NuOscProbExact: a general-purpose code to computeexact two-flavor and three-flavor neutrino oscillation probabilities

Mauricio Bustamante1, ∗

1Niels Bohr International Academy & DARK, Niels Bohr Institute,University of Copenhagen, DK-2100 Copenhagen, Denmark

(Dated: May 7, 2019)

In neutrino oscillations, a neutrino created with one flavor can be later detected with a differentflavor, with some probability. In general, the probability is computed exactly by diagonalizing theHamiltonian operator that describes the physical system and that drives the oscillations. Here weuse an alternative method developed by Ohlsson & Snellman to compute exact oscillation proba-bilities, that bypasses diagonalization, and that produces expressions for the probabilities that arestraightforward to implement. The method employs expansions of quantum operators in terms ofSU(2) and SU(3) matrices. We implement the method in the code NuOscProbExacta, which wemake publicly available. It can be applied to any closed system of two or three neutrino flavorsdescribed by an arbitrary time-independent Hamiltonian. This includes, but is not limited to, os-cillations in vacuum, in matter of constant density, with non-standard matter interactions, and ina Lorentz-violating background.

I. INTRODUCTION

Neutrinos are created and detected in weak interac-tions as flavor states — νe, νµ, ντ — but they propagateas superpositions of propagation states — in vacuum,these are the mass eigenstates ν1, ν2, ν3. Because thesuperposition evolves with time, a neutrino created witha certain flavor has a non-zero probability of being de-tected later with a different flavor [1–4]. The observationof oscillations in solar, atmospheric, reactor, and accel-erator neutrinos has led to the momentous discovery ofneutrino mass and of flavor mixing in leptons [5, 6].

Computing the probabilities of flavor transition is in-tegral to studying oscillations. Computing them exactlytypically involves diagonalizing the Hamiltonian operatorthat drives the time-evolution of neutrinos. But, becausethe expressions involved are often complex, it is notori-ously hard to produce exact analytical expressions for theprobabilities that also provide physical insight. The caseof oscillations in vacuum is an exception [7–9]. Beyondthat, there is a large body of work dedicated to deriv-ing exact probabilities for different scenarios; see, e.g.,Refs. [10–27]. Yet, though some of these expressions aresuperficially elegant, they are seldom used due to theirunderlying complexity, particularly in the case of oscilla-tions amongst three neutrino flavors.

More often, carefully selected perturbative expansionsand approximations are employed to cast the probabili-ties in forms that are amenable to physical interpretation.Many such approximate expressions exist in the literature[23, 28–31], especially for oscillations in matter [26, 32–43] with precisions that reach the per-cent level. Unfortu-nately, there is no systematic way to produce these usefulexpressions, since they are tailored to specific Hamilto-nians (however, see, Ref. [44]), their derivation is not

a https://github.com/mbustama/NuOscProbExact

trivial, or their application is limited to specific ranges ofvalues of a perturbative parameter.

Hence, the best course of action in cases where we seekhigh precision in the computation of probabilities is sim-ply to compute them exactly, often numerically. Thisis a common strategy to explore non-standard oscillationscenarios, i.e., arbitrary Hamiltonians, for which analyticsolutions are in general unavailable. For instance, this isdone when scanning a parameter space without knowinga priori our region of interest, or approximate expressionsof the probabilities that are valid inside that region.

Here, in lieu of diagonalizing the Hamiltonian, we usean alternative method, developed by Ohlsson & Snellman(hereafter, OS) in Refs. [45–48], to compute exact oscil-lation probabilities. We provide a numerical implemen-tation for systems of two and three neutrino flavors. Themethod relies on expanding the quantum operators thatdrive the time-evolution of neutrinos in terms of SU(2)and SU(3) matrices [45–48]. It has two assumptions:

1. The system must be closed, i.e., it must conservethe number of neutrinos summed over all flavors

2. The Hamiltonian must be time-independent (ex-cept in some cases; see Section VI B)

Both conditions are satisfied in many physical scenariosstudied in the literature, e.g., oscillations in vacuum, inmatter of constant density, with non-standard neutrinointeractions, and in diverse new-physics scenarios. Themethod does not apply to scenarios where neutrinos “leakout” of the system, e.g., 3+1 systems of sterile neutrinos[49–53], with neutrino decays into invisible products [54–59], or open systems, like those with decoherence [60–64].

We provide the computer code NuOscProbExact [65], alightweight numerical implementation of the OS methodthat computes exact two- and three-flavor oscillationprobabilities for arbitrary time-independent Hamiltoni-ans. The code can be easily used in oscillation analyses.

arX

iv:1

904.

1239

1v2

[he

p-ph

] 5

May

201

9

2

In Section II, we set the scope, context, and approachof the paper. In Section III, we recap the basics ofneutrino oscillations and establish the concrete goal ofthe computation. In Sections IV and V, we presentthe OS method, in a simplified formulation, for systemsof two and three flavors. In Section VI, we describeNuOscProbExact and show examples of its use. In Sec-tion VII, we conclude.

II. SCOPE, CONTEXT, AND APPROACH

Below, to compute the oscillation probabilities, we fol-low the OS method. References [45–48] introduced ex-pressions applicable to generic oscillation scenarios, andalso found analytic expressions [45] for the probabili-ties in the cases of two-flavor oscillations in matter andthree-flavor oscillations in vacuum and matter. Refer-ence [10] presented an earlier application to three-flavoroscillations in matter. Because the exact analytic expres-sions tailored to three-neutrino oscillations in vacuumand matter are lengthy, and because we are interested inproviding a general-purpose numerical implementation ofthe method, we do not attempt to reproduce analyticalsolutions or find new ones.

Later, we work through the method. Here, we givean overview. We start by expanding the Hamiltonianin terms of 2 × 2 Pauli matrices — in the case of twoneutrino flavors — or of 3× 3 Gell-Mann matrices — inthe case of three flavors. Appendix A shows these matri-ces. When studying neutrino oscillations, these expan-sions are sometimes performed not on the Hamiltonian,but on the associated density matrix. This approach isparticularly useful to study oscillations in the early Uni-verse [66–71] and in supernovae [72–79].

In the OS method, we instead first expand the Hamil-tonianH and then the associated time-evolution operatore−iHt. For the latter, we use the exponential expansionsof Pauli and Gell-Mann matrices [80–82]. These expan-sions are a direct application of the Cayley-Hamilton the-orem, which states that an analytic function of an n× nmatrix can be written as a polynomial of degree (n−1) inthat matrix. The coefficients of the expansion are com-puted using SU(2) and SU(3) invariants, which allowsus to bypass the diagonalization of the Hamiltonian thatwould otherwise be needed to compute the probabilities.

Sophisticated numerical codes exist to compute prob-abilities, either for general application or for particularscenarios, e.g., GLoBES [83], nuCraft [84], NuSQuIDS [85],and Prob3++ [86]. The general-purpose code NuSQuIDS[85, 87] implements the same expansions used in the OSmethod efficiently, and embeds them in a larger formal-ism that can also deal with time-dependent Hamiltoni-ans. The code Prob3++ [86] implements the expansionsfor oscillations in matter, based on Ref. [10].

While it is possible to extend the method to systems ofn > 3 neutrino flavors, the expansions in SU(n) quicklybecome complicated [82, 88]. Since the objective of

NuOscProbExact is to treat the common cases of two-and three-neutrino oscillations, exploring these general-izations is beyond the scope of this paper. However, Ref.[89] applied the OS method to the n = 4 case for four-flavor oscillations in matter and NuSQuIDS [85, 87] imple-ments it for cases up to n = 6 [90].

Below, our approach is expository while condensed: weprovide sufficient detail to present the method and facil-itate its implementation, and refer to earlier works forfurther mathematical detail.

III. NEUTRINO OSCILLATION RECAP

Let ν represent the flavor state of a neutrino. The stateevolves according to the Schrodinger equation

idν

dt= Hν , (1)

where t is the time elapsed since the creation of the neu-trino and H is the Hamiltonian written in flavor space.We use units where c = ~ = 1. By definition, H is Her-mitian. In a system of n neutrinos, we represent H by an × n matrix and ν by a column vector with n entries.Below, we consider the cases n = 2, for two-neutrinooscillations, and n = 3, for three-neutrino oscillations.

We restrict the discussion to time-independent Hamil-tonians, so that the corresponding time-evolution opera-tor is U (t) = e−iHt. Hamiltonians of this type describe,for instance, neutrino propagation in vacuum and in mat-ter of constant density. Because neutrinos are relativistic,we approximate the propagated distance L ' t. Thus,the evolved state of a neutrino born as να (α = e, µ, τ) is

να (L) = U(L)να = e−iHLνα . (2)

SinceH is Hermitian, the evolution operator U is unitary.Because the Hamiltonian in flavor space is non-

diagonal, i.e., because it mixes flavor states, after prop-agating for a distance L, the neutrino of initial flavorνα becomes a superposition of neutrinos of all flavors,

each with a different probability amplitude, ν†βνα(L)

(β = e, µ, τ). The probability of detecting the neutrino

with flavor β is Pνα→νβ (L) = |ν†βνα(L)|2.

In Eq. (2), to compute the action of the evolution op-erator, να must be an eigenstate of H. Yet, this is typi-cally not the case. Thus, the usual procedure to computethe evolved state is to diagonalize the Hamiltonian inEq. (2), compute the evolved state in the space spannedby the eigenvectors of the Hamiltonian, and rotate backto flavor space to obtain να(L). These steps are oftencarried out numerically, especially in the three-neutrinocase, because the expressions quickly become unmanage-able. There are numerical codes that do this efficiently,e.g., GLoBES [83, 91, 92].

Below, we follow instead the OS method, as explainedin Section II, implement it numerically, and show resultsof the implementation.

3

Coefficient Expression

h012

[(H2)11 + (H2)22]

h1 Re [(H2)12]

h2 −Im [(H2)12]

h312

[(H2)11 − (H2)22]

TABLE I. Coefficients in the expansion of the two-neutrinoHamiltonian H2 in Eq. (3). The coefficient h0 does not takepart in the calculation of the flavor-transition probability; weinclude it here for completeness.

IV. TWO-NEUTRINO OSCILLATIONS

We consider first oscillations between only two neu-trino flavors; later, we consider three flavors. This is agood approximation when describing reactor, accelera-tor, and atmospheric neutrinos. We represent the two-neutrino Hamiltonian operator by a 2×2 matrix H2. Thethree traceless, Hermitian Pauli matrices σk (k = 1, 2, 3)— the generators of the SU(2) algebra — plus the iden-tity matrix 1 make up the orthogonal basis of 2× 2 ma-trices. Thus, we expand the Hamiltonian as

H2 = h01 + hkσk , (3)

where, here and below, we assume the Einstein conven-tion of summing over repeated indices. The coefficientsh0 and hk are functions of the components of the Hamil-tonian; we show their explicit expressions in Table I. Inthe two-neutrino case, the neutrino state at any time isν(L) = fα(L)να + fβ(L)νβ , where fα and fβ are, respec-tively, the probability amplitudes of measuring the stateto be a να or a νβ (with α 6= β). We represent the neu-trino state as a two-component column vector; the pure

states are να = (1 0)T

and νβ = (0 1)T

.Thus, the evolution operator is U2 (L) = e−iH2L =

e−i(h01+hkσk)L. We factorize1 this into e−ih01Le−ihkσ

kL.The operator e−ih01L introduces a global phase that doesnot affect the probability, i.e., e−ih01Lν = e−ih0Lν. After

discarding it, we are left with U2(L) = e−ihkσkL.

To compute the action of U2, we use a well-known iden-tity of Pauli matrices that generalizes Euler’s formula,

e±iakσk

= cos (|a|)± iakσk sin (|a|) , (4)

where a is a unit vector in the direction of the vectora = (a1, a2, a3) and |a| is its modulus. So we can writethe evolution operator as in Ref. [45],

U2 (L) = cos (|h|L) 1− i sin (|h|L)

|h| hkσk , (5)

1 We can do this because the commutator C2 ≡[h01, hkσk

]= 0,

so that [h01, C2] =[hkσ

k, C2

]= 0. For analogous reasons, we

can also do this in the three-neutrino case in Section V.

Coefficient Expression

h013

[(H3)11 + (H3)22 + (H3)33

]h1 Re

[(H3)12

]h2 −Im

[(H3)12

]h3

12

[(H3)11 − (H3)22

]h4 Re

[(H3)13

]h5 −Im

[(H3)13

]h6 Re

[(H3)23

]h7 −Im

[(H3)23

]h8

√3

6

[(H3)11 + (H3)22 − 2 (H3)33

]TABLE II. Coefficients in the expansion of the three-neutrinoHamiltonian H3 in Eq. (7). The coefficient h0 does not takepart in the calculation of the flavor-transition probability; weinclude it here for completeness.

where |h|2 ≡ |h1|2 + |h2|2 + |h3|2.The evolved state να(L) of a neutrino that was created

with flavor α, i.e., with fα(0) = 1 and fβ(0) = 0, isνα (L) = U2(L)να. After some manipulation, the flavor-

transition probability Pνα→νβ (L) = |ν†βU2(L)να|2 is

Pνα→νβ (L) =|h1|2 + |h2|2

|h|2sin2 (|h|L) (α 6= β) , (6)

where |h1|2 + |h2|2 = |(H2)12|2 and |h|2 = |(H2)12|2 +|(H2)11−(H2)22|2/4. Because of the conservation of prob-ability, Pνα→να (L) = 1 − Pνα→νβ (L). Appendix B con-tains the derivation of Eq. (6). Appendix C shows asimple application to two-flavor oscillations in vacuum.Equation (6) is our final result in the two-neutrino case.

The key to the calculation of Eq. (6) was to expandthe time-evolution operator via the Pauli-matrix identity,Eq. (4). Later, in the three-neutrino case, we use ananalogous identity for the Gell-Mann matrices.

V. THREE-NEUTRINO OSCILLATIONS

We follow the steps that we used in the two-neutrinocase closely. We represent the three-neutrino Hamilto-nian by a 3 × 3 matrix H3. The eight traceless, Hermi-tian Gell-Mann matrices λk (k = 1, . . . , 8) — with λk/2the generators of the SU(3) algebra — plus the identitymatrix 1 make up the orthogonal basis of 3× 3 matrices.Thus, we expand the Hamiltonian as

H3 = h01 + hkλk , (7)

where h0 and hk are now functions of the componentsof H3; we show their explicit expressions in Table II.The neutrino state at any time is ν(L) = fe(L)νe +fµ(L)νµ + fτ (L)ντ , where fe, fµ, and fτ are, respec-tively, the probability amplitudes of measuring the stateto be a νe, νµ, or ντ . We represent the neutrino stateas a three-component column vector; the pure states are

νe = (1 0 0)T

, νµ = (0 1 0)T

, and ντ = (0 0 1)T

.

4

Tensor component Value

d(118) = d(228) = d(338)1√3

d(146) = d(157) = d(256) = d(344) = d(355)12

d(247) = d(366) = d(377) − 12

d(448) = d(558) = d(668) = d(778) − 1

2√3

d888 − 1√3

TABLE III. All of the non-zero components of the tensor dijk,defined in the main text. The tensor is completely symmetricin its indices. Here, i, j, and k can each take integer valuesbetween 1 and 8. The notation (ijk) represents all permuta-tions of the indices in parentheses. A component vanishes ifthe number of indices in the set {2, 5, 7} is odd.

The evolution operator is U3 (L) = e−iH3L =

e−ih01Le−ihkλkL. Again, after discarding the global

phase, we are left with U3 (L) = e−ihkλkL.

Next we compute the action of U3 on a neutrino state.We wish to expand U3 using an identity for the Gell-Mann matrices that is similar to the identity for the Paulimatrices, Eq. (4), and that allows us to write

U3 (L) = u01 + iukλk , (8)

where the complex coefficients u0 and uk are functionsof L and the hk. Reference [80] introduced and demon-strated such an identity; below, we make use of theirresults, leaving most of the proofs to the reference. Seealso Refs. [81, 82, 88, 93–95] for further details.

The coefficients in Eq. (8) can be trivially written asu0 = 1

3Tr U3 and uk = − i2Tr(λkU3); next we unpack

these forms. An application of Sylvester’s formula [96]to 3 × 3 matrices allows us to express the coefficients interms of the SU(3) invariants

L2 |h|2 ≡ L2hkhk ,

−L3〈h〉 ≡ −L3dijkhihjhk .

The tensor dijk = 14Tr ({λi, λj}λk), where the brackets

represent the anticommutator. It appears in the productlaw of Gell-Mann matrices and its components are thestructure constants of the SU(3) algebra. Table III showsall non-zero components.

Next, we solve the characteristic equation of −hkλkL,i.e., φ3 − (L2 |h|2)φ − 2

3 (−L3〈h〉) = 0. The equationfollows from the Cayley-Hamilton theorem, written con-veniently in terms of invariants [80, 81]. Its three latentroots, or eigenvalues, are φm ≡ ψmL (m = 1, 2, 3), with

ψm ≡2 |h|√

3cos

[1

3(χ+ 2πm)

], (9)

where cos (χ) = −√

3〈h〉/|h|3. The step above is key:writing the eigenvalues in terms of the SU(3) invariantsallows us to bypass an explicit diagonalization [81].

Three-neutrino probability Expression

Pνe→νe

∣∣∣u0 + iu3 + i u8√3

∣∣∣2Pνe→νµ |iu1 − u2|2

Pνe→ντ |iu4 − u5|2

Pνµ→νe |iu1 + u2|2

Pνµ→νµ

∣∣∣u0 − iu3 + i u8√3

∣∣∣2Pνµ→ντ |iu6 − u7|2

Pντ→νe |iu4 + u5|2

Pντ→νµ |iu6 + u7|2

Pντ→ντ

∣∣∣u0 − i 2u8√3

∣∣∣2TABLE IV. Exact three-neutrino oscillation probabilities, foran arbitrary time-independent Hamiltonian. The complex co-efficients u0 and uk are computed in Eqs. (10) and (11).

With this, the coefficients in Eq. (8) are

u0 =1

3

3∑m=1

eiLψm , (10)

uk =

3∑m=1

eiLψmψmhk − (h ∗ h)k

3ψ2m − |h|2

, (11)

where (h ∗ h)i ≡ dijkhjhk. Appendix D contains the

derivation of Eq. (11). Using Eqs. (9), (10), and (11), wewrite the evolution operator concisely as in Ref. [45],

U3(L) =

3∑m=1

eiLψm[1 +

ψmhk − (h ∗ h)k3ψ2

m − |h|2λk]. (12)

Equation 12 was introduced in Refs.[45–47], and appliedto find analytic expressions of the probabilities in thecases of oscillations in vacuum and matter. For a nu-merical implementation of Eq. (12) it is convenient tocalculate the coefficients u0 and uk with Eqs. (10) and(11), and use them to directly expand U3 in Eq. (8). Thisis the strategy that we adopt in NuOscProbExact [65].

The evolved state of a neutrino created as να isνα(L) = U3(L)να. Therefore, the flavor-transition prob-

ability is Pνα→νβ (L) = |ν†βU3(L)να|2. Table IV showsthe expressions for the probabilities in terms of the co-efficients u0 and uk. These are our final results in thethree-neutrino case.

Because the algebra of Gell-Mann matrices is morecomplicated than that of Pauli matrices, the identity thatexpands the exponential of Gell-Mann matrices in Eq. (8)is notoriously more complicated than the identity thatexpands the exponential of Pauli matrices, Eq. (5). Aspointed out by Ref. [80], this may seem a disappoint-ing generalization of Eq. (4) to SU(3). However, whenconstructing an exponential parametrization of SU(3),there is no way to avoid the solution of at least a cu-bic equation. Regardless, following the procedure aboveyields exact three-neutrino flavor-transition probabilitiesfor arbitrary time-independent Hamiltonians.

5

VI. CODE DESCRIPTION AND EXAMPLES

Description.— The code NuOscProbExact that we pro-vide is a lightweight numerical implementation of the OSmethod described above. It computes exact oscillationprobabilities in the often-studied two- and three-flavorcases, for arbitrary time-independent Hamiltonians. (Formore than three flavors and time-dependent Hamiltoni-ans, see NuSQuIDS [85].) NuOscProbExact is fully writtenin Python 3.7; it is open source, and publicly availablein a GitHub repository [65].

The main input to NuOscProbExact is the Hamiltonianmatrix H2 or H3, provided as a 2×2 or 3×3 list. The codeinternally computes the hk coefficients using Table I inthe two-neutrino case and Table II in the three-neutrinocase. To compute two-neutrino probabilities, the codeevaluates Eq. (6). To compute three-neutrino probabili-ties, the code evaluates the expressions in Table IV.

Documentation.— Detailed documentation is in theGitHub repository [65], and is bundled with the code.

Examples.— Listing 1 shows a basic code example ofhow to use NuOscProbExact to compute three-neutrinoprobabilities in four representative oscillation scenarios:in vacuum, in matter of constant density, with non-standard interactions in matter, and with Lorentz invari-ance violation. Bundled with the code we provide furtherexamples, also for two-neutrino oscillations.

Below, we introduce each scenario briefly; we do notexplore their phenomenology, but we provide references.Following our tenet, we do not derive analytic expressionfor the probabilities, only numerically evaluate them.

Figure 1 shows the probabilities Pνe→νe , Pνµ→νe , andPνµ→νµ for the four scenarios, as a function of neutrinoenergy, computed using NuOscProbExact [65]. We set thebaseline to L = 1300 km to match that of the far detectorof the planned DUNE experiment [97]. The parametersand their values used in each example case are introducedbelow. All of the Hamiltonians below are written in theflavor basis. Figure 1 can be generated by running thebundled example file oscprob3nu plotpaper.py.

Appendix E shows the two-neutrino counterparts ofthe example three-neutrino scenarios presented below.We provide implementations of these two-neutrino sce-narios as part of NuOscProbExact [65].

A. Oscillations in vacuum

The Hamiltonian that drives oscillations in vacuum is

Hvac3 (E) =

1

2E

(R3,θM

23R†3,θ

), (13)

where M23 ≡ diag(0,∆m2

21,∆m231) is the mass matrix,

with ∆m221 ≡ m2

2−m21 and ∆m2

31 ≡ m23−m2

1, and the 3×3 complex rotation matrix R3,θ is the Pontecorvo-Maki-Nakagawa-Sakata (PMNS) mixing matrix. We express itin terms of three mixing angles, θ13, θ12, θ23, and oneCP-violation phase, δCP [9].

1 import numpy as np2

3 # NuOscProbExact modules4 import oscprob3nu # Core functionality5 import hamiltonians3nu # Sample Hamiltonians6 from globaldefs import * # Constants (in capitals)7

8 energy = 1.e9 # Neutrino energy [eV]9 baseline = 1.3e3 # Baseline [km]

10

11 # Vacuum Hamiltonian before multiplying by 1/ energy12 # NO: "normal ordering "; IO: inverted ordering13 h_vacuum_energy_indep = hamiltonians3nu .\14 hamiltonian_3nu_vacuum_energy_independent( \15 S12_NO_BF , S23_NO_BF ,16 S13_NO_BF , DCP_NO_BF ,17 D21_NO_BF , D31_NO_BF)18

19 # Hamiltonian for oscillations in vacuum20 h_vacuum = np.multiply( 1./energy ,21 h_vacuum_energy_indep)22

23 # Hamiltonian for oscillations in matter24 # VCC_EARTH_CRUST: Potential [eV], density 3 g cm^{-3}25 h_matter = hamiltonians3nu.hamiltonian_3nu_matter( \26 h_vacuum_energy_indep , energy , VCC_EARTH_CRUST)27

28 # Hamiltonian for non -standard interactions29 # EPS_3 is the list of NSI strength parameters30 # [EPS_EE , EPS_EM , EPS_ET , EPS_MM , EPS_MT , EPS_TT]31 h_nsi = hamiltonians3nu.hamiltonian_3nu_nsi( \32 h_vacuum_energy_indep , energy , VCC_EARTH_CRUST ,33 EPS_3)34

35 # Hamiltonian for Lorentz -invariance violation36 # LIV parameters: SXI12 , SXI23 , SXI13 , DXICP , B1 ,37 # B2, B3, LAMBDA38 h_liv = hamiltonians3nu.hamiltonian_3nu_liv( \39 h_vacuum_energy_indep , energy ,40 SXI12 , SXI23 , SXI13 , DXICP , B1, B2, B3, LAMBDA)41

42 # The routine probabilities_3nu computes probabilities43 for h_matrix in [h_vacuum , h_matter , h_nsi , h_liv ]:44 # CONV_KM_TO_INV_EV converts km to eV^{-1}45 Pee , Pem , Pet , Pme , Pmm , Pmt , Pte , Ptm , Ptt = \46 oscprob3nu.probabilities_3nu( \47 h_matrix ,48 baseline*CONV_KM_TO_INV_EV)49

50 print("Pee = %6.5f, Pem = %6.5f, Pet = %6.5f" \51 % (Pee , Pem , Pet))52 print("Pme = %6.5f, Pmm = %6.5f, Pmt = %6.5f" \53 % (Pme , Pmm , Pmt))54 print("Pte = %6.5f, Ptm = %6.5f, Ptt = %6.5f" \55 % (Pte , Ptm , Ptt))56 print()57

58 # This returns:59 # Pee = 0.92768 , Pem = 0.01432 , Pet = 0.0580060 # Pme = 0.04023 , Pmm = 0.37887 , Pmt = 0.5809061 # Pte = 0.03210 , Ptm = 0.60680 , Ptt = 0.3611062

63 # Pee = 0.95262 , Pem = 0.00623 , Pet = 0.0411564 # Pme = 0.02590 , Pmm = 0.37644 , Pmt = 0.5976665 # Pte = 0.02148 , Ptm = 0.61733 , Ptt = 0.3611966

67 # Pee = 0.92494 , Pem = 0.01758 , Pet = 0.0574968 # Pme = 0.03652 , Pmm = 0.32524 , Pmt = 0.6382469 # Pte = 0.03855 , Ptm = 0.65718 , Ptt = 0.3042770

71 # Pee = 0.92721 , Pem = 0.05299 , Pet = 0.0198072 # Pme = 0.05609 , Pmm = 0.25288 , Pmt = 0.6910373 # Pte = 0.01670 , Ptm = 0.69412 , Ptt = 0.28917

Listing 1. Code snippet to use NuOscProbExact to computethree-neutrino oscillation probabilities in vacuum, matter ofconstant density, with non-standard interactions, and withCPT-odd Lorentz-violating background, for fixed neutrinoenergy E = 1 GeV and baseline L = 1300 km. Constants— with variable names in capitals — are pulled from theglobaldefs module; see the main text for their values.

6

0.85

0.90

0.95

1.00P ν

e→ν e

VacuumMatterNSICPT-odd LIV

0.00

0.05

0.10

P νµ→

ν e

100 101

Neutrino energy [GeV]

0.00

0.25

0.50

0.75

P νµ→

ν µ

FIG. 1. Three-neutrino oscillation probabilities Pνe→νe (top),Pνµ→νe (center), and Pνµ→νµ (bottom), computed usingNuOscProbExact [65]. The scenarios shown are for oscillationsin vacuum, in matter of constant density, with non-standardinteractions (NSI), and in a CPT-odd Lorentz-violating back-ground (LIV). In all cases, the baseline is L = 1300 km. Seethe main text for details.

To compute the probabilities in Fig. 1, we fix the mix-ing parameters to their best-fit values provided by therecent NuFit 4.0 global fit to oscillation data [98, 99],assuming normal mass hierarchy, and including Super-Kamiokande atmospheric neutrino data: sin2 θ12 =0.310, sin2 θ23 = 0.582, sin2 θ13 = 0.02240, δCP = 217◦,∆m2

21 = 7.39 · 10−5 eV2, ∆m231 = 2.525 · 10−3 eV2.

B. Oscillations in matter of constant density

When neutrinos propagate in matter, νe and νe scatteron electrons via charged-current interactions. The inter-

actions introduce potentials that shift the energies of theneutrinos. As a result, the values of the mass-squareddifferences and mixing angles in matter differ from theirvalues in vacuum, and depend on the number densityof electrons [100–104]. Computing oscillation probabili-ties in constant matter is integral to long-baseline exper-iments, where neutrinos traverse hundreds of kilometersin the crust of the Earth to reach the detectors [105, 106].

The Hamiltonian that drives oscillations in matter is

Hmatt3 (E) = Hvac

3 (E) + A3 . (14)

The term A3 ≡ diag(VCC, 0, 0) is due to interactions with

matter, where VCC =√

2GFne is the charged-currentpotential and ne is the number density of electrons.

To compute the probabilities for oscillations in mat-ter (and also with non-standard interactions) in Fig. 1,we consider a constant matter density of ρ = 3 g cm−3,the average density of the crust of the Earth [107]. Thenumber density of electrons is ne = Yeρ/[(mp + mn)/2],where mp and mn are the masses of the proton and neu-tron, respectively, and Ye = 0.5 is the average electronfraction in the crust, which is electrically neutral. SeeRefs. [45, 46] for the analytic form of the probabilities inmatter, deduced with the OS method, and Ref. [35] fora related approximation. In long-baseline experiments,even if there are density changes along the trajectory ofthe neutrino beam, using the average density is a goodapproximation [108, 109].

The result above can be extended to the case whereneutrinos traverse multiple slabs of matter, each of con-stant, different density. See, e.g., Refs. [110, 111], for anoverview of this scenario, and Refs. [46, 48, 112] for stud-ies with the OS method. This applies to long-baselineneutrino experiments that consider a non-uniform matterdensity profile [113–117], and to Earth-traversing neu-trinos that cross multiple density layers inside Earth[118]. The probability amplitudes obtained after travers-ing each slab need to be stitched together [8]. If a neu-trino created as να traverses Nslabs slabs of constant-density matter, each of width Lj , then the evolved state is

να({Lj}) =[∏Nslabs

j=1 U(j)3 (Lj)

]να, where U(j)

3 is Eq. (12)

computed using the matter Hamiltonian evaluated withthe matter density of the j-th slab. The final oscillation

probability is Pνα→νβ ({Lj}) = |ν†βνα({Lj})|2.

C. Oscillations with non-standard interactions

Oscillations in matter might receive sub-leading con-tributions due to new neutrino interactions with thefermions of the medium that they propagate in. Theseare known as non-standard interactions (NSI); see Refs.[119–122] for reviews.

In this case, the Hamiltonian is

HNSI3 (E) = Hvac

3 (E) + A3 + V3 , (15)

7

where V3 ≡ VCCε3 is the matter potential due to NSI andε3 is the matrix of NSI strength parameters, i.e.,

ε3 =

εee εeµ εeτε∗eµ εµµ εµτε∗eτ ε∗µτ εττ

. (16)

The parameters εαβ represent the total strength of theNSI between leptons of flavors α and β interacting withthe electrons, u quarks, and d quarks that make upstandard matter. Following Ref. [122], we write εαβ =εeαβ+(2+Yn)εuαβ+(1+2Yn)εdαβ , with the ratio of the num-

ber densities of neutrons to electrons Yn ≡ nn/ne ≈ 1 inthe Earth. In our simplified treatment, we do not con-sider separately interactions with each fermion type oreach chiral projection of the fermion [119–122].

To compute the probabilities for NSI in Fig. 1, we againconsider propagation in the constant-density crust of theEarth, with VCC evaluated as in Section VI B. BecauseNSI have not been observed, we choose arbitrary valuesfor the strength parameters that are allowed at the 2σlevel by a recent global fit to oscillation (LMA solution)plus COHERENT data [122] (see also Refs. [123–126]):εuee = −εueµ = 0.01, εuµµ = 0.2, εueτ = εuµτ = εuττ = 0, andthe same for d quarks. Like in Ref. [122], we set all εeαβ =

0. Thus, for Fig. 1, the NSI parameters in Eq. (16) areεee = −εeµ = 0.06, εµµ = 1.2, and εeτ = εµτ = εττ = 0.

D. Oscillations in a Lorentz-violating background

Lorentz invariance is one of the linchpins of the Stan-dard Model (SM), but is violated in proposed extensions,some related to quantum gravity; see Refs. [127–130] forreviews. There is no experimental evidence for Lorentz-invariance violation (LIV), but there are stringent con-straints on it [25, 131–133]. The effects of LIV are nu-merous, e.g., changes in the properties and rates of pro-cesses of particles versus their anti-particles, introductionof anisotropies in particle angular distributions, and, inthe case of neutrinos, changes to the effective mixing pa-rameters and, thus, to the oscillation probabilities.

To study LIV, we adopt the framework of the StandardModel Extension (SME) [134], an effective field theorythat augments the SM by adding LIV parameters to allsectors, including neutrinos [15, 19–21, 135–137]. In theSME, LIV is suppressed by a high energy scale Λ, stillundetermined. We focus on CPT-odd LIV, where theCPT symmetry is also broken. This is realized by meansof a new vector coupling of neutrinos to a new LIV back-ground field. Unlike the other oscillation cases presentedabove, the contribution of CPT-odd LIV to the Hamilto-nian grows with neutrino energy. This makes high-energyatmospheric and astrophysical neutrinos ideal for testingLIV [15, 21, 133, 138–144].

The Hamiltonian for CPT-odd LIV is [15, 19, 145]

HLIV3 (E) = Hvac

3 (E) +E

ΛR3,ξB3R†3,ξ . (17)

1 import numpy as np2

3 import oscprob3nu4 import hamiltonians3nu5 from globaldefs import *6

7 energy = 1.e9 # Neutrino energy [eV]8 baseline = 1.3e3 # Baseline [km]9

10 h_vacuum_energy_indep = hamiltonians3nu .\11 hamiltonian_3nu_vacuum_energy_independent( \12 S12_NO_BF , S23_NO_BF ,13 S13_NO_BF , DCP_NO_BF ,14 D21_NO_BF , D31_NO_BF)15

16 h_vacuum = np.multiply( 1./energy ,17 h_vacuum_energy_indep)18

19 # The user -supplied routine hamiltonian_mymodel depends20 # on some parameters represented by mymodel_parameters ,21 # and should return a 3x3 array22 h_mymodel = h_vacuum \23 + hamiltonian_mymodel(mymodel_parameters)24

25 Pee , Pem , Pet , Pme , Pmm , Pmt , Pte , Ptm , Ptt = \26 oscprob3nu.probabilities_3nu( \27 h_mymodel ,28 baseline*CONV_KM_TO_INV_EV)

Listing 2. Template to use NuOscProbExact to computeprobabilities for three-neutrino oscillations with an arbitrary,user-supplied Hamiltonian

The second term on the right-hand side is the ef-fective Hamiltonian that introduces LIV. Here, B3 ≡diag(b1, b2, b3), where bi (i = 1, 2, 3) are the eigenvaluesof the LIV operator B3, and R3,ξ is the 3× 3 mixing ma-trix that rotates it into the flavor basis. It has the samestructure as the PMNS matrix, but different values of themixing angles and phase. In general, there is a relativephase between R3,ξ and R3,θ that cannot be rotated away[21, 138]; in our simplified treatment, we set it to zero.

Because LIV has not been observed, the values of theeigenvalues bi and of the LIV mixing parameters ξ areundetermined. Current upper limits [133] set using high-energy atmospheric neutrinos imply that bi/Λ . 10−28

(this is c(4) in the notation of Ref. [133]). The LIV energyscale is believed to be at least Λ = 1 TeV. However, tocompute the probabilities for LIV in Fig. 1 such that theyexhibit features at the lower energies used in the plot, weset artificially high values: b1/Λ = b2/Λ = 10−21 andb3/Λ = 5 · 10−21. For simplicity, we set all mixing anglesto zero, so that R3,ξ = 1.

E. Oscillations with arbitrary Hamiltonians

The usefulness of NuOscProbExact stems in part fromits ability to compute oscillation probabilities for any ar-bitrary time-independent Hamiltonian. Listing 2 showsa code template to compute three-neutrino oscillationprobabilities using a user-supplied Hamiltonian that isadded to the Hamiltonian for oscillations in vacuum.

8

VII. CONCLUSIONS

We have provided the code NuOscProbExact [65] tocompute exact two-neutrino and three-neutrino oscilla-tion probabilities for arbitrary time-independent Hamil-tonians. The code is a numerical implementation of themethod developed by Ohlsson & Snellman [10, 45–47]and uses exponential expansions of SU(2) and SU(3) ma-trices to bypass the diagonalization of the Hamiltonian.It can be used to compute oscillation probabilities inmany often-studied oscillation scenarios, including, butnot limited to, oscillations in vacuum, in constant mat-ter density, non-standard neutrino interactions, and new-physics scenarios, like Lorentz-invariance violation.

In developing NuOscProbExact, our goal was to pro-vide a general-purpose numerical code to compute exactoscillation probabilities that is also lightweight and canbe easily incorporated into diverse oscillation analysesof standard and non-standard oscillations. This is es-pecially useful in the case of three-neutrino oscillations,where analytic expressions of the probabilities are of-

ten unavailable. The code is suitable for exploring wideparameter spaces where approximate expressions of theprobabilities are not available. We provide it with thisapplication in mind.

ACKNOWLEDGMENTS

We thank Carlos Arguelles, Peter Denton, FranciscoDe Zela, Joachim Kopp, Stephen Parke, Serguey Pet-cov, Jordi Salvado, and Irene Tamborra for helpful dis-cussion and suggestions. We thank especially Shirley Liand Tommy Ohlsson for that and for carefully readingthe manuscript, and the latter for pointing out the refer-ences of the Ohlsson-Snellman method. MB is supportedby the Villum Fonden project no. 13164. Early phases ofthis work were supported by a High Energy Physics LatinAmerican European Network (HELEN) grant and by agrant from the Direccion Academica de Investigacion ofthe Pontificia Universidad Catolica del Peru.

∗ mbustamante@nbi.ku.dk; ORCID: 0000-0001-6923-0865

[1] B. Pontecorvo, Sov. Phys. JETP 26, 984 (1968), [Zh.Eksp. Teor. Fiz. 53, 1717 (1967)].

[2] V. D. Barger, K. Whisnant, and R. J. N. Phillips, Phys.Rev. D 22, 1636 (1980).

[3] S. M. Bilenky, Proc. Roy. Soc. Lond. A 460, 403 (2004).[4] G. Fantini, A. Gallo Rosso, F. Vissani, and V. Zema,

Adv. Ser. Direct. High Energy Phys. 28, 37 (2018),arXiv:1802.05781 [hep-ph].

[5] T. Kajita, Rev. Mod. Phys. 88, 030501 (2016).[6] A. B. McDonald, Rev. Mod. Phys. 88, 030502 (2016).[7] B. Kayser, Phys. Rev. D 24, 110 (1981).[8] C. Giunti and C. W. Kim, Fundamentals of Neutrino

Physics and Astrophysics (2007).[9] M. Tanabashi et al. (Particle Data Group), Phys. Rev.

D 98, 030001 (2018).[10] V. D. Barger, K. Whisnant, S. Pakvasa, and R. J. N.

Phillips, Phys. Rev. D 22, 2718 (1980), [, 300 (1980)].[11] S. T. Petcov, Phys. Lett. B 200, 373 (1988).[12] H. W. Zaglauer and K. H. Schwarzer, Z. Phys. C 40,

273 (1988).[13] E. Torrente Lujan, Phys. Rev. D 53, 4030 (1996),

arXiv:hep-ph/9505209 [hep-ph].[14] A. B. Balantekin, Phys. Rev. D 58, 013001 (1998),

arXiv:hep-ph/9712304 [hep-ph].[15] S. R. Coleman and S. L. Glashow, Phys. Rev. D 59,

116008 (1999), arXiv:hep-ph/9812418 [hep-ph].[16] A. M. Gago, M. M. Guzzo, H. Nunokawa, W. J. C.

Teves, and R. Zukanovich Funchal, Phys. Rev. D 64,073003 (2001), arXiv:hep-ph/0105196 [hep-ph].

[17] K. Kimura, A. Takamura, and H. Yokomakura, Phys.Rev. D 66, 073005 (2002), arXiv:hep-ph/0205295 [hep-ph].

[18] P. F. Harrison, W. G. Scott, and T. J. Weiler, Phys.Lett. B 565, 159 (2003), arXiv:hep-ph/0305175 [hep-

ph].[19] V. A. Kostelecky and M. Mewes, Phys. Rev. D 70,

031902 (2004), arXiv:hep-ph/0308300 [hep-ph].[20] V. A. Kostelecky and M. Mewes, Phys. Rev. D 69,

016005 (2004), arXiv:hep-ph/0309025 [hep-ph].[21] M. C. Gonzalez-Garcıa and M. Maltoni, Phys. Rev. D

70, 033010 (2004), arXiv:hep-ph/0404085 [hep-ph].[22] M. Blennow and T. Ohlsson, J. Math. Phys. 45, 4053

(2004), arXiv:hep-ph/0405033 [hep-ph].[23] D. Meloni, T. Ohlsson, and H. Zhang, JHEP 04, 033

(2009), arXiv:0901.1784 [hep-ph].[24] S. Ando, M. Kamionkowski, and I. Mocioiu, Phys. Rev.

D 80, 123522 (2009), arXiv:0910.4391 [hep-ph].[25] K. Abe et al. (Super-Kamiokande), Phys. Rev. D 91,

052003 (2015), arXiv:1410.4267 [hep-ex].[26] L. J. Flores and O. G. Miranda, Phys. Rev. D 93, 033009

(2016), arXiv:1511.03343 [hep-ph].[27] A. Popov and A. Studenikin, (2018), arXiv:1803.05755

[hep-ph].[28] A. M. Gago, E. M. Santos, W. J. C. Teves, and

R. Zukanovich Funchal, (2002), arXiv:hep-ph/0208166[hep-ph].

[29] M. Blennow and T. Ohlsson, Phys. Rev. D 78, 093002(2008), arXiv:0805.2301 [hep-ph].

[30] J. S. Dıaz, V. A. Kostelecky, and M. Mewes, Phys. Rev.D 80, 076007 (2009), arXiv:0908.1401 [hep-ph].

[31] M. Esteves Chaves, D. R. Gratieri, and O. L. G. Peres,(2018), arXiv:1810.04979 [hep-ph].

[32] S. T. Petcov and S. Toshev, Phys. Lett. B 187, 120(1987).

[33] K. Hirota, Phys. Rev. D 57, 3140 (1998).[34] A. Cervera, A. Donini, M. B. Gavela, J. J. Gomez Ca-

denas, P. Hernandez, O. Mena, and S. Rigolin, Nucl.Phys. B 579, 17 (2000), [Erratum: Nucl. Phys. B 593,731 (2001)], arXiv:hep-ph/0002108 [hep-ph].

[35] M. Freund, Phys. Rev. D 64, 053003 (2001), arXiv:hep-

9

ph/0103300 [hep-ph].[36] E. K. Akhmedov, R. Johansson, M. Lindner, T. Ohls-

son, and T. Schwetz, JHEP 04, 078 (2004), arXiv:hep-ph/0402175 [hep-ph].

[37] A. Friedland and C. Lunardini, Phys. Rev. D 74, 033012(2006), arXiv:hep-ph/0606101 [hep-ph].

[38] A. N. Ioannisian and A. Yu. Smirnov, Nucl. Phys. B816, 94 (2009), arXiv:0803.1967 [hep-ph].

[39] A. D. Supanitsky, J. C. D’Olivo, and G. Medina-Tanco,Phys. Rev. D 78, 045024 (2008), arXiv:0804.1105 [astro-ph].

[40] H. Minakata and S. J. Parke, JHEP 01, 180 (2016),arXiv:1505.01826 [hep-ph].

[41] Y.-F. Li, J. Zhang, S. Zhou, and J.-y. Zhu, JHEP 12,109 (2016), arXiv:1610.04133 [hep-ph].

[42] P. B. Denton, H. Minakata, and S. J. Parke, JHEP 06,051 (2016), arXiv:1604.08167 [hep-ph].

[43] G. Barenboim, P. B. Denton, S. J. Parke, and C. A.Ternes, Phys. Lett. B 791, 351 (2019), arXiv:1902.00517[hep-ph].

[44] P. B. Denton, S. J. Parke, and X. Zhang, Phys. Rev. D98, 033001 (2018), arXiv:1806.01277 [hep-ph].

[45] T. Ohlsson and H. Snellman, J. Math. Phys. 41, 2768(2000), [Erratum: J. Math. Phys. 42, 2345 (2001)],arXiv:hep-ph/9910546 [hep-ph].

[46] T. Ohlsson and H. Snellman, Phys. Lett. B 474, 153(2000), [Erratum: Phys. Lett. B 480, 419 (2000)],arXiv:hep-ph/9912295 [hep-ph].

[47] T. Ohlsson, Dynamics of Quarks and Leptons: Theo-retical Studies of Baryons and Neutrinos, Ph.D. thesis,Royal Inst. Tech., Stockholm (2000).

[48] T. Ohlsson and H. Snellman, Eur. Phys. J. C 20, 507(2001), arXiv:hep-ph/0103252 [hep-ph].

[49] K. N. Abazajian et al., (2012), arXiv:1204.5379 [hep-ph].

[50] A. Palazzo, Mod. Phys. Lett. A 28, 1330004 (2013),arXiv:1302.1102 [hep-ph].

[51] S. Gariazzo, C. Giunti, M. Laveder, Y. F. Li,and E. M. Zavanin, J. Phys. G 43, 033001 (2016),arXiv:1507.08204 [hep-ph].

[52] M. Dentler, A. Hernandez-Cabezudo, J. Kopp, P. A. N.Machado, M. Maltoni, I. Martınez-Soler, andT. Schwetz, JHEP 08, 010 (2018), arXiv:1803.10661[hep-ph].

[53] C. Giunti and T. Lasserre, (2019), arXiv:1901.08330[hep-ph].

[54] A. S. Joshipura, E. Masso, and S. Mohanty, Phys. Rev.D 66, 113008 (2002), arXiv:hep-ph/0203181 [hep-ph].

[55] J. F. Beacom and N. F. Bell, Phys. Rev. D 65, 113009(2002), arXiv:hep-ph/0204111 [hep-ph].

[56] J. F. Beacom, N. F. Bell, D. Hooper, S. Pakvasa,and T. J. Weiler, Phys. Rev. Lett. 90, 181301 (2003),arXiv:hep-ph/0211305 [hep-ph].

[57] M. C. Gonzalez-Garcıa and M. Maltoni, Phys. Lett. B663, 405 (2008), arXiv:0802.3699 [hep-ph].

[58] P. Baerwald, M. Bustamante, and W. Winter, JCAP1210, 020 (2012), arXiv:1208.4600 [astro-ph.CO].

[59] J. M. Berryman, A. de Gouvea, D. Hernandez, andR. L. N. Oliveira, Phys. Lett. B 742, 74 (2015),arXiv:1407.6631 [hep-ph].

[60] F. Benatti and R. Floreanini, JHEP 02, 032 (2000),arXiv:hep-ph/0002221 [hep-ph].

[61] E. Lisi, A. Marrone, and D. Montanino, Phys. Rev.

Lett. 85, 1166 (2000), arXiv:hep-ph/0002053 [hep-ph].[62] T. Ohlsson, Phys. Lett. B 502, 159 (2001), arXiv:hep-

ph/0012272 [hep-ph].[63] J. A. Carpio, E. Massoni, and A. M. Gago, Phys. Rev.

D 97, 115017 (2018), arXiv:1711.03680 [hep-ph].[64] P. Coloma, J. Lopez-Pavon, I. Martınez-Soler, and

H. Nunokawa, Eur. Phys. J. C 78, 614 (2018),arXiv:1803.04438 [hep-ph].

[65] M. Bustamante, https://github.com/mbustama/

NuOscProbExact, NuOscProbExact.[66] B. H. J. McKellar and M. J. Thomson, Phys. Rev. D

49, 2710 (1994).[67] N. F. Bell, R. R. Volkas, and Y. Y. Y. Wong, Phys. Rev.

D 59, 113001 (1999), arXiv:hep-ph/9809363 [hep-ph].[68] S. Hannestad, Phys. Rev. D 65, 083006 (2002),

arXiv:astro-ph/0111423 [astro-ph].[69] Y. Y. Y. Wong, Phys. Rev. D 66, 025015 (2002),

arXiv:hep-ph/0203180 [hep-ph].[70] S. Hannestad, I. Tamborra, and T. Tram, JCAP 1207,

025 (2012), arXiv:1204.5861 [astro-ph.CO].[71] D. Boriero, D. J. Schwarz, and H. Velten, (2017),

arXiv:1704.06139 [astro-ph.CO].[72] J. T. Pantaleone, Phys. Lett. B 342, 250 (1995),

arXiv:astro-ph/9405008 [astro-ph].[73] G. Sigl, Phys. Rev. D 51, 4035 (1995), arXiv:astro-

ph/9410094 [astro-ph].[74] S. Pastor and G. Raffelt, Phys. Rev. Lett. 89, 191101

(2002), arXiv:astro-ph/0207281 [astro-ph].[75] A. B. Balantekin and H. Yuksel, New J. Phys. 7, 51

(2005), arXiv:astro-ph/0411159 [astro-ph].[76] G. L. Fogli, E. Lisi, A. Marrone, and A. Mirizzi, JCAP

0712, 010 (2007), arXiv:0707.1998 [hep-ph].[77] B. Dasgupta and A. Dighe, Phys. Rev. D 77, 113002

(2008), arXiv:0712.3798 [hep-ph].[78] H. Duan, G. M. Fuller, and Y.-Z. Qian, Ann. Rev. Nucl.

Part. Sci. 60, 569 (2010), arXiv:1001.2799 [hep-ph].[79] A. Mirizzi, I. Tamborra, H.-T. Janka, N. Sa-

viano, K. Scholberg, R. Bollig, L. Hudepohl, andS. Chakraborty, Riv. Nuovo Cim. 39, 1 (2016),arXiv:1508.00785 [astro-ph.HE].

[80] A. J. MacFarlane, A. T. Sudbery, and P. H. Weisz,Commun. Math. Phys. 11, 77 (1968).

[81] T. L. Curtright and C. K. Zachos, Rept. Math. Phys.76, 401 (2015), arXiv:1508.00868 [math.RT].

[82] T. S. Van Kortryk, J. Math. Phys. 57, 021701 (2016),arXiv:1508.05859 [math.RT].

[83] P. Huber, J. Kopp, M. Lindner, and W. Winter, https://www.mpi-hd.mpg.de/personalhomes/globes/,GLoBES.

[84] M. Wallraff and C. Wiebusch, https://nucraft.

hepforge.org/, nuCraft.[85] C. A. Arguelles Delgado, J. Salvado, and C. N.

Weaver, https://github.com/arguelles/nuSQuIDS,NuSQuIDS.

[86] R. S.-K. c. Wendell, http://webhome.phy.duke.edu/

~raw22/public/Prob3++/, Prob3++.[87] C. A. Arguelles Delgado, J. Salvado, and C. N.

Weaver, Comput. Phys. Commun. 196, 569 (2015),arXiv:1412.3832 [hep-ph].

[88] D. Kusnezov, J. Mat. Phys. 36, 898 (1995).[89] Y. Kamo, S. Yajima, Y. Higasida, S.-I. Kubota,

S. Tokuo, and J.-I. Ichihara, Eur. Phys. J. C 28, 211(2003), arXiv:hep-ph/0209097 [hep-ph].

[90] J. Salvado, Private communication.

10

[91] J. Kopp, Int. J. Mod. Phys. D 19, 523 (2008),arXiv:physics/0610206 [physics].

[92] P. Huber, J. Kopp, M. Lindner, M. Rolinec, andW. Winter, Comput. Phys. Commun. 177, 432 (2007),arXiv:hep-ph/0701187 [hep-ph].

[93] Y. Lehrer-Ilamed, Proc. Camb. Phil. Soc 60, 61 (1964).[94] S. P. Rosen, J. Math. Phys. 12, 673 (1971).[95] A. J. Torruella, J. Mat. Phys. 16, 1637 (1975).[96] J. J. Sylvester, Phil. Mag. 16, 267 (1883).[97] B. Abi et al. (DUNE), (2018), arXiv:1807.10334

[physics.ins-det].

[98] I. Esteban, M. C. Gonzalez-Garcıa, A. Hernandez-Cabezudo, M. Maltoni, and T. Schwetz, JHEP 01, 106(2019), arXiv:1811.05487 [hep-ph].

[99] NuFit, “Three-neutrino fit based on data available inNovember 2018,” http://www.nu-fit.org/.

[100] L. Wolfenstein, Phys. Rev. D 17, 2369 (1978).[101] S. P. Mikheyev and A. Yu. Smirnov, Sov. J. Nucl. Phys.

42, 913 (1985).[102] H. A. Bethe, Phys. Rev. Lett. 56, 1305 (1986).[103] S. J. Parke, Solar Neutrinos: An Overview, Phys. Rev.

Lett. 57, 1275 (1986).[104] S. P. Rosen and J. M. Gelb, Phys. Rev. D 34, 969 (1986).[105] G. J. Feldman, J. Hartnell, and T. Kobayashi,

Adv. High Energy Phys. 2013, 475749 (2013),arXiv:1210.1778 [hep-ex].

[106] M. V. Diwan, V. Galymov, X. Qian, and A. Rub-bia, Ann. Rev. Nucl. Part. Sci. 66, 47 (2016),arXiv:1608.06237 [hep-ex].

[107] A. M. Dziewonski and D. L. Anderson, Phys. EarthPlanet. Interiors 25, 297 (1981).

[108] S. T. Petcov, Phys. Lett. B 434, 321 (1998), arXiv:hep-ph/9805262 [hep-ph].

[109] M. Freund, M. Lindner, S. T. Petcov, and A. Ro-manino, Nucl. Phys. B 578, 27 (2000), arXiv:hep-ph/9912457 [hep-ph].

[110] M. V. Chizhov and S. T. Petcov, Phys. Rev. Lett. 83,1096 (1999), arXiv:hep-ph/9903399 [hep-ph].

[111] M. V. Chizhov and S. T. Petcov, Phys. Rev. D 63,073003 (2001), arXiv:hep-ph/9903424 [hep-ph].

[112] K. M. Merfeld and D. C. Latimer, Phys. Rev. C 90,065502 (2014), arXiv:1412.2728 [hep-ph].

[113] B. Jacobsson, T. Ohlsson, H. Snellman, and W. Winter,Phys. Lett. B 532, 259 (2002), arXiv:hep-ph/0112138[hep-ph].

[114] T. Ohlsson and W. Winter, Phys. Rev. D 68, 073007(2003), arXiv:hep-ph/0307178 [hep-ph].

[115] C. A. Arguelles, M. Bustamante, and A. M. Gago, Mod.Phys. Lett. A 30, 1550146 (2015), arXiv:1201.6080 [hep-ph].

[116] B. Roe, Phys. Rev. D 95, 113004 (2017),arXiv:1707.02322 [hep-ex].

[117] K. J. Kelly and S. J. Parke, Phys. Rev. D 98, 015025(2018), arXiv:1802.06784 [hep-ph].

[118] M. Freund and T. Ohlsson, Mod. Phys. Lett. A 15, 867(2000), arXiv:hep-ph/9909501 [hep-ph].

[119] T. Ohlsson, Rept. Prog. Phys. 76, 044201 (2013),arXiv:1209.2710 [hep-ph].

[120] O. G. Miranda and H. Nunokawa, New J. Phys. 17,095002 (2015), arXiv:1505.06254 [hep-ph].

[121] Y. Farzan and M. Tortola, Front. in Phys. 6, 10 (2018),arXiv:1710.09360 [hep-ph].

[122] I. Esteban, M. C. Gonzalez-Garcıa, M. Maltoni,

I. Martınez-Soler, and J. Salvado, JHEP 08, 180 (2018),arXiv:1805.04530 [hep-ph].

[123] G. Mitsuka et al. (Super-Kamiokande), Phys. Rev. D84, 113008 (2011), arXiv:1109.1889 [hep-ex].

[124] A. Esmaili and A. Yu. Smirnov, JHEP 06, 026 (2013),arXiv:1304.1042 [hep-ph].

[125] J. Salvado, O. Mena, S. Palomares-Ruiz, and N. Rius,JHEP 01, 141 (2017), arXiv:1609.03450 [hep-ph].

[126] M. G. Aartsen et al. (IceCube), Phys. Rev. D 97, 072009(2018), arXiv:1709.07079 [hep-ex].

[127] N. E. Mavromatos, Planck scale effects in astro-physics and cosmology. Proceedings, 40th Karpacs Win-ter School, Ladek Zdroj, Poland, February 4-14, 2004,Lect. Notes Phys. 669, 245 (2005), arXiv:gr-qc/0407005[gr-qc].

[128] S. Liberati and L. Maccione, Ann. Rev. Nucl. Part. Sci.59, 245 (2009), arXiv:0906.0681 [astro-ph.HE].

[129] S. Liberati, Class. Quant. Grav. 30, 133001 (2013),arXiv:1304.5795 [gr-qc].

[130] J. D. Tasson, Rept. Prog. Phys. 77, 062901 (2014),arXiv:1403.7785 [hep-ph].

[131] V. A. Kostelecky and N. Russell, Rev. Mod. Phys. 83,11 (2011), arXiv:0801.0287 [hep-ph].

[132] R. Abbasi et al. (IceCube), Phys. Rev. D 82, 112003(2010), arXiv:1010.4096 [astro-ph.HE].

[133] M. G. Aartsen et al. (IceCube), Nature Phys. 14, 961(2018), arXiv:1709.03434 [hep-ex].

[134] D. Colladay and V. A. Kostelecky, Phys. Rev. D 58,116002 (1998), arXiv:hep-ph/9809521 [hep-ph].

[135] D. Hooper, D. Morgan, and E. Winstanley, Phys. Rev.D 72, 065009 (2005), arXiv:hep-ph/0506091 [hep-ph].

[136] A. Kostelecky and M. Mewes, Phys. Rev. D 85, 096005(2012), arXiv:1112.6395 [hep-ph].

[137] J. S. Dıaz, Adv. High Energy Phys. 2014, 962410(2014), arXiv:1406.6838 [hep-ph].

[138] M. C. Gonzalez-Garcıa, F. Halzen, and M. Maltoni,Phys. Rev. D 71, 093010 (2005), arXiv:hep-ph/0502223[hep-ph].

[139] L. A. Anchordoqui, H. Goldberg, M. C. Gonzalez-Garcıa, F. Halzen, D. Hooper, S. Sarkar, and T. J.Weiler, Phys. Rev. D 72, 065019 (2005), arXiv:hep-ph/0506168 [hep-ph].

[140] C. A. Arguelles, T. Katori, and J. Salvado, Phys. Rev.Lett. 115, 161303 (2015), arXiv:1506.02043 [hep-ph].

[141] M. Bustamante, J. F. Beacom, and W. Winter, Phys.Rev. Lett. 115, 161302 (2015), arXiv:1506.02645 [astro-ph.HE].

[142] R. W. Rasmussen, L. Lechner, M. Ackermann,M. Kowalski, and W. Winter, Phys. Rev. D 96, 083018(2017), arXiv:1707.07684 [hep-ph].

[143] M. Ahlers, K. Helbing, and C. Perez de los Heros,Eur. Phys. J. C 78, 924 (2018), arXiv:1806.05696 [astro-ph.HE].

[144] M. Ackermann et al., (2019), arXiv:1903.04333 [astro-ph.HE].

[145] A. Dighe and S. Ray, Phys. Rev. D 78, 036002 (2008),arXiv:0802.0121 [hep-ph].

11

Appendix A: Pauli and Gell-Mann matrices

For completeness, and to avoid any ambiguity in themethod presented in the main text, we show here explic-itly all the Pauli and Gell-Mann matrices.

The three Pauli matrices σk = σk are:

σ1 =

(0 11 0

), σ2 =

(0 −ii 0

), σ3 =

(1 00 −1

).

The eight Gell-Mann matrices λk = λk are:

λ1 =

0 1 01 0 00 0 0

, λ2 =

0 −i 0i 0 00 0 0

,

λ3 =

1 0 00 −1 00 0 0

, λ4 =

0 0 10 0 01 0 0

,

λ5 =

0 0 −i0 0 0i 0 0

, λ6 =

0 0 00 0 10 1 0

,

λ7 =

0 0 00 0 −i0 i 0

, λ8 =1√3

1 0 00 1 00 0 −2

.

Appendix B: Derivation of Eq. (6)

We proceed by computing the survival probabilityPνα→να = |ν†αU2(L)να|2. We start by operating on ναwith the linear combination hkσ

k that appears in the ex-pansion of the time-evolution operator U2(L), Eq. (5),

i.e., (hkσk)να. In matrix form, with να = (1 0)

T, this is(

h3 h1 − ih2h1 + ih2 −h3

)(10

)=

(h3

h1 + ih2

).

So, using Eq. (5) for U2(L), the survival probability am-plitude is

ν†αU2(L)να = cos (|h|L)− i h3|h| sin (|h|L) ,

where the coefficient h3 = [(H2)11 − (H2)22]/2; see TableI. Since H2 is Hermitian, its diagonal elements are real,and, hence, h3 is real. Because of this, we can write thesurvival probability |ν†αU2(L)να|2 as

Pνα→να(L) = cos2(|h|L) +|h3|2|h|2 sin2(|h|L) .

Now, because |h3|2 = |h|2 − |h1|2 − |h2|2, this becomes

Pνα→να(L) = 1− |h1|2 + |h2|2|h|2 sin2(|h|L) .

Because of the conservation of probability, Pνα→νβ = 1−Pνα→να , with α 6= β, which gives Eq. (6).

Appendix C: Two-flavor oscillations in vacuum

As a simple example and cross-check of the OS method,we consider two-flavor oscillations in vacuum. These aredriven by the mass-squared difference between two masseigenstates ν1 and ν2, with masses m1 and m2, out ofwhich the flavor states νe and νµ are constructed (or νµand ντ ). The spaces of flavor and mass states are con-nected by a unitary rotation that is parametrized by amixing angle θ, i.e.,

R2,θ =

(cos θ sin θ− sin θ cos θ

). (C1)

In this case, the Hamiltonian in the flavor basis, for aneutrino of energy E, is

Hvac2 (E) =

1

2ER2,θM

22R†2,θ , (C2)

where M22 ≡ diag(∆m2/2,−∆m2/2) is the mass matrix,

and ∆m2 ≡ m22 −m2

1. Using Table I, we identify

|h1|2 =∆m2

2Esin2 (2θ) , |h3|2 =

∆m2

2Ecos2 (2θ) ,

and |h2|2 = 0, so that |h1|2/|h|2 = sin2 (2θ). FromEq. (6), the probability is

P vacνe→νµ(E,L) = sin2 (2θ) sin2

(∆m2

4EL

),

which is the standard expression for two-neutrino oscil-lations in vacuum; see, e.g., Refs. [8, 9].

Appendix D: Derivation of Eq. (11)

First, we write

uk = −1

2

3∑m=1

eiφm∂φm

∂ (−hkL),

where

∂φm∂ (−hkL)

=2[φm (−hkL) + dijk

(−hiL

) (−hjL

)]3φ2m − I2

.

This can be expanded as

uk = −1

2Tr

∂

∂ (−hkL)e−ihkλ

kL

= −1

2

∂

∂ (−hkL)Tr U3

= −1

2

3∑m=1

eiφm∂φm

∂ (−hkL).

Thus, the coefficients uk can be written as

uk = −Lxhk + L2y (h ∗ h)k , (D1)

12

0.00

0.25

0.50

0.75

1.00P ν

e→ν e

VacuumMatterNSICPT-odd LIV

0.00

0.25

0.50

0.75

P νµ→

ν e

100 101

Neutrino energy [GeV]

0.00

0.25

0.50

0.75

P νµ→

ν µ

FIG. 2. Two-neutrino oscillation probabilities Pνe→νe (top),Pνµ→νe (center), and Pνµ→νµ (bottom), computed using themethod presented here, via NuOscProbExact [65]. This figureis the two-neutrino counterpart of Fig. 1. See the main textand Appendix E for details.

where

x = − 1

L

3∑m=1

ψmeiLψm

3ψ2m − |h|2

, (D2)

y = − 1

L2

3∑m=1

eiLψm

3ψ2m − |h|2

. (D3)

Inserting Eqs. (D2) and (D3) into Eq. (D1) results inEq. (11) in the main text.

Appendix E: Sample two-neutrino Hamiltonians

Here we present the two-neutrino Hamiltonians in-cluded as examples in NuOscProbExact. These are thetwo-neutrino counterparts of the three-neutrino examplespresented in Section VI; we refer to that section for a de-scription of each scenario. All of the Hamiltonians beloware written in the flavor basis.

Figure 2 shows the probabilities Pνe→νe , Pνµ→νe , andPνµ→νµ for the same four scenarios as in Fig. 1, computedusing NuOscProbExact [65]. Again, we set the baselineto L = 1300 km. The parameters and their values usedin each example case are introduced below; they are aselection of the ones used in Fig. 1.

For oscillations in vacuum, we use Eq. (C2), i.e.,

Hvac2 (E) =

1

2ER2,θM

22R†2,θ ,

where the rotation matrix R2,θ is given in Eq. (C1), interms of the mixing angle θ. In Fig. 2, we set θ and ∆m2,respectively, to the values of θ12 and ∆m2

21 used in Fig. 1.For oscillations in matter, we use

Hmatt2 (E) = Hvac

2 (E) + A2 ,

where A2 ≡ diag(VCC, 0) and VCC is defined as before. InFig. 2, we set ρ = 3 g cm−3.

For oscillations in matter with non-standard interac-tions, we use [21, 138]

HNSI2 (E) = Hvac

2 (E) + A2 + V2 ,

where V2 ≡ VCCε2 and the matrix of NSI strength pa-rameters is

ε2 =

(εee εeµε∗eµ εµµ

).

In Fig. 2, we set εee = −εeµ = 0.06 and εµµ = 1.2.For oscillations in a CPT-odd Lorentz-violating back-

ground, we use [21, 138]

HLIV2 (E) = Hvac

2 (E) +E

ΛR2,ξB2R†2,ξ ,

where U2,ξ is a 2× 2 rotation matrix, like Eq. (C1), butevaluated at a different mixing angle ξ. In Fig. 2, we setR2,ξ = 1, b1/Λ = 10−21, and b2/Λ = 5 · 10−21.