LIGO II Configuration options K.A. Strain for…

                       LIGO II Configuration options
 K.A. Strain for the LSC Advanced Interferometer Configurations Working Group

                                        April 5, 2000


1       Introduction
An outline/baseline design for the core configuration for LIGO II exists (see the LSC Technical
White Paper and LIGO II project book). There are, however, several areas requiring urgent
attention. This document lists and considers options that are likely to have significant impact
on the cost of the project.

1.1     The configuration
The baseline configuration of the 2, 3 or 4 LIGO II interferometers is a power recycled Fabry-
Perot Michelson system, with signal recycling. The transmittance of ITM and SR mirrors will
be chosen to provide a number of operating modes. Each interferometer will then be tunable
over a pre-set range. It is possible to incorporate a method for varying the transmittance of the
signal recycling mirror. The sensing system will then be designed to achieve low noise compared
to the minimum sensitivity envelope and to support the required operating modes.


2       Top-level options
The project level options must be considered as soon as possible.

2.1     Number of interferometers
How many inferferometers are needed? Possibilities are 2, 3 or 4 with increasing cost and effort.

2.2     Interferometer operating modes
It is necessary to make a tactical decision to maximise the science output. Some example options
are presented.

2.2.1    Conservative option

To make 2 broadband interferometers, more-or-less optimised for NS:NS and other compact
object inspirals. The function of a third interferometer is ambiguous in this context (perhaps
something will be learned by looking at co-incidence data in LIGO, but that is going to be late
information).

2.2.2    Intermediate option

To make 2 broadband interferometers as above, but to add a third system which is optimised
for narrowband performance at mid range frequencies. This provides the chance of carrying
out several searches simultaneously. The tunable system can be set either to look for LMXBs
or to catch anything from the coalescence at the end of an inspiral (or a pulsar survey etc.).

                                               1
2.3    Bold option
To make all interferometers both tunable and of variable bandwidth. This will need variable
signal recycling mirrors on all systems. It will also require a committee to decide what modes
to use.

2.4    Noise requirements
One possibility is to follow the LIGO I model for setting the noise budget, but to omit the
shot noise from the model, and add it in again last. This copes with the differences between
shot noise curves. It might not be appropriate if we make 2 broadband interferometers and 1
narrowband interferometer. It leads to frightening requirements at high frequency, as illustrated
in figure 1. Note that this is not an attempt to actually set the requirements, only to illustrate
how they might be set and what they might look like.

                        -21
                       10




                        -22
                       10
          1/2
           h(f) / Hz




                        -23
                       10




                        -24
                       10




                        -25
                       10
                               1                        2                     3
                              10                      10                     10
                                                    f / Hz




Figure 1: The non-optical noise curve for LIGO II which could form the basis of a set of sensing
requirements (solid) is compared to the total noise curve for LIGO I (upper, dashed). To avoid
unrealistic requirements at high frequencies an estimate of the locus of narrowband response
minima has been used to sketch out the prototype curve (lower, dashed) for the upper limit
of the sum of all noise contributuions not explicity included in the white paper figures. Note
that, thus far, a very rough figure has been included for the mirror thermal noise pending final
beamsize selection.



3     System-level options
These are across sub-system (and working group) boundaries. The test mass material option
is not discussed here as it is properly handled in the lasers and optics working group.




                                                2            5-4-2000   14:14DRAFT - for comment
3.1     ITM choice
The decision to baseline 3% ITMs in the white paper and project book was taken mainly to
allow a fall back non-RSE option. This choice should be revisited (promptly) to place any
decision on a more solid footing ­ and have it properly documented.

3.2     Fundamental sensing options
As well as the need to include sensing and control of the SR mirror there are a few options for
the ISC system. Some of these are considered here.

3.2.1   Main L- sensing

In a detuned optical system the transmission of a single rf sideband will be optimised (some
proposed sensing schemes make use of SSB modulation directly). This leads to an imbalance
between phase-modulation sidebands at the photodetector(s) and so AM rejection and other
benefits normally associated with the RF modulation/demodulation schemes will be degraded.
Given that the modulation schemes are not particularly efficient (it is hard to achieve within
about 3 dB of the idea signal to noise ratio, so half of the power is wasted) it has been suggested
(by Peter Fritschel) that dc offset locking of the interferometer is considered. This can closely
approach the ideal signal to noise ratio. While it also eliminates a few noise-couplings from
having any significance (e.g. oscillator noise) it means that we have full sensitivity to a few
other noise sources. in particular, anything that changes the power at the beamsplitter must
be extremely well controlled. The dc technique eliminates the mixer and its problems (but the
system itself has a square-law nonlinearity).
There are 2 points which encourage the selection of the dc system. Firstly, the large power
recycling factor (160 with finesse 130 arm cavities) will produce a very low coupled cavity pole
in the interferometer (< 1 Hz). The interferometer will therefore tend to reject noise on the
input light. This rejection will be from 2 to 3 orders of magnitude at frequencies where it matters
most (it may not be practical to provide the same filtering for rf of the conventional scheme.
Secondly, the use of an output modecleaner can reduce the total light on the photodetector,
allowing a very modest offset from the dark fringe (probably < 1 pm).
This option simplifies enormously certain laser and input optics requirements, and could double
system efficiency. It may also be noted that it is possible to measure, with sufficient fidelity,
the light power in the beamsplitter region, thus providing either substraction or vito signals for
the main output. We only need to worry about effects in the signal recycling cavity and on to
the photodetector.
An in-depth noise analysis is needed urgently.
Note that it is anticipated that the dc sensing will only work after a conventional (but not
optimised) rf scheme has brought the system to the operating point.

3.3     Modecleaners
3.3.1   Input modecleaners

It is anticipated that laser frequency, relative amplitude and pointing noise will all have to
be at least as good as in LIGO I, and very possibly have to be a little better (based on some
very preliminary estimates). If a conventional rf modulation scheme is used for the main
signal detection, it is almost certain that a post-modulation modecleaner will be required. It
is very likely that a pre-modulation modecleaner (of non-commensurate length with the post-
modulation one) will be needed too. If the system is to be tuned over a wide range of frequencies
it is probable that we will need to vary one or more modulation frequencies to program the
tuning. It might be convenient to have a long post-modulation modecleaner to obtain the
required tuning. An analysis of this is urgently required. A long modecleaner may also provide


                                                3         5-4-2000   14:14DRAFT - for comment
the necessary frequency reference, given that it would have large spot sizes and less thermal
noise. (But does it fit, and is scattering a problem?)

3.3.2    Output modecleaners

See the note below. The effect of output modecleaners will be evaluated using Melody when it
is clear that the signal recycling action is simulated with know detuning.


4       Subsystem-level options
These are (more-or-less) confined to the ISC system since the subsystems do not perfectly map
to the working groups they can still be cross-group concerns.

4.1     Signal recycling/extraction cavity length
In principle 3 general lengths are possible: 0, and 4 km, with 2km being an option wherever
mid-stations are fitted. Although the potential benefits of long SR cavities were recognised
some time ago by Brian Meers, they have not been fully analysed in a modern context. This
analysis is required urgently.

4.2     Output modecleaners
The option of an output modecleaner becomes important whenever the modulation (or dc offset)
at the output of the interferometer is too weak to dominate the higher order modes ­ as it is
likely to be in LIGO II if we cannot modulate the whole input beam.

4.3     Variable signal recycling mirrors
There are 3 options.

4.3.1    Replacement

This is the slow, but technically obvious solution, if the mirror must be changed to make a
particular observation, then change it by physical replacement. The timescale is weeks and
there is a loss of observing time during the change. A juke-box solution might be feasible
although mechanically complex.

4.3.2    Thermally tuned

A solution under test for GEO 600 is to use a fused silica etalon as the signal recycling mirror.
Temperature changes of order 1 K provide tuning of the bandwidth over the full range. Heating
rings remote from the mirror can achieve this. The settling time is many hours (although there
is significant change in less than one hour). This is technically feasible, although manufacturing
requirements for the etalon are quite stringent (the GEO mirror is plane parallel, but the LIGO
one may need to have accurately matched and centred curves, this may be expensive to make).

4.3.3    Separately suspended mirrors

Two mirrors forming a linear cavity make the most versatile solution. Unfortunately there is
no proposed control system. So this option seems unavailable for LIGO II. (LIGO IIa?) The
complexity is in the control (including alignment) of the two mirrors.




                                                4        5-4-2000   14:14DRAFT - for comment
5    Teams
The system and sub-system concerns are being investigated by small teams within the AIC.
These teams will report each month. We will have some e-mail conferences and (as rarely
as possible given our geographical spread) teleconferences. The teams are looking at noise
couplings (Mason), DC locking (Fritschel), Output optics (McClelland), length of extraction
cavity (Mason), input modecleaners (with L&O group, Reitze), long modecleaner configuration
benefits (Strain), ITM optimisation (Strain), and general thermal analysis/Melody qualification
­ from the point of view of sensing efficiency (Strain). (Reports from the 4 small DR/RSE
experiments, the Caltech 40m and Glasgow 10m groups are very welcome too.)


6    Results and recommendations
empty


References
 [1] E. Gustafson, D. Shoemaker, K. Strain and R. Weiss, LIGO Science Collaboration White
     Paper on Detector Research and Development,
     http://www.ligo.caltech.edu/docs/T/T990080-00.pdf.




                                              5        5-4-2000   14:14DRAFT - for comment