## AHFinder

Date

Abstract

Finding Apparent Horizons in a numerical spacetime

### 1 Purpose

Thorn AHFinder ﬁnds Apparent Horizons (AHs) in numerical spacetimes. It calulates various quantities like horizon area and its corresponding mass.

### 2 Using AHFinder

Thorn AHFinder can be used either with a minimization or a ﬂow algorithm.

#### 2.1 Parameters

##### General parameters
• ahfinder::ahf_active (steerable)
To activate the thorn set ahf_active = ”yes”. This parameter is set by default to ”no”.
• ahfinder::ahf_flow (steerable)
By default the minimization algorithm is used. To switch to the ﬂow algorithm one has to set ahf_ﬂow = ”yes”

##### Parameters used in evolutions
• ahfinder::ahf_findevery (steerable)
Speciﬁes how often the ﬁnder is called. The default is to ﬁnd horizons at every iteration.
• ahfinder::ahf_findafter (steerable)
The number of iterations after which the thorn is called the ﬁrst time can be speciﬁed by this parameter. Here the default is 0, calling the thorn also on the initial time slice.
• ahfinder::ahf_findaftertime
Instead of specifying the number of iterations, one can specify after how much coordinate time the thorn is called the ﬁrst time. When diﬀerent from zero, this parameter overrides the value of ahf_ﬁndafter. Default here is also 0.

##### Parameters specifying the expansion of the surface in sperical harmonics.
• ahfinder::ahf_lmax
The maximal number of terms in the expansion in $𝜃$. The default value is 8. The maximal value is 19.
• ahfinder::ahf_phi (steerable)
If axisymmetry is expected the surface does not need to be expanded in phi. This is the default. To look for non-axisymmetric surface use ahf_phi = ”yes”.
• ahfinder::ahf_[xyz]c (steerable)
Sets the x-, y-, and z-coordinates of the center of the expansion. The default is the origin (ahf_xc = 0, ahf_yc = 0, ahf_zc = 0). The center of the expansion should be set inside the expected apparent horizon, otherwise the algorithm will fail.
• ahfinder::ahf_wander
The center of the expansion can also be allowed to move. To do this use ahf_wander = ”yes”. However, this only works with the minimization algorithm. The default is for the center not to move.
• ahfinder::ahf_r0 (steerable)
Sets the radius of the initial sphere. The default is 0.0, forcing the largest sphere possible in the grid.

##### Looking for three horizons

The ﬁnder can also be used to look for three horizons each time. This is done by just running the algorithm three consecutive times with diﬀerent initial guesses, and is useful for simulations of black hole collisions.

• ahfinder::ahf_find3
Set ahf_ﬁnd3 = ”yes” to search for three horizons. The default is to look for only one horizon.
• ahfinder::ahf_[xyz]_ (steerable)
Sets the x-, y-, and z-coordinates of the center of the expansion for horizon 0, 1 and 2. The default in each case is the origin.
• ahfinder::ahf_r0_[0-2] (steerable)
Sets the radius of the initial spheres for horizon 0, 1 and 2. The default in all cases is 0.0, forcing the largest sphere possible in the grid.

##### Further parameters for the initial guess

The initial guess can be furthermore controlled by some parameters which are set to ”no” by default.

• ahfinder::ahf_guessold (steerable)
To use on old horizon as initial guess set ahf_guessold = ”yes”. However, if during the evolution the apparent horizon jumps discontinuously it might be lost by using this option.
• ahfinder::ahf_nn0, ahfinder::ahf_nn2
If no old horizon is used the inital guess can be speciﬁed further for the minimization algorithm. This algorithm is sensitive to the initial guess, so this is important. The initial guess is set up by an expansion in spherical harmonics in the ﬁrst two coeﬃcients (l=0,l=2). The default for both these parameters are 10, in which case the algorithm tests 100 diﬀerent combinations to ﬁnd the best initial guess.
• ahfinder::ahf_sloppyguess
It is also possible to use only a sphere as initial guess. This is much faster and is done by using ahf_sloppyguess = ”yes”. In this case a number of spheres (speciﬁed by ahf_nn0) with diﬀerent radii are tested for the initial guess.
• ahfinder::ahf_inner
If one wants to look for an inner horizon instead of an outer one, use ahﬁnder::ahf_inner = ”yes”. This only works with the minimization algorithm.

##### Parameters for surface intergrals
• ahfinder::ahf_ntheta (steerable)
The number of subdivisions in $𝜃$. Default is 200.
• ahfinder::ahf_nphi (steerable)
The number of subdivisions in $\varphi$. Default is 200.

##### Parameters indicating symmetries
• ahfinder::ahf_ref[xyz] (steerable)
Speciﬁes the existence of reﬂection symmetry on the yz-plane, xz-plane and xy-plane respectively. By default all are set to ”no”.
• ahfinder::ahf_octant (steerable)
Octant symmetry is speciﬁed by using ahf_octant = ”yes”. This is set to ”no” by default. Possible parmeter settings are ”yes” for reﬂection symmetries on all three coordinate planes and ”high” for an additional rotational symmetry of $\pi ∕2$ around the z axis.
• ahfinder::ahf_cartoon (steerable)
A further symmetry can (must) be speciﬁed when running with the ”axisymmetric” mode given by the Cartoon method by using ahf_cartoon = ”yes”. This is set to ”no” by default.

##### Parameters for minimization algorithm
• ahfinder::ahf_tol
Sets the tolerance for the minimization algorithm. The default value is 0.1.
• ahfinder::ahf_maxiter
Sets the maximum number if iterations for each step of the POWELL algorithm. The default value is 10.
• ahfinder::ahf_minarea
Usually the square of the expansion is minimized. To switch to minimization of the area one can use ahf_minarea = ”yes” (default is ”no”). Notice that only for time symmetric data the surface found by area minimization will correspond to an apparent horizon.

##### Parameters for the ﬂow algorithm
• ahfinder::ahf_flow (steerable)
The fow algotithm is used by setting ahf_ﬂow = ”yes”. The default is ”no”
• ahfinder::ahf_flowiter (steerable)
Sets the maximum number of iterations for the ﬂow algorithm. The default value is 200.
• ahfinder::ahf_flowtol
Sets the tolerance for the ﬂow. The default value is 0.0001.
• ahfinder::ahf_flowa
Sets the $\alpha$ parameter for the ﬂow. The default value is 0.01.
• ahfinder::ahf_flowb
Sets the $\beta$ parameter for the ﬂow. The default value is 0.5.
• ahfinder::ahf_flowh
Sets the weight of $H$ ﬂow. The default value is 0.0.
• ahfinder::ahf_flowc
Sets the weight of $C$ ﬂow. The default value is 1.0.
• ahfinder::ahf_flown
Sets the weight of $N$ ﬂow (not yet implemented). The default value is 0.0.

The character of the diﬀerent ﬂows and the $\alpha$ and $\beta$ parameters are described in Carsten Gundlach’s paper on his pseudo-spectral apparent horizon ﬁnder (gr-qc/9707050).

##### Parameters for output
• ahfinder::ahf_logfile
By default no logﬁle for AHFinder is written. To obtain a log ﬁle one must set ahf_logﬁle = ”yes”.
• ahfinder::ahf_verbose
if ahf_verbose = ”yes” messages are printed to screen at the beginning and the end of the algorithm. This is the default.
• ahfinder::ahf_veryverbose
if ahf_veryverbose = ”yes” messages are also printed to screen during the iteration process. The default is ”no”.
• ahfinder::ahf_2Doutput (steerable)
To get 2D output of grid functions use ahf_2Doutput = ”yes” (default is ”no”). The output is controlled be the thorn itself, not by cactus standard output.
• ahfinder::ahf_3Doutput (steerable)
In future versions to get 3D output of grid functions use ahf_3Doutput = ”yes” (default is ”no”).
• ahfinder::ahf_areamap
To get an area map use ahf_areamap = ”yes” (default is ”no”). This is useful for looking at the behaviour of the area for surfaces close to the horizon.

##### Parameters for mask
• ahfinder::ahf_mask (steerable)
The mask is 0 inside the horizon and 1 outside, and is used in black-hole excision techniques. By default the mask is oﬀ. It can be enabled by setting ahf_mask = ”strong”, which sets the mask only if the ﬁnder is sure that a horizon was found, or by setting ahf_mask = ”weak”, which makes the ﬁnder set the mask also if a horizon is probably there but either the resolution or lmax are to low to really resolve it.
• ahfinder::ahf_masktype (steerable)
The mask can be of two types. Either ahf_masktype = ”simple”, which masks a cube contained inside the horizons, or ahf_masktype = ”lego”, which masks a lego-sphere.
• ahfinder::ahf_maskshrink (steerable)
Sets a buﬀer zone between the region inside the horizon and the region where the mask is 0, by limiting the mask to a region smaller than ahf_maskshrink times the horizon radius. The default value is 0.8.
• ahfinder::ahf_maskbuffer (steerable)
Sets a buﬀer zone between the region inside the horizon and the region where the mask is 0 with a width of at least ahf_maskbuﬀer grid points. The default value is 5.

#### 2.2 Minimal parameter settings

Usually only a few of the parameters described above are needed in the parameter ﬁle. The simplest parameter settings for using the ﬂow algorithm for a full 3D horizon with a large sphere as initial guess is

interpolation_order = 2     # Second order interpolation
ahf_active          = "yes"
ahf_flow            = "yes"
ahf_phi             = "yes"

This looks for a non-axisymmetric horizon around the origin with lmax = 8 and using the ﬂow algorithm. It starts with the largest sphere that ﬁts in the 3D grid and outputs 2D grid functions. The other parameters can be used if needed.

#### 2.3 Hints for parameter settings

In full 3D the ﬂow algorithm is faster than the minimization algorithm. However, in cases when there are very few terms in the expansion in spherical harmonics the minimization can be faster. In axisymmetry this typically happens for lmax$\le 10$.

While the default settings usually work ﬁne, they can be changed to meet special purposes:

• If the horizon is expected to be far from spherical the parameter ahf_lmax can be set to a higher value. 12 should be high enough. However, only values up to 20 are supported.
• If the latter parameter is set to a value higher than 8 then the parameter ahf_maxiter can be raised to, e.g., 14. This can be useful since more iterations can be necessary for higher coeﬃcients of the expansion.

#### 2.4 Output to Files

The output of the thorn consists of two gridfunctions and several one dimensional output ﬁles.

• To depict the position of the horizon, the most important ﬁles are ahfgrid_2d_...ieee. These ﬁles contain a 2D gridfunction whose zero level locates the horizon.
• The ﬁles ahf_exp_2d_...ieee show the expansion of outgoing photons on the level set of the gridfunction ahfgrid. The horizon coincides with zeros of the expansion.
• The surface area of the horizon is given in ahf_area.tl
• The mass of the horizon is given in ahf_mass.tl
• The coeﬃcients of the expansion in spherical harmonics are given in ahf_coeﬀ.alm.
• The ﬁles ahf_circ_eq.tl, ahf_meri_p1.tl and ahf_meri_p2.tl contain the equatorial circumference of the surface, the length of the meridian at $\varphi =0$, and the length of the meridian at $\varphi =\pi ∕2$ of the surface, respectively.
• If an output of a logﬁle is set in the parameters, the log ﬁle for the last time the horizon was called is ahf_logﬁle.

#### 2.5 Some results with the ﬁnder

The ﬁnder has been examined with puncture initial data for single and binary-black hole scenarios.

Calculations with diﬀerent grid spacings but constant grid size show convergence of the horizon area.

This has been checked with diﬀerent linear momenta in the z direction ${P}_{z}=\left(0M,2M,5M\right)$ and vanishing spin. Also for ${P}_{z}=2M$ and a spin of $5M$ in the x direction the horizon converges. Figure 1 shows the case with ${P}_{z}=2M$ and vanishing spin. Figure 1: Convergence of the horizon area for ${P}_{z}=2M$

Further, not only the area converges but also the shape of the horizon. For both the minimization and the ﬂow algorithm the horizon converges to the same shape, as can be seen from the coeﬃcients fo the expansion. The order of convergence for the coeﬃcients is between 1.4 and 1.7.

By using the parameters ahf_xc, ahf_yc, ahf_zc it can also be shown that the ﬁnder also locates horizons which are not centered. This works in general as long as the surface can be expanded in spherical harmonics around this point, but the error increases with the oﬀ-centering.

The parameter ahf_r0 can be used e.g. when dealing with two black holes. If one searches for separate horizons one can center the ﬁnder on one of the locations of the holes and use an initial radius ahf_r0 smaller than the coordinate distance of the holes. With this parameter settings the single horizon can be found faster. But also a setup with an initial sphere of maximum radius should work at least for the ﬂow algorithm. This has been checked with puncture data for two holes with vanishing linear and angular momentum for each hole (equivalent to Brill-Lindquist data) and is shown in Figure 2. Here for a coordinate distance of the holes of 1.6M the separated horizons for the holes are found but no common horizon. For a coordinate distance of 1.5M a common horizon is found and also single ones, which are inner surfaces in this case. This coincides with other work where the critical coordinate distance for a single horizon is between 1.53M and 1.56M (gr-qc/9809004). Figure 2: Horizon positions for BL data