Issue 
A&A
Volume 519, September 2010



Article Number  A104  
Number of page(s)  7  
Section  Astronomical instrumentation  
DOI  https://doi.org/10.1051/00046361/201014739  
Published online  20 September 2010 
E/B decomposition of CMB polarization pattern of incomplete sky: a pixel space approach
J. Kim  P. Naselsky
Niels Bohr Institute & Discovery Center, Blegdamsvej 17, 2100 Copenhagen, Denmark
Received 6 April 2010 / Accepted 26 June 2010
Abstract
CMB polarization pattern may be decomposed into gradientlike (E)
and curllike (B) mode, which provide invaluable information
respectively.
We have investigated E/B decomposition in pixel space. We find E/B
mixing due to incomplete sky is localized in pixelspace, and
negligible in the regions far away from the masked area. By estimating
the expected local leakage power, we have diagnosed ambiguous pixels.
Our criteria for ambiguous pixels (i.e. )
is associated with the tensortoscalar ratio of B mode power spectrum, which the leakage power is comparable to.
By setting
to a lower value, we may reduce leakage level, but reduce sky fraction at the same time.
Therefore, we have solved
,
and obtained the optimal ,
which minimizes the estimation uncertainty, given a foreground mask and
noise level. We have applied our method to a simulated map blocked by a
foreground (diffuse + point source) mask. Our simulation shows leakage
power is smaller than primordial (i.e. unlensed) B mode power spectrum
of tensortoscalar ratio
at wide range of multipoles (
), while allowing us to retain sky fraction 0.48.
Key words: methods: data analysis  cosmic background radiation
1 Introduction
Over the past years, CMB polarization has been measured by several experiments and is being measured by the Planck surveyor (Kovac et al. 2002; Leitch et al. 2002; Pryke et al. 2002a,b Halverson et al. 2002; Brown & et al. 2009; Hinderks et al. 2009; Leitch et al. 2005; Pryke & et al. 2009; The Planck Collaboration 2006; Ade & et al. 2008). CMB polarization pattern may be considered as the sum of gradientlike E mode and curllike B mode (Zaldarriaga & Seljak 1997; Kamionkowski et al. 1997). In the standard model, B mode polarization is not produced by scalar perturbation, but solely by tensor perturbation. Therefore, measurement of B mode polarization makes it possible to probe the universe on the energy scale at inflationary period (Dodelson 2003; Zaldarriaga & Seljak 1997; Mukhanov 2005; Kamionkowski et al. 1997; Liddle & Lyth 2000). In most inflationary models, tensortoscalar ratio r is much smaller than one, and the WMAP 7 year data imposes an upper bound on r<0.36 at confidence level (Komatsu et al. 2010; Larson et al. 2010).Besides instrument noise, there are complications, which limits detectability of tensor perturbation. Imperfection in removing foreground and gravitational lensing imposes observational limit on tensorscalarratio: and respectively (Seljak & Hirata 2004; Tucci et al. 2005). Due to the nature of the observation or heavy foreground contamination, reliable of estimation on CMB polarization signal is not available over a whole sky. Incomplete sky coverage leads to E/B mixing, and very significantly limit our capacity to measure tensor perturbation as well (Bunn et al. 2003). Therefore, there have been various efforts to understand and reduce E/B mixing (Bunn et al. 2003; Lewis 2003; Kim 2007b; Smith 2006; Lewis et al. 2002; Kim 2007a).
It is best to implement E/B decomposition in map space, since diffuse foregrounds and point sources are welllocalized in map space, and their spatial information are known relatively better than other properties. In this paper, we investigate E/B decomposition in pixel space. Our investigation shows that E/B mixing is highly localized in pixel space. Therefore, we may reduce E/B mixing effectively by excluding the ambiguous pixels. We have applied our method to simulated maps partially blocked by a foreground (diffuse + point source) mask. After excluding ambiguous pixels, we find that leakage power in retained pixels (sky fraction 0.48) is smaller than primordial (i.e. unlensed) B mode power spectrum of tensortoscalar ratio at wide range of multipoles ( ).
The outline of this paper is as follows. In Sect. 2, we discuss allsky analysis of CMB polarization. In Sect. 3, we derive E/B decomposition in pixel space. In Sect. 4, we discuss the application to cut sky, and the method to diagnose ambiguous pixels. In Sects. 5 and 6, we present our simulation result. In Sect. 7, we summarize our investigation. In Appendix A, we discuss error analysis of pseudo C_{l} estimation, and show interpixel noise correlation may be neglected.
2 Stokes parameters
The state of polarization is described by Stokes parameter (Rohlfs & Wilson 2003; Kraus 1986).
Since Thompson scattering does not generate circular polarization,
Stokes parameter Q and U are sufficient to describe CMB polarization (Dodelson 2003).
Stokes parameter Q and U transform under rotation of an angle
on the plane perpendicular to direction
(Zaldarriaga & Seljak 1997; Zaldarriaga 1998):
Therefore, allsky Stokes parameters may be decomposed into spin spherical harmonics (Zaldarriaga & Seljak 1997) as follows:
where the decomposition coefficients are obtained by:
Though the quantity shown in Eq. (2) has direct association with physical observables (i.e. Stokes parameters), rotational variance leads to computational complication. Therefore, two real scalar quantities, termed ``E'' and ``B'' mode, are often built out of (Zaldarriaga & Seljak 1997; Kamionkowski et al. 1997):
where and refer to lowering and raising operator respectively (Zaldarriaga & Seljak 1997). The explicit expression of and are given as follows (Zaldarriaga & Seljak 1997):
where is an arbitrary spin s function. The decomposition coefficients of E and B mode are related to (Zaldarriaga & Seljak 1997) as follows:
For a Gaussian seed fluctuation model, decomposition coefficients of E and B mode satisfy the following statistical properties:
(8)  
(9) 
where denotes an ensemble average. In Fig. 1, we show unlensed and of the WMAP concordance CDM model for various tensortoscalar ratio r.
Figure 1: The power spectrum of E ( top) and B ( bottom): no lensing, B mode power spectrum is plotted for various tensortoscalar ratio r. 

Open with DEXTER 
3 E/B decomposition in pixel space
In this section, we are going to derive a pixelspace analogue of E/B decomposition.
Using Eqs. (3), (6) and (7), we may easily show Eqs. (4) and (5) are equivalently given by:
where
Therefore, we may identify as pixelspace filters for E/B decomposition. Using the property of spins spherical harmonics , we may show the pair of the filter functions have complex conjugate relation:
Figure 2: Filter function: ( left), ( right) for a fixed , and spanning area, . 

Open with DEXTER 
Figure 3: Filter function: modulus for , highly peaked at . 

Open with DEXTER 
4 Incomplete sky coverage
Due to heavy foreground contamination, CMB polarization signal is not
estimated reliably over a whole sky.
For instance, the WMAP team have subtracted diffuse foregrounds by
templatefitting, and masked the regions that cannot be cleaned
reliably.
In Fig. 7, we show a
foreground mask, which combines the WMAP team's polarization mask with
the point source mask and shall be used for our simulation.
The E/B decomposition coefficients from a masked sky are given by:
where
(16) 
and is a foreground mask. Therefore, E/B maps reconstructed from incomplete sky are given by:
Since filter functions are sharply peaked around , certain pixels far away from masked regions may contain negligible E/B mixing, and vice versa. Equivalently, E/B mixing is localized in pixels close to the masked regions. For higher pixel resolution, Eq. (12) contains summation up to higher l, which makes the peak of the filter function sharper. Therefore, E/B mixing decreases with increase in pixel resolution. The spherical harmonic method (the first line of Eqs. (17) and (18)) are much faster than the pixelspace method (the second line), while they are mathematically equivalent. Therefore, we are going to rely on spherical harmonic transformation method for our simulation in the next section. However, it should be kept in mind that the pixelspace approach have provided useful insights on E/B decomposition of incomplete sky.
Using Eq. (18), we may show the expected power of
is given by:
where denotes an ensemble average, and
and is the separation angle between and , and are the rotation angles respectively, which align at and with the great circle passing through and (refer to Fig. 1 of Ng & Liu 1999, for a geometrical diagram). Taking into account Eqs. (19)(21), we may easily show that the expected leakage power at is given by
where
Therefore, we may diagnose ambiguous pixels (i.e. heavy E/B mixing) by comparing with , where is the local power contributed by B mode and given by replacing ``E'' with ``B'' in Eqs. (22)(24). However, estimating Eqs. (22)(24) is prohibitively complicated. Therefore, we are going to resort to MonteCarlo simulations in order to estimate . Depending , we may classify the pixel at as ``pure'' and ``ambiguous''. To be specific, we may retain pixels satisfying:
(25) 
where r is the assumed tensortoscalar ratio of MonteCarlo simulation, from which is estimated. Therefore, the level of leakage in retained pixels is comparable to the primordial B mode power spectrum of tensortoscalar ratio . In Fig. 4, we show the sky fraction for various , given a foreground mask shown in Fig. 7. Since sky fraction decreases with lower , we may not simply set to a lowest value.
Figure 4: Effective sky fraction for various cut level . 

Open with DEXTER 
=  
(26) 
where N_{l} is noise power spectrum. Note that the leakage does not bias the B mode power spectrum estimation, but increases the variance, when the power spectrum estimation is made by a pseudoC_{l} method and leakage is taken care of (Grain et al. 2009; Hivon et al. 2002a). By requiring , we get
In Fig. 5, we plot the left and right hand side of Eq. (27) for the noise level of Planck HFI instrument, and the multipole l=86, which is the peak multipole of primordial B mode power spectrum. From Fig. 5, we find curves intersect at with weak dependence on r. It should be noted that the weak dependence is due to the low signaltonoise ratio of the considered experiment (i.e. ), and the dependence on r is not weak in general. We are going to use for the simulation in the next section.
Figure 5: Numerical solution of Eq. (27) for various r and the noise level of Planck HFI instrument: two plots represent the left hand side (LHS) and right hand side (RHS) of Eq. (27). 

Open with DEXTER 
5 Application to simulated data
Using the WMAP concordance CDM model, we have simulated Stokes parameter Q and U over a wholesky with a HEALPix pixel resolution ( ) and 10' FWHM beam. We have made the inputmap to contain no B mode polarization. Therefore, any nonzero values in output B map are attributed to leakage. We show our simulated polarization map in Fig. 6, where the orientation and length of headless arrows indicates polarization angle and amplitude respectively. Note that the polarization map shows only gradientlike patterns, because they contain only E mode polarization. It is wellknown that E/B mixing increases with the length of cut sky boundary (Bunn et al. 2003). We have combined the WMAP team's polarization mask with the point source mask, and prograded it to . In order to reduce sharp boundaries, we have smoothed the mask with FWHM Gaussian kernel. We have referred to the WMAP team's boundary smoothing process of Internal Linear Combination map (Hinshaw et al. 2007). Nevertheless, it should be noted that smoothed boundary is not essential to our method, and further improvement may be possible by using more sophisticated smoothing kernel (Das et al. 2009). In Fig. 7, we show our smoothed mask, whose sky fraction amounts to 0.71.
Figure 6: Input polarization map: E mode polarization only. 

Open with DEXTER 
Figure 7: Foreground mask: smoothed with FWHM Gaussian kernel. 

Open with DEXTER 
Figure 8: Output map from incomplete sky. 

Open with DEXTER 
Figure 9: Filerted map: masked by a foreground mask, and pixels of are set to zero. 

Open with DEXTER 
Figure 10: Leakage power spectrum and primordial B mode power of various tensortoscalar ratio r: a blue curve correspond to the leakage power estimated without ambiguous pixel filtering (Fig. 8), a green curve to the leakage power estimated with ambiguous pixel filtering (Fig. 9). 

Open with DEXTER 
6 Scaledependence of leakage
In order to reduce leakage, we have removed ambiguous pixels in a scaleindependent way. On the other hand, it is known that leakage has some dependence on scales as well as realspace. Specifically, leakage of low l extends over large area, while leakage of high l is often confined to the small area nearby the boundary. Therefore, one may argue that our method does not reduce the leakage at low l as effectively as that of high l or leads to unnecessary loss of information. However, as shown in Fig. 8, leakage is relatively localized in pixel space, and we were able to reduced leakage very significantly at wide range of multipoles, while retaining sky fraction 0.48. Besides that, our simulation shows the leakage at lowest l is reduced significantly as well. It is also possible to implement a further leakage reduction in a scaledependent way, after ambiguous pixels are removed. Nevertheless, a hybrid method, which exploits scale and position dependence simultaneously, may be most optimal. A wavelet approach may be promising for such implementation, since wavelet functions are, in general, welllocalized in harmonic space as well as pixel space. We defer a rigorous investigation to a separate publication.7 Discussion
We have investigated E/B decomposition in pixel space, and shown that we may produce E/B decomposed maps by convolving polarization maps with certain filter functions of a sharp peak. We find that E/B mixing due to incomplete sky is localized in pixelspace, and negligible in the regions far away from masked area. By estimating the expected local leakage power and comparing it with the expected pure mode power, we have diagnosed ambiguous pixels and excluded them. Our criteria for ambiguous pixels (i.e. ) is associated with the tensortoscalar ratio of B mode power spectrum, which the leakage power is comparable to. The estimation error may increases with lower , because sky fraction decreases. Therefore, we have solved and obtained the optimal , which minimizes the estimation error, given a foreground mask and noise level. We have applied our method to simulated maps blocked by a foreground mask. Simulation shows that leakage power is subdominant in comparison with unlensed B mode power spectrum of at wide range of multipoles ( ), while pixels of sky fraction 0.48 are retained. We may apply our method equally to small sky patch observation, by treating unobserved sky as masked region. From simulation with sky patch of simple symmetric shape, we have confirmed our method reduce E/B mixing very effectively. A rigorous investigation is deferred to a separate publication.
Noise is slightly correlated from pixel to pixel in E and B maps, even when interpixel correlation is absent in Q and U maps. However, this interpixel noise correlation induced by E/B decomposition is not confined to our method, but E/B decomposition in general. Besides that, we find interpixel noise correlation may be neglected without sacrificing the accuracy of error analysis (refer to Appendix A for details). Therefore, it does not limit the applicability of our method.
Current observations such as WMAP were unable to detect B mode polarization. Therefore, we did not attempt to apply our method to observation data. When Planck polarization data of high SignaltoNoiseRatio (SNR) are available in near future, we may apply our method to the data, and be able to detect B mode polarization.
AcknowledgementsWe are grateful to the anonymous referee for thorough reading and helpful comments, which leads to significant improvement of this work. We acknowledge the use of the Legacy Archive for Microwave Background Data Analysis (LAMBDA) and the HEALPix package (Gorski et al. 1999,2005). This work is supported by FNU grant 272060417, 272070528 and 21040355. This work is supported in part by Danmarks Grundforskningsfond, which allowed the establishment of the Danish Discovery Center.
Appendix A: Error analysis
Power spectrum is usually estimated by pseudoC_{l} method at high multipoles (Hivon et al. 2002b; Efstathiou 2006; Nolta et al. 2009; Larson et al. 2010; Wandelt et al. 2001).
According to pseudoC_{l} method, we may estimate power spectrum as follows (Hivon et al. 2002b; Wandelt et al. 2001):
(1) 
where
(2) 
and
(3) 
with being a foreground mask function. The pseudoquantity and are given by:
where j and k refers to a DAyear combination and refers to data (T, E or B). We may split into signal and noise:
The noise part is given by:
where a pixel index i runs over pixels outside a foreground mask, and N_{i} refers to noise at ith pixel. By central limit theorem (Riley et al. 2006; Arfken & Weber 2000), follows a Gaussian distribution, and is uncorrelated among distinct jth DAyear data (i.e. ).
If cross power spectra are used (i.e. ), noise does not bias estimation, and its statistical properties need to be known only for errors analysis (Hinshaw et al. 2007).
In order to understand the effect of noise on error analysis, let us consider covariance of :
where C_{l} is a theoretical power spectrum and
Using Eqs. (A.4) and (A.6), we find
where A denotes terms irrelevant to noise. Therefore, we need to estimate noise covariance in order to estimate Eq. (A.10). First, let us consider diagonal elements of noise covariance:
The main contribution of Eq. (A.11) comes from the first term, because of cancellation through summation in the second term. Therefore, we find with good approximation:
Offdiagonal elements of noise covariance are given by:
Comparing (A.12) with Eq. (A.11), we may see the magnitude of offdiagonal elements is much smaller than that of diagonal elements, because cancellation through summation arise both in the first and the second term of Eq. (A.12). Therefore, we find noise covariance as follow:
Using Eqs. (A.8)(A.10) and (A.13), we find covariance of :
where C denotes terms irrelevant to noise. As shown in Eq. (A.14), we may neglect interpixel noise correlation in computing covariance of .
References
 Ade, P., Bock, J., Bowden, M., et al. 2008, ApJ, 674, 22 [NASA ADS] [CrossRef] [Google Scholar]
 Arfken, G. B., & Weber, H. J. 2000, Mathematical Methods for Physicists, 5th Ed. (San Diego, CA USA: Academic Press) [Google Scholar]
 Bond, J. R., Jaffe, A. H., & Knox, L. 1998, Phys. Rev. D, 57, 2117 [NASA ADS] [CrossRef] [Google Scholar]
 Brown, M. L., Ade, P., Bock, J., et al. 2009, ApJ, 705, 978 [NASA ADS] [CrossRef] [Google Scholar]
 Bunn, E. F., Zaldarriaga, M., Tegmark, M., & de OliveiraCosta, A. 2003, 67, 023501 [Google Scholar]
 Das, S., Hajian, A., & Spergel, D. N. 2009, Phys. Rev. D, 79, 083008 [NASA ADS] [CrossRef] [Google Scholar]
 Dodelson, S. 2003, Modern Cosmology, 2nd Ed. (Academic Press) [Google Scholar]
 Efstathiou, G. 2006, MNRAS, 370, 343 [NASA ADS] [CrossRef] [Google Scholar]
 Eriksen, H. K., O'Dwyer, I. J., Jewell, J. B., et al. 2004, ApJ, 155, 227 [NASA ADS] [Google Scholar]
 Gorski, K. M., Hivon, E., Banday, A. J., et al. 2005, ApJ, 622, 759 [NASA ADS] [CrossRef] [Google Scholar]
 Gorski, K. M., Wandelt, B. D., Hansen, F. K., Hivon, E., & Banday, A. J. 1999 [arXiv:astroph/9905275] [Google Scholar]
 Grain, J., Tristram, M., & Stompor, R. 2009, Phys. Rev. D, 79, 123515 [Google Scholar]
 Halverson, N. W., Leitch, E. M., Pryke, C., et al. 2002, ApJ, 568, 38 [NASA ADS] [CrossRef] [MathSciNet] [Google Scholar]
 Hinderks, J. R., Ade, P., Bock, J., et al. 2009, ApJ, 692, 1221 [NASA ADS] [CrossRef] [Google Scholar]
 Hinshaw, G., Nolta, M. R., Bennett, C. L., et al. 2007, ApJ, 170, 288 [Google Scholar]
 Hivon, E., Górski, K. M., Netterfield, C. B., et al. 2002a, ApJ, 567, 2 [NASA ADS] [CrossRef] [Google Scholar]
 Hivon, E., Górski, K. M., Netterfield, C. B., et al. 2002b, ApJ, 567, 2 [NASA ADS] [CrossRef] [Google Scholar]
 Kamionkowski, M., Kosowsky, A., & Stebbins, A. 1997, Phys. Rev. D, 55, 7368 [NASA ADS] [CrossRef] [Google Scholar]
 Kim, J. 2007a, MNRAS, 375, 625 [NASA ADS] [CrossRef] [Google Scholar]
 Kim, J. 2007b, MNRAS, 375, 615 [NASA ADS] [CrossRef] [Google Scholar]
 Komatsu, E., Smith, K. M., Dunkley, J., et al. 2010, [arXiv:1001.4538] [Google Scholar]
 Kovac, J., Leitch, E. M., Pryke, C., et al. 2002, Nature, 420, 772 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
 Kraus, J. 1986, Radio Astronomy, 2nd Ed. (Powell, Ohio USA: CygnusQuasar Books) [Google Scholar]
 Larson, D., Dunkley, J., Hinshaw, G., et al. 2010, ApJS, accepted[arXiv:1001.4635] [Google Scholar]
 Leitch, E. M., Kovac, J. M., Pryke, C., et al. 2002, Nature, 420, 763 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
 Leitch, E. M., Kovac, J. M., Halverson, N. W., et al. 2005, ApJ, 624, 10 [NASA ADS] [CrossRef] [Google Scholar]
 Lewis, A. 2003, Phys. Rev. D, 68, 083509 [NASA ADS] [CrossRef] [Google Scholar]
 Lewis, A., Challinor, A., & Turok, N. 2002, Phys. Rev. D, 65, 023505 [NASA ADS] [CrossRef] [Google Scholar]
 Liddle, A. R., & Lyth, D. H. 2000, Cosmological Inflation and LargeScale Structure, 1st Ed. (Cambridge University Press) [Google Scholar]
 Mukhanov, V. 2005, Physical Foundations of Cosmology, 1st Ed. (Cambridge University Press) [Google Scholar]
 Ng, K., & Liu, G. 1999, Int. J. Mod. Phys. D, 8, 61 [NASA ADS] [CrossRef] [Google Scholar]
 Nolta, M. R., Dunkley, J., Hill, R. S., et al. 2009, ApJS, 180, 296 [NASA ADS] [CrossRef] [Google Scholar]
 Pryke, C., Halverson, N. W., Kovac, J. M., et al. 2002a, ApJ, 568, 28 [Google Scholar]
 Pryke, C., Halverson, N. W., Leitch, E. M., et al. 2002b, ApJ, 568, 46 [Google Scholar]
 Pryke, C., Ade, P., Bock, J., et al. 2009, ApJ, 692, 1247 [NASA ADS] [CrossRef] [Google Scholar]
 Riley, K. F., Hobson, M. P., & Bence, S. J. 2006, Mathematical Methods for Physics and Engineering: A Comprehensive Guide, 3rd Ed. (Cambridge University Press) [Google Scholar]
 Rohlfs, K., & Wilson, T. L. 2003, Tools of Radio Astronomy, 4th Ed. (New York, NY USA: SpringerVerlag) [Google Scholar]
 Seljak, U., & Hirata, C. M. 2004, Phys. Rev. D, 69, 043005 [NASA ADS] [CrossRef] [Google Scholar]
 Smith, K. M. 2006, Phys. Rev. D, 74, 083002 [NASA ADS] [CrossRef] [Google Scholar]
 The Planck Collaboration 2006, [arXiv:astroph/0604069] [Google Scholar]
 Tucci, M., MartínezGonzález, E., Vielva, P., & Delabrouille, J. 2005, MNRAS, 360, 935 [NASA ADS] [CrossRef] [Google Scholar]
 Wandelt, B. D., Hivon, E., & Górski, K. M. 2001, Phys. Rev. D, 64, 083003 [NASA ADS] [CrossRef] [Google Scholar]
 Zaldarriaga, M. 1998, ApJ, 503, 1 [NASA ADS] [CrossRef] [Google Scholar]
 Zaldarriaga, M., & Seljak, U. 1997, Phys. Rev. D, 55, 1830 [Google Scholar]
All Figures
Figure 1: The power spectrum of E ( top) and B ( bottom): no lensing, B mode power spectrum is plotted for various tensortoscalar ratio r. 

Open with DEXTER  
In the text 
Figure 2: Filter function: ( left), ( right) for a fixed , and spanning area, . 

Open with DEXTER  
In the text 
Figure 3: Filter function: modulus for , highly peaked at . 

Open with DEXTER  
In the text 
Figure 4: Effective sky fraction for various cut level . 

Open with DEXTER  
In the text 
Figure 5: Numerical solution of Eq. (27) for various r and the noise level of Planck HFI instrument: two plots represent the left hand side (LHS) and right hand side (RHS) of Eq. (27). 

Open with DEXTER  
In the text 
Figure 6: Input polarization map: E mode polarization only. 

Open with DEXTER  
In the text 
Figure 7: Foreground mask: smoothed with FWHM Gaussian kernel. 

Open with DEXTER  
In the text 
Figure 8: Output map from incomplete sky. 

Open with DEXTER  
In the text 
Figure 9: Filerted map: masked by a foreground mask, and pixels of are set to zero. 

Open with DEXTER  
In the text 
Figure 10: Leakage power spectrum and primordial B mode power of various tensortoscalar ratio r: a blue curve correspond to the leakage power estimated without ambiguous pixel filtering (Fig. 8), a green curve to the leakage power estimated with ambiguous pixel filtering (Fig. 9). 

Open with DEXTER  
In the text 
Copyright ESO 2010
Current usage metrics show cumulative count of Article Views (fulltext article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 4896 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.