34 A map domain model to predict crosstalk systematic effects on maps and power-spectra
LiteBIRD, Crosstalk, Systematics, Mapmaking, Power spectra, Scanning strategy, Flat-sky approximation, Curved sky, Spin-weighted fields, Rotation group SO(3)
34 A map domain model to predict crosstalk systematic effects on maps and power-spectra
Crosstalk systematics are a time-domain problem. It mixes the detector TOD according to1,
\[\tilde{d}^{i} = \sum_{j} X^{i j} d^{j} . \label{crosstalk_tod}\]
However, it is equivalent to a map-domain problem if the following holds:
Assumption 1. Assumptions of treating crosstalk systematics as a map-domain problem.
- In scanning strategy:
- each detector has a uniform map coverage (that is, the weight-map per detector is constant over the sphere);
- (for fixed orientation) the focal plane has fixed orientation with respect to the sky locally;
- (or for isotropic orientation) the scanning strategy is isotropic in focal plane orientation2;
- The mapmaking operation:
- is linear in the observed TOD, and
- depends on the pointing matrix only;
- there is no noise such that the mapmaking operation can be considered as the inverse transform of the observation operation;
- where the resultant map is a weighted sum of individual detector maps (independence of detectors);
- each detector can resolve the \(I, Q, U\) components independently;
- has infinite resolution, that is, in the continuum limit;
- In detector layout,
- all detectors are fixed on the focal plane;
- the focal plane angular scale is \(\ll 1\).
Proof. With Assumption 1, it can be proved that crosstalk systematics are map-domain systematics.
Observation and mapmaking are often written as discrete sums:
\[\begin{align*} d_{t} & = \sum_{p} A_{t p} m_{p} + n_{t} \\ \hat{m}_{p} & = \sum_{t} W_{p t} d_{t} \end{align*}\]
Here, this is made abstract under the continuum limit (2f), and noise is ignored using assumption (2c):
\[\begin{align*} \mathcal{O} \mathopen{} \left [ A \right ] \mathclose{} \mathopen{} \left \{ m \right \} \mathclose{} \mathopen{} \left ( t \right ) \mathclose{} & \equiv d \mathopen{} \left ( t \right ) \mathclose{} & & = \oint A \mathopen{} \left ( t, \mathbf{r} \right ) \mathclose{} m \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} \,\mathrm{d}^{2} \mathbf{r} \\ \mathcal{M} \mathopen{} \left [ A \right ] \mathclose{} \mathopen{} \left \{ d \right \} \mathclose{} \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} & \equiv m \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} & & = \int W \mathopen{} \left [ A \right ] \mathclose{} \mathopen{} \left ( \mathbf{r}, t \right ) \mathclose{} d \mathopen{} \left ( t \right ) \mathclose{} \,\mathrm{d} t \end{align*}\]
where \(\mathcal{O} \mathopen{} \left [ A \right ] \mathclose{} , \mathcal{M} \mathopen{} \left [ A \right ] \mathclose{}\)3 are linear transformations of \(m, d\) respectively (assumption 2a) that depend on \(A\) (assumption 2b). They are inverse transformations of each other (assumption 2c). Note that while \(\mathcal{O} \mathopen{} \left [ A \right ] \mathclose{}\) is linear in \(A\), \(\mathcal{M} \mathopen{} \left [ A \right ] \mathclose{}\) is not required to be linear in \(A\), and in practice they are not.
For the map \(m\) , from assumption (2e), we can consider either \(m = T\) as a real scalar field, or \(m = \mathcal{Q} = Q + \mathrm{i} U\) as a complex spin-\(2\) field. Explicitly, we can write \(A \mathopen{} \left ( t, \mathbf{r} \right ) \mathclose{} = \mathrm{e}^{- s \mathrm{i} \alpha \mathopen{} \left ( t \right ) \mathclose{}} \delta^{\left( 2 \right)} \mathopen{} \left ( \mathbf{r} - \mathbf{r} \mathopen{} \left ( t \right ) \mathclose{} \right ) \mathclose{}\) where \(\alpha\) is the orientation of the detector, \(s\) is the spin of the field ( \(s^{T} = 0 , s^{\mathcal{Q}} = 2\) ). This can be used with assumption (3a) to rewrite the detector pointing with respect to the boresight pointing \(A^{i} \mathopen{} \left ( t, \mathbf{r} \right ) \mathclose{} = \mathrm{e}^{- s \mathrm{i} \Delta \alpha^{i}} A \mathopen{} \left ( t, R_{i}^{-1} \mathbf{r} \right ) \mathclose{}\), where \(R_{i} : \mathbf{r} \mathopen{} \left ( t \right ) \mathclose{} \mapsto \mathbf{r}^{i} \mathopen{} \left ( t \right ) \mathclose{} , \Delta \alpha^{i} \equiv \alpha^{i} \mathopen{} \left ( t \right ) \mathclose{} - \alpha \mathopen{} \left ( t \right ) \mathclose{}\)4 are independent of \(t\) by assumption (3a)5. With this, note that we can rewrite an observation at \(A^i\) to an observation at \(A\):
\[\begin{align*} \mathcal{O} \mathopen{} \left [ A^{i} \right ] \mathclose{} \mathopen{} \left \{ m \right \} \mathclose{} \mathopen{} \left ( t \right ) \mathclose{} & = \oint A^{i} \mathopen{} \left ( t, \mathbf{r} \right ) \mathclose{} m \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} \,\mathrm{d}^{2} \mathbf{r} \\ & = \mathrm{e}^{- s \mathrm{i} \Delta \alpha^{i}} \oint A \mathopen{} \left ( t, R_{i}^{-1} \mathbf{r} \right ) \mathclose{} m \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} \,\mathrm{d}^{2} \mathbf{r} \\ & = \mathrm{e}^{- s \mathrm{i} \Delta \alpha^{i}} \oint A \mathopen{} \left ( t, \mathbf{r}^{\prime} \right ) \mathclose{} m \mathopen{} \left ( R_{i} \mathbf{r}^{\prime} \right ) \mathclose{} \,\mathrm{d}^{2} \mathbf{r}^{\prime} \\ & = \mathcal{O} \mathopen{} \left [ A \right ] \mathclose{} \mathopen{} \left \{ \mathcal{R}^{-1}_{i} \mathopen{} \left \{ m \right \} \mathclose{} \right \} \mathclose{} \mathopen{} \left ( t \right ) \mathclose{} \end{align*}\]
where \(\mathcal{R} \mathopen{} \left \{ m \right \} \mathclose{} \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} \equiv \mathrm{e}^{s \mathrm{i} \Delta \alpha^{}} m \mathopen{} \left ( R^{-1} \mathbf{r} \right ) \mathclose{}\) is the active transform of the spin-\(s\) field \(m\)6 (and hence \(\mathcal{R}^{-1} \mathopen{} \left \{ m \right \} \mathclose{} \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} \equiv \mathrm{e}^{- s \mathrm{i} \Delta \alpha^{}} m \mathopen{} \left ( R \mathbf{r} \right ) \mathclose{}\) is the passive transform.)
The resultant map per detector about the crosstalk in TOD would then be,
\[\begin{align} \tilde{m}^{i} & \equiv \mathcal{M} \mathopen{} \left [ A^{i} \right ] \mathclose{} \mathopen{} \left \{ \tilde{d}^{i} \right \} \mathclose{} \nonumber \\ \mathcal{O} \mathopen{} \left [ A^{i} \right ] \mathclose{} \mathopen{} \left \{ \tilde{m}^{i} \right \} \mathclose{} & = \tilde{d}^{i} = \sum_{j} X^{i j} \mathcal{O} \mathopen{} \left [ A^{j} \right ] \mathclose{} \mathopen{} \left \{ m \right \} \mathclose{} \nonumber \\ \mathcal{O} \mathopen{} \left [ A \right ] \mathclose{} \mathopen{} \left \{ \mathcal{R}^{-1}_{i} \mathopen{} \left \{ \tilde{m}^{i} \right \} \mathclose{} \right \} \mathclose{} & = \sum_{j} X^{i j} \mathcal{O} \mathopen{} \left [ A \right ] \mathclose{} \mathopen{} \left \{ \mathcal{R}^{-1}_{j} \mathopen{} \left \{ m \right \} \mathclose{} \right \} \mathclose{} \nonumber \\ \mathcal{R}^{-1}_{i} \mathopen{} \left \{ \tilde{m}^{i} \right \} \mathclose{} & = \sum_{j} X^{i j} \mathcal{R}^{-1}_{j} \mathopen{} \left \{ m \right \} \mathclose{} \nonumber \\ \tilde{m}^{i} & = \sum_{j} X^{i j} \mathcal{R}^{-1}_{i \rightarrow j} \mathopen{} \left \{ m \right \} \mathclose{} \text{ , or explicitly,} \label{eq:crosstalk_map_pf} \tag{34.1}\\ \mathrm{e}^{- s \mathrm{i} \Delta \alpha^{i}} \tilde{m}^{i} \mathopen{} \left ( R_{i} \mathbf{r} \right ) \mathclose{} & = \sum_{j} X^{i j} \mathrm{e}^{- s \mathrm{i} \Delta \alpha^{j}} m \mathopen{} \left ( R_{j} \mathbf{r} \right ) \mathclose{} \nonumber \\ \tilde{m}^{i} \mathopen{} \left ( \mathbf{r} ' \right ) \mathclose{} & = \sum_{j} X^{i j} \mathrm{e}^{- s \mathrm{i} \alpha^{i \rightarrow j}} m \mathopen{} \left ( R_{j} R_{i}^{-1} \mathbf{r} ' \right ) \mathclose{} \nonumber \\ \tilde{m}^{i} \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} & = \sum_{j} X^{i j} \mathrm{e}^{- s \mathrm{i} \alpha^{i \rightarrow j}} m \mathopen{} \left ( R_{i \rightarrow j} \mathbf{r} \right ) \mathclose{} \nonumber \end{align}\]
where \(\alpha^{i \rightarrow j} \equiv \Delta \alpha^{j} - \Delta \alpha^{i} , R_{i \rightarrow j} \equiv R_{j} R_{i}^{-1} , \mathcal{R}_{i \rightarrow j} \equiv \mathcal{R}_{j} \mathcal{R}_{i}^{-1}\). Finally, from assumption (2d) & (1a), the resultant map obtained in the mapmaking operation is,
\[\begin{align} m^{\prime} \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} & = \sum_{i} w^{i} \tilde{m}^{i} \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} , \quad \sum_{i} w^{i} = 1 \nonumber \\ & = \mathcal{X} m \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathcal{R}_{i \rightarrow j}^{-1} \mathopen{} \left \{ m \right \} \mathclose{} \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} \label{crosstalkMapAbstract} \tag{34.2}\\ & = \mathcal{X} m \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathrm{e}^{- s \mathrm{i} \alpha^{i \rightarrow j}} m \mathopen{} \left ( R_{i \rightarrow j} \mathbf{r} \right ) \mathclose{} \label{crosstalkMap} \tag{34.3}\end{align}\]
where \(\mathcal{X}^{i j} \equiv w^{i} X^{i j} , \mathcal{X} \equiv \operatorname{tr} \mathopen{} \left ( \mathcal{X}^{i j} \right ) \mathclose{}\). Note that in the limit that the crosstalk matrix is identity on the diagonal, \(\mathcal{X} = 1\) by definition. That is, the first term above recovers the original map, whereas the 2nd term is a first order correction as we normally have \(X^{i j} \underset{i \neq j}{ \ll } 1\).\(\quad\Box\)
This section is closed by observing when the assumptions are approximately true. Note that the assumption that mapmaking is linear in \(d\) is true for most mapmaking methods,7 while the assumption that it depends on \(A\) only is valid for mapmaking methods 1–3 of Tegmark,8 which include binned mapmaking (also known as naïve mapmaking) and a maximal likelihood mapmaking.
34.1 Flat sky, fixed orientation, isotropic map coverage
To apply (34.3) under flat-sky approximation with a fixed focal plane orientation, one substitutes \(R_{i} \rightarrow T_{i}\), the translation operator, and reinterprets \(\mathbf{r}\) as the 2-dimensional angle in radians. This implies \(R_{i \rightarrow j} \mathbf{r} \rightarrow T_{i \rightarrow j} \mathbf{r} = \mathbf{r} + \mathbf{r}^{i \rightarrow j}\),
\[\begin{align*} m^{\prime} \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} & = \mathcal{X} m \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathrm{e}^{- s \mathrm{i} \alpha^{i \rightarrow j}} m \mathopen{} \left ( \mathbf{r} + \mathbf{r}^{i \rightarrow j} \right ) \mathclose{} \\ & = \left( m \ast P \right) \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} \equiv \int_{\mathbb{R}^{2}} m \mathopen{} \left ( \mathbf{r}^{\prime} \right ) \mathclose{} P \mathopen{} \left ( \mathbf{r} - \mathbf{r}^{\prime} \right ) \mathclose{} \,\mathrm{d}^{2} \mathbf{r} \text{, where} \\ P \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} & = \mathcal{X} \delta^{\left( 2 \right)} \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathrm{e}^{- s \mathrm{i} \alpha^{i \rightarrow j}} \delta^{\left( 2 \right)} \mathopen{} \left ( \mathbf{r} + \mathbf{r}^{i \rightarrow j} \right ) \mathclose{} \end{align*}\]
is the point spread function9. By convolution theorem10, \(m^{\prime} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} = P \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} m \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{}\), where
\[\begin{align*} \mathcal{F} \mathopen{} \left \{ P \right \} \mathclose{} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & = \int_{\mathbb{R}^{2}} P \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} \mathrm{e}^{- \mathrm{i} \mathbf{r} \cdot \mathbf{\ell}} \,\mathrm{d}^{2} \mathbf{r} \\ {}_{{s}} P \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & = \mathcal{X} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathrm{e}^{\mathrm{i} \left( \mathbf{r}^{i \rightarrow j} \cdot \mathbf{\ell} - s \alpha^{i \rightarrow j} \right)} \end{align*}\]
Hence, this is how the fields are transformed under crosstalk,
\[\begin{align*} T^{\prime} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & = {}_{{0}} P \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} T \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \\ \mathcal{Q}^{\prime} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & = {}_{{2}} P \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \mathcal{Q} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \end{align*}\]
With the following standard equations to propagate from \(\mathcal{Q}\) to \(E, B\), with \(\mathcal{E} \equiv E + \mathrm{i} B\),
\[\begin{align*} \mathcal{E} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & = \mathcal{Q} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \mathrm{e}^{-2 \mathrm{i} \phi_{\mathbf{\ell}}} \\ E \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & = \frac{1}{2} \left[ \mathcal{E} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} + \bar{\mathcal{E}} \mathopen{} \left ( - \mathbf{\ell} \right ) \mathclose{} \right] \\ B \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & = \frac{1}{2 \mathrm{i}} \left[ \mathcal{E} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} - \bar{\mathcal{E}} \mathopen{} \left ( - \mathbf{\ell} \right ) \mathclose{} \right] \end{align*}\]
With some derivation and simplification, and defining
\[\mathcal{A} ^\mathsf{T} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} = \begin{bmatrix} T \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & E \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & B \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \end{bmatrix}\]
and expanding the sum inside \(P\)11,
\[\begin{align} \mathcal{A}^{\prime} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & = \begin{bmatrix} {}_{{0}} P \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & 0 & 0 \\ 0 & \Re \mathopen{} \left \{ {}_{{2}} P \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \right \} \mathclose{} & - \Im \mathopen{} \left \{ {}_{{2}} P \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \right \} \mathclose{} \\ 0 & \Im \mathopen{} \left \{ {}_{{2}} P \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \right \} \mathclose{} & \Re \mathopen{} \left \{ {}_{{2}} P \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \right \} \mathclose{} \end{bmatrix} \mathcal{A} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \label{APA} \tag{34.4}\\ & = \left[ \mathcal{X} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathcal{I}^{i j} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \right] \mathcal{A} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \label{ACrosstalk} \tag{34.5}\\ \mathcal{I} \mathopen{} \left ( \mathbf{r} \cdot \mathbf{\ell}, \alpha \right ) \mathclose{} & \equiv \begin{bmatrix} \mathrm{e}^{\mathrm{i} \mathbf{r} \cdot \mathbf{\ell}} & 0 & 0 \\ 0 & \cos \mathopen{} \left ( \mathbf{r} \cdot \mathbf{\ell} - 2 \alpha \right ) \mathclose{} & - \sin \mathopen{} \left ( \mathbf{r} \cdot \mathbf{\ell} - 2 \alpha \right ) \mathclose{} \\ 0 & \sin \mathopen{} \left ( \mathbf{r} \cdot \mathbf{\ell} - 2 \alpha \right ) \mathclose{} & \cos \mathopen{} \left ( \mathbf{r} \cdot \mathbf{\ell} - 2 \alpha \right ) \mathclose{} \end{bmatrix} \nonumber \\ \mathcal{I}^{i j} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & \equiv \mathcal{I} \mathopen{} \left ( \mathbf{r}^{i \rightarrow j} \cdot \mathbf{\ell}, \alpha^{i \rightarrow j} \right ) \mathclose{} \nonumber \end{align}\]
Writing the usual power spectrum \(C_\ell\) in terms of \(\mathcal{A}\),
\[\begin{align*} \mathbb{C} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} & \equiv \mathcal{A} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \mathcal{A} ^\dagger \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \\ \mathbb{C}_{\ell} & \equiv \left\langle \mathbb{C} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \right\rangle_{\mathbf{\hat{\ell}}} \equiv \left\langle \mathcal{A} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \mathcal{A} ^\dagger \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \right\rangle_{\mathbf{\hat{\ell}}} \\ & = \begin{bmatrix} C^{TT}_{\ell} & C^{TE}_{\ell} & C^{TB}_{\ell} \\ C^{ET}_{\ell} & C^{EE}_{\ell} & C^{EB}_{\ell} \\ C^{BT}_{\ell} & C^{BE}_{\ell} & C^{BB}_{\ell} \end{bmatrix} \end{align*}\]
Expanding using (34.5), we have12,
\[\begin{align} \mathbb{C}^{\prime}_{\ell} & = \mathcal{X}^{2} \mathbb{C}_{\ell} + \mathcal{X} \sum_{i , j \neq i} \mathcal{X}^{i j} \left\langle \mathcal{I}^{i j} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \mathbb{C} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} + \text{ h.c.} \right\rangle_{\mathbf{\hat{\ell}}} + \dots \nonumber \\ & = \mathcal{X}^{2} \mathbb{C}_{\ell} + \mathcal{X} \sum_{i , j \neq i} \mathcal{X}^{i j} \left( \mathcal{I}^{i j}_{\ell} \mathbb{C}_{\ell} + \text{ h.c.} \right) + \dots \label{CIC} \tag{34.6}\\ \mathcal{I}^{i j}_{\ell} & \equiv \left\langle \mathcal{I}^{i j} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \right\rangle_{\mathbf{\hat{\ell}}} \nonumber \\ & = J_{0} \mathopen{} \left ( r^{i \rightarrow j} \ell \right ) \mathclose{} \begin{bmatrix} 1 & 0 & 0 \\ 0 & \cos 2 \alpha^{i \rightarrow j} & \sin 2 \alpha^{i \rightarrow j} \\ 0 & - \sin 2 \alpha^{i \rightarrow j} & \cos 2 \alpha^{i \rightarrow j} \end{bmatrix} \nonumber \\ & = J_{0} \mathopen{} \left ( r^{i \rightarrow j} \ell \right ) \mathclose{} R^{-1}_{T} \mathopen{} \left ( 2 \alpha^{i \rightarrow j} \right ) \mathclose{} \nonumber \\ \mathbb{C}^{\prime}_{\ell} & = \mathcal{X}^{2} \mathbb{C}_{\ell} + \mathcal{X} \sum_{i , j \neq i} \mathcal{X}^{i j} J_{0} \mathopen{} \left ( r^{i \rightarrow j} \ell \right ) \mathclose{} \left[ R^{-1}_{T} \mathopen{} \left ( 2 \alpha^{i \rightarrow j} \right ) \mathclose{} \mathbb{C}_{\ell} + \text{ h.c.} \right] + \dots \label{CXJRC} \tag{34.7}\end{align}\]
where the 2nd line has the statistical average applied on the inner \(\mathbb{C} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{}\), \(J_{\alpha}\) is the Bessel functions of the first kind, , \(R_T\) is the rotational matrix over the \(T\)-axis, and only up to first-order in \(\mathcal{X}^{i j}\) is kept. Note that it means that to first order, each of the crosstalk \(i, j\) components is modulated by a Bessel function and rotated in the \(EB\)-plane. Unlike a global mis-calibration, this rotation is not global as the individual contributions are different.
34.2 Flat sky, isotropic orientation, isotropic map coverage
In the previous section, the focal plane is assumed to be fixed in orientation with respect to the (flat-)sky when we replace \(R\) by \(T\). However, in a more realistic situation, it will have different orientations as it scans across the same point on the sky. That effect can be considered analytically if the scanning strategy is assumed to be isotropic in the focal plane orientation (assumption (1c)). Due to this isotropic averaging effect, the effective PSF is now the azimuthal average of the original PSF,
\[\begin{align*} P^{\prime} \mathopen{} \left ( r \right ) \mathclose{} & \equiv \left\langle P \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} \right\rangle_{\mathbf{\hat{r}}} = \frac{1}{2 \pi} \int_{0}^{2 \pi} P \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} \,\mathrm{d} \theta \\ & = \frac{1}{2 \pi} \int_{0}^{2 \pi} \mathcal{X} \frac{\delta \mathopen{} \left ( r \right ) \mathclose{}}{2 \pi r} \\ & + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathrm{e}^{- s \mathrm{i} \alpha^{i \rightarrow j}} \frac{\delta \mathopen{} \left ( r + r^{i \rightarrow j} \right ) \mathclose{} \delta \mathopen{} \left ( \theta + \theta^{i \rightarrow j} \right ) \mathclose{}}{r} \,\mathrm{d} \theta \\ & = \frac{1}{2 \pi r} \left[ \mathcal{X} \delta \mathopen{} \left ( r \right ) \mathclose{} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathrm{e}^{- s \mathrm{i} \alpha^{i \rightarrow j}} \delta \mathopen{} \left ( r + r^{i \rightarrow j} \right ) \mathclose{} \right] \end{align*}\]
\[\begin{align} \mathcal{F} \mathopen{} \left \{ P^{\prime} \right \} \mathclose{} \mathopen{} \left ( \ell \right ) \mathclose{} & = \int_{0}^{2 \pi} \int_{0}^{\infty} P^{\prime} \mathopen{} \left ( r \right ) \mathclose{} \mathrm{e}^{- \mathrm{i} r \ell \cos \theta} r \,\mathrm{d} r \,\mathrm{d} \theta \nonumber \\ {}_{{s}} P^{\prime}_{\ell} & = \mathcal{X} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathrm{e}^{- s \mathrm{i} \alpha^{i \rightarrow j}} J_{0} \mathopen{} \left ( \ell r^{i \rightarrow j} \right ) \mathclose{} \label{sPlAveraged} \tag{34.8}\end{align}\]
Note that as \(P\) is azimuthally symmetric, it depends on \(\ell\) only. Substituting (34.8) into (34.4),
\[\begin{equation} \mathcal{A}^{\prime} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} = \left[ \mathcal{X} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathcal{I}^{i j}_{\ell} \right] \mathcal{A} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{} \label{ACrosstalkAveraged} \tag{34.9}\end{equation} \]
Following the same procedure as in last section, it is found that the resultant power-spectra are the same as those in (34.7) up to first order.
Comparing the results of this section with the last, it is found that the 2d-power-spectra in the previous section (34.5) is not statistically isotropic, while here (34.9) is. Furthermore, while the resultant 1d-power-spectra are the same up to 1st order, the 2nd order term is different, where in the previous section it is an average of products of \(\mathcal{I}^{i j} \mathopen{} \left ( \mathbf{\ell} \right ) \mathclose{}\), here it is a product of \(\mathcal{I}^{i j}_{\ell}\) which is already averaged. This means the 2nd order term can be written explicitly,
\[\begin{multline*} \mathbb{C}^{\prime}_{\ell} = \mathcal{X}^{2} \mathbb{C}_{\ell} + \mathcal{X} \sum_{i , j \neq i} \mathcal{X}^{i j} J_{0} \mathopen{} \left ( r^{i \rightarrow j} \ell \right ) \mathclose{} \left[ R^{-1}_{T} \mathopen{} \left ( 2 \alpha^{i \rightarrow j} \right ) \mathclose{} \mathbb{C}_{\ell} + \text{ h.c.} \right] \\ + \sum_{i , j \neq i , m , n \neq m} \mathcal{X}^{i j} \mathcal{X}^{m n} J_{0} \mathopen{} \left ( r^{i \rightarrow j} \ell \right ) \mathclose{} J_{0} \mathopen{} \left ( r^{m \rightarrow n} \ell \right ) \mathclose{} \times \\ R^{-1}_{T} \mathopen{} \left ( 2 \alpha^{i \rightarrow j} \right ) \mathclose{} \mathbb{C}_{\ell} R_{T} \mathopen{} \left ( 2 \alpha^{m \rightarrow n} \right ) \mathclose{} \end{multline*} \]
34.3 Curved sky, isotropic orientation, isotropic map coverage
To generalize the above results to curved sky, it is necessary to digress to some definitions and theorems.
First, CMB power-spectra in curved sky are expanded in spin-weighted spherical harmonics (SWSH). Using a view that is popularized by,13 define
\[\begin{align*} {}_{s} \mathcal{Y}_{\ell m} \mathopen{} \left ( \phi, \theta, \psi \right ) \mathclose{} & \equiv \mathrm{e}^{- \mathrm{i} s \psi} {}_{s} Y_{\ell m} \mathopen{} \left ( \theta, \phi \right ) \mathclose{} \\ {}_{s} \mathcal{Y}_{\ell m} \mathopen{} \left ( R \right ) \mathclose{} & \equiv \mathrm{e}^{- \mathrm{i} s \psi} {}_{s} Y_{\ell m} \mathopen{} \left ( R \mathbf{\hat{z}} \right ) \mathclose{} ,\quad R \equiv R \mathopen{} \left ( \phi, \theta, \psi \right ) \mathclose{} \\ & = \left( -1 \right)^{m} \sqrt{\frac{2 \ell + 1}{4 \pi}} D^{\left( \ell \right)}_{- m s} \mathopen{} \left ( R \right ) \mathclose{} \end{align*}\]
\({}_{s} Y_{l m}\) is a spin-\(s\) function on the 2-sphere, that is, it changes upon a local transformation. \({}_{s} \mathcal{Y}_{l m}\) is defined instead as a scalar function such that \({}_{s} \mathcal{Y}_{l m}\): \(\operatorname{SO(3)} \rightarrow \mathbb{C}\)14. \(R\) is a rotation parameterized by the Euler angles15, and \(D\) is the Wigner-\(D\) matrix in Wigner’s original sign convention.
The key to deriving a curved-sky solution is to define a delta function on this \(\operatorname{SO} \mathopen{} \left ( 3 \right ) \mathclose{}\) space. Rewriting the delta function in \(\mathbb{R}^{2}\) as \(\delta^{R^{\prime}} \mathopen{} \left ( \mathbf{\hat{n}} \right ) \mathclose{} \equiv \delta^{\left( 2 \right)} \mathopen{} \left ( R^{\prime} \mathbf{\hat{n}} - \mathbf{\hat{z}} \right ) \mathclose{}\) where \(R'\) is the pointing. Define a \(2\)-dimensional spin-\(s\) delta function in \(\operatorname{SO} \mathopen{} \left ( 3 \right ) \mathclose{}\),
\[{}_{{s}} \delta^{R^{\prime}} \mathopen{} \left ( R \right ) \mathclose{} \equiv \delta^{\left( 2 \right)} \mathopen{} \left ( R^{\prime} R = R_{z} \mathopen{} \left ( \alpha \right ) \mathclose{} \right ) \mathclose{} \mathrm{e}^{- \mathrm{i} s \psi}\]
Note that \(α\) is an arbitrary angle, and hence after an integral over \(\operatorname{SO} \mathopen{} \left ( 3 \right ) \mathclose{}\), there remains a \(1\)-dimensional integral.
The convolution with this delta function will then be16,
\[\begin{align*} \left( {}_{{s}} f \ast {}_{{s}} \delta^{R^{\prime}} \right) \mathopen{} \left ( R \right ) \mathclose{} & = \frac{1}{2 \pi} \int_{R^{\prime \prime} \in \operatorname{SO} \mathopen{} \left ( 3 \right ) \mathclose{}} {}_{{s}} f \mathopen{} \left ( R^{\prime \prime} \right ) \mathclose{} {}_{{s}} \delta^{R^{\prime}} \mathopen{} \left ( R^{\prime \prime -1} R \right ) \mathclose{} \,\mathrm{d} R^{\prime \prime} \\ {}_{{s}} \delta^{R^{\prime}} \mathopen{} \left ( R^{\prime \prime -1} R \right ) \mathclose{} & = \delta^{\left( 2 \right)} \mathopen{} \left ( R^{\prime} R^{\prime \prime -1} R = R_{z} \mathopen{} \left ( \alpha \right ) \mathclose{} \right ) \mathclose{} \mathrm{e}^{- \mathrm{i} s \psi^{\prime \prime \prime}} \\ R^{\prime \prime \prime} & \equiv R^{\prime \prime -1} R \\ & = R^{\prime -1} R_{z} \mathopen{} \left ( \alpha \right ) \mathclose{} \\ R \mathopen{} \left ( \phi, \theta, \psi \right ) \mathclose{} & = R_{z} \mathopen{} \left ( \phi \right ) \mathclose{} R_{y} \mathopen{} \left ( \theta \right ) \mathclose{} R_{z} \mathopen{} \left ( \psi \right ) \mathclose{} \Rightarrow \psi^{\prime \prime \prime} = - \phi^{\prime} + \alpha \\ R^{\prime \prime} & = R R_{z} \mathopen{} \left ( - \alpha \right ) \mathclose{} R^{\prime} \\ \left( {}_{{s}} f \ast {}_{{s}} \delta^{R^{\prime}} \right) \mathopen{} \left ( R \right ) \mathclose{} & = \frac{1}{2 \pi} \oint {}_{{s}} f \mathopen{} \left ( R R_{z} \mathopen{} \left ( - \alpha \right ) \mathclose{} R^{\prime} \right ) \mathclose{} \mathrm{e}^{\mathrm{i} s \left( \phi^{\prime} - \alpha \right)} \,\mathrm{d} \alpha \\ & = \frac{1}{2 \pi} \oint {}_{{s}} f \mathopen{} \left ( R R^{\prime} \mathopen{} \left ( \phi^{\prime} = \phi^{\prime \prime} \right ) \mathclose{} \right ) \mathclose{} \mathrm{e}^{\mathrm{i} s \phi^{\prime \prime}} \,\mathrm{d} \phi^{\prime \prime} \end{align*}\]
Specializing at \(R = I\),
\[\begin{align*} \left( {}_{{s}} f \ast {}_{{s}} \delta^{R^{\prime}} \right) \mathopen{} \left ( I \right ) \mathclose{} & = \frac{1}{2 \pi} \oint {}_{{s}} f \mathopen{} \left ( R^{\prime} \mathopen{} \left ( \phi^{\prime} = \phi^{\prime \prime} \right ) \mathclose{} \right ) \mathclose{} \mathrm{e}^{\mathrm{i} s \phi^{\prime \prime}} \,\mathrm{d} \phi^{\prime \prime} \\ & = \frac{1}{2 \pi} \oint {}_{{s}} f \mathopen{} \left ( R_{z} \mathopen{} \left ( \phi^{\prime \prime} \right ) \mathclose{} R^{\prime} \mathopen{} \left ( \phi^{\prime} = 0 \right ) \mathclose{} \right ) \mathclose{} \mathrm{e}^{\mathrm{i} s \phi^{\prime \prime}} \,\mathrm{d} \phi^{\prime \prime} \end{align*}\]
The above shows that it is performing a spin-weighted azimuthal average around \(R'\). That is, with assumption (1c, 3b) and eq. (34.1), (34.3), the convolved map is, using the \(\operatorname{SO} \mathopen{} \left ( 3 \right ) \mathclose{}\) notation,
\[\begin{align*} {}_{{s}} m^{\prime} \mathopen{} \left ( R \right ) \mathclose{} & = \left( m \ast {}_{{s}} P \right) \mathopen{} \left ( R \right ) \mathclose{} \text{, where} \\ {}_{{s}} P \mathopen{} \left ( R \right ) \mathclose{} & \equiv \mathcal{X} {}_{{s}} \delta^{I} \mathopen{} \left ( R \right ) \mathclose{} + \sum_{i , j \neq i} \mathcal{X}^{i j} {}_{{s}} \delta^{i \rightarrow j} \mathopen{} \left ( R \right ) \mathclose{} \\ {}_{{s}} \delta^{i \rightarrow j} & \equiv {}_{{s}} \delta^{R^{i \rightarrow j}} \end{align*}\]
Assumption (1c) is implicit when the convolution is used which has an azimuthal averaging effect as explained above, corresponding to a scanning strategy with isotropic focal plane orientation.
Then, the harmonic expansion of this delta function can be derived,
\[\begin{align*} {}_{{s}} \delta^{R^{\prime}}_{\ell m} & = \oint {}_{{s}} \delta^{R^{\prime}} \mathopen{} \left ( R \right ) \mathclose{} {}_{{s}} \bar{\mathcal{Y}}_{\ell m} \mathopen{} \left ( R \right ) \mathclose{} \,\mathrm{d} \Omega \\ & = \oint \delta^{\left( 2 \right)} \mathopen{} \left ( R^{\prime} R = R_{z} \mathopen{} \left ( \alpha \right ) \mathclose{} \right ) \mathclose{} {}_{{s}} \bar{Y}_{\ell m} \mathopen{} \left ( \theta, \phi \right ) \mathclose{} \,\mathrm{d} \Omega \\ & = {}_{{s}} \bar{Y}_{\ell m} \mathopen{} \left ( R^{\prime -1} \hat{z} \right ) \mathclose{} = {}_{{s}} \bar{\mathcal{Y}}_{\ell m} \mathopen{} \left ( R^{\prime -1} \right ) \mathclose{} \vert_{\phi^{\prime} = 0} \\ & = \left( -1 \right)^{- m} \sqrt{\frac{2 \ell + 1}{4 \pi}} d^{\ell}_{s , - m} \mathopen{} \left ( \theta^{\prime} \right ) \mathclose{} \mathrm{e}^{- \mathrm{i} s \psi^{\prime}} \end{align*}\]
And that of \(P\),
\[{}_{{s}} \left( {}_{{s}} f \ast {}_{{s}} P \right)_{\ell m} = {}_{{s}} f_{\ell m} \left( \mathcal{X} + \sum_{i , j \neq i} \mathcal{X}^{i j} d^{\ell}_{s s} \mathopen{} \left ( \theta^{i \rightarrow j} \right ) \mathclose{} \mathrm{e}^{- \mathrm{i} s \alpha^{i \rightarrow j}} \right)\]
where \(d\) is the Wigner \(d\) matrix.
Equipped with this, the effect of crosstalk on the spherical harmonics can be derived,
\[\begin{align*} a^{\prime}_{\ell m} & \equiv {}_{{0}} \left( T \ast {}_{{0}} P \right)_{\ell m} = a_{\ell m} \left( \mathcal{X} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathcal{P}_{\ell} \mathopen{} \left ( \cos \theta^{i \rightarrow j} \right ) \mathclose{} \right) \\ {}_{{2}} a^{\prime}_{\ell m} & \equiv {}_{{2}} \left( \mathcal{Q} \ast {}_{{2}} P \right)_{\ell m} = {}_{{2}} a_{\ell m} \left( \mathcal{X} + \sum_{i , j \neq i} \mathcal{X}^{i j} d^{\ell}_{22} \mathopen{} \left ( \theta^{i \rightarrow j} \right ) \mathclose{} \mathrm{e}^{-2 \mathrm{i} \alpha^{i \rightarrow j}} \right) \\ {}_{{-2}} a^{\prime}_{\ell m} & = {}_{{-2}} a_{\ell m} \left( \mathcal{X} + \sum_{i , j \neq i} \mathcal{X}^{i j} d^{\ell}_{22} \mathopen{} \left ( \theta^{i \rightarrow j} \right ) \mathclose{} \mathrm{e}^{2 \mathrm{i} \alpha^{i \rightarrow j}} \right) \end{align*}\]
Defining \(\mathcal{A} ^\mathsf{T}_{\ell m} = \begin{bmatrix} a^{T}_{\ell m} & a^{E}_{\ell m} & a^{B}_{\ell m} \end{bmatrix}\), the effect of crosstalk on \(\mathcal{A}\) would be,
\[\begin{align*} \mathcal{A}_{\ell m}^{\prime} & = \left[ \mathcal{X} + \sum_{i , j \neq i} \mathcal{X}^{i j} \mathcal{I}^{i j}_{\ell} \right] \mathcal{A}_{\ell m} \\ \mathcal{I}_{\ell} \mathopen{} \left ( \theta, \alpha \right ) \mathclose{} & \equiv \begin{bmatrix} \mathcal{P}_{\ell} \mathopen{} \left ( \cos \theta \right ) \mathclose{} & 0 & 0 \\ 0 & d^{\ell}_{22} \mathopen{} \left ( \theta \right ) \mathclose{} \cos 2 \alpha & d^{\ell}_{22} \mathopen{} \left ( \theta \right ) \mathclose{} \sin 2 \alpha \\ 0 & - d^{\ell}_{22} \mathopen{} \left ( \theta \right ) \mathclose{} \sin 2 \alpha & d^{\ell}_{22} \mathopen{} \left ( \theta \right ) \mathclose{} \cos 2 \alpha \end{bmatrix} \\ \mathcal{I}^{i j}_{\ell} & \equiv \mathcal{I} \mathopen{} \left ( \theta^{i \rightarrow j}, \alpha^{i \rightarrow j} \right ) \mathclose{} \end{align*}\]
Where \(\mathcal{P}\) is the Legendre polynomials17. Defining the 2-point power-spectra as,
\[\begin{align*} \mathbb{C}_{\ell} & \equiv \left\langle \mathcal{A}_{\ell m} \mathcal{A} ^\dagger_{\ell m} \right\rangle_{m} \\ & \equiv \begin{bmatrix} C^{TT}_{\ell} & C^{TE}_{\ell} & C^{TB}_{\ell} \\ C^{ET}_{\ell} & C^{EE}_{\ell} & C^{EB}_{\ell} \\ C^{BT}_{\ell} & C^{BE}_{\ell} & C^{BB}_{\ell} \end{bmatrix} \end{align*}\]
The resultant power-spectra with crosstalk is computed as,
\[\begin{equation} \mathbb{C}^{\prime}_{\ell} = \mathcal{X}^{2} \mathbb{C}_{\ell} + \left[ \mathcal{X} \sum_{i , j \neq i} \mathcal{X}^{i j} \mathcal{I}^{i j}_{\ell} \mathbb{C}_{\ell} + \text{h.c.} \right] + \sum_{i , j \neq i , m , n \neq m} \mathcal{X}^{i j} \mathcal{X}^{m n} \mathcal{I}^{i j}_{\ell} \mathbb{C}_{\ell} \mathcal{I}^{\dagger m n}_{\ell} \label{eq:crosstalk_spectra_curve_sky} \tag{34.10}\end{equation} \]
This can be compared with the derivation in sec. 34.1 and the result in sec. 34.2. Their relationship with each other is more apparent, given \(\mathcal{P}_\ell(\cos θ) = d^\ell_{00}(θ)\) and18
\[\mathcal{P}_{\ell} \mathopen{} \left ( \cos \theta \right ) \mathclose{} = \sqrt{\frac{\theta}{\sin \theta}} J_{0} \mathopen{} \left ( \left( \ell + \frac{1}{2} \right) \theta \right ) \mathclose{} \left( 1 + \operatorname{O} \mathopen{} \left ( \frac{1}{\ell} \right ) \mathclose{} \right) ,\]
where \(θ \ll 1\) from assumption (3b).
Lastly, note that keeping \(\mathbb{C}'_\ell\) up to first order is easier to understand the crosstalk effect on the resultant power-spectra qualitatively, while \(\mathcal{A}'_{\ell m}\) is more useful to evaluate the exact19 crosstalk effect. Notice how the crosstalk effect is independent of the signal and is completely captured by the hardware information only, which includes the crosstalk matrix, detector noise, focal plane configuration, etc. This means that it can be computed very fast without performing computationally expensive TOD simulations. This prediction can be used to design the hardware to minimize crosstalk effect with very fast iterations.
Below we will expand eq. (34.10) in a special case which will be useful later.
2-detectors with equal weights
If we take a 2-detector case with symmetric crosstalk matrix and equal weighting between them,
\[X = \begin{bmatrix} 1 & 2 x \\ 2 x & 1 \end{bmatrix} \Rightarrow \mathcal{X}^{i j} = \begin{bmatrix} \frac{1}{2} & x \\ x & \frac{1}{2} \end{bmatrix} \Rightarrow \mathcal{X} = 1 .\]
Assuming \(C^{TB}_\ell = C^{EB}_\ell = 0\), expanding eq. (34.10),
\[\begin{equation}\begin{aligned} C_{\ell }^{\prime\text{TT}} &= C_{\ell }^{\text{TT}} \left(x P_{\ell }+1\right){}^2 \\ C_{\ell }^{\prime\text{EE}} &= d_{22}^2 x^2 \sin ^2(2 \alpha ) C_{\ell }^{\text{BB}}+C_{\ell }^{\text{EE}} \left(d_{22} x \cos (2 \alpha )+1\right){}^2 \\ C_{\ell }^{\prime\text{BB}} &= C_{\ell }^{\text{BB}} \left(d_{22} x \cos (2 \alpha )+1\right){}^2+d_{22}^2 x^2 \sin ^2(2 \alpha ) C_{\ell }^{\text{EE}} \\ C_{\ell }^{\prime\text{TE}} &= C_{\ell }^{\text{TE}} \left(x P_{\ell }+1\right) \left(d_{22} x \cos (2 \alpha )+1\right) \\ C_{\ell }^{\prime\text{TB}} &= -x d_{22} \sin (2 \alpha ) C_{\ell }^{\text{TE}} \left(x P_{\ell }+1\right) \\ C_{\ell }^{\prime\text{EB}} &= d_{22} x \sin (2 \alpha ) \left(C_{\ell }^{\text{BB}}-C_{\ell }^{\text{EE}}\right) \left(d_{22} x \cos (2 \alpha )+1\right) \label{eq:2-detector-crosstalk-spectra-expanded} \end{aligned}\tag{34.11}\end{equation}\]
with some of the functions suppressed, such as \(P_\ell \equiv P_\ell(\cos θ)\), \(d_{22} \equiv d^\ell_{22}(θ)\), where \(θ\) is the angular distance on sky between the 2 detectors, and \(α\) is the angle between the polarization of the 2 detectors. The expression is more simplified in the case of flat-sky approximation where \(P_\ell(\cos θ) \sim d^\ell_{22}(θ) \sim J_0(\ell θ)\) as mentioned above.
There are a few features worth pointing out here. \(α\) does not affect \(TT\), as expected. For \(EE, BB\), there is a symmetry in the equations in swapping between them. Lastly, the \(E \rightarrow B\) leakage induced by crosstalk systematics is 2nd order in \(x\), which can still be significant as \(C^{EE}_\ell \gg C^{BB}_\ell\).
Footnotes
Matrices are specified with explicit indices throughout unless stated otherwise. Also, upper indices and lower indices are used for detectors and the discrete sampling points respectively.↩︎
That is, (1a) assumes the scan is \(\mathbb{S}^{2}\) invariant, and (1c) in addition to (1a) assumes it is \(\operatorname{SO} \mathopen{} \left ( 3 \right ) \mathclose{}\) invariant.↩︎
\(O[f]\{g\}(x)\) means an operator \(O\) is a higher-order function of \(f\) which transforms a function \(g\) to another function \(O\{g\}\) evaluated at \(x\).↩︎
Lower indices are used for the detector index in \(R\) and its variants for ease of typesetting.↩︎
\(R\) can depend on \(\mathbf{r}\) by assumption (1b).↩︎
This fixes the remaining d.o.f. of \(R\) in terms of \(\Delta \alpha\) .↩︎
Methods 1–8 in Tegmark, “How to Make Maps from Cosmic Microwave Background Data Without Losing Information”, table 1.↩︎
“How to Make Maps from Cosmic Microwave Background Data Without Losing Information”.↩︎
Note that the point spread function convolution here is analogous to beam convolution with \(P \mathopen{} \left ( \mathbf{r} \right ) \mathclose{} = B \mathopen{} \left ( - \mathbf{r} \right ) \mathclose{}\) .↩︎
Physicists’ convention is followed here. The same symbol to represent the function before or after Fourier Transform depends on the argument, and the Fourier Transform convention used is standard in the CMB literature, namely, defined with angular frequency in the non-unitary gauge where only the inverse transform has a factor of \(2\pi\).↩︎
\(\Re , \Im\) below are the real part and imaginary part operators.↩︎
h.c. stands for Hermitian Conjugate.↩︎
Boyle, “How Should Spin-Weighted Spherical Functions Be Defined?”↩︎
\(\operatorname{Spin(3)}\) for half-integer spin.↩︎
Using convention commonly used in quantum mechanics, namely, the z-y-z, right-handed, active transformation.↩︎
See Generalized convolution theorem on \(\operatorname{SO} \mathopen{} \left ( 3 \right ) \mathclose{}\).↩︎
\(\mathcal{P}\) is used instead of the usual \(P\) as it is already used for the PSF above.↩︎
Technology (U.S.), NIST Handbook of Mathematical Functions Hardback and CD-ROM, eq. 14.15.11.↩︎
Exact as long as the assumptions are valid. But it is still an approximation if the assumptions are only approximately true.↩︎