Abstract
The simplest molecules in nature, molecular hydrogen ions in the form of H_{2}^{+} and HD^{+}, provide an important benchmark system for tests of quantum electrodynamics in complex forms of matter. Here, we report on such a test based on a frequency measurement of a vibrational overtone transition in HD^{+} by laser spectroscopy. We find that the theoretical and experimental frequencies are equal to within 0.6(1.1) parts per billion, which represents the most stringent test of molecular theory so far. Our measurement not only confirms the validity of highorder quantum electrodynamics in molecules, but also enables the long predicted determination of the protontoelectron mass ratio from a molecular system, as well as improved constraints on hypothetical fifth forces and compactified higher dimensions at the molecular scale. With the perspective of comparisons between theory and experiment at the 0.01 partperbillion level, our work demonstrates the potential of molecular hydrogen ions as a probe of fundamental physical constants and laws.
Introduction
The possibility of accurate quantum electrodynamics (QED) calculations and the presence of narrow optical transitions between longlived vibrational levels make laser spectroscopy of H_{2}^{+} and HD^{+} a sensitive tool to test fundamental physics at the molecular scale. Examples include tests of relativistic quantum mechanics and QED^{1,2,3,4}, and searches for physics beyond the Standard Model at the molecular scale^{5,6} and beyond General Relativity^{7,8}. Furthermore, spectroscopy of the molecular hydrogen ion has long been proposed as a means to determine the value of fundamental physical constants^{1,2,4,9}, and to realise precise molecular clocks^{8,10}.
Nearly all the above applications involve a comparison of stateoftheart molecular theory with accurate results from (most often laser) spectroscopy. The most accurate results so far have been obtained using HD^{+} ions stored in radiofrequency (r.f.) traps, sympathetically cooled by lasercooled Be^{+} ions^{3,11}, allowing tests at a relative uncertainty of a few parts per billion (p.p.b.). However, the most precise experimental result^{3}, a frequency measurement of the (v,L): (0,0)–(1,1) rovibrational line at 5.1 μm in HD^{+} with a relative uncertainty of 1.1 p.p.b., was found to disagree by 2.7 p.p.b. with a more accurate theoretical value obtained from stateoftheart ab initio molecular theory^{1,2}. This disagreement has thus far been unresolved, and additional highprecision experimental data are needed to draw conclusions about the validity of the theoretical framework, and to open up the wide range of applications of molecular hydrogen ion spectroscopy mentioned above.
Here, we report on an optical frequency measurement of the (v,L): (0,2)–(8,3) vibrational overtone transition in HD^{+} at 782 nm. Our experimental result is found to be in agreement with the theoretical prediction to within 0.6(1.1) p.p.b., thereby confirming the validity of relativistic quantum mechanics and QED in a vibrating molecular system at an unprecedented level. We subsequently exploit the agreement between theory and experiment for the first determination of the proton–electron mass ratio from a molecular system, and to put tighter constraints on the strength and range of ‘fifth forces’ at the molecular scale.
Results
Experimental procedure
Our experimental apparatus and procedure for HD^{+} spectroscopy are described in detail in the Methods section. In brief, trapped HD^{+} molecular ions are cooled to ∼10 mK by storing them together with lasercooled Be^{+} ions, and we excite and detect the (v,L): (0,2)–(8,3) transition by resonanceenhanced (1+1’) multiphoton dissociation (REMPD) spectroscopy, see Fig. 1a. We acquire a spectrum by observing the loss of HD^{+} because of REMPD, inferred from the change in the monitored Be^{+} fluorescence induced by r.f. excitation of the HD^{+} secular motion, for different values of the 782 nm spectroscopy laser frequency (Methods and Fig. 2a). The latter is stabilized and counted using an optical frequency comb laser (Methods).
Line shape model for nonlinear leastsquares fitting
Hyperfine interactions lead to a manifold of 83 lines (Fig. 1b,c), the ∼25 strongest of which are located within the range (−110 MHz, 140 MHz) around the hyperfineless rovibrational frequency, ν_{0}. Each hyperfine component is Doppler broadened to ∼16 MHz, and the hyperfine structure is only partially resolved. Using a realistic spectral line shape function (Methods), we employ nonlinear least squares fitting to extract not only the transition frequency ν_{0,fit}, but also the intensity of the 782 nm laser, I_{L}, the motional temperature of the HD^{+} ions, T_{HD+}, and the temperature of the Be^{+} ions during secular excitation, T_{i}, which are relevant parameters for the spectral analysis (see Methods for details). The 782nm intensity we find agrees (within the fit error) with the intensity estimated from the laser beam waist and beam alignment uncertainty. Likewise, the fitted temperature agrees well with the results of molecular dynamics (MD) simulations.
The relevance of the additional fit parameters is illustrated by the correlation matrix, Ω, of the estimated parameters, which reveals significant correlations between ν_{0,fit} and T_{HD+}, and ν_{0,fit} and I_{L} (Fig. 2c). Adding fit parameters results in an increased error in the fitted value ν_{0,fit}, which rises from 0.25 MHz (0.65 p.p.b. relative to ν_{0,fit}) to 0.33 MHz (0.85 p.p.b.) when T_{i}, T_{HD+} and I_{L} are added as free fit parameters. Even so, the spectral fit result presented here marks the first time that a subp.p.b. resolution is achieved in molecularion spectroscopy.
Systematic effects and frequency value
The fit result ν_{0,fit} is corrected for various systematic frequency offsets because of electric and magnetic fields. We calculate the a.c. and d.c. Stark effect ab initio with high accuracy following the approach of Karr^{10} to find the frequency shift because of blackbody radiation (BBR), the r.f. trap field and the electric fields of the lasers. The total Stark shift of −1.3(1) kHz is dominated by the shifts due to the 313, 532 and 782 nm lasers (which are all on during excitation), with individual contributions of 0.008(1) kHz, −0.45(7) kHz and −0.87(13) kHz, respectively. The uncertainty of the Stark shift is caused almost entirely by the uncertainties of the laser beam intensities. We also evaluate the Zeeman effect, which stems from the static quantization field of 0.19 mT directed along the z axis of the trap (Methods). The polarization of the 782 nm laser is such that it induces σ^{−} and σ^{+} transitions at equal rates. The Zeeman effect leads to a shift to ν_{0,fit} by −16.9(3.2) kHz, the uncertainty of which is due to a possible 2% imbalance between the σ^{−} and σ^{+} rates caused by imperfect polarization optics and depolarization because of the windows of the vacuum chamber. Another uncertainty stems from the accuracy of the theoretical hyperfine structure^{12}, which enters through our spectral line shape function. We repeated the fit procedure with a spectral line shape function based on the hyperfine structure obtained with slightly altered spin coefficients (within their uncertainty range), for which we observe essentially no shift of ν_{0,fit}. Compared with the 0.33 MHz statistical fit uncertainty of ν_{0,fit}, the contribution of the above line shifts to the total frequency value and uncertainty is negligible, as are the frequency shifts due to the secondorder Doppler effect^{13} (<5 Hz) and the electricquadrupole shift^{14} (<0.1 kHz).
During REMPD, the motional dynamics in the lasercooled Coulomb crystal are significantly influenced by several (laserinduced) chemical processes. For example, REMPD of an HD^{+} ion leads to the formation of energetic fragments, and previously it was observed that the two reactions, HD^{+}+hν_{1}+hν_{2}→H(1s)+D^{+}+k.e. and HD^{+}+hν_{1}+hν_{2}→D(1s)+H^{+}+k.e., occur with about equal probability (here k.e. stands for kinetic energy, and ν_{1} and ν_{2} for the frequencies of the 782 and 532 nm lasers, respectively)^{15}. Owing to their relatively high chargetomass ratio, the produced protons are not stably trapped in our apparatus. Indeed, we observe only cold, trapped deuterons after REMPD, indicating that each deuteron transferred most of its 0.41 eV kinetic energy to the ions in the Coulomb crystal (which itself contains only 2 meV of thermal energy at 10 mK), causing (at least transiently) significant heating of the crystal. Other reactions with energetic ionic products are the result of collisions with H_{2} molecules in the background vapour, and we also take the reactions Be^{+}(^{2}P_{3/2})+H_{2}→BeH^{+}+H(1s)+k.e., HD^{+}+H_{2}→H_{2}D^{+}+H(1s)+k.e. and HD^{+}+H_{2}→H_{3}^{+}+D(1s)+k.e. (with ionic product kinetic energies of 0.25, 0.36 and 0.66 eV, respectively) into account here. We determine reaction rates from measured loss rates of Be^{+} and HD^{+}, which are in agreement with expected Langevin reaction rates given the pressure of 1 × 10^{−8} Pa in our vacuum apparatus^{16}. We include these processes in realistic MD simulations (Methods) employing leapfrog integration with an adaptive time step, to ensure that collisions between energetic ions are correctly handled. Under these conditions, our MD simulations reveal average HD^{+} zvelocity distributions, which deviate significantly from thermal (Gaussian) distributions, an effect that has hitherto been neglected by the widespread assumption that lasercooled ion ensembles exhibit thermal velocity distributions. As shown in Fig. 3, we empirically find that the velocity distributions are better represented by qGaussians^{17}, which have the additional advantage that the effect of a given chemical reaction can be parameterized by a corresponding qvalue (with q=1 corresponding to a Gaussian distribution). Taking all reactions and reaction rates into account, we find that the velocity distribution is best characterized by a qGaussian with q ranging between 1.00 and 1.07, depending on the assumed thermalization rate and the REMPD rate. We therefore include qGaussians in our spectral line shape function (Methods). Comparing scenarios with minimum and maximum expected thermalization rates (corresponding to q=1.07 and q=1.00, respectively), we find a mean shift of −0.25(25) MHz with respect to the case of q=1.00. The origin of this shift lies in the overlap and saturation of Dopplerbroadened hyperfine components in the spectrum (Fig. 2d). A similar frequency shift may occur when micromotion sidebands are present. Here we make a distinction between excess radial micromotion caused by radial static offset fields, and axial micromotion that could arise from an axial r.f. field because of geometrical imperfections of our ion trap. A finiteelement analysis of the trap’s electric field reveals that such an axial field will be approximately constant over the extent of the Coulomb crystal, which implies that the corresponding micromotion contribution to the line shape will be homogeneous. We use the fluorescence correlation method of Berkeland et al.^{13} to find a small micromotion amplitude of 11(4) nm amplitude along the 782nm laser beam as follows. Measurements of the r.f. field amplitude are performed on the Be^{+} ions using the 313nm laser beam (which copropagates with the 782 nm laser beam). Repeating such measurements for various values of the trap r.f. voltage allows separating the axial and the radial component (the latter being due to the residual projection of the laser beam onto the radial direction). We find that the axial micromotion component is dominant, which is explained as follows. First, the 782nm laser is aligned parallel to the trap axis, so that the radial micromotion amplitude is suppressed by the nearzero angle between the 782nm laser and the radial directions. Second, as the surrounding Be^{+} ions shield the HD^{+} from static radial offset fields, the HD^{+} ions are trapped symmetrically about (and close to) the nodal line of the r.f. field. These two conditions limit the contribution of the radial micromotion to well below the measured axial amplitude of 11(4) nm. The resulting sidebands are included in the spectral line shape function, leading to an additional frequency shift of −55(20) kHz with respect to the case of zero micromotion.
After correcting for all systematic frequency shifts (Table 1), we finally obtain ν_{0}=383,407,177.38(41) MHz.
Discussion
Our frequency value ν_{0} of the (v,L): (0,2)–(8,3) transition is in good agreement with the more accurate theoretical value^{2} ν_{0,theo}=383,407,177.150(15) MHz. The contribution of the QED terms^{1,2,18} to this frequency amounts to −1547.225(15) MHz (−4035.46(4) p.p.b.), and our measurement therefore confirms the validity of QED in a molecular system at an unprecedented level of 2.7 × 10^{−4}. We furthermore note that our measurement is sensitive to (and in agreement with) QED terms up to order m_{e} α^{7}, given that the m_{e} α^{7} term contributes 780(15) kHz (that is, 2.03(4) p.p.b.) to the transition frequency.
Salumbides et al.^{5} showed that spectroscopic tests of molecular QED can be used to set upper bounds on a hypothetical ‘fifth force’, acting between hadrons at separations of the order of 1 Å, and arising from the exchange of unknown, massive virtual particles. Modelling such a hadron–hadron interaction with a Yukawatype potential of the form ħc α_{5} e^{−r/λ}/r and computing the resulting frequency shift to ν_{0}, we can exploit the 1 p.p.b. agreement between theory and experiment to rule out (at the 90% confidence level) interactions at a range λ=1–2 Å (corresponding to forcecarrying particles with mass m_{Y}=1–2 keV c^{−2}) with an interaction strength relative to the finestructure constant, α_{5}/α, larger than 5–8 × 10^{−10} (Fig. 4). In a similar way, applying the ArkaniHamed–Dimopoulos–Dvali formulation to probe the effect of compactified higher dimensions on energy levels in molecules^{6}, we place an improved upper bound of 0.5 μm on the compactification radius of higher dimensions in 11dimensional Mtheory.
Four decades ago, Wing et al.^{4} suggested that molecular theory could one day be used to translate measured vibrational frequencies of HD^{+} to a new value of the proton–electron mass ratio, μ. The high accuracy of our result and the good agreement with theory, which assumes the 2010 Committee on Data for Science and Technology (CODATA) recommended value μ_{CODATA10}, now allow the first determination of μ from molecular vibrations. Previously^{19} we had derived the sensitivity relation δμ/μ=−2.66 δν_{0}/ν_{0}, which we employ to adjust μ_{CODATA10} to a new value, μ_{HD+}, such that the theoretical frequency matches our experimental value. We thus find μ_{HD+}=1,836.1526695(53), with a relative uncertainty of 2.9 p.p.b., which approaches that of the values taken into account in the 2010 CODATA adjustment^{20}. For example, the value reported by Farnham et al. is only a factor of 1.3 more precise than our result^{21}. Although the precision of a very recent determination by Sturm et al. is still 31 times higher^{22}, we point out that our method yields μ as a single parameter from molecular vibrations, whereas most other μ values are ratios of individual determinations of the electron and proton relative atomic masses (exceptions are μ determinations from atomic laser spectroscopy^{23,24}). Therefore, the agreement of μ_{HD+} with all other values of μ forms an additional consistency check of the various (and vastly different) methods used (Fig. 5b). In particular, it implies that relativistic quantum mechanics and QED consistently describe at the few p.p.b. level such diverse systems as the bound electron^{22,25,26,27}, antiprotonic helium^{23} and the molecular hydrogen ion. Furthermore, of all the methods that produce μ as a single parameter, our approach is surpassed only by spectroscopy of antiprotonic helium, which is 2.3 times more precise but additionally requires charge, parity and time reversal invariance^{23} (Fig. 5a).
In principle, the transition frequency ν_{0} depends on the value of other fundamental constants as well, such as the deuteron–proton mass ratio^{7}, m_{d}/m_{p}, the fine structure constant, and the proton charge radius^{1,2}. However, the sensitivity of ν_{0} to changes in μ is known to be three times larger than the second largest one, the m_{d}/m_{p} sensitivity^{7}. Moreover, if we propagate the uncertainties of the 2010 CODATA values of the fundamental constants through the sensitivity relations, we find that the relative contributions by μ, m_{d}/m_{p}, α and the proton radius to the theoretical uncertainty of ν_{0} are 154, 11.6, 0.004 and 5.13 parts per trillion, respectively. We therefore conclude that μ is the correct parameter to constrain.
The error budget in Table 1 reveals that the experimental uncertainty is limited by Doppler broadening. To overcome this, more involved Dopplerfree twophoton spectroscopy of HD^{+} has been proposed^{9}, which should reduce the uncertainty to below 1 × 10^{−12}. This should enable a comparison between theory and experiment that would not only test the QED description of chemically bonded particles at an unprecedented level, but also produce a new value of μ with an uncertainty below 0.1 p.p.b., surpassing the most precise determination of μ obtained from independent electron and proton relative atomic mass measurements^{22}, which represent a completely different method (Fig. 5c). Our work demonstrates the potential of molecular hydrogen ions for the determination of mass ratios of fundamental particles, as well as stringent tests of QED, and searches for new physics.
Methods
Experimental procedure
We typically store 40–85 HD^{+} ions in a linear r.f. trap, together with ∼750 Be^{+} ions, which are laser cooled to a temperature of 5–10 mK using a continuouswave (CW) 313nm laser beam propagating along the symmetry (z) axis of the trap. As a consequence, only the axial motion of the Be^{+} ions is lasercooled directly. However, the threedimensional extent of the Coulomb crystal ensures good coupling of the axial motion to the radial motion of the Be^{+} and HD^{+} ions, so that both Be^{+} and HD^{+} are efficiently cooled in all three dimensions. Although larger numbers of ions may be trapped, smaller ion numbers ensure better reproducibility of the experimental conditions. The stronger confinement of HD^{+} by the pseudopotential leads to the formation of a string or zigzag structure of HD^{+} ions along the nodal line of the r.f. field, which coincides with the z axis. Be^{+} ions arrange themselves in a threedimensional ellipsoidal structure surrounding the HD^{+}. At temperatures below 0.1 K, the twospecies ion ensemble solidifies into a Coulomb crystal. 313 nm fluorescence photons emitted by Be^{+} are imaged onto an electron multiplied chargecoupled device camera and a photomultiplier tube. We obtain a measure of the number of trapped HD^{+} ions by resonantly driving their radial secular motion (∼800 kHz) using an a.c. electric field^{19}. MD simulations indicate that this ‘secular excitation’ heats up and melts the ion ensemble, heating the Be^{+} ions to an average temperature (in the zdirection) T_{Be+,av} proportional to the number of HD^{+}, N_{HD+}. Typically, T_{Be+,av}=2–4 K. For a fixed coolinglaser detuning Δ=−18 Γ (with Γ=2π × 19.4 MHz, the natural linewidth of the 313 nm cooling transition), this temperature rise leads to Doppler broadening and, thus, to a significant rise in the 313nm fluorescence rate with average value F. Whereas previous work approximated the fluorescence rise versus N_{HD+} by a linear dependence^{28}, we realistically model the (nonlinear) dependence of F on N_{HD+}, and take this into account in our analysis (see Spectral line shape function for fitting).
To excite the (v,L): (0,2)–(8,3) rovibrational transition at 782 nm, we send a CW 782nm laser beam along the z axis, counterpropagating the 313nm laser beam. The 782nm radiation is obtained from a titanium:sapphire laser with a linewidth of 0.5 MHz, which is frequency locked to an optical frequency comb laser as follows. The frequency of the optical beat note of the 782nm laser with a nearby mode of the comb is measured by a counter every 30 ms. From the measured beatnote frequency, the comb repetition rate, comb carrierenvelope offset frequency and the comb mode number, we determine the laser’s optical frequency. The counter output is fed into a digital feedback loop, which controls the 782nm laser so as to stabilize the beatnote frequency to a preset value. The comb is fully referenced to a global positioning system (GPS) disciplined rubidium atomic clock (providing longterm relative accuracy at the level of 2 × 10^{−12}), and the frequency uncertainty of our optical frequency measurement system (10 kHz) is limited solely by the frequency instability of the locked 782 nm laser averaged over the 10 s of REMPD. To detect excitation to the v=8 state by the 782 nm laser, we overlap this beam with a CW 532nm laser beam, leading to REMPD of HD^{+} (Fig. 1a). An experimental cycle looks as follows. We first load a fresh sample of HD^{+} and find a measure, F_{i}, of the initial ion number, N_{i}, by secular excitation. We subsequently lower the 313nm cooling laser intensity and detuning to reduce the ion temperature to ∼10 mK, and expose the ions to the 313, 782 and 532 nm lasers for 10 s. Afterwards, we apply secular excitation to determine the fluorescence level F_{f} corresponding to the remaining ion number, N_{f}, and we define a signal S=(F_{i}−F_{f})/F_{i}. Repeating this cycle for different values of the 782nm laser frequency, ν, we obtain a spectrum consisting of 1,772 points (Fig. 2a). Noise in the spectrum is primarily due to the stochastic nature of REMPD of small HD^{+} ensembles with a mean rotationalhyperfine occupancy of the order of one ion per state.
Spectral line shape function for fitting
Hyperfine rate equations allow computing the evolution of a sample of HD^{+} ions, with an initial thermal rotational distribution^{29} corresponding to T=300(5) K, under the influence of REMPD lasers, BBR and losses due to chemical reactions with backgroundgas molecules (Fig. 1b,d). Population in rotational states with L=0–5 is included, thus ignoring the 2.4% population in L=6 and higher. Accurate hyperfine line strengths at the magnetic substate level are obtained by extending the approach of Koelemeij^{30} so as to include hyperfine structure and the Zeeman effect. The response to the 782nm laser of each hyperfine level is modelled using a Dopplerbroadened profile based on qGaussians. For the assessment of the Zeeman effect, the hyperfine rate equations are adapted to include Zeeman line shifts and broadening due to hyperfine line splitting (which remains much smaller than the 16MHz Doppler width). The q parameter is determined from realistic MD simulations (which take into account the timedependent trap potential and momentum changes by scattering of photons from the cooling laser) with an uncertainty limited by the minimum and maximum expected rates of heating events. Solving the rate equations allows predicting the fractional loss of HD^{+}, γ, as a function of the 782nm laser frequency relative to the hyperfineless frequency, ν−ν_{0}, the intensity, I_{L}, of the 782nm laser and the HD^{+} motional temperature, T_{HD+}. We obtain a smooth, continuous spectral line shape function for fitting as follows. First, we compute γ on a dense threedimensional grid of values (ν, I_{L}, T_{HD+}), which we subsequently interpolate to obtain a fourdimensional function γ(ν−ν_{0,fit}, I_{L}, T_{HD+}). Using the shorthand notation γ(ν−ν_{0,fit}, I_{L}, T_{HD+})=γ, and making use of the linear dependence of T_{Be+,av} on N_{HD+} found from MD simulations, the average Be^{+} temperatures during secular excitation of HD^{+} before and after REMPD are given by T_{i}≡T_{Be+,av}(N_{i}) and T_{f}≡T_{Be+,av}(γN_{i}), respectively. We consecutively insert these temperature values in the scattering rate function g, defined as
which includes the beryllium mass, m_{Be+}, and the fixed 313 nm laser detuning, wave vector, k, and saturation parameter, s. The integration is performed over the distribution of Be^{+} velocities v along k, that is, v_{k}≡k·v/k. This function is used to model the fluorescence rise F observed in the experiment, and allows us to construct a fivedimensional fit function G(ν−ν_{0,fit}, I_{L}, T_{HD+}, T_{i})≡[g(T_{i})—g(T_{f})]/[g(T_{i})–g_{0}], with g_{0} the steadystate scattering rate before secular excitation. The function G realistically models the signal S (Fig. 2c), and is used for fitting.
We employ the above model also to estimate the uncertainties due to several systematic effects. For example, MD simulations predict a slight increase of T_{HD+} and q with increased REMPD rates in the scenario of inefficient thermalization of fast D^{+}. We can estimate the frequency shift because of this effect by making both T_{HD+} and q REMPDrate dependent. This leads to an additional frequency uncertainty of 61 kHz, which is included in the error associated with Doppler effects due to chemistry (Table 1). From our model, we furthermore deduce that ignoring states with L≥6 introduces an uncertainty of 32 kHz, whereas the possible 5 K error in the BBR temperature estimate (which takes into account daytoday temperature variations and the possibility that the trap electrodes may be at a slightly higher temperature due to r.f. dissipation) translates to a 5kHz effect.
Additional information
How to cite this article: Biesheuvel, J. et al. Probing QED and fundamental constants through laser spectroscopy of vibrational transitions in HD^{+}. Nat. Commun. 7:10385 doi: 10.1038/ncomms10385 (2016).
References
 1.
Korobov, V. I., Hilico, L. & Karr, J.Ph. mα^{7}Order corrections in the hydrogen molecular ions and antiprotonic helium. Phys. Rev. Lett. 112, 103003 (2014).
 2.
Korobov, V. I., Hilico, L. & Karr, J.Ph. Theoretical transition frequencies beyond 0.1 p.p.b. accuracy in H_{2}^{+}, HD^{+}, and antiprotonic helium. Phys. Rev. A 89, 032511 (2014).
 3.
Bressel, U. et al. Manipulation of individual hyperfine states in cold trapped molecular ions and application to HD^{+} frequency metrology. Phys. Rev. Lett. 108, 183003 (2012).
 4.
Wing, W. H., Ruff, G. A., Lamb, W. E. Jr. & Spezeski, J. J. Observation of the infrared spectrum of the hydrogen molecular ion HD^{+}. Phys. Rev. Lett. 36, 1488–1491 (1976).
 5.
Salumbides, E. J. et al. Bounds on fifth forces from precision measurements on molecules. Phys. Rev. D 87, 112008 (2013).
 6.
Salumbides, E. J., Schellekens, A. N., GatoRivera, B. & Ubachs, W. Constraints on extra dimensions from precision molecular spectroscopy. N. J. Phys. 17, 033015 (2015).
 7.
Schiller, S. & Korobov, V. Tests of time independence of the electron and nuclear masses with ultracold molecules. Phys. Rev. A 71, 032505 (2005).
 8.
Schiller, S., Bakalov, D. & Korobov, V. I. Simplest molecules as candidates for precise optical clocks. Phys. Rev. Lett. 113, 023004 (2014).
 9.
Tran, V. Q., Karr, J.Ph., Douillet, A., Koelemeij, J. C. J. & Hilico, L. Twophoton spectroscopy of trapped HD^{+} ions in the LambDicke regime. Phys. Rev. A 88, 033421 (2013).
 10.
Karr, J.Ph. H_{2}^{+} and HD^{+}: candidates for a molecular clock. J. Mol. Spectrosc. 300, 37–43 (2014).
 11.
Koelemeij, J. C. J., Roth, B., Wicht, A., Ernsting, I. & Schiller, S. Vibrational spectroscopy of HD^{+} with 2p.p.b. accuracy. Phys. Rev. Lett. 98, 173002 (2007).
 12.
Bakalov, D., Korobov, V. I. & Schiller, S. Highprecision calculation of the hyperfine structure of the HD^{+} ion. Phys. Rev. Lett. 97, 243001 (2006).
 13.
Berkeland, D. J., Miller, J. D., Bergquist, J. C., Itano, W. M. & Wineland, D. J. Minimization of ion micromotion in a Paul trap. J. Appl. Phys. 83, 5025–5033 (1998).
 14.
Bakalov, D. & Schiller, S. The electric quadrupole moment of molecular hydrogen ions and their potential for a molecular ion clock. Appl. Phys. B 114, 213–230 (2014).
 15.
Roth, B., Koelemeij, J. C. J., Daerr, H. & Schiller, S. Rovibrational spectroscopy of trapped molecular hydrogen ions at millikelvin temperatures. Phys. Rev. A 74, 040501(R) (2006).
 16.
Wineland, D. J. et al. Experimental issues in coherent quantumstate manipulation of trapped atomic ions. J. Res. Natl Inst. Stand. Technol. 103, 259 (1998).
 17.
Umarov, S., Tsallis, C. & Steinberg, S. On a qcentral limit theorem consistent with nonextensive statistical mechanics. Milan J. Math 76, 307–328 (2008).
 18.
Korobov, V. I. Relativistic corrections of mα^{6} order to the rovibrational spectrum of H_{2}^{+} and HD^{+} molecular ions. Phys. Rev. A 77, 022509 (2008).
 19.
Koelemeij, J. C. J., Noom, D. W. E., De Jong, D., Haddad, M. A. & Ubachs, W. Observation of the v’=8  v=0 vibrational overtone in cold trapped HD^{+}. Appl. Phys. B 107, 1075–1085 (2012).
 20.
Mohr, P. J., Taylor, B. N. & Newell, D. B. CODATA recommended values of the fundamental physical constants: 2010. Rev. Mod. Phys. 84, 1527–1605 (2012).
 21.
Farnham, D. L., Van Dyck, R. S. Jr. & Schwinberg, P. B. Determination of the electron's atomic mass and the proton/electron mass ratio via Penning trap mass spectroscopy. Phys. Rev. Lett. 75, 3598–3601 (1995).
 22.
Sturm, S. et al. Highprecision measurement of the atomic mass of the electron. Nature 506, 467–470 (2014).
 23.
Hori, M. et al. Twophoton laser spectroscopy of antiprotonic helium and the antiprotontoelectron mass ratio. Nature 475, 484–488 (2011).
 24.
De Beauvoir, B. et al. Absolute frequency measurement of the 2S8S/D transitions in hydrogen and deuterium: new determination of the Rydberg constant. Phys. Rev. Lett. 78, 440–443 (1997).
 25.
Beier, T. et al. New determination of the electron’s mass. Phys. Rev. Lett. 88, 011603 (2002).
 26.
Verdu, J. et al. Electronic g factor of hydrogenlike oxygen ^{16}O^{7+}. Phys. Rev. Lett. 92, 093002 (2004).
 27.
Mohr, P. J., Taylor, B. N. & Newell, D. B. CODATA recommended values of the fundamental physical constants: 2006. Rev. Mod. Phys 80, 633–730 (2008).
 28.
Schneider, T., Roth, B., Duncker, H., Ernsting, I. & Schiller, S. Alloptical preparation of molecular ions in the rovibrational ground state. Nat. Phys. 6, 275–278 (2010).
 29.
Koelemeij, J. C. J., Roth, B. & Schiller, S. Blackbody thermometry with cold molecular ions and application to ionbased frequency standards. Phys. Rev. A 76, 023413 (2007).
 30.
Koelemeij, J. C. J. Infrared dynamic polarizability of HD^{+} rovibrational states. Phys. Chem. Chem. Phys. 13, 18844 (2011).
Acknowledgements
We are indebted to J. Bouma, T. Pinkert and R. Kortekaas for technical assistance, and to V. Korobov, E. Hudson and R. Gerritsma for fruitful discussions. We thank E. Salumbides for providing the curve for D_{2} in Fig. 4. This research was funded through the Netherlands Foundation for Fundamental Research on Matter (FOM), the COST action MP1001 IOTA and the DutchFrench bilateral Van Gogh Programme. SURFsara (http:// www.surfsara.nl) is acknowledged for the support in using the Lisa Compute Cluster for MD simulations. J.C.J.K. thanks the Netherlands Organisation for Scientific Research (NWO) and the Netherlands Technology Foundation (STW) for support.
Author information
Affiliations
LaserLaB, Department of Physics and Astronomy, VU University, De Boelelaan 1081, Amsterdam 1081 HV, The Netherlands
 J. Biesheuvel
 , K. S. E. Eikema
 , W. Ubachs
 & J. C. J. Koelemeij
Laboratoire Kastler Brossel, UPMCSorbonne Universités, CNRS, ENSPSL Research University, Collège de France, 4 place Jussieu, Paris 75005, France
 J.Ph. Karr
 & L. Hilico
Département de Physique, Université d’Evry Val d’Essonne, Boulevard François Mitterrand, 91025 Evry, France
 J.Ph. Karr
 & L. Hilico
Authors
Search for J. Biesheuvel in:
Search for J.Ph. Karr in:
Search for L. Hilico in:
Search for K. S. E. Eikema in:
Search for W. Ubachs in:
Search for J. C. J. Koelemeij in:
Contributions
J.B. and J.C.J.K built the apparatus, and acquired and analysed data. K.S.E.E. provided the optical frequency measurement system. J.Ph.K., L.H., J.B. and J.C.J.K. evaluated systematic frequency shifts and performed simulations. W.U. and J.C.J.K. conceived and supervised the experiment. J.B. and J.C.J.K. wrote the manuscript, which was edited and approved by all coauthors.
Competing interests
The authors declare no competing financial interests.
Corresponding author
Correspondence to J. C. J. Koelemeij.
Rights and permissions
This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
About this article
Further reading

1.
Rotational spectroscopy of cold and trapped molecular ions in the Lamb–Dicke regime
Nature Physics (2018)

2.
From a quantumelectrodynamical light–matter description to novel spectroscopies
Nature Reviews Chemistry (2018)

3.
Quantum electrodynamics, highresolution spectroscopy and fundamental constants
Applied Physics B (2017)

4.
Applied Physics B (2017)

5.
Highaccuracy deepUV Ramseycomb spectroscopy in krypton
Applied Physics B (2017)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.