Is the proton radius puzzle evidence of extra dimensions?

The proton charge radius inferred from muonic hydrogen spectroscopy is not compatible with the previous value given by CODATA-2010, which, on its turn, essentially relies on measurements of the electron-proton interaction. The proton's new size was extracted from the 2S-2P Lamb shift in the muonic hydrogen, which showed an energy excess of 0.3 meV in comparison to the theoretical prediction, evaluated with the CODATA radius. Higher-dimensional gravity is a candidate to explain this discrepancy, since the muon-proton gravitational interaction is stronger than the electron-proton interaction and, in the context of braneworld models, the gravitational potential can be hugely amplified in short distances when compared to the Newtonian potential. Motivated by these ideas, we study a muonic hydrogen confined in a thick brane. We show that the muon-proton gravitational interaction modified by extra dimensions can provide the additional separation of 0.3 meV between 2S and 2P states. In this scenario, the gravitational energy depends on the higher-dimensional Planck mass and indirectly on the brane thickness. Studying the behavior of the gravitational energy with respect to the brane thickness in a realistic range, we find constraints for the fundamental Planck mass that solve the proton radius puzzle and are consistent with previous experimental bounds.

The proton charge radius is defined as < r 2 p >= r 2 ρ E (r) d 3 r, where ρ E is the normalized electric charge density of the proton. Based on the standard theory of bound-state quantum electrodynamics (QED), the effects of the proton internal structure on atomic energy spectrum can be predicted. For instance, in the muonic hydrogen, it is expected that the contribution for the 2S 1/2 − 2P 1/2 Lamb shift is given by [2,15]: where r p must be expressed in femtometer. According to this formula, the energy shift is ∆E th L r CD p = 202.0416(469) meV, when it is calculated with the CODATA-2010 radius. On the other hand, the experimental value is extracted from the measurement of the (2P F =1 3/2 − 2S F =0 1/2 ) and (2P F =2 3/2 − 2S F =1 1/2 ) transitions frequencies, ν s and ν t respectively, and from the formula [2,15]: where the numeric term comes from the explicit calculation of the 2P fine and hyperfine splitting. By using the measured frequencies, ν s = 54611. 16 (1.05) GHz [2,15] and have been proposed to explain the energy excess [16][17][18][19][20][21][22][23][24][25][26][27][28][29], but there is no final conclusion yet.
Here we want to discuss an alternative explanation. As the muon is around 207 times heavier than the electron, it is reasonable to conjecture that gravity is the missing piece in this puzzle. The problem is that the Newtonian potential is negligible in atomic system.
However, in the context of the braneworld with large extra dimension, the gravitational potential can be much greater in short distances. This fact has motivated us to address this issue in the context of the braneworld models.
In the braneworld scenario, our visible Universe is a submanifold with three spatial dimensions (the 3-brane) embedded in an ambient space of higher dimensions (the bulk) [30][31][32][33].
Matter and standard model fields are confined to the brane while gravity can propagate in every direction of the bulk. Although gravity has access to whole ambient space, the existence of a bound zero-mode (due to a compact topology or to an appropriate curvature of the bulk), guarantees that the Newtonian behavior is recovered for distances greater than a characteristic length scale ℓ of the extra space, making the model phenomenologically viable. In the case of compact topology, ℓ is the size of the supplementary space, while in the case of non-compact topology, ℓ is related to the curvature radius of the ambient space.
It follows from this picture that gravity may feel directly the effects of extra dimensions in a length scale that could be much greater than the scale in which matter and other fields experience the influence of extra dimensions.
Tests of the inverse square law in laboratory establish that the radius of the extra dimension should be smaller than 44 µm [34][35][36][37][38]. This is the tightest constraint for models with only one extra dimension. When the number of extra dimensions is greater, the most stringent constraints come from astrophysics [39,40] and high energy particle collisions [41][42][43][44].
If the gravitational field obeys the Gauss law in the bulk, in the weak field limit, then the gravitational potential of a point-like mass behaves as (G n m) /r n+1 for r << ℓ, where G n is the gravitational constant defined in a space with n extra dimensions. The relation between the Newtonian constant, G, and G n is given by G n ∼ Gℓ n , in magnitude order.
Therefore, in comparison with the Newtonian potential, the extra-dimensional version is amplified by a factor of the order of (ℓ/r) n in short distances. This property has motivated the study of the gravitational interaction in atomic and molecular systems as a way of obtaining empirical bounds for the number and size of extra dimensions [45][46][47][48][49][50][51]. Considering that the gravitational interaction is a small term of the atomic Hamiltonian, we find that the gravitational energy is proportional to the mean value of r −(n+1) in the atomic state.
However, this average diverges for S−states, when n > 2. To avoid this problem, some authors introduce a cut-off radius to perform the calculations [45][46][47][48][49]. As a consequence, the results become dependent on an arbitrary parameter. Previous attempts of solving the proton radius puzzle by means of the extra-dimensional gravity also resorted to a cut-off radius [47,48].
In thick brane scenario the divergence problem is naturally solved. The origin of the divergences is the fact that a delta-like confinement in the brane is a singular distribution from the viewpoint of the bulk [51,52]. However, in a thick brane scenario, the confined particles are described by a regular wave function with a non-null width in the transversal directions. This width should be less than the brane thickness and its value is related to the strength of the confinement. As the width is non-null, the divergence problem naturally disappears.
Considering the muonic hydrogen in this scenario, we find the energy shift of the atomic states caused by the muon-proton gravitational interaction. Based on these calculations, we show that the gravitational energy can account for the energy excess of the measured Lamb shift, solving, in this way, the proton radius puzzle. This condition determines some constraints for the higher-dimensional Planck mass which are consistent with previous empirical bounds.

II. THE GRAVITATIONAL ENERGY OF AN ATOM IN A THICK BRANE
In the field-theory framework, the brane can be seen as a topological defect capable of trapping matter inside its core [53]. As an illustration, we can mention a domain wall in (4+1)-dimensions that separates two vacuum states of a scalar field φ along the extra dimension z [53]. In this configuration, the scalar field can confine matter in the center of the wall by means of a Yukawa-type interaction with Dirac spinors. Under the influence of this interaction, the zero-mode state is described by the following wave function: where β is the coupling constant, ψ (x) represents a free spinor in the (3 + 1)-dimensions, φ 0 = η tanh (z/ε) is the scalar in a domain wall configuration interpolating between two vacua ±η of the scalar field. This wave function has a peak at the center of the brane (z = 0) and decreases exponentially in the transverse direction. The parameter ε can be seen as a measure of the brane thickness, which must be smaller than 10 −19 m to be consistent with current experimental constraints [30,42].
Confinement mechanism for matter in topological defects of greater codimension can be also formulated in a similar away. Based on the previous example, it is reasonable to expect that the wave function of localized particles can be written as Ψ (r, where χ (z) is some normalized function defined in the supplementary space of n-dimensions, concentrated around the origin.
In this context, let us now study the gravitational potential produced by a confined particle in the thick brane. As we are assuming that ℓ >> ε, then we have to consider the direct effects of the extra dimensions on the gravitational potential. To take this into account, we will admit that the static gravitational field satisfies the Gauss law in the bulk.
Thus, in the case of a flat supplementary space with a compact topology, the exact potential of a point-like mass M lying in the origin of the coordinate system and evaluated at the position R = (r, z) can be written as [54]: where the sum spans the topological images of M in the covering space of the extradimensional manifold and R = |R|. The exact position R ′ i of the mirror images depends on the topology of the supplementary space. For instance, in the case of a flat n−torus with size ℓ, we have R i = ℓ (0, 0, 0, k i ), where k i is a vector with n integer number as components.
Regarding the influence of the gravitational potential on the energy spectrum of the muonic hydrogen, the topological images can be neglected, since the contribution they give is lesser than the empirical error of the µp experiment (see appendix). Therefore, to calculate the proton gravitational potential, φ, we may use the approximate Green function −GM/R n+1 , which is weaker than the real potential of a point-like mass. So, assuming that the proton mass m p is distributed on the spatial extension of the nucleus, the proton gravitational potential is where the mass density is ρ M = |Ψ p | 2 m p and Ψ p (r, z) = χ p (z) ψ p (r) is the higher-dimensional wave function of the proton.
The muon-proton gravitational interaction, which is described by the Hamiltonian H G = m µ φ (where m µ is the muon mass), modifies the muonic hydrogen spectrum. Assuming that H G is a small term of the atomic Hamiltonian, the energy shift can be calculated by the perturbation method for each state. In the first order, the energy correction is m µ φ Ψ , i.e., the mean value of the gravitational energy in the state Ψ. By using Eq. (5), we can write the energy shift as: where the higher-dimensional wave function of the muon (more precisely, the reduced par- is the product of the extra-dimensional part χ µ (z) and the solutions ψ µ (r) of the Schrödinger equation for the muonic hydrogen.
To calculate (6), we shall assume that the proton mass is uniformly distributed inside the nucleus. This means that the 3-dimensional part, ψ p (r), is constant in the spatial extension of the nucleus and zero outside r > r CD p . In equation (6), the major contribution comes from the integral in the interior region of the nucleus. For S−states, equation (6) yields where a 0 is the Bohr radius of the muonic hydrogen, ψ S (0) is the wave function of a S-state evaluated in the origin and γ n is a numeric factor whose value depends on the number of extra dimension. For instance, γ 3 = 2π 3/2 , γ 4 = 4π/3, γ 5 = π 3/2 /3 and γ 6 = 4π/15. The gravitational energy depends on how tight is the confinement in the thick brane. In fact, σ is associated to the spatial distribution of the particles in the transverse direction. This parameter is defined as: where m is a positive integer that should satisfy the condition m ≤ (n − 1) and Γ stands for gamma function. If χ is a Gaussian function, then σ coincides with the standard deviation of the Gaussian distribution. For the sake of consistency, σ should be smaller than the brane thickness.
Equation (7) is valid for n > 2. Here we do not discuss the cases n = 1 and n = 2, once the atomic gravitational energy are not strong enough to explain the proton radius puzzle in those dimensions. The integral of equation (6) in the external region is smaller than (7) by a factor of the order of σ/r p , which is lesser than 10 −5 for realistic branes with ε ≤ 10 −20 m.
On its turn, for P -states, the gravitational contribution is smaller than (7) , according to equation (7). On the other hand, the effect on 2P −level is smaller by a factor of the order of 10 −5 , therefore, it is negligible within the precision of 10 −7 eV of the muonic hydrogen experiment [15]. Thus, the gravitational interaction is responsible for an additional enlargement between the levels 2P − 2S given by |δE g 2S |. The puzzle would be solved if |δE g 2S | = 0.3290(469) meV. This condition implies a relation between G n and σ, which, in terms of the fundamental Planck mass M D of the higher-dimensional space, as defined in ref. [41], can be written as: where G n was substituted by the term in the bracket. Figure (1) shows a numerical analysis of equation (9)  In these experiments, the modified gravitational potential is parameterized as GM/r 1 + αe −r/λ , where, in the ADD model, α = 8n/3 and λ is equal to the radius R of extra dimensions [35][36][37][38]. From the empirical constraints on α and λ, upper bounds for R are inferred for each value of n. For instance, for n = 1 and n = 2, the data imply that R < 44 µm and R < 37 µm, respectively, which corresponds (see, the relation between R and M D in the appendix) to M D > 3.6 TeV for n = 2 [37,55]. For greater codimensions, the experimental limits are much below than TeV scale and, therefore, compatible with constraints of Figure 1. If the modification of the gravitational potential is due to radion exchange between matter, instead of graviton exchange, the parameters have different meaning. In this case, α = n/ (n + 2) and λ is the Compton wavelength of the radion, which is related to M * (the unification scale [56]) by the formula λ 2 ∼ ( 3 /cGM 4 * ) [56][57][58]. According to Ref. [56], the limits go from M * > 5.7 TeV (n = 1) to M * > 6.4 TeV (n = 6). Although the exact relation between λ and the fundamental Planck mass depends on the stabilization mechanism of the radion [58], there is plenty space to accommodate these bounds in Figure 1, for n > 3.
Astrophysical and cosmological constraints are strong for n ≤ 4 and are derived from the implications of the supposed production of the KK-gravitons in stars [39,40,55,59].
In this context, the most stringent bound is obtained from the analysis of this process in supernovae explosions. In the ADD higher-dimensional model, an old remnant neutron star is surrounded by trapped KK gravitons which slowly decay into photons. A fraction of them is absorbed by the neutron star causing its heating. As the excess heat is not observed, constraints can be obtained. Data from PSR J09521+0755 demand that M D > 76 TeV for n = 3 [40,55]. In principle, this limit would rule out the case n = 3 of our analysis in Figure   1. However, it is important to have in mind that astrophysics bounds could be evaded by some mechanism that provides an extra mass for KK gravitons [59,60]. Finally let us now compare our results with other spectroscopy data. In a previous work [51], considering a hydrogen stuck in a thick brane, we determined lower bounds for the higher-dimensional Planck mass from the (2S − 1S)-transition. The limits from H spectroscopy are weaker than those necessary to solve the proton radius puzzle. This means that, considering the current constraints of M D as shown in Figure 1, the gravitational energy is capable of explaining the additional difference between the 2S 1/2 and 2P 1/2 states of µp, but it is still hidden in the H spectrum. The reason is that, according to Eq. (7), the atomic energy due to the gravitational interaction depends on the lepton mass to the fourth order, m 4 , approximately, since the energy is proportional to m/a 3 0 and a 0 is defined in terms of atomic reduced mass. Thus, the gravitational energy of the hydrogen is almost (200) 4 times smaller than that of µp, assuming that the confinement of both atoms is similar, i.e., σ H ≃ σ µp . Therefore, the constraints from the muonic hydrogen are also compatible with the most precise spectroscopic data available, which is provided by the hydrogen spectrum.
So, based on these considerations, we can conclude that there are regions in Figure 1 in which the required values of M D to solve the proton radius puzzle satisfy all previous experimental bounds.
At this point, it is important to emphasize that the calculations we have done here are based on the classical behavior of gravity. However, as pointed out in Ref. [41], quantumgravity effects may become significant in a length scale of the order of l D (the Planck length defined in the higher-dimensional space, which is given by l D = ( /c) M −1 D ) or even in a greater scale, depending on the fundamental theory of gravity, not yet known. If this is the case, then unpredicted phenomena could distort or even overshadow the classical effects we have investigated here.
However, according to [61], if the theory of General Relativity is considered as an effective theory, then it is possible to estimate quantum corrections to the gravitational potential energy. In three-dimensional space, if d is the distance between particles with mass M and m, then the classical potential energy is given by the Newtonian term GMm/d and quantum contributions are smaller by a factor of the order of (l p /d) 2 , where l p is the usual Planck length. In the higher-dimensional case, the classical term is G n Mm/d n+1 , and, according to dimensional analysis, the quantum corrections would be of the order of (l D /d) n+2 . Of course, in the muonic hydrogen, proton and muon cannot be considered as point-like particles.
Nevertheless, it is instructive to define an effective distance between them, in the extradimensional space, d ef f , by writing the atomic gravitational energy as G n m p m µ /d n+1 ef f . Now, from equation (9), d ef f can be estimated. Comparing it with the fundamental Planck length, we verify that the ratio (l D /d ef f ) n+2 depends on n and σ, but, for any dimension and for any value of σ investigated here, it is smaller than 10 −4 . Thus, if d ef f is the relevant characteristic length scale of the system concerning its gravitational interaction, then we can expect that the classical contribution will be the leading gravitational influence in this system within the braneworld scenario we are considering here. However, as the fundamental quantum gravity theory is not known, only experiments can answer this question.

IV. FINAL REMARKS
In the thick brane scenario, the direct influence of extra dimensions on gravity arises in a length scale ℓ that may be much greater than the scale in which standard model fields feel directly the effects of supplementary space. It happens that the modified gravitational potential is amplified in small distances (r << ℓ) when compared to the Newtonian potential.
In this context, we found that the proton-muon gravitational interaction can explain the excess of 0.3 meV in the Lamb shift of muonic hydrogen, provided that the fundamental Planck mass satisfies some constraints. In Figure (1), we can find constraints for M D which solve the proton radius puzzle without violating any previous empirical bound.
In the muonic hydrogen experiment, the 2S hyperfine splitting (2S-HFS) was investigated too [15]. In the leading order, the proton structure affects 2S-HFS by means of the so-

V. APPENDIX
Let us consider equation (4). Here we want to demonstrate that the potential generated by the topological images, V im , is negligible within the µp experiment precision. The value of V im depends on the point R = (r, z). We obtain an upper bound for |V im | by evaluating each term of the series in an appropriate point R min of the ambient space of least distance from the corresponding topological image (located in the covering space). First, notice that for all the points R min , we should have r = 0. Thus: If the supplementary space is a flat n-torus with sides of length ℓ, then −ℓ/2 ≤ z i ≤ ℓ/2. Let us consider that the real mass M is in the center of this space, which we will denote by T 0 (ℓ).
The covering space, R n , can be viewed as if it were filled by cells that are copies of T 0 (ℓ). Now consider a mirror image i of M that belongs to another cell. We want to determine the least distance from i to T 0 (ℓ). Notice that the fundamental cell T 0 (ℓ) is inside a ball B of radius d = √ nℓ/2 (the semi-diagonal of T 0 (ℓ)). For n < 8, only the first neighbors, whose distance to the center of T 0 (ℓ) is ℓ, are inside this ball. And, with respect to them, the least distance to T 0 (ℓ) is ℓ/2. Now let us consider images which are outside B. The distance from i, whose position is ℓk i , to any point of T 0 (ℓ) is greater than the radial distance from i to the surface of B, i.e., |z − ℓk i | ≥ ℓk i − d, for any z ∈ T 0 and k i = |k i | > 1. Therefore, ℓk i − d is a lower estimate of the least distance from i to T 0 (ℓ). There are 2n images with k i = 1. The next neighbors have k i = √ 2. Thus, separating the contributions from first neighbors (k i = 1), we may write: To estimate this quantity, let us define T i (1) − the symmetric n-torus of unity size with center at the image i. Each term within the summation sign can be interpreted as the volume of a column above T i (1) and whose height is given by the step function f (k i ) ≡ where x is the position vector in the covering space. The semi-diagonal of T i (1) measures √ n/2. Therefore, for every x ∈ T i (1), x ≤ (k i + √ n/2). Now, as g (x) is a decreasing function, then g (x) ≥ g (k i + √ n/2) = f (k i ) inside the cell T i (1). Thus, the integral of g (x) in the region T i (1) is an upper estimate for f (k i ). For the sake of consistency, we should have x > √ n, according to the definition of g (x). On the other hand, as x ≥ (k i − √ n/2) for x ∈ T i (1), then we may conclude that the previous analysis are valid for k i > 3 √ n/2, i.e., only for cells whose center is separated from the origin by a distance greater than 3 √ n/2. A possible choice is k i = 2 √ n. Closer cells should be taken separately. Thus, computing the contributions given by the first neighbors (k i = 1), by the images at intermediary positions √ 2 ≤ k i ≤ 2 √ n and from the most distant images (k i > 2 √ n), we may write the potential V im as: where we have employed the following estimation: 2π n/2 Γ (n/2) (3 n − 1) (13) Therefore, in the case of muonic hydrogen, the gravitational energy due to the topological images is lesser than (G n m p m µ /ℓ n+1 ) F n , where F n is the function defined from (12), which depends only on the number of extra dimensions and that can be explicitly calculated for n < 8. In magnitude order, G n ∼ Gℓ n , then, in terms of the Newtonian gravitational constant, we may write |V im | ≤ (Gm p m µ /ℓ)F n . As the precision of the muonic hydrogen experiment Empirical bounds for the torus radius (ℓ/2π) in the cases n = 3,..., 6 are shown in Figure   2. By using these values, we can explicitly check that the contribution of the topological images is negligible indeed.