*ad hoc*basis. We rigorously demonstrate that optimization strategies in Stokes polarimetry based on three common metrics, namely the Frobenius condition number of the instrument matrix, the determinant of the associated Gram matrix, or the equally weighted variance, are frequently formally equivalent. In particular, using each metric, we derive the same set of constraints on the measurement states, correcting a previously reported proof, and show that these can be satisfied using spherical 2 designs. Discussion of scenarios in which equivalence between the metrics breaks down is also given. Our conclusions are equally applicable to optimization of the illumination states in Mueller matrix polarimetry.

## 1.

## Introduction

Quantitative analysis of the state of polarization of light provides a powerful tool in modern science. Applications vary from microscopy, biomedical diagnosis, and astrophysics^{1}^{–}^{3} to crystallographic, material, and single-molecule studies.^{4}^{,}^{5} While the polarization state of light itself can be used to transmit information, hence presenting new opportunities in optical data storage and communications,^{6}^{–}^{9} changes in polarization induced by a material can alternatively be used for object detection^{10} or to characterize sample properties, such as chirality or molecular orientation.^{11}^{–}^{13}

Stokes polarimeters, which allow a complete characterization of the polarization state of input light as described by the associated $4\times 1$ Stokes vector $\mathbf{S}$, comprise of $N$ ($\ge 4$) distinct measurements that can be multiplexed in time,^{14} frequency,^{15} or space.^{16} Fundamentally, each constituent measurement outputs an intensity ${I}_{j}$ ($j\in [1,N]$), which is proportional to the projection of the incident Stokes vector onto an analysis state vector ${\mathbf{W}}_{j}$, i.e., ${I}_{j}={\mathbf{W}}_{j}^{\mathrm{T}}\mathbf{S}$. Central to the description and design of Stokes polarimeters is hence the so-called instrument or measurement matrix $\mathbb{W}={({\mathbf{W}}_{1},{\mathbf{W}}_{2},\dots ,)}^{\mathrm{T}}$ formed from stacking the set of analysis vectors. In order to obtain an estimate of the Stokes vector from the set of projections ${I}_{j}$, the measurement matrix must be inverted. So as to limit noise propagation through this inversion process, optimization of the measurement matrix is hence frequently performed. Optimization in this vein has been performed using different metrics including the associated information content,^{17}^{–}^{19} matrix determinant,^{20}^{–}^{22} signal-to-noise ratio,^{23} equally weighted variance (EWV),^{24}^{,}^{25} and condition number.^{21}^{–}^{23}^{,}^{25}^{–}^{29}

Mueller matrix polarimeters, on the other hand, combine a Stokes polarimeter with use of multiple incident polarized states so as to measure the full Mueller matrix of an object. Variation of the probing polarization states (as can be described using an analogous illumination matrix), therefore, introduces additional degrees of freedom, hence admitting further optimization.^{17}^{,}^{28}^{–}^{32} Application specific optimization of polarimeters has also been reported, for example, in detection and imaging problems the polarization contrast is a more suitable metric.^{33}^{,}^{34}

Recently, the equivalence of a number of optimization metrics, namely the EWV, the condition number of $\mathbb{W}$, and the determinant of the associated Gram matrix, was discussed by Foreman et al.^{35} Additionally, Foreman et al. proved that a Stokes polarimeter is optimal (as characterized by these metrics) when the set of analysis states defines a spherical 2 design^{36} on the unit Poincaré sphere. A re-examination of the equivalence between these metrics is, however, necessary due to an error in the proof presented in Ref. 35. The goal of this paper is, therefore, to provide a rigorous proof that the conclusions of Ref. 35 hold. Our derivations also elicit greater insight into the optimization of nonideal Stokes polarimeters, which is hence discussed. We additionally note that our results are equally applicable to optimization of the probing states used in Mueller matrix polarimetry due to the similar matrix structure of the problem.^{31}^{,}^{37}

## 2.

## Optimal Polarimetry with Spherical 2 Designs

The instrument matrix $\mathbb{W}$ of a polarimeter is an $N\times 4$ matrix, the rows of which are the Stokes vectors of the $N$ polarization states being analyzed, normalized such that the polarimeter is passive. Accordingly, the instrument matrix has the parametric form:

## Eq. (1)

$$\mathbb{W}=\frac{1}{2}\left[\begin{array}{cc}1& {\mathbf{w}}_{1}^{\mathrm{T}}\\ 1& {\mathbf{w}}_{2}^{\mathrm{T}}\\ \vdots & \vdots \\ 1& {\mathbf{w}}_{N}^{\mathrm{T}}\end{array}\right]\triangleq \frac{1}{2}(\begin{array}{cc}\mathbf{r}& \mathbb{Q}\end{array}),$$In Stokes polarimetry, one performs $N$ intensity measurements ${I}_{j}$, $j\in [1,N]$ by projecting the input Stokes vector $\mathbf{S}$ onto each of the $N$ analyzers described by the $N$ rows of the matrix $\mathbb{W}$. If these measurements are stacked in an $N$-dimensional vector $\mathbf{I}={({I}_{1},{I}_{2},\dots ,{I}_{N})}^{\mathrm{T}}$, and if we assume that the measurements are perturbed by white additive noise, we obtain

where $\mathbf{\Delta}$ is an $N\times 1$ random vector with covariance matrix ${\sigma}^{2}{\mathbb{I}}_{N}$ and ${\mathbb{I}}_{n}$ denotes the $n\times n$ identity matrix. The maximum-likelihood estimate of $\mathbf{S}$ is obtained by where denotes the pseudoinverse matrix. The estimate $\widehat{\mathbf{S}}$ is a random vector of mean $\mathbf{S}$ (i.e., the estimator is unbiased) and of covariance matrix^{17}

^{,}

^{23}

^{,}

^{24}The estimation variances of each element of the Stokes vector estimate are the diagonal elements of this matrix. A natural goal of polarimeter optimization is to find the measurement matrix $\mathbb{W}$ that minimizes the sum of these variances, i.e., the trace of ${\mathbb{K}}_{\mathbf{S}}$. The corresponding performance metric is called the EWV, i.e.,

## Eq. (6)

$$\mathrm{EWV}=\mathrm{tr}({\mathbb{K}}_{\mathbf{S}})={\sigma}^{2}\mathrm{tr}({\mathbb{G}}^{-1}),$$To optimize the EWV, we first express the Gram matrix $\mathbb{G}$ in block format, viz.

## Eq. (8)

$$\mathbb{G}=\frac{1}{4}\left(\begin{array}{cc}N& {\mathbf{r}}^{\mathrm{T}}\mathbb{Q}\\ {\mathbb{Q}}^{\mathrm{T}}\mathbf{r}& {\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}\end{array}\right)\triangleq \left(\begin{array}{cc}A& {\mathbf{B}}^{\mathrm{T}}\\ \mathbf{C}& \mathbb{D}\end{array}\right).$$^{38}

## Eq. (9)

$${\mathbb{G}}^{-1}=\left[\begin{array}{cc}{A}^{-1}+{A}^{-1}{\mathbf{B}}^{\mathrm{T}}{\mathbb{M}}^{-1}\mathbf{C}{A}^{-1}& -{A}^{-1}{\mathbf{B}}^{\mathrm{T}}{\mathbb{M}}^{-1}\\ -{\mathbb{M}}^{-1}\mathbf{C}{A}^{-1}& {\mathbb{M}}^{-1}\end{array}\right],$$## Eq. (11)

$$\mathrm{tr}({\mathbb{G}}^{-1})={A}^{-1}+{A}^{-1}{\mathbf{B}}^{\mathrm{T}}{\mathbb{M}}^{-1}\mathbf{C}{A}^{-1}+\mathrm{tr}({\mathbb{M}}^{-1}).$$## Eq. (12)

$$\mathbb{M}=\frac{1}{4}({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}-\frac{{\mathbf{qq}}^{\mathrm{T}}}{N}),$$^{38}

## Eq. (13)

$${(\mathbb{Z}+{\mathbf{xy}}^{\mathrm{T}})}^{-1}={\mathbb{Z}}^{-1}-\frac{{\mathbb{Z}}^{-1}{\mathbf{xy}}^{\mathrm{T}}{\mathbb{Z}}^{-1}}{1+{\mathbf{y}}^{\mathrm{T}}{\mathbb{Z}}^{-1}\mathbf{x}},$$## Eq. (14)

$${\mathbb{M}}^{-1}=4{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}+4\frac{{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}{\mathbf{qq}}^{\mathrm{T}}{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}}{N-{\mathbf{q}}^{\mathrm{T}}{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}\mathbf{q}}.$$## Eq. (15)

$$\mathrm{tr}({\mathbb{G}}^{-1})=4\{\frac{1}{N}+\mathrm{tr}[{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}]\phantom{\rule{0ex}{0ex}}+\frac{{\mathbf{q}}^{\mathrm{T}}[N{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-2}+{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}]\mathbf{q}}{N[N-{\mathbf{q}}^{\mathrm{T}}{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}\mathbf{q}]}\},$$^{38}

Noting that $N>0$ and that ${\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}$ is positive definite, it follows immediately that the first two terms in Eq. (15) are positive. We show in Sec. 6 that the third term is also positive. Consequently, the trace in Eq. (15) is minimal when its three terms are minimal. The first term is constant, and the third is minimal when it is null, i.e., when $\mathbf{q}={\mathbb{Q}}^{\mathrm{T}}\mathbf{r}=\mathbf{0}$ or equivalently

Importantly, Eq. (16) expresses a polynomial constraint that must be satisfied by an optimal measurement matrix and is equivalent to that given in Eq. (4) of Ref. 35. When Eq. (16) holds, minimizing $\mathrm{tr}({\mathbb{G}}^{-1})$ is equivalent to minimizing $\mathrm{tr}[{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}]$. This optimization has to be done under the constraint that the trace of the matrix ${\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}$ is constant as follows from the normalization of ${\mathbf{w}}_{j}$. Indeed, since each row of the matrix $\mathbb{Q}$ is a unit-norm vector, we have## Eq. (17)

$$\mathrm{tr}({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})=\mathrm{tr}(\mathbb{Q}{\mathbb{Q}}^{\mathrm{T}})=N.$$## Eq. (18)

$$\mathrm{\Psi}(\mathbb{Q})=\mathrm{tr}[{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}]-\mu [\mathrm{tr}({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})-N],$$## Eq. (19)

$$\mathrm{\Psi}(\mathit{\beta})=\sum _{j=1}^{3}\frac{1}{{\beta}_{j}}+\mu (\sum _{j=1}^{3}{\beta}_{j}-N),$$## Eq. (20)

$${\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}=\sum _{j=1}^{N}{\mathbf{w}}_{j}{\mathbf{w}}_{j}^{\mathrm{T}}=\frac{N}{3}{\mathbb{I}}_{3}.$$## Eq. (21)

$$\mathbb{G}={\mathbb{W}}^{\mathrm{T}}\mathbb{W}=\frac{N}{12}\left[\begin{array}{cccc}3& 0& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right].$$## Eq. (22)

$${\mathbb{K}}_{\mathbf{S}}=\frac{4}{N}{\sigma}^{2}\left[\begin{array}{cccc}1& 0& 0& 0\\ 0& 3& 0& 0\\ 0& 0& 3& 0\\ 0& 0& 0& 3\end{array}\right].$$Finally, the conditions expressed by Eqs. (16) and (20) are satisfied when the set of measurement states on the normalized Poincaré sphere, defined by $\{{\mathbf{w}}_{j}\}$, $j\in [1,N]$, constitute a spherical 2 design (see Sec. 7 for a proof) as reported in Ref. 35. A spherical $t$ design is defined as a collection of $N$ points $\{{\mathbf{w}}_{j}\}$ on the surface of the unit sphere (in our case in ${\mathcal{R}}^{3}$) for which the normalized integral of any polynomial function, $f(\mathbf{w})$, of degree $t$ or less is equal to the average taken over the $N$ points. The Platonic solids, i.e., the regular tetrahedron ($N=4$), the octahedron ($N=6$), the cube ($N=8$), the icosahedron ($N=12$), and the dodecahedron ($N=20$), are well-known examples of spherical 2 designs. A geometric scheme to construct optimal polarimeters for any even $N$, any factorable odd value of $N$, and for prime $N>5$ has also been described in Ref. 35. Further examples of spherical designs and construction strategies can be found in Refs. 3940.–41. Critically, spherical 2 designs are known to exist for any $N\ge 4$, with the important exception of $N=5$.^{39}^{,}^{41} In the context of optimal polarimetry, this implies that for $N=5$ the constraints described by Eqs. (16) and (20) cannot be fully satisfied. Recalling Eq. (15), this arises because the second and third terms cannot be simultaneously minimized. Although the resulting measurement states do not form a spherical 2 design, the sum of these two terms, and hence the EWV, can nevertheless be minimized yielding a value of $8.119{\sigma}^{2}$. The corresponding analysis states define a square pyramid inscribed by the unit Poincaré sphere.

## 3.

## Equivalence of Optimization Metrics

We will now demonstrate that the optimization of two other popular metrics, namely the condition number and the determinant of the Gram matrix, lead to exactly the same measurement frames as the EWV so that these three criteria are strictly equivalent.

## 3.1.

### Condition Number

The condition number $\kappa $ of the instrument matrix is defined by $\kappa =\Vert \mathbb{W}\Vert \Vert {\mathbb{W}}^{+}\Vert $, where ${\mathbb{W}}^{+}$ is the pseudoinverse matrix and $\Vert \cdots \Vert $ denotes the matrix norm. In principle, any choice of matrix norm can be made, however, within the context of polarimetry, the most common choices are those of either the 2-norm,^{42}^{,}^{43} defined as the maximum singular value of $\mathbb{W}$, or the Frobenius norm,^{27}^{,}^{35}^{,}^{43} given by^{38}

## Eq. (23)

$${\Vert \mathbb{P}\Vert}_{F}={[\mathrm{tr}({\mathbb{P}}^{\mathrm{T}}\mathbb{P})]}^{1/2}={[\mathrm{tr}(\mathbb{P}{\mathbb{P}}^{\mathrm{T}})]}^{1/2}.$$^{43}In this paper, we exclusively consider the Frobenius norm (and henceforth drop the subscript $F$). This selection is motivated by the resulting equivalence between the condition number and EWV. To prove this equivalence [for polarimeters with instrument matrix of the form of Eq. (1)], we first note that our choice of normalization of the measurement states ${\mathbf{W}}_{j}={[1,{\mathbf{w}}_{j}]}^{\mathrm{T}}/2$ implies that

## Eq. (24)

$${\Vert \mathbb{W}\Vert}^{2}=\mathrm{tr}({\mathbb{W}}^{\mathrm{T}}\mathbb{W})=\frac{N}{2}.$$## Eq. (25)

$${\Vert {\mathbb{W}}^{+}\Vert}^{2}=\mathrm{tr}[{({\mathbb{W}}^{+})}^{\mathrm{T}}{\mathbb{W}}^{+}]=\mathrm{tr}[{({\mathbb{W}}^{\mathrm{T}}\mathbb{W})}^{-1}]=\frac{\mathrm{EWV}}{{\sigma}^{2}}.$$## 3.2.

### Determinant of the Gram Matrix

The first works on Stokes polarimeter optimization considered devices with a minimal number ($N=4$) of measurement vectors.^{26} Optimization of such systems used the determinant of the matrix $\mathbb{W}$ (which for this value of $N$ is square and nonsingular) as a performance metric. In this case, the optimal structure found dictated that the measurement vectors defined a regular tetrahedron on the Poincaré sphere, a result that we also found above by optimizing the EWV. We show in this section that this result comes from the strict equivalence of these two optimization metrics. This equivalence can be generalized to any value of $N$ if one considers the optimization of the determinant of the Gram matrix $\mathbb{G}$ since for $N>4$ the matrix $\mathbb{W}$ itself is rectangular and its determinant is thus not defined. Notice that this equivalence was mentioned in Ref. 35, but there was an erroneous step in the logic presented in that work (see Sec. 8 for more details).

We intend here to show that maximization of the determinant $|\mathbb{G}|$ yields the same polynomial constraints embodied in Eqs. (16) and (20). Considering the block form of the Gram matrix in Eq. (8), its determinant can be written as^{38}

## Eq. (27)

$$|\mathbb{G}|=|A-{\mathbf{B}}^{\mathrm{T}}{\mathbb{D}}^{-1}\mathbf{C}||\mathbb{D}|\phantom{\rule{0ex}{0ex}}=\frac{1}{256}[N-{\mathbf{r}}^{\mathrm{T}}\mathbb{Q}{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}{\mathbb{Q}}^{\mathrm{T}}\mathbf{r}]|{\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}|.$$For the second factor, we note that $|{\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}|=\prod _{j=1}^{3}{\beta}_{j}$ where ${\beta}_{j}$, $j\in [\mathrm{1,3}]$, are again the eigenvalues of the matrix ${\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}$, which are positive since ${\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}$ is positive definite. Moreover, according to Eq. (17), the matrix ${\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}$ has constant trace. Maximization of $|{\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}|$ is thus once again a constrained optimization problem, which can be solved using the method of Lagrange multipliers. We will consider maximization of $\mathrm{ln}|{\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}|$, which is equivalent since the logarithmic function is monotonically increasing. The Lagrange function then becomes

## Eq. (28)

$$\mathrm{\Psi}(\mathit{\beta})=\sum _{j=1}^{3}\mathrm{ln}\text{\hspace{0.17em}}{\beta}_{j}-\mu (\sum _{j=1}^{3}{\beta}_{j}-N).$$## 4.

## Discussion

The main conclusion from the analysis presented in the Secs. 2 and 3 is that among all measurement matrices of the form described by Eq. (1), those that maximize the condition number, the EWV and the determinant are exactly the same. Our result can thus be said to unify many previous works on polarimeter optimization, e.g., the early work of Azzam et al.^{26} (which optimized based on the instrument matrix determinant), Ambirajan and Look^{22} (based on the condition number and determinant), Sabatke et al.^{24} (based on the EWV and determinant), and Tyo^{44} (based on condition number), among many others.

Modeling of $\mathbb{W}$ based on Eq. (1) implies that the transmittance of each polarization analyzer and the degree of polarization of the transmitted light are both equal to one. This assumption is frequently made in polarimetry, however, it is interesting to consider the case where it is not fulfilled. In the general case, each analyzer, as described by each row of the measurement matrix, may have a different transmission ${t}_{i}$, $i\in [1,N]$, and a different resulting degree of polarization ${P}_{i}$, $i\in [1,N]$, such that the measurement matrix can be expressed in the form:

## Eq. (29)

$$\mathbb{W}=\frac{1}{2}\left[\begin{array}{cc}{t}_{1}& {t}_{1}{P}_{1}{\mathbf{w}}_{1}^{\mathrm{T}}\\ {t}_{2}& {t}_{2}{P}_{2}{\mathbf{w}}_{2}^{\mathrm{T}}\\ \vdots & \vdots \\ {t}_{N}& {t}_{N}{P}_{N}{\mathbf{w}}_{N}^{\mathrm{T}}\end{array}\right]\triangleq \frac{1}{2}(\begin{array}{cc}\mathbb{T}\mathbf{r}& \mathbb{TPQ}\end{array}),$$## Eq. (31)

$$\kappa =\frac{\sqrt{\sum _{i=1}^{N}(1+{P}_{i}^{2}){t}_{i}^{2}}}{2\sigma}\sqrt{\mathrm{EWV}}.$$## Eq. (32)

$$\mathrm{EWV}=4{\sigma}^{2}\{\frac{1}{T}+\mathrm{tr}[{({\mathbb{Y}}^{\mathrm{T}}\mathbb{Y})}^{-1}]\phantom{\rule{0ex}{0ex}}+\frac{{\mathbf{y}}^{\mathrm{T}}[T{({\mathbb{Y}}^{\mathrm{T}}\mathbb{Y})}^{-2}+{({\mathbb{Y}}^{\mathrm{T}}\mathbb{Y})}^{-1}]\mathbf{y}}{T[T-{\mathbf{y}}^{\mathrm{T}}{({\mathbb{Y}}^{\mathrm{T}}\mathbb{Y})}^{-1}\mathbf{y}]}\},$$## Eq. (33)

$$\mathrm{EWV}=\frac{4{\sigma}^{2}}{{\tau}^{2}}\{\frac{1}{N}+\frac{1}{{\rho}^{2}}\mathrm{tr}[{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}]\phantom{\rule{0ex}{0ex}}+\frac{{\mathbf{q}}^{\mathrm{T}}[(N/{\rho}^{2}){({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-2}+{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}]\mathbf{q}}{N[N-{\mathbf{q}}^{\mathrm{T}}{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}\mathbf{q}]}\},$$Another important practical question is which of the three considered metrics is the most appropriate for evaluating the performance of a polarimeter under more general conditions. Indeed, from this point of view, the metrics are not necessarily equivalent, particularly in complex noise regimes or when nonideal polarization state analyzers are used. This is most easily seen by noting that the three metrics can be expressed in the form:

## Eq. (35)

$$\kappa ={\left(\sum _{i=1}^{4}{\nu}_{i}\right)}^{1/2}{\left(\sum _{i=1}^{4}\frac{1}{{\nu}_{i}}\right)}^{1/2},$$## Eq. (37)

$${\mathbb{K}}_{\mathbf{S}}={\mathbb{G}}^{-1}{\mathbb{W}}^{\mathrm{T}}{\mathbb{K}}_{\mathbf{I}}\mathbb{W}{\mathbb{G}}^{-\mathrm{T}},$$Another strong advantage of the EWV is that it can be used for polarimeter optimization in the presence of nonadditive noises sources. The EWV has been used to determine the optimal measurement frames in the presence of Poisson shot noise.^{45}^{,}^{46} In this case, the covariance matrix of the Stokes estimate takes a different form to that of Eq. (5). Consequently, the EWV is no longer given by Eq. (6), and thus not proportional to the square of the condition number. Furthermore, when measurements are simultaneously affected by several types of statistically independent noise sources, the total EWV is simply the sum of the individual EWVs for each noise source. This additive property has been recently employed to characterize the actual performance of microgrid-based polarimetric cameras in the presence of both additive detection noise and Poisson shot noise.^{47}

In conclusion, the key finding of the present work is that when optimizing the estimation performance of a polarimeter in the presence of additive Gaussian noise, the Frobenius condition number of the instrument matrix, the Gram determinant, and EWV are three strictly equivalent metrics. When evaluating and comparing the performance of different polarimeters however, or when optimizing polarimeters in the presence of nonadditive, non-Gaussian noise sources, the EWV has strong advantages compared with the other two metrics.

## 5.

## Conclusions

We have shown that optimization of the EWV, of the Frobenius condition number, or of the determinant of the Gram matrix of a Stokes polarimeter leads to the same optimal measurement structures, namely, spherical 2 designs. These structures yield a very simple closed-form expression for the covariance matrix of the Stokes vector estimator and thus of the variances of each element of the Stokes vector. These expressions constitute the fundamental limit of the estimation variance that can be reached by a Stokes polarimeter in the presence of additive noise.

As a conclusion, we would like to stress that although the three considered metrics are equivalent for polarimeter optimization in the presence of additive noise, the EWV has the simplest physical interpretation since it corresponds to an estimation variance, which has a clear and useful statistical meaning. As a consequence, in contrast to the two other metrics, the EWV can be used for polarimeter optimization in the presence of noise sources with nonadditive, non-Gaussian, or mixed statistics. As discussed above, this problem has already been addressed by optimizing the EWV obtained after application of the pseudoinverse estimator.^{45}^{,}^{46} Although this procedure gives satisfying results in practice,^{48} it is not strictly optimal. Indeed, in the presence of nonadditive and non-Gaussian noise, by virtue of the Cramér-Rao lower bound, the appropriate criterion is the trace of the inverse Fisher information matrix.^{17}^{,}^{18} The value of this metric corresponds to the EWV of an efficient estimator (where “efficient” is meant here in the precise sense used in estimation theory^{49}), whereas in general the pseudoinverse estimator is not efficient. The interesting problem of analyzing the differences between the optimal measurement structures found using a Fisher information-based metric and the spherical 2 designs remains as future work.

## 6.

## Appendix A: Positivity of the Third Term of Eq. (15)

We demonstrate in this section that the third term of the expression of $\mathrm{tr}({\mathbb{G}}^{-1})$ in Eq. (15) is positive definite. Since the matrix ${\mathbb{Q}}^{\mathrm{T}}\mathbb{Q}$ is by definition a positive matrix, the numerator of this term is also positive. We, therefore, need only analyze the denominator. Considering then the singular value decomposition $\mathbb{Q}=\mathbb{UF}{\mathbb{V}}^{\mathrm{T}}$, where $\mathbb{U}$ and $\mathbb{V}$ are unitary matrices and $\mathbb{F}$ is diagonal, it is easily seen that

## Eq. (38)

$$\mathbb{Q}{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}{\mathbb{Q}}^{\mathrm{T}}=\mathbb{UF}{\mathbb{U}}^{\mathrm{T}},$$## Eq. (39)

$${\mathbf{q}}^{\mathrm{T}}{({\mathbb{Q}}^{\mathrm{T}}\mathbb{Q})}^{-1}\mathbf{q}={\mathbf{v}}^{\mathrm{T}}\mathbb{F}\mathbf{v}=\sum _{i=1}^{3}{v}_{i}^{2},$$## Eq. (40)

$$\sum _{i=1}^{3}{v}_{i}^{2}\le \sum _{i=1}^{N}{v}_{i}^{2}={\Vert \mathbf{v}\Vert}^{2}={\Vert \mathbf{r}\Vert}^{2}=N,$$## 7.

## Appendix B: Satisfying Eqs. (16) and (20) with Spherical $t$ Designs

Consider a finite set of points $\{{\mathbf{w}}_{j}\}$ ($j\in [1,N]$), which lie on the surface of the three-dimensional unit sphere. The set of points $\{{\mathbf{w}}_{j}\}$ are said to constitute a spherical $t$ design if for any polynomial function $f(\mathbf{w})$ of order $t$ or lower:

where $\mathrm{d}{\sigma}_{\mathbf{w}}$ is the normalized surface area element of the unit sphere.Proof that Eqs. (16) and (20) can be satisfied using spherical 2 designs follows by showing that we can generate the constraints through appropriate choice of polynomial functions $f(\mathbf{w})$ of second-order degree or less in Eq. (42). Considering first the case $f(\mathbf{w})={w}_{s}$ ($s\in [\mathrm{1,3}]$), substitution into Eq. (42) yields:

where ${w}_{sj}$ is the value of the $s$’th element of ${\mathbf{w}}_{j}$. We can express $\mathbf{w}$ in terms of the usual spherical polar coordinates, i.e., $\mathbf{w}={(\mathrm{sin}\text{\hspace{0.17em}}\theta \text{\hspace{0.17em}}\mathrm{cos}\text{\hspace{0.17em}}\varphi ,\mathrm{sin}\text{\hspace{0.17em}}\theta \text{\hspace{0.17em}}\mathrm{sin}\text{\hspace{0.17em}}\varphi ,\mathrm{cos}\text{\hspace{0.17em}}\theta )}^{\mathrm{T}}$ such that $4\pi \mathrm{d}{\sigma}_{\mathbf{w}}=\mathrm{sin}\text{\hspace{0.17em}}\theta \mathrm{d}\theta \text{\hspace{0.17em}}\mathrm{d}\varphi $. It is then simple to show that $\int {w}_{s}\mathrm{d}{\sigma}_{\mathbf{w}}=0$ for $s\in [\mathrm{1,3}]$ such that Eq. (43) reduces to Eq. (16). Similarly, using the polynomial function $f(\mathbf{w})={w}_{s}{w}_{t}$ for $\{s,t\}=\{\mathrm{1,2},3\}$, Eq. (42) becomes: Evaluating the integral on the right-hand side yields ${\delta}_{st}/3$, such that Eq. (44) reduces to Eq. (20), therefore, completing our proof. Although we have proven that Eqs. (16) and (20) can be satisfied by a spherical 2 design, it is worthwhile to note that it automatically follows that they can also be satisfied by a spherical design of higher order, $t\ge 2$, because a spherical $t$ design is also a $t-1$ design.## 8.

## Appendix C: Previous Derivation

The constraints derived in Sec. 2 through direct minimization of the EWV were first derived by Foreman et al. exploiting a claimed equivalence between minimizing the trace of ${\mathbb{G}}^{-1}$ and maximizing the determinant of $\mathbb{G}$. Specifically, using the definition of the matrix inverse and Jacobi’s formula, it was first shown that the condition number can be expressed in the form:^{35}

## Eq. (45)

$${\kappa}^{2}=\frac{N}{2}\mathrm{tr}[{({\mathbb{W}}^{\mathrm{T}}\mathbb{W})}^{-1}]=\frac{N}{2}\sum _{i=1}^{4}\frac{\partial \text{\hspace{0.17em}}\mathrm{ln}|\mathbb{G}|}{\partial {G}_{ii}},$$## Acknowledgments

The authors would like to thank Dr. A. Favaro for useful discussions. M. R. F. also acknowledges financial support from the Royal Society through a Royal Society University Research Fellowship. The authors declare they have no conflicts of interest.

## References

**,” Adv. Opt. Photonics, 3 205 –271 (2011). https://doi.org/10.1364/AOP.3.000205 AOPAC7 1943-8206 Google Scholar**

*Polarization-resolved nonlinear microscopy: application to structural molecular and biological imaging***,” Astron. Astrophys., 517 A86 (2010). https://doi.org/10.1051/0004-6361/201014167 AAEJAF 0004-6361 Google Scholar**

*Polarimetric observations of comet 67P/Churyumov-Gerasimenko during its 2008–2009 apparition***,” Opt. Express, 20 14090 –14099 (2012). https://doi.org/10.1364/OE.20.014090 OPEXFF 1094-4087 Google Scholar**

*Polarization-resolved second harmonic generation microscopy with a four-channel Stokes polarimeter***,” Chem. Soc. Rev., 33 514 –525 (2004). https://doi.org/10.1039/b201314m CSRVBR 0306-0012 Google Scholar**

*Polarimetric imaging of crystals***,” Opt. Lett., 33 1020 –1022 (2008). https://doi.org/10.1364/OL.33.001020 OPLEDP 0146-9592 Google Scholar**

*Determination of the three-dimensional orientation of single molecules***,” Nature, 459 410 –413 (2009). https://doi.org/10.1038/nature08053 Google Scholar**

*Five-dimensional optical recording mediated by surface plasmons in gold nanorods***,” Opt. Express, 22 26240 –26245 (2014). https://doi.org/10.1364/OE.22.026240 OPEXFF 1094-4087 Google Scholar**

*Polarization-multiplexed encoding at nanometer scales***,” Light: Sci. Appl., 7 17148 (2018). https://doi.org/10.1038/lsa.2017.148 Google Scholar**

*Direct fiber vector eigenmode multiplexing transmission seeded by integrated optical vortex emitters***,” Appl. Opt., 43 274 –282 (2004). https://doi.org/10.1364/AO.43.000274 APOPAI 0003-6935 Google Scholar**

*Target detection with a liquid-crystal-based passive Stokes polarimeter***,” Adv. Mater., 11 895 –905 (1999). https://doi.org/10.1002/(ISSN)1521-4095 ADVMEW 0935-9648 Google Scholar**

*Polarized luminescence from oriented molecular materials***,” Nature, 514 76 –79 (2014). https://doi.org/10.1038/nature13680 Google Scholar**

*Evanescent-wave and ambient chiral sensing by signal-reversing cavity ringdown polarimetry***,” Appl. Opt., 45 5470 –5478 (2006). https://doi.org/10.1364/AO.45.005470 APOPAI 0003-6935 Google Scholar**

*Dual-field imaging polarimeter using liquid crystal variable retarders***,” J. Opt. Soc. Am. A, 31 1013 –1022 (2014). https://doi.org/10.1364/JOSAA.31.001013 JOAOD6 0740-3232 Google Scholar**

*Generalized channeled polarimetry***,” Appl. Opt., 37 5938 –5944 (1998). https://doi.org/10.1364/AO.37.005938 APOPAI 0003-6935 Google Scholar**

*Broadband division-of-amplitude polarimeter based on uncoated prisms***,” Opt. Express, 16 15212 –15227 (2008). https://doi.org/10.1364/OE.16.015212 OPEXFF 1094-4087 Google Scholar**

*A priori information and optimisation in polarimetry***,” Phys. Rev. A, 82 043835 (2010). https://doi.org/10.1103/PhysRevA.82.043835 Google Scholar**

*Information and resolution in electromagnetic optical systems***,” University of Arizona, (2015). Google Scholar**

*Matrix structure for information driven polarimeter design***,” J. Opt. Soc. Am., 20 955 –958 (2003). https://doi.org/10.1364/JOSAA.20.000955 JOSAAH 0030-3941 Google Scholar**

*Optimal beam splitters for the division-of-amplitude photopolarimeter***,” Opt. Eng., 34 1656 –1658 (1995). https://doi.org/10.1117/12.202098 Google Scholar**

*Optimum angles for a polarimeter: part II***,” Opt. Eng., 34 1651 –1655 (1995). https://doi.org/10.1117/12.202093 Google Scholar**

*Optimum angles for a polarimeter: part I***,” Appl. Opt., 41 619 –630 (2002). https://doi.org/10.1364/AO.41.000619 APOPAI 0003-6935 Google Scholar**

*Design of optimal polarimeters: maximization of signal-to-noise ratio and minimization of systematic error***,” Opt. Lett., 25 802 –804 (2000). https://doi.org/10.1364/OL.25.000802 OPLEDP 0146-9592 Google Scholar**

*Optimization of retardance for a complete Stokes polarimeter***,” Opt. Express, 18 9815 –9830 (2010). https://doi.org/10.1364/OE.18.009815 OPEXFF 1094-4087 Google Scholar**

*Optimization and performance criteria of a Stokes polarimeter based on two variable retarders***,” J. Opt. Soc. Am. A, 5 681 –689 (1988). https://doi.org/10.1364/JOSAA.5.000681 JOAOD6 0740-3232 Google Scholar**

*General analysis and optimization of the four-detector photopolarimeter***,” Opt. Eng., 41 965 –972 (2002). https://doi.org/10.1117/1.1467361 Google Scholar**

*Optimisation and structuring of the instrument matrix for polarimetric measurements***,” Appl. Opt., 41 2488 –2493 (2002). https://doi.org/10.1364/AO.41.002488 APOPAI 0003-6935 Google Scholar**

*Optimization of a dual-rotating-retarder Mueller matrix polarimeter***,” Opt. Express, 16 11589 –11603 (2008). https://doi.org/10.1364/OE.16.011589 OPEXFF 1094-4087 Google Scholar**

*Optimization of Mueller matrix polarimeters in the presence of error sources***,” Proc. SPIE, 2265 314 –326 (1995). https://doi.org/10.1117/12.186680 PSISDG 0277-786X Google Scholar**

*Optimum angles for a Mueller matrix polarimeter***,” Opt. Express, 20 20466 –20481 (2012). https://doi.org/10.1364/OE.20.020466 OPEXFF 1094-4087 Google Scholar**

*Optimum selection of input polarization states in determining the sample Mueller matrix: a dual photoelastic polarimeter approach***,” J. Opt. Soc. Am. A, 34 1054 –1062 (2017). https://doi.org/10.1364/JOSAA.34.001054 JOAOD6 0740-3232 Google Scholar**

*Optimal configuration of static Mueller imagers for target detection***,” Opt. Lett., 39 6759 –6762 (2014). https://doi.org/10.1364/OL.39.006759 OPLEDP 0146-9592 Google Scholar**

*Contrast optimization in broadband passive polarimetric imaging***,” J. Opt. Soc. Am. A, 33 9 –16 (2016). https://doi.org/10.1364/JOSAA.33.000009 JOAOD6 0740-3232 Google Scholar**

*Optimal configuration of static polarization imagers for target detection***,” Phys. Rev. Lett., 115 263901 (2015). https://doi.org/10.1103/PhysRevLett.115.263901 PRLTAO 0031-9007 Google Scholar**

*Optimal frames for polarisation state reconstruction***,” Geometria Dedicata, 6 363 –388 (1977). https://doi.org/10.1007/BF03187604 Google Scholar**

*Spherical codes and designs***,” Opt. Lett., 42 2153 –2156 (2017). https://doi.org/10.1364/OL.42.002153 OPLEDP 0146-9592 Google Scholar**

*Optimal Mueller matrix estimation in the presence of additive and Poisson noise for any number of illumination and analysis states***,” Graphs Combi., 6 369 –372 (1990). https://doi.org/10.1007/BF01787704 Google Scholar**

*A construction of spherical 2-design***,” Geometriae Dedicata, 43 167 –179 (1992). https://doi.org/10.1007/BF00147866 GEMDAT 0046-5755 Google Scholar**

*Construction of spherical t-designs***,” Discrete Comput. Geom., 15 429 –441 (1996). https://doi.org/10.1007/BF02711518 Google Scholar**

*McLaren’s improved snub cube and other new spherical designs in three dimensions***,” Proc. SPIE, 4133 75 –81 (2000). https://doi.org/10.1117/12.406613 PSISDG 0277-786X Google Scholar**

*Figures of merit for complete Stokes polarimeter optimization***,” Opt. Express, 16 2091 –2108 (2008). https://doi.org/10.1364/OE.16.002091 OPEXFF 1094-4087 Google Scholar**

*Noise reduction in a laser polarimeter based on discrete waveplate rotations***,” Opt. Lett., 25 1198 –1200 (2000). https://doi.org/10.1364/OL.25.001198 OPLEDP 0146-9592 Google Scholar**

*Noise equalization in Stokes parameter images obtained by use of variable-retardance polarimeters***,” Opt. Lett., 34 647 –649 (2009). https://doi.org/10.1364/OL.34.000647 OPLEDP 0146-9592 Google Scholar**

*Noise minimization and equalization for Stokes polarimeters in the presence of signal-dependent Poisson shot noise***,” Opt. Lett., 41 5772 –5775 (2016). https://doi.org/10.1364/OL.41.005772 OPLEDP 0146-9592 Google Scholar**

*Equalized estimation of Stokes parameters in the presence of Poisson noise for any number of polarization analysis states***,” Opt. Express, 26 29968 –29982 (2018). https://doi.org/10.1364/OE.26.029968 OPEXFF 1094-4087 Google Scholar**

*Polarimetric precision of micropolarizer grid-based camera in the presence of additive and Poisson shot noise***,” Opt. Lett., 42 1899 –1902 (2017). https://doi.org/10.1364/OL.42.001899 OPLEDP 0146-9592 Google Scholar**

*Performance comparison of pseudo-inverse and maximum-likelihood estimators of Stokes parameters in the presence of Poisson noise for spherical design-based measurement structures*## Biography

**Matthew R. Foreman** received his MPhys degree from the University of Oxford in 2006 and his PhD from Imperial College London in 2010. He has held research posts at the UK National Physical Laboratory (Teddington) and the Max Planck Institute for the Science of Light (Erlangen), where he held an Alexander von Humboldt Fellowship. Currently, he is a Royal Society University Research Fellow at Imperial College London. His research interests include theoretical aspects of nanophotonics, plasmonics, polarimetry, and random scattering and sensing.

**François Goudail** graduated from the École Supérieure d’Optique (Orsay) in 1992 and received his PhD in 1997 from the University of Aix-Marseille III. He was an associate professor at Fresnel Institute (Marseille) until 2005. He is now a professor at the Institut d’Optique Graduate School (Palaiseau). His research topics include information extraction in images from different types of passive and active sensors (hyperspectral, SAR, polarimetric), wavefront engineering and joint design of optical systems, and image processing algorithms.