 154,
"sha256": "bbeebd879e1dff6918546dc0c179fdde505f2a21591c9a9c96e36b054ec5af83"
},
{
"index": 29,
"display": false,
"tex": "[G,L]\\neq0",
"line": 156,
"sha256": "58d4003ef16e0a40b1b05bb5128ad0c9589d81a6031d6c78fb049a3d37e5f3ef"
},
{
"index": 30,
"display": false,
"tex": "G",
"line": 156,
"sha256": "333e0a1e27815d0ceee55c473fe3dc93d56c63e3bee2b3b4aee8eed6d70191a3"
},
{
"index": 31,
"display": false,
"tex": "\\mathcal S",
"line": 156,
"sha256": "dc5c5f4f378f97aa75a34c1738527bcbf978f83318fbd0b263e41e37be306cb9"
},
{
"index": 32,
"display": false,
"tex": "L_k^\\dagger L_j",
"line": 156,
"sha256": "d9236c26dbb604df38ed22d86ff953d2b6dbab4a3048229e18b8e327fcfb537d"
},
{
"index": 33,
"display": false,
"tex": "L_k",
"line": 156,
"sha256": "dfc293305a0744ec99aee29ab95600c388a3aaa8605435547265ad0dd4de4239"
},
{
"index": 34,
"display": true,
"tex": "G=G_{\\parallel}+G_\\perp,\n\\qquad\nG_{\\parallel}\\in\\mathcal S,\n\\qquad\nG_\\perp\\perp\\mathcal S.",
"line": 164,
"sha256": "7899069a580715cd0e034099730f2d3ebf5351dad20cc4727ffb5ff974dfa0df"
},
{
"index": 35,
"display": false,
"tex": "G_\\perp\\neq0",
"line": 172,
"sha256": "dd7070ceacdf3845fe7e963dc464b4040d273c129293edb8d1369a6c45a0f94b"
},
{
"index": 36,
"display": false,
"tex": "G_\\perp",
"line": 174,
"sha256": "9ee301a3921003108e83dd3037578409416e5d475d9dd160122f3614448d0aae"
},
{
"index": 37,
"display": false,
"tex": "G_\\perp",
"line": 181,
"sha256": "9ee301a3921003108e83dd3037578409416e5d475d9dd160122f3614448d0aae"
},
{
"index": 38,
"display": false,
"tex": "N",
"line": 189,
"sha256": "8ce86a6ae65d3692e7305e2c58ac62eebd97d3d943e093f577da25c36988246b"
},
{
"index": 39,
"display": false,
"tex": "i",
"line": 189,
"sha256": "de7d1b721a1e0632b7cf04edf5032c8ecffa9f9a08492152b926f1a5a7e765d7"
},
{
"index": 40,
"display": false,
"tex": "S_i",
"line": 189,
"sha256": "698cd4222a287d75136e89e22b63e6d542ecd847bb8b5bcb00babf5305a71e27"
},
{
"index": 41,
"display": false,
"tex": "\\mathbf s",
"line": 189,
"sha256": "5d12ad1226b6955d6a1dbd88e1598d3667a15e33c6bed3a2595eec5d41538749"
},
{
"index": 42,
"display": false,
"tex": "\\mathbf n_a",
"line": 189,
"sha256": "b9ac8c8bdf022c14ed226545d012cb0ff8b9e08929a88dc122461b3852d6b530"
},
{
"index": 43,
"display": true,
"tex": "H(t)=\\gamma\\sum_{i=1}^N\n\\left[\n\\theta s_i+\\sum_{a=1}^r\\xi_a(t)n_{ai}\n\\right]S_i.",
"line": 191,
"sha256": "bb6386476041025dfaf4c27e58b850df758756290f3f4f8cbca07f99eb209aa9"
},
{
"index": 44,
"display": true,
"tex": "\\mathbf w=(w_1,\\ldots,w_N),",
"line": 200,
"sha256": "7b51f571f7585fd24b073766f6a0a3124b31e0411495e0f3db5d20393090e87d"
},
{
"index": 45,
"display": true,
"tex": "\\ket{\\psi_{\\mathbf w}}\n=\\frac{\\ket{\\mathbf w}+\\ket{-\\mathbf w}}{\\sqrt2}.",
"line": 206,
"sha256": "b61d598a3f59885529450e117199602fe748c83308c20a133fa95619b22f03f0"
},
{
"index": 46,
"display": true,
"tex": "\\mathbf w\\cdot\\mathbf n_a=0\n\\quad\\text{for every }a.",
"line": 213,
"sha256": "b18d32b08a4070c147daaafe957b14da6a32a38ad5774159fc4822c55d73c07c"
},
{
"index": 47,
"display": true,
"tex": "\\mathbf w\\cdot\\mathbf s\\neq0.",
"line": 220,
"sha256": "53e4e0cbcaa9ed08e6ea3c3a511abd42e9998c426168f1acac2ed2f1cc85c4ba"
},
{
"index": 48,
"display": false,
"tex": "N_{\\rm noise}",
"line": 224,
"sha256": "344cef4330364b6419d8751f1feabe3ff4598e3dd04f46d783860e7d6c7b12f3"
},
{
"index": 49,
"display": true,
"tex": "\\mathbf s_\\perp\n=\\left[I-N_{\\rm noise}\n(N_{\\rm noise}^{\\mathsf T}N_{\\rm noise})^+\nN_{\\rm noise}^{\\mathsf T}\\right]\\mathbf s,",
"line": 226,
"sha256": "82fa62275fb9e61e586b58e0329c26726bb7ff64d8a9d4d2eadb5e56b89e9969"
},
{
"index": 50,
"display": false,
"tex": "+",
"line": 233,
"sha256": "a318c24216defe206feeb73ef5be00033fa9c4a74d0b967f6532a26ca5906d3b"
},
{
"index": 51,
"display": false,
"tex": "\\mathbf w",
"line": 235,
"sha256": "7c7d2c840b19811fdcd00ea34f98eeae1678b591b5b3707f90d197d6fcd480b3"
},
{
"index": 52,
"display": false,
"tex": "L_k^\\dagger L_j",
"line": 237,
"sha256": "d9236c26dbb604df38ed22d86ff953d2b6dbab4a3048229e18b8e327fcfb537d"
},
{
"index": 53,
"display": true,
"tex": "\\mathbf x=(-1,0,1)",
"line": 243,
"sha256": "c41e579c9006741813883b20fd57da00e4100ba101acf8935e15165d3b42e81a"
},
{
"index": 54,
"display": true,
"tex": "B(x)=b_0+b_1x+\\frac{b_2}{2}x^2.",
"line": 249,
"sha256": "9018d78b95b30dea9a2f7cac5be22e8f5a89310385fc7ed4ece7a5664942645b"
},
{
"index": 55,
"display": true,
"tex": "\\mathbf n_0=(1,1,1),\n\\quad\n\\mathbf n_1=(-1,0,1),\n\\quad\n\\mathbf s=\\left(\\frac12,0,\\frac12\\right).",
"line": 255,
"sha256": "6fe9d484abcfa6e786aef822c649641b20550ed0bf1cbfefcaf91eb41850231b"
},
{
"index": 56,
"display": true,
"tex": "\\mathbf w\\cdot\\mathbf n_0\n=\\mathbf w\\cdot\\mathbf n_1=0",
"line": 265,
"sha256": "28185fe711c658f2ed611d730f0e0766c9b06680b67415914abcc9569856dd64"
},
{
"index": 57,
"display": true,
"tex": "\\mathbf w=(1,-2,1)",
"line": 272,
"sha256": "f7b9b18400c565e975ea1dfa866c789e45babf093b461ede92b1dbc36fc85fb1"
},
{
"index": 58,
"display": false,
"tex": "\\mathbf w\\cdot\\mathbf s=1",
"line": 276,
"sha256": "6e1aa57fed6a75890f7291549d2f0fe115daa7da5d7fd8b106c29caebc7a842b"
},
{
"index": 59,
"display": false,
"tex": "^{40}\\mathrm{Ca}^+",
"line": 278,
"sha256": "1cc240034381489a93059cc43431804c40fd0f1b49b014f46bab14fd8c926fe7"
},
{
"index": 60,
"display": false,
"tex": "2.6(1)",
"line": 282,
"sha256": "4e4965c48316b6fd784be59d7f33fbfdc216ddf2f5e38875b0fb34159e0fcadc"
},
{
"index": 61,
"display": false,
"tex": "1.49(6)",
"line": 282,
"sha256": "29777de1fba47b2e0cc46dd5a13bcdf767f0133ff0ac8bd006c0054b05a2c96d"
},
{
"index": 62,
"display": false,
"tex": "-2",
"line": 287,
"sha256": "cf3bae39dd692048a8bf961182e6a34dfd323eeb0748e162eaf055107f1cb873"
},
{
"index": 63,
"display": false,
"tex": "Z",
"line": 287,
"sha256": "bbeebd879e1dff6918546dc0c179fdde505f2a21591c9a9c96e36b054ec5af83"
},
{
"index": 64,
"display": false,
"tex": "\\xi_a(t)",
"line": 291,
"sha256": "18d74b0a0fa2a665e1a8343dad6411c696369f939849a35c54b10c8c8a974754"
},
{
"index": 65,
"display": false,
"tex": "T^2",
"line": 311,
"sha256": "c0614079049a77737e98768544e9b07a028500457fc706d1fd13d05e3e23e9ec"
},
{
"index": 66,
"display": false,
"tex": "R",
"line": 311,
"sha256": "8c2574892063f995fdf756bce07f46c1a5193e54cd52837ed91e32008ccf41ac"
},
{
"index": 67,
"display": false,
"tex": "O(\\kappa T/R^c)",
"line": 313,
"sha256": "71525039b9b8b1de18fa2e92e2e6bf3ab780af89fae8010765dfe78e207462bd"
},
{
"index": 68,
"display": false,
"tex": "c",
"line": 313,
"sha256": "2e7d2c03a9507ae265ecf5b5356885a53393a2029d241394997265a1a25aefc6"
},
{
"index": 69,
"display": true,
"tex": "\\text{identify nuisance modes}\n\\to\n\\text{find the surviving signal}\n\\to\n\\text{choose a code and state}\n\\to\n\\text{choose a measurement}\n\\to\n\\text{audit total resources and bias}.",
"line": 333,
"sha256": "4333fe4f0acda3837c76869b3dc45a366d8119e75a707fa3c295bacee0335a6d"
},
{
"index": 70,
"display": false,
"tex": "G\\propto L\\propto Z",
"line": 349,
"sha256": "062ae66af03ba06a5520890f94f6b24b8453ab9c52217dc7dd0e0953dec8054a"
},
{
"index": 71,
"display": false,
"tex": "F_{\\rm Q}\\propto N^2",
"line": 357,
"sha256": "4af1e2459b3837c8ca2ca1756e24c832ee721927ab6bcce2e32b37fd8a11b548"
},
{
"index": 72,
"display": false,
"tex": "F_{\\rm Q}\\propto T^2",
"line": 358,
"sha256": "78ab9c83aa0f6457e1df80e9870fac20f10d2f8c8431ba910ca943b07d681e3a"
},
{
"index": 73,
"display": true,
"tex": "x=(-1,0,1).",
"line": 369,
"sha256": "444f224184ccc92efb8d5f03c60ea915cb43848326c3f0d0a048b271e696b444"
},
{
"index": 74,
"display": true,
"tex": "H=\\gamma\\sum_{i=1}^3B(x_i)S_i,\n\\qquad\nB(x)=b_0+b_1x+\\frac{b_2}{2}x^2.",
"line": 375,
"sha256": "8f10271007a60d8ab4bd6aa06939bfbe893c2e80e5c3759cf5356e126fc4d371"
},
{
"index": 75,
"display": false,
"tex": "S_i\\ket{w_i}=w_i\\ket{w_i}",
"line": 381,
"sha256": "aa8b5f10ee9a108777c419a281eaf7de41995044e1c5e22e3d59b7b82f328511"
},
{
"index": 76,
"display": true,
"tex": "\\ket\\psi\n=\\frac{\\ket{\\mathbf w}+\\ket{-\\mathbf w}}{\\sqrt2}.",
"line": 383,
"sha256": "d5d90493c2b8ee9913967743c59d2613c4b7dd838074204afa1217a6ee939bad"
},
{
"index": 77,
"display": false,
"tex": "\\mathbf w",
"line": 390,
"sha256": "7c7d2c840b19811fdcd00ea34f98eeae1678b591b5b3707f90d197d6fcd480b3"
},
{
"index": 78,
"display": false,
"tex": "b_0",
"line": 390,
"sha256": "c73740af47bea4cff696936de19f88b836576203c45f90a224130e72a5804ee9"
},
{
"index": 79,
"display": false,
"tex": "b_1",
"line": 390,
"sha256": "74bf6daef0c8b4f8cdba0f626389263961ee5fed9fa3af83a2afde69f2aa9f5e"
},
{
"index": 80,
"display": false,
"tex": "b_2",
"line": 390,
"sha256": "bfdaa023faec0e4ec7f3169a5d2a1dfacd662d171267b69c36ba6c7b139f2d85"
},
{
"index": 81,
"display": false,
"tex": "t",
"line": 394,
"sha256": "e3b98a4da31a127d4bde6e43033f66ba274cab0eb7eb1c70ec41402bf6273dd8"
},
{
"index": 82,
"display": false,
"tex": "b_2",
"line": 398,
"sha256": "bfdaa023faec0e4ec7f3169a5d2a1dfacd662d171267b69c36ba6c7b139f2d85"
},
{
"index": 83,
"display": false,
"tex": "(-d,0,d)",
"line": 398,
"sha256": "e762ca78aa6ddcce808589654de02dd7dc85b8a138cf37479bd20e3d26179b9a"
},
{
"index": 84,
"display": false,
"tex": "b_2",
"line": 402,
"sha256": "bfdaa023faec0e4ec7f3169a5d2a1dfacd662d171267b69c36ba6c7b139f2d85"
},
{
"index": 85,
"display": false,
"tex": "[G,L_k]",
"line": 410,
"sha256": "1749a26db5b8e82ab378e77051161a3d9fe52dcf087fc59555071ec5b774f9f6"
},
{
"index": 86,
"display": false,
"tex": "\\mathbf w\\cdot(1,1,1)=0",
"line": 415,
"sha256": "8114424462b79fba9957eefe75c6fb3f18fa14b705302796737a4f5d3454bc3a"
},
{
"index": 87,
"display": false,
"tex": "\\mathbf w\\cdot(-1,0,1)=0",
"line": 415,
"sha256": "84068441f003c7f3b56078d7b787595b6603b6f5278ca922a701a83966377543"
},
{
"index": 88,
"display": false,
"tex": "e^{-\\ii b_2K}\\ket\\psi",
"line": 416,
"sha256": "cbfaf6abec65a70c2d1d5ce95014287f02432fd75fb72bbf8eb311ce44150d5a"
},
{
"index": 89,
"display": false,
"tex": "F_{\\rm Q}=4\\operatorname{Var}_\\psi K",
"line": 416,
"sha256": "a4df434698edda4d257765e63b686b9869c31de0836ec090ee7cb6d10be19b25"
},
{
"index": 90,
"display": true,
"tex": "-w_1+w_3=0,",
"line": 426,
"sha256": "2c4834e469e355e10e9a7bac882aa2ee693625ad00936df53d5820e4f47d7561"
},
{
"index": 91,
"display": false,
"tex": "w_3=w_1",
"line": 430,
"sha256": "c3066e651169b17c78272af56bc0f9781c60c73b6223d9053479e425e8a184de"
},
{
"index": 92,
"display": true,
"tex": "w_1+w_2+w_3=0,",
"line": 432,
"sha256": "2c9efa69f6ef0cbd31e11dbb8f02c64a353f18a687ee469d5da01d7ccbf2ef32"
},
{
"index": 93,
"display": false,
"tex": "w_2=-2w_1",
"line": 436,
"sha256": "c91dfd18f3ed57666bd28c9ad8e0d65c15648625bffaf5d5f98d4c15e240e4a6"
},
{
"index": 94,
"display": true,
"tex": "\\boxed{\\mathbf w=(1,-2,1)}.",
"line": 438,
"sha256": "ae197fcf4338258c8e58a57eb4fb5bacae2f781a7d73810d87fae08bcc46276a"
},
{
"index": 95,
"display": true,
"tex": "\\mathbf w\\cdot\n\\left(\\frac12,0,\\frac12\\right)=1.",
"line": 444,
"sha256": "bbb42683c37483e9c1012e9b8224a698013eaa6c0405567b3800304df23eaa01"
},
{
"index": 96,
"display": true,
"tex": "E_+=\\gamma b_2,\n\\qquad\nE_-=-\\gamma b_2.",
"line": 451,
"sha256": "0ec456f1d3678203f16bdbb5fb5905f2615fccc8ab9c55cec5e92963d4c50a5e"
},
{
"index": 97,
"display": true,
"tex": "\\boxed{\\Delta\\phi=2\\gamma b_2t}.",
"line": 459,
"sha256": "cc82fed9b06473e401831ac74cbf56efb7201639a2d54a3745cafa1f5d6ded5a"
},
{
"index": 98,
"display": true,
"tex": "K=\\frac{\\gamma t}{2}\n\\sum_i x_i^2S_i.",
"line": 467,
"sha256": "1253d32eb1709ab39c39fd9d30e2b81dd30b71c6ef4b0fc68dc6e0f4d4aee8f6"
},
{
"index": 99,
"display": false,
"tex": "K",
"line": 472,
"sha256": "86be9a55762d316a3026c2836d044f5fc76e34da10e1b45feee5f18be7edb177"
},
{
"index": 100,
"display": false,
"tex": "+\\gamma t",
"line": 472,
"sha256": "8d0ddc9258c301b6ae4fa39be4af0c54a8d5b9b1ee46456ca6e4cd3e0d352962"
},
{
"index": 101,
"display": false,
"tex": "-\\gamma t",
"line": 472,
"sha256": "e0e44b5d993f8abe0cc9dbfcd86b6d85f97e676dd2474b0005cec5888b441203"
},
{
"index": 102,
"display": true,
"tex": "\\operatorname{Var}_\\psi K=(\\gamma t)^2",
"line": 474,
"sha256": "f70e921774932ba28eb7262b5c3a2d1e7785def0f18135021a94c8bd6f1f450d"
},
{
"index": 103,
"display": true,
"tex": "\\boxed{F_{\\rm Q}(b_2)=4\\gamma^2t^2}.",
"line": 480,
"sha256": "509c8fc7a47b4b4c9a746f1d3a248ced2cce86acf86b2d5fbc29ff852a976983"
},
{
"index": 104,
"display": false,
"tex": "x=(-d,0,d)",
"line": 484,
"sha256": "1d9847d6a66d4776134d6e90459ca482dbf504e044274ca7f47062afbd4e0830"
},
{
"index": 105,
"display": false,
"tex": "\\pm\\gamma td^2",
"line": 484,
"sha256": "77417c7b81b7e6eef15beee900440fb4d557c489817c165460e2cc724b906a75"
},
{
"index": 106,
"display": true,
"tex": "F_{\\rm Q}(b_2)=4\\gamma^2t^2d^4.",
"line": 486,
"sha256": "3230498640d90208be3ef0c28e7e8d2cef9ced61b6c0635c7f1e88731e3d478e"
},
{
"index": 107,
"display": false,
"tex": "S_i",
"line": 490,
"sha256": "698cd4222a287d75136e89e22b63e6d542ecd847bb8b5bcb00babf5305a71e27"
},
{
"index": 108,
"display": false,
"tex": "b_2",
"line": 490,
"sha256": "bfdaa023faec0e4ec7f3169a5d2a1dfacd662d171267b69c36ba6c7b139f2d85"
},
{
"index": 109,
"display": false,
"tex": "G",
"line": 498,
"sha256": "333e0a1e27815d0ceee55c473fe3dc93d56c63e3bee2b3b4aee8eed6d70191a3"
},
{
"index": 110,
"display": false,
"tex": "G'",
"line": 498,
"sha256": "5e1a1a36d46632a4ba5bafa01b9b31d16528e3fe1ca22ed8c1f74209c19332ec"
},
{
"index": 111,
"display": false,
"tex": "T^2",
"line": 500,
"sha256": "c0614079049a77737e98768544e9b07a028500457fc706d1fd13d05e3e23e9ec"
},
{
"index": 112,
"display": false,
"tex": "G_\\perp",
"line": 507,
"sha256": "9ee301a3921003108e83dd3037578409416e5d475d9dd160122f3614448d0aae"
},
{
"index": 113,
"display": false,
"tex": "L_k^\\dagger L_j",
"line": 529,
"sha256": "d9236c26dbb604df38ed22d86ff953d2b6dbab4a3048229e18b8e327fcfb537d"
},
{
"index": 114,
"display": false,
"tex": "(1,-2,1)",
"line": 532,
"sha256": "fd08a0717a97a0aad6f3fcdd7f1d06692c8e99b5d09ac648a4abc884d99e6548"
},
{
"index": 115,
"display": false,
"tex": "L_k^\\dagger L_j",
"line": 538,
"sha256": "d9236c26dbb604df38ed22d86ff953d2b6dbab4a3048229e18b8e327fcfb537d"
}
]
}