1.  Delensing :

1.1  Variance spectrum

muKarcmin	0.1	1.0	5.0
Bicep
LiB

1.2  Window matrices

Delens	0.0	0.1	0.2	0.3	0.4	0.5
Bicep, 0.1 muK
LiB, 0.1 muK

2.  Modes mixing :

2.1  Bicep, Window matrices ns 64 Vs ns 128 :

Pixelization effect due to lmax :

0.1	1.0	5.0	10.0	20.0	50.0

• bservations :
  • The last bins suffer from the same E/B leakage than for ns128. We conclude this is due to the pixelization effect.

2.2  E to B leakage

Modes

Modes contributions to the BB-spectrum reconstruction, computed from the renormalized modes-mixing matrix {$\bar F_{\ell\ell'} \equiv F_{\ell\ell'} / \sum_{\ell''} F_{\ell\ell''}$}. All {$\ell'$}-modes contributions are shown for to three bins {$\ell$}'s respectively selected at lower (blue), intermediate (red), and higher (green) angular scales. Each upper panel quantifies the E contributions (E-to-B leakage) computed from the $EB$ block of {$\bar F_{\ell\ell'}$}. Lower panels quantify the BB contributions. Tegmark explains that {$\bar F_{\ell\ell'} $} quantifies the E-B leakage, and is equals to a kronecked delta function on full sky.

	0.1	1.0	5.0	10.0	20.0	50.0
Bicep
LiB

EB diagonals mixing

	0.1	1.0	5.0	10.0	20.0	50.0
Bicep
LiB

• observation :
  • The E-to-B leakage is mostly dominant at large angular scales for the reionization survey, then decreases with {$\ell$}.
  • For the recombination survey, we observe the same beahaviour at high noise level, and the opposite behaviour for low noise levels (the leakage increases with $\ell$).
  • The E-B leakage increases with noise.

2.3  B to E leakage

Same as for E to E

	0.1	1.0	5.0	10.0	20.0	50.0
Bicep
LiB

• observation :
  • The E-to-B leakage is mostly dominant at large angular scales for the reionization survey, then decreases with {$\ell$}.
  • For the recombination survey, we observe the same beahaviour at high noise level, and the opposite behaviour for low noise levels (the leakage increases with $\ell$).
  • The E-B leakage increases with noise.

2.4  E to B variance leakage

We basically do the same as above, but with the correlation matrix computed from the spectra of covariance matrix {$V_{\ell\ell'}$} (which is basically very close to {$F^{-1}_{\ell\ell'}$} )

modes :

	0.1	1.0	5.0	10.0	20.0	50.0
Bicep
LiB

EB diagonals (variance leakage)

taking the absolute value of correlation for log scale plotting (their are actually negatives)

	0.1	1.0	5.0	10.0	20.0	50.0
Bicep
LiB

Correlation matrix :

	0.1	1.0	5.0	10.0	20.0	50.0
Bicep
LiB

• observation :

3.  Variance leakage

3.1  Surveys results :

• area is the difference between the {$ C_\ell{EE} \neq 0 $} and {$ C_\ell{EE} =0 $} (leakage VS no-leakage)
• the absolute leakge is compute as {$ (\Delta \hat C^{\text{leak}}_\ell - \Delta \hat C^{\text{no-leak}}_\ell)/\Delta \hat C^{\text{no-leak}}_\ell $}

spectrum + absolute leakage