In this paper
AbstractIntroductionMethodologyGauge ensembles and statisticsComputation of correlatorsAnalysis of correlatorsRenormalizationFitting ProcedureThe two-step approachExtraction of ground state matrix elementsContinuum extrapolationSystematic uncertaintiesComparison with other resultsSummary and ConclusionsACKNOWLEDGMENTSReferencesWorksPhysics and Astronomy
Flavor decomposition of the nucleon axial and tensor charges from lattice QCD
Abstract
We determine the nucleon axial and tensor charges using four ensembles of twisted-mass clover-improved fermions with approximately equal physical volumes of about 5.5~fm and four lattice spacings, ~fm, 0.06~fm, 0.07~fm, and 0.08~fm. The mass of the degenerate up and down, strange and charm quarks are tuned to approximately their physical values. We compute both connected and disconnected contributions to determine the isovector and isoscalar charges and their flavour decomposition. For the nucleon isovector axial charge we find , in agreement with the experimental value. For the isoscalar axial charge we find , while for the corresponding tensor charges we find and .
Introduction
The nucleon axial and tensor charges are fundamental quantities in the Standard Model and provide important insights into nucleon structure and dynamics. The axial charge, , determines the strength of neutron beta decay and provides a direct probe of chiral symmetry breaking in hadron physics. It also enters the interpretation of neutrinoless double-beta decay and plays a central role in precision tests of the unitarity of the Cabibbo–Kobayashi–Maskawa (CKM) matrix. Flavor-diagonal axial charges, , characterize the contribution of quark flavor to the intrinsic spin of the nucleon, , and are constrained by deep-inelastic scattering (DIS) experiments at facilities, such as Jefferson Lab and CERN. Extending these measurements to the low-Bjorken- region is one of the key physics goals of the future Electron-Ion Collider (EIC), where improved sensitivity to the quark and gluon spin structure of the nucleon is expected.
The isovector tensor charge, , is less well known but plays an important role in constraining beyond-the-Standard-Model (BSM) interactions [1]. It provides essential theoretical input for the interpretation of results from neutrino scattering experiments, including DUNE [2], COHERENT [3], and GEMMA [4], as well as from direct dark matter detection experiments [5, 6, 7]. An accurate determination of the tensor charge is also important for phenomenological studies of the nucleon transversity parton distribution function [8], where it can be used as input in such analyses.
In this work, we compute the nucleon axial and tensor charges using twisted-mass, clover-improved fermion ensembles at four different lattice spacings. The ensembles are generated by the Extended Twisted Mass Collaboration (ETMC) with a mass degenerate up and down quark doublet and strange and charm quarks (), with all quark masses tuned to approximately their physical values. This setup, which we refer to as the physical-mass point, allows for a precise determination of both the isovector and flavor-diagonal charges without the need for chiral extrapolations. Our results provide important benchmarks for lattice QCD calculations and deliver essential theoretical input for precision studies probing BSM interactions, CP violation, and dark-matter interactions.
Methodology
The nucleon charges for each quark flavor , denoted as , are derived from the nucleon matrix elements of the corresponding axial and tensor operators at zero momentum transfer. They are defined as
where is the nucleon spinor and for the axial-vector and for the tensor operators.
Gauge ensembles and statistics
| Ensemble | Abrv. | [fm] | [MeV] | ||
| cB211.072.64 | B64 | 1.778 | 0.07948(11) | 140.2(2) | |
| cC211.060.80 | C80 | 1.836 | 0.06819(14) | 136.7(2) | |
| cD211.054.96 | D96 | 1.900 | 0.056850(90) | 140.8(2) | |
| cE211.044.112 | E112 | 1.960 | 0.04892(11) | 136.5(2) |
Table 1. Parameters of the ensembles analyzed in this work. In the first column, we give the name of the ensemble, in the second the abbreviated name, in the third the lattice volume, in the fourth with the bare coupling constant, in the fifth the lattice spacing and in the last the pion mass. Lattice spacings and pion masses are taken from Ref. [9].
To compute the nucleon matrix elements, we examine gauge configurations produced using the twisted-mass fermion discretization, which provides automatic improvement [10, 11]. The action incorporates a clover term [12], which helps suppress isospin-breaking effects associated with the twisted mass fermion formulation.
The isosymmetric pion mass, MeV [13, 14], is reproduced by tuning the bare light quark parameter . For the heavy quarks, the parameters and are tuned using the kaon mass, a suitably defined ratio of the D-meson mass to its decay constant, and a phenomenologically motivated ratio between the strange and charm quark masses, following Refs. [13, 14]. Table I lists the parameters of the ensembles employed in this study. The lattice spacings extracted from the meson sector and the pion masses are taken from Ref. [9], with the former being consistent with those obtained from the nucleon sector in Ref. [15].
| cB211.072.64 | cC211.060.80 | cD211.054.96 | cE211.044.112 | |||||||||||
| 749 configurations | 400 configurations | 496 configurations | 478 configurations | |||||||||||
| [fm] | [fm] | [fm] | ||||||||||||
| 8 | 0.64 | 1 | 6 | 0.41 | 1 | 8 | 0.46 | 1 | 8 | 0.39 | 1 | |||
| 10 | 0.80 | 2 | 8 | 0.55 | 2 | 10 | 0.57 | 2 | 11 | 0.54 | 2 | |||
| 12 | 0.96 | 5 | 10 | 0.69 | 4 | 12 | 0.68 | 4 | 14 | 0.69 | 4 | |||
| 14 | 1.12 | 10 | 12 | 0.82 | 10 | 14 | 0.80 | 8 | 17 | 0.83 | 8 | |||
| 16 | 1.28 | 32 | 14 | 0.96 | 22 | 16 | 0.91 | 16 | 20 | 0.98 | 16 | |||
| 18 | 1.44 | 112 | 16 | 1.10 | 48 | 18 | 1.03 | 32 | 23 | 1.13 | 32 | |||
| 20 | 1.60 | 128 | 18 | 1.24 | 45 | 20 | 1.14 | 64 | 26 | 1.27 | 64 | |||
| 20 | 1.37 | 116 | 22 | 1.25 | 16 | 29 | 1.42 | 128 | ||||||
| 22 | 1.51 | 246 | 24 | 1.37 | 32 | |||||||||
| 26 | 1.48 | 64 | ||||||||||||
| Nucleon 2pt | Nucleon 2pt | Nucleon 2pt | Nucleon 2pt |
Table 2. Statistics used for the connected two- and three-point functions. In the first row we give the name of the ensemble and in the second the number of configurations analyzed per ensemble. For each ensemble, we provide the sink-source time separations in lattice units (first column) and physical units (second column) and the number of source positions per configuration (third column). The last row indicates the number of source positions used for the two-point functions.
| Flavor | \multicolumn{4}{c}{cB211.072.64} | \multicolumn{4}{c}{cC211.060.80} | \multicolumn{4}{c}{cD211.054.96} | \multicolumn{4}{c}{cE211.044.112} | ||||||||||||
| \multicolumn{4}{c}{749 configurations} | \multicolumn{4}{c}{400 configurations} | \multicolumn{4}{c}{494 configurations} | \multicolumn{4}{c}{450 configurations} | |||||||||||||
| Light | 200 | 1 | 512 | 6144 | 450 | 1 | 512 | 6144 | 0 | 8 | 512 | 49152 | 530 | 1 | 512 | 6144 |
| Strange | 0 | 2 | 512 | 12288 | 0 | 4 | 512 | 24576 | 0 | 4 | 512 | 24576 | 0 | 2 | 512 | 12288 |
| Charm | 0 | 12 | 32 | 4608 | 0 | 1 | 512 | 6144 | 0 | 1 | 512 | 6144 | 0 | 1 | 512 | 6144 |
Table 3. Statistics used for the disconnected loops. In the first row we give the name of the ensemble and in the second the number of configurations analyzed per ensemble. For each ensemble, we give from left to right: i) the number of deflated eigenvectors , ii) the number of stochastic sources , iii) the number of Hadamard vectors , and iv) the total number of computed vectors, , which after color and spin dilution are obtained via .
In order to extract the nucleon charges, we compute two- and three-point nucleon correlation functions and perform fits to isolate the desired ground-state matrix element, as described in detail in Sec. IV. In Tables II and III, we present the statistics used for the calculation of the correlation functions, providing the number of configurations analyzed, the number of source positions per configuration as well as details on the stochastic noise and Hadamard vectors used in the computation of disconnected quark loops. The connected three-point functions are evaluated using the so-called fixed sink approach [16, 17], in which a separate inversion is performed for every source-sink separation . As reported in Table II, the number of source positions per configuration is increased for larger values of , ensuring that the statistical uncertainties of the connected three-point functions remain approximately constant.
The disconnected quark loops are calculated for the light, strange, and charm quarks. To enhance the signal-to-noise ratio of these disconnected loops, several noise-reduction techniques are applied. These include the one-end trick [18], exact deflation of low modes [19, 20], and hierarchical probing [21], as explained in Ref. [22] for the specific application to twisted mass fermions. The one-end trick is applied to all quark loops for all four ensembles. We also use hierarchical probing for all quark loops with a probing distance of 8 for the light and strange quark loops and with 4 for the charm-quark loops. Deflation of low modes is applied to the light quark loops for the cB211.072.64 and cC211.060.80 ensembles. For the cD211.054.96 ensemble, a higher number of stochastic sources are used since using deflation for this lattice, which scales with the square of the volume, was prohibitively expensive with our original implementation and the then accessible hardware. With our improved implementation and latest hardware we are able to employ deflation for the cE211.044.112 ensemble.
Computation of correlators
The determination of the nucleon matrix elements requires the evaluation of both three- and two-point Euclidean time correlation functions. The two-point function is given by
where we take the source at the origin, are the sink coordinates, is the unpolarized, positive parity projector and the interpolating field for the nucleon is
To enhance the overlap of the interpolating field with the nucleon ground state and suppress excited-state contamination, thereby allowing the ground state to dominate at the smallest possible time separations, we apply Gaussian smearing [23, 24] to the quark fields that enter the interpolating field.
where is either or of Eq. (3) and is the hopping matrix
The parameters and are adjusted [25, 26] to achieve a nucleon smearing radius of approximately 0.5 fm. To suppress statistical noise arising from ultraviolet fluctuations, APE smearing [27] is applied to the gauge links used in the hopping matrix.
The three-point function for e.g. the axial-vector for an arbitrary momentum transfer is given by
where is the polarized projector, is the sink momentum and , is the axial operator, which is defined as
Analogous expressions hold for the tensor three-point functions, where the tensor operator is
with . These tensor matrix elements are obtained using the same polarized projector . In the following, we will adopt the axial-vector case for our discussion, noting that the generalization to tensor operator insertions is straightforward. Since we extract nucleon charges in the forward limit, we take and work in the rest frame of the nucleon.
Analysis of correlators
The spectral decomposition of the two- and three-point functions is given respectively by
The coefficients of the two-point function are overlap terms given by
and coefficients in the three-point function are given by
where, denotes the nucleon matrix element of interest in the axial case, and an analogous expression applies to the tensor three-point function. The interpolating field creates from the vacuum, i.e. , hadron states carrying the quantum numbers of the nucleon. The states for are excited states with energy . To isolate the matrix elements from , these unknown overlaps must be removed and for the ground-state matrix element, we build the ratio of the three-point function to the two-point function
We note that one requires a sequential propagator in the construction of the three-point function for each projector matrix . In the limit of large time separations, and , with , the energy difference between the first excited state and the ground state, this ratio gives the desired axial charge
The only non-zero contribution arises when the current is in the same direction as the projector i.e. when . We then average over all spatial directions . We use the parameters and extracted from our analysis and substitute Eqs. (9–10) into Eq. (13) to obtain the charges via
The corresponding ratio for the tensor operator yields the tensor charge,
where and , are renormalization constants that are computed non-perturbatively as discussed in Sec. III and no summation over the index is implied. In practice, one must determine the smallest value of at which excited-state contributions are adequately suppressed. The rate at which ground-state dominance is reached depends both on the smearing applied to the interpolating fields and on the energy gap separating the ground state from the excited states. Because the noise grows exponentially with , it is crucial to establish convergence to the ground state as early as possible. To reliably extract the ground-state matrix elements, and thus determine the values of and , we adopt a multi-state fitting strategy, as described in Sec. IV.
Renormalization
The matrix elements of the axial-vector and tensor currents are renormalized nonperturbatively in the RI-MOM scheme [28], followed by perturbative conversion to the MS scheme at the reference scale GeV. We determine the renormalization factors for both flavor-nonsinglet () and flavor-singlet () operators, which are needed to renormalize the isovector and isoscalar combinations. Building upon the calculation of Ref. [29], we extend the analysis to the E112 ensemble with the smallest lattice spacing and further improve the methodology following Refs. [30, 31]. A comprehensive analysis of the nonsinglet renormalization factors will be presented in a forthcoming ETMC publication [32]. While both singlet and nonsinglet renormalization factors are employed, we describe here the analysis of renormalization factors for singlet operators and refer the reader to Ref. [32] for the details of the analysis for nonsinglet ones. Nevertheless, in this work, we use the values of both singlet and nonsinglet renormalization constants and report their values below.1
The RI-MOM scheme is defined on amputated vertex functions of the operators under study with external offshell quark states in Landau gauge,
where . Following continuum perturbation theory [33], we project onto the “tree-level” structures of and and impose the following conditions to extract the corresponding renormalization factors,
where, by considering only the non-zero components of the momentum , the projectors read
with . Furthermore, denotes the RI-MOM scale, and is the renormalization factor of the quark field defined by [15]
where is the quark propagator in the momentum space.
The vertex function and quark propagator are calculated using momentum sources [34] with 10-30 configurations. The singlet vertex function has an additional disconnected contribution, which is calculated using the stochastic methods described in Sec. II A.
We employ gauge ensembles specifically generated for our renormalization program [29], using mass-degenerate, clover-improved twisted-mass fermions. These ensembles cover the same four lattice spacings as the physical-point ensembles used in the calculation of the matrix elements. At each lattice spacing, we consider four ensembles with different values of the twisted-mass parameter , enabling an extrapolation to the chiral limit. The simulation parameters for the three coarsest lattice spacings are given in Ref. [29], while those for the lattice spacing of the E112 ensemble are listed in Table IV. We find that the dependence on is mild, consistent with previous studies by our collaboration (see, e.g., Ref. [35]). We account for this dependence by performing both constant and linear fits in and combining the results through model averaging based on the Akaike information criterion (AIC) [36, 37]. A typical example is illustrated in Figure 1 at , for the singlet renormalization factors.
Figure 1. Results for the renormalization factors (left) and (right) in the RI-MOM scheme at the scale , as a function of . Blue circles, yellow triangles, green diamonds and red squares show, respectively, results for the ensembles at the lattice spacings of the B64, C80, D96 and E112 gauge ensembles, accompanied by the same-colored band showing the AIC-average value using constant and linear fits in . The extrapolated values in the chiral limit are given at .
TABLE IV. ensembles with the lattice spacing of the E112 ensemble and their parameters used for the renormalization: , lattice volume , twisted-mass parameter , hopping parameter , and clover coefficient . For the B64, C80 and D96 ensembles the corresponding parameters are given in Table III of Ref. [29].
| Ensemble | |||||
| cE4.035.48 | 1.960 | 0.0035 | 0.13738600 | 1.5792 | |
| cE4.043.48 | 1.960 | 0.0043 | 0.13738600 | 1.5792 | |
| cE4.056.48 | 1.960 | 0.0056 | 0.13738600 | 1.5792 | |
| cE4.070.48 | 1.960 | 0.0070 | 0.13738600 | 1.5792 |
Table IV.
To suppress rotational breaking lattice effects, we employ spatially isotropic momenta satisfying the “democratic” criterion: . In addition, we improve the nonperturbative results by subtracting one-loop discretization effects from both and , computed in lattice perturbation theory to all orders in the lattice spacing (see, e.g., [35]).
After chiral extrapolation and artifact subtraction, the resulting renormalization factors are evolved from to a large reference scale using continuum perturbation theory [33, 38, 39, 40], as described in Ref. [29]. Residual lattice artifacts depending on the renormalization scale are removed through polynomial fits in . Compared to our analysis in Ref. [29], in the current work, we perform a joint fit in using all four lattice spacings, rather than fitting each ensemble separately. To accommodate the combined dataset, higher-order terms in are included in the fit Ansatz. Specifically, we consider
where is the evolution function defined in Ref. [29]. These simultaneous fits provide an improved description of the data.
We employ several fit ranges within for the nonsiglet , while we use for the singlet and the tensor renormalization functions. Lower momenta are excluded from the analysis, as they suffer from significant hadronic contamination and the perturbative evolution is not reliable in this low-momentum region. The extrapolated values at from all fits are combined using model averaging based on AIC. Figure 2 displays the extrapolation to for the singlet renormalization factors and .
Figure 2. Results for the renormalization factors (left), (right) in the RI-MOM scheme at the scale , as a function of . Blue circles, yellow triangles, green diamonds and red squares show, respectively, results for the ensembles with the lattice spacings of B64, C80, D96 and E112 gauge ensembles, accompanied by the same-colored band showing the AIC-average value. Polynomial fits in are employed in the data for several fit ranges as described in the text. The extrapolated values for each lattice spacing are given at .
The extrapolated values at are converted to the scheme at the reference scale , using an intermediate Renormalization Group Invariant (RGI) scheme (see Ref. [29]). The final values for both singlet and nonsinglet renormalization factors in the scheme at and for each lattice-spacing, are reported in Table V.
TABLE V. Renormalization factors in the scheme at for the nonsinglet and singlet axial (second and fourth rows) and tensor (third and fifth rows) operators used for the B64, C80, D96 and E112 ensembles. The numbers in parentheses are the total errors, obtained by adding the statistical and systematic uncertainties in quadrature. The systematic uncertainty is determined from the AIC procedure.
| B64 | C80 | D96 | E112 | |
| 0.7626(10) | 0.7738(11) | 0.7851(11) | 0.7944(12) | |
| 0.8405(71) | 0.8632(39) | 0.8864(29) | 0.9056(29) | |
| 0.7579(74) | 0.7725(73) | 0.7911(50) | 0.7993(37) | |
| 0.8433(40) | 0.8645(40) | 0.8872(40) | 0.9058(36) |
Table V.
Fitting Procedure
The two-step approach
To extract the ground-state matrix elements, we employ a two-step fitting strategy.
In the first step, we fit the effective mass,
constructed from the highest-statistics two-point functions, using Eq. (23) and Eq. (9) with
which we refer to as a two-state fit. We look for convergence as a function of the smallest time separation used in the fit. The extracted values of the ground-state energy and the energy gap , together with the corresponding ratio of overlap factors entering Eq. (24) are subsequently used as fixed inputs in the analysis of the ratio.
In the second step, we construct the ratio of Eq. (13) by dividing each three-point function by the corresponding two-point function computed with the same statistics in order to take advantage of the correlations between them. The ratio is then fitted using either a two-state fit or to a constant when there is no detectable contamination due to excited states. We refer to the latter as a plateau fit.
For the two-state fit we use
where the charge and , the ratios of overlap factors of the three-point function. As mentioned, we fix the values of , , from the effective mass fit in the first step.
For the isovector axial case, our fit incorporates the temporal component of the three-point function carrying one unit of momentum transfer, , where denotes the spatial extent of the lattice and is a unit vector along the -th spatial direction, together with the corresponding two-point function with one unit of momentum. Including in the analysis the three-point function of the temporal axial current is motivated by chiral perturbation theory [41, 42], which predicts an enhancement of the overlap between a nucleon and a pion-nucleon state, namely the and coefficients. In our setup, the temporal component of the axial operator comes at no extra computational cost, and because it displays a strong sensitivity to the excited state, it enables a more precise determination of the excited state energy of the lowest state and hence a better extraction of the desired ground-state matrix element.
Specifically, we form a ratio using and the matched statistics two-point functions at and , and the effective energy at maximum statistics using Eq. (23) with . The two-state fit function for the ratio of the temporal component we use is
where the three-point function is
The two-point functions entering the ratio are
and
with
In this case, the first-excited-state energy gap of the three-point function at zero momentum transfer is shared with the corresponding one for the three-point function for temporal axial current. For the two-point function, is fixed from the effective mass fit.
In all cases, we compare our approach to the results from the summation method where we retain only the ground state. Specifically, we perform a linear fit in using the following summed ratio
for the axial charge and similarly for for the tensor.
Extraction of ground state matrix elements
A detailed description of the analysis of the nucleon effective mass using the approach explained in Sec. IV can be found in Ref. [43]. We use the same procedure here to obtain the needed parameters for the first step in our approach. We illustrate our results in Figure 3 for the B64 ensemble. We observe that after fm the extracted values of the ground-state energy and , are stable. This is valid for all gauge ensembles. We thus fix fm.

Figure 3. Two-state fit results for the nucleon effective mass, for the B64 ensemble. The upper panel shows results on the effective nucleon mass as a function of the source-sink time separation . The lower-left panel shows results on the nucleon mass extracted from two-state fits as a function of the minimum fit time . The lower-right panel shows the corresponding first excited-state energy gap . Open symbols indicate the selected fit results, while the red bands show the value from the selected two-state fit.
Results on the fits carried out for the extraction of the nucleon ground state matrix element for the isovector, isoscalar connected and disconnected, strange and charm axial charges are shown, respectively, in Figure 4-8 for the four ensembles analyzed here. As mentioned previously, we obtain the desired charge from the matrix element of the appropriate current at zero momentum transfer.

Figure 4. Results for the isovector axial charge . From top to bottom we show results for the B64, C80, D96, and E112 ensembles. Left panels show the ratio as a function of for different source-sink time separations . The middle panels show the midpoint value of the ratio as a function of using the same symbols for each time separation as in the left panels. The right panels show the values extracted from two-state fits (red squares) as a function of the smallest value of , , used in the fits. The open symbols indicate the selected fit. The horizontal bands correspond to the chosen fit result. In the left and middle panels, the gray bands show the resulting curves using the selected two-state fit parameters. The filled symbols in the middle panels represent the time separations used in the best fit that is selected. Results from the summation method (green circles) are also shown for comparison.
As can be seen in Figure 4, the ratio with numerator the connected three-point function, from which the isovector combination to the axial charge is determined, exhibits a considerable dependence on , which indicates sizable excited state contributions. For this case, we therefore perform two-state fits, following the specific fitting approach of Sec. IV. We perform different fits within the range , eliminating time slices from the source and sink to identify convergence to a time-independent value. We find that taking a time distance of fm from the source and sink is an appropriate choice. We also vary the source-sink time separations satisfying . We find that fm is a suitable value for all ensembles.
In Figure 5, we show results on the ratio from which the connected contribution of the isoscalar axial charge is determined. We observe that, apart from some time-dependence near the source and sink, it does not exhibit strong dependence on and thus we perform a plateau fit. This can be seen from fitting to a constant at each time separation in order to probe the dependence on , which is shown in the middle panels. We use fm to remove the time-dependence near the sink and source. We observe that the extracted values for fm are consistent. Thus, we perform a final plateau fit to all values of , starting from fm, while using the same . As can be seen, the values are stable as we vary , and they agree with the obtained values when fitting each separately, for our picked range. We also perform fits using the summation method as an al-

Figure 5. Results for the ratios with the three-point ratio of the connected isoscalar axial current (left panels). The middle panels show the values extracted from plateau fits to each source-sink separation individually as a function of . The right panels show the results from plateau fits (red squares) including all source-sink separations satisfying , where the open symbols indicate the selected fit. The band corresponds to the chosen fit. Results from the summation method (green circles) are also shown for comparison. The rest of the notation is the same as in Fig. 4.
alternative procedure. This gives statistically noisier, but compatible results with the plateau fit results.
The ratios constructed from disconnected three-point functions, namely for the isoscalar and the strange and charm quarks, are shown in Figure 6, Figure 7 and Figure 8. Since no significant excited state effects are observed, we perform plateau fits using the same ranges as for the ratios with the isovector three-point function, namely, we take and .

Figure 6. Results for the disconnected contribution of the isoscalar axial charge . The notation is the same as in Fig. 5.

Figure 7. Results for the strange contribution to the axial charge. The notation is the same as in Fig. 5.

Figure 8. Results for the charm contribution to the axial charge. The notation is the same as in Fig. 5.
In Figure 9 and Figure 10 we show our results from the analysis of the ratio for the connected isovector and isoscalar contributions to the tensor charge respectively. We observe strong excited state contamination in both quantities, thus we perform two-state fits, using and for all ensembles.

Figure 9. Results for the isovector tensor charge . The notation is the same as in Fig. 4.

Figure 10. Results for the connected contribution of the isoscalar tensor charge . The notation is the same as in Fig. 4.
The disconnected contributions to the tensor charge are shown in Figure 11–Figure 13. We observe no detectable

Figure 11. Results for the disconnected contribution of the isoscalar tensor charge . The notation is the same as in Figure 5.

Figure 13. Results for the charm contribution to the tensor charge. The notation is the same as in Figure 5.
excited state contamination within the accuracy of our results. Thus, we perform plateau fits using the same and .
Continuum extrapolation
In the twisted mass formulation and after renormalization, the leading lattice discretization effect are of . Thus, we perform linear extrapolations in in order to extract continuum results. In the cases where we observe no statistically significant -dependence in our data, so we also include constant fits in our analysis and take the model average using the Akaike information criterion (AIC). Namely, we use two linear fits in , either using all four ensembles or omitting the coarser ensemble, and three constant fits using all ensembles or the two and three finer ones. A strong dependence on the lattice spacing will result in a model average favoring the linear fits, while a mild dependence will lead to a model average favoring the constant fits.

Figure 12. Results for the strange contribution to the tensor charge. The notation is the same as in Figure 5.
We form the non-singlet combinations , , and the singlet . We renormalize multiplicatively with the factors of Table V, using and for the non-singlets and and for the singlet. An example extrapolation for the isovector tensor charge is shown in Figure 14, along with the weights of the five models used. In the left panels of Figure 15 and Figure 16 we show the extrapolations for all combinations, for the axial and tensor charge respectively.
Figure 14. Continuum extrapolation of the isovector tensor charge . In the left panel the filled points are the renormalized values for each of the four ensembles. The red band and the star show the AIC model averaged results using linear and constant fits. In the right panel, we show the weights of each extrapolation model as a %-probability. The red band is the AIC model averaged result at the continuum.
Figure 15. Continuum extrapolations of the non-singlet and the singlet quark combinations for the axial charge. The left panel shows the results using the analysis as described in Sec. IV C, with the stars showing the values at , where the inner error is the statistical and the outer the total uncertainty after adding the systematic error in quadrature. In the right column we show the results of the analysis from which the systematic error is obtained, as explained in Sec. IV D.
Figure 16. Continuum extrapolations of the non-singlet and the singlet quark combinations for the tensor charge. The notation is the same as in Figure 15.
Using results for the three non-singlet and the singlet quark combinations at the continuum limit, we reconstruct the individual quark-flavor contributions , , , . In the left panels of Figure 17 and Figure 18, we show the results for the flavor decomposition, for the axial and tensor charges, respectively.
Figure 17. Continuum results for the flavor decomposition of the axial charge. The stars show the values at , obtained using the continuum values of the non-singlet and singlet combinations. The filled symbols are the decomposed values per ensemble, used only for visualization. The bands show the linear combination of the fit bands of the respective isovector and isoscalar combinations in Figure 15. The left and right panels show results from the analyses as described in Figure 15.
Figure 18. Continuum results for the flavor decomposition of the tensor charge. The notation is the same as in Figure 17.
Systematic uncertainties
We investigate systematics arising from the analysis of excited states, where no AIC average was done. For the analysis of connected contributions we will use the summation method to estimate the systematic error. We take the value at fm where the results from two-state fits show a general agreement. For the connected isoscalar axial case , where convergence to the two-state results occurs earlier due to the absence of excited state contamination, we take the value at fm. For the disconnected contributions, instead of performing a single plateau fit to all selected values of simultaneously, we fit to a constant at each value and take the weighted average after convergence, which is typically observed for the same values of as the common fit to all separations. The results from this varied analysis are shown in the right panels of Figure 15–Figure 18.
The systematic uncertainty on the continuum value is computed by taking the absolute difference between the central value of the preferred analysis and the one when changing the procedure as described above, after continuum extrapolation.
Comparison with other results
In Tables VI and Table VII, we give our results for each gauge ensemble as well as our final continuum limit extrapolated results for the axial and tensor charges, respectively. The continuum values are given with statistical and systematic errors.
| B64 | 0.913(24) | 0.517(17) | 0.520(18) | 0.515(18) | 0.514(18) |
| C80 | 0.918(23) | 0.535(15) | 0.538(15) | 0.532(16) | 0.533(15) |
| D96 | 0.942(17) | 0.548(12) | 0.550(12) | 0.553(13) | 0.546(13) |
| E112 | 0.945(15) | 0.549(10) | 0.552(10) | 0.557(11) | 0.545(11) |
| 0.951(21)(5) | 0.556(16)(5) | 0.558(16)(6) | 0.570(22)(5) | 0.550(15)(8) | |
| B64 | 0.715(19) | -0.1986(70) | -0.0018(17) | -0.0003(21) | |
| C80 | 0.726(19) | -0.1916(58) | -0.0014(12) | 0.0003(21) | |
| D96 | 0.745(14) | -0.1971(45) | -0.0007(10) | -0.0018(22) | |
| E112 | 0.747(12) | -0.1980(45) | -0.00150(95) | -0.0032(21) | |
| 0.753(18)(5) | -0.1973(61)(3) | -0.0010(15)(4) | -0.0050(37)(31) |
Table VII. Values for the tensor non-singlet and singlet combinations (top) and the extracted single flavor charges (bottom) for each ensemble and in the continuum limit. The notation is the same as in Table VI.
Our continuum extrapolated results are compared with other recent lattice QCD calculations. We emphasize that our results are the only ones among those considered here for which the continuum limit is obtained exclusively from ensembles simulated directly at the physical pion mass. In contrast, the other studies rely either entirely on ensembles with heavier-than-physical pion masses or on a combination of one or two ensembles simulated at physical pion mass with ensembles at heavier pion masses.
combination of one or two ensembles simulated at physical pion mass with ensembles at heavier pion masses. Since simulations at physical pion mass generally have larger statistical uncertainties, the corresponding continuum and chiral extrapolations can be more strongly constrained by the heavier-than-physical ensembles than by the data using the ensembles with physical pion mass. Given that chiral extrapolations of nucleon observables are known to be less reliable than those in the meson sector [56], such an approach can introduce additional systematic uncertainties that are difficult to quantify. Our analysis avoids this source of uncertainty by using ensembles simulated directly at the physical pion mass and therefore does not require a chiral extrapolation.
In Figure 19, Figure 17 and Figure 18 we compare our results for the nucleon charges with values from other collaborations. Below we give some details on these recent lattice QCD studies.
Figure 19. Comparison of the results of this work with other lattice QCD results, for the isovector axial (top) and tensor (bottom) charges. Our results are shown with the red square and red error band. The pink cross shows the result from CalLat [44], the magenta triangles from the Mainz group [45], the green triangles from PNDME [46], the brown star from CLQCD [47], the cyan triangles from RQCD [48], the black diamonds from the QCDSF/UKQCD/CSSM collaboration [49], the orange crosses from NME [50] and the yellow hexagon from QCD [51]. For , the dashed line and gray band represents the PDG value [52].
| B64 | 1.315(39) | 0.473(12) | 0.5426(62) | 0.459(10) | 0.434(18) |
| C80 | 1.230(17) | 0.475(15) | 0.5350(44) | 0.452(11) | 0.442(24) |
| D96 | 1.241(25) | 0.461(18) | 0.5337(43) | 0.439(12) | 0.417(32) |
| E112 | 1.277(29) | 0.454(19) | 0.5273(58) | 0.432(13) | 0.409(34) |
| 1.263(34)(26) | 0.458(16)(1) | 0.5280(77)(101) | 0.429(19)(15) | 0.417(25)(12) | |
| B64 | 0.895(21) | -0.420(19) | -0.0341(48) | -0.0062(28) | |
| C80 | 0.852(12) | -0.378(11) | -0.0301(64) | -0.0023(42) | |
| D96 | 0.850(16) | -0.391(15) | -0.0371(85) | -0.0055(60) | |
| E112 | 0.865(18) | -0.413(17) | -0.0375(88) | -0.0059(65) | |
| 0.860(19)(7) | -0.404(19)(19) | -0.0360(70)(10) | -0.0030(46)(8) |
Table 6. Values for the axial non-singlet and singlet combinations (top) and the extracted single flavor charges (bottom) for each ensemble and in the continuum limit, using the model average strategy described in the text. For the continuum values, the first error is statistical and the second the systematic due to the excited state analysis.
The CalLat collaboration determined using an approach based on the Feynman-Hellmann theorem [44]. They used a mixed-action setup with highly improved staggered quarks (HISQ) in the sea, generated by the MILC collaboration, and Möbius domain-wall fermions as valence quarks, with the gauge links smoothed using gradient flow. The analysis includes eighteen ensembles spanning three lattice spacings (, , fm) and pion masses in the range of approximately 130–400 MeV. The nucleon axial charge is extracted from multi-state fits to the nucleon two-point and Feynman-Hellmann correlation functions. The final value, denoted as CalLat25 in Figure 19, was obtained through a model-averaged analysis accounting for chiral, discretization, and finite-volume effects.
The PNDME collaboration analyzed thirteen ensembles, which belong to the same set used by CalLat, using highly improved staggered quarks (HISQ) in the sea and clover-improved valence quarks in a mixed-action setup for the computation of the nucleon correlation functions. The ensembles cover four lattice spacings, 0.06 fm, 0.09 fm, 0.12 fm and 0.15 fm, three pion masses (135 MeV, 220 MeV and 310 MeV), and volumes with . Two of these ensembles are at the physical pion mass. A combined chiral, continuum and finite-volume fit was performed to obtain the isovector nucleon charges at the physical point. Their results are denoted as PNDME23 [46] in Figure 19.
The PNDME collaboration has recently reported updated results for the flavor-diagonal up, down and strange axial and tensor charges [53]. The calculation includes both connected and disconnected contributions and incorporates an analysis of excited-state effects including low-lying states. The final results are obtained from a simultaneous chiral, continuum and finite-volume analysis. These results are denoted as PNDME25 in Figure 20 and Figure 21.
Figure 20. Comparison of our results for the up- (left), down- (center), and strange-quark (right) contributions to the axial charge with other lattice QCD studies. Our results are shown with the red square and red error band. The green triangles show results from PNDME [53] and the yellow hexagons from QCD [51]. For the Mainz group the magenta filled triangles show results from [54], while the magenta open triangles from conference proceedings [55].
Figure 21. Comparison of our results for the up- (left), down- (center), and strange-quark (right) contributions to the tensor charges with other lattice QCD analyses. Our results are shown with the red square and red error band. The green triangles show results from PNDME [53] and the magenta open triangles from the Mainz group from conference proceedings [55].
The RQCD collaboration performed an extensive study of nucleon charges [48], analyzing forty-seven ensembles with clover-improved fermions, most of which were generated within the Coordinated Lattice Simulations (CLS) effort. The ensembles cover pion masses from 130 MeV to 430 MeV, six lattice spacings ranging from approximately 0.039 fm to 0.098 fm, and volumes with , including two ensembles with pion masses close to the physical value. Extrapolations to the physical point were performed using leading-order chiral expressions for the pion-mass dependence, the leading asymptotic form for finite-volume corrections, and terms up to in the lattice spacing. Their final values are denoted by RQCD23 in Figure 19.
The Mainz collaboration investigated the isovector charges in Ref. [45] using fifteen CLS ensembles, covering pion masses from 130 MeV to 360 MeV, four lattice spacings, and volumes with . A simultaneous extrapolation in the lattice spacing, pion mass, and volume was performed, with one ensemble having a pion mass slightly below the physical value. Their isovector charge results are denoted by Mainz24 in Figure 19.
In addition, the Mainz collaboration reported a determination of the up, down, and strange axial and tensor charges in conference proceedings [55] on a subset of eight ensembles. These ensembles have four lattice spacings between 0.050 fm and 0.086 fm, volumes with , and pion masses in the range 200 MeV–360 MeV. No ensembles with physical pion mass were included in this analysis. A combined chiral and continuum-limit fit was used to obtain the physical-point results. These results are denoted by Mainz19 in Figure 20 and Figure 21. More recently, the Mainz collaboration published a determination of the strange axial form factor [54]. This calculation uses a subset of fourteen ensembles with pion masses in the range 130 MeV–350 MeV and four lattice spacings ranging from 0.050 fm to 0.086 fm. The strange axial charge is obtained from the strange axial form factor in the limit. In Figure 20, we denote this newer strange axial result as Mainz26.
The QCDSF/UKQCD/CSSM collaboration [49] analyzed twenty-one gauge ensembles, employing a tree-level Symanzik improved gluon action together with nonperturbatively clover-improved Wilson fermions. The ensembles cover pion masses in the range , five lattice spacings, and three lattice volumes. Combined extrapolations in the pion mass, lattice spacing, and volume were used to obtain their isovector charges, shown in Figure 19.
The NME collaboration determined the isovector charges in Ref. [50] using seven ensembles with clover-improved Wilson fermions. The ensembles span pion masses of , lattice spacings from 0.071 fm to 0.127 fm, and volumes satisfying . The physical-point results were obtained through combined chiral, continuum, and infinite-volume extrapolations. These results are denoted by NME21 in Figure 19.
The CLQCD collaboration, recently, reported a high-precision determination of the isovector tensor charge [47]. They analyzed fifteen lattice ensembles, employing tadpole-improved tree-level Symanzik and Clover actions for the gauge and fermion sectors, respectively. The ensembles span five lattice spacings, 0.052 fm 0.105 fm, with pion masses in the range 134 MeV $\le m_\pi \le
341 MeV and volumes with . Five sets of ensembles have the same quark masses and lattice spacing but different volumes, including three ensembles at the physical pion mass. Their final result includes systematic uncertainties from the continuum, infinite-volume and chiral extrapolations, as well as excited-state contamination and renormalization, and are denoted as CLQCD25 in Figure 19.
The QCD collaboration determined the isovector as well as the up, down, and strange axial charges [51] using a hybrid setup with domain wall fermions in the sea and overlap fermions in the valence sector. Their analysis included three RBC/UKQCD gauge ensembles at three lattice spacings, fm, fm, and fm, two volumes, and three pion masses of 171 MeV, 302 MeV, and 337 MeV. Simultaneous continuum, chiral, and volume extrapolations were performed, with the resulting values denoted by QCD18 in Figure 19 and Figure 20.
In Figure 19, we compare our results for the isovector axial and tensor charges with recent lattice QCD calculations. For the isovector axial charge, our result is in good agreement with the results from the other collaborations and compatible with the experimental determination [52]. The isovector tensor charge shows a somewhat larger spread among the lattice calculations. Our result agrees with all other results within one standard deviation, except with the CLQCD value, which is larger than ours.
For the flavor-diagonal charges, shown in Figure 20 and Figure 21, we find overall good agreement with the available lattice results. The tensor charges show particularly good consistency among the different calculations. For the axial charges, our results are also generally consistent with the other determinations. The main visible difference is in the up-quark axial charge, where the PNDME result is smaller than ours. The PNDME collaboration obtains their quoted up and down axial charges by averaging the results of their “standard” and “” analyses, a procedure not followed by the other collaborations.
In Figure 22, we present our result for the non-singlet axial charge combination , together with the corresponding value obtained exclusively from the connected contributions. The latter allows for a comparison with the value of extracted from hyperon decays [57, 58], using the Cabibbo Model and under the assumption of SU(3)-flavor symmetry. In the SU(3)-symmetric limit, the disconnected contributions cancel in this combination, motivating the comparison with the result obtained from the connected only. This provides a useful consistency check between our lattice determination and the phenomenological extraction from hyperon decays, while showing the effects of SU(3)-flavor symmetry breaking.
Figure 22. Comparison of our result for the non-singlet axial charge combination with the value obtained only from the connected contributions, and the value of extracted from hyperon decays using the Cabibbo Model and assuming SU(3)-flavor (SU(3)) symmetry [57, 58]. The red squares show our results for , while the blue circle shows the phenomenological value of .
Summary and Conclusions
The nucleon axial and tensor charges are determined using four twisted-mass clover-improved ensembles with quark masses tuned closely to their physical values. The calculation includes both connected and disconnected contributions, allowing us to determine the isovector, isoscalar, and individual flavor contributions to the axial and tensor charges. These results provide an update to our previous analysis [29] using three ensembles, yielding smaller errors and more stability to the continuum extrapolation.
Our results show overall good agreement with other recent lattice QCD determinations. In particular, the isovector axial charge is consistent with both the available lattice results and the experimental determination. The flavor-diagonal axial and tensor charges are also generally consistent with the results from other lattice calculations, while the isovector tensor charge shows a somewhat larger spread among the different determinations.
The continuum limit is obtained using a model average of constant and linear-in- fits. Although the lattice-spacing dependence is mild, the extrapolation to has a non-negligible effect and is therefore important for obtaining the physical values, together with a comprehensive treatment of the systematic uncertainties due to excited-state contamination. Importantly, the continuum extrapolation is performed using ensembles simulated at approximately the physical pion mass, avoiding the need for a chiral extrapolation and the possible associated uncontrolled systematics.
ACKNOWLEDGMENTS
We thank all members of the ETM Collaboration for the most enjoyable collaboration. We thank Roberto Frezzotti, Giuseppe Gagliardi and Vittorio Lubicz, for useful discussions and crosschecks on the analysis of the renormalization factors. C.A., S.B, A.E, C.I., F.P. and G.S. acknowledge partial support from the projects IMAGE-N (EXCELLENCE/0524/0459), MuonHVP (EXCELLENCE/0524/0017), PulseQCD (EXCELLENCE/0524/0269), DeNuTra (EXCELLENCE/524/0455), HiQO (POST-DOC/0925-ROandOther/0148), Baryon8 (POSTDOC/0524/0001), BSM (CULTURE/AWARD-YR/0524 B/0002), StrongILA (EXCELLENCE/0524/0001), and partonWF (VISION ERC/0525/0010), funded by the European Regional Development Fund and the Republic of Cyprus through the Cyprus Research and Innovation Foundation. C.A. acknowledges partial support from the European Joint Doctorate AQTIVATE that received funding from the European Union’s research and innovation program under the Marie Sklodowska-Curie Doctoral Networks action, Grant Agreement No 101072344. The authors gratefully acknowledge the Gauss Centre for Supercomputing e.V. (www.gauss-centre.eu) for funding this project by providing computing time through the John von Neumann Institute for Computing (NIC) on the GCS Supercomputers JUWELS, JUWELS Booster [59] and JUPITER Booster at Jülich Supercomputing Centre (JSC). The authors acknowledge access to the JUPITER supercomputer, which is funded by the EuroHPC Joint Undertaking, the German Federal Ministry of Research, Technology and Space, and the Ministry of Culture and Science of the German state of North Rhine-Westphalia, through the JUPITER Research and Early Access Program (JUREAP) as part of the EuroHPC project application EHPC-EXT-2023E02-052.
References
References
- [1]T. Bhattacharya, V. Cirigliano, S. D. Cohen, A. Filippuzzi, M. Gonzalez-Alonso, M. L. Graesser, R. Gupta, and H.-W. Lin, Probing Novel Scalar and Tensor Interactions from (Ultra)Cold Neutrons to the LHC, Phys. Rev. D85, 054512 (2012), arXiv:1110.6448 [hep-ph].DOI
- [2]I. Bischer and W. Rodejohann, General Neutrino Interactions at the DUNE Near Detector, Phys. Rev. D 99, 036006 (2019), arXiv:1810.02220 [hep-ph].
- [3]D. Akimov et al. (COHERENT), Observation of Coherent Elastic Neutrino-Nucleus Scattering, Science 357, 1123 (2017), arXiv:1708.01294 [nucl-ex].DOI
- [4]A. Beda, V. Brudanin, V. Egorov, D. Medvedev, V. Pogosov, M. Shirchenko, and A. Starostin, The results of search for the neutrino magnetic moment in GEMMA experiment, Adv. High Energy Phys. 2012, 350150 (2012).
- [5]T. Marrodán Undagoitia and L. Rauch, Dark matter direct-detection experiments, J. Phys. G 43, 013001 (2016), arXiv:1509.08767 [physics.ins-det].
- [6]M. V. Chizhov, Search for tensor interactions in kaon decays at DAPHNE, Phys. Lett. B 381, 359 (1996), arXiv:hep-ph/9511287.DOI
- [7]L. Dhargyal, New tensor interaction as the source of the observed CP asymmetry in τ → K_Sπν_τ, Springer Proc. Phys. 203, 329 (2018), arXiv:1716.06293 [hep-ph].
- [8]C. Couzza, A. Metz, D. Pitonyak, A. Prokudin, N. Sato, and R. Seidl (JAM), Transversity Distributions and Tensor Charges of the Nucleon: Extraction from Dihadron Production and their Universal Nature, Phys. Rev. Lett. 132, 091901 (2024), arXiv:2306.12998 [hep-ph].
- [9]C. Alexandrou et al. (Extended Twisted Mass), Strange and charm quark contributions to the muon anomalous magnetic moment in lattice QCD with twisted-mass fermions, Phys. Rev. D 111, 054502 (2025), arXiv:2411.08852 [hep-lat].arxiv.org/abs/2411.08852
- [10]R. Frezzotti, P. A. Grassi, S. Sint, and P. Weisz (Alpha), Lattice QCD with a chirally twisted mass term, JHEP 08, 058, arXiv:hep-lat/0101001 [hep-lat].
- [11]R. Frezzotti and G. C. Rossi, Chirally improving Wilson fermions. 1. O(a) improvement, JHEP 08, 007, arXiv:hep-lat/0306014 [hep-lat].
- [12]B. Sheikholeslami and R. Wohlert, Improved Continuum Limit Lattice Action for QCD with Wilson Fermions, Nucl. Phys. B259, 572 (1985).DOI
- [13]C. Alexandrou et al., Simulating twisted mass fermions at physical light, strange and charm quark masses, Phys. Rev. D98, 054518 (2018), arXiv:1807.00495 [hep-lat].DOI
- [14]J. Finkenrath et al., Twisted mass gauge ensembles at physical values of the light, strange and charm quark masses, PoS LATTICE2021, 284 (2022), arXiv:2201.02551 [hep-lat].arxiv.org/abs/2201.02551
- [15]C. Alexandrou et al. (Extended Twisted Mass), Quark masses using twisted-mass fermion gauge ensembles, Phys. Rev. D 104, 074515 (2021), arXiv:2104.13408 [hep-lat].DOI
- [16]W. Detmold, W. Melnitchouk, J. W. Negele, D. B. Renner, and A. W. Thomas, Chiral extrapolation of lattice moments of proton quark distributions, Phys. Rev. Lett. 87, 172001 (2001), arXiv:hep-lat/0103006.DOI
- [17]C. W. Bernard, T. Draper, G. Hockney, and A. Soni, CALCULATION OF WEAK MATRIX ELEMENTS: SOME TECHNICAL ASPECTS, in Workshop on Lattice Gauge Theories (Wuppertal, Germany), 1995.
- [18]C. McNeile and C. Michael (UKQCD), Decay width of light quark hybrid meson from the lattice, Phys. Rev. D73, 074506 (2006), arXiv:hep-lat/0603007 [hep-lat].DOI
- [19]A. S. Gambhir, A. Stathopoulos, and K. Orginos, Deflation as a Method of Variance Reduction for Estimating the Trace of a Matrix Inverse, SIAM J. Sci. Comput. 39, A532 (2017), arXiv:1603.05988 [hep-lat].DOI
- [20]C. Alexandrou, S. Bacchio, A. Evangelista, R. Frezzotti, F. Margari, F. Sanfilippo, and C. Schneider, Performance of Low Mode Averaging on Twisted-Mass Fermion Ensembles at the physical pion mass point, (2026), arXiv:2606.27072 [hep-lat].arxiv.org/abs/2606.27072
- [21]A. Stathopoulos, J. Laeuchli, and K. Orginos, Hierarchical Probing for Estimating the Trace of the Matrix Inverse on Toroidal Lattices, SIAM J. Sci. Comput. 35, S299 (2013), arXiv:1302.4018 [hep-lat].arxiv.org/abs/1302.4018
- [23]C. Alexandrou, S. Gusken, F. Jegerlehner, K. Schilling, and R. Sommer, The Static approximation of heavy - light quark systems: A Systematic lattice study, Nucl. Phys. B414, 815 (1994), arXiv:hep-lat/9211042 [hep-lat].DOI
- [24]S. Gusken, A Study of smearing techniques for hadron correlation functions, Nucl. Phys. B Proc. Suppl. 17, 361 (1990).DOI
- [25]C. Alexandrou, S. Bacchio, M. Constantinou, J. Finkenrath, K. Hadjiyiannakou, K. Jansen, G. Koutsou, and A. Vaquero Aviles-Casco, Proton and neutron electromagnetic form factors from lattice QCD, Phys. Rev. D 100, 014509 (2019), arXiv:1812.10311 [hep-lat].DOI
- [26]C. Alexandrou et al., Moments of nucleon generalized parton distributions from lattice QCD simulations at physical pion mass, Phys. Rev. D 101, 034519 (2020), arXiv:1908.10706 [hep-lat].DOI
- [27]M. Albanese et al. (APE), Glueball Masses and String Tension in Lattice QCD, Phys. Lett. B 192, 163 (1987).DOI
- [28]G. Martinelli, C. Pittori, C. T. Sachrajda, M. Testa, and A. Vladikas, A General method for nonperturbative renormalization of lattice operators, Nucl. Phys. B445, 81 (1995), arXiv:hep-lat/9411010 [hep-lat].DOI
- [29]C. Alexandrou, S. Bacchio, J. Finkenrath, C. Iona, G. Koutsou, Y. Li, and G. Spanoudes, Nucleon charges and σ-terms in lattice QCD, Phys. Rev. D 111, 054505 (2025), arXiv:2412.01535 [hep-lat].arxiv.org/abs/2412.01535
- [30]C. Alexandrou, S. Bacchio, G. Koutsou, B. Prasad, and G. Spanoudes, Proton and neutron electromagnetic form factors in the continuum limit using lattice QCD ensembles with physical pion masses, Phys. Rev. D 113, 114524 (2026), arXiv:2507.20910 [hep-lat].DOI
- [31]C. Alexandrou, S. Bacchio, M. Bode, J. Finkenrath, A. Herten, C. Iona, G. Koutsou, F. Pittler, B. Prasad, and G. Spanoudes, Nucleon strange electromagnetic form factors from N_f = 2 + 1 + 1 lattice QCD (2026), arXiv:2603.26591 [hep-lat].arxiv.org/abs/2603.26591
- [32]Extended Twisted Mass Collaboration, Non-perturbative renormalisation of quark bilinear operators with N_f = 2 + 1 + 1 Wilson-clover twisted mass fermions, In preparation.
- [33]J. A. Gracey, Three loop anomalous dimension of nonsinglet quark currents in the RI-prime scheme, Nucl. Phys. B 662, 247 (2003), arXiv:hep-ph/0304113.DOI
- [34]M. Gockeler, R. Horsley, H. Perlt, D. Petters, P. E. Rakow, A. Schafer, G. Schierholz, and A. Schiller, Nonperturbative renormalization of composite operators in lattice QCD, Nucl. Phys. B 544, 699 (1999), arXiv:hep-lat/9807044.DOI
- [35]C. Alexandrou, M. Constantinou, and H. Panagopoulos (ETM), Renormalization of nonlocal operators for Nf=2 and Nf=4 twisted mass fermions, Phys. Rev. D95, 034505 (2017), arXiv:1509.00213 [hep-lat].DOI
- [36]W. I. Jay and E. T. Neil, Bayesian model averaging for analysis of lattice field theory results, Phys. Rev. D 103, 114502 (2021), arXiv:2008.01069 [stat.ME].DOI
- [37]E. T. Neil and J. W. Sitison, Improved information criteria for Bayesian model averaging in lattice field theory, Phys. Rev. D 109, 014510 (2024), arXiv:2208.14983 [stat.ME].DOI
- [38]K. G. Chetyrkin, Quark mass anomalous dimension to O (alpha-s**4), Phys. Lett. B 404, 161 (1997), arXiv:hep-ph/9703278.DOI
- [39]J. A. Gracey, Tensor current renormalization in the RI’ scheme at four loops, Phys. Rev. D 106, 085008 (2022), arXiv:2208.14527 [hep-ph].arxiv.org/abs/2208.14527
- [40]M. Constantinou, M. Hadjantonis, H. Panagopoulos, and G. Spanoudes, Singlet versus nonsinglet perturbative renormalization of fermion bilinears, Phys. Rev. D94, 114513 (2016), arXiv:1610.06744 [hep-lat].
- [41]O. Bär, Nπ-state contamination in lattice calculations of the nucleon axial form factors, Phys. Rev. D 99, 054506 (2019), arXiv:1812.09191 [hep-lat].DOI
- [42]O. Bär, Nucleon-pion-sector contribution in lattice calculations of the nucleon charges g_A, g_T and g_S, Phys. Rev. D 94, 054505 (2016), arXiv:1606.09385 [hep-lat].DOI
- [43]C. Alexandrou, S. Bacchio, J. Finkenrath, C. Iona, G. Koutsou, C. Kummer, Y. Li, B. Prasad, and G. Spanoudes, Nucleon unpolarized second Mellin moments using lattice QCD ensembles with physical quark masses and in the continuum limit (2026), arXiv:2607.20337 [hep-lat].arxiv.org/abs/2607.20337
- [44]Z. B. Hall et al., Signs of nonmonotonic finite-volume corrections to g_A, Phys. Rev. C 113, 055206 (2026), arXiv:2503.09891 [hep-lat].arxiv.org/abs/2503.09891
- [45]D. Djukanovic, G. von Hippel, H. B. Meyer, K. Ottnad, and H. Wittig, Improved analysis of isovector nucleon matrix elements with Nf=2+1 flavors of O(a) improved Wilson fermions, Phys. Rev. D 109, 074507 (2024), arXiv:2402.03024 [hep-lat].arxiv.org/abs/2402.03024
- [46]Y.-C. Jang, R. Gupta, T. Bhattacharya, B. Yoon, and H.-W. Lin (Precision Neutron Decay Matrix Elements (PNDME)), Nucleon isovector axial form factors, Phys. Rev. D 109, 014503 (2024), arXiv:2305.11330 [hep-lat].arxiv.org/abs/2305.11330
- [47]J.-H. Wang, Z.-C. Hu, X. Ji, X. Jiang, Y. Su, P. Sun, and Y.-B. Yang (CLQCD), Precision determination of nucleon iso-vector scalar and tensor charges at the physical point (2025), arXiv:2511.02326 [hep-lat].arxiv.org/abs/2511.02326
- [48]G. S. Bali, S. Collins, S. Heybrock, M. Löffler, R. Rödl, W. Söldner, and S. Weißhäupl (RQCD), Octet baryon isovector charges from Nf=2+1 lattice QCD, Phys. Rev. D 108, 034512 (2023), arXiv:2305.04717 [hep-lat].DOI
- [49]R. E. Smail et al. (QCDSF/UKQCD/CSSM), Constraining beyond the standard model nucleon isovector charges, Phys. Rev. D 108, 094511 (2023), arXiv:2304.02866 [hep-lat].
- [50]S. Park, R. Gupta, B. Yoon, S. Mondal, T. Bhattacharya, Y.-C. Jang, B. Joó, and F. Winter (Nucleon Matrix Elements (NME)), Precision nucleon charges and form factors using (2+1)-flavor lattice QCD, Phys. Rev. D 105, 054505 (2022), arXiv:2103.05599 [hep-lat].DOI
- [51]J. Liang, Y.-B. Yang, T. Draper, M. Gong, and K.-F. Liu, Quark spins and Anomalous Ward Identity, Phys. Rev. D 98, 074505 (2018), arXiv:1806.08366 [hep-ph].DOI
- [52]F. Takahashi et al. (Particle Data Group), Review of Particle Physics, Int. J. Mod. Phys. A 41, 2630011 (2026).
- [53]S. Park, R. Gupta, T. Bhattacharya, F. He, S. Mondal, H.-W. Lin, and B. Yoon, Flavor diagonal nucleon charges using clover fermions on MILC HISQ ensembles, Phys. Rev. D 112, 054508 (2025), arXiv:2503.07100 [hep-lat].arxiv.org/abs/2503.07100
- [54]A. Barone, D. Djukanovic, G. von Hippel, H. B. Meyer, K. Ottnad, and H. Wittig, Strange and flavor-singlet axial form factors of the nucleon from lattice QCD, Phys. Rev. D 114, 034520 (2026), arXiv:2605.06559 [hep-lat].arxiv.org/abs/2605.06559
- [55]D. Djukanovic, H. Meyer, K. Ottnad, G. von Hippel, J. Wilhelm, and H. Wittig, Strange nucleon form factors and isoscalar charges with N_f = 2 + 1 O(a)-improved Wilson fermions, PoS LATTICE2019, 158 (2019), arXiv:1911.01177 [hep-lat].arxiv.org/abs/1911.01177
- [56]P. E. Shanahan, Chiral Effective Theory Methods and their Application to the Structure of Hadrons from Lattice QCD, J. Phys. G 43, 124001 (2016), arXiv:1606.08812 [hep-lat].DOI
- [57]F. E. Close and R. G. Roberts, Consistent analysis of the spin content of the nucleon, Phys. Lett. B 316, 165 (1993), arXiv:hep-ph/9306289.DOI
- [58]A. Deur, S. J. Brodsky, and G. F. De Téramond, The Spin Structure of the Nucleon, Rept. Prog. Phys. 82, 076201 (2019), arXiv:1807.05250 [hep-ph].DOI
- [59]Jülich Supercomputing Centre, JUWELS Cluster and Booster: Exascale Pathfinder with Modular Supercomputing Architecture at Juelich Supercomputing Centre, Journal of large-scale research facilities 7, 10.17815/jlsrf-7-183 (2021).









