betaFeedback

WorksPhysics and Astronomy

Probing new bosons and nuclear structure with ytterbium isotope shifts

Abstract

In this Letter, we present mass-ratio measurements on highly charged Yb42+^{42+} ions with a precision of 4×10−124\times 10^{-12} and isotope-shift measurements on Yb+^{+} on the 2^{2}S1/2_{1/2} →\to 2^{2}D5/2_{5/2} and 2^{2}S1/2_{1/2} →\to 2^{2}F7/2_{7/2} transitions with a precision of 4×10−94\times 10^{-9} for the isotopes 168,170,172,174,176^{168,170,172,174,176}Yb. We present a new method that allows us to extract higher-order changes in the nuclear charge distribution along the Yb isotope chain, benchmarking ab-initio nuclear structure calculations. Additionally, we perform a King plot analysis to set bounds on a fifth force in the keV/c2/c^2 to MeV/c2/c^2 range coupling to electrons and neutrons.

Theories beyond the Standard Model (SM) of particle physics are typically probed by high-energy colliders or astrophysical and cosmological observations. Competitive complementary tests can be performed with high-precision atomic and molecular physics experiments at low energies [1]. In particular, isotope-shift spectroscopy, commonly used to study nuclear charge radii in exotic isotopes [2], is also sensitive to shifts in atomic energy levels induced by hypothetical new bosons that mediate an additional interaction between neutrons and electrons [3, 4]. Such measurements can be analyzed via the King-plot method, where different atomic transitions are combined in such a way that common nuclear and atomic uncertainties are eliminated. Deviations from the linearity of the King plot indicate effects from new physics or higher-order atomic and nuclear structure. This powerful technique has been successfully used to put bounds on physics beyond the SM for example with isotope shifts measured in ytterbium [5, 6, 7]. With increasing precision of the frequency measurements, the uncertainties of the nuclear masses [5, 8] become a limiting factor for distinguishing between higher-order SM effects and new physics.

In this Letter, we present high-precision mass-ratio and isotope-shift measurements of five stable, spinless ytterbium isotopes. Both the mass spectrometry and the isotope-shift spectroscopy are up to two orders of magnitude more precise than previous measurements [7, 8, 9]. The isotope mass ratios are determined using highly charged Yb ions in the Penning-trap mass spectrometer PENTATRAP [10], reaching a relative precision of a few 10−1210^{-12} contributing to a relative uncertainty of 10−1010^{-10} for the mass-normalized isotope shifts. The isotope-shift spectroscopy is performed on Yb+\mathrm{Yb}^{+} on the 2S1/2→2D5/2{}^{2}\mathrm{S}_{1/2} \to{}^{2}\mathrm{D}_{5/2} and the 2S1/2→2F7/2{}^{2}\mathrm{S}_{1/2} \to{}^{2}\mathrm{F}_{7/2} transitions with a relative precision as low as 10−910^{-9}. Our results deviate significantly from some former mass-ratio and isotope-shift measurements. Using our improved data, we construct a generalized King plot [11] and extract a competitive spectroscopic exclusion bound on the coupling strength of potential new bosons to electrons and neutrons.

Combining these precise measurements with atomic structure calculations allows us to investigate higher-order nuclear structure effects in Yb isotopes [6] and extract changes in the quartic charge radius δ⟨r4⟩\delta\langle r^4\rangle along the isotopic chain, providing a new window into nuclear deformation. Building on advances in ab initio nuclear structure calculations [12, 13, 14, 15], we provide a first microscopic description of Yb nuclei starting from chiral effective field theory interactions [16] based on quantum chromodynamics. This method can provide direct insights into the evolution of nuclear charge distributions along isotopic chains towards exotic, neutron-rich nuclei.

Theoretical framework.– An isotope shift is the difference in the frequencies of a given atomic transition in two different isotopes of the same element. Here, we consider the 2S1/2→2D5/2{}^{2}S_{1/2}\rightarrow{}^{2}D_{5/2} electric quadrupole and the highly forbidden 2S1/2→2F7/2{}^{2}S_{1/2}\rightarrow{}^{2}F_{7/2} electric octupole transitions, denoted as α\alpha and γ\gamma, in singly charged Yb+^{+} ions. We consider five stable, even Yb isotopes with mass numbers A∈{168,170,172,174,176}A \in\{168,170,172,174,176\} containing four neighboring isotope pairs (A,A′)(A,A^\prime) with A′=A+2A^\prime=A+2. The corresponding isotope shifts ναA,A′=ναA−ναA′\nu_{\alpha}^{A,A^\prime}=\nu_{\alpha}^{A}-\nu_{\alpha}^{A^\prime} make up the entries of a four-component vector να\boldsymbol{\nu}_{\alpha} for transition α\alpha (and similarly for γ\gamma). This vector να\boldsymbol{\nu}_{\alpha} can be written as a linear combination of the field and mass shifts [17], additional higher-order SM-based shifts, and a term induced by the interaction with the proposed boson. Each of these terms can be decomposed into an electronic factor (with subscript α\alpha) and a vector encoding the nuclear structure:

να=Fαδ⟨r2⟩+Kαw+Gα(2)δ⟨r2⟩2+Gα(4)δ⟨r4⟩+αNPαEMDαh+…(1)\boldsymbol{\nu}_{\alpha}=F_{\alpha}\delta\langle r^{2}\rangle+K_{\alpha}\boldsymbol{w}+G_{\alpha}^{(2)}\delta\langle r^{2}\rangle^{2}+G_{\alpha}^{(4)}\delta\langle r^{4}\rangle+\frac{\alpha_{\mathrm{NP}}}{\alpha_{\mathrm{EM}}}D_{\alpha}\boldsymbol{h}+\ldots \tag*{(1)}

FF, KK, G(2)G^{(2)}, G(4)G^{(4)}, and DD are transition-dependent factors for the multiplicative electronic contribution to the field shift, mass shift, quadratic field shift, quartic shift, and a shift induced by a new boson, respectively. The components of w\boldsymbol{w} are wA,A′=m172/mA−m172/mA′=1/ηA−1/ηA′w^{A,A^\prime}=m_{172}/m_{A}-m_{172}/m_{A^\prime}=1/\eta_{A}-1/\eta_{A^\prime}, the inverse nuclear mass differences of the isotopes AA and A′A^\prime with respect to the nuclear mass of 172Yb{}^{172}\mathrm{Yb}. δ⟨rn⟩\delta\langle r^{n}\rangle is the four-vector with elements δ⟨rn⟩A,A′=⟨rn⟩A−⟨rn⟩A′\delta\langle r^{n}\rangle^{A,A^\prime}=\langle r^{n}\rangle^{A}-\langle r^{n}\rangle^{A^\prime}, the differences between the nn-th moments of the nuclear charge distributions. δ⟨r2⟩2\delta\langle r^{2}\rangle^{2} has the components (δ⟨r2⟩2)A,A′=(δ⟨r2⟩A,176)2−(δ⟨r2⟩A′,176)2(\delta\langle r^{2}\rangle^{2})^{A,A^\prime}=(\delta\langle r^{2}\rangle^{A,176})^{2}-(\delta\langle r^{2}\rangle^{A^\prime,176})^{2} constructed from squared radius differences. Additional higher-order SM contributions may also contribute to Eq. (1) at a given experimental accuracy. αNP=(−1)s+1ynye/(4πℏc)\alpha_{\mathrm{NP}}=(-1)^{s+1}y_{n}y_{e}/(4\pi\hbar c) is the product of the coupling constants yny_{n} and yey_{e} of the new boson (with mass mϕm_{\phi} and spin ss) to the neutron and electron. It enters the Yukawa potential Vne(r)=ℏc⋅αNP⋅exp⁡(−rmϕc/ℏ)/rV_{ne}(r)=\hbar c\cdot\alpha_{\mathrm{NP}}\cdot\exp(-rm_{\phi}c/\hbar)/r [4] generated by the new boson. We normalize αNP\alpha_{\mathrm{NP}} by the fine-structure constant αEM=e2/(4πℏc)\alpha_{\mathrm{EM}}=e^{2}/(4\pi\hbar c), where ee is the elementary charge. h=−(2,2,2,2)\boldsymbol{h}=-(2,2,2,2) is the vector of neutron number differences for the neighboring isotope pairs.

Usually, only the first two terms in Eq. (1) contribute significantly. With isotope-shift measurements for two transitions α\alpha and γ\gamma, one can eliminate δ⟨r2⟩\delta\langle r^{2}\rangle from Eq. (1) so that the first two terms yield the linear King relation [4, 17]

ν~γ≈Fγαν~α+Kγα1,(2)\widetilde{\nu}_{\gamma}\approx F_{\gamma\alpha}\widetilde{\nu}_{\alpha}+K_{\gamma\alpha}\mathbf{1}, \tag*{(2)}

with the mass-normalized ν~γ=νγ/w\widetilde{\nu}_{\gamma}=\boldsymbol{\nu}_{\gamma}/\boldsymbol{w}, 1=(1,1,1,1)\mathbf{1}=(1,1,1,1), Fγα=Fγ/FαF_{\gamma\alpha}=F_{\gamma}/F_{\alpha}, and Kγα=Kγ−FγαKαK_{\gamma\alpha}=K_{\gamma}-F_{\gamma\alpha}K_{\alpha}. Deviations from Eq. (2) indicate the presence of additional terms beyond the leading-order field and mass shifts, as in Eq. (1).

Experimental results.– In Figure 1 we present an overview over the two experimental setups as well as a King plot of the isotope shifts of the γ\gamma transition with respect to the isotope shifts of the α\alpha transition normalized with the inverse mass-ratio difference wA,A′w^{A,A^\prime}.

Scheme of experimental setups and the King-plot analysis

Figure 1. Scheme of experimental setups and the King-plot analysis. a Laser spectroscopy setup for the optical frequency ratio measurements of the transitions near 411 nm (α\alpha) and 467 nm (γ\gamma). The laser fields with wavelengths near 822 nm and 934 nm are first stabilized to an ultra-low expansion (ULE) cavity for short-term frequency stability and then locked to a cryogenic silicon cavity (Si-cavity). The second harmonic conversions of the lasers are focused to interrogate the ions. b King plot of the isotope shifts of the γ\gamma transition with respect to the isotope shifts of the α\alpha transition normalized with the inverse mass-ratio difference wA,A′w^{A,A'}, with insets magnified by a factor of 1.25×1051.25 \times10^{5}. c Residuals of the linear fit in the King plot. For visibility, the uncertainties on the residuals are multiplied by a factor of 200. d Penning-trap setup for the determination of cyclotron frequency ratios. Three highly charged ions (blue and red), produced by an electron beam ion trap, are transported and stored in an A′−A−A′A'-A-A' sequence. By shuttling the set of ions up and down between neighbouring traps (alternating between positions 1 and 2), cyclotron frequency ratios are determined from sequential measurements in the two measurement traps, referred to trap 2 and 3 in the figure. The tunable image-current detection system allows for the determination of the axial eigenfrequencies of different isotopes at equal charge state and equal trapping potential [18]. The radial eigenfrequencies are determined indirectly using the same detection system.

The nuclear mass ratios ηA\eta_{A} of ytterbium isotopes are determined from the cyclotron frequency ratios of highly charged Yb42+^{42+} ions measured at the cryogenic Penning-trap mass spectrometer PENTATRAP [10, 19, 20] and their calculated electron binding energies. The determination of the cyclotron frequencies νc\nu_{c} of two isotope ions allows one to extract their ionic mass ratio via RCA,A′=νc(172Yb42+)/νc(AYb42+)=m(AYb42+)/m(172Yb42+)\mathrm{R}_{\mathrm{C}}^{A,A^\prime}=\nu_{c}({}^{172}\mathrm{Yb}^{42+})/\nu_{c}({}^{A}\mathrm{Yb}^{42+})=m({}^{A}\mathrm{Yb}^{42+})/m({}^{172}\mathrm{Yb}^{42+}). Determining the free cyclotron frequency ν\nu requires measuring all three eigenfrequencies of the trapped ion via νc2=ν+2+νz2+ν−2\nu_{c}^{2}=\nu_{+}^{2}+\nu_{z}^{2}+\nu_{-}^{2} [21], namely the trap-modified cyclotron frequency ν+\nu_{+}, the axial frequency νz\nu_{z}, and the magnetron frequency ν−\nu_{-}.

From the calculated electron binding energies E172(28)=350 773(5) eVE_{172}^{(28)}=350\,773(5)\ \mathrm{eV} of the 28 electrons in the 172Yb42+{}^{172}\mathrm{Yb}^{42+} ion and E172(70)=382 301(16) eVE_{172}^{(70)}=382\,301(16)\ \mathrm{eV} of the 70 electrons in the 172Yb{}^{172}\mathrm{Yb} atom [22], the neutral mass m(172Yb)m({}^{172}\mathrm{Yb}) [9, 23], and the electron mass mem_{e} [24], one can derive the necessary nuclear mass ratios ηA\eta_{A} to similar accuracies. The mass ratios are used instead of single mass values in atomic mass units, since the former are much less sensitive to the uncertainties of the binding energies and the reference mass m(172Yb)m({}^{172}\mathrm{Yb}). Their final values are given in Tab. I with relative uncertainties of 4×10−124\times10^{-12}, corresponding to uncertainties of 0.3 Hz on the isotope shifts. For comparison, previous mass determinations affected the King-plot analysis at a level of 3–30 Hz.

To make use of the new mass uncertainties, we improve the uncertainties of the isotope shifts by performing absolute frequency measurements of the α\alpha and γ\gamma transitions for the five isotopes. Singly charged ytterbium ions are trapped in a segmented, linear radio-frequency Paul trap [25, 26]. The excitation lasers near wavelengths of 411 nm and 467 nm are locked to ultra-low-expansion cavities and to a cryogenic silicon cavity [27] via a frequency comb.

The absolute transition frequencies are obtained with optical frequency ratio measurements by referencing to the γ\gamma transition between the F=0F=0 and F=3F=3 hyperfine states of the 171Yb+{}^{171}\mathrm{Yb}^{+} ion [28]. From the measurements, we obtain isotope shifts shown in Tab. I, with uncertainties below 6 Hz and 16 Hz for the α\alpha and the γ\gamma transitions, respectively.

For details on the mass-ratio measurements, spectroscopic measurements, systematic shifts, and uncertainties, see the Supplemental Material.

King-plot analysis.– A King plot using our mass-normalized isotope-shift measurements of the γ\gamma transition (denoted ν~γPTB\widetilde{\nu}_{\gamma\mathrm{PTB}}) and the α\alpha transition (denoted ν~αPTB\widetilde{\nu}_{\alpha\mathrm{PTB}}) is given in the End Matter. From the King-plot analysis, we observe deviations from linearity averaging to 20.17(2) kHz.

To determine the origin of the nonlinearity in the King plot, we perform a nonlinearity decomposition analysis of all available (sub-)kHz-precision isotope-shift data. Apart from αPTB\alpha_{\mathrm{PTB}} and γPTB\gamma_{\mathrm{PTB}} presented in this work these are αMIT\alpha_{\mathrm{MIT}} and γMIT\gamma_{\mathrm{MIT}} from Ref. [7], the 2S1/2→2D3/2{}^{2}S_{1/2} \rightarrow{}^{2}D_{3/2} transition in Yb+^{+} [5] (denoted β\beta), 1S0→3P0{}^{1}S_{0} \rightarrow{}^{3}P_{0} in Yb [29] (denoted δ\delta), and 1S0→1D2{}^{1}S_{0} \rightarrow{}^{1}D_{2} in Yb [30] (denoted ϵ\epsilon). In total, we construct 7 mass-normalized isotope-shift vectors ν⃗τ\vec{\nu}_{\tau}, with τ∈{αPTB,αMIT,β,γPTB,γMIT,δ,ϵ}\tau\in\{\alpha_{\mathrm{PTB}}, \alpha_{\mathrm{MIT}}, \beta, \gamma_{\mathrm{PTB}}, \gamma_{\mathrm{MIT}}, \delta, \epsilon\}.

Since these vectors are four-vectors, they are uniquely described by their projections onto four basis vectors. We choose the basis vectors ν~δ\tilde{\nu}_{\delta} and 11, which span the plane of King linearity Eq. (2), with α→δ\alpha\rightarrow\delta, and Λ+\Lambda_{+} and Λ−\Lambda_{-} (defined in the Supplemental Material), which are orthogonal to this plane [7]. We obtain ν~τ=Fτδν~δ+Kτδ1+λ+(τ)Λ++λ−(τ)Λ−\tilde{\nu}_{\tau} = F_{\tau\delta}\tilde{\nu}_{\delta} + K_{\tau\delta}1 + \lambda_{+}^{(\tau)}\Lambda_{+} + \lambda_{-}^{(\tau)}\Lambda_{-} with FτδF_{\tau\delta} and KτδK_{\tau\delta} as given in Eq. (2) and the coordinates (λ+(τ),λ−(τ))(\lambda_{+}^{(\tau)}, \lambda_{-}^{(\tau)}) characterizing the deviation of the isotope shift ντ\nu_{\tau} from the linear relation in Eq. (2). As shown in Figure 2a, the data points (λ+(τ),λ−(τ))(\lambda_{+}^{(\tau)},\lambda_{-}^{(\tau)}) lie to a good approximation on the solid black line through the origin of the (λ+,λ−)(\lambda_{+},\lambda_{-}) plane. This implies that the tension of the Yb isotope-shift data with respect to King linearity can to a large extent be explained by a nonlinearity source (new physics, for example) with the appropriate slope λ−/λ+\lambda_{-}/\lambda_{+} in the decomposition plot. We compare the slope of the linear fit to the slopes predicted by the new physics term (dash-dotted line) and the quadratic field shift δ⟨r2⟩\delta\langle r^{2}\rangle (dotted line, from experimental δ⟨r2⟩\delta\langle r^{2}\rangle data [31]). Both have uncertainties not visible in Figure 2a, so we conclude that neither can be the leading source of the nonlinearity in the Yb King plot. Details of the nonlinearity decomposition are provided in the Supplemental Material. Another candidate is nuclear deformation [6, 7], in particular δ⟨r4⟩\delta\langle r^{4}\rangle, which we predict using an ab initio approach.

Nonlinearity decomposition and extracted trend of nuclear deformation

Figure 2. Nonlinearity decomposition and the extracted trend of nuclear deformation, δ⟨r4⟩\delta\langle r^{4}\rangle. (a) Observed and predicted nonlinearities in the (λ+,λ−)(\lambda_{+},\lambda_{-}) plane. The nonlinearities of the transitions αPTB\alpha_{\mathrm{PTB}} (this work, see Table I), β\beta [5], γPTB\gamma_{\mathrm{PTB}} (this work, see Table I), γMIT\gamma_{\mathrm{MIT}} [5, 7], and ϵ\epsilon [30] with respect to the δ\delta [29] transition are plotted in the single-source linear fit (solid line) whose slope is compared to the predicted slopes for the δ⟨r2⟩\delta\langle r^{2}\rangle (dotted), δ⟨r4⟩\delta\langle r^{4}\rangle (dashed lines and gray band based on predictions in b), and new physics (dash-dotted) nonlinearities. The residuals of the single-source linear fit are shown in the lower panel. (b) Solid line: δ⟨r4⟩A,A−2\delta\langle r^{4}\rangle^{A,A-2} values relative to δ⟨r4⟩176,174=7 fm4\delta\langle r^{4}\rangle^{176,174}=7\,\mathrm{fm}^{4} (star) extracted from isotope shifts of the αPTB\alpha_{\mathrm{PTB}} transition using atomic theory. Dashed lines: ab initio nuclear theory predictions [1.8/2.0 (EM), VS1 and VS2; Δ\DeltaN2^{2}LOGO, VS1]. The estimated nuclear theory uncertainties (68% confidence interval) are given by the gray bands.

Nuclear structure effects.– Recent developments have made heavy nuclei accessible to ab initio nuclear structure calculations [15, 32]. To predict δ⟨r4⟩\delta\langle r^{4}\rangle, we use the valence-space in-medium similarity renormalization group (VS-IMSRG) [12, 13] together with the quasiparticle vacua shell model (QVSM) [32] to solve the many-body Schrödinger equation. We employ nucleon-nucleon and three-nucleon interactions from chiral effective field theory, using the so-called 1.8/2.0 (EM) [33] and Δ\DeltaN2^{2}LOGO [34] interactions, which differ in their construction and how they are fit to data, to give insight into interaction uncertainties. To assess many-body uncertainties, we solve the VS-IMSRG at the two- and three-body truncations [35] and employ two different valence spaces, VS1 with a 132^{132}Sn core and VS2 with a 154^{154}Gd core, in the QVSM. In Figure 2a, we show the prediction of the nonlinearity from our nuclear structure calculations, where the uncertainty is represented by the gray band. This uncertainty stems from a correlated statistical model accounting for interaction and many-body uncertainties including correlations between isotope pairs. Details of our nuclear structure calculations and the uncertainty quantification are provided in the Supplemental Material. The two sets of calculations using the 1.8/2.0 (EM) Hamiltonian with valence spaces VS1 and VS2 serve as representative samples of our nuclear theory predictions. Since the best-fit line is compatible with the ab initio calculations for δ⟨r4⟩\delta\langle r^{4}\rangle, we assume δ⟨r4⟩\delta\langle r^{4}\rangle to be the leading King-plot nonlinearity in Yb.

Combining isotope-shift measurements, nuclear mass-ratio measurements, and charge radius measurements [31] with atomic structure calculations of Gα(4)G_{\alpha}^{(4)} [see Eq. (1)] using AMBiT [36], we recast the Yb King-plot analysis into a measurement of nuclear deformation, which can be used to benchmark nuclear structure calculations. The procedure is given in the Supplemental Material, and the extracted changes in δ⟨r4⟩\delta\langle r^{4}\rangle relative to the reference value of δ⟨r4⟩176,174=7 fm4\delta\langle r^{4}\rangle^{176,174}=7\,\mathrm{fm}^{4} (star) are shown in Figure 2b. This reference value is based on input from both our ab initio results and density functional theory calculations [7] that all predict δ⟨r4⟩176,174=6\delta\langle r^{4}\rangle^{176,174}=6–8 fm48\,\mathrm{fm}^{4}. The experimental data show that the evolution of δ⟨r4⟩A,A−2\delta\langle r^{4}\rangle^{A,A-2} along the isotope chain is nearly flat, remarkably consistent with our ab initio calculations within uncertainties. Nonetheless, from the residuals for the transitions αPTB\alpha_{\mathrm{PTB}}, β\beta, γPTB\gamma_{\mathrm{PTB}}, γMIT\gamma_{\mathrm{MIT}} and ϵ\epsilon, shown in Figure 2a, we deduce a 23σ23\sigma preference for more than one linearity, leaving open the possibility of a new boson being responsible for the next-to-leading King nonlinearity. This strengthens the prior two-source hypothesis [7] by a factor of more than 5.

Bounds on new physics.– To extract bounds on the hypothetical new boson, we combine our isotope-shift measurements and nuclear mass measurements with the isotope-shift measurements of Ref. [29]. This allows us to eliminate both the charge radius variance δ⟨r2⟩\delta\langle r^2\rangle and δ⟨r4⟩\delta\langle r^4\rangle from the system of isotope-shift equations without requiring theoretical input. Assuming higher-order SM terms beyond δ⟨r4⟩\delta\langle r^4\rangle to be negligible, the generalized King-plot [11] can be used to set an upper bound on the new physics coupling αNP\alpha_{\mathrm{NP}} as a function of the mass mϕm_\phi of the new boson. Note that King-plot bounds should always be understood in the context of a given dataset since they are highly sensitive, not only to the central values and uncertainties of the frequency and mass measurements, but also to the unknown nuclear and electronic effects that are reflected in the isotope shift data. These aspects are discussed in more detail in the Supplemental Material.

The new bound provided by this work is shown in Figure 3 in red. If the second King-plot nonlinearity were to be explained by new physics only, the couplings of the new boson would be expected to reside between the 2σ2\sigma upper bound (solid red line) and the 2σ2\sigma lower bound (dotted red line), which are only distinguishable in the inset. The new bound supersedes, both in precision and in magnitude, the Ca+^+ King plot bounds [37, 38] and the generalized King plot bounds using previous isotope shift data in Yb+^+ [5, 7, 29] in combination with the new mass-ratio measurements provided by this work (see Table I). Moreover, it takes into account the significant shifts in the mass-ratio measurements ηA\eta_A and the isotope shifts ναA,A+2\nu_{\alpha}^{A,A+2} highlighted in Table I. For instance, the differences in the ναA,A+2\nu_{\alpha}^{A,A+2} measurements entering the bound from the dataset (αMIT,γMIT,δ)(\alpha_{\mathrm{MIT}},\gamma_{\mathrm{MIT}},\delta) (shown in black) and the new bound (red) lead to the displacement of the characteristic peaks at the high end of the plotted mϕm_\phi values, resulting in different slopes. A detailed comparison of our new bound with respect to the bounds presented in Refs. [5, 7], alongside a discussion of the competing astrophysical and laboratory bounds, which are shown in Figure 3 as exclusion regions and green dashed curves, can be found in the Supplemental Material.

Exclusion plot for the new boson coupling to electrons and neutrons

Figure 3. Exclusion plot for the new boson ϕ\phi coupling to electrons and neutrons. Solid lines: 2σ2\sigma upper bounds from Ca+^+ King plots (cyan: Aarhus [37], blue: Williams [38]), Yb+^+ King plots (black, maroon: [5, 7, 29]), and from this work (red) on the product of couplings yeyn/fc=4παNPy_e y_n/fc = 4\pi\alpha_{\mathrm{NP}} to electrons and neutrons depending on the new boson’s mass mϕm_\phi. The inset shows an extract of the 2σ2\sigma upper bound (solid red line) and the 2σ2\sigma lower bound (dotted red line) produced by this work. The shaded areas are disfavored by the constraint on yey_e from (g−2)e(g-2)_e [39, 40], times the constraints on yny_n from neutron optics [41] and neutron scattering experiments [41, 42, 43, 44] (yellow), hydrogen and deuterium spectroscopy using electron scattering or the Lamb shift in muonic atoms to determine the charge radius [45] (see also Ref. [46]), fifth force searches [47, 48], stellar evolution in the globular cluster [49, 50], and energy loss in the supernova SN1987A [51]. The dashed curves show the constraint on yey_e from (g−2)e(g-2)_e times the constraint on yny_n from K→π+invisibleK \to\pi+ \text{invisible}, assuming the new boson ϕ\phi couples only to the top-quark (via yty_t) or only to the up-quark (via yuy_u). The purple band labeled “Be” indicates the coupling range required for a protophobic boson with mϕ=17 MeV/c2m_\phi= 17\,\mathrm{MeV}/c^2 to explain the ATOMKI anomaly [52].

Discussion and conclusion.— We measured the isotope shifts for the 2S1/2→2D5/2^2S_{1/2} \to{}^2D_{5/2} electric quadrupole transition and the 2S1/2→2F7/2^2S_{1/2} \to{}^2F_{7/2} electric octupole transition on five stable, spinless Yb+^+ isotopes and with relative uncertainties on the order of 10−810^{-8} to 10−910^{-9}, as well as mass ratios of these isotopes to a precision of 4×10−124 \times10^{-12}.

Our measurements, combined with atomic structure calculations, have enabled a first direct extraction of the evolution of δ⟨r4⟩\delta\langle r^4\rangle across the ytterbium isotope chain. To understand whether the change in the nuclear charge distribution is consistent with the strong interaction, we have performed ab initio calculations based on chiral effective field theory interactions. Our results reproduce the experimental δ⟨r4⟩\delta\langle r^4\rangle remarkably well for such heavy nuclei. The ⟨r4⟩\langle r^4\rangle nuclear charge moment can provide information about both nuclear deformation [6] (because Yb nuclei are prolate deformed) and the surface thickness of the nuclear density [53, 54]. It has also been shown that this charge moment can provide insights to experimentally estimate the neutron-skin thickness [54, 55]. An exciting future direction is to advance this work to the neutron-rich calcium isotopes. This can widen the search for the new boson and shed light on the puzzling increase of charge radii towards 52^{52}Ca [2].

Our isotope-shift and nuclear mass measurements, combined with the isotope-shift measurements of Ref. [29] allow us to set competitive spectroscopy bounds on new bosonic mediators between neutrons and electrons using the generalized King-plot method [11]. With increasing experimental precision, data from more isotopes will be needed to overcome atomic and nuclear structure uncertainties. In ytterbium, a possible candidate is 166^{166}Yb, with a half-life of 54 hours. Another option is to use elements Sn [56] or Xe [57], each with seven spinless, stable isotopes and suitable clock transitions that could be found in different ionization stages [58].

Acknowledgments.— We thank the referees for very valuable feedback on the manuscript. We thank Joonseok Hur, Martin Steinel and Vladan Vuletić for helpful discussions. This work is supported by the Max Planck Society (MPG), the International Max Planck Research Schools for Precision Tests of Fundamental Symmetries (IMPRS-PTFS), the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme under Grant Agreements 832848 (FunI) and 101020842 (EUSTRONG), and by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project-ID 273811115 – SFB 1225 ISOQUANT as well as under Germany’s Excellence Strategy – EXC-2123 QuantumFrontiers – 390837967 (RU B06) and through Grant No. CRC 1227 (DQ-mat, projects B02 and B03). This work has been supported by the Max-Planck-RIKEN-PTB-Center for Time, Constants, and Fundamental Symmetries. L.S.D. acknowledges support from the Alexander von Humboldt Foundation. This work comprises parts of the Ph.D. thesis work of M.D. submitted to Heidelberg University, Germany, parts of the Ph.D. thesis work of C.-H.Y. submitted to Leibniz Universität Hannover, Germany, and parts of the Ph.D. thesis work of M.H. submitted to the TU Darmstadt, Germany. M.H., T.M., and A.S. gratefully acknowledge the computing time provided on the high-performance computer Lichtenberg at the NHR Centers NHR4CES at TU Darmstadt. N.S. acknowledges the support of the “Program for promoting research on the supercomputer Fugaku”, MEXT, Japan (JPMXP1020230411), and the MCRP program of the Center for Computational Sciences, University of Tsukuba (NUCLSM). E.F. and F.K. thank CERN for their hospitality during the early phase of this work.

References

Experimental setup for the mass-ratio determinations

The experimental setup of PENTATRAP as well as the measurement schemes and techniques are similar to those in prior measurements. Details can be found in [10, 19, 59, 60, 61, 62] and just a short overview and details of this measurement are given.

The mass-ratio determinations to the precision presented rely on the use of highly charged ions to increase the cyclotron frequency. The highly charged AYb42+{}^{A}\mathrm{Yb}^{42+} ions are produced in a room-temperature electron beam ion trap via in-trap laser ablation [63], allowing isotope selection via multi-position targets with isotope-enriched samples of YbO. A cloud of highly charged ions is extracted with a kinetic energy of approximately 4 keV/qq and transported through an electrostatic beamline equipped with a Bradbury-Nielsen Gate [64] for time-of-flight selection of a single charge state. The initial kinetic energy is reduced to a few eV/qq via two pulsed drift tubes, one above

ParameterTrap 2Trap 3
r0r_0 [mm] [19]5.000 (5)5.000 (5)
TRT_R [1]0.87966 (15)0.879002 (7)
U0U_0 [V]−28.14-28.14−12.85-12.85
TzT_z [K]5 (2)8 (2)
ρ+,exc\rho_{+,exc} [µm]12 (2)12 (2)
RLC QQ-factor [1]≈ 3300≈ 9800
ν+\nu_+ [MHz]≈ 26.26≈ 26.27
νz\nu_z [kHz]≈ 707.3≈ 478.1
ν−\nu_- [kHz]≈ 9.5≈ 4.4
B0B_0 [T]7.00215 (2)7.00216 (2)
B2B_2 [mT/m2\mathrm{mT}/\mathrm{m}^2]28 (2)−5-5 (2)
c2c_2 [10−3 mm−210^{-3}\ \mathrm{mm}^{-2}]−14.88576-14.88576 (1)−14.89708-14.89708 (1)
c4/c2c_4/c_2 [10−5 mm−210^{-5}\ \mathrm{mm}^{-2}]−3-3 (8)−7-7 (8)
c6/c2c_6/c_2 [10−6 mm−410^{-6}\ \mathrm{mm}^{-4}]−4-4 (6)−2-2 (6)

Table 2. TABLE II. Trap parameters relevant in the analysis of the systematic effects. The approximate eigenfrequencies are given for the common reference ion 172Yb42+{}^{172}\mathrm{Yb}^{42+}. The magnetic field was calculated using the cyclotron frequency νc\nu_c and ion’s charge-to-mass ratio from literature [23].

the magnet and one right above the Penning-trap assembly inside the magnet. At this energy, the ions can be caught in the upper trap using a low-voltage potential-lift scheme.

The cryogenic setup with a stack of five identical Penning traps is placed in a superconducting magnet with a field strength of 7 T. The traps and detection electronics, see Figure 4, are cooled to the temperature of liquid helium (4 K) providing a low noise environment for ion detection and cooling. The ions are loaded with deliberate low transport efficiency to load only a single ion per shot into the trap tower.

Overview of the Pentatrap measurement setup and axial detection spectra

Figure 4. Overview of the PENTATRAP measurement setup with a sectional view of the trap tower, approximate schematics of the detection system, and example spectra of the axial detection in trap 2 for the determination of the cyclotron frequency ratio of 176Yb42+{}^{176}\mathrm{Yb}^{42+} over 172Yb42+{}^{172}\mathrm{Yb}^{42+}. The lower spectrum shows the detection system resonance frequency tuned to match one of the two axial frequencies of the respective ion during the measurement in position 1 and position 2. The upper spectrum shows a zoom-in of one axial dip spectrum of 172Yb42+{}^{172}\mathrm{Yb}^{42+} with the model fit given as a red line.

For maximal relative precision (highest possible cyclotron frequency), we used one of the highest charge states of 42+ produced by our ion source. Given the field strength of the superconducting magnet and the resonance frequencies of the detection systems, the eigenfrequencies of trapped AYb42+{}^{A}\mathrm{Yb}^{42+} ions are approximately 26 MHz for the trap-modified cyclotron frequency, 700 kHz and 475 kHz for the axial frequencies, and 9.5 kHz and 4.3 kHz for the magnetron frequencies for traps 2 and 3, respectively. A summary of all trapping parameters is given in Tab. II.

The axial eigenmotion is resistively cooled via the detection system: The induced image current by an ion creates a voltage drop across the high impedance of the connected superconducting resonance circuit (LCR). This voltage signal acts as negative feedback dampening the ion’s axial motion, limited by the thermal noise of the detection systems at approximately 4.2 K. Radial modes are indirectly cooled by energy exchange to the axial mode by sideband coupling through a quadrupolar RF field [65].

The supposedly single ions are cleaned from possible contaminant ions by using broadband magnetron excitation with simultaneous magnetron sideband cooling of the ion of interest and trap potential variation. Following this procedure, a set of three ions is prepared in an AA-A′A^\prime-AA sequence in the three central traps, and these are reloaded on average every second day due to charge exchange with rest gas atoms.

For the measurement of the axial frequencies, the dip method is applied [66]: When the ion reaches thermal equilibrium with the detection circuit, the trapped ion acts equivalent to a parallel connected series LC-circuit with a low impedence at the axial frequency. With the ions axial frequency matched to the resonance frequency of the detection circuit, a dip in the resonance can be fitted to extract the axial frequency, see Figure 4. Because of the difference in charge-to-mass ratio, the axial frequencies of different isotopes in the same charge state, stored at the same trapping potential, are separated by multiple kHz. This difference is bridged by tuning the resonance frequency of the LCR circuit using varactor diodes (voltage controlled capacitances) [18], see Figure 4. In this way, the trap potential settings can be kept identical for two ions with a charge-to-mass ratio difference of up to 3%, ensuring identical field imperfections for both ions and suppressed systematic shifts on the determined cyclotron frequency ratio.

For the radial modes the double-dip method is used and for the precise determination of the trap-modified cyclotron frequency, the phase-sensitive Rabi-type Pulse and Phase (PnP) technique [67] is applied. The initial phase imprint for the PnP method is realized by dipolar excitation of the trap-modified cyclotron mode, increasing its amplitude ρ+\rho_{+} on the order of 12 μ\mum. This is followed by a free evolution of the phase and finally coupling of the radial and axial modes to measure the evolved phase using the axial detection system. A reference phase at a short phase evolution time is subtracted from the long evolution time phase to cancel phase offsets in the measurement chain. The equality of excitation radii for the different ions has been tested to a level of 2%, which ensures a low impact of systematic shifts dependent on the excitation radii on the final cyclotron frequency ratio.

During the long phase evolution time, the dip spectrum is recorded, resulting in a simultaneous determination of the two higher eigenfrequencies. The ion species is switched inside both measurement traps by transporting the three ions up or down via potential variation, with one ion always in one of the two storage traps 1 or 4.

Traps 2 and 3 are used simultaneously as measurement traps which increases statistics and allows for cross-checks of trap and detection system related systematic shifts. The determination of the cyclotron frequencies of two ions is done sequentially. To switch ion species in the measurement traps the set of ions is adiabatically transported to the neighboring traps, see positions 1 and 2 in Figure 4.

At the beginning of each measurement cycle, all eigenfrequencies are determined roughly via dip and double-dip methods for both types of ions to determine measurement settings for the PnP method and tune the detection system on resonance with the axial frequencies. The automated measurement cycle is split into two parts: The trap-modified cyclotron frequency is determined via multiple PnP-type phase measurements with a set of up to 10 different evolution times between 0.1 and 100 s, each phase measurement repeated for 4 to 6 times, allowing one to unwrap the absolute frequency. In the second part, typically running for 12 hours before repeating the absolute frequency determination, only the shortest and longest of these evolution times are measured to follow the change of the trap-modified cyclotron frequency due to changes in the magnetic field. This is possible due to our liquid helium level and pressure-stabilized superconducting magnet, resulting in a shot-to-shot ν+\nu_{+} frequency stability on the level of 33–5×10−115 \times10^{-11}. During these 12 hours, the ion species is switched between 20 and 30 times in each trap.

The cyclotron frequency νc=qB/(2πm)\nu_{c}=qB/(2\pi m) of an ytterbium ion with mass mm and charge qq in a magnetic field BB is determined from all measured eigenfrequencies with νc2=ν+2+νz2+ν−2\nu_c^2 = \nu_+^2 + \nu_z^2 + \nu_-^2 [21], with the magnetron frequency ν−\nu_- determined in the preparation of the measurement cycle with sufficient precision due to the frequency hierarchy and respective error propagation. Every subsequent cyclotron frequency measurement of the two ions yields a cyclotron frequency ratio. The ratio is calculated from interpolated values of the cyclotron frequency measurement of one ion species to the time-stamp of the other ion species to adjust to first-order magnetic field drifts.

With the measured RACFR_A^{\mathrm{CF}} the nuclear mass ratios are derived via

ηA=RACF+(RACF−1)[28me−E172(28)/c2]−ΔEA(28)/c2m(172Yb)−70me+E172(70)/c2.(3)\eta_A = R_A^{\mathrm{CF}} + \frac{\left(R_A^{\mathrm{CF}} - 1\right)\left[28m_e - E_{172}^{(28)}/c^2\right] - \Delta E_A^{(28)}/c^2}{m(^{172}\mathrm{Yb}) - 70m_e + E_{172}^{(70)}/c^2}. \tag*{(3)}

Here, m(172Yb)m(^{172}\mathrm{Yb}) [23] is the mass of the neutral atom and mem_e [24] the electron mass. E172(28)E_{172}^{(28)} and E172(70)E_{172}^{(70)} are the binding energies of the 28 electrons in the 172Yb42+^{172}\mathrm{Yb}^{42+} ion and 70 electrons in the 172Yb^{172}\mathrm{Yb} atom, respectively. Thus, the denominator in Eq. (3) represents the nuclear mass of 172Yb^{172}\mathrm{Yb}, i.e., m(172Yb70+)m(^{172}\mathrm{Yb}^{70+}), or m172m_{172} in the main text. The necessary electron binding energies and their dependence on isotopes ΔEA(28)=E172(28)−EA(28)\Delta E_A^{(28)} = E_{172}^{(28)} - E_A^{(28)} are calculated with the GRASP2018 code [22] discussed in later sections.

While the mass ratio can be used in full precision for the King-plot analysis, the neutral masses of the isotopes can also be calculated using the presented binding energies. The resulting uncertainty is limited by the literature value for m(172Yb)m(^{172}\mathrm{Yb}) with a relative uncertainty of 8×10−118 \times10^{-11}. All neutral masses are given in Table III and are compared to the current literature values of the atomic mass evaluation 2020 [23]. The masses of three isotopes agree within 1.5σ1.5\sigma and the mass of 168Yb^{168}\mathrm{Yb} shows a deviation by 3.5σ3.5\sigma. The uncertainty of m(168Yb)m(^{168}\mathrm{Yb}) is improved compared to the literature value by a factor of 7.

Isotope AAm(A)m(A) (this work)m(A)m(A) [23]
168167.933 890 939 (14)167.933 891 297 (100)
170169.934 767 218 (14)169.934 767 242 (11)
174173.938 867 541 (14)173.938 867 545 (11)
176175.942 574 697 (14)175.942 574 706 (15)

Table III. Neutral masses in atomic mass units of the five even ytterbium isotopes determined from the measured cyclotron frequency ratios RACF=νc,172/νc,AR_A^{\mathrm{CF}} = \nu_{c,172}/\nu_{c,A}, the binding energy to of the 42 missing electrons E172(70)−E172(28)E_{172}^{(70)} - E_{172}^{(28)}, and the literature mass of isotope m(172Yb)m(^{172}\mathrm{Yb}) [23]. For comparison, we also provide the literature mass values listed in the atomic mass evaluation 2020 [23].

Systematic effects and analysis of the mass-ratio measurement data

A summary of all systematic shifts on the determined ratios can be found in Table IV for both traps and the different mass ratios. Some of the given shifts are commonly known shifts and uncertainties on the real motional frequencies. Among these are the relativistic shift [68], shifts due to field imperfections [68, 69], and the image charge shift [70, 71]. The trap parameters necessary to calculate these and other effects are given in Tab. II. The dipole excitation necessary for the phase imprint on the trap-modified cyclotron motion for the PnP technique was compared between two isotopes to check for equally increased amplitudes ρ+\rho_+ of this radial mode and yielded a maximum deviation of 2%. Although the absolute excitation radius cannot be determined with such precision, the precision of the excitation radius ratio allows for a significant reduction in systematic uncertainties.

EffectATrap 2Trap 3
Field imperfect (Bi,ci)(B_i, c_i)all0.0 (0.5)0.0 (0.5)
Nonlinear phase readoutall0.0 (1.0)0.0 (1.5)
Magnetron frequencyall0.0 (1.0)0.0 (1.0)
Dip lineshapeall0.0 (4.0)0.0 (5.0)
Common totalall0.0 (4.4)0.0 (5.3)
Image charge shift176−10.0-10.0 (0.5)
Image charge shift174−4.9-4.9 (0.2)
Image charge shift1704.8 (0.2)
Image charge shift1689.5 (0.5)
Special relativity1761.0 (1.5)
Special relativity1740.5 (1.5)
Special relativity170−0.5-0.5 (1.5)
Special relativity168−1.0-1.0 (1.5)

Table IV. Systematic corrections ΔR\Delta R for the determined ratio RCF=R′+ΔRR^{\mathrm{CF}} = R' + \Delta R with R′=νc,172/νc,AR' = \nu_{c,172}/\nu_{c,A} being the measured ratio. All values are expressed in parts per trillion (1⋅10−121 \cdot10^{-12}). For more details see text.

The uncertainty of the image charge shift effect originates from the most precise test to date of the respective analytical estimation of this effect, which agreed to a level of 5% limited by the experimental uncertainty [71]. In addition to these, there are systematic uncertainties mainly originating from our detection system and measurement methods. The highest uncertainty originates in the determination of the axial frequency via the dip determination, which is sensitive to the precision with which the resonance circuit can be tuned to the axial frequency [72]. The determination of the resonance frequency is precise to only 2 Hz in our setup. The given uncertainties on the determined ratio due to this effect are extracted from measurements of the axial frequency in dependence on the resonance frequency as well as full ratio determinations with deliberate detuning of the RLC resonance by 10–20 Hz for only one ion.

The nonlinear phase readout originates from a nonlinear transfer function of the ions cyclotron phase at the time of the PnP coupling pulse to the axial mode. The residual of the ideally linear transfer function follows a sinusoidal behavior with amplitudes of 0.25 and 0.5 degrees in trap 2 and 3, respectively. These phase offsets average out for the long accumulation time measurement, but create constant offsets for the short reference phase, which results in the given uncertainties.

The magnetron frequency systematic uncertainty originates from the calculation of the magnetron frequencies of one ion from the measured magnetron frequency of the other, using the literature mass values. This is sufficient, since the cyclotron frequency ratio is more sensitive to the magnetron frequency difference than to the absolute value, which is only determined to low precision before each measurement.

All systematic shifts are corrected in the weighted mean cyclotron frequency ratio results of the individual traps. None of the dominant systematic shifts and uncertainties are correlated between the two measurement traps. The final value is determined from the weighted mean of the results of the two traps.

Ab initio binding energy calculations for mass-ratio correction

The electron binding energies of neutral Yb and Ni-like Yb42+\mathrm{Yb}^{42+} can be obtained by summing the ionization potentials (IPs) of all the charge states listed in the NIST atomic database [73]. This results in the values 382,457 (614) and 350,722 (233) eV for the binding energies of the 70 and 28 electrons, respectively. To reduce uncertainties in these values, advanced atomic structure calculations were performed in this work. The procedure is straightforward for the case of highly charged Yb42+\mathrm{Yb}^{42+}. However, for neutral Yb, many close-lying levels render it difficult to treat the electron correlations accurately. Nevertheless, considering that the IPs of Yb and Yb+\mathrm{Yb}^{+} are experimentally known to be 6.254160 (12) eV and 12.179185 (25) eV [73], respectively, we can partially avoid this problem by calculating the binding energy of Yb2+\mathrm{Yb}^{2+}.

The calculations were performed via the ab initio fully relativistic multiconfiguration Dirac-Hartree-Fock (MCDHF) and relativistic configuration interaction (RCI) methods [74, 75, 76] implemented in the GRASP2018 code [22, 77]. The many-electron atomic state function (ASF) is constructed as a linear combination of configuration state functions (CSFs) with common total angular momentum (JJ), magnetic (MM), and parity (PP) quantum numbers: ∣ΓPJM⟩=∑kck∣γkPJM⟩|\Gamma^{P}JM\rangle= \sum_k c_k|\gamma_k^{P}JM\rangle. Each CSF ∣γkPJM⟩|\gamma_k^{P}JM\rangle is built from products of one-electron orbitals (Slater determinants), jjjj-coupled to the appropriate angular symmetry and parity, and γk\gamma_k represents orbital occupations, together with orbital and intermediate quantum numbers necessary to uniquely define the CSF. Γ\Gamma collectively denotes all the γk\gamma_k involved in the representation of the ASF. ckc_k is the corresponding mixing coefficient. We first solve the MCDHF equations self-consistently [74, 75, 76] under the Dirac-Coulomb Hamiltonian to obtain an ASF, represented by the set of ckc_k, together with the set of radial orbitals. Then the RCI method is employed to calculate the contributions from mass shift (or nuclear recoil effect), Breit interaction, frequency-dependent transverse photon interaction, and QED effects.

The ground state of 172Yb42+{}^{172}\mathrm{Yb}^{42+} is described by the configuration [Ar]3d10 1S0[\mathrm{Ar}]3d^{10}\,{}^{1}S_0. In the single-configuration Dirac-Hartree-Fock calculation, one obtains a binding energy of 351,379.33 eV for a point-like nuclear charge. The corresponding mass shift equals −0.87-0.87 (5) eV with the uncertainty of the order of (me/M)(αZ)4mec2(m_e/M)(\alpha Z)^4m_ec^2 (MM is the mass of the nucleus). The Fermi model for nuclear-charge distribution gives rise to a correction of −45.06-45.06 (8) eV due to the finite nuclear size effect, with the uncertainty resulting from the nuclear radius predictions. The Breit interaction further contributes energy of −367.66-367.66 eV, and the frequency-dependent transverse-photon interaction adds 6.65 eV to the total binding energy. The residual Coulomb-Breit interaction, the so-called correlation energy, is significant as well but more complicated to account for. Thus, its contribution and uncertainty are discussed in detail in later paragraphs. Similar calculations were also performed for neutral 172Yb{}^{172}\mathrm{Yb}, with the corresponding values being summarized in Table V.

IonGround stateDHF0\mathrm{DHF}_{0}MSFNSBreitωTP\omega\mathrm{TP}QEDSDcHOcTotal
172Yb42+{}^{172}\mathrm{Yb}^{42+}[Ar]3d10 1S03d^{10}\ {}^{1}S_{0}351379.33−0.87-0.87 (3)−45.06-45.06 (8)−367.66-367.666.65249.4 (3.5)48.76 (1)1.4 (1.4)350773 (5)
172Yb2+{}^{172}\mathrm{Yb}^{2+}[Xe]4f14 1S04f^{14}\ {}^{1}S_{0}382828.68−0.90-0.90 (3)−45.31-45.31 (8)−384.95-384.956.79251.1 (3.5)117.26 (9)12 (12)382283 (16)

Table V. Different contributions to the total binding energies of 172Yb42+{}^{172}\mathrm{Yb}^{42+} and 172Yb2+{}^{172}\mathrm{Yb}^{2+}: DHF0\mathrm{DHF}_{0}, the DHF energy assuming a point-like nuclear charge; MS, the mass shift; FNS, the finite nuclear size effect; Breit, the frequency-independent transverse photon interaction; ωTP\omega\mathrm{TP}, the frequency-dependent transverse photon interaction; QED, the QED contribution based on screened-hydrogenic model; SDc, the correlation energies arising from single and double electron substitutions; and HOc, the systematic effect from all unaccounted correlation effects. The values of the DHF0\mathrm{DHF}_{0}, Breit, ωTP\omega\mathrm{TP}, and SDc terms are basis dependent. After taking into account all correlation effects, the basis dependency will be resolved. Thus, their uncertainties are accounted for as a whole in the high-order correlation energy (HOc). The total binding energies are rounded up to integer numbers. All entries are shown in units of eV.

Furthermore, the QED effects are implemented in the GRASP2018 package via a screened-hydrogen-like model for vacuum polarization (VP) and self-energy (SE) contributions, respectively [76]. With values of −54.61-54.61 eV and 307.45 eV for the VP and SE corrections, respectively, they reduce the binding energy by 252.84 eV for the case of 172Yb42+{}^{172}\mathrm{Yb}^{42+}. By comparing these values with respect to the results obtained from a hydrogen-like model without screening, we find that the many-electron QED effects contribute of the order of −15.65-15.65 eV. The uncertainty of these QED calculations can be inferred from ab initio QED calculations in Be-like Xe50+\mathrm{Xe}^{50+} and U88+\mathrm{U}^{88+} with a sub-eV uncertainty [78, 79]. We find that the results obtained by the GRASP2018 code equal 99.03 eV and 625.02 eV, respectively, for the two Be-like ions, and they are larger by 0.85 eV and 8.61 eV, respectively, compared to the ab initio QED results [78, 79]. Assuming a similar systematic error, we obtain a QED contribution of −249.4-249.4 (3.5) eV to the total binding energy of 172Yb42+{}^{172}\mathrm{Yb}^{42+}. However, since QED effects are significant mainly for inner-shell electrons, a similar QED correction of −251.1-251.1 (3.5) eV is obtained for the case of 176Yb2+{}^{176}\mathrm{Yb}^{2+}.

To derive the electron correlation energy, we systematically expand the size of the CSF basis set by allowing single and double (SD) substitutions of electrons from the occupied orbitals of the ground-state configuration to the systematically increasing set of correlation orbitals. These correlation orbitals are added and optimized layer-by-layer [80] up to n=10n = 10 (nn is the principle quantum number), with the highest orbital angular momentum in each layer equals n−1n - 1. It is found that the increment of the correlation energy at each layer decreases exponentially as a function of nn [81]. This allows us to extrapolate the electron correlation energy to n=∞n = \infty. With this scheme, we obtain an SD correlation energy of 48.76 (1) eV and 117.26 (9) eV for 172Yb42+{}^{172}\mathrm{Yb}^{42+} and 172Yb2+{}^{172}\mathrm{Yb}^{2+}, respectively. The uncertainties are derived from extrapolations based on different numbers of data points.

Nevertheless, the full binding energy contains correlations from electron exchanges beyond the SD substitution scheme, as well as from basis functions outside the model space. These contributions are difficult to evaluate but can be estimated from the ionization potentials (IPs) of low-charged ions. We found that, based on SD substitutions from the 4s orbital, the calculated IPs for Yb, Yb+^{+}, and Yb2+^{2+} are 0.80 eV, 0.40 eV, and 0.65 eV smaller than the experimental values (we note that the SD substitutions from orbitals below the 4s orbital generate more than 5 million CSFs for Yb+^{+} and Yb3+^{3+}, which is beyond the capacity of available computing power at our disposal). As a conservative estimation (taken here as the difference between calculated and experimental values, as discussed above), one can assume that the average shift of the calculated IPs for ions from Yb2+^{2+} to Yb3+^{3+} will not be larger than 0.8 eV. To cover this effect, we add a shift of 8.8 (8.8) eV to the correlation-energy difference between Yb2+^{2+} and Yb24+^{24+}. Since this shift is obtained based on the SD substitution from the 4s orbital, for the total-binding-energy calculations based on SD substitution from the 1s orbital, one has to add another 1.0 (1.0) eV shift to account for the contributions from core-core and core-valence correlations. Furthermore, the average shifts due to electron correlations in the theoretical IPs for ions between Yb24+^{24+} and Yb41+^{41+} can be constrained based on the calculations of Pd-like Xe8+^{8+} and Cu-like Kr7+^{7+}, respectively. Since the QED effects are negligible for the IPs of Xe8+^{8+} and Kr7+^{7+}, the discrepancy between experimental and calculated values arises from the unaccounted correlation effects. With SD substitutions from the 1s orbital, we arrive at an upper limit of 0.1 eV [61] for the average shift in IPs. Thus, a shift of 0.9 (9) eV has to be added to the correlation-energy difference between Yb24+^{24+} and Yb42+^{42+}. To be conservative, we also assume an average of 0.1 eV correlation-energy shift in the IPs for ions with charge states higher than Yb42+^{42+}. In total, the unaccounted high-order correlation effects are derived to be 1.4 (1.4) eV and 12 (12) eV for Yb42+^{42+} and Yb2+^{2+}, respectively.

Therefore, electron correlation effects contribute 50.2 (1.4) eV and 129 (12) eV to the total binding energy of Yb42+^{42+} and Yb2+^{2+}, respectively. Altogether, we obtain E172(28)=350,773(5)E_{172}^{(28)} = 350,773 (5) eV and E172(68)=382,283(16)E_{172}^{(68)} = 382,283 (16) eV, where we have rounded up the numbers to integer values of eV. Adding up the IPs of the outermost two electrons, one arrives at E172(70)=382,301(16)E_{172}^{(70)} = 382,301 (16) eV for the total binding energy of the neutral atom. These results agree with the values obtained from the NIST atomic database but with the uncertainties reduced by more than 2 orders of magnitude.

Since each Yb isotope bears a different mass and nuclear size, its total binding energy entails isotope shifts arising from different nuclear recoil and field effects. To investigate the isotope dependence of ΔEA(28)\Delta E_{\mathrm{A}}^{(28)}, see Eq. (3), we start with a Dirac-Hartree-Fock (DHF) and RCI calculation where only the ground state configuration is considered for each isotope and charge state. With 172Yb42+{}^{172}\mathrm{Yb}^{42+} as a reference, we obtain ΔEA(28)=−0.50(1)\Delta E_{\mathrm{A}}^{(28)} = -0.50 (1) eV, −0.25(1)-0.25 (1) eV, 0.25 (1) eV, and 0.50 (1) eV for isotopes A=168A = 168, 170, 174 and 176, respectively.

Experimental details for frequency measurements

We determined the isotope shifts from absolute frequencies να,γA\nu_{\alpha,\gamma}^{A} of the 2S1/2→2D5/2{}^{2}S_{1/2} \rightarrow{}^{2}D_{5/2} electric quadrupole (α\alpha) and the 2S1/2→2F7/2{}^{2}S_{1/2} \rightarrow{}^{2}F_{7/2} electric octupole (γ\gamma) transitions in the stable even isotopes of AYb+{}^{A}\mathrm{Yb}^{+} (A∈{168,170,172,174,176}A \in\{168,170,172,174,176\}). For this, we measured the optical frequency ratios Rα,γ=να,γ171/νγ171R_{\alpha,\gamma} = \nu_{\alpha,\gamma}^{171}/\nu_{\gamma}^{171} with νγ171\nu_{\gamma}^{171} corresponding to the frequency of the γ\gamma transition between the F=0F = 0 and F=3F = 3 hyperfine states of the 171Yb+{}^{171}\mathrm{Yb}^{+} isotope [28]. We determined the absolute frequencies by multiplying RR to the recommended frequency value νγ171=642,121,491,966,772,645.12(12)\nu_{\gamma}^{171} = 642,121,491,966,772,645.12 (12) Hz [82] and correcting for systematic frequency shifts.

The experimental setup is sketched in Figure 5. Two separate experiments are involved in the R\mathcal{R} measurements, one is based on a segmented linear radio-frequency (rf) Paul trap [25, 26], which is used to trap even isotopes of Yb+^{+}, and the other one is based on a ring style rf Paul trap [83] trapping 171^{171}Yb+^{+}. Two stable laser systems are employed for probing the α\alpha and γ\gamma transitions. A laser at 822 nm is frequency-doubled to interrogate the α\alpha transition near 411 nm in the even isotopes. This laser is stabilized to an ultra-low expansion (ULE) optical cavity, obtaining short-term frequency stability of 5×10−165 \times10^{-16} at 10 s averaging time [84]. The larger frequency shifts between the isotopes are bridged with an electro-optic modulator (EOM) in the infrared, while smaller frequency shifts are applied with a double-pass acousto-optic modulator (AOM) in the blue.

Simplified schematic representation of the experimental setup for the interrogation of the alpha and gamma transitions

Figure 5. Simplified schematic representation of the experimental setup used for the interrogation of the α\alpha and γ\gamma transitions near 411 nm and 467 nm, respectively. Even isotopes of Yb+^{+} are trapped in a segmented linear radio-frequency (rf) Paul trap [25, 26] and the odd natural 171^{171}Yb+^{+} is trapped in a ring style rf Paul trap [83]. Probe lasers in the infrared are referenced to ultra-low expansion (ULE) optical cavities for short-term stability, then further stabilized to a cryogenic silicon cavity [27]. Electro-optic modulators (EOMs) are used to bridge the isotope frequency differences and acousto-optic modulators (AOMs) are used to fine-tune the laser frequencies. A frequency comb setup is used to perform the optical frequency ratio measurements.

A stable laser at 934 nm is frequency-doubled to 467 nm and continuously steered to be on resonance with the γ\gamma transition of the 171^{171}Yb+^{+} isotope [28]. Part of the infrared light is frequency offset via an EOM and used to stabilize the frequency of a second laser at 934 nm. The frequency-doubled output of this laser is used for the interrogation of the γ\gamma transitions in the even isotopes. In both setups, the short-term frequency stability of the lasers is improved by locking to a cryogenic silicon cavity [27]. A frequency comb setup is used to perform the optical frequency ratio measurements.

In the segmented, linear rf Paul trap setup used for the even isotopes, the quantization axis is defined by a static magnetic field of about 65 μT\mu\mathrm{T} oriented in the xzxz-plane at an angle of β=26.8(4.0)∘\beta= 26.8(4.0)^\circ with respect to the trap axis zz (shown in Figure 1a in the main text), obtained from geometric measurements. The lasers to interrogate the α\alpha and the γ\gamma transitions are aligned in the radial trapping direction yy. Since the electronic states of the even Yb+^{+} isotopes do not contain sublevels with zero magnetic moment m=0m = 0, the α\alpha and γ\gamma transitions are first-order magnetic-field sensitive. We employ active magnetic field stabilization [85, 86] to increase the maximum coherent interrogation time for these transitions.

AOMs are used to address pairs of Zeeman components with opposite magnetic field sensitivity and the center transition frequencies are obtained via averaging in post-processing. For the α\alpha and the γ\gamma transitions, the ∣2S1/2,mJ=±1/2⟩→∣2D5/2,mJ=±5/2⟩|^{2}S_{1/2},m_J = \pm1/2\rangle\rightarrow|^{2}D_{5/2},m_J = \pm5/2\rangle and the ∣2S1/2,mJ=±1/2⟩→∣2F7/2,mJ=±1/2⟩|^{2}S_{1/2},m_J = \pm1/2\rangle\rightarrow|^{2}F_{7/2},m_J = \pm1/2\rangle transitions are interrogated, respectively.

The basic interrogation sequence starts with 5 ms of laser cooling via the 2S1/2→2P1/2^{2}S_{1/2} \rightarrow{}^{2}P_{1/2} electric dipole transition near 370 nm, assisted by laser light near 935 nm for repumping via the 2D3/2→3[3/2]1/2^{2}D_{3/2} \rightarrow{}^{3}[3/2]_{1/2} transition. It is followed by state preparation to either of the ∣2S1/2,mJ=±1/2⟩|^{2}S_{1/2},m_J = \pm1/2\rangle Zeeman sublevels via the ∣2S1/2,mJ=∓1/2⟩→∣2D5/2,mJ=±3/2⟩|^{2}S_{1/2},m_J = \mp1/2\rangle\rightarrow|^{2}D_{5/2},m_J = \pm3/2\rangle transitions, assisted by the 2D5/2→2P3/2^{2}D_{5/2} \rightarrow{}^{2}P_{3/2} transition near 1650 nm [87]. A Rabi pulse of tπ=2.1t_{\pi} = 2.1 ms (tπ=π/ΩRt_{\pi} = \pi/\Omega_R, where ΩR\Omega_R is the Rabi frequency) or of about 50 ms is employed to interrogate the α\alpha or the γ\gamma transition. The maximum excitation probability is Pα=45%P_{\alpha} = 45\% and Pγ=80%P_{\gamma} = 80\% for the α\alpha and the γ\gamma transitions.

After this, laser light near 370 nm is applied for 2.5 ms for state-selective fluorescence detection. Successful excitation results in the absence of fluorescence. After state detection, repumpers near wavelengths of 638 nm and 1650 nm for the 2F7/2→1[5/2]5/2^{2}F_{7/2} \rightarrow{}^{1}[5/2]_{5/2} and the 2D5/2→2P3/2^{2}D_{5/2} \rightarrow{}^{2}P_{3/2} transitions, respectively, are switched on to return the ion to the cooling cycle. A valid experimental cycle requires the detection of fluorescence after repumping. This interrogation sequence is continuously repeated with 50 cycles at each point of a four-point servo lock. Here, the servo points run over fixed positive and negative detunings from each component of the Zeeman pair, as illustrated in Figure 6. After interrogating at all four points, the mean AOM frequency is updated to obtain the same excitation probability for equal positive and negative detunings. The γ\gamma transition is additionally interrogated at a lower laser intensity using an independent servo. From the high- and low-intensity servos, the transition frequency that is free from probe light-induced ac-Stark shift is derived via linear extrapolation.

Schematic representation of the four-point servo lock

Figure 6. Schematic representation of the four-point servo lock used to determine the transition frequencies. The solid (red) and dashed (blue) curves correspond to the spectra of the ΔmJ,+:∣2S1/2,mJ=∓1/2⟩→∣2D5/2,mJ=±5/2⟩\Delta m_{J,+}: |^{2}S_{1/2},m_J = \mp1/2\rangle\rightarrow|^{2}D_{5/2},m_J = \pm5/2\rangle or the ΔmJ,+:∣2S1/2,mJ=±1/2⟩→∣2F7/2,mJ=±1/2⟩\Delta m_{J,+}: |^{2}S_{1/2},m_J = \pm1/2\rangle\rightarrow|^{2}F_{7/2},m_J = \pm1/2\rangle transitions, respectively. The numbers denote the order of the interrogation frequencies. “Servo A” and “Servo B” are performed in an interleaved manner.

Systematic and statistical uncertainties for the frequency measurements

For each of the two transitions in the even isotopes, one systematic frequency shift dominates and causes most of the systematic uncertainty. The electric quadrupole shift stemming from the interaction of the quadrupole moment with the electric field gradient at the ion position dominates for the α\alpha transition. The electric quadrupole shift can be calculated from the electric field gradient, the electric quadrupole moment, and the angle β\beta between the symmetry axis of the gradient and the applied magnetic field. The electric field gradient is determined to be dEz/dz=3.1610 (15) V/mm2dE_z/dz = 3.1610\ (15)\ \mathrm{V/mm^2} from an axial secular frequency of νax=211.97 (5) kHz\nu_{\mathrm{ax}} = 211.97\ (5)\ \mathrm{kHz}. A potential electric field gradient of 0.2 V/mm20.2\ \mathrm{V/mm^2} from stray electric fields is added in quadrature to the uncertainty of dEz/dzdE_z/dz. With the electric quadrupole moment of 12.5 (4) a0212.5\ (4)\ a_0^2 [88], we obtain an electric quadrupole shift of −9.3 (1.8) Hz-9.3\ (1.8)\ \mathrm{Hz}. For the γ\gamma transition, this systematic shift is negligible due to a more than 400 times smaller quadrupole moment [89].

The dominant systematic effect for the γ\gamma transition is the probe-light induced ac-Stark shift. Even though Rabi excitation at two different laser intensities is employed to determine the transition frequency free from the ac-Stark shift, imperfect light path design caused the beam pointing to change with the rf power applied to the AOM. This shift effect is estimated by comparing the slope of the extrapolation to zero ac-Stark shift during the measurement campaign to that of a measurement where the change in the probe light power is realized independent of the AOM power. From the comparison, a maximum shift of −9.2 (11.1) Hz-9.2\ (11.1)\ \mathrm{Hz} is determined. In contrast, for the α\alpha transition, the probe-light induced ac-Stark shift even with a shorter pulse duration of tπ=2.1 mst_\pi= 2.1\ \mathrm{ms} is calculated from the differential polarizability in Ref. [85] and causes a negligible shift of only 2 (1) mHz2\ (1)\ \mathrm{mHz}.

In addition to the major systematic shifts discussed so far, Table VI includes smaller shift effects and their uncertainties. These shift effects will be discussed in the following.

Systematic shiftδνα\delta\nu_{\alpha} [Hz]uB,αu_{B,\alpha} [Hz]δνγ\delta\nu_{\gamma} [Hz]uB,γu_{B,\gamma} [Hz]
Electric quadrupole shift−9.31.8−0.0160.003
Probe-light induced ac-Stark shift0.0020.001−9.211.1
Second-order Doppler shift−0.0020.004−0.0020.004
Second-order Stark shift−0.0150.047−0.0030.010
Blackbody radiation shift−0.260.13−0.0450.001
Servo error induced shift–0.8–0.1
Total−9.62.0−9.311.1

Table VI. Leading systematic shifts δνα\delta\nu_{\alpha}, δνγ\delta\nu_{\gamma} and corresponding uncertainties uB,αu_{B,\alpha}, uB,γu_{B,\gamma} of the 2S1/2→2D5/2{}^{2}S_{1/2}\to{}^{2}D_{5/2} electric quadrupole (α\alpha) and the 2S1/2→2F7/2{}^{2}S_{1/2}\to{}^{2}F_{7/2} electric octupole (γ\gamma) transition for all even isotopes.

The ion is confined with radial secular frequencies of about 600 kHz and an axial secular frequency of around 200 kHz and laser-cooled to the Lamb-Dicke regime reaching a temperature of 0.60 (6) mK0.60\ (6)\ \mathrm{mK}. Minimization and measurements of the excess micromotion (EMM) using the photon correlation technique [90] were performed only before and after the measurement campaign. Based on these measurements, we estimate the mean rf field causing EMM to be {Ex,Ey,Ez}={100 (500),100 (500),150 (50)} V/m\{E_x,E_y,E_z\} = \{100\ (500),100\ (500),150\ (50)\}\ \mathrm{V/m}. Here, the uncertainties are set to the maximum variations seen over the three months prior to the measurements. Both effects from uncompensated EMM and intrinsic micromotion due to the finite ion temperature are considered to calculate the second-order Doppler shift and the second-order Stark shift. The corresponding second-order Doppler shift is calculated as −0.002 (4) Hz-0.002\ (4)\ \mathrm{Hz} and the second-order Stark shift is −0.015 (47) Hz-0.015\ (47)\ \mathrm{Hz} and −0.003 (10) Hz-0.003\ (10)\ \mathrm{Hz} for the α\alpha and the γ\gamma transition, respectively. The involved static differential polarizability for the α\alpha transition is found in Ref. [85] and for the γ\gamma transition in Ref. [91].

The blackbody radiation shift is calculated to be −0.26 (13) Hz-0.26\ (13)\ \mathrm{Hz} for the α\alpha transition and −0.046 (1) Hz-0.046\ (1)\ \mathrm{Hz} for the γ\gamma transition considering thermal radiation with an effective temperature of TC=297.1 (1.4) KT_C = 297.1\ (1.4)\ \mathrm{K} perturbing the ion [85, 91]. The temperature uncertainty assumed the maximum temperature difference measured at the chamber and the ion trap during the measurement campaign.

Magnetic field variations can cause servo errors because of the long dead time and non-randomized four-point servo lock sequence [92, 93]. These are estimated to be 0.8 Hz and 0.1 Hz for the measurements of the α\alpha and the γ\gamma transitions. By adding the uncertainties to the systematic shifts in quadrature, we determine the total systematic uncertainty for the α\alpha and the γ\gamma transition frequencies to be uB,α=2.0 Hzu_{B,\alpha} = 2.0\ \mathrm{Hz} and uB,γ=11.1 Hzu_{B,\gamma} = 11.1\ \mathrm{Hz}, respectively.

The systematic uncertainty for the realization of the γ\gamma transition in 171Yb+^{171}\mathrm{Yb}^{+} has been evaluated to 1.7 mHz [94]. For the frequency ratio measurements, a gravitational redshift between the experimental setups for the even isotopes of Yb+\mathrm{Yb}^{+} and the 171Yb+^{171}\mathrm{Yb}^{+} of −3.26 (4)×10−17-3.26\ (4) \times10^{-17} has been taken into account.

The statistical uncertainties uAu_A from the measurements are listed in Table VII. The values correspond to the last point in the Allan deviation σy\sigma_y of the transition frequency measurements. Examples are shown in Figure 7 for Rα,γR_{\alpha,\gamma} of the α\alpha and the γ\gamma transitions in 168Yb+^{168}\mathrm{Yb}^{+}. The statistical uncertainty is smaller for the γ\gamma transition because we operate at a different resolved linewidth. With pulsed Rabi spectroscopy, we resolve the α\alpha transition with good contrast at a linewidth of about 300 Hz, limited by the natural decay, while the resolved linewidth of the resonance of the γ\gamma transition is below 20 Hz in these measurements. The fit to the Allan deviation for the α\alpha transition frequency measurement yields σfit,α=101.7 (1.5) Hz/τ\sigma_{\mathrm{fit},\alpha} = 101.7\ (1.5)\ \mathrm{Hz}/\sqrt{\tau}, and that of the Allan deviation for the γ\gamma transition is σfit,γ=13.36 (7) Hz/τ\sigma_{\mathrm{fit},\gamma} = 13.36\ (7)\ \mathrm{Hz}/\sqrt{\tau}. Both of the fits agree with the instability expected from quantum projection noise.

Isotope AAuA,αu_{A,\alpha} [Hz]uA,γu_{A,\gamma} [Hz]
1682.70.4
1703.60.2
1721.80.3
1743.50.3
1761.30.3

Table VII. Statistical uncertainties uA,αu_{A,\alpha}, uA,γu_{A,\gamma} for determining the α\alpha and the γ\gamma transition frequencies.

Allan deviations of the alpha and gamma transition frequency measurements

Figure 7. Allan deviations σy\sigma_y of the α\alpha (a) and the γ\gamma (b) transition frequency measurements in 168^{168}Yb+^{+}. The fits to the data yield σfit,α=101.7 (1.5) Hz/τ\sigma_{\mathrm{fit},\alpha}=101.7\,(1.5)\,\mathrm{Hz}/\sqrt{\tau} and σfit,γ=13.36 (7) Hz/τ\sigma_{\mathrm{fit},\gamma}=13.36\,(7)\,\mathrm{Hz}/\sqrt{\tau}, agreeing with the instabilities expected from quantum projection noise.

The total uncertainties of the transition frequencies for each isotope are given by σνk=uA,k2+uB,k2\sigma_{\nu_k}=\sqrt{u_{A,k}^{2}+u_{B,k}^{2}}, with k=α,γk=\alpha,\gamma. We determine the absolute frequencies of the α\alpha and the γ\gamma transitions for all five stable even isotopes of Yb+^{+} with the frequency ratios, the literature value for the absolute frequency of the γ\gamma transition in 171^{171}Yb+^{+} [82], and the systematic shifts. Values are shown in Table VIII. For determining the isotope shifts, we do not consider any common mode rejection and take into account the full uncertainties as given in Table VIII. The deviation δν=νTW−νRef.\delta\nu=\nu_{\mathrm{TW}}-\nu_{\mathrm{Ref.}} of the isotope shifts determined in this work (TW) and in prior publications are given in Table 1 in the main text for the α\alpha transition [5], and in Table IX for the γ\gamma transition [7].

AARCFA=νc,172/νc,AR_{\mathrm{CF}}^{A} = \nu_{\mathrm{c},172}/\nu_{\mathrm{c},A}ηA=mA/m172\eta_A = m_A/m_{172}ΔηA\Delta\eta_A [10−12][10^{-12}]ναPTBA,A+2\nu_{\alpha\mathrm{PTB}}^{A,A+2} [Hz]ΔναA,A+2\Delta\nu_{\alpha}^{A,A+2} [Hz]νγPTBA,A+2\nu_{\gamma\mathrm{PTB}}^{A,A+2} [Hz]
1680.976 717 951 145 (4)0.976 715 921 749 (4)−1890-1890 (780) [8]2179098868.0 (5.3)−62-62 (210) [5]−4438159671.1-4438159671.1 (15.7)
1700.988 356 814 144 (4)0.988 355 799 258 (4)−88-88 (108) [9]2044851281.0 (4.9)−3499-3499 (340) [5]−4149190501.1-4149190501.1 (15.7)
172–––1583064149.3 (4.8)−4271-4271 (360) [5]−3132320458.1-3132320458.1 (15.7)
1741.011 648 196 817 (4)1.011 649 212 140 (4)153 (122) [9]1509053195.8 (4.7)−2094-2094 (280) [5]−2976320458.3-2976320458.3 (15.7)
1761.023 303 526 697 (4)1.023 305 557 965 (4)68 (173) [9]–––

Table 1. Measured values of the mass ratios and isotope shifts. Columns 2 and 3 show the Yb42+^{42+} cyclotron frequency ratios and nuclear mass ratios of the stable, even ytterbium isotopes relative to the nuclear mass of isotope A=172A = 172, with the differences to Refs. [8, 9] given in Column 4. Columns 5 and 7 show the isotope shifts ναA,A+2=ναA−ναA+2\nu_{\alpha}^{A,A+2} = \nu_{\alpha}^{A} - \nu_{\alpha}^{A+2} of the α\alpha transition and the γ\gamma transition in units of Hz, with the differences to Ref. [5] for α\alpha given in Column 6. Our isotope shifts for the γ\gamma transition are compatible with those of Ref. [7].

Isotope AAνα−729 THz\nu_{\alpha}-729\,\mathrm{THz} [Hz]νγ−642 THz\nu_{\gamma}-642\,\mathrm{THz} [Hz]
168481 090 977 315.9 (3.4)108 197 800 720.0 (11.1)
170478 911 878 447.9 (4.1)112 635 960 391.1 (11.1)
172476 867 027 166.9 (2.7)116 785 150 892.3 (11.1)
174475 283 963 017.6 (4.0)119 917 471 350.4 (11.1)
176473 774 909 821.9 (2.4)122 893 863 395.6 (11.1)

Table VIII. Absolute frequencies να,γ\nu_{\alpha,\gamma} of the α\alpha and the γ\gamma transitions for all five stable even isotopes of Yb+^{+}.

AAΔνγA,A+2\Delta\nu_{\gamma}^{A,A+2} [Hz]
168−629 (500) [7]
170−121 (450) [7]
172−1142 (500) [7]
174−445 (480) [7]

Table IX. Differences of the measured γ\gamma transitions in this work compared to Ref. [7] in units of Hz.

King-plot analysis

From the King plot shown in Figure 1 b,Figure 1c in the main text, we can extract the slope and offset, which correspond to the constants FγαF_{\gamma\alpha} and KγαK_{\gamma\alpha} in

ν~γ≈Fγαν~α+Kγα1,(4)\tilde{\nu}_{\gamma}\approx F_{\gamma\alpha}\tilde{\nu}_{\alpha}+K_{\gamma\alpha}1, \tag*{(4)}

which result in −2.22131 (5) and 0.033363 (8) Hz, respectively. We also resolve deviations from linearity on the average of 20.17 (2) kHz. From the extracted slope, the nuclear mass-ratio uncertainties can be translated into frequency uncertainties. Here, the maximum translated frequency uncertainty is around 0.3 Hz, stemming from the (168, 170) isotope pair, improving on the 28 Hz uncertainty in Ref. [7].

Nonlinearity decomposition

Given nn isotope pairs, n−2n-2 King-plot nonlinearities can be resolved by projecting the isotope shift data on nn nn-vectors. For simplicity, we use the same reference transition δ\delta for all King plots entering our nonlinearity analysis and choose our basis in such a way that two basis vectors, ν~δ\tilde{\nu}_{\delta} and 1\mathbf{1} [see Eq. (4) with α→δ\alpha\to\delta], lie within the plane of King linearity, whereas the other two, Λ+∝(ν~δc−ν~δb,ν~δa−ν~δd,ν~δd−ν~δa,ν~δb−ν~δc)\Lambda_{+} \propto(\tilde{\nu}_{\delta}^{c}-\tilde{\nu}_{\delta}^{b},\tilde{\nu}_{\delta}^{a}-\tilde{\nu}_{\delta}^{d},\tilde{\nu}_{\delta}^{d}-\tilde{\nu}_{\delta}^{a},\tilde{\nu}_{\delta}^{b}-\tilde{\nu}_{\delta}^{c}) and Λ−∝(ν~δd−ν~δb,ν~δa−ν~δc,ν~δb−ν~δd,ν~δc−ν~δa)\Lambda_{-} \propto(\tilde{\nu}_{\delta}^{d}-\tilde{\nu}_{\delta}^{b},\tilde{\nu}_{\delta}^{a}-\tilde{\nu}_{\delta}^{c},\tilde{\nu}_{\delta}^{b}-\tilde{\nu}_{\delta}^{d},\tilde{\nu}_{\delta}^{c}-\tilde{\nu}_{\delta}^{a}), normalized to 1 Hz and with a=(168,170)a=(168,170), b=(170,172)b=(170,172), c=(172,174)c=(172,174), d=(174,176)d=(174,176), are orthogonal to this plane and directly calculable from the experimental data. In this way, we can use the same basis for the nonlinearity decomposition of the experimental data and for the ab initio predictions. The coordinates (λ+(τ),λ−(τ))(\lambda_{+}^{(\tau)},\lambda_{-}^{(\tau)}), τ∈{αPTB,αMIT,β,γPTB,γMIT,δ,ϵ}\tau\in\{\alpha_{\mathrm{PTB}},\alpha_{\mathrm{MIT}},\beta,\gamma_{\mathrm{PTB}},\gamma_{\mathrm{MIT}},\delta,\epsilon\}, defined via

ν~τ=Fτδν~δ+Kτδ1+λ+(τ)Λ++λ−(τ)Λ−(5)\tilde{\nu}_{\tau}=F_{\tau\delta}\tilde{\nu}_{\delta}+K_{\tau\delta}\mathbf{1}+\lambda_{+}^{(\tau)}\Lambda_{+}+\lambda_{-}^{(\tau)}\Lambda_{-} \tag*{(5)}

characterize the nonlinearities in the isotope shift ντ\nu_{\tau}.

In the presence of one dominant source of King nonlinearity, the electronic coefficients G(2)G^{(2)}, G(4)G^{(4)}, (αNP/αEM)D(\alpha_{\mathrm{NP}}/\alpha_{\mathrm{EM}})D, . . . [see Eq. (6)] drop out of the ratio λ−(τ)/λ+(τ)=λ−/λ+\lambda_{-}^{(\tau)}/\lambda_{+}^{(\tau)}=\lambda_{-}/\lambda_{+}, making λ−/λ+\lambda_{-}/\lambda_{+} a transition-independent quantity. This motivates a linear fit through the origin of the (λ+,λ−)(\lambda_{+},\lambda_{-}) plane to the six data points (λ+(τ),λ−(τ))(\lambda_{+}^{(\tau)},\lambda_{-}^{(\tau)}) derived for the six transitions listed in Table X. Indeed, the tension of the Yb isotope-shift data for King linearity can be drastically reduced by introducing a King nonlinearity satisfying λ−/λ+=−1.777(14)\lambda_{-}/\lambda_{+}=-1.777(14), leaving a residual nonlinearity at the level of 23 σ23\,\sigma.

NotationTransitionλ\lambda [nm]Refs.
αMIT,αPTB\alpha_{\mathrm{MIT}},\alpha_{\mathrm{PTB}}2S1/2→2D5/2{}^{2}S_{1/2} \to{}^{2}D_{5/2} E2 in Yb+^{+}411[5, 7], t.w.
β\beta2S1/2→2D3/2{}^{2}S_{1/2} \to{}^{2}D_{3/2} E2 in Yb+^{+}435[5]
γMIT,γPTB\gamma_{\mathrm{MIT}},\gamma_{\mathrm{PTB}}2S1/2→2F7/2{}^{2}S_{1/2} \to{}^{2}F_{7/2} E3 in Yb+^{+}467[5, 7], t.w.
δ\delta1S0→3P0{}^{1}S_{0} \to{}^{3}P_{0} in Yb578[29]
ϵ\epsilon1S0→1D2{}^{1}S_{0} \to{}^{1}D_{2} in Yb361[30]

Table X. Notation and references for the transitions used in the King-plot analysis and the nonlinearity decomposition plot. “t.w.” abbreviates “this work”.

We compare this slope to predictions of individual nonlinearity terms in

να=Fαδδ⟨r2⟩+Kαw+Gα(2)δ⟨r2⟩2+Gα(4)δ⟨r4⟩+αNPαEMDαh+…(6)\nu_{\alpha}=F_{\alpha\delta}\delta\langle r^{2}\rangle+K_{\alpha}w+G_{\alpha}^{(2)}\delta\langle r^{2}\rangle^{2}+G_{\alpha}^{(4)}\delta\langle r^{4}\rangle+\frac{\alpha_{\mathrm{NP}}}{\alpha_{\mathrm{EM}}}D_{\alpha}h+\ldots \tag*{(6)}

to identify the leading King-plot nonlinearity. The new physics prediction for λ−/λ+\lambda_{-}/\lambda_{+} follows straightforwardly from Eqs. (6) and (5), and has a relative uncertainty of approximately 3×10−73\times10^{-7}. The λ−/λ+\lambda_{-}/\lambda_{+} slope corresponding to the quadratic field shift δ⟨r2⟩2\delta\langle r^{2}\rangle^{2} [see Eq. (6)], dotted line in Figure 2a in the main text) can be derived from experimental data for δ⟨r2⟩\delta\langle r^{2}\rangle [31] with a relative uncertainty of approximately 2%. If the new physics (quadratic field shift) term were to dominate the nonlinearity, the experimental data points would be expected to lie on the dash-dotted (dotted) line in Figure 2a in the main text. We conclude that neither the new physics term nor the quadratic field shift term can be the dominant nonlinearity in the ytterbium King plot (Figure 1b,Figure 1c in the main text).

Our ab initio calculations of δ⟨r4⟩\delta\langle r^{4}\rangle fit better to data: In Figure 2a in the main text we show the projections of our chiral effective field theory calculations using the 1.8/2.0 (EM) interactions with valence spaces VS1 (dashed, blue) and VS2 (dashed, orange). Both agree within uncertainties (the 68% confidence interval is given by the gray band) with the linear fit to data, suggesting that δ⟨r4⟩\delta\langle r^{4}\rangle can explain the leading King nonlinearity in ytterbium. Our ab initio calculations are discussed in more detail in Section H.

Ab initio nuclear structure calculations

The ground-state properties of atomic nuclei, such as ⟨r2⟩\langle r^{2}\rangle and ⟨r4⟩\langle r^{4}\rangle, can be computed ab initio using effective field theories for the strong interactions between nucleons and systematically improvable many-body methods to solve the many-body Schrödinger equation. Yb isotopes are challenging to describe ab initio due to being heavy and open-shell, and our work leverages recent developments improving the treatment of three-nucleon forces in heavy nuclei [15] and performing large-scale diagonalizations of the many-body Hamiltonian [32].

Our calculations are performed with the valence-space IMSRG [12, 13, 95], which solves the many-body Schrödinger equation for a given input Hamiltonian HH via a unitary transformation to a block-diagonal form, Hˉ=UHU†\bar{H}=UHU^{\dagger}. The unitary transformation U=eΩU=e^{\Omega} is formulated with respect to a reference state ∣Φ0⟩|\Phi_{0}\rangle for the system of interest and can as a result be efficiently approximated at the normal-ordered two-body level, the IMSRG(2). We solve the IMSRG(2) in a model space of 15 oscillator shells based on an underlying harmonic oscillator basis with frequency ℏω=12\hbar\omega=12 MeV, including three-nucleon forces with a truncation of E3max=28E_{\mathrm{3max}}=28 made possible by Ref. [15]. For the transformed Hamiltonian, an effective valence-space Hamiltonian is decoupled and subsequently diagonalized via large-scale shell-model methods.

We employ two nuclear Hamiltonians with two- and three-nucleon interactions derived within chiral effective field theory (EFT). The 1.8/2.0 (EM) Hamiltonian [33] is constructed from the N3LO\mathrm{N}^{3}\mathrm{LO} nucleon-nucleon potential of Entem and Machleidt (EM) [96], transformed using the similarity renormalization group to a resolution scale of λ=1.8 fm−1\lambda= 1.8\,\mathrm{fm}^{-1}, and N2LO\mathrm{N}^{2}\mathrm{LO} three-nucleon interactions with a regulator cutoff Λ3N=2.0 fm−1\Lambda_{3\mathrm{N}} = 2.0\,\mathrm{fm}^{-1}. It is fit to few-body systems (up to 4He{}^{4}\mathrm{He}) and predicts ground-state energies, spectra, and differential radius trends well in medium-mass and heavy nuclei [97, 98, 99, 100]. The ΔN2LOGO\Delta\mathrm{N}^{2}\mathrm{LO}_{\mathrm{GO}} Hamiltonian [34] is constructed at N2LO\mathrm{N}^{2}\mathrm{LO} with explicit inclusion of Δ\Delta isobars in the EFT and is fit to few-body data and nuclear matter properties and optimized to reproduce bulk properties in medium-mass nuclei. Differences in results obtained by the two Hamiltonians reflect the underlying EFT uncertainty for nuclear forces.

We employ two valence spaces: VS1 with a 132Sn{}^{132}\mathrm{Sn} core and an active valence space consisting of 1g7/21g_{7/2}, 2d5/22d_{5/2}, 2d3/22d_{3/2}, 3s1/23s_{1/2}, 1h11/21h_{11/2} proton orbitals and 2f7/22f_{7/2}, 1h9/21h_{9/2}, 1i13/21i_{13/2}, 2f5/22f_{5/2}, 3p3/23p_{3/2}, 3p1/23p_{1/2} neutron orbitals; and VS2 with a 154Gd{}^{154}\mathrm{Gd} core and an active valence space consisting of 2d3/22d_{3/2}, 3s1/23s_{1/2}, 1h11/21h_{11/2} proton orbitals and 1h9/21h_{9/2}, 1i13/21i_{13/2}, 2f5/22f_{5/2}, 3p3/23p_{3/2}, 3p1/23p_{1/2} neutron orbitals. Varying the valence space allows us to assess some of the uncertainty due to the employed many-body approximation. Recent developments have made IMSRG calculations at the normal-ordered three-body level available [35]. These calculations can currently not be converged in Yb, but we performed restricted calculations to estimate the order of magnitude of truncated contributions in the IMSRG, another source of many-body uncertainties.

To compute ⟨r2⟩\langle r^{2}\rangle and ⟨r4⟩\langle r^{4}\rangle, we evaluate the ground-state expectation values of the translationally invariant point-proton radius operators Rp2R_{p}^{2} and Rp4R_{p}^{4}, with the definitions

Rp2=∑i[(1+τi)12Z(1−2A)+1A2]ri2+∑i<j[2A2−2AZ(1+τi+τj2)]ri⋅rj(7)R_{p}^{2} = \sum_i \left[(1+\tau_i)\frac{1}{2Z}\left(1-\frac{2}{A}\right)+\frac{1}{A^{2}}\right]r_i^{2} +\sum_{i<j}\left[\frac{2}{A^{2}}-\frac{2}{AZ}\left(1+\frac{\tau_i+\tau_j}{2}\right)\right]\mathbf{r}_i\cdot\mathbf{r}_j \tag*{(7)}
Rp4=∑i[(1+τi)12Z(1−4A)]ri4+∑i<j−2A2[(1+τi)ri2+(1+τj)rj2]ri⋅rj+O(A−2)(8)R_{p}^{4} = \sum_i \left[(1+\tau_i)\frac{1}{2Z}\left(1-\frac{4}{A}\right)\right]r_i^{4} +\sum_{i<j}\frac{-2}{A^{2}}\left[(1+\tau_i)r_i^{2}+(1+\tau_j)r_j^{2}\right]\mathbf{r}_i\cdot\mathbf{r}_j+\mathcal{O}(A^{-2}) \tag*{(8)}

We neglect the indicated higher-order contributions in Rp4R_{p}^{4}, which would also include three- and four-body parts but are strongly suppressed. For ⟨r2⟩\langle r^{2}\rangle, we also include the spin-orbit [101] and relativistic Darwin-Foldy [102] corrections and account for the finite size of nucleons.

Our computed values of ⟨r2⟩\langle r^{2}\rangle, ⟨r4⟩\langle r^{4}\rangle, δ⟨r2⟩\delta\langle r^{2}\rangle, and δ⟨r4⟩\delta\langle r^{4}\rangle are given in Table XI and Table XII. We note that we obtained nonphysical results for 168Yb{}^{168}\mathrm{Yb} with the ΔN2LOGO\Delta\mathrm{N}^{2}\mathrm{LO}_{\mathrm{GO}} Hamiltonian, with an inversion of proton 2d3/22d_{3/2} and 1h11/21h_{11/2} single-particle orbitals at the Hartree-Fock level relative to 170–176Yb{}^{170\text{--}176}\mathrm{Yb}. For the δ⟨r2⟩\delta\langle r^{2}\rangle and δ⟨r4⟩\delta\langle r^{4}\rangle values in Table XII, we obtain extrapolated values for ΔN2LOGO\Delta\mathrm{N}^{2}\mathrm{LO}_{\mathrm{GO}} based on the differences observed for (172,170)(172,170) and (170,168)(170,168) for the 1.8/2.01.8/2.0 (EM) Hamiltonian in the same valence space.

Isotope AA⟨r2⟩\langle r^2\rangle [fm2^2] 1.8/2.0 (EM) VS11.8/2.0 (EM) VS2Δ\DeltaN2^2LOGO_{\mathrm{GO}} VS1
16824.23324.287(*)
17024.38724.45726.609
17224.53424.62126.711
17424.67424.78126.810
17624.76124.86826.906
Isotope AA⟨r4⟩\langle r^4\rangle [fm4^4] 1.8/2.0 (EM) VS11.8/2.0 (EM) VS2Δ\DeltaN2^2LOGO_{\mathrm{GO}} VS1
168710.4719.5(*)
170719.3728.9850.9
172727.8738.4857.4
174736.2747.8863.6
176743.3755.5869.6

Table XI. Absolute values of ⟨r2⟩\langle r^2\rangle and ⟨r4⟩\langle r^4\rangle for the 1.8/2.0 (EM) Hamiltonian using valence spaces VS1 and VS2 and for the Δ\DeltaN2^2LOGO_{\mathrm{GO}} Hamiltonian using VS1. Missing results for 168^{168}Yb with the Δ\DeltaN2^2LOGO_{\mathrm{GO}} Hamiltonian are discussed in the main text.

Isotope pair A,A′A,A'δ⟨r2⟩(A,A′)\delta\langle r^2\rangle^{(A,A')} [fm2^2] 1.8/2.0 (EM) VS11.8/2.0 (EM) VS2Δ\DeltaN2^2LOGO_{\mathrm{GO}} VS1
(170,168)0.1530.1700.108(*)
(172,170)0.1480.1640.103
(174,172)0.1400.1600.099
(176,174)0.0870.0870.096
Isotope pair A,A′A,A'δ⟨r4⟩(A,A′)\delta\langle r^4\rangle^{(A,A')} [fm4^4] 1.8/2.0 (EM) VS11.8/2.0 (EM) VS2Δ\DeltaN2^2LOGO_{\mathrm{GO}} VS1
(170,168)8.859.486.81(*)
(172,170)8.519.486.47
(174,172)8.359.346.21
(176,174)7.117.706.00

Table XII. δ⟨r2⟩\delta\langle r^2\rangle and δ⟨r4⟩\delta\langle r^4\rangle values for the 1.8/2.0 (EM) Hamiltonian using valence spaces VS1 and VS2 and for the Δ\DeltaN2^2LOGO_{\mathrm{GO}} Hamiltonian using VS1. Extrapolated results (*) for (A,A′)=(170,168)(A,A')=(170,168) for the Δ\DeltaN2^2LOGO_{\mathrm{GO}} Hamiltonian are discussed in the main text.

Our results are compared with experimental measurements of δ⟨r2⟩\delta\langle r^{2}\rangle from Ref. [31] and our extraction of δ⟨r4⟩\delta\langle r^{4}\rangle trends in Figure 10 and Figure 9, respectively.

Experimental fourth nuclear charge-radius moment differences compared with nuclear structure theory predictions

Figure 9. Experimental δ⟨r4⟩A,A−2\delta\langle r^{4}\rangle^{A,A-2} values relative to δ⟨r4⟩176,174=7 fm4\delta\langle r^{4}\rangle^{176,174}=7\ \mathrm{fm}^{4} extracted from isotope shifts from the α\alpha transitions using atomic theory (fiducial, core holes) are compared to nuclear theory predictions from our ab initio calculations [top, 1.8/2.0 (EM), VS1 and VS2; ΔN2LOGO\Delta N^{2} \mathrm{LO}_{\mathrm{GO}}, VS1] and from density functional theory calculations [7] [bottom, SV-min, RD-min, UNEDF1, Fy(Δr\Delta r)] for A∈{170,172,174,176}A\in\{170,172,174,176\}. The gray bands give estimated uncertainties of the theory results.

Comparison of experimental ytterbium isotope-shift values with ab initio and density functional theory predictions

Figure 10. Experimental δ⟨r2⟩A,A−2\delta\langle r^{2}\rangle^{A,A-2} values [31] are compared to nuclear theory predictions from our ab initio calculations [top, 1.8/2.0 (EM), VS1 and VS2; Δ\DeltaN2^{2}LOGO, VS1] and from density functional theory calculations [7] [bottom, SV-min, RD-min, UNEDF1, Fy(Δr\Delta r)] for A∈{170,172,174,176}A \in\{170, 172, 174, 176\}. The gray bands give estimated uncertainties of the theory results.

To assess theoretical uncertainties, we employ a correlated statistical uncertainty model. For the propagation of uncertainties in δ⟨r2⟩\delta\langle r^{2}\rangle and δ⟨r4⟩\delta\langle r^{4}\rangle to King-plot analyses, it is essential to account for correlations across isotope pairs. Errors due to truncations in ab initio nuclear theory are very systematic in nature, producing similar errors in neighboring isotopes. As a result, strong correlations arise, and we attempt to quantify and account for these in the following.

We assume that a prediction for given observable o=(o(170,168),o(172,170),o(174,172),o(176,174))\mathbf{o} = (o^{(170,168)},o^{(172,170)},o^{(174,172)},o^{(176,174)}) (true value unknown) can be understood as

o=o~+εEFT+εVS+εMB,(9)\mathbf{o} = \tilde{\mathbf{o}} + \boldsymbol{\varepsilon}_{\mathrm{EFT}} + \boldsymbol{\varepsilon}_{\mathrm{VS}} + \boldsymbol{\varepsilon}_{\mathrm{MB}}, \tag*{(9)}

where o~\tilde{\mathbf{o}} is the approximate prediction and εEFT\varepsilon_{\mathrm{EFT}}, εVS\varepsilon_{\mathrm{VS}}, and εMB\varepsilon_{\mathrm{MB}} are errors made due the necessary truncations in the EFT for the Hamiltonians, the valence space, and the many-body method, respectively. We choose

o~=0.75 o1.8/2.0  (EM),VS1+0.25 oΔN2LOGO,VS1(10)\tilde{\mathbf{o}} = 0.75\,\mathbf{o}_{1.8/2.0\;(\mathrm{EM}),\mathrm{VS1}} + 0.25\,\mathbf{o}_{\Delta\mathrm{N}^{2}\mathrm{LO}_{\mathrm{GO}},\mathrm{VS1}} \tag*{(10)}

based on the reproduction of experimental δ⟨r2⟩\delta\langle r^{2}\rangle values. Additionally, we model the EFT, valence-space, and many-body errors as random variables distributed as multivariate normal distributions,

εEFT∼N(0,ΣEFT),(11)\boldsymbol{\varepsilon}_{\mathrm{EFT}} \sim\mathcal{N}(\mathbf{0},\Sigma_{\mathrm{EFT}}), \tag*{(11)}
εVS∼N(0,ΣVS),(12)\boldsymbol{\varepsilon}_{\mathrm{VS}} \sim\mathcal{N}(\mathbf{0},\Sigma_{\mathrm{VS}}), \tag*{(12)}
εMB∼N(0,ΣMB).(13)\boldsymbol{\varepsilon}_{\mathrm{MB}} \sim\mathcal{N}(\mathbf{0},\Sigma_{\mathrm{MB}}). \tag*{(13)}

We use full covariance matrices Σ\Sigma to allow us to consider correlations across isotope pairs in our error model, which we discuss below.

We estimate the variances [i.e., σi2=diag⁡(Σi)\sigma_i^{2} = \operatorname{diag}(\Sigma_i)] for our EFT and valence-space errors as

σEFT2=(o1.8/2.0  (EM),VS1−oΔN2LOGO,VS1)2,(14)\sigma_{\mathrm{EFT}}^{2} = \left(\mathbf{o}_{1.8/2.0\;(\mathrm{EM}),\mathrm{VS1}}-\mathbf{o}_{\Delta\mathrm{N}^{2}\mathrm{LO}_{\mathrm{GO}},\mathrm{VS1}}\right)^{2}, \tag*{(14)}
σVS2=(o1.8/2.0  (EM),VS1−o1.8/2.0  (EM),VS2)2.(15)\sigma_{\mathrm{VS}}^{2} = \left(\mathbf{o}_{1.8/2.0\;(\mathrm{EM}),\mathrm{VS1}}-\mathbf{o}_{1.8/2.0\;(\mathrm{EM}),\mathrm{VS2}}\right)^{2}. \tag*{(15)}

To estimate σMB2\sigma_{\mathrm{MB}}^{2}, we performed IMSRG(3)-N7N^{7} calculations for 172Yb{}^{172}\mathrm{Yb} [35]. IMSRG(3) calculations cannot currently be converged in Yb, but by including restricted three-body operators in the calculations, we gain insight into the magnitude of the IMSRG(3) corrections. Moreover, calculations in carbon and calcium have shown that the three-body corrections are strongly correlated across systems (with a correlation coefficient r≥0.99r \geq0.99). We find the IMSRG(3) corrections for ⟨r2⟩\langle r^{2}\rangle and ⟨r4⟩\langle r^{4}\rangle in 172Yb{}^{172}\mathrm{Yb} to be on the order of 0.06 fm20.06\,\mathrm{fm}^{2} and 2 fm42\,\mathrm{fm}^{4}, respectively. Assuming a size extensive scaling of radius corrections and exploiting the strong correlation between neighboring isotopes, we find small many-body uncertainties σMB2\sigma_{\mathrm{MB}}^{2} for the differential quantities δ⟨r2⟩\delta\langle r^{2}\rangle and δ⟨r4⟩\delta\langle r^{4}\rangle. All employed standard deviations are given in Table XIII.

Isotope pair A,A′A,A'δ⟨r2⟩(A,A′)\delta\langle r^2\rangle^{(A,A')} [fm2^2] σEFT(A,A′)\sigma_{\mathrm{EFT}}^{(A,A')}σVS(A,A′)\sigma_{\mathrm{VS}}^{(A,A')}σMB(A,A′)\sigma_{\mathrm{MB}}^{(A,A')}
(170,168)0.0450.0160.005
(172,170)0.0450.0170.005
(174,172)0.0410.0200.005
(176,174)0.0090.0010.005
Isotope pair A,A′A,A'δ⟨r4⟩(A,A′)\delta\langle r^4\rangle^{(A,A')} [fm4^4] σEFT(A,A′)\sigma_{\mathrm{EFT}}^{(A,A')}σVS(A,A′)\sigma_{\mathrm{VS}}^{(A,A')}σMB(A,A′)\sigma_{\mathrm{MB}}^{(A,A')}
(170,168)2.040.630.15
(172,170)2.040.970.16
(174,172)2.140.980.16
(176,174)1.110.580.16

Table XIII. Assessed standard deviations for δ⟨r2⟩\delta\langle r^2\rangle and δ⟨r4⟩\delta\langle r^4\rangle EFT, valence-space, and many-body errors.

Finally, the covariance between εi(A,A−2)\varepsilon_i^{(A,A-2)} and εi(B,B−2)\varepsilon_i^{(B,B-2)} (i∈{EFT,VS}i \in\{\mathrm{EFT},\mathrm{VS}\}) is estimated as

cov⁡(εi(A,A−2),εi(B,B−2))=ri(A−B)/2σi(A,A−2)σi(B,B−2),(16)\operatorname{cov}(\varepsilon_i^{(A,A-2)},\varepsilon_i^{(B,B-2)})=r_i^{(A-B)/2}\sigma_i^{(A,A-2)}\sigma_i^{(B,B-2)}, \tag*{(16)}

assuming an exponentially decaying correlation with a correlation rir_i. Based on tests of different correlation values against our computed values, we found rEFT=0.99r_{\mathrm{EFT}}=0.99 and rVS=0.97r_{\mathrm{VS}}=0.97. For the many-body uncertainties, we conservatively assumed no correlations.

We emphasize that these uncertainties and the underlying variances are not meant to be interpreted statistically but rather as an expert assessment of underlying parameters in our model to estimate correlated uncertainties and propagate those to King-plot nonlinearity analyses. We find this model to work well for δ⟨r2⟩\delta\langle r^2\rangle when compared to experimental values and expect it to work similarly well for δ⟨r4⟩\delta\langle r^4\rangle.

Ab initio atomic calculations of electronic coefficients

The dependence of isotope shifts on the nuclear structure parameters considered in this work is quantified by the electronic coefficients FF, KK, G(2)G^{(2)}, G(4)G^{(4)}, and DD of Eq. (6). These can be calculated from atomic theory, however the accuracy is severely limited because of the strong many-body correlations. Worse, because parameters that depend on nuclear size and shape all depend on the wavefunction at the nucleus, the coefficients FF and G(4)G^{(4)} (and DD at large mϕm_\phi) are nearly proportional to each other for different transitions. That means that quantities such as Gγα(4)=Gγ(4)−FγαGα(4)G_{\gamma\alpha}^{(4)}=G_\gamma^{(4)}-F_{\gamma\alpha}G_\alpha^{(4)} that appear in

the King linearity see Eq. (F1) contain strong cancellations and therefore have large errors. A key advantage of the approach outlined in Section J is that the changes in δ⟨r4⟩\delta\langle r^4\rangle are extracted with the accuracy of Gα(4)G_\alpha^{(4)} rather than Gγα(4)G_{\gamma\alpha}^{(4)}.

We have used AMBiT [36] to perform a particle-hole configuration interaction (CI) calculation of these parameters for both Yb and Yb+^+. For both atom and ion, we start with a self-consistent Dirac-Fock calculation for the closed-shell core with 68 electrons (up to the filled 4f144f^{14} shell). We then generate 6s6s, 6p6p, and 5d5d valence orbitals by solving the Dirac-Fock equations in the potential of this core. An orbital basis is then constructed by multiplying these orbitals with simple polynomials and orthogonality [103, 104]. The nuclear charge is represented by a Fermi-Dirac distribution.

For the case of transitions in Yb+^+ (α\alpha, β\beta, γ\gamma), we freeze

the core 5sp4d5sp4d orbitals [105]. The 4f4f shell is treated as a (hole) valence shell. We allow single and double electron excitations up to 8spdf8spdf from leading configurations 6s6s, 5d5d, 6p6p, 4f−16s24f^{-1}6s^2, 4f−16s 5d4f^{-1}6s\,5d, 4f−15d24f^{-1}5d^2, and 4f−16p24f^{-1}6p^2, with the additional restriction that a maximum of two holes in the 4f4f shell were allowed. Configuration state functions, with defined angular momentum JJ and projection M=JM=J were formed from these configurations, and all were included in the CI Hamiltonian, which was then diagonalized to obtain level energies and wavefunctions. Coefficients calculated with this method are listed as “Fiducial” in Table XIV.

Transition τ\tauFτF_\tau [GHz/fm2^2] Fiducial FFFτF_\tau [GHz/fm2^2] Fiducial ⟨F⟩\langle F\rangleFτF_\tau [GHz/fm2^2] Core holes FFFτF_\tau [GHz/fm2^2] Core holes ⟨F⟩\langle F\rangleGτ(2)G_\tau^{(2)} [MHz/fm4^4] Fiducial FFGτ(2)G_\tau^{(2)} [MHz/fm4^4] Fiducial ⟨G(2)⟩\langle G^{(2)}\rangleGτ(2)G_\tau^{(2)} [MHz/fm4^4] Core holes FFGτ(2)G_\tau^{(2)} [MHz/fm4^4] Core holes ⟨G(2)⟩\langle G^{(2)}\rangleGτ(4)G_\tau^{(4)} [MHz/fm4^4] Fiducial ⟨G(4)⟩\langle G^{(4)}\rangleGτ(4)G_\tau^{(4)} [MHz/fm4^4] Core holes ⟨G(4)⟩\langle G^{(4)}\rangle
α\alpha-14.69-13.09-16.22-16.1481.872.490.389.28.7610.80
β\beta-14.91-13.07-16.51-16.4583.072.391.890.98.7511.00
γ\gamma37.7829.0933.2445.45-209.7-160.8-184.5-250.7-19.47-30.38
δ\delta-9.73-10.62-9.82-11.354.258.854.962.27.127.57
ϵ\epsilon-13.54-14.82-13.69-15.5575.381.976.385.99.9110.41

Table XIV. Electronic field shift coefficients FτF_\tau, quadratic field shift coefficients Gτ(2)G_\tau^{(2)}, and quartic shift coefficients Gτ(4)G_\tau^{(4)} of the transitions τ\tau defined in Table X.

For the fiducial calculations of Yb transitions (δ\delta and ϵ\epsilon) we can keep the 4f4f shell within the frozen core and treat the atom as having two valence electrons above closed shells. We use the same Dirac-Fock potential as in Yb+^+ (in the case of Yb this is a VN−2V^{N-2} potential). We expand the virtual basis up to 12spdf12spdf and include all possible configurations of the two valence electrons within this basis.

To calculate electronic coefficients, we used both a finite field approach and an operator approach. In the finite field approach the nuclear radius is varied around the physical point and the coefficient is extracted by numerical derivative. For example, Fτ=dωτ/d⟨r2⟩F_\tau=d\omega_\tau/d\langle r^2\rangle where ωτ\omega_\tau is the calculated transition frequency and rr is the nuclear root-mean-square radius. This method is listed as “FF” in Table XIV. The operator approach simply evaluates the expectation value of an effective operator at the physical point, for example, Fτ=⟨δVnuc⟩iF_\tau=\langle\delta V_{\mathrm{nuc}}\rangle_i, where δVnuc\delta V_{\mathrm{nuc}} is the change in the nuclear potential caused by δ⟨r2⟩\delta\langle r^2\rangle. This method is listed as “⟨O⟩\langle O\rangle” in Table XIV. We see a significant difference between the two, which is expected since the operator approach does not include core relaxation. Nevertheless, we include it here as a “worst case” indication of theoretical uncertainty.

To provide a more reliable uncertainty measure, we have performed very large CI calculations by including, in addition, all configurations that can be formed by taking single electron excitations from core orbitals and up to two excitations of valence electrons from the leading configurations. Here we are inspired by [106] who suggest using single core excitations to treat valence and core-valence correlations more efficiently. We also expect that by unfreezing the core, the calculations of short-range matrix elements could be improved. The results of our calculations that include these additional core-excited configurations are listed in Table XIV as “Core holes”. We see that generally, the difference between finite field and operator approaches to the matrix elements is smaller when using core holes, reflecting that the relaxation of the core is partially included using this method. The exception is the γ\gamma transition for which we see a large difference: this transition directly involves a change in the occupancy of the 4f4f valence state, and the relaxation of this shell is large. Again we stress that our core hole approach is used as a measure of the theoretical uncertainty associated with having frozen cores in our fiducial CI method.

The DD coefficients that enter the bounds on new physics via the generalized King plot [11] are computed as described above (see also Ref. [7]), with the addition of a finite Yukawa potential λe−mϕrc/ℏ/r\lambda e^{-m_\phi r_c/\hbar}/r. For each new boson mass mϕm_\phi we repeat the calculation with several values of λ\lambda and take the numerical derivative. Figure 8 shows the values of the DD coefficients DτD_\tau, τ∈{α,β,γ,δ,ϵ}\tau\in\{\alpha,\beta,\gamma,\delta,\epsilon\}, as a function of the mass mϕm_\phi of the new boson, whereas Table XV provides a list of DD coefficients for all five transitions at selected values of mϕm_\phi. Note the difference in convention with respect to Ref. [7], which does not normalize the new physics coupling αNP\alpha_{\mathrm{NP}} by the fine structure constant αEM\alpha_{\mathrm{EM}}. The largest uncertainty in our calculation is due to the neglect of core polarization. Generally this will increase the size of the DD coefficients by up to 20%, which is an estimate of our error. We found that not including core-polarization improved numerical stability and gave more conservative bounds on new physics.

Electronic coefficients as a function of the mass of the new boson

Figure 8. Electronic coefficients DτD_\tau, τ∈{α,β,γ,δ,ϵ}\tau\in\{\alpha,\beta,\gamma,\delta,\epsilon\}, introduced in Eq. (1) of the main text, as a function of the mass mϕm_\phi of the new boson. The differences between DαD_\alpha and DβD_\beta are barely visible. The linear thresholds of the symmetric log-arithmetic plot are indicated with black dashed lines. Note the difference in convention with respect to Fig. S14 in Ref. [7].

mϕm_\phi [eV/c2c^2]iiDiD_i [Hz]
1α\alpha314941096838754.4
β\beta354896899928149.1
γ\gamma-2571409426668460.5
δ\delta-312730940277077.6
ϵ\epsilon41363423961625.695
10210^2α\alpha313713223754497.7
β\beta353538697268367.2
γ\gamma-2563969680971037.5
δ\delta-310519664492137.25
ϵ\epsilon41801954702786.555
10410^4α\alpha-53135793622491.055
β\beta-53749008165471.24
γ\gamma68784591261563.305
δ\delta-8440014196795.924
ϵ\epsilon-40882557593630.83
10610^6α\alpha-7939928663244.745
β\beta-8128695194140.711
γ\gamma20539381981880.473
δ\delta-4930494151686.425
ϵ\epsilon-7245790266413.627
10810^8α\alpha-4923184224.180642
β\beta-5007700585.950998
γ\gamma12117018040.460941
δ\delta-3251693925.4536247
ϵ\epsilon-4526134700.21715

Table XV. Electronic DD coefficients of transitions ii (see Eq. (1) in the main text) for selected values of the mediator mass mϕm_\phi, calculated using AMBiT [36].

Determining the evolution of r4r^{4} along the isotope chain from King-plot data

Assuming that the leading source of nonlinearity in the King plot is δ⟨r4⟩\delta\langle r^4\rangle, we can use isotope-shift measurements to extract information on this quantity. We proceed in two steps: Using the precise nuclear mass and frequency measurements presented in this work, we construct the quantities

ν^τα,r=ντα−wαwrντr,(17)\hat{\nu}^{\alpha,r}_{\tau}=\nu^\alpha_{\tau}-\frac{w^\alpha}{w^r}\nu^r_{\tau}, \tag*{(17)}

where τ\tau is the transition index, r=(174,176)r=(174,176) denotes the reference isotope pair, and a∈{(168,170),(170,172),(172,174)}a\in\{(168,170),(170,172),(172,174)\} is an isotope pair index. Approximating Eq. (6) (with α→τ\alpha\to\tau) by

ντa≈Fτδ⟨r2⟩a+Kτwa+Gτ(4)δ⟨r4⟩a,(18)\nu_\tau^a \approx F_\tau\delta\langle r^2\rangle^a+K_\tau w^a+G_\tau^{(4)}\delta\langle r^4\rangle^a, \tag*{(18)}

we obtain

ν^τa,r≈FτDa,r+Gτ(4)Qa,r,(19)\hat{\nu}_\tau^{a,r}\approx F_\tau D^{a,r}+G_\tau^{(4)}Q^{a,r}, \tag*{(19)}

where the isotope-pair dependent quantity

Da,r=δ⟨r2⟩a−wawrδ⟨r2⟩r,(20)D^{a,r}=\delta\langle r^2\rangle^a-\frac{w^a}{w^r}\delta\langle r^2\rangle^r, \tag*{(20)}

can be constructed from the charge radius measurements tabulated in Ref. [31], and

Qa,r=δ⟨r4⟩a−wawrδ⟨r4⟩r(21)Q^{a,r}=\delta\langle r^4\rangle^a-\frac{w^a}{w^r}\delta\langle r^4\rangle^r \tag*{(21)}

describes the evolution of the nuclear deformation parameter δ⟨r4⟩\delta\langle r^4\rangle along the isotope chain. Following the King-plot approach, we perform a linear fit in isotope pair space, i.e., we fit FτF_\tau to

ν^τr≈FτDr,(22)\hat{\nu}_\tau^r\approx F_\tau\mathbf{D}^r, \tag*{(22)}

where xr=(xa,r,xb,r,xc,r,xd,r)\mathbf{x}^r=(x^{a,r},x^{b,r},x^{c,r},x^{d,r}), x∈{ν^τ,D}x\in\{\hat{\nu}_\tau,\mathbf{D}\}. In this way, the precision of the isotope-shift measurements and nuclear mass measurements is exploited and field shift coefficients FτF_\tau with relative uncertainties at the level of ≈0.3%\approx0.3\% can be obtained.

Next, we subtract the fit results from the ν^τa,r\hat{\nu}^{a,r}_\tau and obtain the residuals

ν^τa,r−FτDa,r=Gτ(4)Qa,r.(23)\hat{\nu}^{a,r}_\tau-F_\tau D^{a,r}=G_\tau^{(4)}Q^{a,r}. \tag*{(23)}

Making use of the calculated electronic coefficients Gτ(4)G_\tau^{(4)}, which are listed in Table XIV we deduce three independent objects Qa,rQ^{a,r} per isotope-shift transition τ\tau. Fixing r=(174,176)r=(174,176) and τ=αPTB\tau=\alpha\mathrm{PTB} and choosing a value δ⟨r4⟩r=−7 fm4\delta\langle r^4\rangle^r=-7\,\mathrm{fm}^4, we are able to extract δ⟨r4⟩\delta\langle r^4\rangle values for the remaining isotope pairs, given in Table XVI. We choose the reference value δ⟨r4⟩r=−7 fm4\delta\langle r^4\rangle^r=-7\,\mathrm{fm}^4 informed by predictions from ab initio nuclear structure theory and density functional theory shown in Figure 9, which all predict this value to lie between 66–8 fm48\,\mathrm{fm}^4. Beyond this chosen reference value, uncertainties and covariances are obtained by propagating uncertainties from experimental input (ντr,wa,and δ⟨r2⟩a)(\nu_\tau^r,w^a,\text{and }\delta\langle r^2\rangle^a) and atomic theory (Gτ(4))(G_\tau^{(4)}) assuming independent normal distributions. The dominant source of uncertainty is δ⟨r2⟩(170,168)\delta\langle r^2\rangle^{(170,168)}, and we also understand this uncertainty to lead to correlations in the fit in Eq. (22), leading to the considerable anti-correlation between δ⟨r2⟩(170,168)\delta\langle r^2\rangle^{(170,168)} and δ⟨r4⟩(172,170)\delta\langle r^4\rangle^{(172,170)}.

Isotope pair a=(A,A′)a=(A,A')δ⟨r4⟩a\delta\langle r^{4}\rangle^{a} [fm4^{4}]cov⁡(δ⟨r4⟩a,δ⟨r4⟩b)\operatorname{cov}(\delta\langle r^{4}\rangle^{a},\delta\langle r^{4}\rangle^{b}) [fm8^{8}]
Isotope pair b=(A,A′)b=(A,A')
(170,168)(172,170)(174,172)
(170,168)7.33 (27)0.076−0.088−0.010
(172,170)7.53 (32)−0.0880.1040.006
(174,172)6.97 (28)−0.0100.0060.079
(176,174)7 (reference)

Table XVI. Experimental δ⟨r4⟩A,A−2\delta\langle r^{4}\rangle^{A,A-2} values relative to δ⟨r4⟩176,174=7 fm4\delta\langle r^{4}\rangle^{176,174}=7\ \mathrm{fm}^{4} extracted from isotope shifts from the αPTB\alpha_{\mathrm{PTB}} transitions using atomic theory with uncertainties and covariances propagated from input uncertainties.

We note that the approach outlined here to extract information on δ⟨r4⟩\delta\langle r^4\rangle from a King-plot nonlinearity is unable to give information on the absolute scale of δ⟨r4⟩\delta\langle r^4\rangle. This is simply because we are considering only the nonlinearity, but δ⟨r4⟩\delta\langle r^4\rangle also produces an isotope-shift contribution parallel to the linear field shift. We are unable to disentangle this contribution from the dominant field shift, so our approach is only able to make statements about how δ⟨r4⟩\delta\langle r^{4}\rangle changes across the isotopic chain. These values are compared with predictions for ab initio nuclear structure theory and density functional theory in Figure 9. Our extracted δ⟨r4⟩\delta\langle r^{4}\rangle values show a weak decrease between δ⟨r4⟩(172,170)\delta\langle r^{4}\rangle^{(172,170)} and δ⟨r4⟩(174,172)\delta\langle r^{4}\rangle^{(174,172)} and otherwise flat trend in δ⟨r4⟩\delta\langle r^{4}\rangle across the ytterbium isotopes studied here. Comparing this to the discussed nuclear structure calculations, we find that the flatter δ⟨r4⟩\delta\langle r^{4}\rangle trends of ab initio calculations more closely reproduce the extracted trends.

Bounds on the new boson

King-plot bounds

The ytterbium spectroscopy bounds in Figure 3 in the main text were derived by constructing generalized King plots [11], which require the isotope-shift and mass measurements of nn pairs of stable spinless isotopes on n−1n-1 transitions to eliminate n−2n-2 nuclear factors from the system of isotope-shift equations. In ytterbium, it is currently the number of suitable isotope pairs, n=4n=4, that fixes the maximal dimension of the generalized King plot: Using the frequency measurements on 3 transitions, the charge radius variance δ⟨r2⟩\delta\langle r^{2}\rangle, as well as the leading King nonlinearity, induced by δ⟨r4⟩\delta\langle r^{4}\rangle, can be eliminated. The remaining King nonlinearity is used to evaluate the sensitivity of the isotope-shift data to the coupling αNP\alpha_{\mathrm{NP}} of the new boson to electrons and neutrons. This method is particularly useful since it only requires theoretical input in the form of the DD-coefficients that appear in Eq. (6) (see also Table XV and Figure 8). Note that the subtraction of additional calculated SM effects such as δ⟨r2⟩\delta\langle r^{2}\rangle from the remaining King nonlinearity would reduce the precision of the generalized King-plot bound, since this would introduce unknown theory uncertainties. Indeed, the main benefit of the generalized King plot method over direct comparisons between theory and experiment is that it is data-driven, taking advantage of the experimental precision.

The red curve in Figure 3 in the main text shows the bound on the new boson obtained by applying the generalized King-plot method to the frequency and mass measurements presented in this work, combined with the frequency measurements presented in Ref. [29], and by computing

max⁡∣αNP±2σαNP∣/αEM,(24)\max\left|\alpha_{\mathrm{NP}}\pm2\sigma_{\alpha_{\mathrm{NP}}}\right|/\alpha_{\mathrm{EM}}, \tag*{(24)}

where σαNP\sigma_{\alpha_{\mathrm{NP}}} denotes the uncertainty on the new physics coupling αNP\alpha_{\mathrm{NP}} and is estimated using a simple Monte Carlo. (Simple uncertainty propagation leads to similar results.) The red curve in Figure 3 in the main text thus corresponds to the absolute maximum of the upper and lower 2σ2\sigma bounds on αNP/αEM\alpha_{\mathrm{NP}}/\alpha_{\mathrm{EM}}. This choice is motivated by the fact that without full understanding of nuclear effects, King plots can only place bounds on new physics and thus should never exclude the case of zero new physics (αNP→0\alpha_{\mathrm{NP}}\to0). Although the sign of αNP=(−1)s+1ynye/(4πℏc)\alpha_{\mathrm{NP}} = (-1)^{s+1}y_n y_e/(4\pi\hbar c) decides whether the new force is attractive or repulsive, showing the generalized King-plot bounds in the (αNP/αEM,mϕ)(\alpha_{\mathrm{NP}}/\alpha_{\mathrm{EM}},m_{\phi}) plane rather than the (∣αNP/αEM∣,mϕ)(|\alpha_{\mathrm{NP}}/\alpha_{\mathrm{EM}}|,m_{\phi}) plane could lead to the false conclusion that the bounds on αNP\alpha_{\mathrm{NP}} become infinitely stringent in the case where both the upper and the lower bounds are on one side of αNP=0\alpha_{\mathrm{NP}} = 0. Let us simply note that the bound that we are showing in Figure 3 in the main text (and in Figure 11) would, in the (αNP/αEM,mϕ)(\alpha_{\mathrm{NP}}/\alpha_{\mathrm{EM}},m_{\phi}) plane, be negative for mϕm_{\phi} values below the peak at around 10 keV and positive for the plotted mϕm_{\phi} values above the peak.

Comparison of the generalized King-plot bounds for the transition frequencies $\alpha_{\mathrm{MIT}}$, $\gamma_{\mathrm{MIT}}$, $\delta$, $\alpha_{\mathrm{PTB}}$, and $\gamma_{\mathrm{PTB}}$

Figure 11. Comparison of the generalized King-plot bounds for the transition frequencies αMIT\alpha_{\mathrm{MIT}}, γMIT\gamma_{\mathrm{MIT}} [5, 7], δ\delta [29] and αPTB\alpha_{\mathrm{PTB}}, γPTB\gamma_{\mathrm{PTB}} presented in this work. The bounds employing the inverse mass differences stated in Ref. [7] are labeled with “2022”, the bounds employing the mass-ratio measurements presented in this work are labeled with “MPIK”. All bounds are computed using Eq. (24).

In Figure 3 of the main text, our new bound is compared with bounds constructed using the same methodology and mass-ratio measurements, but with Yb+^{+} isotope-shift data from Refs. [7, 29] (black, maroon curves), and with King-plot bounds obtained by applying the projection method of Ref. [37] to the Ca+^{+} data of Refs. [37, 38] (cyan, blue curves). In the (αNP/αEM,mϕ)(\alpha_{\mathrm{NP}}/\alpha_{\mathrm{EM}},m_{\phi}) plane, the generalized King-plot bound derived for the set of transitions αMIT\alpha_{\mathrm{MIT}}, γMIT\gamma_{\mathrm{MIT}}, and δ\delta (black curve) would have the opposite sign to the bound for αPTB\alpha_{\mathrm{PTB}}, γPTB\gamma_{\mathrm{PTB}}, and δ\delta (red).

This is due to the differences in the isotope-shift data ναA,A+2\nu_{\alpha}^{A,A+2} presented in Table 1 in the main text and provides yet another argument for interpreting the upper bound on ∣αNP∣|\alpha_{\mathrm{NP}}| as a sensitivity of the respective dataset to the new boson rather than focusing on the sign of the bound. Indeed, the fact that for the datasets (αPTB,γPTB,δ)(\alpha_{\mathrm{PTB}},\gamma_{\mathrm{PTB}},\delta) and (αMIT,γMIT,δ)(\alpha_{\mathrm{MIT}},\gamma_{\mathrm{MIT}},\delta) the upper and the lower bounds on αNP\alpha_{\mathrm{NP}} are on the same side of αNP=0\alpha_{\mathrm{NP}} = 0 implies that a King-plot nonlinearity is dominating the system of isotope shifts, and should ideally be removed by constructing a higher-dimensional generalized King plot from a larger dataset. The aforementioned offsets between the αMIT\alpha_{\mathrm{MIT}} and αPTB\alpha_{\mathrm{PTB}} measurements also affect the locations of the characteristic peaks in the generalized King-plot bounds. This explains the difference in the slopes of the black and red curves at the upper end of the plotted mϕm_{\phi} region. When these offsets are taken into account, these two slopes agree.

Figure 11 shows how the improved accuracy and precision of the frequency and mass-ratio measurements translate into sensitivity to the new physics coupling αNP\alpha_{\mathrm{NP}}: Whereas higher precision leads to smaller uncertainties σαNP\sigma_{\alpha_{\mathrm{NP}}} on the new physics coupling, and thus have the tendency to lead to stronger bounds on ∣αNP/αEM∣|\alpha_{\mathrm{NP}}/\alpha_{\mathrm{EM}}|, a shift in the central value of a mass or frequency measurement will generally lead to a shift of the central value αNP\alpha_{\mathrm{NP}} predicted by the generalized King-plot construction. This effect can be observed by comparing the black and the gray curves in Figure 11: Since the bound indicated in black was constructed using more precise mass measurements, one would naively expect it to be more stringent than the bound shown in gray that employs the same mass values as Ref. [7]. However, the shift in the central value of m168m_{168} between the old and the new mass-ratio measurements translates into an upward shift of the bound, which cannot be compensated by the increased experimental precision. In other words, the bound shown in gray appears to be more stringent than the bound shown in black since it underestimates a hitherto unknown systematic uncertainty. Comparing the new bound presented by this work (red curve) against the gray and the black and the gray curves, we see that both the improvement of the isotope-shift measurements and of the mass-ratio measurements have an impact on the final bound, and that the achieved sensitivity was only made possible by the combination of these improvements.

Comparing the orange and maroon curves in Figure 12, which show the generalized King-plot bounds obtained from the combination of the isotope-shift frequencies associated to the transitions β\beta [5], γMIT\gamma_{\mathrm{MIT}} [5, 7], and δ\delta [29], we observe a similar upward shift to that between the gray and the black curves. The dashed orange curve in Figure 12 corresponds to the absolute maximum of the upper and lower bounds on the new physics coupling presented in Ref. [7]. This is not a generalized King-plot bound, but a bound obtained from a combination of a fit and the theoretical input for the electronic coefficients DρστD_{\rho\sigma\tau}, where ρ\rho, σ\sigma and τ\tau are transition indices [7]. Since the theoretical uncertainty on these triple-index electronic coefficients DρστD_{\rho\sigma\tau} is expected to be significantly larger than on the single-index electronic coefficients DτD_{\tau}, and since we aim to compare the experimental results on equal footing, we do not show this bound in Figure 3 in the main text but instead show the generalized King-plot bound computed using the same data.

Comparison of the King-plot bounds for the transition frequencies $\alpha_{\mathrm{MIT}}$, $\gamma_{\mathrm{MIT}}$, $\beta$, and $\delta$

Figure 12. Comparison of the King-plot bounds for the transition frequencies αMIT\alpha_{\mathrm{MIT}}, γMIT\gamma_{\mathrm{MIT}} [5, 7], β\beta [5] and δ\delta [29]. The bounds employing the inverse mass differences stated in Ref. [7] are labeled with “2022”, the bounds employing the mass-ratio measurements presented in this work are labeled with “MPIK”. All curves apart from the orange dashed curve show generalized King-plot bounds. The orange dashed curve is taken from Ref. [7]. All bounds are computed using Eq. (24).

Competing bounds

The competing 2σ2\sigma bounds are shown as exclusion regions. The combination of the bound on yey_e from the measurement of the anomalous magnetic moment of the electron (g−2)e(g-2)_e [39, 40] and the bound on yny_n [45] from neutron optics [41] and neutron scattering experiments [41, 42, 43, 44] is indicated in yellow. For masses mϕ<102 eV/c2m_\phi< 10^2\,\mathrm{eV}/c^2, fifth force searches via the Casimir effect provide particularly strong bounds (green) [47, 48].

In green we show the bounds from hydrogen-deuterium (HD) isotope shifts [107] (see also Refs. [46, 108] for bounds on new physics from hydrogen and deuterium spectroscopy). For light elements, there are not sufficiently many isotopes to construct a King plot, but the precision of the theory predictions reaches the level of the experimental precision, such that a comparison of the two can be used to set a bound. This bound, however, depends sensitively on experimental input for the proton and deuteron charge radii. The hydrogen-

deuterium bounds in Figure 3 in the main text are obtained using charge radii determined via electron scattering and muonic atom spectroscopy, respectively. The former leads to a weaker bound than the bound from (g−2)e(g-2)_e times neutron scattering, and the latter could be affected by the new boson, if the latter couples to muons.

Beam dump experiments provide competitive bounds for 2me≲mϕ2m_e \lesssim m_\phi. These are not shown in Figure 3 in the main text due to their dependence on the lifetime of the new boson ϕ\phi, which is a function of all couplings of ϕ\phi, not only of the couplings to electrons and neutrons.

The blue-shaded areas in Figure 3 in the main text are disfavored by astrophysical considerations, such as bounds from star cooling in horizontal branch stars (globular cluster) [49, 50] and from energy loss in the core of the supernova SN1987a [51], which is subject to an O(1)\mathcal{O}(1) uncertainty [109]. Complementary bounds from laboratory experiments would provide valuable contributions to the search for new bosons also in these regions of parameter space.

Finally, the green dashed lines in Figure 3 in the main text show the bounds on yey_e from the anomalous magnetic moment of the electron (g−2)e(g-2)_e [39, 40] multiplied by the bounds on yny_n from the anomalous meson decay K→π+invisibleK \to\pi+ \text{invisible}, which can be estimated using the bounds on the new boson couplings to quarks and the first coefficient of the QCD β\beta-function [110]. These bounds are highly dependent on the flavor structure of the couplings of the new boson ϕ\phi to the quarks. For illustration, Figure 3 in the main text shows the bound on αNP/αEM\alpha_{\mathrm{NP}}/\alpha_{\mathrm{EM}} in the presence of a coupling to top-quarks, yty_t, and the bound in presence of a coupling to up-quarks, yuy_u. They differ by approximately seven orders of magnitude.

This leaves open regions of parameter space where King plots have the potential of providing the most competitive bounds on ϕ\phi that are agnostic both of its decay widths into dark matter and of the flavor structure of its couplings to quarks. In the future, the King-plot bounds can be further improved by increasing the dimension of the generalized King plots. This can be achieved by expanding the experimental program to isotopes with long half-lives (e.g., 166Yb{}^{166}\mathrm{Yb}) or by moving to new atomic systems such as Sn [56] or Xe [57], both of which have seven, spinless stable isotopes and suitable clock transitions.

References

  1. [1]M. S. Safronova et al., Rev. Mod. Phys. 90, 025008 (2018).
  2. [2]R. F. Garcia Ruiz et al., Nat. Phys. 12, 594 (2016).
  3. [3]C. Delaunay et al., Phys. Rev. D 96, 093001 (2017).
  4. [4]J. C. Berengut et al., Phys. Rev. Lett. 120, 091801 (2018).
  5. [5]I. Counts et al., Phys. Rev. Lett. 125, 123002 (2020).
  6. [6]S. O. Allehabi et al., Phys. Rev. A 103, L030801 (2021).
  7. [7]J. Hur et al., Phys. Rev. Lett. 128, 163201 (2022).
  8. [8]D. A. Nesterenko et al., Int. J. Mass Spectrom. 458, 116435 (2020).
  9. [9]R. Rana, M. Höcker, and E. G. Myers, Phys. Rev. A 86, 050502(R) (2012).
  10. [10]J. Repp et al., Appl. Phys. B 107, 983 (2012).
  11. [11]J. C. Berengut et al., Phys. Rev. Res. 2, 043444 (2020).
  12. [12]H. Hergert et al., Phys. Rep. 621, 165 (2016).
  13. [13]S. R. Stroberg et al., Ann. Rev. Nucl. Part. Sci. 69, 307 (2019).
  14. [14]H. Hergert, Front. Phys. 8, 379 (2020).
  15. [15]T. Miyagi et al., Phys. Rev. C 105, 014302 (2022).
  16. [16]E. Epelbaum, H. W. Hammer, and U. G. Meißner, Rev. Mod. Phys. 81, 1773 (2009).
  17. [17]W. H. King, J. Opt. Soc. Am. 53, 638 (1963).
  18. [18]F. Heiße et al., Phys. Rev. Lett. 119, 033001 (2017).
  19. [19]C. Roux et al., Appl. Phys. B 107, 997 (2012).
  20. [20]A. Rischka et al., Phys. Rev. Lett. 124, 113001 (2020).
  21. [21]L. S. Brown and G. Gabrielse, Phys. Rev. A 25, 2423 (1982).
  22. [22]C. Froese Fischer et al., Comput. Phys. Commun. 237, 184 (2019).
  23. [23]W. J. Huang et al., Chinese Phys. C 45, 030002 (2021).
  24. [24]E. Tiesinga et al., Rev. Mod. Phys. 93, 025010 (2021).
  25. [25]K. Pyka et al., Appl. Phys. B 114, 231 (2014).
  26. [26]J. Keller et al., Phys. Rev. Appl. 11, 011002 (2019).
  27. [27]D. G. Matiei et al., Phys. Rev. Lett. 118, 263202 (2017).
  28. [28]R. Lange et al., Phys. Rev. Lett. 126, 011102 (2021).
  29. [29]K. Ono et al., Phys. Rev. X 12, 021033 (2022).
  30. [30]N. L. Figueroa et al., Phys. Rev. Lett. 128, 073001 (2022).
  31. [31]I. Angeli and K. P. Marinova, At. Data Nucl. Data Tables 99, 69 (2013).
  32. [32]N. Shimizu et al., Phys. Rev. C 103, 014312 (2021).
  33. [33]K. Hebeler et al., Phys. Rev. C 83, 031301 (2011).
  34. [34]W. G. Jiang et al., Phys. Rev. C 102, 054301 (2020).
  35. [35]M. Heinz et al., Phys. Rev. C 103, 044318 (2021).
  36. [36]E. V. Kahl and J. C. Berengut, Comput. Phys. Commun. 238, 232 (2019).
  37. [37]C. Solaro et al., Phys. Rev. Lett. 125, 123003 (2020).
  38. [38]T. T. Chang et al., Phys. Rev. A 110, L030801 (2024).
  39. [39]D. Hanneke, S. Fogwell, and G. Gabrielse, Phys. Rev. Lett. 100, 120801 (2008).
  40. [40]X. Fan et al., Phys. Rev. Lett. 130, 071801 (2023).
  41. [41]H. Leeb and J. Schmiedmayer, Phys. Rev. Lett. 68, 1472 (1992).
  42. [42]V. V. Nesvizhevsky, G. Pignol, and K. V. Protasov, Phys. Rev. D 77, 034020 (2008).
  43. [43]Y. N. Pokotilovski, Phys. Atom. Nucl. 69, 924 (2006).
  44. [44]R. Barbieri and T. E. O. Ericson, Phys. Lett. B 57, 270 (1975).
  45. [45]C. Frugiuele et al., Phys. Rev. D 96, 015011 (2017).
  46. [46]R. M. Potvliege, A. Nicolson, M. P. A. Jones, and M. Spannowsky, Phys. Rev. A 108, 052825 (2023).
  47. [47]M. Bordag, U. Mohideen, and V. M. Mostepanenko, Phys. Rep. 353, 1 (2001).
  48. [48]M. Bordag et al., Advances in the Casimir Effect (Oxford University Press, Oxford, 2009).
  49. [49]J. A. Grifols and E. Masso, Phys. Lett. B 173, 237 (1986).
  50. [50]J. A. Grifols, E. Masso, and S. Peris, Mod. Phys. Lett. A 4, 311 (1989).
  51. [51]G. Raffelt, Phys. Rev. D 86, 015001 (2012).
  52. [52]A. J. Krasznahorkay et al., Phys. Rev. Lett. 116, 042501 (2016).
  53. [53]P.-G. Reinhard, W. Nazarewicz, and R. F. Garcia Ruiz, Phys. Rev. C 101, 021301 (2020).
  54. [54]H. Kurasawa, T. Suda, and T. Suzuki, Prog. Theor. Exp. Phys. 2021, 013D02 (2021).
  55. [55]T. Naito, G. Colò, H. Liang, and X. Roca-Maza, Phys. Rev. C 104, 024316 (2021).
  56. [56]D. R. Leibrandt, S. G. Porsev, C. Cheung, and M. S. Safronova, Nature Communications 15 (2024)
  57. [57]N.-H. Rehbehn et al., Phys. Rev. Lett. 131, 161803 (2023).
  58. [58]C. Lyu, C. H. Keitel, and Z. Harman, (2023), arXiv:2305.09603.DOI
  59. [59]R. X. Schüssler et al., Nature 581, 42 (2020).
  60. [60]P. Filianin et al., Phys. Rev. Lett. 127, 072502 (2021).
  61. [61]K. Kromer et al., Eur. Phys. J. A 58, 202 (2022).
  62. [62]K. Kromer et al., Phys. Rev. C 109, L021301 (2024).
  63. [63]C. Schweiger et al., Rev. Sci. Instrum. 90, 123201 (2019).
  64. [64]N. E. Bradbury and R. A. Nielsen, Phys. Rev. 49, 388 (1936).
  65. [65]E. A. Cornell et al., Phys. Rev. A 41, 312 (1990).
  66. [66]D. J. Wineland and H. G. Dehmelt, J. Appl. Phys. 46, 919 (1975).
  67. [67]E. A. Cornell et al., Phys. Rev. Lett. 63, 1674 (1989).
  68. [68]L. S. Brown and G. Gabrielse, Rev. Mod. Phys. 58, 233 (1986).
  69. [69]J. Ketter et al., Int. J. Mass Spectrom. 358, 1 (2014).
  70. [70]R. S. Van Dyck et al., Phys. Rev. A 40, 6308 (1989).
  71. [71]M. Schuh et al., Phys. Rev. A 100, 023411 (2019).
  72. [72]S. Rau et al., Nature 585, 43 (2020).
  73. [73]A. Kramida, Y. Ralchenko, J. Reader, and NIST ASD Team, “NIST Atomic Spectra Database (version 5.11),” Online (2023).
  74. [74]I. P. Grant, Adv. Phys. 19, 747 (1970).
  75. [75]J. P. Desclaux, D. F. Mayers, and F. O’Brien, J. Phys. B: Atom. Mol. Phys. 4, 631 (1971).
  76. [76]I. P. Grant, Relativistic quantum theory of atoms and molecules: theory and computation (Springer New York, 2007).
  77. [77]P. Jönsson et al., Atoms 11, 7 (2023).
  78. [78]A. V. Malyshev et al., Phys. Rev. A 90, 062517 (2014).
  79. [79]A. V. Malyshev et al., Phys. Rev. Lett. 126, 183001 (2021).
  80. [80]P. Jönsson et al., Atoms 11, 68 (2023).
  81. [81]C. Lyu et al., Mol. Phys. 0, e2252105 (2023).
  82. [82]Bureau International des Poids et Mesures (BIPM), https://www.bipm.org/documents/20126/69375142/171Yb+_642THz_2021.pdf/a22c77ad-0c15-e054-ee6c-b7114a23a7a3 (2022).
  83. [83]E. C. Beaty, J. Appl. Phys. 61, 2118 (1987).
  84. [84]J. Keller et al., Appl. Phys. B 116, 203 (2014).
  85. [85]H. A. Fürst et al., Phys. Rev. Lett. 125, 163001 (2020).
  86. [86]C.-H. Yeh et al., New J. Phys. 25, 093054 (2023).
  87. [87]C. F. Roos et al., Nature 443, 316 (2006).
  88. [88]T. R. Tan et al., Phys. Rev. A 104, L010802 (2021).
  89. [89]R. Lange et al., Phys. Rev. Lett. 125, 143201 (2020).
  90. [90]J. Keller et al., J. Appl. Phys. 118, 104501 (2015).
  91. [91]N. Huntemann et al., Phys. Rev. Lett. 116, 063001 (2016).
  92. [92]E. Peik, T. Schneider, and C. Tamm, J. Phys. B: At. Mol. Opt. Phys. 39, 145 (2005).
  93. [93]T. Lindvall et al., Metrologia 60, 045008 (2023).
  94. [94]C. Sanner et al., Nature 567, 204 (2019).
  95. [95]K. Tsukiyama, S. K. Bogner, and A. Schwenk, Phys. Rev. Lett. 106, 222502 (2011).
  96. [96]D. R. Entem and R. Machleidt, Phys. Rev. C 68, 041001 (2003).
  97. [97]J. Simonis et al., Phys. Rev. C 96, 014303 (2017).
  98. [98]S. R. Stroberg et al., Phys. Rev. Lett. 126, 022501 (2021).
  99. [99]K. Hebeler, Phys. Rep. 890, 1 (2021).
  100. [100]P. Arthuis, K. Hebeler, and A. Schwenk, arXiv:2401.06675.DOI
  101. [101]A. Ong, J. C. Berengut, and V. V. Flambaum, Phys. Rev. C 82, 014320 (2010).
  102. [102]J. L. Friar, J. Martorell, and D. W. L. Sprung, Phys. Rev. A 56, 4579 (1997).
  103. [103]P. Bogdanovich and G. Žukauskas, Sov. Phys. Collect. 23, 13 (1983).
  104. [104]M. G. Kozlov, S. G. Porsev, and V. V. Flambaum, J. Phys. B: At. Mol. Opt. Phys. 29, 689 (1996).
  105. [105]In this notation the integers refer to the highest principal quantum numbers of the orbitals that follow. Here it refers to 1s − 5s, 2p − 5p, and 3d − 4d.
  106. [106]M. G. Kozlov et al., Phys. Rev. A 105, 052805 (2022).
  107. [107]C. Delaunay et al., Phys. Rev. D 96, 115002 (2017).
  108. [108]M. P. A. Jones, R. M. Potvliege, and M. Spannowsky, Phys. Rev. Res. 2, 013244 (2020).
  109. [109]K. Blum and D. Kushnir, Astrophys. J. 828, 31 (2016).
  110. [110]S. Knapen, T. Lin, and K. M. Zurek, Phys. Rev. D 96, 115021 (2017).

Paper details

vlnc:W_msmk9n2k
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents