1. Measurement of water optical properties in ANTARES
MANTS
Paris, September 25th 2010
Juande Zornoza (IFIC, Valencia)

2. Introduction
The knowledge of the water optical properties is essential for the correct simulation of the performance of the detector (efficiency and angular resolution)
Compared to the ice, the big advantage of deep sea water is the larger scattering length, critical for the angular resolution.
Moreover, the uniformity of the optical parameters makes life easier (no dust layers…)
On the other hand, absorption length (better efficiency) plays in favour of ice detectors (in addition to the lack of K40 and bioluminescence)
After the installation of the first lines, optical beacons (LEDs and laser) can be used for the measurement of optical properties

3. Water model: Partic Scattering phase function
Morel and Loisel approach
Scattering phase function
Average cosine of global distribution
Let me remind you…
Probability of molecular scattering
(Rayleigh)

4. ,P.Amram et al.,  Background light in potential sites for the ANTARES undersea neutrino telescope  Astroparticle Physics 13 (2000)  
R&D phase
Several tests with autonomous lines were done during the R&D phase.

5. LED Optical beacons Their main goal is timing calibration.
LED Beacons: four beacons per line (storeys 2, 9, 15, 21)
36 LEDs: 6 faces x 6 LEDs/face
Light almost uniform 50º <  < 120º.
Their main goal is timing calibration.
For water properties studies, a single upward-pointing LED is used in each run.
1-2 runs for optical water properties studies taken per month.

6. LED characteristics
The default type of LED emits blue light (=470 nm, 15 nJ per pulse)
Other wavelengths:
A few UV (=400 nm) already working
To be deployed in Nov.: = 385, 400, 440, 470, 505, 518 nm
Light output can be regulated
Firing different LEDs
For each LED changing base voltage

7. Information available
To estimate the water optical parameters we have two pieces of information:
Dependence of Nhits on distance
Delay in hit arrival times
Nhits R2 (m2)
distance (m)

8. Montecarlo simulation
25 combinations for each absorption length:
Absorption Length (m) = 55, 60, 63, 65, 70
L scattering (m) = 30, 40, 50, 60, 70
η = 0.17, 0.15, 0.12, 0.10, 0.05
The way we proceed is to scan some reasonable range of the water optical parameters with MC, so that we can compare this with our measurements. For each absorptyion…..
125 sets of optical water parameters simulated

9. Analysis procedure
storey 2
storey 9
storey 15
Time
Line
Line
Emission with ONE top LED of a beacon at the lowest floor
The amount of light is measured at the PMTs above in the same line
Nhits * (distance)2 is plotted versus distance
storey 21
storey 15
storey 9
* This “L” is not the absorption length
storey 2

10. Experimental procedure
The starting point for the fit is limited by the loss of charge due to dead time
The choice of the ending point is driven by the level of statistics
The differences in the OM efficiencies are taken into account

11. Extraction of signal Determine the peak  Gaussian fit
Nnoise
Tmax
Tmin
Nsignal
Determine the peak  Gaussian fit
Choose fixed time window [Tmin,Tmax] and select the hits in this time window.
Tmin = Tpeak – 3s.
Tmax = Tpeak ns.
Calculate their overall Nhit

12. Effect of scattering: R technique
R+ΔR
The effect of scattering is to increase the photon path length in ΔR on average.
Perform the exponential fit using (mean photon path), R + ΔR instead of R.
ΔR can be extracted from time distributions Dt
ΔR is related to scattering.
Exponential fit with R + ΔR will be more insensitive to scattering, making the LDR parameter closer to the real absorption length

13. R+∆R vs R in MC R + DR (m) R (m)
Once we have our simulation, the first thing to do is to parameterize the dependence of R+AR vsR. Here yoou can see one example…
R (m)
MC simulation allows to estimate the mean delay introduced by scattering (for different combinations of parameters).
For each model, a different parameterization of R+DR vs R is found.
The more the scattering, the steeper the slope (a) of this parameterization.

14. Example of calculating L
Input parameters: Labs=65 m, Lscat=50 m, η=0.05
Before DR correction
After DR correction
Nhits R2 (m2)
L = 53.5 m
Nhits R2 (m2)
L∆R = 60.8 m
L = 57.2 m
L = 57.3 m
Once we have parameterize this dependence we can plot Nhitstimes Rsquare vs r plus Ar. You can see that now the new value of L is closer to input value of the absorption length but still not exactly the same.
Of course this new L is just alpha times L for a linear parametrization.
R (m)
R + DR (m)

15. Labs=55 m Labs=60 m Labs=65 m Labs=70 m
Difference between Labs and LDR
Labs=55 m
Labs=60 m
Mean = 4.2 ± 1.1 m
Mean = 5.1 ± 1.5 m
Labs=65 m
Labs=70 m
Mean = 5.3 ± 1.6 m
Mean = 6.1 ± 1.7 m

16. Comparison LDR vs Labs Difference = 5.2 ± 1.7 m
Labs-LDR (m)
In the range studied for the model parameters, the right value of Labs is recovered by adding a shift of about 5 meters to LDR.

17. Slope of the fit R+∆R vs R with L∆R
Scattering
This method will be based on plotting the slope of the fit of R+AR vs R and the corrected L.
This variable here is related with the scattering while this other variable is related with the absorption. So our measurement of these variables will allows us constrain the space parameter.
L_∆R (m)
L_∆R (m)
Absorption

18. Data Now we look at the data:
R+∆R (m)
Now we look at the data and again the first thing we do is to study the dependence of R+AR vs R. Here we have merge the data for ten runs whith the optical beacons. But we plan to include more soon. So you can seee that we have the measured value of the plot I¡ve just shown and an estimation of its error.
R (m)
Fit between 135 and 240 meters. Slope (a) = ± 0.014

19. Calculation of L and LDR in data
For one run:
Before DR
With DR
L = 53.5 m
L = 57.2 m
An now we can calculate the other variable which is the corrected L. This is for one of the runs,
R+DR (m)

20. L∆R for 10 runs L∆R = 58.0 ± 1.3 m (L = 54.3 ± 1.3 m) LDR (m)
And if we do this for the ten runs we get this distribution. And again this also give us an estimation of the error. Just from this plot we have a first estimation of the absorption length which would be about 63 m
LDR (m)
L∆R = 58.0 ± 1.3 m
(L = 54.3 ± 1.3 m)

21. Stability of L (and LDR)
Color code: OB used
(forget about that)
~2 years
Blue UV
Green

22. Comparing data with MC (I)
Slope of the fit R+∆R vs R
But as we said our method is tto use this plot which shows us which are the preferred combination of parameters. In this ellipse the vertical error comes from the error of the fit of R+ARvsR and the x error comes from the spread of the corrected L for ten runs.
Anyway a more careful threathment of this errors is needed in order to take in to account correlations and things like that. In Anycase we can see that our first estimation of landa absorption equal to 63 seems to be quite accurate.
L_∆R (m)
Labs~63 m

23. Comparing data with MC (II)
Leff scatt
Slope of fit R+DR vs R
LDR (m)

24. Comparing data with MC (III)
We can also estimate L_eff_scatt from a, assuming Labs=63 m:
Slope of fit R+DR vs R
data
Leffscat ~ m

25. Measurement of speed of light
Small chromatic dispersion because of small wavelength range (~10nm)
Time distribution is fitted to the convolution of gaussian and exponential

26. Comparison of parametric formula and measurement

27. Summary
The medium is part of the detector, so it is necessary to study its properties for a correct simulation and estimation of the systematic effects.
It is a convoluted problem: models include several parameters and it is complex to disentangle scattering and absorption.
The system of Optical Beacons is providing results to constrain the parameter space (Labs~63 m, Leffscat~300 m)
Further steps:
Checks on errors (correlations, data/MC agreement…)
More blue LED runs will be included
Other wavelengths
Refraction index measured