Quantum electrodynamic theory of the cardiac excitation propagation I: construction of quantum electrodynamics in the bidomain
aa r X i v : . [ phy s i c s . g e n - ph ] O c t Quantum electrodynamic theory of the cardiac excitation propagation I:construction of quantum electrodynamics in the bidomain
Sehun Chun a) African Institute for Mathematical Sciences and Stellenbosch University, 5 Melrose road, Muizenberg, Cape Town,South Africa 7945
To provide a unified theoretical framework ranging from a cellular-level excitation mechanism to organic-levelgeometric propagation, a new theory inspired by quantum electrodynamic theory for light propagation isproposed by describing the cardiac excitation propagation as the continuation of absorption and emission ofcharged ions by myocardial cells. By the choice of gauge and the membrane current density, a set of Maxwell’sequations with a charge density and a current density is constructed in macroscopic bidomain and is shownto be equivalent to the diffusion-reaction system with the B. van der Pol oscillator. The derived Maxwell’sequations for the excitation propagation obeys the conservational laws of the number of the cations, energyand momentum, but the total charge is not conserved. The Lagrangian is derived to reveal that the trajectoryand wavefront of the excitation propagation are the same as the electrodynamic wave if ion channels workuniformly. From the second quantization, the Hamiltonian is also derived to explain the excitation mechanismof the myocardial cell by Feynman’s diagram and the mechanism of the refractory period in the perspective ofpositron. The effects of the external electromagnetic field are explained both from the action of the Lagrangianand the interaction by the Hamiltonian.PACS numbers: 87.19.ld, 87.19.Hh, 32.80.-t, 42.50.-p
I. INTRODUCTION: WHY QED THEORY FORELECTROPHYSIOLOGY?
Quantum electrodynamics (QED) is the theory of in-teraction between light and matter, mostly about the ab-sorption and emission of photons. QED lies in the hearthof quantum physics on the scale of one thousandth of pi-cometer (10 − ∼ − m ) and seems to be not on theright scale to be correlated with the mechanism of thecardiac excitation propagation on the tissue scale of mi-crometer (10 − ∼ − m ). This irrelevance seems toreflect more than the axiomatic schisms between physicsand biology. An intriguing nomenclature such as quan-tum biology has been used by some scholars to explainthe photosynthesis of leaf and the bird’s eye campus ,but what is proposed in this paper is not the direct ap-plications of quantum theory such as the effect of theelectromagnetic field on the ions in the heart. Instead,by adapting some analogies on the implicit changes ofscales, we propose that QED also provides the funda-mental mechanism of the cardiac excitation propagation,saying that QED is also the theory of interaction betweenthe electrical signal and the cardiac tissue concerning theabsorption and emission of the propagating cations .As most of the new theories are proposed to explainthe unexplainable phenomena in view of classical the-ories, the introduction of QED to cardiac electrophysi-ology is caused by the same motivation. On a macro-scopic scale, the propagation of the cardiac excitation, orthe cardiac action potential or membrane potential, hasbeen often modeled as waves by a system of diffusion-reaction equations. But this wave model often fails to a) Electronic mail: [email protected] provide theoretical explanations on fundamental mecha-nisms, for example, the necessary conditions of conduc-tion failure. Some disruptions of the excitation prop-agation have been explained by the curvature of thewavefront or the subsequent changes of propagationalvelocities , but these explanations are only applied tothe simplest cases and often fail to be applied to actualgeometry with anisotropy and complex curvature. More-over, as soon as the excitation propagation looses theproperties of the wave or meets the discontinuity of theexcitable media such as myocardial farction or the dis-continuities of myocardial fibre, the mathematical mod-eling or its computational simulation fails to reflect thereal electrophysiological phenomena. The most signifi-cant and required mathematical studies for electrophys-iological pathologies seem to be in the realm of the non-wavelike properties of the cardiac excitation propagationsuch as partial propagation through damaged myocardialtissue or, the movement of ions after the collision of twopropagations, thus we should be prepared to accept the particle aspect of the excitation propagation on a macro-scopic scale. Contrary to the popular ion models on acellular scale such as the Luo-Rudy model providingno theoretical explanation on an organic scale, the incor-poration of the particle aspect on an organic scale fromthe original wave model will provide both explanationsfor phenomena ranging from the cellular level to the or-ganic level.The quanta aspect of the cardiac excitation propaga-tion is more naturally accepted than that of light in theearly 1900’s by Plank , Dirac and Einstein . Maybethe wave aspect of the excitation propagation was neveraccepted by cardiologists since they understand, by ob-servations or textbooks, that myocardial cells are excitedby the influx of charged ions. Then, we may ask ourselveson whether the quanta of the excitation propagation trav-els continuously following the law of classical mechanicspossibly under the influence of an electromagnetic field.Looking at the molecular propagation of the cations suchas potassium ( K + ) reveals that they propagate in thecontinuous procedure of being absorbed and emitted bymyocardial cells. The classical concept of trajectory canbe mathematically constructed from the orthogonal di-rection (in the Riemannian sense) to the wavefront, butclinical observations have never confirmed the existenceof such solid object traveling continuously from the ini-tial time to the final time. Substituting matter in theaether with the myocardial cell , we find a close similar-ity between the interaction between light and matter andits equivalence in the excitation of the myocardial cells.Briefly stated, as light propagates through the space in acontinuous absorption and emission of photons by elec-tron in the aether, we claim in this paper that the electricsignal propagates through the heart tissue in a continu-ous absorption and emission of the propagational cationsby myocardial cells. The only noticeable difference isthat the electron moves in space, but the myocardial cellis stationary. This analogy may be no surprise if light isclassified as a (the fastest) signal in an aether , the sameas the cardiac excitation propagation.All the motivations for the quantization theory of theexcitation propagation, however, have stemmed from un-deniable demands to express the diffusion-reaction equa-tions of the propagation in the language of Maxwell’sequations. Then, the quanta theory of the excitationpropagation is a direct consequence of the second quan-tization of Maxwell’s equations, not to mention the ge-ometric theory of the propagation with a very high fre-quency. The use of Maxwell’s equations as the governingequations for the propagation has the following advan-tages: (i) The first is its versatile expressions in terms ofthe field or the potential. Thus, we can directly incor-porate the external electromagnetic field or the externalpotential. (ii) The second advantage is the convenientderivation of Lagrangians and Hamiltonians possibly inthe simplest form because the governing equations canbe expressed under conservational laws. The Lagrangiancan be used to trace the trajectory of the propagationof the excitation and the simple expression of the La-grangian or, the comparison of it to that of the classicalMaxwell’s equations will shed more light on the behav-ior of the excitation propagation even in complex curvedanisotropic space.(iii) The third is the geometric expression of the gov-erning equations. The eikonal equation for the diffusionequation has been derived by Keener to trace the wave-front of the excitation propagation, but the expression istoo complicated to be practically useful for clinical stud-ies, not to mention that it is derived from the FitzHugh-Nagumo (FHN) equations, or just the diffusion operator,thus, subsequently, inherits the restriction of the wave-like properties. As popularly used in geometric optics,Maxwell’s equations for the excitation propagation may yield the simplest form of the eikonal equation with avery high frequency. (iv) The last, but the most impor-tant advantage is the quantization of the field inducedby the excitation propagation. This means that the ex-citation propagation is considered as the movement ofcorpuscular positive ions of spin 1 (or bosons) which sat-isfy Bose-Einstein statistics, such as photons. This isnecessary in view of cellular-level dynamics because itis well known that the movement of electrically chargedions such as K + , Ca , N a + , Cl − across the intercellularspace or membrane induces the excitation propagation.In spite of subsequent simplifications of ion-pumping pro-cesses, the benefit of quantization can serve as a powerfultool for unanswered phenomena. Well-known examplescould be the interaction of the magnetic field with thepropagating ions, the influence of the magnetic field onthe resting state which we may call a vacuum , and thecollisions of multiple wavelets of the excitation propaga-tion, all of which cannot be explained by the wave theoryof the propagation.Nevertheless, the QED theory for the cardiac excita-tion propagation opens up many fundamental questionsin the perspective classical or quantum electrodynam-ics. (i) The first is whether the (electric) excitation canbe viewed as an electrodynamic wave . The main differ-ence from classical electrodynamic waves lies in the factthat the excitation propagation always requires the me-dia for propagation, for example, the cardiac or nervetissue. This may be analogous to aether which was ab-stractly used to explain the propagation of light by theearly 1900s’ . As the concept of aether has become re-dundant by the field theory of classical electromagneticwaves, can a similar procedure be also legitimately ap-plied to the excitation propagation to eliminate the bio-logical media by adapting the field theory?(ii) Secondly, by the second quantization, we obtain thecreation and annihilation operator representing the inter-action between photon and matter. In the biological sys-tem, matter can be naturally replaced by the basic unitof the media such as the cardiac cell or nerve cell, and thephoton can be replaced by the cation or the positively-charged ion in the tissue. The first replacement requiresus to change only the notion of matter , but the secondreplacement requires us to additionally change the sizeof the propagating particle. Consequently, the absorp-tion or emission of photons by matter should be trans-lated as entrance or exit of the cation thorough a biologi-cal cell representing the similar mechanisms of light, buton a enormously large scale compared to it. A questionrises whether this replacement or translation is legitimatein Maxwell’s equations and its subsequent quantizationsuch that the measure of quantum is not absolute, butmay depend on the type of media.(iii) The last question is on the use of the bidomain space to represent the biological space, instead of theclassical mono-domain space to represent the physicalworld. The bidomain space means that one point on amacroscopic scale always represents two separate pointsin different microscopic spaces. The introduction of thebidomain is inevitable for the construction of a conser-vational system for the excitation propagation. For ex-ample, consider the forest fire: Energy increases in thedomain consisting of the trees only, but energy is pre-served in the domain consisting of the trees and the airaround it. Interpreting this bidomain into the languagesof the modo-domain of the physical world seems to bevalid and produces an unconventional concept of time-varying point charge with respect to time which will beelaborated in the latter part of this paper, but does thisexistence of the two domains at every point of the worldviolate any axiom of physical laws?In the remainder of the paper, we will not prove orjustify these axiomatic questions and leave them for laterdiscussions and publications. This is a reasonable excusebecause answering these rather philosophical questionsseems not to be required for the analysis and explana-tion shown in this paper. In the next section, for readerswho are not familiar with the classical diffusion-reactionmodel, for example, the FHN equations for the excitationpropagation, the diffusion-reaction equations and its lim-itations will be explained in brief. FIG. 1. Illustrations of the cardiac cell (left) and the actionpotential (right).
A. Brief review and restrictions of the diffusion-reactionmodel
Inspired by the FitzHugh’s model , adapting thedesign of the Nagumo’s electric circuit for excitation innerves, the excitation propagation has been most widelycharacterized by the Bonhoeffer van der Pol (BvP) modelfor a relaxation oscillator which can be expressed foran oscillating quantity x such as ¨ x + a ( x −
1) ˙ x + x = 0 , (1)where a ∈ R + and the damping coefficients depend on x quadratically. By introducing the variable y from theLi´enard’s transformation , this oscillator is alternatively expressed as ˙ x = a (cid:18) y + x − x (cid:19) , (2)˙ y = − c ( x − a + a y ) , (3)where a , a ∈ R + . The biological tissue consists of twokinds of media: One is the intracellular space such as themyocardial cell in the heart and the other is the intersti-tial space such as the ambient medium surrounding thecell known as the bath (left of Figure 1).Measured by the electric potential difference betweenthe intracellular space and the interstitial space, the ac-tion potential, as illustrated in the right of Figure 1, isdiffused to the neighboring cells. This propagation mayor may not excite all of them because the mechanism ofexcitation is well characterized by the BvP oscillator ofequation (1) which is only activated by a certain mag-nitude of the membrane potential or a pre-determinedvoltage threshold. The FHN model is derived as an one-dimensional oscillator, but has been widely used as thereaction function of equations (2) and (3) in the diffusion-reaction model which was first proposed as a similar classby Hodgkin and Huxley and inspired by Turing’s mon-umental work for animal coats . Tung extended thismechanism to divide the domain into two separate, butcommunicating domains and included the diffusion pro-cess in the interstitial space while maintaining the BvPoscillator for the intracellular space. Nevertheless, whenthe conductivity ratio between the two media remainsrelatively constant, the Tung’s model, also known as the bidomain model, reduces to a simpler model in order todepict the mechanism of the BvP oscillator in one space,known as the mono-domain model . There are severalvariations of the FHN equations which fit better withthe real shape of the action potential, for example, theRogers-McCulloch model or the Aliev-Panfilov model for the cardiac action potential, but all of them naturallyshare the same critical properties as the FHN model withthe BvP oscillator.In the multi-dimensional space, the FHN model is oftenexpressed as the system of the diffusion-reaction equa-tions as ∂φ∂t = ∇ · ( D∇ φ ) + F ( φ, φ , ψ ) , (4) ∂ψ∂t = G ( φ, ψ ) , (5)where φ is the membrane potential as an activator and ψ is the refractoriness as an inhibitor. F ( φ, φ , ψ ) and G ( φ, ψ ) are reaction functions such that F, G : R × R → R . The diffusivity tensor D represents the conductivityand directionality of myocardial fibre. With the empha-sis on the wave-like property of the excitation propaga-tion, the diffusion-reaction (DR) equations have enjoyedunprecedented success in the modeling of electrophysio-logical phenomena in nerves and in the heart, but at thesame time, they have also revealed some restrictions inanalyzing diverse and complex electrophysiological phe-nomena.The first restriction comes from the fact that (i) theDR model does not obey conservational laws for energyand momentum. The variables φ and ψ only indicate thedifference between two variables measured at the differ-ent spaces, thus energy or momentum is generally notconserved in a physical domain as intuitively being rec-ognized from the equivalent phenomena of forest fire.Consequently, many useful physical concepts and math-ematical devices remain out of reach for the analysis ofthe excitation propagation due to the non-conservationalproperties of the DR model. (ii) Moreover, the analysis ofthe DR model is restricted with the given scale, thus themathematical analysis of the different scale cannot be an-alyzed. This happens because the microscopic DR modelshares the same diffusion operator with the macroscopicDR model, but its reaction functions are significantly dif-ferent. Computationally, a large sum of the microscopicDR model can be an approximation of the macroscopicDR model, but mathematically, they are not equivalent.This inconsistency between different scales prevents theunderstanding of the phenomena occurring on the differ-ent scale.The third restriction arises because (iii) the DR equa-tions have actually one physical variable, the membranepotential denoted as φ in equations (4) and (5). The FHNequations are written with two variables φ and ψ , but thesecond variable ψ , expressed as a function of the mem-brane potential and its time derivative as ψ = f ( φ, ˙ φ ),works as the inhibitor of the membrane potential anddoes not represent a substantially different field in theperspective of classical electrodynamics. If the excita-tion is regarded as the electrodynamic field in three-dimensional space and the membrane potential as thescalar potential, then the above DR equations containonly the scalar potential φ without three components ofthe vector potential. This restriction results in the unde-termined electric and magnetic field even with the time-dependent solution of the DR equations. The underde-termined electromagnetic field induced by the excitationpropagation from the governing equations directly meansthe lack of important tools in the perspective of field inthe study of complex cardiac electrophysiology as wellas the ignorance of the coupling effect of the externalelectromagnetic field.Clinical problems related to the external electromag-netic field can be briefly described as follows: The inter-nal electric current in the heart has been widely studied in vivo or in vitro . The clinical studies of the externalelectric current are not as active as those of the internalelectric current, but the original research of the formermay date back to the 1930s . After the seminal papersdemonstrating the effects of the external electric currentsfor producing effective cardiac beats , mainly for the ter-mination of ventricular tachycardia or fibrillation , thisprocedure has become one of the most effective and pop- ular treatments for cardiac patients. However, its mech-anism remains largely unknown on both microscopic andmacroscopic scales. The biggest difficulty arises when wetry to answer how the exterior electric current is coupledwith the membrane potential that does not completelydetermine the internal electric current up to a constant.If we assume that the time variation of the vector po-tential A is approximately zero, then the gradient of themembrane potential is the same as that of the electriccurrent, but this assumption may not represent reality ifvarying magnetic fields are present in the heart internallyor externally.Similar arguments can be applied to the magnetic field.Since Baule and McFee first reported the magnetic fieldof the heart by magneto-cardiogram in 1962, the mag-netic field of frog-heart muscle and a single axon invitro seemed to validate the intrinsic magnetic field in-duced by the excitation. To date, no clinical implementa-tions have been devised for the use of the magnetic field.Moreover, the effect of the external magnetic field, whichhas never gained substantial attraction in either of theneurology or cardiology communities, remains largely un-known as well. One may argue that this ignorance is dueto the dependency of the magnetic field on the electricfield such that an independent consideration of the mag-netic field is negligible. In 1982’s publication, Plonseymathematically showed the similar claim that the mag-netic field is completely determined by the electric field ,but Plonesy implicitly used the aforementioned assump-tion on vector potential. Roth and Wikswo provided acounter-example to this claim by showing that the mag-netic field induced by the excitation may exist withoutthe presence of the electric field. B. Goals, notations and order of this paper
The goals of the paper can be summarized accordingto two different perspectives. The first is focused on thepractical aspect of this study for clinical applications: (i)The derivation of a mathematical expression to show thatthe functionality of ion channels reflecting the shape ofthe action potential can change the direction and velocityof the propagation. (ii) The derivation of a mathematicalexpression on the effect of the external electromagneticfields for the propagation. (iii) The derivation of thegeometrical action potential propagation and its eikonalequation.The second perspective is on the theoretical aspectof this study: (i) The diffusion-reaction system withthe BvP oscillator can be equivalently expressed byMaxwell’s equations in the bidomain with an appropriatechoice of gauge and the membrane current density. (ii)The one-dimensional BvP oscillator in reciprocal spacedirectly contributes to the reaction function of the ex-citation in multi-dimensional space. (iii) The Maxwell’sequations for the excitation propagation conserve the to-tal number of the cations, the total energy and the mo-mentum. (iv) The Lagrangian of the Maxwell’s equa-tions for the excitation propagation is the same as theLagrangian of the classical Maxwell’s equations if ionchanges work uniformly in all the media. (v) The Hamil-tonian of the excitation can be expressed with quantumoperators and the refractory region can be described asan analogy of a positron.The derivation and quantization of Maxwell’s equa-tions do not go beyond the level of textbooks, especiallyfollowing the book by C. Cohen-Tannoudji et. al. tocompare its results from the classical Maxwell’s equationswith the Coulomb gauge. The description of the excita-tion and its subsequent derivation of a set of Maxwell’sequations also can be applied to the nerve cell in neuro-science, but for the sake of convenience and consistency,we mainly consider cases from cardiac electrophysiology.The most important terminology in this paper is thecation, but it could mean ambiguously multiple objects,probably the same as the ambiguity of the meaning of aphoton. Mostly, the cation means a positively-chargedion traveling in myocardial cells for excitation. Thecations are of a single kind, identical and indistinguish-able, obeying the laws of Bose-Einstein statistics, thesame as the properties of a photon. Among several ionssuch as K + , Ca , N a + , Cl − being involved in the exci-tation mechanism of myocardial cells, the potassium K + could have the closest properties to the cation. However,it may be more accurate to say that the cation meansthe corpuscular of energy and momentum delivered bythe propagation , not a specific type of charged ion. Thus,we may call it a photon as well, but to avoid confusion,we stick to its nomenclature as the cation.Consequently, ion channels are only related to the in-flux and efflux of the cations and this means that we onlypay our attention to the changes of the membrane po-tential induced by the membrane current of the cations.This could be an excessive simplification for complex ionchannels with several ions, but may reveal the funda-mental functions and goals of ion channels on a macro-scopic scale. For example, the excitation of the cardiaccell is mainly aimed to induce calcium Ca for muscularcontraction, thus it may be not relevant to include theion channels for Ca to understand the effects of ionchannels for the excitation propagation. On the otherhand, the sodium N a + channels plays a critical role es-pecially in the depolarization phase, but in this paperthe sodium channels are regarded as the channels of thecations, which is likely to pose no serious problem be-cause they are all positively charged. The chloride Cℓ − is similarly substituted by the channels of the cations, butdue to the different signs, the influx of Cℓ − is consideredas the efflux of the cation and vice versa .This paper is organized as follows: In Section II,Maxwell’s equations are constructed from a microscopicbidomain to a macroscopic bidomain. In Section III, thechoice of gauge and the membrane current density aredescribed and the BvP oscillator is constructed in recip-rocal space. Section IV shows that the diffusion-reaction system for the excitation propagation is equivalent to thederived Maxwell’s equations and its meaning is explainedin the perspective of the semiclassical theory of radia-tion. In Section V, the conservation of the total numberof the cations is proved, but the total charge is shownnot to be conserved. Also, for the system of the particleand the field, the total energy and the total momentumare shown to be conserved. Section VI proposes the La-grangian for the derived Maxwell’s equations and showsthat the Lagrangian is the same as the electromagneticwaves in homogeneous and isotropic media. The effectsof the external electromagnetic field on the trajectoryof the propagation are also shown. In Section VII, theHamiltonian of the Maxwell’s equations is derived. More-over, the excitation mechanism and the refractory periodare described by Feynman’s diagram and the transitionamplitude. The effects of the external electromagneticfield on the excitation mechanism are also shown. Ap-pendix is organized as follows: Appendix I provides theproof of proposition 4 (A), proposition 5 (B). AppendixII provides the proof of lemma 2 (A) and Appendix IIIprovides the proof of lemma 4 (A) and proposition 8 (B). TABLE I. List of notations π i Microscopic intercellular domain π o Microscopic interstitial domain π i ∩ π o Membrane in the microscopic domainΠ Macroscopic domain V i Field or variable in π i V o Field or variable in π o V k Field or variable in reciprocal space V k Parallel component to the wave vector kV ⊥ Perpendicular component to the wave vector k II. FROM MICROSCOPIC TO MACROSCOPICBI-DOMANA. Microscopic domain
Let us begin with Maxwell’s equations on the micro-scopic domain being described as follows: Suppose thatthe microscopic domain contains the collection of my-ocardial cells, each of which is typically 100 µm long and15 µm in diameter as well as the surrounding bath .The intracellular space is denoted as π i representing my-ocardial cells, while the interstitial space is denoted as π o representing the bath. For simplicity, we assume thateach domain is homogeneous. By the microscopic scalefor the excitation propagation, we mean that π i and π o are microscopically separable with a clear boundary as π i ∩ π o = 0. Thus, one point in the microscopic domainbelongs to either of π i or π o while disregarding the thinmembrane of π i ∩ π o . Suppose that, in each microscopicdomain, the dynamics of electromagnetic field induced bythe presence or the movement of point charges are wellexpressed by Maxwell’s equations as follows. In SI units,for the intracellular space π i , ∇ · e i = ̺ i ε i , ∇ · b i = 0 , (6) ∇ × e i = − ∂ b i ∂t , µ i ∇ × b i = ε i ∂ e i ∂t + j i , (7)and for the interstitial space π o , ∇ · e o = ̺ o ε o , ∇ · b o = 0 , (8) ∇ × e o = − ∂ b o ∂t , µ o ∇ × b o = ε o ∂ e o ∂t + j o , (9)where the superscript i and o indicate the variables andfields belonging to the intercellular domain π i and in-terstitial domain π o , respectively. The use of Maxwell’sequations for the electric signal propagation in the heartor the brain has been widely accepted theoretically andexperimentally , not to mention light scattering andbirefringence and light absorption by the action po-tential. Thus we will not discuss the further justificationof the biological electrodynamic field. Representing thepropagation of the electric signal in the resting state,equations (6) - (9) are written in the same expression asthose of the classical electrodynamics that representingthe propagation of light in the space devoid of matter.But they should be interpreted in the different contextbecause we consider different kinds of signal propagationin different media.Let ̺ be the charge density and j be the currentdensity in π i or π o . The magnitude of ̺ and j in eachmicroscopic domain π i and π o are not trivial for almosteverywhere because ion-pumps being attached to π i and π i generate sources by transferring charged ions fromthe other domain. But they are only non-trivial in theduration of the excitation of the myocardial cell whenion pumps are activated. Then, e and b are the electricand magnetic fields induced by them in the cardiactissue , not necessarily meaning the same kind of theelectromagnetic field of the classical electrodynamicsin the physical space . Similarly, the permittivity ε andpermeability µ should also be redefined correspondingto those of the classical electromagnetics. Let us definethe resting state as the condition of the myocardialcell where the membrane potential, or the difference ofthe scalar potential between π i and π o , is stable andno macro-dynamics of the ions occurs. Let c be the maximum speed of the cardiac excitation propagation inthe resting state of the cardiac tissue which is knownto be approximately 1 m/s . Then, for permeability µ of the vacuum state, the permittivity ε is definedas c − /µ , equivalently, ε µ = c and we suppose thatthey remain constant in each domain unless mentionedotherwise. Then we propose the following axiom whichnaturally holds in the classical electrodynamics: Axiom 1 : In the vacuum state, the macroscopicphase velocity by the intercellular space π i is thesame as that by the macroscopic interstitial space π o as ε i µ i = ε o µ o and consequently the same at the membrane.If we consider the generation of the electromag-netic field as the consequences of the moving of chargedions, the equality ε i µ i = ε o µ o means that the maximumspeed of the electric signal in π i and π o is the same.Since this equality is well accepted in the classicalelectrodynamics since the maximum speed of light isconstant everywhere in the relativistic sense, we mayapply the same principle that the electric signal travelsat the same speed in the resting state π i and π o . Infact, this axiom is supported by more fundamentalobservations on the existence of the membrane potential,or the scalar potential difference between in π i and π o (otherwise the membrane potential would collapse)and its constant speed in homogeneous resting media.Then the time variable t corresponding to this signal isaccordingly defined as t = ℓ/c for any length ℓ , but theremaining analysis is non-relativistic, thus for the sakeof simplicity the time t is just set as the physical time t = ℓ/ (speed of light). The charge density ̺ and thecurrent density j are considered discretely with pointcharge ι iα in π i or ι oα in π o and are expressed as ̺ i,o ( r , t ) = X α ι i,oα δ [ r − r α ( t )] , , (10) j i,o ( r , t ) = X α ι i,oα ν i,oα ( t ) δ [ r − r α ( t )] , , (11)where α is the index of each point charge and δ is theDirac-delta function. r α indicates the location of pointcharge indexed as α , while ν iα and ν oα is the velocity ofpoint charge α in each domain π i and π o . In electro-static conditions where there is no movement of chargedparticles and consequently no excitation occurs, the con-servation of the electric charge and the electric currentholds such that ˆ ∪ π i ∂̺ i ∂t dx + ˆ ∪ π o ∂ρ o ∂t dx = 0 , ˆ ∪ ∂π i j i · n ds + ˆ ∪ ∂π i j o · n ds = 0 . In words, the first equality means that the total chargeis conserved in π i ∪ π o . Charged ions can change the do-main but always stays in π i ∪ π o . The second equalitymeans the net current is zero in π i ∪ π o . The zero net cur-rent becomes more obvious by introducing the membranecurrent j m which measures the electric current throughthe membrane where each electric current is expressed as j i = j m and j o = − j m in electrostatic conditions. FIG. 2. From microscopic domain π i and π o to macroscopicdomain Π. p i and p o are the microscopic point in π i and π o ,respectively, and ¯ p i and ¯ p o are the macroscopic point that areobtained as the mean value. B. Macroscopic domain
Next, we integrate the two sets of Maxwell’s equa-tions (6) - (9) of a microscopic domain into the equiv-alent equations on a macroscopic scale. The macroscopicscale corresponds to the domain where one point alwaysrepresents one point in π i and one point in π o . This ispossible because we assume that the size of the myocar-dial cell is much smaller than the unit of the macroscopicunit. Consequently, there is no spacial measurement in Πto distinguish between π i and π o . For example a macro-scopic domain means a myocardial tissue on an organicscale consisting of hundreds and thousands of myocardialcells. The boundary between the macroscopic domainand the microscopic domain could be ambiguous, but weroughly regard the macroscopic scale as the equivalenceof the organic scale by which the cell and the bath cannotbe differentiated as shown in Figure (2). Let us denotethis macroscopic domain as Π. Shifting from the mi-croscopic domain to the macroscopic domain follows theclassical mean value approach first used by H. A. Lorentzfor macroscopic Maxwell’s equations ; the macroscopicfield component is obtained as the average of the micro-scopic field components. For example, the electric field E and the magnetic field B on the macroscopic scale areobtained as E = ¯ e ≡ V ˆ e dV, B = ¯ b ≡ V ˆ b dV, where V is the volume of the sphere centered at eachpoint in the microscopic domain and the bar notation in-dicate that the corresponding quantity is obtained fromthe mean value. Suppose that the sphere is sufficientlylarge so that the sphere does not divide point charge ι iα in π i . By the mean value procedure, ι iα becomes macro-scopic with the corresponding new index α for macro-scopic point charges, but ι iα in the intercellular space π i is not added with ι oα in the interstitial space π o . Themacroscopic point charge q iα and q oα is the average ofpoint charge ι iα and ι oα in each microscopic domain π i and π o such as q iα ≡ (1 /V ) ´ ι iα dV and q oα ≡ (1 /V ) ´ ι oα dV where the sphere V contains both π i and π o , but ι iα and ι oα only exist in π i and π o , respectively.The macroscopic domain Π can also be constructed from π i and π o . One macroscopic point represents eachmicroscopic point in the two different microscopic spacesand subsequently, all parameters may have two differentvalues at each macroscopic point. Note that, as afore-mentioned in the Introduction, this is no more than thebidomain premise which is most popularly used in bio-logical modeling . As a result, Maxwell’s equations forthe intracellular space and the interstitial space are writ-ten in the same macroscopic domain Π: From Maxwell’sequations in π i (6) and (7), ∇ · E i = ρ i ε i , ∇ · B i = 0 , (12) ∇ × E i = − ∂ B i ∂t , µ i ∇ × B i = ε i ∂ E i ∂t + J i , (13)and from Maxwell’s equations in π o (8) and (9), ∇ · E o = ρ o ε o , ∇ · B o = 0 , (14) ∇ × E o = − ∂ B o ∂t , µ o ∇ × B o = ε o ∂ E o ∂t + J o , (15)where the capital letter of the fields such as E , B , and J indicates that the corresponding field is macroscopicand the permittivity and permeability in the macroscopicdomain Π are obtained similarly, but they are the same asthose of the microscopic domain due to the homogeneousassumption of the media such that they are constant ineach microscopic domain. With a point charge q iα and q oα being derived from ι iα and ι oα , the charge density ρ i and ρ o and the current density J i and J o are expressed as ρ i,o ( r , t ) = X α q i,oα δ [ r − r α ] , J i,o ( r , t ) = X α q i,oα v i,oα δ [ r − r α ] , where the velocity of the macroscopic particles v iα and v oα are defined as the weighted average velocity of ι iα ν iα and ι oα ν oα , respectively, such that v i,oα ≡ q i,oα (cid:18) V ˆ ι i,oα ν i,oα dV (cid:19) . (16)Both of sets of Maxwell’s equations (12) - (15) are de-fined in the same macroscopic domain Π. This unusualco-existence does not mean that two electromagneticfields interfere with each other in the near-field of thedomain r α . They appear to be at the same locationmacroscopically, but each field actually lies in the differ-ent space microscopically. Consequently, the only placethey may interact is at the membrane, the boundariesof the intercellular space and the interstitial space, i.e., π o ∩ π i . For every point of the domain Π, there aretwo distinct fields which do not interact with each otherin the near field such as inside the myocardial tissue.However, the point charge in each field may work asa dipole moment, thus they are likely to interact inthe far-field such as outside the heart. This can besummarized as the following axiom: Axiom 2 : The field ( E i , B i ) in π i does not inter-fere with the field ( E o , B o ) in π o at the near-field of themacroscopic domain Π except at the membrane π i ∩ π o .The electromagnetic field is only generated by apoint charge which only lies in either π i or π o . Letus denote q i and q o as point charges staying in π i and π o , respectively. However, we suppose that a pointcharge never stays in the membrane π i ∩ π o because themembrane is relatively thin on the scale of nanometer(10 − m ) and the movement through ion pumps isrelatively instantaneous. In other words, we do notconsider the membrane as the space such that pointcharges only travel in π i and π o . Axiom 3 : The membrane is sufficiently thin ev-erywhere relative to π i and π o . Thus, point charges canonly stay in and travel through either π i or π o , but notin π i ∩ π o . C. Weighted difference of the field and potential
In order to retrieve the well-known observables such asthe membrane potential, we will express the governingMaxwell’s equations as the weighted difference betweenthe field in π i and π o . Since equations (12) - (15) arein the same domain, we can multiply equations (14) and(15) with p ε o /ε i and subtract from equations (12) and(13). Using axiom 1, we obtain ∇ · E = ρε i , (17) ∇ · B = 0 , (18) ∇ × E = − ∂ B ∂t , (19)1 µ i ∇ × B = ε i ∂ E ∂t + J , (20)where the new fields and parameters are defined as theweighted difference such as E ≡ E i − λ E o , B ≡ B i − λ B o ,ρ ≡ ρ i − λ − ρ o , J ≡ J i − λ − J o , where λ is a scalar defined as λ = p ε o /ε i = p µ i /µ o .Since the imaginary component of permittivity is the con-ductivity divided by the frequency, the ratio p σ o /σ i isonly proportional to the imaginary part of λ without be-ing related to the real part. But this does not implythat λ = 1 .
0. The experimental value of λ is unknown,but the presumed value from the well-known phenomenawill be discussed in section VI. This weight differencecan be practically measured at the membrane, i.e. at theboundaries of the two spaces, but we prefer to maintain the intracellular permittivity and permeability constants µ i and ε i instead of µ i − µ o and ε i − ε o . Moreover, sup-pose that q iα and q oα are non-negligible at every locationof r α as physiologically measured, as in ref . Then, ρ and J are expressed as ρ ( r , t ) = X α χ α δ [ r − r α ] , J ( r , t ) = X α χ α v α δ [ r − r α ] , (21)where the new point charge χ α , namely point charge dif-ference , and the velocity v α , namely velocity difference ,are defined in Π as χ α ≡ q iα − λ − q oα , v α ≡ χ α (cid:0) q iα v iα − λ − q oα v oα (cid:1) . (22)Contrary to q iα and q α , χ α is defined only in Π dueto the property that the magnitude of χ α ( r ) can bechanged from the definition of χ α . Consequently, χ α does not explicitly obey axiom 3, but its variation isclosely related to it. It is important to note that ionpumps can significantly change χ α . The operations ofion pumps to change χ α will be discussed in detail inthe later part of this paper. Before proceeding further,we need to briefly mention that the Maxwell’s equations(17) - (20) and the fields E , B , ρ, J are well defined in Π. Proposition 1 : The Maxwell’s equations (17) -(20) with the weighted difference fields are well definedeverywhere in Π.
Proof : By axiom 3, point charge q lies either in π i or π o , thus the Maxwell’s equations (17) - (20) onlyrepresent the well-defined Maxwell’s equations (12) -(15). For example, consider point charge q iα lies in π i . Then, by axiom 2, the Maxwell’s equations turnout to be equations (12) - (13) since E o and B o arezero in π i . A similar argument exists for q oα lying in π o (cid:3) .An additional advantage of the expression of (17) -(20) is that it represents the field value at the membrane π i ∩ π o which is crucial for the initiation of the membranecurrent density J m . If we consider the vector potential A and the scalar potential φ being derived from equations(17) - (20) such as B = ∇ × A , (23) E = −∇ φ − ∂ A ∂t , (24)then we can verify that A and φ are defined as A ≡ A i − λ A o and φ ≡ φ i − λφ o . Note that the classicalmembrane potential is now generalized as φ , the weighteddifference by λ to the potentials in π o . Moreover, substi-tuting equation (17) into the divergence of equation (20)yields the conservation of charge density difference as ∂ρ∂t + ∇ · J = 0 , or ∂ρ i ∂t + ∇ · J i = 1 λ (cid:18) ∂ρ o ∂t + ∇ · J o (cid:19) . (25)The conservation of the first equality means that the timevariation of the charge density difference ρ is only causedby the current density difference J . On the other hand,the second equality only implies that charge density isconserved in π i ∪ π o . We notice that J is zero even withthe significant current density J i in the intracellular spaceif there is the same magnitude and direction of the cur-rent density J o in the interstitial space. If we consider themembrane current density J m , then the current densitiesfor each microscopic domain are expressed as J i = J m and J o = − J m and the current density difference J isexpressed as J = J i − λ − J o = (1 + 1 /λ ) J m . In the nextsection, the construction of the BvP oscillator from theMaxwell’s equations (17) - (20) will show that the mem-brane current density J m is a function of the scalar poten-tial and its time derivative. Then, ρ is also the functionof a scalar potential such as J m = J m ( φ, ˙ φ ) , ρ = ρ ( φ ).Each variable is naturally a function of permittivity ε i and conductivity σ i , but we drop the notations for sim-plicity. TABLE II. Fields and variables by weighted differenceSymbol Definition Symbol Definition
E E i − λ E o B B i − λ B o ρ ρ i − λ − ρ o J J i − λ − J o A A i − λ A o φ φ i − λφ o χ α q iα − λ − q oα v α ( q iα v iα − λ − q oα v oα ) /χ α III. CHOICE OF GAUGE AND MEMBRANE CURRENTDENSITY
The FHN model, a diffusion-reaction model with theBvP oscillator, is popularly used for mathematical mod-eling of the excitation propagation, thus the derivationof the FHN model from the Maxwell’s equations (17) -(20) mean that the two equations are actually equivalentor one system of equations are a subsystem of the otherand may show that the Maxwell’s equations (17) - (20)can also represent the dynamics of the cardiac excitationpropagation. This derivation consists of two procedures:one is to derive the diffusion operator and the other is toderive the BvP oscillator for the reaction.
A. Gauge choice
Firstly, the diffusion operator is easily obtained bygauge choice. By applying the divergence operatorto equation (24) and using equation (17), we obtain ρ ( φ ) /ε i = −∇ φ − ∂ ( ∇ · A ) /∂t . In Maxwell’s equations,the choice of ∇ · A is known as gauge and remains redun-dant for the same fields E and B , but is rather chosenaccording to the type of electromagnetic propagation . For example, with Coulomb gauge ∇ · A = 0, the aboveequation becomes ∇ φ = ρ ( φ ) /ε i , which describes theinstantaneous distribution of a scalar potential. Lorentzgauge defined as ∇ · A = (1 /c ) ∂φ/∂t transforms theabove equation into ∇ φ − (1 /c ) ∂ φ/∂t = − ρ ( φ ) /ε i which gives the special solution of the time dependentPoisson equation describing the retarded radiation .But, since neither of them seems to represent the dy-namics of the propagation, we propose the new gauge tobe defined as ∇ · A = − φ. (26)By using this gauge, we obtain ∂φ∂t = ∇ φ + ρ ( φ ) ε i . (27)Note that the isotropic elliptic operator is obtained by theuse of the new gauge. The physical meaning of the newgauge can be understood in several ways: If we integrateequation (26) over a small region Ω ∈ Π, then by thedivergence theorem, we obtain ´ ∂ Ω A · n dS = − ´ Ω φdV .This equality means that the scalar potential φ is deter-mined by the flux of the vector potential A across theboundaries. Moreover, if we decompose the vector po-tential A into the longitudinal component ( A k ) and thetransverse component ( A ⊥ ) such as A = ∇ A k + ∇ × A ⊥ ,then by substituting this expression into equation (26),we obtain ∇ A k = − φ to imply that the longitudinalcomponent of A determines φ , but the transverse com-ponent still remains undetermined and independent of φ .In general, we do not assume that the transverse compo-nent is zero because equation (18) is B = ∇ × ( ∇ × A ⊥ )implying that the magnetic field for the gauge choice (26)is also zero. However, such a strong restriction is not re-quired for the remaining analysis of this paper.Nevertheless, the new gauge choice still does not deter-mine the potentials because various potentials can pro-duce the same electromagnetic field. For example, thefollowing gauge transformation also yields the same elec-tromagnetic field in equations (17) - (20); A → A + ∇ Λand φ → φ − ∂ Λ /∂t for a scalar function Λ knownas gauge function . Substituting the above transforma-tions into equation (26) reveals that the gauge functionfor the gauge (26) satisfies the simple diffusion equation ∂ Λ /∂t − ∇ Λ = 0. Therefore, for any function Λ satis-fying the above equality, multiple scalar potentials beingadded by −∇ Λ and its corresponding vector potentialsproduce the same electric field E and magnetic field B .The existence of gauge function and subsequence gaugeinvariance are also well described in quantum mechanicsfor example from the invariance of the Pauli equation .As aforementioned, this gauge choice (26) does notchange the electromagnetic field and its potential up to aconstant, but reflects the different mechanism of the lightoscillator for excitation, or signal oscillator for excitationin general. Consider the following wavelike equation be-ing obtained by substituting the gauge choice (26) into0equation (20): ∇ A − c ∂ A ∂t = −∇ φ + 1 c ∂ ( ∇ φ ) ∂t + µ i J ( φ, ˙ φ ) . As a similar procedure done by Heitler , let us con-sider the vector potential A as a series of orthogonalplane waves with the wave number k α and frequency ν α such as A = P α a α ( t ) A α ( r ) where a α depends only on t , and A α depends only on r . Expand φ similarly as φ = P α b α ( t ) φ α ( r ). For simplicity, we only consider thetransverse component of A α with the velocity of the elec-tron as v α = v cos ν t , then the above equation turnsout to be¨ a α + ν α a α = k α ( c b α + ˙ b α ) + f α cos ν t, (28)where c is the velocity of electron e equivalent to 4 πc ρ such that c = ν α /k α and f α = ( e/c ) v | a α ( k ) | cos Θ forthe angle Θ between the polarization and the oscilla-tor. For Lorentz gauge with no b α and ˙ b α in equa-tion (28), a α is analytically given at time t = 0 as a α = ( f α / ( ν α − ν ))(cos ν t − cos ν α t ), thus the oscillatorsare only excited at the wave with the same frequency asthe electron. On the other hand, the new gauge (26) gen-erates additional terms involving the coefficients of φ , b α and ˙ b α , as a source term added to f α cos ν t . Thus, theexcitation of the oscillators now significantly depends onthe scalar potential φ α and its time variation ˙ φ α . More-over, the energy of the oscillator H α after the time t subsequently depends on b α and ˙ b α , and subsequently φ α and ˙ φ α as H α ( k, t ) = ˆ t ˙ a α (cid:16) k α ( c b α + ˙ b α ) + f α cos ν t (cid:17) dt. The presence of φ α and ˙ φ α in the energy of the oscillatorimplies that (1) the energy of the oscillators having afrequency between ν and ν + dν is no more proportionalto the time t and (2) the amount of energy transferred tothe oscillators is not the same as that of the energy flowout of the oscillators. B. Choice of the membrane current density
Instead of assigning the charge density ρ ( φ ) for theBvP oscillator directly, an oscillator will be first con-structed in reciprocal space for the membrane currentdensity J m ( φ, ˙ φ ) to retrieve the original form of the BvPoscillator. Then, the reaction function in real space willbe subsequently determined. This procedure will yieldsimilar results to those in the construction of ρ ( φ ) forthe BvP model in real space, but will verify the connec-tion between the oscillators in reciprocal space and thereaction function in real space. Let’s consider all the fields and variables as running plain waves such as A ( r , t ) = 1(2 π ) / ˆ a k ( k , t ) e i k · r d k,φ ( r , t ) = 1(2 π ) / ˆ φ k ( k , t ) e i k · r d k, J ( r , t ) = 1(2 π ) / ˆ j k ( k , t ) e i k · r d k,ρ ( r , t ) = 1(2 π ) / ˆ ρ k ( k , t ) e i k · r d k, E ( r , t ) = 1(2 π ) / ˆ e k ( k , t ) e i k · r d k, B ( r , t ) = 1(2 π ) / ˆ b k ( k , t ) e i k · r d k, where r is the position vector, k is the wave vector and t is the time variable. This is also known as the Fourierspatial transformation . The fields A , J , E , B , φ , and ρ are in real space, while a k , j k , e k , b k , φ k and ρ k are inthe space, known as reciprocal space or frequency domain ,where the subscript k represents the coefficient of theplane wave with the wave number k . We only considerthat all the fields are real such as e ∗ k ( k , t ) = e k ( − k , t ) , φ ∗ k ( k , t ) = φ k ( − k , t ) , j ∗ k ( k , t ) = j k ( − k , t ) , ρ ∗ k ( k , t ) = ρ k ( − k , t ) , e ∗ k ( k , t ) = e k ( − k , t ) , b ∗ k ( k , t ) = b k ( − k , t ) , where the superscript ∗ means the complex conjugate.We often decompose vectors into the longitudinal vectorfields and transverse vector fields: the longitudinal vectorfield v k is parallel to the wave vector k and is definedas v k k ( k ) ≡ κ [ κ · v k ( k )] for the normalized wave vector κ = k /k . The transverse vector field v ⊥ is perpendicularto the wave vector k and is defined as v ⊥ k ( k ) ≡ v k − v k k .Then it can be easily shown that i k · v ⊥ k = 0 and i k × v k k =0. In reciprocal space, the Maxwell’s equations (17) - (20)are written as i k · e k = ρ k ( φ k ) ε i , (29) i k · b k = 0 , (30) i k × e k = − ˙ b k , (31)1 µ i i k × b k = ε i ˙ e k + 1 + λλ j mk ( φ k , ˙ φ k ) , (32)and equations (23) - (24) and the conservation of charge(25) are given by b k = i k × a k , (33) e k = − i k φ k − ˙ a k , (34)˙ ρ k = − i λλ k · j mk , (35)where the dot notation is used to represent the differ-entiation with respect to the time variable t in recip-1rocal space. Then, the gauge choice (26) is also ex-pressed as i k · a k = − φ k , or, in the transverse direc-tion, a k k = − ( i k /k ) φ k . Similarly, equation (27) is wellexpressed in reciprocal space and by differentiating thisequation with respect to t and by substituting equation(35), we obtain¨ φ k + k ˙ φ k = − iε i λλ k · j mk . (36)The construction of j mk will be conveniently achieved bydecomposing it into two components: One is the currentdensity induced by the electric field e k and the other isthe current density induced by the BvP oscillator. Letthe former component be denoted by j ck , namely the con-ducting membrane current density where c stands forconducting, and the latter by j rk , namely the reactivemembrane current density , where r stands for reaction.Thus, j mk is given by j mk = λ λ ( j ck + j rk ) . (37)The conducting membrane current density j ck is simplycaused by the membrane potential difference between π i and π o . This is similar to the early model on ion channelsbased on electro-diffusion described by the Nernst-Planckequation . j ck can be decomposed into two directions:One is in the parallel direction to k and the other isin the perpendicular direction such as j ck = j c k k + j c ⊥ k where j c k k can be expressed as j c k k = − σ i e k k for the electricconductivity σ i . Since e k k can be expressed in terms of φ k and ˙ φ k from equations (33), (34), and the gauge choice,the conducting current difference j ck is given by j ck = − iσ i k ( φ k − k ˙ φ k ) + j c ⊥ k . (38)On the other hand, the reactive membrane current den-sity j rk is controlled by the macroscopic mechanism of theion channels featured as a resilient oscillator. The choiceof j rk is obviously not defined in the classical electrody-namics because the physical domain is not a bidomain.Thus, we resort to the previous modeling of the excitationmechanism in the biological tissue. For example, in orderto reflect the biological mechanism of the membrane cur-rent flow as first modeled by FitzHugh , we adapt theBvP oscillator as shown in equation (1). Various waysof constructing BvP oscillators are possible, but for thesake of simplicity, we construct the simple j rk as j rk = iε i k k ( φ th − φ k ) ˙ φ k , (39)where φ th is called the threshold potential as the lowestlevel of the electric potential for excitation. The mem-brane current density j rk is not defined as the weighteddifference, but we let the positive sign of j rk be the influxinto π i (or efflux of π o ) and the negative sign of j rk be the influx into π o (or efflux of π i ). If φ k is larger thanthe threshold potential φ th , then the membrane currentoccurs in the negative direction of the wave vector k andis added to the magnitude of j c k k for the rapid increase ofthe potential difference φ k . On the other hand, if φ k isless than the threshold potential φ th , then it flows in thedirection of the wave vector k and it is likely to cancelout j c k k which normally occurs in the opposite direction.Consequently, the cardiac cell is only excited when φ k is sufficiently larger than φ th . In reciprocal space, themembrane current density j rk is constructed in the direc-tion of the wave vector k such that j rk is proportional to˙ φ and φ ph − φ k . In real space, by the inverse Fouriertransform, we obtain the following expression as J r ( r , t ) = 14 πε i ˆ ( φ th − φ ) ˙ φ r − r ′ | r − r ′ | d r ′ . (40)This equality means that the reactive membrane currentdensity J r ( r , t ) is just the Coulomb field of which mag-nitude is proportional to ( φ th − φ ) ˙ φ . This is a directconsequence of the construction of j rk as the longitudi-nal wave in the direction of wave vector k in reciprocalspace. This mechanism becomes more obvious when themembrane current density j mk is expressed as the sum of j ck and j rk such as j mk = − iσ i k λ λ ( φ k − k ˙ φ k ) + iε i k k ( φ th − φ k ) ˙ φ k + j c ⊥ k . (41)Drawn from equation (41), Figure 3 demonstrates thatthe membrane current density j mk increases almostquadratically as the membrane potential φ k increases.Note this phenomenon is almost universal for every ˙ φ k and k ignoring its diverse magnitude. The reason that j mk is not exactly zero at φ k = φ th is due to the scalarpotential induced by the conducting membrane currentdensity j ck . Substituting the expression (41) of j mk intoequation (36) yields¨ φ k + h φ k − (cid:16) φ th − k + η i c (cid:17)i ˙ φ k + k η i φ k = 0 , (42)where η i ≡ σ i /ε i and ω ≡ kc .The threshold potential φ th is an arbitrary scalar quan-tity for the membrane current density j mk in real spacesolely depending on the type of the excitable media. Inreciprocal space, φ th can almost be randomly chosen de-pendent on the type of excitable media, but the follow-ing argument shows its first-order dependency on thewave number k . Suppose that φ th is a constant or atmost a function of k δ where δ <
1. Then, as the wavenumber k is sufficiently large in equation (42), whichroughly implies that the wave is highly fluctuate in timesand can be interpreted as a motion in a shorter dis-tance space, the equation for the oscillator converges to˙ φ k + η i φ k = 0. Therefore, φ k has the following analyticsolution as φ k = c k exp( − η i t ) for an arbitrary constant c k . But, this behavior of the solution contradicts the well2known fact that the membrane potential φ k is zero in theresting state independent of time t and the wave number k . On the contrary, if φ th is a function of k , then equa-tion (42) for a sufficiently high frequency yields φ k = 0satisfying the fundamental conditions of φ k , though themeaning of a high frequency in the excitation propaga-tion remains largely unknown. C. Constructing BvP oscillator
As a consequence, we may choose φ th as φ th ( k ) ≡ p k − η i /c . Then, the equation (42) reduces to asimpler expression as¨ φ k + (cid:0) φ k − (cid:1) ˙ φ k + k η i φ k = 0 . (43)Equivalently by introducing the variable ψ k from theLi´enard’s transformation such as ψ k = ˙ φ k /c + φ k / − (cid:0) φ th − k + η i k/c ) φ k , the BvP oscillator in reciprocalspace caused by j mk is given as: for a, b > φ k = ψ k + (cid:16) φ th − k + η i c (cid:17) φ k − φ k , (44)˙ ψ k = − k ( η i ψ k − a + bφ k ) . (45)Note the similarity between equations (43) and (1), or(44) - (45) and (2) - (3). Equation (43) has the ad-ditional component of k partly because they lie in thedifferent spaces, but the corresponding oscillators are inprinciple the same kind as the BvP oscillator of equa-tion (1) because they have the same quadratic dampingfactor . More analysis can be drawn from the vector po-tential that is also written as an oscillator in reciprocalspace. With equations (32), (34), and (41), the dynamicsof the vector potential a k in reciprocal space is given by¨ a ⊥ k + c k a ⊥ k = c µ i j ⊥ k , (46)¨ a k k − c k ˙ a k k + 2 c k a k k = c µ i j k k . (47)The BvP oscillator does not change the dynamics of a ⊥ k which remains the same as a harmonic oscillator of classi-cal electrodynamical waves. But it significantly changesthe dynamics of a k k crucial for the absorption and emis-sion of the propagating charged particles. Comparingequation (47) with equation (28) immediately revealsthat the equivalent term of ( c b α + ˙ b α ) in reciprocal spaceis only substituted by k ˙ a k k . Roughly stated, this meansthat the membrane potential and its current actually con-tribute to the oscillators representing a k as a dampingfactor . This result is in accord with the BvP oscillator(1). FIG. 3. The current density j k versus the membrane potential φ for a constant wave number k (left) and for a constant timevariation of the scalar potential ˙ φ . σ i = ε i = 1 . k = 1 forthe left plot and ˙ φ = 1 . IV. MAXWELL’S EQUATIONS WITH THE BVPOSCILLATORA. Constructing the reaction function
To obtain a set of Maxwell’s equations equivalent tothe FHN equations, the charge density difference ρ onlyneeds to be derived from the current density difference j k (41) representing the BvP oscillator in reciprocal space.By substituting equation (41) into equation (35), thetime derivative of the charge density difference is givenby˙ ρ k = ˙ ρ ck + ˙ ρ rk = − σ i k (cid:18) φ k − k ˙ φ k (cid:19) + ε i ( φ th − φ k ) ˙ φ k . (48)The first term is clearly induced by the conducting cur-rent density j ck and is called the conducting membranecharge current , denoted by ˙ ρ ck . The second term is sim-ilarly induced by the reactive membrane current density j r and is called the reactive membrane charge current denoted by ˙ ρ rk . Then, equation (48) can be naturally de-composed into the two components of j ck and j rk such as˙ ρ k = ˙ ρ ck + ˙ ρ rk , but we use the following decompositionfor ˙ ψ k and ˙ ξ k such as ˙ ρ k = (1 /ε i ) ˙ ψ k + ˙ ξ k leading to theFHN equations for the BvP oscillator: For an arbitrarytime-dependent function f ( t ) ∈ R ,˙ ψ k = − ˙ ρ ck ε i − η i ˙ φ k − η i φ k f ( t ) η i φ k , ˙ ξ k = ˙ ρ rk − f ( t ) η i φ k + η i φ k σ i ˙ φ k . Then the charge density difference ρ = ρ ( r , t ) in realspace is simply expressed as ∂ρ∂t = ∂ψ∂t − η i (cid:18) f ( t ) φ + φ (cid:19) + ε i f ∂φ∂t − ε i ∂ φ∂t , where f ( r ) = (2 /π ) − / ´ (cid:0) φ th + 2 σ i (cid:1) e i k · r d k . If φ th isa constant independent of k , then f is just φ th + 2 σ i .Nevertheless we consider the general function of φ th ( k ),3thus we maintain this general expression of f ∈ R be-ing independent of the time t . Since f ( t ) is an arbitraryfunction and φ ( t ) can be best approximated as a poly-nomial, it is always possible to choose f ( t ) such that ´ t φ ( r , t )( f ( t ) + φ ( r , t ) / dt ′ is a constant independentof time. Let the value of this integration be f ∈ R . Sup-pose that φ and ψ are all zero at t = 0. By considering φ and σ i as being time independent, the above equationis integrated with respect to time to obtain ρ ( r , t ) = ε i (cid:20) ψ ( r , t ) + f + f φ ( r , t ) − φ ( r , t )3 (cid:21) . (49)Substituting equation (49) and the expression of ψ inreal space into equation (27), we obtain the followingdiffusion-reaction system: ∂φ∂t = ∇ φ + ψ + f + f φ − φ , (50) ∂ψ∂t = − η i ( ψ + f + ( f − f ) φ ) . (51)Compare the above equations with the FHN equations(4) and (5). The reaction functions of the FHN equa-tions vary from model to model, but we may say thatthe above equations share the same properties of the re-action function with those of the FHN equations becausethe function has the components of φ , ψ , and φ repre-senting the BvP properties of the reaction function. Forexample, if we let f = 0 and f = −
1, then the reac-tion functions (50) - (51) are exactly the same as those ofthe original FHN equations . The diffusion-reactionsystem with the BVP oscillator, which displays the dy-namics of the scalar potential φ without the other threecomponents of the vector potential A , is the subsystemof the Maxwell’s equations (52) - (56). But, all the fourequations of the Maxwell’s equations (52) - (56) shouldbe used to derive the diffusion-reaction system (50) - (51)for the excitation propagation in real space. This meansthat neither of the Maxwell’s equations is redundant forthe diffusion-reaction system.The derivation of the anisotropic diffusion-reaction sys-tem with the BvP model can be obtained with a similarprocedure. Applying the divergence operator and tensorproduct with the electric conductivity tensor σ to theexpression of equation (34) in real space, we obtain thefollowing equation: ∇ · ( σ E ) = −∇ · σ ∇ φ − ∂ ( ∇ · ( σ A )) ∂t . Then, the gauge choice and the Coulomb’s law are mod-ified as ∇ · ( σ A ) = − φ, ∇ · ( σ E ) = ρ ani ( φ ) ε i , where ρ ani is the new charge density depending both on E and σ . This means that the conductivity tensor σ ,which is a non-identity tensor due to the cylindrical shape of the cardiac fibre, can significantly increase or decreasethe charge density ρ ani and consequently the potentials φ and A . Deriving the charge density from the BvPmodel for the above equation is beyond the scope of thispaper, so we can simply use the same reaction functionsof equation (49) again to obtain ∂φ∂t = ∇ · σ ∇ φ + ψ + f + f φ − φ ,∂ψ∂t = − η i ( ψ + f + ( f − f ) φ ) . What remains is to obtain the expression for theMaxwell’s equations with ρ m and J m . With φ th ( k ) = p k − η i /c as before and by substituting ρ m fromequation (48) and J m from equation (41) into theMaxwell’s equations (17) - (20), we finally obtain thefollowing Maxwell’s equations with the BvP oscillators: ∇ · E = η i ˆ t ∇ φdt ′ + Γ i φ − ∇ φ − φ , (52) ∇ · B = 0 , (53) ∇ × E = − ∂ B ∂t , (54)1 µ i ∇ × B = ε i ∂ E ∂t + σ i ∇ φ − ε i ∂ ∇ φ∂t (55) − ε i π ˆ (cid:20) Γ i ∂φ∂t + ∂∂t (cid:18) φ (cid:19)(cid:21) r − r ′ | r − r ′ | d r ′ , (56)where we introduced the new variable Γ i ≡ η i − η i /c . Proposition 2 : The Maxwell’s equations (52) -(56), which are derived from equations (17) - (20) withthe new gauge (26) and the choice of the membrane cur-rent density (41), are equivalent to the diffusion-reactionsystem with the BvP oscillator for the membranepotential φ . Proof : See sections II - IV. (cid:3)
The Maxwell’s equations (52) - (56) are now com-pletely described in the macroscopic domain Π ofphysical space similar to those of classical electromag-netic waves, but have two unique properties: (i) The firstproperty is the presence of the additional point chargeand the current density that are not directly induced byconductivity. In the bidomain, these terms represent themembrane current and ion-pumped point charge, butfor the classical electrodynamics in the physical domain,they are often regarded as the sources of charges. Thena question arises on how we interpret these sources. (ii)The second property is the presence of a varying pointcharge χ α for the electromagnetic field. Depending onthe activation of ion pumps, the magnitude of pointcharge changes or even the signs of it changes. Thisphenomenon may not exist in the real world since it isnot likely to obey the conservation of charges. Then,4what is the role of point charge χ α if it is not to perturbthe conservation of total energy or momentum of thetotal system?These non-classical components can be made intoanalogies of several physical phenomena. Leaving therole of χ α to quantum theory, the first unique propertywill be explained in the next subsection in the perspectiveof semiclassical theory of radiation . B. Interpretation by the semiclassical theory of radiation
Let us consider the excitation propagation as the elec-tromagnetic field generated by a traveling solid cation .Let the radiation of the excitation propagation mean thesame as the classical electromagnetic waves such that thecell can be excited only by radiation, or the transverseelectromagnetic waves excluding the weak Coulomb in-teraction between the cations. Then we soon realize thatsome energy should be radiated by this cation to excitethe media. Thus, the total energy of this cation shoulddiminish as it travels. In other words, the energy loss oc-curs at every change of velocity of the cation, thus evenif the states of motion of the particle is the same, the en-ergy of it may be different. But, the energy of the cationremains the same all the time in a perfect homogeneousmedia. Otherwise, the excitation should depend on thedistance from the initial source.A similar argument is applied to the radiation of pho-ton. If we assume that a photon is a solid particle, thenlight is radiated by an accelerating photon, thus a subse-quent energy loss occurs. But, similar to the excitationpropagation, the energy of the photon does not dependon the distance from the source in non-dissipative media.This contradiction gives birth to an idea of reaction or self-force to compensate for the energy loss. Consider aparticle with mass m traveling with acceleration ˙ v . Sincethe energy radiated per unit time by this particle is givenby (2 / e ˙ v /c , the external force K should be expressedwith the self-force K s for this energy loss such as K + K s = m ˙ v , where K s can be easily obtained as (2 / e /c )¨ v . Inthe excitation propagation, the acceleration of the cationdoes not require the existence of the solid object of thecation, but can be deduced by the shape of the wavefrontfrom the geometric relation between trajectory and wavefront . The source of the energy of K s has remainedlargely unknown in a vacuum, but should respond im-mediately to the acceleration of the moving particle tomaintain its total energy.A similar idea has been adapted to the semiclassicaltheory in the self-consistent equations . The onlymodification is the use of a dipole moment and polariza-tion induced by the electromagnetic field. This dipolemoment being induced by the field interacting with theatom yields the generation of polarization density P ( R , t ) which acts as a source in Maxwell’s equations such that ∇ × ( ∇ × E ( r , t )) + 1 c ∂ E ( r , t ) ∂t = − µ i ∂ P ( R , t ) ∂t . (57)Comparing this equation with equations (54) and (56)confirms that the membrane current density is only sub-stituted by the second derivative of polarization densitywith respect to time. Moreover, since polarization den-sity can be expressed by the membrane current density,we may say that equation (57) is actually in the sameform as equations (54) and (56). Let us consider theequality relation that the polarization current ˙ P is thesame as the polarization current as the difference betweenthe current density J and the magnetization current J g such as ˙ P = J − J g , where J g = ∇ × M , where M is the magnetization density derived as P α ´ uχ α r α × ˙ r α δ ( r − r α ) du for the discretized chargedensity. The membrane current density is just the por-tion of the current density except for the conducting cur-rent and the magnetic polarization such that J m = λ λ ( J − J g ) . (58) V. CONSERVATION OF TOTAL CHARGE,MOMENTUM AND ENERGY
The Maxwell’s equations (52) - (56) are now com-pletely described and represent the diffusion-reaction sys-tem with the BvP oscillator in real space. But, it remainsin question whether the system of these equations (52)- (56) is under conservational laws in views of charge,momentum, and energy.
A. Conservation of total charge
In principle, conservational laws fails if only one do-main of the bidomain is considered, but the total num-ber of point charges should remain constant in the bido-main Π if no flux occurs at the boundaries. For sim-plicity, consider that only the cations are propagatedalong the velocity vector of the excitation propagationand the cations are transported through the membrane.For point charge χ α at r α , let us consider the operator m + : χ α → I + and m : χ α → I + returning the total num-ber of the cation in π i and π o , respectively, such that m + χ α ( r ) = N i , mχ α = N o , where N i and N o are the total number of the cations in π i or π o , respectively. Let Q be the electric charge ofone cation both for π i and π o . Then it is easy to verifythat Q ( m + − λ − m ) χ α = χ α . Moreover, let us introduce5the operator τ + which transports one cation from π o to π i , called the membrane influx operator. Similarly, theoperator τ which transports one cation from π i to π o ,is called the membrane efflux operator. Then it can beeasily verified that m + ( τ + χ α ) = N i + 1 , m ( τ χ α ) = N i − . Let P be the ion pump operator which transports a num-ber of the cations from π i to π o or vice versa. Note that P is no more than the combinations of τ + and τ − suchthat P = ( τ + ) N + +( τ ) N − for the number of the transport N + and N − of each operator τ + and τ , respectively. Forexample, after ion pumping, a new point charge χ newα at r α can be obtained such that χ newα = Q ( N i − λ − N o + N + − λ − N − ) . As we assume, χ α > N i > N o , but the signs of χ newα may vary such that χ newα = ( > , if N − − λN + < λN i − N o ,< , if N − − λN + > λN i − N o . Moreover, N − − N + is mainly determined by the BvPoscillator depending on the membrane potential φ ( r α , t ),thus we can say that χ α is also a function of themembrane potential φ such that χ α = χ α ( φ, r , t ). Thenthe conservation of total number of the cations and totalcharge is expressed as follows: Proposition 3 : Suppose that there is no flux ofcharged particles at the boundaries of domain Π. Then,in the dynamical system for the Maxwell’s equations(52) - (56), the total number of the cation for the finitenumber N of point charge is conserved in Π such that N X α =1 ( m + + m ) χ α ( t ) = N X α =1 ( m + + m ) χ α (0) , ∀ t > . (59)Let Q i be the electric charge of the cation in π i and let Q o be the electric charge of it in π o . If Q o is differentfrom − λQ i , then the total charge P α χ α is not conservedsuch that N X α =1 χ α ( t ) = N X α =1 χ α (0) , any t > . (60) Proof : The total number of point charges N is fixed andthe only changes can be made by the ion pump operator P . Thus, it is enough to show that the above quantitiesare conserved by the operation of the membrane flux op-erator τ + and τ because P is solely a function of τ + and τ . By applying the membrane influx operator τ + to anumber of the cations at r α of equation (59), we obtainthe conservation of the total number of the cations as( m + + m ) τ + χ α = ( N i + 1 + N o −
1) = ( m + + m ) χ α . But note that the total difference of the number of thecation, i.e. P α ( m + − m ) χ α , is not preserved as( m + − m ) τ + χ α = ( m + − m ) χ α + 2 . Similarly, by applying τ + to the total charge of thecations at r α (60) and by using Q o = − λQ i , we obtainthe conservation of total charge as τ + χ α = (cid:0) Q i m + − λ − ( − λQ i ) m (cid:1) τ + χ α = χ α . If Q o = − λQ i , then we can easily verify that the totalcharge is not conserved. For example, with the sameelectric charge Q for one cation both in π i and π o , totalpoint charge P α χ α is not preserved as τ + χ α = N X α Q ( m + − λ − m ) τ + χ α = χ α − − λ, though the magnitude is bounded due to the conserva-tion of the total number of the cation. Similar argumentscan be easily shown for the membrane efflux operator τ (cid:3) .Proposition 3 implies the different interpretation ofion pumps in the macroscopic domain Π. In the biolog-ical tissue, ion pumps simply transport a charged ionfrom π i to π o or vice versa. Thus, ion pumps initiatethe change of locations of the ion while the electriccharge of the ion remains unchanged. In Π, however,ion pumping only changes the sign and the electriccharge of the ion, but does not change the locationof it. Moreover, the equality condition ( Q o = − λQ i )leads to the conservation of point charge such that χ α never changes its signs and magnitude independent ofion pump P . But, in the bidomain, it is natural to set Q o as the same sign and magnitude of Q i , thus it isinevitable to violate the conservation of charge in Π,which causes many peculiar properties in the mechanismof the propagation of the biological waves different fromphysical waves. In the next sections, we will observe theeffects of this varying point charge χ α on the Lagrangianand Hamiltonian of the Maxwells’ equations (52) - (56). B. Conservation of total energy and momentum
Before proceeding further, we will first verify that theNewton-Lorentz equation still holds for the Maxwell’sequations (52) - (56) which are necessary for the proofof the conservation of total energy and momentum.
Lemma 1 : Let v α be the velocity of the particleindexed α which has mass m α and point charge χ α .Then, the Newton-Lorentz equation is valid for theMaxwell’s equations (52) - (56) such that m α d v α dt = χ α [ E + v α × B ] . (61)6 Proof : Since the point charge lies microscopically eitherin π i or π o , axiom 2 and 3 imply that it is sufficient toshow that equation (61) expresses the Newton-Lorentzequation for each microscopic domain π i or π o . If thepoint charge lies in π i , then by equation (22), the aboveequation reduces to m α d v iα dt = q iα (cid:2) E i + v α × B i (cid:3) , which is just the Newton-Lorentz equation in the inter-cellular space π i . On the other hand, if the point chargelies in π o , then, by equation (22), equation (61) reducesto m α d v oα dt = q oα [ E o + v α × B o ] , which is just the Newton-Lorentz equation in theinterstitial space π o (cid:3) .Using the Newton-Lorentz equation (61), we ob-tain two propositions on the conservation of the totalenergy and the total momentum of the closed dynamicalsystem. These conservational laws are actually the sameas the classical Maxwell’s equations with the Coulombgauge . The conservation of total energy and momen-tum is the direct consequence of the Newton-Lorentzequation on the supposition of axiom 2 and 3, and theintact form of equations (54) and (56) resulting from thefact that the membrane current density is a point-wisecurrent which only changes point charge χ α withoutadding charge current χ α v α . Thus, the proofs are similarand will be provided in Appendix I for interested readers. Proposition 4 . Consider a closed domain Π suchthat no flux occurs at the boundary. Then the energyof moving particles with mass m α traveling in theelectromagnetic field for the Maxwell’s equations (52) -(56) is well defined as U = X α m α v α + ε i ˆ (cid:2) E + c B (cid:3) d r (62)and is conserved in Π independent of time. Proof : With the Newton-Lorentz equation (61),see Appendix IA.
Proposition 5 . Consider a closed domain Π suchthat no flux occurs at the boundary. Then the totalmomentum of moving particles with mass m α travelingin the electromagnetic fields for the Maxwell’s equations(52) - (56) is well defined as P = X α m α v α + ε i ˆ [ E × B ] d r (63)and is also conserved in Π independent of time. Proof : With the Newton-Lorentz equation (61),see Appendix IB.
VI. LAGRANGIAN
To study the effects of the time-varying point charge χ α and the membrane current density J m on the classicalor quantum mechanical paths of the cations in Maxwell’sequations (52) - (56), we consider the most popularly-used Lagrangian L , known as the standard Lagrangian ,for the system of the particles and the electromagneticfield : L ( r ) = X α m α v α + ˆ L ( r ) d r, (64)where L ( r ) is called the Lagrangian density and is ex-pressed as L ( r ) = ε i (cid:2) E ( r ) − c B ( r ) (cid:3) + J ( r ) · A ( r ) − ρ ( r ) φ ( r ) . (65)In equation (65), the first bracket represents the La-grangian of the moving particles, J ( r ) · A ( r ) the La-grangian of the electromagnetic fields, and ρ ( r ) φ ( r ) theinteraction between the charge particles and the field.Note that the Lagrangian (64) only holds for π i and π o ,not for π i ∩ π o because the particles do not stay in themembrane as mentioned in axiom 3. In this section,we will study the difference between the standard La-grangian of the Maxwell’s equations (52) - (56) and thatof the classical electrodynamical waves with the Coulombgauge ∇· A = 0 . This Lagrangian is known to be gaugeinvariance, thus the use of the new gauge ∇ · A = − φ does not change the Lagrangian, while J m can signifi-cantly change it. We are particularly interested in therole of ion channels on the change of the trajectory, orequivalently the wavefront, of the excitation propagation,which is equivalently represented in the Maxwell’s equa-tions (52) - (56) as the dependency and sensitivity ofthe standard Lagrangian on χ α . The Coulomb gauge ispreferred over the Lorentz gauge because the excitationpropagation is considered from the non-relativistic pointsof view in this paper. The validity of the standard La-grangian in bidomain Π can be easily verified by showingthat the standard Lagrangian is valid for each π i and π o ,but this will not be shown here. A. Contribution of χ α and J m on the Lagrangian In order to study whether the Lagrangian is modifiedby the gauge choice and J m , the Lagrangian should beexpressed as L ( r ) = X α m α v α + L ( k ) d k, where ffl indicates the integration over the domain for Re ( k ) >
0. Using the fact that all the fields are real,we can also express the Lagrangian density in reciprocal7space as L ( k ) = ε i h | e k ( k ) | − c | b k ( k ) | i + j ∗ k ( k ) · a k ( k )+ j k ( k ) · a ∗ k ( k ) − ρ ∗ k ( k ) φ k ( k ) − ρ k ( k ) φ ∗ k ( k ) . (66)Then, a lemma follows immediately. Lemma 2 : The Lagrangian of the Maxwell’s equations(52) - (56) can be expressed in reciprocal space as L ( k ) = − ρ k ρ ∗ k ε i k + ε i (cid:2) ˙ a ⊥∗ k · ˙ a ⊥ k − c k a ⊥∗ k · a ⊥ k (cid:3) (67)+ (cid:2) j ⊥∗ k · a ⊥ k + j ⊥ k · a ⊥∗ k (cid:3) , (68)which is the same as that of the classical Maxwell’sequations with the Coulomb gauge. Proof : See Appendix II.Let us pay our attention to the first term of theLagrangian density in equation (68), known as the
Coulomb energy of a system of charges . As mentionedin equation (48), charge density can be divided into twocomponents: The charge density caused by conductingand the reactive membrane charge density such as ρ = ρ c + ρ r . Substituting this decomposition into theCoulomb energy yields ρ k ρ ∗ k ε i k = 1 ε i k [ ρ ck ( ρ ck ) ∗ + ρ ck ( ρ rk ) ∗ + ρ rk ( ρ ck ) ∗ + ρ rk ( ρ rk ) ∗ ] , which shows the effects of the reactive membrane chargedensity ρ r caused by j r in the Lagrangian. Moreover,using the discrete expression of ρ in reciprocal space suchas ρ k ( k ) = P α χ α ( t )(2 π ) − / e − i k · r α , the integration ofthis term is given by1 ε i ˆ ρ ∗ k ( k ) ρ k ( k ) k d k = 18 πε i X α χ α ( t ) r (2 π ) / + X α = β χ α ( t ) χ β ( t ) | r α − r β | . (69)The first term represents the Coulomb self energy ofthe particle α and the second term represents theCoulomb interaction between the particles α and β .Therefore, the changes of the Lagrangian due to themembrane current density J m is implicitly expressed inthe Coulomb energy. The membrane current densityonly changes the magnitude and the signs of χ α withoutmodifying the total number of them, thus the effectof the membrane current density is reflected in the qualitative characteristics of χ α . Nevertheless, theaction of the above Coulomb potential can be regardedthe same as that with a constant point charge as shownin the following lemma with a new definition: Definition : Suppose there exists a scalar function F αβ : R + → R + such that the time integration of χ α χ β is equal to F αβ ( t ) as ˆ tt χ α ( t ) χ β ( t ) dt ′ = F αβ ( t ) , ∀ α, β, (70)where the time t is the minimum value of the latesttime for the constant resting value of χ α and χ β . Then,the varying point charge χ α is called the time-integrable . Lemma 3 : Suppose that χ α ( t ) is time-integrablefor all indexes α . Then the action of the Coulombenergy of a system of charges (69) with χ α ( t ) is the sameas the action with a time-independent χ α (0). Proof : Differentiating equation (70) with respectto t , we obtain χ α ( t ) χ β ( t ) = χ α ( t ) χ β ( t ) + dF αβ ( t ) dt . By substituting the above equality into equation (69)and by using the fact that the extremes of the actionremain the same by the factor of dF/dt , we reach theconclusion (cid:3) .The condition (70) is actually valid only if the ac-tion potential can be approximated as a polynomial. Infact, the action potential is sufficiently smooth, thusthe polynomial approximation of the action potential iswidely used explicitly or implicitly in most mathematicaland computational modeling. Therefore, we may acceptthis condition naturally without more restrictions. As aconsequence, the time-varying χ α of the Coulomb energydoes not change the action, but the following propositionshows that it contributes to the change of the action bymodifying the Lagrangian of the electrodynamic field. Proposition 7 : Suppose that χ α ( t ) is time-integrablefor all indexes α . Then the standard Lagrangian (71)and (72) of the Maxwell’s equations (52) - (56) is givenby L ( r ) = X α m α v α − πε i X α χ α ( t ) r (2 π ) / − πε i X α = β χ α ( t ) χ β ( t ) | r α − r β | + ˆ L ( r ) d r, (71)where the Lagrangian density L ( r ) is L ( r ) = ε i h(cid:0) E ⊥ (cid:1) − c B i + ( J c ) ⊥ · A ⊥ , (72)which is independent of the reactive membrane currentdensity J r , but depends on the time variation of pointcharge χ α . Moreover, the action induced by χ α ( t ) witha constant velocity v α (0) is the same as that by constantpoint charge χ α (0) with a time-dependent velocity v ′ α ( t ).8 Proof : The derivation of the Lagrangian (71) with (72)is obtained directly from equation (68) of Lemma 2and equation (69). The remaining task is to prove theindependency of the Lagrangian density on J r . But,this is also a direct result from the choice of j rk (39)because j rk is only in the direction of k and consequently,( j rk ) ⊥ ( k ) = 0 or ( J r ) ⊥ ( r ) = 0. Since the first termin the Lagrangian density remains constant indepen-dent of χ α , we only need to study the second term( J c ) ⊥ · A ⊥ . Let us decompose χ α into two componentsas χ cα , or the point charge induced by the conducingcharge density ρ c , and χ rα , or the point charge inducedby the reactive membrane charge density ρ r , such as χ α = χ cα + χ rα . Then, with the discrete expression of J c = P α χ cα v α δ [ r − r α ], the action by the Lagrangiandensity for the electrodynamic field S is given by S = ˆ t t ˆ ( J c ) ⊥ · A ⊥ d rdt = ˆ t t X α χ cα ( t ) (cid:0) v α · A ⊥ (cid:1) dt. Since χ α is time-integrable, it is easy to show that thereexists a function G α ( t ) : R → R such that G α ( t ) ≡ ˆ tt χ cα ( t ) χ cα ( t ) dt ′ , (73)where the time t is the latest time for the constant rest-ing value of χ α . Then the above equation reduces to S = ˆ t t X α χ cα (0) (cid:0) v ′ α ( t ) · A ⊥ (cid:1) dt, where we introduced the new velocity v ′ α =( dG α ( t ) /dt ) v α . The proof is done only by observ-ing that the above equation is the action of theLagrangian of electrodynamic field with a constant pointcharge χ α (0) with velocity v ′ ( t ) (cid:3) .The following corollary also may show the practi-cal interpretation of proposition 7. Corollary : Suppose that χ α ( t ) is time-integrablewith G α for all indexes α . If G α is the same for allindexes α , then the Lagrangian (71) with (72) is thesame as the Lagrangian of the classical electrodynamicsin homogeneous media. On the other hand, if G α isdifferent for all indexes α , the Lagrangian (71) with(72) corresponds to the Lagrangian of the classicalelectrodynamics in inhomogeneous media.Preposition 7 and corollary imply mostly two cru-cial characteristics of the excitation propagation: (i)The first characteristics is obviously that the operationof ion channels can be translated as the changes ofmaterial properties. The time variation of point chargeis only induced by ion channels, but proposition 7implies that this time dependency of point charge canbe shifted to the time-dependent velocity that can be realized as the varying conductivity property of media.In the context of the original definition of geometry,any object to change the trajectory of the propagation,we may say that ion channels can be also regardedgeometry, in addition to the shape of the domain andthe conductivity property of media. (ii) The secondcharacteristics is that the membrane current densitycan only change the Lagrangian of the electrodynamicfield. In other words, this means that the membranecurrent does not modify the Lagrangian of the Coulombenergy, moving particles, or interaction between theparticles and the fields. As we observe later from theHamiltonian, the non-interference of the membranecurrent density especially to the interaction betweenthe particles and the fields, gives birth to the simplestexcitation system, the same as that of light propagation. B. Huygens’ principle and the eikonal equation
Proposition 7 and corollary 1 suggest that the trajec-tory of the excitation propagation is the same as thetrajectory of light propagation in the homogeneous andisotropic media with normal ion channels with a propercondition as mentioned as a supposition. But this factcould turn out to be of no surprise when we compare thefundamental mechanism of the excitation propagationwith that of light propagation, known as the Huygens’principle saying : Each element of a wave-front may be regarded asthe centre of a secondary disturbance which gives rise tospherical wavelets; and moreover, that the position of thewave-front at any later time is the envelope of all suchwavelets.
However, no better description can be given thanthe above principle to the mechanism of the diffusion-reaction system, such as the classical FHN equationsfor the excitation propagation. If we replace secondarydisturbance and spherical wavelets with reaction and diffusion , respectively, without losing its meaning, theabove description of the propagation remains intactfor the diffusion-reaction system. Consequently, with-out considering the velocity of the propagation, thetrajectory and wavefront should remain same for bothpropagations.If the two different systems share the same propaga-tion mechanism, then their eikonal equation should becoincident. In geometric optics, the surface of light prop-agation is provided by the optical path S satisfying |∇S| = (cid:18) ∂ S ∂x (cid:19) + (cid:18) ∂ S ∂y (cid:19) + (cid:18) ∂ S ∂z (cid:19) = √ ε i µ i . (74)This equation is derived for regions that are sufficientlyfar from the sources, or equivalent for a sufficiently largevalue of the wave number when the electrodynamic field9is considered as a time-harmonic field. The Maxwell’sequations (52) - (56) equivalent to the FHN equations(4) - (5) cannot be written without source terms becauseof the presence of ion channels almost everywhere. Thus,it is very difficult to prove mathematically that theeikonal equation of Maxwell’s equations (52) - (56) orthe FHN equations (4) - (5) can be written the same asequation (74). This is mostly because nor charge density ρ nor the membrane current density J m being dividedby the wave number converges to zero even at a veryhigh frequency. Even the computational study for thecoincidence of the two eikonal equations are not trivial.Thus, in spite of strong inference from proposition 7 andcorollary 1, we put it as a conjecture for validation inthe later publications such that Conjecture : Suppose that χ α ( t ) is time-integrablefor all indexes α and the media is homogeneous andisotropic with constant conductivity. Then the eikonalequation of the Maxwell’s equations (52) - (56) is thesame as the classical eikonal equation of optics (74). C. With external electromagnetic field
We also can consider the effect of the external electro-magnetic field on the excitation propagation in the heart,especially focusing on the changes of trajectory and ve-locity. Consider that the electric field E ′ e and B ′ e beingmeasured in the vacuum are applied to the myocardialtissue. Let A ′ e and φ ′ e be the corresponding potentials inthe vacuum. To be represented in the same myocardialmedium, the field and potential measured in the vacuumshould be expressed as those in the microscopic domain π i or π o . Consider that the transformation of the fieldand potential between the vacuum and π i or π o can besimply performed by a linear projection operation H suchthat H E ′ e | π i = E ie , H E ′ e | π o = E oe , H B ′ e | π i = B ie , H B ′ e | π o = B oe . Similar operations can be applied to the potentials toyield A ie ( A oe ) and φ ie ( φ oe ). For example, if we consider thefield as the consequential phenomena from the movingcharge, then we may consider the operator H as the lineartransformation caused by the proportional changes of thevelocity of the moving charge from the vacuum to thebidomain or vice versa. But, according to axiom 1, themaximum velocity of the signal is the same in π i and π o ,thus the operator H should be the same operator for thefield and potential in π i and π o , but only differentiateaccording to the location of q α . The weighted differenceof the external field and the external potential in Π istherefore defined as E e ≡ E ie − λ E oe , B e ≡ B ie − λ B oe , A e ≡ A ie − λ A oe , φ e ≡ φ ie − λφ oe . According to the above definitions, the strength of theexternal field or the external potential depends on theparameter λ defined as λ = p ε i /ε o = p µ o /µ i . For ex-ample, if λ = 1, E e and B e are always zero because weregard that the operator H performs the same for π i and π o . Consequently A e and φ e are zero. Thus, there willbe no effect of the external electromagnetic field. How-ever, this does not reflect the real phenomena as shownin refs. , but it is more reasonable to choose λ dif-ferent from 1 . E e and B e are roughly proportional to E ′ e and B ′ e ,respectively, and similarly A e and φ e to A ′ e and φ ′ e , re-spectively. When the external field is applied to Π, thenew Lagrangian density is given by L ( r ) = ε i h(cid:0) E ⊥ (cid:1) − c B i + ( J c ) ⊥ · ( A ⊥ + A ⊥ e ) − ρφ e . (75)Or, by means of the standard procedure of the Power-Zienau-Woolley transformation , L ( r ) = ε i h(cid:0) E ⊥ (cid:1) − c B i + M · ( B + B e ) + P · ( E ⊥ + ( E e ) ⊥ ) − ρφ e , (76)where P ( r ) is the polarization density and M ( r ) is themagnetization density such as P ( r ) = X α ˆ χ α r α δ [ r − p r α ] dp, M ( r ) = X α ˆ pχ α r α × ˙ r α δ [ r − p r α ] dp. If the external field and potential is sufficiently large, A e modifies the Lagrangian of the electrodynamic field and φ e modifies the Lagrangian of the interaction betweenthe particle and the field. Leaving the effects of φ e on the Lagrangian of the interaction to the study ofthe Hamiltonian in the next section, the rest of thissection focuses on the effect of A e on the Lagrangianof the electrodynamics field. Consider the followingproposition: Proposition 8 : Suppose that χ α is time-integrablefor all α . Suppose that the perpendicular componentto the wave vector of the external vector potential A e during time t to t is non-trivial for χ α located at r α . Then, applying the electric potential A e to thebidomain Π causes the changes of the propagationalvelocity of χ α . Moreover, if A e is in the oppositedirection to A , then there exists the critical magnitudeof the electric potential A ∗ e to stop the propagation of χ α . Proof : For the Lagrangian density (75), the ac-tion by the Lagrangian of the electrodynamic field is0given by S = ˆ t t X α χ cα ( t ) v α · ( A ⊥ + A ⊥ e ) dt = ˆ t t X α χ cα ( t ) v ′ α · A ⊥ dt, where v ′ α ≡ v α · (1 + A ⊥ · A ⊥ e ). Note that this actionis the same as the action without the external field,but with a different velocity. Thus, it can be deducedthat applying A e only leads to the changes of velocityof χ α without considering the interaction between theparticle and the field. The existence of the critical A ∗ e for the stopping of the propagation can be directlyinferred from the existence of A ∗ e to satisfy the equality A ⊥ · A ⊥∗ e = − .
0, which implies that A ⊥∗ e should be inthe opposite direction to A ⊥ (cid:3) .If A ⊥ e is in the same direction as A ⊥ , then it is conjec-tured that the propagational velocity increases up to themaximum velocity c , defined as c = √ ε i µ i = √ ε o µ o . Asposed in axiom 1, the propagational velocity is assumedto remain constant not exceeding c independent of A ⊥ e that is greater than a certain magnitude, though whatactually happens in vivo is unknown. VII. HAMILTONIAN
The Hamiltonian of the Maxwell’s equations (52) -(56), an operator associated with the total energy of thesystem, also may help us to enhance the understandingof (i) the excitation mechanism of the excitation propa-gation and (ii) the effect of the interaction between theintrinsic or external electrodynamic field and the cationfor the excitation of the myocardial cell. Let us beginwith the following lemma.
Lemma 4 : The Hamiltonian of the Maxwell’s equations(52) - (56) is expressed as H = X α m α [ p α − χ α A α ( r α )] (77)+ X α πε i (2 π ) χ α r + ε i H ( k ) d k, (78)where the Hamiltonian density H in reciprocal space isderived as H ( k ) = (cid:0) ˙ a ⊥ k (cid:1) ∗ ˙ a ⊥ k + c k ( a ⊥ k ) ∗ · a ⊥ k . (79)Proof: See Appendix IIIA.The first term of the Hamiltonian (78) representsthe kinetic energy of the particles located at r α withmomentum p α = ( ~ /i ) ∇ α where the value of theconstant ~ is unknown. Note that ~ is different from the Planck constant 6 . × − m kg/s and shouldbe defined such that the energy of the monochromaticwave of the Maxwell’s equations (52) - (56) with thefrequency ω is expressed as E = n ~ ω for an integer n . Intuitively, this modification is required because thetraveling particle is the cation which carries a possiblylarger energy and momentum than those of photon.The second term corresponds to the Coulomb energy,and the last term corresponds to the radiation energyof the transverse field, or the energy generated by theperpendicular component of the field to the wave vector k . Similar to the Lagrangian, the only difference of theHamiltonian (78) to that of the Maxwell’s equations withCoulomb gauge is the presence of the time-varying pointcharge χ α ( t ). To understand the impact of χ α ( t ) on theHamiltonian, the Hamiltonian (78) will be expressedwith more distinguishable components by solving thebracket and using the normal variables from the secondquantization . Proposition 8 : The Hamiltonian of the Maxwell’sequations (52) - (56) is given by H = H + H R + H C + H I + H I + H I , (80)where H p = X α p α m α , (81) H R = ~ ω (cid:18) a + a + 12 (cid:19) , (82) H C = 18 πε i X α χ α ( φ, t )(2 π ) / r + X α = β χ α ( φ, t ) χ β ( φ, t ) | r α − r β | , (83) H I = − X α χ α m α p α · A ( r α ) , (84) H I = − X α g α χ α m α S α · B ( r α ) , (85) H I = X α χ α m α A ( r α ) , (86)where S α is the spin of the particle α and g α is theLand´e factor. Proof : See Appendix IIIB.The particle Hamiltonian H p represents the kineticenergy of the particle with mass m α and momentum p α = ( ~ /i ) ∇ α , independent of χ α . The radiation fieldHamiltonian H R depending on the operator a + and a , known as the creation operator and annihilationoperator of the cation in the single mode, representsthe energy of the radiation field with frequency ω . Ifthe propagating cations have various kinds of ions withvarious polarizations, then equation (82) should bewritten as the sum over all the mode j for corresponding1 ~ j , but both for simplicity and reflecting reality, wesuppose that all the propagational cations are the samekind with the same polarization. Note that H p and H R are independent of χ α ( t ) and consequently independentof the reactive membrane current by ion channels. The Coulomb Hamiltonian H C is in the same form as that ofthe Lagrangian. H p , H R and H C contain the dynamical variables of theparticle or the transverse field, but the other Hamiltoni-ans, known as the interaction Hamiltonian , contain bothdynamic variables of the particle and the transverse fieldto indicated the interaction between them. The inter-action Hamiltonian H I consists of three different com-ponents: (1) H I represents the energy caused by themomentum of the cation α in the direction of the poten-tial A . (2) H I represents the spin energy of the cation α caused by the magnetic field B . (3) H I representsthe kinetic energy of the oscillatory forced motion by A .Note that H C and H I are all dependent on χ α ( t ), thus,consequently, dependent on the reactive membrane cur-rent density by ion channels.In the next subsections, the Hamiltonian H will be di-vided into two components: One is the unperturbed H and the perturbed H I where H = H − H I . The motiva-tions for this decomposition are well described in ref. ,but will be described here in brief. The unperturbed H contains the radiation field Hamiltonian H R and the par-ticle Hamiltonian H p , thus the quantum state or energystate of H , representing the free particle in the radiationfield, remains constant during the propagation. On theother hand, the quantum state of the perturbed Hamilto-nian H I changes according to the interaction between theradiational field and the myocardial cell. The CoulombHamiltonian is the only undetermined component, butwill be assorted as the unperturbed Hamiltonian to pro-vide the lowest quantum number for the resting state. FIG. 4. Feynman diagram of the cardiac excitation withoutthe membrane current. π i is stationary. k and k ′ are thewave vectors. n and n ′ are the number of the incident andemitting cations, respectively. A. Transition amplitude without the membrane current
In order to understand how the Hamiltonian is relatedto the cardiac excitation propagation, the transition am- plitudes for the excitation mechanism of the myocardialcells will be derived from Feynman diagram . Let usstate the following actions of the interaction betweenthe cations and the myocardial cells reflecting the realbiological phenomena:(ACTION i) The cation travels in the space π i and time t > π i is stationary for allthe time t > π o fluctuates in thespace π o and exchanges the cations with the myocardialcells.(ACTION iv) The cardiac cell absorbs and emits thecations.Note that these actions are similar to the interac-tion between electrons and photons in quantum opticswhere (ACTION iii) is analogous to the interaction be-tween protons and electrons, but are strikingly differentin (ACTION ii) because electrons also travels in space.(ACTION i) and (ACTION ii) are obvious. (ACTIONiii) is also clear since the exchanges are induced by ionchannels. The expression of (ACTION iv) could be lessfamiliar, but is equivalent to other popular terminologiessuch as the safety factor (SF) which measures the ratiobetween the inward axial current ( I in ) with the capacitycurrent ( I c ) and the outward axial current ( I out ) of amyocardial cell defined as SF ≡ (cid:18) ˆ I c dt + ˆ I out (cid:19)(cid:30) ˆ I in dt, (87)where each current is integrated over the time inter-val only when ρ is positive. Successful conduction thusmeans SF > π i . Then, (ACTION i) and (ACTION iv)correspond to multiple wavy lines in Figure 4A if themotion of each cation is represented by a single wavyline. Since the time marches from bottom to top in thediagram, two bottom wavy lines represent two cationstraveling from other places and being absorbed by a my-ocardial cell. Accordingly, the points 1 and 2 correspondto the annihilation of a cation denoted by a term in equa-tion (82). On the other hand, three upper wavy linesrepresent three cations being emitted by the same my-ocardial cell and traveling to other places. Thus, thepoints 3 , , a + term in equation (82). Without consideringthe capacity current, we can say that the safety factorfor this myocardial cell is 1 . (cid:12)(cid:12) ψ i (cid:11) and | ψ o i be the2state vector representing the energy spectral of π i and π o , denoted by the quantum number { n i } and { n o } ,respectively. Let us define the relative state vector | ψ i isdefined as the superpositions of these two states such that | ψ i ≡ (cid:12)(cid:12) ψ i (cid:11) − λ | ψ o i where the weight factor λ is obtainedfrom the normal variable α for the operators a and a + .In quantum number, it is equivalent to n i − λn o . Then,the relative state vector | ψ i can be categorized as follows:(STATE i) Resting state: | ψ i = . (STATE ii) Excited state: | ψ i > n , for a fixed n ≫ . (STATE iii) Refractory state: | ψ i ≤ . The resting state (STATE i), equivalent to the vacuumstate in electrodynamics, is the direct consequence ofequation (82) with aa + = n = 0. For (STATE ii), n is related to the threshold of the membrane potential,but is fixed and constant in homogeneous media. Therefractory state (STATE iii) looks like an unnaturalphenomenon, but is in fact a natural one, even fromthe point of view of physics, and will be explained indetail in the next subsection. Let | ψ i i and | ψ f i bethe initial and final state vector. Then the transitionamplitude is given by h ψ f | U ( t f , t i ) | ψ i i for the evolutionoperator U ( t f , t i ) which transforms the initial state | ψ i i at t i into the final state | ψ f i at t f . For example,exp[ − iE ( t f − t i ) / ~ ], the solution of the Schr¨odingerequation i∂ψ/∂t = Eψ , represents the free evolution ofthe energy state E from time t i to t f .For the sake of simplicity, multiple absorption oremission cations will be represented by a single wavyline, but with different energy, as shown in Figure 4B.But, this requires the following axiom: Axiom 4 : The propagating cations are identicaland indistinguishable, obeying the Bose-Einstein statis-tics.This axiom is obvious because the cations with differentpolarization are either rare in biological tissue or makeno difference in functionality, especially in generating themembrane potential. Let k and k ′ be the wave vector ofan incident and emitting cation, respectively. Let n and n ′ be the number of the incident cations and emittingcations, respectively. The polarization of each cations isdisregarded. Let ω and ω ′ be its corresponding angularfrequency such that ω = c | k | and ω ′ = c | k | ′ for thespeed c of the signal. Let t be the resting phase, t bethe time when the cations are absorbed, t be the timewhen the cations are emitted and t be the time whenit is back to the resting phase. The letters r (restingstate), s (excited state), and f (refractory state) nextto the solid line indicate the quantum number for eachprocedure. Then Figure 4B is expressed by the following evolution operator:exp (cid:20) − i ~ ( E f + ~ n ′ ω ′ )( t − t ) (cid:21) h f |H I | s i× exp (cid:20) − i ~ E s ( t − t ) (cid:21) h s |H I | r i (88) × exp (cid:20) − i ~ ( E r + ~ nω )( t − t ) (cid:21) . The last component, exp [ − i/ ~ ( E a + ~ nω )( t − t )], in-dicates the unitary transformation with respect to the en-ergy level E r + ~ nω from t to t . The fourth component, h s |H I | r i , corresponds to the change of the states from r to s by the interaction Hamiltonian H I , due to the ab-sorption of the cation at time t , in the Schr¨odinger rep-resentation followed by the unitary transformation withrespect to the energy level E s from t to t . The secondcomponent, h f |H I | s i , similarly describes the change ofthe states from s to f again by H I , due to the emissionof the cation at time t , followed by the unitary transfor-mation exp [ − i/ ~ ( E f + ~ n ′ ω ′ )( t − t )] with respect tothe energy level E f + ~ n ′ ω ′ .The total energy of the incident cations ~ nω andthe total energy E r of the myocardial cell depend onthe microscopic coherence of the incident cations whichcould be a reflection of the macroscopic geometry ofthe neighboring cells (This will be discussed in part IIof this series of papers). Since the energy E s is thesole parameter for the excited states, our only concernwill be on the interacting Hamiltonian H I changing thequantum state r into the quantum state s . Thus, it is nosurprise to find that each component of H I is a functionof χ α corresponding to the action of ion channels forthe induction of the membrane current density. Forexample, if χ α is constant, the total amount of energy H I remains constant and consequently h s |H I | r i is closeto zero if s ≫ r . Thus, the only possible way to changethe quantum states from r to s could be achieved bymodifying the spin energy H I such that the perturba-tion Hamiltonian is only restricted to H I + H I to yieldthat h s | ( H I + H I ) | r i is not trivial even if s ≫ r . But,if χ α varies according to the influx of ion channels, thenthe total energy of H I varies as well, thus the transitionamplitude h s |H I | r i is not trivial for any s and r . Wecan reach the similar argument if we suppose that ionchannels only response to the membrane potential andequivalently to H I , not to the momentum ( H I ) or thespin energy ( H I ). In summary, Lemma 4 : The interaction Hamiltonians (84) -(86) correspond to the activation of ion channels beyondthe threshold. Moreover, if ion channels are activated bythe (static) membrane potential, then the Hamiltonian H I (86) is only involved for the excited states.Note that the Hamiltonian H I (86) is proportionalto χ α , thus its energy changes quadratically as themembrane current occurs.3 FIG. 5. Feynman diagram of the cardiac excitation with themembrane current. π i is stationary and q o fluctuates. k and k ′ are the wave vectors. n and n ′ are the number of theincident and emitting cations, respectively, while m and m ′ are the number of the influx and efflux cations, respectively. B. Transition amplitude with the membrane current
Figure 4 and the transition amplitude (88) display thesimplest form of interactions between the cations and themyocardial cells, but in order to reflect more detail ofinteraction for more crucial behavior, another factor willbe included in Feynman’s diagram; point charge q o in π o .Similar results may be obtained with π o , but the use of q o seems to better reflect the complex mechanism of thecardiac excitation. According to (ACTION iii), the linesfor q o are slightly wavy as shown in Figure (5) since q o fluctuates in π o . The biggest advantage of introducing q o in the diagram is the strategical representation of themembrane current.Let t be the time when the incident cations are ab-sorbed, t be the time when the influx membrane currentoccurs, t be the time when the efflux membrane currentoccurs, and t be the time when cations are emitted.The wavy line between 2 and 5 indicate the influx mem-brane current because point 2 occurs later than point 5.Similarly, the wavy line between 3 and 6 indicates the ef-flux membrane current because point 6 occurs later thanpoint 3. The left arrow besides the line of π o indicatethat the time flows forward. Contrary to the depolariza-tion phase at time 2, the repolarization phase involvesboth of influx and efflux membrane current. Figure 5Billustrates this fact. But, if we consider each absorptionand emission by means of energy and momentum, Figure5A and 5B can be displayed by the same plot Figure 5C.Let n and n ′ be the number of incident cations, re- spectively. Let m and m ′ be the number of the cationsthrough the influx and efflux membrane currents, re-spectively. Suppose that the wave vector k and k ′ arethe same for the incident cations in π i and the cationsthrough the membrane. Also, we suppose that the num-ber n of the incident cations is sufficiently large suchthat the absorbed cations induce a sufficiently large elec-tric potential in π i to create a larger membrane potentialthan the voltage threshold ( φ th ). The most critical stepis to set up the quantum number between time 1 and 4when an active membrane current occurs. For simplic-ity we let this period share the same quantum number s , like a plateau resulting from the equivalence betweeninflux and efflux, but all the influx occurs prior to theefflux. Then, the evolution operator for Figure 5C willbe obtained asexp (cid:20) − i ~ ( E f + ~ ′ m ′ ω ′ )( t − t ) (cid:21) h f |H I | s i m ′ Y ℓ =1 × exp (cid:20) − i ~ ( E f + ~ ( n + ℓ ) ω ′ )∆ t ′ (cid:21) × m Y j =1 exp (cid:20) − i ~ ( E r + ~ ( n + j ) ω )∆ t (cid:21) h s |H I | r i× exp (cid:20) − i ~ ( E r + ~ nω )( t − t ) (cid:21) , where ∆ t ′ = (2 t − t − t ) / m ′ and ∆ t =( t + t − t ) / m . The evolution operator for theexcited state without any perturbation Hamiltonianimplies that the excitation state is considered the free evolution of various discrete energy levels of themyocardial cell induced by the membrane current . Thismodeling could be an excessive simplification of variousion movements through ion channels during the cardiacexcitation, but this better characterizes the criticalproperties of the excitation. C. Refractory period in QED
Another important application of Feynman diagramand the transition amplitude is the representation of therefractory period in the perspective of quantum electro-dynamics. The refractory period, indicated as the mem-brane potential below the resting potential as shown inFigure 6B, is one of the unique features of the action po-tential and has been regarded as the possible causes ofmany unexplained nonlinear phenomena in cardiac elec-trophysiology. In a region during the refectory periodshortly after the excitation propagates, the region be-comes inactive to any excitation ( absolution refractoryperiod ) or requires more excitation than normal ( rela-tive refractory period ). This inactivation is biologicallycaused by the inactivation of a voltage-gated sodiumchannel and the slowly closing potassium channel , butthe refractory period will be described only by quantum4electrodynamical concepts. This means that the refrac-tory period can be represented by Feynman’s diagramwithout introducing ion channels. The goal of this de-scription is to reveal the functionality of ion channels togenerate the refractory region and its easier mathemati-cal expression for important nonlinear phenomena, suchas atrial reentry, caused by the refractory region.In Figure 6A, the influx of the membrane current oc-curs at time 5 later than the emission of the propagationcations at time 4. But, if we consider this happens before the time 4, then everything looks similar except the timetravels backwards between times 4 and 5. Since the eventsduring times 3 - 5 happen almost continuously, the orderof these events may change. Then the sequence of eventsoccurs as follows: At time 3 when the efflux membranecurrent occurs, it marches to time 4 when the emissionof the propagation cations occurs. Then, suddenly timetravels backward to reach time 5 when the influx mem-brane current occurs and proceeds to the final time torestore the resting potential. The backward traveling intime looks impossible, but is a very natural phenomenonwhich has been beautifully recognized as positrons byFeynman .The concept of the positron has been devised to explainthe wave traveling backward in time to be annihilated toyield photons. This positron has often been observed inthe laboratory, which reveals the same as the electron,but is attracted to normal electrons . This is possiblebecause electrons can have both positive and negativecharges; the positron has only the positive charge, con-trary to the negative charge of the normal electrons. Thisphenomenon is also explained by negative energy statescreated by scattering in a potential, equivalent to Dirac’sHole theory that the vacuum is the sea of the nega-tive energy states except one hole that is occupied bypositively-charged particles. But, the positron can beannihilated if it collides with an electron, emitting pho-tons as a result. This is why the positron is known asthe anti-particle .However, this annihilation by collision is not likely tooccur in the process of the cardiac excitation because themyocardial cells are stationary and are separated by themembrane. Thus, the negative energy state exists rela-tively for a long period to account for the refractory pe-riod. Let us relate each time in Figure 6A to each phaseof the action potential in Figure 6B. Time 1 for the inci-dent cations is obviously related to the initiation of thedepolarization phase. Time 2 corresponds to the rapidincrease of the membrane potential above the threshold.Time 3 corresponds to the beginning of the repolarizationphase. The membrane potential continues to decreaseuntil it reaches the resting potential again at time 4, butthe emission of the propagation cations into π i resultsin the lower membrane potential than the resting poten-tial. However, the influx membrane current restores themembrane current up to the resting potential at time 5.These relations between points in the diagram and thephase of the action potential again reveal that the back- ward time marching from time 4 to time 5 corresponds tothe refractory period of which energy state can be con-sidered negative as (STATE iii) when the energy state ofthe resting potential is 1 / refractory region . It’s natural to say that the restingstate is in a positive energy state equivalent to a negatively-charged electron . According to the insightsfrom Feynman’s diagram, we may regard that themyocardial cell changes into a negative energy stateequivalent to positively-charged electron during therefractory period. Since the cation is also positivelycharged, the positively-charged cation cannot propa-gate easily into the positively-charged myocardial cell,contrary to the normal absorption of the cations bythe negatively-charged myocardial cell. This repulsionof the cations due to the same signs of charges as themyocardial cell leads to the non-excitability propertyof the refractory region. Note the similarities with theDirac’s Hole theory asserting the existence of a holeconsisting of the positively-charged electrons in the seaof negatively-charge electrons. Thus, independent of therestoring frequency of the muscle fiber, the myocardialcell cannot be excited during the refractory period. Thisleads to the following lemma: Lemma 4 : The refractory region corresponds tothe Dirac’s Hole filled with positively-charged electronsin the perspective of QED.However, the positively-charged electrons do notmean that the myocardial cell is filled with morepositively-charged ions. What actually happens is tothe contrary. The refractory period has more negativemembrane potential than the resting potential, thus theintercellular space π i is likely to be filled with a smallernumber of positive ions than that of the resting state.Instead, a lower membrane potential that the restingpotential should be interpreted as a negative energystate and correspondingly a cell in the refractory periodas an anti-(excitable)-cell . The term anti makes moresense when one particle and anti-particle collides to beannihilated to yield new particles, but in the myocardialsystem, the cells do not move and there are no chance ofcollision between a cell and an anti-cell, thus the use of anti may be not appropriate. D. Negative energy state in quantum operators
In this subsection, the relative state vector | ψ i willbe revisited here to better understand the meaning ofthe negative-state states for the refractory period. Theinflux membrane flux operator ( τ + ) and the efflux mem-brane flux operator ( τ ) that are introduced in SectionV will be used here again, but will be defined more rig-5 FIG. 6. Feynman diagram to represent the refractory period(A) and its correspondence to the action potential (B). orously. Similar to the annihilation a and creation a + operator , suppose that the membrane flux operators τ and τ + satisfy the canonical commutation relations;[ τ, τ ] = 0 , [ τ + , τ + ] = 0 , [ τ, τ + ] = 1 , where 1 means the identity in the space. Let us constructthe eigenstate of τ + τ as follows: First let the resting state | / i represent the relative quantum state 1 / n i − λn o is 1 /
2. The state | / i does not mean that n i = n o = 0, thus | / i is not the ground state either for π i and π o . Applying τ and τ + to this resting state | / i yields (cid:12)(cid:12)(cid:12)(cid:12)
32 + λ (cid:29) = τ + (cid:12)(cid:12)(cid:12)(cid:12) (cid:29) , (cid:12)(cid:12)(cid:12)(cid:12) − (cid:18)
12 + λ (cid:19)(cid:29) = τ (cid:12)(cid:12)(cid:12)(cid:12) (cid:29) . Note that the relative quantum state is no longer an inte-ger because λ is generally not an integer. The quantumstate | / λ i simply means that the quantum state of π i is larger than that of π o by 3 / λ . Similarly, thequantum state | − / λ i simply means that the quan-tum state of π o is larger than that of π i by 1 / λ . Thus,the positive or negative quantum states are well definedwith the relative quantum states. In general, the relativeeigenstates | n i and | − n i are defined as (cid:12)(cid:12)(cid:12)(cid:12) n + 12 + nλ (cid:29) = 1 √ n τ + n (cid:12)(cid:12)(cid:12)(cid:12) (cid:29) , (89) (cid:12)(cid:12)(cid:12)(cid:12) − n − − nλ (cid:29) = √ nτ n (cid:12)(cid:12)(cid:12)(cid:12) (cid:29) , (90)and subsequently, τ + τ (cid:12)(cid:12)(cid:12)(cid:12) n + 12 + nλ (cid:29) = (cid:18) n + 12 + nλ (cid:19) (cid:12)(cid:12)(cid:12)(cid:12) n + 12 + nλ (cid:29) . On the other hand, because the annihilation ( a ) and cre-ation ( a + ) operator are applied to π i only, the relativeeigenstate | n i and | − n i are derived as (cid:12)(cid:12)(cid:12)(cid:12) n + 12 (cid:29) = 1 √ n a + n (cid:12)(cid:12)(cid:12)(cid:12) (cid:29) , (cid:12)(cid:12)(cid:12)(cid:12) − n + 12 (cid:29) = √ na n (cid:12)(cid:12)(cid:12)(cid:12) (cid:29) . Then, by using above equations, the final quantum statecan be also expressed at t = 5 in Figure 6A such that (cid:12)(cid:12)(cid:12)(cid:12)
12 + n + m (1 + λ ) − m ′ (1 + λ ) − n ′ (cid:29) = √ n ′ m ′ √ nm a n ′ τ m ′ τ + m a + n (cid:12)(cid:12)(cid:12)(cid:12) (cid:29) . (91)Simple calculus reveals that the negative energy state oc-curs if n ′ − n > (1 + λ )( m − m ′ ). Suppose that n ′ /n isfixed as the constant safety factor. Then the ratio be-tween m and m ′ can lead to the negative energy statefor the following cases: (i) If m ′ is larger than m , whichmeans the efflux membrane current is larger than the in-flux membrane current. (ii) If m is larger than m ′ , but itsdifference m − m ′ is smaller than ( n ′ − n ) / (1+ λ ). The case(i) is obvious since both n − n ′ and m − m ′ are all nega-tive, but the case (ii) is worthy of being noticed because,in order to prevent negative energy states such as therefractory period, the influx membrane current shouldbe sufficiently larger than the efflux membrane current.But, if the number m is relatively close to the number m ′ to maintain the propagation from the conservation of thetotal number of cations as shown in proposition 3, thenthe negative energy state is likely to occur in the cardiacexcitation propagation. E. With the external electromagnetic field
As the similar studies of the changes of the Lagrangianwith the external electromagnetic field, we also investi-gate the changes of the Hamiltonian with the externalelectromagnetic field. Let the magnetic field B e be theexternal magnetic field and let A e and φ e be the exter-nal electromagnetic potentials which are constructed inthe same way as in Section IV.C. Then, the Hamiltonianof the Maxwell’s equations (52) - (56) with the externalfield is expressed as H e = H e + H R + H C + H eI + H eI + H I + H eI , (92)where H R , H C , and H I are the same as in the com-ponents of the Hamiltonian (82), (83), and (86). OtherHamiltonians with the superscript e are defined as fol-lows: H ep = X α ( p eα ) m α , (93) H eI = − X α χ α m α p eα · A ( r α ) , (94) H eI = − X α g α χ α m α S α · ( B + B e )( r α ) , (95) H eI = X α χ α φ e ( r α , t ) , (96)where the momentum p e is the new momentum affectedby the external potential A e defined as p eα ( r α , t ) = p α − χ α A e ( r α , t ).6Observe that the external field significantly affects theparticle Hamiltonian ( H p ), due to the additional parti-cle Hamiltonian ( H p ) caused by the external field. Onthe contrary, the Hamiltonian for the transverse field H R remains independent of the external field, which meansthat the trajectory of the cations remains unchanged evenunder the influence of the external field. However, dra-matic changes occur for the interaction Hamiltonian H I .The kinetic energy of the oscillatory forced motion H I remains unchanged, but H I and H I are significantlychanged because of the changes of the momentum p α andthe additional effect of the external magnetic field B e , re-spectively. The effects of H I and H I on the membranecurrent may be negligible since the membrane currentmostly is known to be sensitive only to the membranepotential, not to the momentum or the spin energy.On the other hand, the new addition of H I to theinteraction Hamiltonian dramatically changes the mech-anism of the cardiac excitation in the following ways: (i) H I depends on χ α being related to ion channels, butthe impact of H I is applied to any myocardial cell withnon-zero χ α . Reflecting that the resting state has thequantum number 1 / χ α could be negligible in the resting state, but is not zeroalmost everywhere independent of the phase of the exci-tation. Then, H I may cause the excitation of the my-ocardial cells independent of the excitation propagation.(ii) Secondly, H I is proportional to charge density φ e ,thus even for the region where the charge density ρ issmall, the myocardial cell can be excited by a sufficientlylarge external scalar potential φ e . The presence of H I may provide explanations on the effects of the externalelectric currents to terminate fibrillations. This is be-cause the excitation of all the myocardial cells by a hugeexternal field can trigger them into the resting state atthe same time shortly after the electric shock, as a bet-ter condition for the normal propagation from a naturalinitiator such as the sinoatrial node. VIII. CONCLUSIONS AND DISCUSSIONS
The strength of the proposed QED theory for thecardiac excitation propagation lies in the fact that itprovides analytical explanations on many electrophys-iological phenomena which have been unexplained bypreviously-developed theories. This is mainly becausethe governing equations are Maxwell’s equations underconservational laws. A simple expression of the La-grangian provides many insights such as which factorsare critical for the changes of the propagation and whatthe role of ion channels is in the action potential. Also,the Hamiltonian simplifies the excitation mechanism el-igible for simpler mathematical analysis. It is also en-couraging to see the clinically-supported explanations onthe effects of the electromagnetic field generated by thecardiac excitation or the external electrodynamic fields. The validation of this theory is mostly self-sufficient,especially by using the following proofs: (i) The deriva-tion of a set of Maxwell’s equations equivalent to thediffusion-reaction system. (ii) For the Lagrangian, thetrajectory of the diffusion-reaction system is shown to bethe same as the electrodynamic wave when χ α changesnormally everywhere, which can be also deduced easilybecause they share the same propagational mechanism.But the validation for the theories by the Hamiltonianseems to be only possible by future experimental stud-ies. The validation for the effects of the electromagneticfields also seems to be supported by showing the consis-tency between clinical observations and what the theoryexplains.However, there are also some drawbacks as mentionedin the Introduction. (i) The first is related to the propa-gating cations. If the propagating particle is more thanone kind, then calculations become too complicated orbecomes impossible. Also, the concept of the cationscould be concrete, but at the same time could be ab-stract. (ii) The biggest drawback is the simplification ofion channels. The second quantization of Maxwell’s equa-tions does not show the existence of numerous ion chan-nels for several charged ions. Consequently, the use of acorresponding quantum operator may remain restrictedwithout describing in detail the complex dynamics ofthe real phenomena. One possible way is to considerthe different charged ions as different modes. But, thenegatively-charged ions are not relevant to this case andvarious values of the Planck constant ~ may only lead tomuch more complicated analysis which may be beyondour understanding. (iii) The last is the difficulty of usingthe Maxwell’s equations (52) - (56) for computationalsimulations because the expressions for charge densityand current density are too complicated. But, this canbe easily solved by using diffusion-reaction equations asusual, but using, additionally, Maxwell’s equations (52)- (56) to obtain E , B , and A from φ .In the future publication as a continuing effort of de-veloping the QED theory for the cardiac excitation prop-agation, another quantum optical concept known as co-herence will be introduced in order to understand someimportant problems such as (i) when conduction fails (ii)what is the role of geometry in conduction failure (iii)how conduction failure can be prevented in the perspec-tive of optical coherence, etc. IX. APPENDIX I: PROOFS IN SECTION IVA. Proof of Proposition 4
The total energy (62) is well defined in the macroscopicdomain Π, since for the particle in each space, the aboveequation will reduce to the classical energy for π i or π o ;if the particle lies in π i , then equation (62) with equation7(22) reduces to U = X α m α ( v oα ) + ε i ˆ (cid:2) ( E i ) + c ( B i ) (cid:3) d r, and if the particle lies in π o , then it reduces to U = X α m α ( v oα ) + ε i ˆ h ( p ε o /ε i E o ) + c ( p µ i /µ o B o ) i d r = X α m α ( v oα ) + ε o ˆ (cid:2) ( E o ) + c ( B o ) (cid:3) d r. Thus, the total energy (62) is well defined in Π. Thedifferentiation of the above equation with respect to thetime t yields ∂ U ∂t = X α m α v α · d v α dt + ε i ˆ (cid:20) E · ∂ E dt + c B · ∂ B dt (cid:21) d r. By substituting the Maxwell’s equations (54) and (56)and the Newton-Lorentz equation (61), we obtain ∂ U ∂t = X α v α · ( χ α E ( r α , t )) − ˆ E · J d r + ε i c ˆ [ E · ( ∇ × B ) − B · ( ∇ × E )] d r. Due to the discrete expression of J as shown in equation(21), we notice that the first two terms cancel out. Theintegrand in the last integration can be simplified as ∇ · ( E × B ). Thus, the above equation reduces to ∂ U ∂t = ε i c ˆ ∇ · ( E × B ) d r = ε i c ˆ S ( E × B ) · n dS, where the last equation is obtained by the divergence the-orem. Since no flux of the electromagnetic fields occursat the boundaries in the closed system, the right handside is zero, thus ∂ U /∂t = 0 (cid:3) . B. Proof of Proposition 5
The momentum (63) is also well defined in Π, since forthe particle in each microscopic domain, equation (63)will reduce to the classical momentum for π i or π o ; if theparticle lies in π i , then equation (63) with equation (22)reduces to P = X α m α v iα + ε i ˆ (cid:2) E i × B i (cid:3) d r, and if the particle lies in π o , then it reduces to P = X α m oα v α + ε i ˆ h ( − p ε o /ε i E o ) × ( − p µ i /µ o B o ) i d r = X α m α v oα + ε o ˆ [ E o × B o ] d r. Thus, the total momentum (63) is well defined in Π. Af-ter differentiating the above equation with respect to t ,let us substitute again Maxwell’s equations (54) and (56)and the Newton-Lorentz equation (61) in equation (63)to obtain ∂ P ∂t = X α χ α E ( r α , t ) + χ α v α × B ( r α , t ) − ˆ J × B d r + ε i ˆ (cid:2) c ( ∇ × B ) × B − E × ( ∇ × E ) (cid:3) d r. (97)Substituting equation (21) into the above equation willcancel out the second and the third terms. For the inte-gration term, we use V × ( ∇ × V ) = 12 ∇ ( V ) − X j e j ∇ · ( V j V ) + V ( ∇ · V ) , where e j is the directional vector of the Cartesian co-ordinates, then for the first part of the integration weobtain ˆ ( ∇ × B ) × B d r = 0 , (98)where the first and the second terms on the right handside are zero because there is no flux of B across theboundaries in the closed system and the third term iszero because of equation (53). Similarly, for the secondpart of the integration, we obtain ˆ E × ( ∇ × E ) d r = − ˆ S E · n dS − ˆ S B j E · n dS − ˆ E ( ∇ · E ) d r = X α χ α E ( r α , t ) , (99)where the first and second terms on the right hand sideare zero because there is no flux of E across the bound-aries in the closed system and the third term is obtainedby the use of equation (52). Finally, substituting equal-ities (98) and (99) into equation (97) yields ∂ P /∂t = 0 (cid:3) . X. APPENDIX II: PROOF IN SECTION VA. Proof of Lemma 2
Substituting equation (33) and (34) into equation (66)yields L ( k ) = ε i h | ˙ a k ( k ) + i k φ k ( k ) | − c | k × a k ( k ) | i + [ j ∗ k ( k ) · a k ( k ) + j k ( k ) · a ∗ k ( k ) − ρ ∗ k ( k ) φ k ( k ) − ρ k ( k ) φ ∗ k ( k )] . To eliminate the scalar potential φ k in this equation, wesubstitute the equality φ k = (1 /k )( ik ˙ a k k + ( ρ k /ε i )) being8obtained from equations (29) and (34). Then, we obtain L ( k ) = − ρ k ρ ∗ k ε i k + ε i (cid:2) ˙ a ⊥∗ k · ˙ a ⊥ k − c k a ⊥∗ k · a ⊥ k (cid:3) + ik ddt h ρa k∗ k − ρ ∗ a k k i , (100)where we used ˙ a ⊥ k = ˙ a k − ( k /k ) ˙ a k k and used the new vari-ables a k k = κ · a k and j k k = κ · j k . With the conservation ofcharge (25), or equivalently ˙ ρ = − ikj k in the k -space, thelast total time derivative is obtained from the followingequality: j k∗ k a k k + j k k a k∗ k + ik (cid:16) ρ k ˙ a k∗ k − ρ ∗ k ˙ a k k (cid:17) = ik ddt h ρa k∗ k − ρ ∗ a k k i . Then the only difference of Lagrangian with L ( k ) (100) toLagrangian with L ( k ) (68) is the above total derivativeterm with respect to time, thus we only need to showthat this term can be subtracted without changing theextremes of the action integral from the Lagrangian ofour system. Let S be the action integral for the classicalMaxwell’s equations to the Lagrangian L . Then theaction integral S for the Lagrangian density L (100) isexpressed as S = ˆ t t L = S + 14 π ˆ t t (cid:20) ddt ˆ ρ ( r , t ) φ ( r ′ , t ) | r − r ′ | d r (cid:21) dt, thus, S = S + 14 π | r − r ′ | [ ρ ( r , t ) φ ( r ′ , t ) − ρ ( r , t ) φ ( r ′ , t )] . Observe that S and S are only different in terms of aconstant, thus have the same extreme values. Therefore,the total derivative term with respect to t is redundantand can be subtracted from the Lagrangian density inequation (100) without changing its extremes. (cid:3) XI. APPENDIX III: PROOF IN SECTION VIA. Proof of Lemma 4
Consider the conjugate momentum p α for the particle α and the conjugate momenta ξ for ˙ A ⊥ defined as p α ( r α ) ≡ ∂ L ∂ ˙ r α , ξ ( k ) ≡ ∂ L ∂ ˙ a ⊥ k ! ∗ , (101)where the differentiation with respect to a vector V isjust considered as a vector whose component is the dif-ferentiation with respect to each component V i . Then,the Hamiltonian for the Lagrangian (71) and (72) can bederived as H = X α p α · ˙ r α + h ξ ∗ · ˙ a ⊥ k + ξ · (cid:0) ˙ a ⊥ k (cid:1) ∗ i d k − L . (102)Substituting the Lagrangian (71) and (72) with the con-jugate momenta (101) into the Hamiltonian (102), weobtain the Hamiltonian (78) with (79) (cid:3) . B. Proof of Proposition 8
Using the normal variable α = − i/ (2 N ( k )) (cid:2) e ⊥ k − c ( k /k ) × b k (cid:3) which satisfies˙ α ( k ) + iω α ( k ) = i ε i N j ⊥ k , (103)where ω = ck and N is the normalization coefficient, nor-mally chosen as N ( k ) = p ~ ω/ ε i , we can express e ⊥ k and b k with α and its conjugate α ∗ , thus the Hamiltoniandensity is expressed as H ( k ) = N [ α ∗ ( k ) · α ( k ) + α ( k ) · α ∗ ( k )] . (104)In fact, this is the same as the classical Maxwell’s equa-tions with the Coulomb gauge because the gauge choiceand the reactive membrane current density (39) does notaffect the Maxwell’s equations (31) and (32) and con-sequently the oscillator of the normal variable α (103)remains the same. For j ⊥ k = 0, the equation (103)boils down to a Schr¨odinger equation for the wave func-tion α such as i ~ ˙ α ( k , t ) = ~ ω α ( k , t ) as a Schr¨odinger’sform for the equation of motion . This similarity of-ten leads to the substitution of the normal variable α and its conjugate α ∗ with the annihilation operator a i and a + i , respectively . As a consequence, by using[ a, a + ] = aa + − a + a = 1 and by introducing the spinmagnetic moment, we have the following Hamiltonian(78) and (79) (cid:3) . R. R. Aliev and A. V. Panfilov. A simple two-variable model ofcardiac excitation.
Chaos, Solitons and Fractals , 7(3):293–301,1996. Philip Ball. Physics of life: The dawn of quantum biology.
Na-ture , 474:272–274, 2011. G. M. Baule and R. Mcfee. Detection of the magnetic field of theheart.
Amer. Heart J. , 66:95–96, 1963. C. S. Beck, W. H Pritchard, and H. S. Feil. Ventricular fibrillationof long duration abolished by electric shock.
J. Amer. Medi.Asso. , 135(15):985–986, 1947. M. Born and E. Wolf.
Principle of Optics . Cambridge UniversityPress, seventh edition, 1999. D. Burstein and D. Cohen. Comparison of magnetic field andelectric potential produced by frog heart muscle.
J. Appl. Phys. ,57(7):2640–2646, 1985. S. Chun. Geometric analysis on the role of the pv as unidirec-tional block to cause atrial reentry.
Submitted , 2013. L. B. Cohen, B. Hille, and R. D. Keynes. Light scattering andbirefringence changes during activity in the electric organ of elec-trophorus electricus.
J. Physiol. , 203:489–509, 1969. C. Cohen-Tannoudji, B. Diu, and F. Lalo¨e.
Quantum mechanics .Wiley-Interscience, 1992. C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg.
Pho-tons and atoms: Introduction to quantum electrodynamics . JohnWiley and Sons, 1989. C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg.
Atom-photon interactions: basic processes and applications . John Wi-ley and Sons. Inc, 1992. H. Curtis and N. Sue Barness.
Biology . W. H. Freeman, 1989. V. A. Davydov, V. S. Zykov, and A. S. Mikhailov. Kinematics ofautowave structures in excitable media.
Sov. Phys. Usp. , 34:665–684, 1991. P. A. M. Dirac. The quantum theory of the emission and absorp-tion of radiation.
Proc. Roy. Soc. Lond. Ser. A. , 115(767):243–265, 1927. P. A. M. Dirac. Quantised singularities in the electromagneticfield.
Proc. Roy. Soc. Lond. Ser. A. , 133:60–72, 1931. P. A. M. Dirac.
The principles of quantum mechanics . Oxford,fourth edition, 1958. A. Einstein. On the present status of the radiation problem.
Phys. Zeits , 10:185–193, 1909. L. P. Ferris, B. G. King, P. W. Spence, and H. B. Williams. Effectof electric shock on the heart.
Transactions of the AmericanInstitute of Electrical Engineers , 55(5):498–515, 1936. R. P. Feynman. The theory of positrons.
Phys. Rev. , 76(6):749–759, 1949. R. P. Feynman.
Quantum Electrodynamics: A lecture note andreprint volume . W. A. Benjamin, Inc., 1961. R. P. Feynman.
QED: the strange theory of light and matter .Penguin books, 1985. R. P. Feynman.
Statistical mechanics: A set of lectures . West-view Press, second edition, 1998. R. P. Feynman.
Quantum mechanics and path integrals . Doverpublications, Inc, emended edition, 2005. R. FitzHugh. Impulses and physiological states in theoreticalmodels of nerve membrane.
Biophys. J. , 1(6):445–466, 1961. R. FitzHugh.
Mathematical models of excitation and propagationin nerve. Chapter 1 . McGraw-Hill Book Co., 1969. John E. Hall.
Guyton and Hall Textbook of Medical Physiology .Saunders, 12th revised edition edition, 2010. W. Heitler.
The quantum theory of radiation . Dover publications,third edition, 2010. A. L. Hodgkin and A. F. Huxley. A quantitative description ofmembrane current and its application to conduction and excita-tion in nerve.
J. Physiol. (London) , 117:500–544, 1952. C. Huygens.
English translation of Trait´e de la Lumi`ere . Leyden,1690. J. D. Jackson.
Classical Electrodynamics . John Wiley and Sons,third edition, 1998. J. D. Jackson. From lorenz to coulomb and other explicit gaugetransformations.
Am. J. Phys. , 70:917, 2002. A. M. Katz.
Physiology of the heart . Lippincott Williams andWilkins, fifth edition, 2010. J. Keener and J. Sneyd.
Mathematical Physiology . Springer,1998. J.P. Keener. An eikonal-curvature equation for action potentialpropagation in myocardium.
J. Math. Biol. , 29(7):629–651, 1991. A. G. Kl´eber and Y. Rudy. Basic mechanisms of cardiac impulsepropagation and associated arrhythmias.
Physiol. Rev. , 84:431–488, 2004. L. D. Landau and E. M. Lifshitz.
The classical theory of fields .Butterworth-Heinemann, fourth edition, 1980. A. Li´enard. ´Erude des oscillations entretenues.
Rev. g´en. ´elec ,23:901–912, 946–954, 1928. H. A. Lorentz.
The theory of electrons and its applications to thephenomena of light and radiant heat . Leipzig: B. G. Teubner,second edition, 1916. R. Loudon.
The quantum theory of light . Oxford, third edition,2000. J. Nagumo, S. Arimoto, and S. Yoshizawa. An active pulse trans-mission line simulating nerve axon.
Proc. IRE , 50:2061–2070, 1962. M. Planck.
Waermestralung . P. Blakiston Son & Co., 1914. R. Plonsey. The nature of sources of bioelectric and biomagneticfields.
Biophys. J. , 39:309–312, 1982. E. A. Power and S. Zienau. Coulomb gauge in non-relativisticquantum electro-dynamics and the shape of spectral lines.
Philos.Trans. Roy. Soc. A , 251(999):427–454, 1959. E. Purves, G. J. Augustine, D. Fitzpatrick, W. C. Hall, A.-S.LaMantia, and L. E. White.
Neuroscience . Sinauer Associates,Inc., fifth edition, 2011. P. M. Ragan, W. Weiping, and S. R. Eisenberg. Magneticallyinduced currents in the canine heart: A finite element study.
IEEE Trans. Bio. Eng. , 42(11):1110–1116, 1995. J. M. Rogers and A. D. McCulloch. A collocation-galerkin finiteelement model of cardiac action potential propagation.
IEEETrans. Biomed. Eng. , 41(8):743– 757, 1994. W. N. Ross, B. M. Salzberg, L. B. Cohen, and H. V. Davila.A large change in dye absorption during the action potential.
Biophys. J. , 14:983–986, 1974. B. J. Roth and J. P. Wikswo. The magnetic field of a single axon.
Biophys. J. , 48:93–109, 1985. B. J. Roth and J. P. Wikswo. Electrically silent magnetic fields.
Biophys. J. , 50:739–745, 1986. C. H. Ruo and Y. Rudy. A model of the ventricular cardiac actionpotential. depolarization, repolarization, and their interaction.
Circ. Res. , 68:1501–1526, 1991. H. P. Schwan and C. F. Kay. The conductivity of living tissues.
N.Y. Acad. Sci. , 65:1007, 1957. M. O. Scully.
Quantum Optics . Cambridge University Press,1997. M. O. Scully and M. Sargent III. The concept of the photon.
Phys. Today , 25(3):38–47, 1972. L. Tung.
A bidomain model for describing ischemic myocardialD.C. potential . Ph.D. thesis, MIT, 1978. A. M. Turing. The chemical basis of morphogenesis.
Phil. Trans.Roy. Soc. London B , 237(641):37–72, 1952. B. van der Pol. On relaxation oscillations.
Phil. Mag. , 2:978,1926. E. T. Whittaker.
A history of the theories of aether and elec-tricity: from the age of Descartes to the close of the Nineteenthcentury . Forgotten books, 1910. J. P. Wikswo, J. P. Barach, and J. A. Freeman. Magnetic field ofa nerve implies: first measurements.
Science , 208(4439):53–55,1980. R. G. Woolley. Molecular quantum electrodynamics.
Proc. Roy.Soc. Lond. A. , 321(1547):557–572, 1971. P. M. Zoll, A. J. Linenthal, W. Gibson, M. H. Paul, and L. R.Norman. Termination of ventricular fibrillation in man by exter-nally applied electric countershock.
N. Engl. J. Med. , 254:727–732, 1956. P. M. Zoll, M. H. Paul, A. J. Linenthal, L. R. Norman, andW. Gibson. The effects of external electric currents on the heart:control of cardiac rhythm and induction and termination of car-diac arrhythmias.
Circ. , 14:745–756, 1956. V. S. Zykov. Kinematics of rigidly rotating spiral waves.