Re: Light Concentrators

From: Hans Jostlein (
Date: Wed Jun 08 2005 - 12:12:33 CDT

Thank you for your detailed and useful summary , Steve.
I am very pleased that someone is running honest models on this.
I am not surprised that some 30% gains are possible (the strawman cone I
showed in my note would do that, most likely),
which is certainly worthwhile if it does not compromise our ability to
locate events.
Since 3/4 of the area would be covered with reflectors, one might actually
realize (theoretically, and in the absence of absorption)
a fourfold gain if we boldly sum the geometric series (1 + 3/4 + (3/4)^2 . .

I have a few concerns which you may already have thought of and have the
answers to:

a. Do you have a sketch of your concentrator design?
I would very much like to see that

b. Do the concentrators stick out to where they deflect light that
otherwise would hit a PMT directly?

c. Braidwood uses scintillation light, not Cerenkov.
     Also, the detector is quite bit smaller than SNO.
In view of those factors, do you expect we can differentiate, by timing,
between first light and reflected light?
Does our electronics design (such as it is) allow us to do that?
If we can't differentiate, I suspect that the ability to locate the event is
compromised by the reflected light, which tends to illuminate PMT's more
uniformly than the direct light.
Can your model predict localization with and without integrating over all
light, or up to some delay time ?

d. It has been pointed out that the PMT's (Hamamatsu R5912's) have a
significant drop off in QE at their curved edges:
Byron Roe wrote:

"It is also true that the efficiency of the PMT's for light coming into
the side is
quite low. Laura Gladstone, my REU student last year is just finishing a
NIM paper and finds for the miniBooNE 8" tubes (including old and new)
relative efficiency = 1 + a_2theta^2 + a_4theta^4 + a_6theta^6,
a_2 = -1.11e-4
a_4 = 4.19e-9
a_6= -5.23e-14
and theta is in degrees wrt normal to tube.

The paper is not yet public, but soon will be and these numbers will
change little
if at all.

It will be interesting to see what this roll-off does to the concentrator



----- Original Message -----
From: "Steve Biller" <>
To: "Braidwood Collaboration" <>
Sent: Wednesday, June 08, 2005 11:02 AM
Subject: Light Concentrators
----- Original Message -----
From: "Steve Biller" <>
To: "Braidwood Collaboration" <>
Sent: Wednesday, June 08, 2005 11:02 AM
Subject: Light Concentrators

> It doesn't seem to us that there should be any problems with
> light concentrators. In fact, preliminary investigations suggest
> that these ought to work extremely well for Braidwood. Here
> are some details:
> The optimal shape of a light collectors for hemispherical
> PMTs is not a flat cone, but can be best described as approximately
> segments of a Winston cone arranged in a ring about the perimeter of
> the PMT. In accordance with Lioville's theorem, the effective gain
> in light collection (or 'concentration factor'), has several terms:
> First, 1/sin^2(theta), where theta is the half-angle of the cone opening.
> Secondly, the transmission function at a given incident angle
> (i.e. cone reflectivity and probablity to release a photoelectron
> at a given incidence angle on the PMT). Thirdly, because the PMT surface
> is curved, there is an extra factor for the ratio of the actual
> photocathode
> area to the projected area at zero incidence angle. For typical, 8-inch,
> hemispherical tubes, this last factor amounts to a gain of about 30%.
> The SNO concentrators were optimised for a cut-off half-angle
> of about 56 degrees, so as to include 1m of light water outside
> of the acrylic vessel. For Braidwood, if we just wanted to set this
> angle at the acrylic vessel (which may be very reasonable),
> the angle is almost identical. Obviously, these concentrators have
> worked extremely well in SNO (we couldn't have done without them),
> so there is a clear proof of principle that this does all work.
> The theoretical maximum concentration factor (taking the transmission
> function as being 1) for a 56 degree opening angle and standard 8-inch,
> hemispherical PMTs is therefore:
> C_max = (1/sin^2(56 deg))*(1.0)*(1.33) = 1.9
> In practise, light cones have something closer to a 90% transmission
> function averaged over angles with an angular response that rolls-off
> at the edges. In SNO, we managed to achieve an average concentration
> of about 1.7. There is every reason to believe that a similar factor
> can be achieved for Braidwood. If we wished to extend the cut-off angle
> to encorporate a view of 20cm into the buffer region, this concentration
> factor would go down to about 1.5.
> Two issues that arise in the use of concentrators are: 1) the fact that
> the angular acceptance of the PMTs is clearly affected and needs to be
> modelled, and 2) some photons will bounce off the reflectors, into the
> detector to produce late light. In the case of the former, PMTs have
> a wavelength-dependent angular response that has to be considered in any
> case, and this is all part of what you see from your calibration data.
> In SNO, while we've gone through the process of modelling and questioning
> and re-modelling etc. the angular response function of cones, the fact is
> that this really hasn't been a problem as the calibrations nail things
> anyway.
> And, in the case of Braidwood, we're talking about a relative measurement
> of identical detectors so all this is second order in any case.
> Regarding the
> late light issue, once again, reflections off the PMT glass etc. are there
> anyway. The additional light, whether late or not, improves the
> measurement of total charge. Also, the increased sampling statistics of
> each PMT+concentrator improves the resolution of the prompt-light peak
> thus, enhances the ability to reconstruct events based on timing.
> The late-timing tail due to reflections will only last tens of nanoseconds
> and, again, while there a some aspects of this tail which are not
> modelled in SNO, this has not really affected us and the concentrators
> definately been a positive (in fact, necessary) gain.
> Preliminary simulations of the Braidwood geometry using the SNO
> Monte Carlo both confirm a concentration factor in the expected
> regime and indicates that the additional late-light tail based on the
> arrival time of the 1st photon at each PMT appears to be minimal.
> However, we are still studying details of the timing and charge
> resolutions and particularly need to look more closely at these
> near the edge of the fiducial volume. We hope to produce a more detailed
> report of all this in the near future.
> A final note: the context in which we discussed the use of light
> concentrators previously was not as a means to necessarily decrease the
> number of PMTs used, but to enhance the light collection for the same
> number of tubes in a cost-efficient manner so as to further improve the
> vertex reconstruction and energy resolution of the detector (after all,
> a gain of 1.7 in light yield stands to improve our energy resolution
> by up to 30%). In addition, they may allow us to potentially pick up the
> weaker Cherenkov signal for use in additional background discrimination.
> - Steve

This archive was generated by hypermail 2.1.6 : Fri Jun 10 2005 - 03:10:15 CDT