Content uploaded by Bing He

Author content

All content in this area was uploaded by Bing He on May 20, 2018

Content may be subject to copyright.

Superluminal propagation of evanescent modes as a quantum

effect

Zhi-Yong Wang1*, Cai-Dong Xiong1, Bing He2

1School of Optoelectronic Information, University of Electronic Science and Technology of China,

Chengdu 610054, CHINA

2Department of Physics and Astronomy, Hunter College of the City University of New York, 695 Park

Avenue, New York, NY 10021, USA

*E-mail: zywang@uestc.edu.cn

Abstract

Contrary to mechanical waves, the two-slit interference experiment of single photons

shows that the behavior of classical electromagnetic waves corresponds to the quantum

mechanical one of single photons, which is also different from the quantum-field-theory

behavior such as the creations and annihilations of photons, the vacuum fluctuations, etc.

Owing to a purely quantum effect, quantum tunneling particles including tunneling photons

(evanescent modes) can propagate over a spacelike interval without destroying causality.

With this picture we conclude that the superluminality of evanescent modes is a quantum

mechanical rather than a classical phenomenon.

Keywords: Evanescent modes, superluminal propagation, quantum effect, virtual photons

PACS: 03.65.Xp, 41.20.Jb, 42.50.Nn, 03.65.Pm

1. Introduction

Nowadays, both theoretical and experimental investigations have presented a

conclusion that the evanescent modes of the electromagnetic field can superluminally

propagate [1-10]. At the level of quantum mechanics, via tunneling analogy the

superluminal propagation of evanescent modes has been described as the quantum tunneling

behavior of photons, which implies that the superluminality of evanescent modes is due to a

quantum effect. In this paper, at the level of quantum field theory, we will further show that

the superluminality of evanescent modes is due to a purely quantum effect, and clarify some

misunderstandings on the physical properties of evanescent modes. As an application, we

conclude that a recent objection [11-13] to the superluminality of evanescent modes is

invalid, because the objection is completely based on classical mechanics by regarding

electromagnetic waves as mechanical waves.

To avoid misunderstanding our argumentation (e.g., in spite of the genuine text of Refs.

[9, 10], in Ref. [13] the author made improper claim that “they mistake a non-zero

1

propagator for a non-zero commutator”), one should not confuse the following two issues:

(1) whether a particle can propagate over a spacelike interval? (2) whether such propagation

destroys causality (i.e., whether it means a measurement performed at one point can affect

another measurement at a point separated from the first with a spacelike interval), if a

particle does propagate over a spacelike interval? According to quantum field theory, a

non-zero propagator or non-zero transition probability amplitude for a spacelike interval

implies that a particle can propagate over the spacelike interval [14], but this spacelike

propagation does not destroy causality provided that the commutator of two observables

with a spacelike interval vanishes, that is, a measurement performed at one point does not

affect another measurement at a point separated from the first with a spacelike interval.

On the other hand, the commutator between two field operators located at spacelike

distance does not always vanish if the field operators are not observables. For example, in

the Coulomb gauge, the commutator between electromagnetic potentials does not vanish for

spacelike distances [15]. Moreover, in the quantization theory of evanescent modes [16], by

assuming the high-frequency behavior of the refractive index, one can find that the

commutator of evanescent field operators between two space-like separated points does not

vanish, whose physical meaning and the related causality problem have been discussed by

Stahlhofen and Nimtz [17].

2. Quantum field theory naturally explains superluminality of particles

A main reason to object all the existing theoretical and experimental investigations on

the superluminality of evanescent modes lies in the fact that such superluminal propagation

is in conflict with special relativity. However, special relativity has been developed on the

basis of classical mechanics without taking into account any quantum-mechanical effect. On

the other hand, because quantum field theory combines quantum mechanics with special

relativity, such that it can give us such a conclusion [14, 18, 19]: owing to

quantum-mechanical effect, a particle can propagate over a spacelike interval (without

destroying Einstein’s causality), which corresponds to the quantum tunneling phenomenon

[18].

For example, just as S. Weinberg discussed [19] (with some different notations and

conventions): “Although the relativity of temporal order raises no problems for classical

physics, it plays a profound role in quantum theories. The uncertainty principle tells us that

when we specify that a particle is at position at time , we cannot also define its

velocity precisely. In consequence there is a certain chance of a particle getting from

1

x1

t

2

11

(, )tx to even if the spacetime interval is spacelike, that is,

22

(, )tx12 12

ct t−> −xx .

To be more precise, the probability of a particle reaching if it starts at is

nonnegligible as long as (we call Eq. (1) Weinberg’s formula)

22

(, )tx11

(, )tx

22 2

12 12

0( ) ( ) ( )ct t mc<− − − ≤xx =2

, (1)

where is Planck’s constant (divided by ), c is the velocity of light in vacuum, and

is the particle’s mass, then

=2π

mmc= is the particle’s Compton wavelength. We are thus

faced with our paradox: if one observer sees a particle emitted at , and absorbed at

, and if is positive (but less than or equal to

11

(, )tx

22

(, )tx22

12 12

()(ct t−−−xx 2

)2

(mc=)), then

a second observer may see the particle absorbed at at a time before the time it

is emitted at . There is only one known way out of this paradox. The second observer

must see a particle emitted at and absorbed at . But in general the particle seen by

the second observer will then necessarily be different from that seen by the first observer (it

is the antiparticle of the particle seen by the first observer)”. In other words, to avoid a

possible causality paradox, one can resort to the particle-antiparticle symmetry. The process

of a particle created at and annihilated at as observed in a frame of

reference, is identical with that of an antiparticle created at and annihilated at

as observed in another frame of reference.

2

x2

t1

t

1

x

2

x1

x

11

(, )tx22

(, )tx

22

(, )tx

11

(, )tx

In fact, Weinberg’s argument is equivalent to the usual argument in

quantum-field-theory textbooks: let ()

x

φ

stand for a scalar field operator, 0 denote the

vacuum state, then 0()()0xy

φφ

represents the transition probability amplitude from the

state ()0y

φ

to the state ()0x

φ

[14], such that 2

0()()0xy

φφ

corresponds to the

probability for a scalar particle to propagate over the spacetime interval 2

()

x

y−. In

particular, if the probability amplitude for a scalar particle propagating over a spacelike

interval is denoted as

2

()xy−<0()0()()Dx y x y

φφ

−= 0

, then according to quantum

3

field theory, ()0()()0Dy x y x

φφ

−= represents the probability amplitude for the

corresponding antiparticle propagating backwards over the spacelike interval. Because

for , the two spacelike processes are undistinguishable

and the commutator

()(Dx y Dy x−= −)0

2

()xy−<

[(), ()] 0[(),()]0 0xy xy

φφ φφ

=

=, such that the causality is

maintained. Therefore, Weinberg has provided another way of looking at the statement that

“a measurement performed at one point does not affect another measurement at a point

separated from the first with a spacelike interval”. Studying a quantum Lorentz

transformation one can also obtain Eq. (1) [20].

In fact, Eq. (1) is just an approximation of a more rigorous result. For our purpose, let

us derive the rigorous result within the framework of quantum field theory. The transition

probability amplitude for any particle to propagate over the spacetime interval 2

()

x

y− can

be expressed in terms of ()0()()0Dx y x y

φφ

−= , and then we can ignore the spin

degree of freedom and take the scalar field ()

x

φ

for example. For convenience let

, , and denote

(0,0,0,0)y=( , ,0,0)xtr=(, ) ( )Dtr Dx y

=

−, according to quantum field

theory, one has (up to a constant factor)

d

( , ) exp[ i ( ) ]

2π2p

p

pc

Dtr Et pr

E

+∞

−∞

=−−

∫=, (2)

where 22 24

p

Epcm=+c

. Let (2)

0()

H

z denote the zero-order Hankel function of the

second kind, as the spacetime interval is spacelike (i.e., 22 2 0ct r

−

<), one can prove that,

(2) 2 2 2

0

(, ) ( i4) ( i )Dtr H r ct=− − − , (3)

where mc== is the Compton wavelength. Therefore, the asymptotic behaviors of

are governed by the Hankel function of imaginary argument: falls off like

(, )Dtr (, )Dtr

1 exp( )

z

z− for 222

zrct=− →+∞, while falls off faster than 1 exp( )

z

z

−

for

the other 222

zrct=− . In the observable sense, is always ignored for

(, )Dtr

222 1zrct=− >, that is, one always takes the approximate as follows:

4

22 2 2 2

22 2 2 2

0, for ( )

(, ) 0, for 0 ( )

ct r mc

Dtr ct r mc

⎧=−<−=−

⎪

⎨

≠

>−≥− =−

⎪

⎩

=

=

. (4)

According to the approximate given by Eq. (4), for the spacelike interval , the

probability amplitude for the particle to propagate from to

is nonnegligible as long as the Weinberg’s formula given by Eq. (1) is

satisfied. In other words, the Weinberg’s formula given by Eq. (1) is just an approximate of

the rigorous result given by Eq. (3). The rigorous result (3) implies that there are in

principle no limitations to the spacelike interval of

22 2 0ct r−<

(, )Dtr (0,0,0,0)y=

( , ,0,0)xtr=

222

rct−, but the probability 2

(, )Dtr

falls off rapidly for large 222

rct−, and the spacelike process cannot be observed

provided that the probability is too small.

Taking an electron for example, the Compton wavelength of the electron is

10

3.87 10mc −

≈×= (millimeter, mm). For any spacelike interval there is always an inertial

reference in which one has , using Eq. (1) one has

1

tt=2

22 2 10

12 12 12

( ) ( ) 3.87 10 mmct t mc −

−−−=−≤ ≈×xx xx =. (5)

The spacelike process occurring within such a spatial region is difficult to be observed. In

fact, within the Compton wavelength of the electron, the many-particle effects arising from

the creations and annihilations of virtual electron-positron pairs cannot be ignored. On the

other hand, for large spacelike interval the corresponding probability becomes so small that

the spacelike process cannot be observed. In a word, Eq. (1) as the approximate of the

rigorous result (3), describes a spacelike process with a sufficiently large probability (in the

observable sense).

3. Superluminal propagation of evanescent modes as a quantum effect

Photons inside a hollow waveguide can be treated as free massive particles with an

effective mass 2

eff c

m

ω

==c

[9, 21], where c

ω

is the cut-off frequency of the waveguide,

then the argument in Section 2 is also valid for the guided photons. In fact, quantum field

theory tells us that [9, 10]: owing to a quantum effect, photons inside a waveguide can

5

propagate over a spacelike interval, which corresponds to the fact that the evanescent modes

can propagate superluminally through an undersized waveguide (i.e., the photonic tunneling

phenomenon). Likewise, here the Einstein causality is preserved via the particle-antiparticle

symmetry, but for the moment the antiparticle of a photon is the photon itself, such that the

process that a photon propagates superluminally from A to B as observed in an inertial

frame of reference, is equivalent to that the photon propagates superluminally from B to A

as observed in another inertial frame of reference. Moreover, via quantum Lorentz

transformation [20] or by developing special relativity on the basis of quantum mechanics

[21], another theoretical evidence for the superluminality of evanescent modes can be

obtained, which also shows that the superluminal behavior of evanescent modes arises from

a quantum effect, i.e., the Heisenberg's uncertainty. However, here the theoretical evidence

is obtained at the level of quantum mechanics, it is just an approximate of the more rigorous

result given by Eq. (3), i.e., the related spacelike interval in a spacelike process is just the

one with a sufficiently large propagation probability (i.e., with

2(2) 2

0

( , ) (1 16)[ ( i)]Dtr H≥−

).

As an example, let the cut-off frequency of a waveguide be c9.49GHz

ω

=, the

effective Compton wavelength of photons inside the waveguide is

eff c 31.6mmmc c

ω

=≈=, which is far too larger than (i.e., the Compton

wavelength of the electron). That is, for tunneling photons, the spacelike interval with a

sufficiently large propagation probability is

10

3.87 10 mm

−

×

22 2

12 12 eff c

( ) ( ) 31.6mmct t mcc

ω

−−−≤ =≈xx =. (6)

Therefore, contrary to the superluminal behavior of tunneling electrons, the superluminal

behavior of evanescent modes can be easily observed experimentally. It is important to

mention that, Eq. (6) does not conflict with those experimental results with the largest

tunneling distance larger than 31.6 mm, because: 1) for a given spacelike interval

22

rct−2

, the propagation distance 12

r=−xx

is related to the propagation time

; 2) Eq. (6) as an approximate of the more rigorous result given by Eq. (3), just

corresponds to the result with a sufficiently large propagation probability (i.e., satisfying

12

(ttt=−)

6

2(2) 2

0

( , ) (1 16)[ ( i)]Dtr H≥−

), while Eq. (3) tells us that there are in principle no limitations

to the spacelike interval, though the probability 2

(, )Dtr falls off rapidly for large

spacelike interval.

Moreover, contrary to electrons, photons are bosons and do not carry any charge, by

increasing the number of tunneling photons the observable spacelike interval can be

augmented ad lib. Eq. (3) shows that the probability amplitude falls off

exponentially (but does not vanish) as the spacelike interval

(, )Dtr

222

rct

−

→+∞; on the other

hand, for the spacelike interval there is always an inertial reference in which one has 0t

=

.

Therefore, even if the propagation distance , the propagation time can be

arbitrarily small, which is in agreement with the Hartman effect [22].

r→+∞

In frustrated total internal reflection, evanescent modes as near field consist of virtual

photons [23-24], these virtual photons correspond to the elementary excitations of

electromagnetic interactions. Now we show that evanescent modes inside an undersized

waveguide are also identical with virtual photons. As we know, the near fields of a dipole

antenna fall off with the distance from the antenna like r1n

r ( ). However, if we

assume that an aerial array formed by an infinite set of infinite-length line sources arranging

in a periodic manner (with the period ), then the near fields of the aerial array falls off

like

2n≥

0

r

0

exp( )rr−. With respect to the TE10 mode, an undersized waveguide is equivalent to

such an aerial array [25] and evanescent modes inside the undersized waveguide are

equivalent to the near fields of the aerial array, which implies that evanescent fields inside

the waveguide can also be described by virtual photons. As we know, the propagation of

virtual photons is due to a purely quantum-mechanical effect, which also implies that one

cannot understand the propagation of evanescent waves via classical mechanics. To show

the evanescent TE10 mode (with the frequency c

ω

ω

<

, where c

ω

is the cut-off frequency)

is equivalent to the near field of the aerial array, basing on Ref. [25], one ought to make the

following replacements:

22

cc

0

ωω

=−→22

c

ω

ω

−, exp(i 0 ) 1t

⋅

⋅=→exp(i )t

ω

, (7)

22

22

()(,)0

→

y

Exz

xz

∂∂

+=

∂∂

22 2

2222

()(,

y

Exzt

xzct

∂∂ ∂ ,)0

+

−=

∂∂ ∂ , (8)

7

c

0

π

(,) sin( )exp( )

yz

x

Exz E ac

ω

=−

→0

π

( , , ) sin( )exp(i )

yx

E

xzt E t z

a

ω

κ

=−

, (9)

where 0c

πza c

ω

== , and 22

cc

κωω

=− . The presence of the decay factor exp( )z

κ

−

implies that the field (,,)

y

E

xzt mainly exists in the neighborhood of the aerial array, and

then is the near field of the aerial array.

4. Some misunderstandings about the physical properties of evanescent modes

To show that the superluminal propagation of evanescent modes is due to a purely

quantum effect, in addition to the argument presented above, some misunderstandings on

the physics properties of evanescent modes should be clarified. In particular, these

misunderstandings appear in the objection [11-13] to the superluminality of evanescent

modes.

Firstly, classical electromagnetic waves are conceptually different from mechanical

waves such as water waves, acoustic waves, waves on a string, etc. (e.g., only via media can

mechanical waves propagate, while electromagnetic waves can propagate in vacuum). For

example, the wave nature of mechanical waves is usually described by classical mechanics,

while the wave-particle dualism of light tells us that the wave nature of classical

electromagnetic waves is essentially a quantum mechanical issue, and historically quantum

mechanics arises from extending the wave-particle dualism of light to that of massive

particles. In terms of the spinor representation of electromagnetic field [26, 27], one can

obtain the quantum-mechanical theory of single photons (note that the usual quantum theory

of electromagnetic field is referred to the field-quantized theory rather than quantum

mechanics). In fact, to interpret the two-slit interference experiment of single photons

one-by-one emitted from a light source, one has to regard the behaviors of classical

electromagnetic waves as the quantum-mechanical ones of single photons, and here have

nothing to do with the quantum-field-theory effects such as the creations and annihilations

of photons, or the zero-point fluctuations of quantum electromagnetic field.

From the point of view of classical mechanics, inside an undersized waveguide

evanescent waves have support everywhere (through exponential damping) along the

undersized waveguide, and "the propagation of evanescent modes" is not a well-defined

concept. However, this classical picture just provides us with a phenomenological

description. From the point of view of quantum mechanics, now that some fraction of an

electromagnetic wave beam entering in the input side of an undersized waveguide with

8

finite length will come out of the exit of the waveguide, it indicates that there must have a

physical process that some photons in the evanescent wave beam propagate through the

undersized waveguide. As a quantum tunneling phenomenon, this physical process is due to

a purely quantum-mechanical effect without any classical correspondence, such that only

via quantum theory can one explain the propagation of evanescent modes. In other words, if

evanescent waves could not propagate, likewise any other quantum tunneling phenomenon

could not occur, because the wavefunctions of tunneling particles within a potential barrier

are similar to evanescent electromagnetic waves: they possess imaginary wave-numbers;

they do not describe propagating waves but evanescent waves.

On the other hand, plug in a signal into a tunnel and as long as it can be read out at the

other end, Einstein causality is violated [17].

In the objection to the superluminality of evanescent modes, evanescent waves are by

mistake regarded as “exponentially attenuated standing waves”. To clarify such a

misunderstanding, let us assume that a hollow waveguide is placed along the direction of

z-axes, and the waveguide is a straight rectangular pipe with the transversal dimensions

and (a, the cross-section of the waveguide lies in

a

bb>0

x

a

≤

≤ and ). Inside

the waveguide, for electromagnetic waves with the frequency

0yb≤≤

ω

and wave-number vector

(, , )

x

yz

kkk=k, take the electric field component

x

E for example, it can be written as

( , , , ) cos( )sin( )exp(i i )

xxyz

E

xyzt A kx ky t kz

ω

=

−, (10)

where A is a constant factor, π

x

kna

=

and π

y

klb

=

( ) are the

wavenumbers along the x-axis and y-axis directions, respectively. In Eq. (10),

, 0,1, 2,...nl=

(, ) cos( )sin( )

xy

f

xy A kx ky≡ represents a standing-wave factor (in which the wavenumbers

x

k and

y

k are real), which implies that the electromagnetic waves inside the waveguide

form a standing-wave structure along the

x

y plane (i.e., along the cross-section of the

waveguide) for both propagation and evanescent modes. On the other hand, in Eq. (10), for

a real ,

z

kexp(i i )

z

tkz

ω

− is a propagation factor and then represents

propagation modes; while for an imaginary

( , , , )

x

Exyzt

i

z

k

κ

=

− (

κ

is a real number),

exp(i i )

z

tkz

ω

− becomes the attenuation factor of exp(i )tz

ω

κ

−

and then

represents evanescent modes. Therefore, along the z-axis direction (i.e., along direction of

the waveguide), the electromagnetic waves form propagating- and evanescent-wave

structures for the propagation and evanescent modes, respectively. In other words, along the

( , , , )

x

Exyzt

9

direction of the waveguide, there is no standing-wave structure, such that the evanescent

modes cannot be regarded as “exponentially attenuated standing waves”. If one insists on

regarding the evanescent modes as “exponentially attenuated standing waves” for the reason

that they contain the standing-wave factor (, ) cos( )sin( )

xy

f

xy A kx ky

≡

, he would have to

call the propagation modes “propagating standing waves”, which is a self-contradictory

appellation.

A reason for denying that evanescent modes have quantum mechanical behaviors is

that “evanescent modes can completely be described by Maxwell’s equations”. However, it

is well known that Maxwell’s equations not only describe the classical electromagnetic field,

but also the quantum one (at the level of quantum field theory). On the other hand, just as

mentioned above, the two-slit interference experiment of single photons shows that the

behavior of classical electromagnetic waves corresponds to the quantum mechanical one of

single photons (in the first-quantized sense).

5. Conclusions

The two-slit interference experiment of single photons tells us that, contrary to

mechanical waves described by classical mechanics, the wave nature of electromagnetic

waves is essentially a quantum mechanical issue. Just as all other quantum tunneling

phenomena, the propagation of evanescent modes attributes to the quantum-mechanical

behavior of photons and cannot be understood via classical mechanics. The superluminal

propagation of evanescent modes can be interpreted by the quantum-mechanical behavior of

single photons (in terms of photonic quantum tunneling), or by the quantum-field-theory

behavior of the electromagnetic field (in terms of a non-zero transition probability

amplitude for a spacelike interval, or in terms of spacelike virtual photons). In a word, the

superluminal propagation of evanescent modes is a purely quantum mechanical

phenomenon without any classical correspondence. As a result, any objection to the

superluminality of evanescent modes is invalid provided that the objection is completely

based on classical mechanics.

Acknowledgments

The first author (Z. Y. Wang) would like to thank Professor G. Nimtz for his many

helpful discussions. This work was supported by the National Natural Science Foundation

of China (Grant No. 60671030) and Project supported by the Scientific Research Starting

Foundation for Outstanding Graduate, UESTC , China (Grant No. Y02002010501022).

10

References

[1] A. M. Steinberg, P. G. Kwiat, and R. Y. Chiao, Phys. Rev. Lett.71 (1993) 708.

[2] Ch. Spielmann, R. Szipöcs, A. Stingl, and F. Krausz, Phys. Rev. Lett. 73 (1994) 2308.

[3] J. J. Carey, J. Zawadzka, D. A. Jaroszynski, and K. Wynne, Phys. Rev. Lett. 84 (2000) 1431.

[4] A. Haibel, G. Nimtz, and A. A. Stahlhofen, Phys. Rev. E 63 (2001) 047601.

[5] G. Nimtz and W. Heitmann, Prog Quant Electr. 21 (1997) 81.

[6] G. Nimtz, A. Haibel, and R.-M. Vetter, Phys. Rev. E 66 (2002) 037602.

[7] A. P. Barbero, H. E. Hernández-Figueroa and E. Recami, Phys. Rev. E 62 (2000) 8628.

[8] S. Longhi, P. Laporta, M. Belmonte, and E. Recami, Phys. Rev. E 65 (2002) 046610.

[9] Zhi-Yong Wang, Cai-Dong Xiong, and B. He, Phys. Rev. A 75 (2007) 013813.

quant-ph/0612069.

[10] Zhi-Yong Wang and Cai-Dong Xiong, Phys. Rev. A 75 (2007) 042105.

[11] H. G. Winful, Phys. Rev. E 68 (2003) 016615.

[12] H. G. Winful, Phys. Reports. 436 (2006) 1.

[13] H. G. Winful, Phys. Rev. A 76 (2007) 057803.

[14] M. E. Peskin and D. V. Schroeder, An Introduction to Quantum Field Theory, Addison-Wesley,

New York (1995), pp. 27-28.

[15] W. Greiner, J. Reinhardt, FIELD QUANTIZATION, Springer-Verlag Berlin Heidelberg (1996),

pp.208-209.

[16] C. K. Carniglia and L. Mandel, Phys. Rev. D 3 (1971) 280.

[17] A. A. Stahlhofen and G. Nimtz, Europhys. Lett. 76 (2006) 189.

[18] W. Greiner and J. Reinhardt, Quantum Electrodynamics, Springer-Verlag, New York (1992),

pp. 55-60.

[19] S. Weinberg, Gravitation and Cosmology, John Wiley & Sons, New York (1972), p. 61.

[20] Zhi-Yong Wang and Cai-Dong Xiong, Phys. Lett. B 659 (2008) 707.

[21] Zhi-Yong Wang and Cai-Dong Xiong, http://xxx.lanl.gov/abs/0708.3519.

[22] Th. Hartman, J. Appl. Phys. 33 (1962) 3427.

[23] S. T. Ali, Phys. Rev. D 7 (1972) 1668.

[24] M. Ohtsu and H. Hori, Near-Field Nano-Optics, Kluwer Academic/plenum Publishers, New

York (1999).

[25] R. P. Feynman, R. B. Leighton, and M. Sands, Feynman Lectures on Physics, Addison-Wesley,

New York, 1964, Vol. 2, Chap. 24-8.

[26] Ole Keller, Phys. Rep. 411 (2005) 1.

[27] Zhi-Yong Wang , Cai-Dong Xiong and Ole Keller, Chin. Phys. Lett. 24 (2007) 418.

11