2. Neutrino Dynamics in the Free Field Approximation
Over the past 20 years, we have seen an intensive development in the physics of materials. One of the theoretical problems in this area is the construction of models of the interaction between 2D objects and fields of QED. In order to find possible methods for its solution, it was suggested to employ the Symansik approach [
22] for construction of the QED models with space-time inhomogeneities interpreted as a description of material environments [
23,
24]. On this basis, modifications of QED were developed for modeling the interaction between QED fields and 2D materials. Some effects of this interaction were investigated and are presented in [
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38,
39,
40,
41,
42,
43].
To describe the processes called neutrino oscillations, one uses a model of a system with three pairs of four-component spinor fields
,
,
which a free-action functional reads as [
6,
11,
14,
15]
where
are the Kronecker delta-symbols,
M is a Hermitean
- matrix with three eigenvalues
and corresponding normalized three-component eigenvectors
:
We assume that
. In (
1), we used the notation
,
, and
is a Lorentz invariant scalar product of four-differential vector with a four-Dirac matrix vector
where
is the unique
matrix, and
are the Pauli matrices
It is supposed that the spinor fields and the matrix are convenient for direct description of neutrino physics and its experimentally observed features.
Using notations
we can write the free action (
1) of the model in terms of the fields
,
as
One says that the system is considered in a lepton (also-called flavor) representation, if fields and the non-diagonal mass matrix M are used for its description. In the so-called mass representation, the system states are characterized by the fields and diagonal mass matrix (i.e., by the masses ). For writing indices, we will use the letter in the lepton representation and the letter in the mass one.
The considered system of spinor fields can be characterised by the local, independent from representation, bilinear function
defined by a
—matrix
as follows
However, properties of the components of and appear to be essentially different.
In the stationarity point of
the fields
satisfy the Dirac equations
If one chooses
as their plane wave solutions
then
does not depend on the spase-time point
x. For a similar quantity of flavor representation, one obtains
The dependence on the point x of this expression is determined by the factors . If space parts of the four moments coincide by : , then and do not depend on the space coordinates of . For given , the moment component is defined as , and is a periodic function of the time coordinate with period . Thus, the function describes an evolution of the system which is characterized by three periods, . It is an example of a typical process called neutrino oscillations within the flavor description of the system.
If
, and
, then
for small
and
for small
. The free field approximation of the action functional enables to describe the propagation of neutrino in vacuum.
For processes in which the influence of the material environment is significant, it was proposed to represent this in the model by an additional potential in the Hamiltonian. In this way, models with constant and adiabatically varying density of the matter were constructed and studied by Mikheev, Smirnov, and Wolfenstein [
44,
45,
46,
47,
48,
49,
50,
51]. It was shown that the effective masses of neutrino are changed by their interaction with material media. This can cause resonance effects in the processes of neutrino oscillations (MSW resonance), which significantly change their characteristics.
The problem of modeling the interaction of neutrinos with external media attracts the attention of many researchers. It remains actual at the present time. In developing the methods used in [
44,
45,
46,
47,
48,
49,
50], many models describing the interactions of neutrino and matter with constant and adiabatically distributed density have been constructed [
14,
16,
17,
18,
19,
20,
21].
However, little attention has been paid to the study of boundary effects and phenomena generated by the strong inhomogeneous medium, for modeling of which it is necessary to take into account the interaction of neutrinos with singular density distribution concentrated in a -dimensional subspace of the Minkowski space-time. In this paper, we will demonstrate the possibility of applying the methods of quantum field theory to such problems.
3. Interaction of Neutrinos with Matter
The main idea of Symanzik’s approach in constructing renormalizable models of quantum field theory in a non-uniform space-time is to use the possibility of modifying the action functional of a usual renormalizable quantum field model which is invariant in respect to the space-time translations and Lorentz transformations by appending an additional so-called defect action functional (DAF) obeying some general requirements [
22].
The most important of these is that the modified model must remain renormalizable. It is a formal mathematical requirement that imposes strong restrictions on the possible form of the DAF. It should naturally also be assumed that the basic physical principles of interaction laws in the original model also remain non-broken in the modified one. In the gauge theory models, these could be the basic postulate about locality and local gauge invariance. In addition to that, some common physical requirements can be taken into account. For example, the DAF does not break the unitarity of the scattering matrix.
In the framework of Symansik’s approach, one constructed the model describing the interaction of the QED fields with two-dimensional material, which form is defined by the solution of equation
. The full action functional of that reads as
where
is the usual action functional of QED
with an electromagnetic field
A and spinor fields
, electron charge
e, and mass
m. The DAF is the sum of two terms:
written as
We used here the notation
for the totally antisymmetric Levi-Civita tensor (
),
a and the elements of the matrix
Q are dimensionless parameters. The matrix
Q satisfies the condition
. The parameter
a is a real number. The delta-function
describes a subspace
of
-space-time filled with 2D material [
23,
24,
33,
34]. Any
matrix can be represented as a linear combination of 16 linearly independent matrices with complex coefficients. As such basic elements, we will use the matrices
of the following form
where
I is the
identity matrix. These
can be considered as matrices that form a basis for a linear (reducible) representation of the Lorentz group. The Dirac matrices
are transformed as components of a Lorentz contravariant vector,
I is the scalar, and
is the pseudoscalar. The matrices
are represented as contravariant components of the pseudovector and antisymmetric tensor of the second rank, respectively.
Thus, the differential matrix operator and the QED action functional are invariant in respect to Lorentz transformations, and the describing the interaction of the QED fields with the extended object is breaking this symmetry. The remaining symmetry properties of the system are defined by the form of the surface and the choice of parameters of the matrix .
The action functional (
3) was proposed in [
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38,
39,
40,
41,
42,
43] as a realization of the opportunity to construct a model of the interaction of QED fields with two-dimensional materials within the framework of the Symanzik approach, unless the electron mass,
does not contain other dimensional parameters. This model is local, gauge-invariant, and renormalizable. For a material with a given shape (function
) and with given material properties (parameters
of matrix
Q and
a), it is possible to investigate theoretically within the framework of the model a large class of various problems. For example, it can be used for calculating the characteristics of scattering processes and bound states of particles. In this case, using conventional methods of QED and corresponding to their specific problem modifications, one can obtain quantitative results with a high degree of accuracy that are suitable for experimental verification and various predictions.
In our work, we propose to generalize this approach to the case of interaction of neutrino fields with matter whose distribution of density would be a local function concentrated in a subspace with dimension of the four-dimensional Minkowski space-time.
Since, by definition, the action functionals are dimensionless and the dimension of the product of two spinor fields is equal to three, for the DAF, it is necessary by
to have parameters with negative dimensions. Adding such a DAF to the basic action of the model will violate the renormalizability of that. Therefore, the only valid value for
is
. Therefore, a possible generalization of the QED functional
for a description of interaction of neutrino fields with a singularly distributed medium could be proposed for mass representation, as
Here, the elements of a hermitian
) matrix
and
matrix
Q are constant dimensionless parameters. The matrix
Q is supposed to be presented as
with 16 complex numbers
and linear independent matrices
of the form (
4). The solution of equation
describes a region of Minkowski space filled with the matter that interact with neutrinos. Its properties are presented by the parameters
. In this paper, we consider as the extended material object the plane
. It corresponds to choosing
.
We put on the matrix Q the restriction , which is necessary for the scattering matrix unitarity. It follows from , by that , , . Therefore, the coefficient by the matrix in the representation is imaginary, and all other coefficients are real.
The matrix
Q is simplified if there is a symmetry in the interaction of plane
and spinor fields. If it is assumed that the material plane is isotropic and homogeneous, that is, the DAF (
5) is invariant with respect to rotation about the
—axis and to translations along
—directions, then
Q has the form [
34]
where
, are real numbers.
The free action functional (
1) has the form
Here,
is the Hermitian
matrix and
is the Dirac
operator matrix. The proposed contribution in DAF of spinor fields
(
5) is the quadratic form of the same structure with
. From this point of view, (
5) can be considered as a minimal model of interaction of neutrinos with material planes. We will use it in this paper.
This model contains 17 real parameters: 9 of them define the matrix
, and there are 8 in
Q. The unitary flavor transformation of fields
enables one to diagonalize the matrix
in (
5). In this presentation, the eigenvalues of
appear to be the coupling constants of three neutrino mixes with the plane
.
In order to obtain an analogue of the Chern-Simons potential describing the coupling of electromagnetic fields with 2D material for the theory of weak interactions, it is sufficient to use the fact that the strength tensor of the non-abelian Yang-Mills field
has the form
, and changes as
under the local gauge transformation with the matrix
. It means that
where
is the Heaviside step function of
and
are constant parameters, is a gauge invariant functional.
If the field
disappears at large
x, then
, where
In the framework of the perturbation theory the functional is also equivalent to , and the last one can be used as DAF in modeling the interaction of the Yang-Mills field with 2D material concentrated on the surface . It can be considered as a possible generalization of the abelian Chern-Simons action functional for the non-abelian gauge vector field. If it disagrees with non-perturbative results obtained by using as an alternative versions of DAFs, this situation will impose a special investigation.
It is important to note that in the theory of gauge interactions of bosonic vector and fermionic spinor fields, their interaction with 2D materials is described within the framework of the proposed approach by the sum of functionals, each of which contains only bosonic or fermionic fields. Therefore, the influence of fields of one type on the effects of interaction of fields of another type with extended 2D objects is, in the main approximation, insignificant.
We assume that this is also true for the processes of interaction of neutrinos with a strongly inhomogeneous medium, and to study their features, we will use a model with DAF (
5) which contains only neutrino fields.
The invariant in respect to all not affecting the axis
transformations of the Lorentz group interaction of plane
with a Dirac field was considered in [
33]. For symmetry of such a kind, one needs to put
in (
6) and the matrix
Q obtain the form
If one takes into account only the properties of the plane material which are invariant in respect to all rotations and busts, then one can put
and obtain
This matrix depends on two real parameters
. If the parity symmetry is supposed to not be broken by the DAF, then
and
Thus, the full action functional describing the interaction of the material plane
with the system of Dirac fields
in the mass representation reads as
Here, the fields
have three mass components
, and
, and each of them has four spinor components. For notations of spinor and flavor indices, we will use the Latin and Greek letters, respectively. The matrices
, and
Q do not depend on
x coordinates. The
one is diagonal on the spinor and mass indices:
, but for
Q it is so only in the special case (
9). The matrix
is supposed to be Hermitian of general form.
The action functional (
10) describes three free Dirac particles with masses
interacting on the plane
. The matrix
Q represents the properties that are material of this plane which are essential for its interaction with spinor fields. The diagonal part of the matrix
defines, for each particle, its interaction constant with the plane. The non-diagonal elements of
can be considered as induced by the plane coupling constants between different
-components of the fields
.
In the flavor representation, the action functional (
10) is written as
For
, it is presented in (1) and
The matrix elements of
M,
L and
,
are connected by relations
Thus, we constructed the model of interaction of neutrino fields with strong inhomogeneous matter based on the Symanzik approach in quantum field theory. The DAF (
5) is supposed to be used as the addition term to the action functional of renormalized models describing neutrino physics. The constructed model is an analog of the model of interaction of the QED fields with 2D matter, of which the investigation in the Gaussian approximation enabled one to obtain non-trivial theoretical results about Casimir and Casimir-Polder effects [
23,
24,
27,
31,
42], scattering processes [
32,
40,
41], and the bound state of photons and Dirac particles [
37,
39,
43]. Within the Gaussian approximation of the proposed model, we consider the scattering of neutrinos on the material plane and analyze the influence of collisions with it on their oscillations.
4. Statement of the Problem
Although the action functional (
10) is Gaussian, the processes, which it describes, are nontrivial. We will study the scattering on the plane
of particles, which are presented by the fields
, by using the modified Dirac equations
characterizing the point of stationarity of the functional
. The ordinary way to do it is to find the solution of (
11) and (
12) and applying that to construct the currents of incident, reflected, and transmitted particles. It enables one to calculate the characteristics of the scattering process. Such a problem was solved for the interaction of one Dirac field with the plane
defined by the DAF with matrices
Q of the form (
6), (
7). Our task is to obtain such results for the model with an action functional (
10).
If
, and
denote solutions of (
11) and (
12) by
and
, respectively; then they must satisfy the free Dirac equations
and conditions on the plane
:
with matrix
S corresponding to the symmetry of considered interaction defined by the matrix
Q, and a
flavor matrix
.
We suppose that in the scattering process, the incident and reflected particles are in the subspace
, and the transmitted ones are in the region
. The incident and transmitted particles move in the positive direction of the
axes, and we denote by
and
the describing them spinors. Reflected particles moving in the opposite direction will be represented by the spinor
. Thus, the fields
in (
14) have the form:
,
.
If the reflection of particles exists, then
, the functions
are not continuous by
, and
in (
14). It seems to be in contradiction with (
11), since (
11) is correctly defined if
is a continuous by
function.
This problem is solved by an auxiliary regularization of the
-function in the interaction action [
33,
39]. It enables one to construct a regularized version of the conditions (
14), and it is possible after removing regularization in this expression, to obtain a finite limit for
S in terms of the coupling constants of the plane with a Dirac field. For the matrix
Q defining this interaction in the model with one spinor field, one received
[
39]. However, in the framework of other regularization schemes, it appears to be
[
33].
Thus, the matrix
S which must be expressed in terms of
Q and used in such an approach depends on choosing the regularization, but it is essential that both
and
obey the requirement
If (
14) is fulfilled, and
, then the equality (
15) ensures that
that is, no additional current is created on the plane
along the
axis.
In constructing a solution to the problem proposed by us, we assume that its determining parameters are the elements of the matrices
S and
. In this case, it is supposed that there is a regularization procedure for the delta function in the action of the model, which makes it possible to establish a one-to-one relationship between elements of matrices
and
. In this respect,
S and
can be considered to be directly related to the observables and independent from the choice of the regularization procedure. Calculations based on the use of boundary condition (
14) do not require any additional regularization scheme. Therefore, values of elements
S and
can be expressed in terms of experimental data. Values of the elements of the matrices
may depend on the choice of regularization. It can be compared with renormalization in models of quantum field theory, where the observable values are expressed in terms or renormalized parameters which are considered as independent from bare parameters of Lagrangian and used regularization.
We suppose to calculate the characteristic of the scattering process by using the boundary condition (
14) and to obtain results in terms of matrices
. The problem here is that for our model, we are given matrix
Q, but we do not know the matrix
S independent of the choice of regularization. In this situation, it is natural to try to construct this matrix by analyzing the properties of the matrices
and
.
To reveal their structural features, which may also be the same for
S, we introduce convenient notation that was used in [
37,
39]. If
M is a
matrix with elements
,
, then
are the
matrices
We will use the notations
,
for the
matrices corresponding to the unit
matrix
and Pauli matrices
,
. The matrix
Q (
6) can be written as
where
, and
.
In virtue of
, we receive
As
, the condition
is written for the
-matrices
as
. This means that
are imaginary and
are real numbers, since
, and
with
. It is fulfilled for real parameters
,
in (
6). It follows from (
16) that the parameters
are imaginary and
are real.
If
M and
N are
-matrices, and
is an analytical function at
, then the following relations are fulfilled for the matrices
The matrix
S from the boundary condition (
14) can be obtained in the regularised model. It appears to be dependent on the regularisation scheme function of the matrix
. Examples of such could be
and
.
For the description of neutrino scattering on the plane
, we will use the matrix
S of the form
where
and
,
are real numbers. Employing this parameterization does not generate difficulties for constructing a complete solution to the problem posed by us. On the other hand, there is no reason to expect that the appearance of results obtained in this way cannot be received within the framework of the approach using regularization (see the
Appendix A). To present formulas in compact form, it is convenient to also use the following notations
The condition is written for these parameters as .
It can also be useful to present
,
as
Here, and there are not conditions connecting with each other.
5. Scattering of Plane Waves
The solution of free Dirac equations (
13) can be presented in the form
Here, we used the notation
and
to obey the condition
. If
, then the spinor
in the integrand of (
21) describes by
the particle moving in the positive direction of
-axes, and
corresponds to movement in the opposite direction. The spinor
fulfills the Dirac equation
For the scattering process described by (
13) and (
14), the most general plane wave presentation of spinors
can be chosen as
Here,
where
,
and
are arbitrary complex parameters. Functions
describe the incidend and transmitted particles moving in the positive direction of
-axes, and
corresponds to reflected particles moving from the plane
in the negative direction of
-axes. The boundary condition (
14) is written for
as
with matrix
S presented in (
17) and (
18) and
Thus, (
22) is a system of 12 linear equations which enables one to express the amplitudes
,
,
of reflected and transmitted particles in terms of amplitudes
of incident ones. Substituting the spinors
in (
22) and using the notations
for
, we obtain six equations of the form
It follows from (
24) that
where we used the notations
,
Excluding
,
in (
25), one receives the equations
with (
)-matrices
The solution for
and
can be written as
and the problem is to construct the matrices
in an explicit form.
It can be solved in the more general formulation. Let
be (
)-matrices,
-
n-component vectors,
, and
One needs to construct the solution of the system of equations (
27)
with an explicit form of the
-matrices
,
.
Finding the components
of the vector
V directly from Equation (
27), one obtains them in the form
Here, the sets
of the indexes are assumed to be chosen as
. Comparing (
28) and (
29), we receive the following expressions for the matrices
Using
and (
26)–(
29), one can calculate the matrices
and obtain the right-hand sides in the representations (
26) of
in an evident form. However, for the unitary
matrix
of general form, they turn out to be rather cumbersome and inconvenient for analyzing their properties. Therefore, to study the most simple effects of the neutrino interaction with planes, we restrict ourselves in this paper to the case of a diagonal matrix,
.
6. Explicit Results in a Simplified Model
In virtue of
, a diagonal part of
has the form
with real parameters
. For
of such a kind, the matrices
are diagonal, and (
26) is written as
Using the notations (
19) and (
20), one can present (
30) and (
31) as
with
and
where
.
The currents
of incident, reflected, and transmitted particles are of the form
Here, are components of the momentum vectors , , . In the used parametrization, .
In virtue of (
23) and
,
Thus, the continuity of the current component
at the plane
implies that
. Taking into account the direction of the current
, we come to the conclusion that the equality
must be fulfilled. To verify (
32), it is sufficient to use the above given expressions for the matrices
and to take into account the relations
for the parameters on which they depend. It follows from (
32) that if one denotes the reflection and transmission coefficients for the considered scattering process of the
-th particle as
then
The matrices
are Hermitian. The elements
,
of the matrix
can be presented as
where
. The elements
of the matrix
for
,
are the following
This is a consequence of the relation (
32).
We see that in the mass representation, the obtained results for characteristics of the scattering process do not depend on parameters
of diagonal matrix
. However, they influence the oscillation events of transmitted and reflected neutrinos in the flavor description of the system, taking the contribution to the oscillating part of (
2) in the form
It can essentially change the characteristics of neutrino oscillations after their collision with the material plane.
In the model with diagonal matrix
and matrix
S of the form (
17), we constructed a plane wave solution of Equations (
13) and (
14), describing the motion of particles with an arbitrary angle of incidence to the plane
. Now, we consider a special process of neutrino collision with a plane with an angle of incidence equal to zero.
7. Moving of Particles along the -Axes
If the particles move orthogonally to the plane’s
direction, then
,
. This scattering process is described by reflection and transition matrices
where
The matrices
,
are also diagonal. Their elements
fulfill the relations
. The reflection and transition coefficient could be written as
Here, the angle is defined by relation .
Let us denote
where
are constant parameters determining this function, which will be used to analyze the scattering process under consideration. It follows directly from definition that
. For
, this function reaches its maximal value
by
:
The equation
has, by
and
, two solutions,
:
The neighbourhood
of the point
can be considered as characterizing the function
region. We denote an estimation of its extension as
. Then,
The parameters
can be expressed in terms of
as
Substituting them in
, we receive the following presentation of this function
Let us put
, where
are real constants which can be parameterized by
as follows:
Then,
and
The function
has the form
, where
and
. Since this relation between parameters
,
, and
is symmetric in respect to replacement
, there are two possibilities of inverting it, as
Since , , the relation 2) imposes the restriction on , and in case 1), the inequality must be fulfilled. Thus, the function can be presented in the form with , , . It is even: has the maximum value by , and the extension of the neighborhood of is . By given , the function has the minimal value of by . If , then .
The graphs of the functions
by different
are shown on
Figure 1. With an increase of
c, the vicinity of the maxima becomes increasingly more flat, and the positions of the maxima become increasingly less noticeable. Using the notation
one can present the functions
of
in the form
and obtain the following expression for the transmission coefficient:
Let us consider, as an example, a scattering process for which
and
. In this case,
and
The maximal value of
is
,
,
. The parameters
(
34) are very close to ones defining the function presented by Graphic 1 on the
Figure 1. The difference between it and the graph of the function
(
33) is inessential. Note that
(
34) fulfill the restrictions
and
. However,
and the inequality
is not satisfied. Therefore, there is only one
-parametrization
of the function
in this case.
Outside the neighborhood,
(
) of the point
(
,
(
33) is 10 times less than its maximum value. For
, it decreases monotonically from
to
and has the maximal value
at the point
(
34).
The graph of
(
33) differs significantly from the (graphs (2)–(6)) shown in
Figure 2 for
by
. These ones can have two maxima (graphs (2) and (3)), one smoothed maximum (graph (6)), or change very little on the most part of the interval
(graphs (4) and (5)).
In the general case, the process of neutrino scattering on the plane is described by one universal function parameterized by constants : with factors . In the terms of parameters used above, , , , . The maximal value of is . The argument k of this function is , where m is the mass of the particle and its energy. Hence, by the physical values of . If , then has the maximum on the interval . If , then grows from to with an increase of k by .
Since, in the considered scattering process m, and k fulfill the relations , , one can present , as , . Hence, if , then and , .
Corresponding to (
33) function
, the transmission coefficient as functions of
has the maximum by
m, and as the function of
, it is maximal by
.
Now, we analyze the features of system dynamics in the framework of the flavor representation. As the matrices
used in (
2) when discussing the oscillation process of neutrinos, we consider
. The function
is interpreted as the scalar (pseudo-scalar) density of the
-
particles with momentum
,
(
) is the four-current vector (axial four-current pseudo-vector) of the
-
particles moving along the
-axis. We will use the plane wave presentation for spinors
describing the particles with momentum
moving orthogonal to the plane
:
By means of notation
, the results obtained for the
are written in the form
and
Oscillations of incident waves are presented by
as
and
For transmitted waves, one needs to replace
. Thus, changing in (
35) and (
36)
one receives the corresponding results for transmitted neutrino oscillations. If
and momentum
is the same for all particles, then
. For given
, we denote
,
. If
monotonically increases from zero to
by
, then
, since
. In virtue of
the greater the derivatives of the function
are at points
and
for given
that is, the faster
grows on the interval
, the more there will be an influence of the neutrino collisions with the plane on the process of their oscillations.
If there is a maximum of by , then each , can be maximal, and the following situations are possible:
- (1)
is maximal, ;
- (2)
is maximal,
or ;
- (3)
is maximal, .
The more the function resembles an approximation of the delta-function and the nearer the maximal is to the maximum of , the greater is the difference of maximal from non-maximal , and the more significant will be the changes in neutrino oscillations as a result of their collisions with the plane.
We considered the model with 8 real parameters in the matrix
Q (
6). It could be interesting to compare its properties with features of the system with matrices
Q of the form (
7)–(
9). In the case (
7), there are four real parameters defining
Q and
Here,
,
are real parameters fulfilled the conditions
,
and
,
. Hence, for such materials,
Putting
in (
37)–(
40), one obtains for the model given by matrix
Q (
8) with two real parameters
with real
,
fulfilling the conditions
,
, and
,
. In this model,
If one sets in (
41) and (
42)
, then one receives, for the model with a matrix
Q (
9) containing one real parameter,
Here,
,
are real parameters,
,
and
. In this case,
The essential difference between the models with matrix
Q of the form (
6)–(
9) is that the functions
, in the models (
6), (
7) can have the maximum by
, but
in models (
8), (
9) are, by
, the monotonously growing functions.
8. Conclusions
In our work, we considered the problem of neutrino interaction with matter. Using our experience in constructing models of QED in singular background fields, we have proposed a quantum-field approach, which may be useful for the theoretical description of neutrino propagation in a highly inhomogeneous medium. It assumes taking into account the basic symmetry principles of modern physics of fundamental interactions that underlie the Standard Model and can, in principle, be generalized to describe the interaction of all lepton fields with the external environment. Mainly, attention was paid to the problem of neutrino scattering on a material plane, considered as a simplest example of process in the space with a strongly inhomogeneous distribution of substance.
In a general form, for a model with an off-diagonal unitary matrix mixing Dirac fields in the mass representation, expressions for the reflected and transmitted waves were obtained. For them, in the model with a diagonal , an explicit solution was obtained, which was used to analyze the oscillations of neutrinos in the case of their motion, orthogonally to the plane . It was shown that the parameters that determine the material properties of the plane can be chosen so that its effect on the neutrino flux is similar to a filter that transmits particles in a narrow interval of low energies and almost completely reflects all other ones. As a result of the neutrino collision with the plane, the parallel component of the momentum does not change, and the orthogonal one does not change in absolute value. Only the amplitudes of fields can change significantly.
The example we have considered with parameters of the model (
33) and (
34) shows that the interaction of neutrinos with a plane can effect their filtration. A characteristic feature of the filtration process of particles upon collision with a plane is the possibility of essentially different transmission coefficients for neutrinos of different masses. In this case, due to filtration of their flux, the regimes of the neutrino oscillations before the plane and behind it can be strongly different. This phenomenon can be used to estimate the masses of neutrinos of various types in carrying out analyses of experimental data.
Although the filtering and MSW-resonance results are similar in many ways, their mechanics are not the same. The MSW effect is formed non-locally in space and time. This requires a certain volume of matter and a certain period of time, generally speaking, that are different for various substances. In order for the filtration of the neutrino flux to occur, their collision with the plane is sufficient, which (in the framework of the considered model) occurs instantaneously and locally at . From the point of view of the possibility of verifying the adequacy of the proposed model, it would be interesting to reveal in the dynamics of neutrino oscillations an effect, which cannot arise as a consequence of MSW resonance and is a manifestation of the filtration process.
One of the current theoretical problems in neutrino astro-physics is constructing numerical models of dynamics of supernova explosions. Many research teams have been working on this issue [
52,
53,
54,
55,
56,
57,
58,
59,
60,
61,
62,
63,
64,
65,
66,
67]. Perhaps, taking into account the filtering mechanism in such models will be useful to achieve a better understanding of the features of the process of collapse of the super-heavy star core.
In general terms, a possible scenario of its evolution can be presented as follows. If, in the core of the star, its shell filters neutrinos by energies, they can be divided into two classes. Particles with energies from the narrow range of the low-energy region leave the core unhindered. These neutrinos can be called free. For all the others, which we will call bound, the core shell is impermeable.
In the process of the star’s evolution, its core is subjected to the pressure of the increasing gravitational forces. In it, neutrinos are born, the free ones are emitted, and the bound neutrinos are accumulated in the core. This can go on until the main features of the interaction between the core shell and neutrinos changes, the class of free neutrinos expands, and the star will emit them with further contraction without a significant change in its structure. In our model, this can be described by changing the function
. For instance, if the plots of possible
, is shown in
Figure 2, then within the change
, a large fraction of the bound high-energy neutrinos become free, they will leave the star, essentially changing the intensity and spectrum of its neutrino emission.
If the main interaction features of neutrinos with the core shell are not changed and at least some of the bound neutrinos do not become free, then the enormous energy accumulated by them in the core destroys its shell sooner or later. After that, the core and the star are exploded.
There is great interest among experimentalists to determine neutrino masses directly [
68,
69,
70,
71,
72]. One can assume that the employment of 2D materials and special surface treatment techniques in the construction of neutrino detectors would enable one to efficiently use the filtering mechanism in experiments of such a kind.
The maximal number of parameters in the model, the properties of which we have studied in detail, was eight for the matrix Q and three for the diagonal matrix . Not all of them are included in our results, and the question arises whether it is possible to reduce the number of parameters in the model without limiting its area of applicability. We considered versions of the model with four, two, and one parameters in the Q matrix, simplified for symmetry reasons, and found a difference in the properties of their predicted transmission coefficients. This raises the question of whether one can confine oneself to using the simplest models to describe real neutrino oscillation data.
We suppose that the proposed method for modeling the processes of interaction of neutrinos with matter can be useful for theoretical studies and analysis of the obtained experimental data.