The Astrophysical Journal, 918:15 (8pp), 2021 September 1 https://doi.org/10.

3847/1538-4357/ac0970
© 2021. The American Astronomical Society. All rights reserved.

On the (Mis)Interpretation of the Scattering Polarization Signatures in the Ca II 8542 Å

Line through Spectral Line Inversions
Rebecca Centeno1 , Jaime de la Cruz Rodríguez2 , and Tanausú del Pino Alemán3
1
High Altitude Observatory (NCAR), 3080 Center Green Dr., Boulder, CO, USA; [email protected]
2
Institute for Solar Physics, Dept. of Astronomy, Stockholm University, AlbaNova University Centre, SE-10691 Stockholm, Sweden
3
Instituto de Astrofísica de Canarias, C. Vía Láctea s/n, La Laguna, Tenerife, Spain
Received 2021 April 5; revised 2021 June 4; accepted 2021 June 7; published 2021 August 30

Abstract
Scattering polarization tends to dominate the linear polarization signals of the Ca II 8542 Å line in weakly
magnetized areas (B  100 G), especially when the observing geometry is close to the limb. In this paper, we
evaluate the degree of applicability of existing non-LTE spectral line inversion codes (which assume that the
spectral line polarization is due to the Zeeman effect only) at inferring the magnetic field vector and, particularly,
its transverse component. To this end, we use the inversion code STiC to extract the strength and orientation of the
magnetic field from synthetic spectropolarimetric data generated with the Hanle-RT code. The latter accounts for
the generation of polarization through scattering processes as well as the joint actions of the Hanle and the Zeeman
effects. We find that, when the transverse component of the field is stronger than ∼80 G, the inversion code is able
to retrieve accurate estimates of the transverse field strength as well as its azimuth in the plane of the sky. Below
this threshold, the scattering polarization signatures become the major contributors to the linear polarization signals
and often mislead the inversion code into severely over- or underestimating the field strength. Since the line-of-
sight component of the field is derived from the circular polarization signal, which is not affected by atomic
alignment, the corresponding inferences are always good.
Unified Astronomy Thesaurus concepts: Solar magnetic fields (1503); Radiative transfer (1335); Starlight
polarization (1571)

1. Introduction between the Zeeman splitting and the line widths, instead of
linearly, as the circular polarization features do.
Spectropolarimetry, the science of measuring the intensity
But even in the absence of magnetic fields, there are other
and polarization of light as a function of wavelength, is a very
mechanisms that can generate polarization in spectral lines.
powerful tool for remote sensing of thermodynamic and When the magnetic sub-levels of a given energy level are
magnetic properties of astrophysical plasmas. unevenly populated (i.e., atomic-level polarization), this results
There are several physical mechanisms that can generate in an imbalance between the number of π and σ transitions per
polarization in spectral lines. They are all related to the lack of unit volume and time, which leads to emergent polarization
symmetry in the absorption and excitation mechanisms of the signatures without the need to invoke a wavelength shift (see
atom. The mere presence of spectral line polarization tells us Trujillo Bueno 2001 for a short introductory review to
something about the conditions of the plasma where the light scattering polarization and the Hanle effect and Landi
was emitted or absorbed. The subtleties of the intensity and Degl’Innocenti & Landolfi 2004 for a detailed explanation of
polarization spectra encode a wealth of information about how the origin of polarization in spectral lines). The population
the atoms were excited and thus about the conditions imbalance and quantum coherence among the degenerate
dominating the atmosphere where the observed radiation was magnetic sub-levels, known as atomic-level polarization, lead
generated. to an imbalance between the rates of different |ΔM| transitions,
A spectral line arises from the absorption or emission of a which, in turn, produces emergent linear polarization signals,
photon when an electron transitions from one bound atomic (or commonly referred to as “scattering polarization.”
molecular) energy level to another. These energy levels are, in In the Sun’s atmosphere, the most common mechanism for
general, degenerate. In the presence of an external magnetic creating atomic-level polarization is the so-called optical
field, these magnetic sub-levels (represented by the magnetic pumping, that is, the anisotropic illumination of the atoms
quantum number M) split and the degeneracy is broken, leading leads to selective emission or absorption of the different
to a wavelength shift between the π (ΔM = 0) and σ spectral line components and to scattering linear polarization,
(ΔM = ± 1) components of the spectral line. The splitting of without the need for a magnetic field. Elastic collisions have a
strong depolarizing effect, and tend to equalize the populations
the energy levels is, in general, proportional to the magnetic
of the magnetic sub-levels. So for this mechanism to be
field strength, and this polarization mechanism is commonly
effective, the radiative transitions have to dominate the atomic
known as the Zeeman effect. If the magnetic field is very weak excitation and deexcitation processes. In the rarefied environ-
(relative to the width of the spectral line), or absent, the ment of the chromosphere, collisions are infrequent and the
wavelength shift of the magnetic sub-levels is very small and radiation is weakly coupled to the local thermodynamic
the ensuing polarization signatures are negligible. This is conditions (this is commonly referred to as non-local thermo-
especially true for the linear polarization signatures of dynamic equilibrium, or non-LTE). This sets a perfect stage for
chromospheric lines, which scale with the square of the ratio optical pumping to act effectively, creating atomic-level

1
The Astrophysical Journal, 918:15 (8pp), 2021 September 1 Centeno, de la Cruz Rodríguez, & del Pino Alemán

polarization that leads to the generation of polarization To date, the only spectral line inversion codes that account
signatures in spectral lines in the absence of an external for the physics of scattering polarization and the combined
magnetic or electric field. In chromospheric quiet-Sun regions, action of the Hanle and Zeeman effects are limited to the
where the magnetic fields are weak and the Zeeman effect interpretation of some neutral He lines, where computing the
generates small polarization signatures, the linear polarization radiative transfer in a slab of constant properties can be a
of chromospheric lines is typically dominated by scattering suitable approximation (see, for instance, Asensio Ramos et al.
processes (Manso Sainz & Trujillo Bueno 2003, 2010). 2008).
The Hanle effect is the relaxation of the atomic-level This work attempts to evaluate the inaccuracy in the
coherences due to the presence of a magnetic field. When a chromospheric magnetic field values retrieved from non-LTE
magnetic field is present, the emergent scattering polarization inversions of the Ca II 8542 Å line when neglecting the
signals will change with respect to the zero-field case, leaving a scattering polarization signals and their modification due to the
measurable imprint on the polarization spectrum. The Hanle Hanle effect. Section 2 describes the computer codes and
effect is typically sensitive to relatively weak magnetic field numerical setups used to synthesize and invert the spectral
strengths (depending on the spectral line) and allows us to lines. In Section 3 we present the analysis of the inversion
access regimes where the Zeeman effect does not have useful results, followed by some brief thoughts and conclusions in
Section 4.
diagnostic capabilities.
The Ca II 8542 Å line has been deemed one of the most
promising lines for chromospheric magnetic field diagnostics 2. Synthesis and Inversion of Spectral Lines
(Quintero Noda et al. 2016, 2017; Lagg et al. 2017) due to its Purely a numerical experiment, this work aims at quantifying
accessibility from ground-based observations and the relative the biases incurred by non-LTE spectral line inversion codes
ease of its interpretation—that is, when modeling it, it is safe to when analyzing Ca II 8542 Å Stokes spectra. Currently, the
assume complete redistribution (Uitenbroek 1989) and to non-LTE spectral line inversion codes capable of interpreting
ignore non-equilibrium ionization effects (Wedemeyer-Bohm the spectral line radiation in the Ca II IR triplet do not account
& Carlsson 2011) and full 3D radiative transfer computations for the generation of polarization induced by scattering
are unnecessary in certain scenarios (de la Cruz Rodríguez et al. processes and its modification due to the Hanle effect. There
2012). It is of interest to note that neglecting 3D radiative are, however, forward-modeling radiative transfer software
transfer effects can have a large impact on the scattering packages that account for the subtle quantum-mechanical
polarization signals of this and other spectral lines (see Jaume effects responsible for these processes, which can realistically
Bestard et al. 2021 for an analysis of their impacts on Ca I model the polarization signatures of Ca II 8542 Å.
4227 Å). The presence of horizontal inhomogeneities in the Hanle-RT (see, for instance, del Pino Alemán et al. 2016) is
physical properties of the Sun’s atmosphere can also break the a 1.5D radiative transfer code able to solve the polarized
axial symmetry of the radiation field, fueling an additional radiation transfer problem in a plane-parallel geometry under
source of atomic-level polarization. Štěpán & Trujillo Bueno non-LTE conditions for a multi-term or multi-level atom in the
(2016) quantified the error in the emergent linear polarization presence of an arbitrary magnetic field, taking into account the
of Ca II 8542 Å when neglecting said horizontal inhomogene- effects of partial frequency redistribution (PRD) as well as the
ities, and concluded that it could be as large as the linear contribution of inelastic and elastic collisions. The rate of
polarization amplitudes themselves. This study was carried out depolarizing elastic collisions with neutral hydrogen are the
on the polarization signals emerging from a high-resolution result of ab initio quantum-mechanical calculations by Manso
quiet-Sun radiation magneto-hydrodynamic simulation and Sainz et al. (2014). Spectral line polarization is generated by
probably represents an upper bound for these scattering means of scattering processes as well as by the combined action
polarization signals. Spatial degradation of the spectral profiles of the Hanle and Zeeman effects.
and stronger magnetic fields will likely render these 3D
radiative transfer effects less important. 2.1. Synthesis
Despite it being commonly used to extract information about
The Hanle-RT code was used to synthesize the Stokes
the chromospheric magnetic field, whether the interpretation is
profiles of the Ca II 8542 Å line in a number of magnetic field
done by applying the weak-field approximation (WFA, see, for
scenarios and observing geometries. The spectra were synthe-
instance Asensio Ramos et al. 2017; Centeno 2018; Kuridze
sized in the 1D semi-empirical C model atmosphere of
et al. 2019; Morosin et al. 2020) or using non-LTE spectral line
Fontenla et al. (1993), hereafter, FALC. Constant magnetic
inversion codes (Socas-Navarro et al. 1998; Pietarila et al. fields with varying strengths (from 0 to 100 G in steps of 20 G
2007; Milić & van Noort 2018; de la Cruz Rodríguez et al. and from 100 G to 1000 G in steps of 100 G), horizontal with
2019), these diagnostic tools do not account for the physics of respect to the local solar surface, were introduced ad hoc in the
scattering polarization or its modification via the Hanle effect. model to generate the corresponding Zeeman/Hanle polariza-
However, scattering polarization signatures typically dominate tion signatures. The spectra were generated for lines-of-sight
the linear polarization profiles of Ca II 8542 Å in weak-field (LOS) with five different heliocentric angles (θ), from disk
areas (Manso Sainz & Trujillo Bueno 2003, 2010), particularly center to close to the limb (μ = 1, 0.9, 0.6, 0.5, 0.1, where
close to the limb, where the observing geometry maximizes m = cos q ). In all geometries, the magnetic field was kept
these polarization signatures. When combined with the horizontal to the local solar surface, so the angle between the
enhancing effect of shocks, the linear polarization signatures magnetic field direction and the LOS adopts different values
can reach amplitudes of up to 1% of the continuum intensity (and so does the ratio of linear to circular Zeeman-induced
(Carlin et al. 2012, 2013), which would dominate over Zeeman polarization from the observer’s perspective). Figure 1 depicts
signatures induced by magnetic fields in the hecto-Gauss range. three of the LOS geometries listed above, where the black

2
The Astrophysical Journal, 918:15 (8pp), 2021 September 1 Centeno, de la Cruz Rodríguez, & del Pino Alemán

Figure 1. Representation of three of the five LOS geometries used in the Hanle-RT forward-modeling calculations (μ = 1, μ = 0.9, and μ = 0.1 shown here). This
figure shows a top view of a cross-section through the Sun’s equator, where (0, 0) represents the center of the Sun. The magnetic field vector, B, is always tangent to
the solar surface, and its LOS component points toward the observer for the observing geometries away from the disk center. The Sun’s equator, the magnetic field,
and the LOS are in the same plane.

arrow represents the magnetic field vector, always tangential to sight (see Figure 1). In all other geometries, there is a V/I
the solar surface, and the red dashed line shows the line-of- signal as long as the magnetic field is non-zero. Stokes I and
sight. Stokes V are not affected by this form of atomic-level
The atomic model of Ca II used in this work has five bound polarization (atomic alignment), so the Zeeman-only and full
energy levels as well as the Ca III continuum, allowing for five calculation profiles look identical.
bound–bound radiative transitions (i.e., the Ca II H & K lines in
the near-UV and the Ca II infrared triplet; see Figure 1 in 2.2. Inversion
Centeno 2018). Ca II 8542 Å is a magnetically sensitive line
that arises from the transition between the 32D5/2 and the As stated earlier, there is currently no non-LTE spectral line
42P3/2 levels. Both in the Hanle-RT synthesis as well as in the inversion code capable of treating the physical processes that
spectral line inversion, the IR triplet is computed assuming lead to atomic-level polarization and the Hanle effect in the Ca
complete frequency redistribution (CRD), while the H & K UV II 8542 Å line. In order to assess how “standard” non-LTE
lines are treated in PRD accounting for Raman scattering (also spectral line inversion codes interpret the polarized signatures
referred to as “cross-redistribution” in some works) from the IR due to these subtle quantum-mechanical effects, we inverted
triplet to the UV doublet. The spectral lines were synthesized the synthetic Hanle-RT profiles with the STockholm inversion
on a regular wavelength grid with a spectral sampling of 1 mÅ Code (STiC, de la Cruz Rodríguez et al. 2016, 2019) and
and a coverage that spanned ±1.2 Å from the line core. No compared the inversion results to the magnetic field of the
additional broadening mechanisms (aside from the natural, model used to generate the original spectra.
thermal, and collisional broadenings) were included in the STiC is a parallelized non-LTE full-Stokes spectral line
inversion code that uses a modified version of the Rybycki and
calculations, which allowed for a fair amount of continuum in
Hummer radiative transfer code (RH, Uitenbroek 2001) to
the polarization spectra.
solve for the atomic population densities assuming statistical
Two sets of synthetic Stokes spectra were computed for each
equilibrium in a plane-parallel atmosphere. The code can
magnetic field strength and each observing geometry, one in
account for PRD effects (Leenaarts et al. 2012) which are
which the linear polarization signatures were a consequence of
important for the modeling of many chromospheric lines. The
scattering phenomena as well as the joint actions of the Hanle polarized radiative transfer equation is solved using cubic
and Zeeman effects and another one in which the atomic-level Bezier solvers (de la Cruz Rodríguez & Piskunov 2013). The
polarization was “turned off” and only the signals due to the inversion engine of STiC is based on a Levenberg–Marquardt
Zeeman effect were permitted. This allowed us to evaluate the algorithm (Press et al. 1992) and includes an equation of state
performance of the spectral line inversion of purely Zeeman- extracted from the Spectroscopy Made Easy (SME, Piskunov
induced signatures against the inversion of the more realistic & Valenti 2017) code. Polarization is exclusively generated by
scenario including scattering polarization and the Hanle effect. the Zeeman effect, therefore STiC cannot account for
The total number of synthetic Stokes profiles is 150 (15 values polarization due to scattering processes and its modification
of the field strength for 5 observing geometries and 2 via the Hanle effect.
calculation modes—with and without atomic-level As in the synthesis, the Ca II model atom used in the
polarization). inversions contains five bound energy levels and the Ca III
Figure 2 shows a few examples of the synthetic Stokes continuum, which allows for the bound–bound radiative
profiles for several field strengths and observing geometries. As transitions of Ca II H & K and the Ca IR triplet. As in the
expected, the Stokes spectra for the LOS with μ = 0.1 (the two previous section, the H & K lines are treated in the PRD
examples on the right side of the figure) show the largest regime, while the IR triplet is computed assuming CRD.
differences between the Zeeman-only case (blue solid line) and The inversions were carried out in two cycles, following the
the full calculation with scattering polarization (magenta dots). strategy of Vissers et al. (2021), with four nodes for the
This is because observing geometries very close to the limb temperature in the first cycle and seven in the second cycle. The
favor larger contributions of scattering polarization to the number of nodes for the rest of the parameters stayed constant
Stokes profiles. Note that for the disk-center geometry throughout the two cycles, with three for the microturbulent
(μ = 1.0), V/I is always zero because the magnetic field is velocity, and one for each one of the remaining physical
tangential to the surface and therefore transverse to the line-of- parameters (the LOS velocity, the LOS component of the

3
The Astrophysical Journal, 918:15 (8pp), 2021 September 1 Centeno, de la Cruz Rodríguez, & del Pino Alemán

Figure 2. Examples of sets of synthetic Stokes profiles for different magnetic field strengths and different observing geometries. The magenta dots correspond to the
profiles computed taking into account scattering polarization and the joint action of the Hanle and Zeeman effects, while the blue solid line corresponds to the
calculations where only Zeeman-induced polarization was permitted. The field strength and the observing geometry, μ, are labeled at the top of each Stokes I panel.

magnetic field, and the transverse component of the magnetic et al. 2014 for instance):
field and its azimuth in the plane of the sky). In the case of the
viewing angle with μ = 0.1, additional nodes for the temper- ws2
c2 = å ss l
å [IsOBS (l) - IsSYN (l)]2 (1 )
ature and the microturbulent velocity were needed in the s
second cycle in order to obtain a good fit to the Stokes I profile.
where the weights, ws, were chosen to approximately equalize
However, STiC implements a regularization of the merit
the amplitudes of the synthetic Hanle-RT Stokes Q, U, and V to
function that minimizes artifacts in the atmospheric gradients
that result from the excess freedom in the parameters of the that of Stokes I. In this expression, IOBS represents the Hanle-
model atmosphere (see de la Cruz Rodríguez et al. 2019, for RT Stokes spectra to be inverted and ISYN represents the best fit
details). found by the STiC inversion code. σs is the expected noise of
The initial guess model atmosphere was the same for all each one of the Stokes parameters. The data that we were
inversions. It consisted of a version of the FAL-C model with a inverting were noiseless, so we artificially set to σs ∼ 10−3
perturbed temperature profile and a constant value for the LOS times the continuum intensity to indicate to STiC the
velocity and the magnetic field vector. Each set of Stokes magnitude of the acceptable error in fitting Stokes I. The
spectra was paired with a set of weights to ensure a good indices s and λ run through the four Stokes parameters and the
inversion fit for all Stokes parameters simultaneously (Del Toro wavelength, respectively. Note that the merit function inside
Iniesta & Ruiz Cobo 2016). The inversion weights are a set of the STiC code has additional terms that apply the regularization
four values used to emphasize or de-emphasize the relative scheme and the normalization of χ2.
importance of a given Stokes parameter with respect to Stokes I
in the evaluation of the merit function, χ2, that the inversion 3. Analysis
algorithm is trying to minimize (see explanation in Centeno 3.1. Disk Center, μ = 1
Figure 3 summarizes the results for the observing geometry
with μ = 1.0. The top row shows inversion results for the

4
The Astrophysical Journal, 918:15 (8pp), 2021 September 1 Centeno, de la Cruz Rodríguez, & del Pino Alemán

Figure 3. Inversion results for the LOS with μ = 1. The top row shows results for the cases in which atomic polarization was accounted for while the bottom row
concerns itself with the Zeeman-only case. The first column shows the retrieved magnetic field versus the model value, where the solid line represents the one-to-one
expected correspondence. The second and third columns show the fits (cyan lines) to the Hanle-RT Stokes I and Q/I (blue dots) for the case of a magnetic field of
∼100 G. The last two columns present the retrieved temperature and magnetic field (cyan line) as a function of optical depth and how they compare to the model used
for the Hanle-RT synthesis (blue dots) as well as the initial guess for the inversion (red line).

profiles computed with scattering polarization and the joint with different sensitivities to temperature (see, for instance, da
action of the Hanle and Zeeman effects, while the bottom row Silva Santos et al. 2018; de la Cruz Rodríguez et al. 2019). That
presents the results for the Zeeman-only case. The left-most said, the discrepancies are larger at optical depths where Ca II
column compares the values of the horizontal field retrieved by 8542 Å lacks sensitivity to the temperature, especially above
the STiC inversion, B^STiC, to the field in the model, B^mod (with log t = -5.5 (Quintero Noda et al. 2016). In the regions of the
which the synthetic Hanle-RT profiles were generated). The atmosphere where the Ca II 8542 Å line should have sensitivity
retrieved magnetic field values in the Zeeman-only case to the temperature, some of the discrepancy between the
(bottom) are always extremely close to the real values (black original model and the inversion results comes from the fact
solid line). The initial atmospheric guess as well as the choice that the Hanle-RT syntheses were carried out in the FALC
of weights for the Stokes parameters impact the solution to model atmosphere with electron density and hydrogen popula-
some degree, resulting in very small deviations from the actual. tions calculated in non-LTE. For the inversion, on the other
For the case of the full calculation (top), the agreement is hand, the STiC code was set up to solve the statistical
remarkable and not qualitatively different from the Zeeman- equilibrium for the Ca II atom only, while the H populations
only case when the field strength is above ∼100 G. Significant and electron densities were computed internally under the
disagreements between the retrieved field values and the assumption of LTE. Additional sources of discrepancy, more
expected ones appear below this threshold. In this regime, the difficult to quantify, stem from the differences between the
signatures of the Hanle effect alter the shape of the linear Hanle-RT and STiC codes in the formulation of the equation of
polarization profiles significantly, resulting in a misinterpreta- state, the algorithm that computes the formal solution of the
tion from the STiC code. equation of transfer, and other factors that might affect the
The second and third columns show the best fits found by the convergence of the statistical equilibrium. In these last two
inversion (cyan line) to the Hanle-RT Stokes I and Q/I (blue columns of Figure 3, the red line shows the guess model
dots), synthesized for a magnetic field of 100 G. In this atmosphere used to initialize the inversion.
particular geometry, U/I and V/I are exactly zero. It is obvious
that the STiC inversions are not able to fit the signatures
3.2. Other Geometries
characteristic of the Hanle effect (Q/I panel in top row), while
it does a good job at fitting the Zeeman-only profile (bottom As we move to observing geometries away from the disk
row). Despite the misfit of the former, STiC is still able to infer center, the magnetic field presents both a longitudinal as well as
an approximate value of the magnetic field by virtue of a transverse component in the LOS reference frame. This leads
attempting to fit the width and amplitude of the Q/I profile. It is to the appearance of Stokes V signals exclusively due to the
worth noting that, in general, the quality of the fits becomes longitudinal Zeeman effect. Furthermore, the fraction of
increasingly better as the field strength increases and the Hanle Zeeman-induced linear polarization becomes smaller and the
contribution becomes less and less important. Below 100 G, the effects of scattering polarization become more important as the
inversion fits are poor and the inferred magnetic field values are heliocentric angle of the “observation” increases (i.e., μ
rather inaccurate. decreases).
Columns 4 and 5 compare the retrieved temperature and Figure 4 shows a comparison between the components of the
horizontal magnetic field stratifications as a function of optical vector magnetic field retrieved by the STiC inversions and
depth (solid cyan lines) to the model atmosphere used for the those used for the Hanle-RT syntheses, for the LOS away from
synthesis (blue dots). The disagreements between the original the disk center. The left column corresponds to the inversion of
and the inverted temperature profiles are partly due to a Zeeman-only profiles while the right column depicts the results
degeneracy between this physical quantity and the micro- for the full forward calculation with atomic polarization.
turbulent velocity. This degeneracy can be mitigated by The top row quantifies the relative (percent) difference
carrying out simultaneous inversions of multiple spectral lines between the inferred (B^STiC) and model (B^mod ) values of the

5
The Astrophysical Journal, 918:15 (8pp), 2021 September 1 Centeno, de la Cruz Rodríguez, & del Pino Alemán

Figure 4. Inversion results for observing geometries away from the disk center. The panels show the relative (percent) differences between the magnetic field values
retrieved by the STiC inversions and those in the model atmospheres used for the Hanle-RT syntheses. Left and right correspond, respectively, to the Zeeman-only and
the full calculation with atomic polarization cases. The top two rows show the results for the strength and the azimuth of the transverse component of the magnetic
field, while the bottom panels present the results for the LOS magnetic field.

transverse magnetic field, as a function of the model value. The 3.3. The Case of μ = 0.1
x-axis is on a logarithmic scale to allow the reader to clearly see Special consideration should be given to the case of the LOS
the results at low field strengths (B^mod < 100 G), and each line with μ = 0.1. This corresponds to the observing geometry
of sight is represented by a colored symbol. For transverse closest to the limb, and therefore has the smallest transverse
magnetic field strengths above 80 G, the inversion results are magnetic fields in the observer’s reference frame (B^mod ranging
always within 5% of the true value, regardless of whether STiC from 0 to 100 G). Also, the linear polarization profiles in this
is interpreting the spectra from the full scattering polarization scenario will be the most affected by scattering polarization and
calculation or the Zeeman-only signals. Below 80 G (marked the Hanle effect.
by the vertical gray line), the inferences derived from the Figure 5 shows the best inversion fits for the case
spectra with scattering polarization (right) start to deviate B mod = 200 G in the observing geometry μ = 0.1. In the
significantly from the true values, while in the case of the LOS reference frame, this corresponds to a transverse magnetic
Zeeman-only signals (left), the inversion inferences are still field of B^mod = 20 G and an LOS component of Bmod = 199
accurate. The middle row shows the difference between the G. The top row corresponds to the full calculation, including
inferred and the model magnetic field azimuths (in degrees) as scattering polarization and the Hanle effect, while the bottom
a function of B^mod . Just as in the case of the horizontal field row shows the results for the Zeeman-only calculations. The
strength, the inferences from the scattering polarization open circles represent the synthetic profiles calculated with
signatures become less reliable below the 80 G mark. Hanle-RT (from left to right, I, Q/I, U/I, V/I) while the cyan
The bottom row presents the inversion results for the solid line shows the best fit found by the STiC inversion code.
This figure epitomizes how the inversion code interprets the
longitudinal component of the magnetic field, BSTiC, and how
signatures introduced by atomic polarization and the Hanle
they compare to the model values (Bmod ). Because the atomic- effect in the linear polarization profiles. In particular, the
level polarization only carries atomic alignment, it does not second panel of the top row shows how the Stokes Q/I profile
affect the circular polarization signals, which are exclusively is affected by these subtle quantum-mechanical effects
due to the Zeeman effect. Therefore, the Stokes V signals are (magenta circles), which is to be compared to its Zeeman-
correctly interpreted by STiC, and the BSTiC inferences are only counterpart in the bottom row (blue circles). In both cases,
always rather accurate. the best fits found by the inversion code (cyan lines) look rather
similar, and yield comparable values of the transverse magnetic

6
The Astrophysical Journal, 918:15 (8pp), 2021 September 1 Centeno, de la Cruz Rodríguez, & del Pino Alemán

Figure 5. Inversion fits for the case of B mod = 200 G and observing geometry with μ = 0.1. The open circles represent the Hanle-RT synthetic profiles, while the cyan
line shows the best fit found by the inversion code. The top row shows the case of the full calculation while the bottom row corresponds to the Zeeman-only spectra.

Figure 6. Inversion fit (black line) of Q/I for the zero-field case when scattering polarization is accounted for (open circles). STiC attempts to fit the non-magnetic
scattering polarization signals as if they were of Zeeman origin, leading to non-zero inferences of the transverse component of the magnetic field (the value retrieved in
each case is listed in the legend).

field: the inversions of the Zeeman-only case yields 20.2 G, not account for all the physical processes that can generate
while the full calculation results in 22.0 G. The model value polarization in these lines. Although the Zeeman effect is taken
was B^mod = 20 G in this case. Even though the Hanle effect into account in these codes (Socas-Navarro et al. 1998; Milić &
strongly depolarizes the π component of the Stokes Q/I signal, van Noort 2018; de la Cruz Rodríguez et al. 2019), the physics
there might be enough information in the inner wings for STiC for the generation and transfer of polarization due to scattering
to extract information about the magnetic field through the and its modification via the Hanle effect are not. This work
Zeeman effect. The resulting B^STiC is only over-estimated by assesses the errors made by one of these inversion codes (STiC)
10% in this case. when interpreting the polarization signatures of Ca II 8542 Å.
To estimate the errors incurred by STiC when interpreting
3.4. The Zero-field Case the Ca II 8542 Å spectra, we invert sets of Stokes profiles
synthesized with the Hanle-RT code in the FALC model
The case of the full calculation for B = 0 G misleads the atmosphere with ad hoc magnetic fields of different strengths.
inversion code, since the purely scattering polarization Stokes
Hanle-RT computes the spectral line polarization generated by
Q signals are interpreted by STiC as Zeeman-induced
scattering processes as well as its modification due to the
transverse magnetic fields of tens to a hundred Gauss.
combined actions of the Hanle and Zeeman effects. The spectra
Figure 6 shows the STiC fits (black solid lines) to the Hanle-
RT Q/I profiles (open circles) for the case of the full were synthesized simulating five observing geometries, from
calculation with B = 0 G. From left to right, the observing disk center (μ = 1) to close to the solar limb (μ = 0.1).
geometry goes from μ = 0.9 to μ = 0.1. As the figure suggests, We find that, for most field strengths and observing
STiC is not able to fit the Q/I profiles, yet it attempts to mimic geometries, STiC does a good job of retrieving both the
the scattering polarization signals in order to minimize the longitudinal and the transverse components of the magnetic
merit function that drives the inversion algorithm. This leads to field. Even when the amplitude of the scattering polarization
non-zero values of the magnetic field, which in the case of the and Hanle signatures are comparable to that of the Zeeman
most extreme observing geometry yields B^STiC ~ 120 G. signature in the linear polarization profiles, STiC is able to
retrieve the transverse component of the magnetic field with
remarkable accuracy, overestimating its value by only a few
4. Conclusions percent. When the transverse magnetic field component is
Most spectral line inversion codes that are able to interpret weaker than 80 G, the scattering polarization and Hanle
chromospheric spectra formed under non-LTE conditions do signatures start dominating over the Zeeman-induced signals.

7
The Astrophysical Journal, 918:15 (8pp), 2021 September 1 Centeno, de la Cruz Rodríguez, & del Pino Alemán

Naturally, STiC is not able to reproduce the spectral signatures Jaime de la Cruz Rodríguez https://orcid.org/0000-0002-
induced by scattering polarization and the Hanle effect, leading 4640-5658
to errors in the retrieved transverse component of the magnetic Tanausú del Pino Alemán https://orcid.org/0000-0003-
field and its azimuth in the plane of the sky. 1465-5692
In the case where the model magnetic field is B mod = 0 G
and polarization signatures are exclusively due to scattering References
processes, STiC still attempts to fit the linear polarization by
interpreting it as a Zeeman signal. This leads to spurious Asensio Ramos, A., de la Cruz Rodríguez, J., Martínez González, M. J., &
Socas-Navarro, H. 2017, A&A, 599, A133
transverse magnetic field values, which, in the case of μ = 0.1, Asensio Ramos, A., Trujillo Bueno, J., & Landi Degl’Innocenti, E. 2008, ApJ,
is just over 120 G. However, by visually inspecting the original 683, 542
spectra as well as the inversion fits, it is possible to determine Carlin, E. S., Asensio Ramos, A., & Trujillo Bueno, J. 2013, ApJ, 764, 40
that the agreement is poor and that the inversion results are Carlin, E. S., Manso Sainz, R., Asensio Ramos, A., & Trujillo Bueno, J. 2012,
ApJ, 751, 5
questionable. Centeno, R. 2018, ApJ, 866, 89
This is a relatively simple experiment that only addresses Centeno, R., Schou, J., Hayashi, K., et al. 2014, SoPh, 289, 3531
spectral profiles synthesized in a 1D static semi-empirical da Silva Santos, J. M., de la Cruz Rodríguez, J., & Leenaarts, J. 2018, A&A,
model atmosphere with constant magnetic fields in the absence 620, A124
of noise in the Stokes parameters. The one-dimensional nature de la Cruz Rodríguez, J., Leenaarts, J., & Asensio Ramos, A. 2016, ApJL,
830, L30
of this study omits, by design, the effects of spatial de la Cruz Rodríguez, J., Lenaarts, J., Danilovic, S., & Uitenbroek, H. 2019,
inhomogeneities and horizontal radiative transfer on the A&A, 623, A74
atomic-level polarization. This likely leads to an under- or de la Cruz Rodríguez, J., & Piskunov, N. 2013, ApJ, 764, 33
overestimate of the typical linear polarization signals expected de la Cruz Rodríguez, J., Socas-Navarro, H., Carlsson, M., & Leenarts, J. 2012,
in very quiet areas of the (real) Sun, which will become less A&A, 543, A34
del Pino Alemán, T., Casini, R., & Manso Sainz, R. 2016, ApJL, 830, L24
important in more strongly magnetized regions. Moreover, as Del Toro Iniesta, J. C., & Ruiz Cobo, B. 2016, LRSP, 13, 4
mentioned in the introduction, strong velocity gradients can Fontenla, J. M., Avrett, E. H., & Loeser, R. 1993, ApJ, 406, 319
lead to much larger scattering polarization signatures in Jaume Bestard, J., Trujillo Bueno, J., Štěpán, J., & del Pino Alemán, T. 2021,
observing geometries close to the limb. The interpretation of ApJ, 909, 183
Kuridze, D., Mathioudakis, M., Morgan, H., et al. 2019, ApJ, 874, 126
such signals with a non-LTE inversion code could lead to Lagg, A., Lites, B. W., Harvey, J., Gosain, S., & Centeno, R. 2017, SSRv,
larger errors in the retrieved magnetic field values. Never- 210, 37
theless, this work provides a first attempt at setting a threshold Landi Degl’Innocenti, E., & Landolfi, M. 2004, Polarization in Spectral Lines
for the applicability of non-LTE spectral line inversion codes, (Dordrecht: Kluwer)
Leenaarts, J., Pereira, T., & Uitenbroek, H. 2012, A&A, 543, A109
and shows that, when the contribution of scattering polarization Manso Sainz, R., Roncero, O., Sanz-Sanz, C., et al. 2014, ApJ, 788, 118
to the linear polarization signals is smaller than that of its Manso Sainz, R., & Trujillo Bueno, J. 2003, PhRvL, 91, 111102
Zeeman-induced counterpart, the inversion results are as Manso Sainz, R., & Trujillo Bueno, J. 2010, ApJ, 722, 1416
accurate as when the polarization signals are due to the Milić, I., & van Noort, M. 2018, A&A, 617, A24
Zeeman effect only. Morosin, R., de la Cruz Rodríguez, J., Vissers, G. J. M., & Yadav, R. 2020,
A&A, 642, A210
Pietarila, A., Socas-Navarro, H., & Bogdan, T. 2007, ApJ, 670, 885
This material is based upon work supported by the National Piskunov, N., & Valenti, J. A. 2017, A&A, 597, A16
Center for Atmospheric Research, which is a major facility Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. 1992,
sponsored by the National Science Foundation under Coopera- Numerical Recipes in C: The Art of Scientific Computing (2nd ed.;
Cambridge: Cambridge Univ. Press)
tive Agreement No. 1852977. This project has received Quintero Noda, C., Shimizu, T., de la Cruz Rodríguez, J., et al. 2016, MNRAS,
funding from the European Research Council (ERC) under 459, 3363
the European Union’s Horizon 2020 research and innovation Quintero Noda, C., Shimizu, T., Katsukawa, Y., et al. 2017, MNRAS,
program (SUNMAG, grant agreement 759548). T.d.P.A. 464, 4534
acknowledges the funding received from the European Socas-Navarro, H., Trujillo Bueno, J., & Ruiz Cobo, B. 1998, ApJ, 507, 470
Štěpán, J., & Trujillo Bueno, J. 2016, ApJ, 826, L10
Research Council (ERC) under the European Union’s Horizon Trujillo Bueno, J. 2001, in ASP Conf. Ser. 236, Advanced Solar Polarimetry:
2020 research and innovation program (ERC Advanced grant Theory, Observations and Instrumentation, Proc. 20th Int. Sac. Peak
agreement No. 742265). Summer Workshop, ed. M. Sigwarth (San Francisco, CA: ASP), 161
Uitenbroek, H. 1989, A&A, 213, 360
Uitenbroek, H. 2001, ApJ, 557, 389
ORCID iDs Vissers, J. G. M., Danilovic, S., de la Cruz Rodríguez, J., et al. 2021, A&A,
Rebecca Centeno https://orcid.org/0000-0002-1327-1278 645, A1
Wedemeyer-Bohm, S., & Carlsson, M. 2011, A&A, 528, A1