Example 4 - extracting geometrical parameters¶
LIGO document DCC-T1900708 Cavity Tilts & Shifts [12] discusses the differences between the geometrical model and the modal model used by Finesse 2. This example replicates this work in Finesse 3 and discusses how to extract these geometrical parameters.
See also
See also tests/validation/hom/test_geometric_model.py which contains this code, reworked into a unit test.
Geometrical parameters¶
Chapter 19 of [13] provides the following relations for the cavity axis tilt and shift as a function of the mirror alignments.
where the nomenclature follows [12] and [10]. As in [12], we consider the case of a hemispherical cavity, where the input coupler is the flat mirror and the output coupler is the curved mirror. A graphical representation of the possible combinations is shown in [12].
When the curved mirror rotates, the cavity axis shifts, but remains parallel to the input beam. When the flat mirror rotates, the cavity axis rotates.
The cavity axis rotation can be expressed in terms of the difference in spot positions,
where L is the length of the cavity. These geometric relations can be expressed using the following python code,
import numpy as np
# Cavity parameters
L = 0.1 # Cavity Length
RcM2 = 0.5 # Radius of curvature of M2 mirror = 0.5 m
g1 = 1 # g1 = 1 - L/RcM1 = 1 since RcM1 = infinity
g2 = 1 - L/RcM2 # g2 for M2 mirror
print('g1g2 =', g1*g2) # Stability check
# Mirror Tilt:
npoints = 200
theta_low = -1e-6
theta_high = 1e-6
def geometric(mirror, theta_low, theta_high, npoints):
if mirror == 1:
# M1 tilt
theta_1 = np.linspace(theta_low, theta_high, npoints+1, endpoint=True)
# M2 tilt
theta_2 = 0
elif mirror == 2:
# M1 tilt
theta_1 = 0
# M2 tilt
theta_2 = np.linspace(theta_low, theta_high, npoints+1, endpoint=True)
else:
raise NotImplementedError('Mirror should be 1 or 2 and an integer')
# Resultant translation
## Displacement on M1
delta_x1 = (g2/(1-g1*g2))*L*theta_1 - (1/(1- g1*g2))*L*theta_2
## Displacement on M2
delta_x2 = (1/(1-g1*g2))*L*theta_1 - (g1/(1-g1*g2))*L*theta_2
## Angular displacement of optical axis
delta_theta = (delta_x2 - delta_x1)/L
print('Results for analytical solutions:')
print('∆x1: ', max(delta_x1), min(delta_x1), ' m')
print('∆x2: ', max(delta_x2), min(delta_x2), ' m')
print(r'∆θ: ', max(delta_theta), min(delta_theta), ' m')
return delta_x1, delta_x2, delta_theta
g1g2 = 0.8
where we have functionalised the geometric comparison for later analysis.
Finesse model¶
The system can be modeled in Finesse using the following code,
import matplotlib.pyplot as plt
import finesse
from finesse import init_plotting
init_plotting()
T = 1e-6
# Build Finesse model
kat = finesse.Model()
kat.parse(
f"""
laser laser P=1.0 f=0 phase=0
# Mirrors
mirror M1 R=(1-&M1.T) T={T} L=0.0 phi=0 Rc=[inf, inf] xbeta=0 ybeta=0
mirror M2 R=(1-&M1.T) T=&M1.T L=0.0 phi=0 Rc=[{RcM2}, {RcM2}] xbeta=0 ybeta=0
# Spaces
space s0 portA=laser.p1 portB=M1.p1 L=1 nr=1 gouy_x=0 gouy_y=0
space s_cav portA=M1.p2 portB=M2.p1 L=0.1 nr=1 gouy_x=0 gouy_y=0
# Amplitude Detectors
amplitude_detector det00_1 node=M1.p2.i f=0 n=0 m=0
amplitude_detector det00_2 node=M2.p1.o f=0 n=0 m=0
amplitude_detector det10_1 node=M1.p2.i f=0 n=1 m=0
amplitude_detector det10_2 node=M2.p1.o f=0 n=1 m=0
# Beam Property Detectors
beam_property_detector bp_n2 prop=0 node=M1.p2.i direction=x q_as_bp=false
beam_property_detector bp_n3 prop=0 node=M2.p1.o direction=x q_as_bp=false
beam_property_detector cavwaist prop=1 node=M1.p2.i direction=x q_as_bp=false
# Photodiodes
power_detector_dc pcirc node=M1.p2.o pdtype=none
# Config
cavity cavity1 source=M1.p2.o via=M2.p1.i priority=1
lambda(1.064e-06)
modes(maxtem=3)
"""
)
# Print beam tracing information
tsy = finesse.tracing.tools.propagate_beam(
to_node=kat.M2.p1.o, from_node=kat.M1.p2.o, direction='y'
)
print(tsy.make_table(tablefmt='simple'))
Name z [m] w0 [mm] zr [m] w [mm] RoC [m] Acc. Gouy [deg] q
------- ------- --------- -------- -------- --------- ----------------- ---------------
M1.p2.o 0 0.260262 0.2 0.260262 inf 0 0.000 + 0.200j
M2.p1.i 0.1 0.260262 0.2 0.290982 0.5 26.5651 0.100 + 0.200j
M2.p1.o 0.1 0.260262 0.2 0.290982 -0.5 26.5651 -0.100 + 0.200j
Note
Notice that only the transmissivity of M1 is specified. The reflectivity of M1 and the parameters of M2 are specified by reference, so that we can later vary M1 transmission and all other parameters will update. See KatScript for more information.
Here Python f-strings are used to substitute parameters into the KatScript.
Relating the models¶
We can relate the spot position shifts and tilts derived from the geometric optics model, to the mode amplitudes in the modal model. Starting from the relations in Table 1 of [14] we can rearrange to obtain (See section 4.4.2 of [15] for a recent derivation),
We now define a function run_simulation which handles executing the the simulation for
us. Additionally we can define a function process_sim which takes the
ArraySolution object and extracts the geometric parameters,
using the above relations, from our amplitude_detector and
beam_property_detector. See [12] for a precise derivation.
def run_simulation(mirror, theta_low, theta_high, npoints):
kat.parse(
f"""
# Actions
xaxis(
parameter=M{mirror:.0f}.xbeta,
mode=lin,
start={theta_low},
stop={theta_high},
steps={npoints},
pre_step=none,
post_step=none
)
"""
)
return process_sim(kat.run())
def process_sim(out):
# Comparison
acirc = np.sqrt(out['pcirc'].real) # circulating amplitude
# intermodal phase HG10 -> HG00
# at mirror 1
intermodal_phase_1 = np.angle(out['det10_1']) - np.angle(out['det00_1'])
# at mirror 2
intermodal_phase_2 = np.angle(out['det10_2']) - np.angle(out['det00_2'])
# Real part of \power normalised HG10 mode ampltiude
# at M1
a_shift_1 = (np.abs(out['det10_1'])/acirc)*np.cos(intermodal_phase_1)
# at M2
a_shift_2 = (np.abs(out['det10_2'])/acirc)*np.cos(intermodal_phase_2)
# Imaginary part of power normalised HG10 mode ampltiude
# at M1
a_tilt_1 = (np.abs(out['det10_1'])/acirc)*np.sin(intermodal_phase_1)
# at M2
a_tilt_2 = (np.abs(out['det10_2'])/acirc)*np.sin(intermodal_phase_2)
# Spot position shift
# at M1
dx_finesse_1 = out['bp_n2'].real*a_shift_1
# at M2
dx_finesse_2 = out['bp_n3'].real*a_shift_2
# Wavefront tilt at
# M1
dtheta_1 = a_tilt_1*kat.lambda0/(np.pi*out['cavwaist'].real)
# M2
dtheta_2 = a_tilt_2*kat.lambda0/(np.pi*out['cavwaist'].real)
print('Results for finesse solutions:')
print('Intermodal_1: ', max(intermodal_phase_1/np.pi), min(dx_finesse_1/np.pi), ' pi')
print('Intermodal_2: ', max(intermodal_phase_2/np.pi), min(dx_finesse_2/np.pi), ' pi')
print('∆x1: ', max(dx_finesse_1), min(dx_finesse_1), ' m')
print('∆x2: ', max(dx_finesse_2), min(dx_finesse_2), ' m')
print(r'∆θ1: ', max(dtheta_1), min(dtheta_1), ' m')
print(r'∆θ2: ', max(dtheta_2), min(dtheta_2), ' m')
return (
intermodal_phase_1,
intermodal_phase_2,
dx_finesse_1,
dx_finesse_2,
dtheta_1,
dtheta_2,
out.x[0]
)
Discussion¶
Lastly, we can define a function to process these results and plot the output.
def comparison_plot(mirror, theta_low, theta_high, npoints, ax=None):
delta_x1, delta_x2, delta_theta = geometric(
mirror, theta_low, theta_high, npoints
)
(
intermodal_phase_1,
intermodal_phase_2,
dx_finesse_1,
dx_finesse_2,
dtheta_1,
dtheta_2,
xaxis
) = run_simulation(mirror, theta_low, theta_high, npoints)
if ax is None:
fig, ax = plt.subplots(nrows=3, sharex=True, figsize=(4,6/3*4))
xaxis = 1e6 * xaxis
ax2col = 'darkgreen'
# Plot dx1
l3, = ax[0].plot(xaxis,1e6*delta_x1, '-', label='Geometric')
l4, = ax[0].plot(xaxis, 1e6*dx_finesse_1, '--',label='Modal')
ax[0].plot([], [], '--', color=ax2col, label='Difference')
ax[0].set_ylabel('$\Delta x_1$ [um]')
# Plot dx1 err
ax1_2 = ax[0].twinx()
l5, = ax1_2.plot(xaxis, 1e12*(delta_x1 - dx_finesse_1), '--', color=ax2col)
ax1_2.tick_params(axis='y', colors=ax2col)
ax1_2.set_ylabel('Diff. [pm]',color=ax2col)
# plot dx2
ax[1].plot(xaxis, 1e6*delta_x2, '-', label='Geometric')
ax[1].plot(xaxis, 1e6*dx_finesse_2, '--', label='Modal')
ax[1].plot([], [], '--', color=ax2col, label='Difference')
ax[1].set_ylabel('$\Delta x_2$ [um]')
# plot dx2 err
ax2_2 = ax[1].twinx()
ax2_2.plot(xaxis, 1e12*(delta_x2 - dx_finesse_2), '--', color=ax2col)
ax2_2.tick_params(axis='y', colors=ax2col)
ax2_2.set_ylabel('Diff. [pm]',color=ax2col)
# plot dtheta1
ax[2].plot(xaxis, 1e6*delta_theta, '-', label='Geometric')
ax[2].plot(xaxis, 1e6*dtheta_1, '--', label='Modal')
ax[2].plot([], [], '--', color=ax2col, label='Difference')
ax[2].set_ylabel(r'$\theta^x_0$ at M1 [urad]')
# plot dtheta1 err
ax3_2 = ax[2].twinx()
ax3_2.plot(xaxis, 1e12*(dtheta_1 - delta_theta), '--', color=ax2col)
ax3_2.tick_params(axis='y', colors=ax2col)
ax3_2.set_ylabel('Diff. [prad]',color=ax2col)
ax[0].legend()
ax[2].set_xlabel(f'M{mirror} Misalignment [urad]')
fig, ax = plt.subplots(nrows=3, ncols=2, sharex=True, figsize=(8,6/3*4))
comparison_plot(1, theta_low, theta_high, npoints, ax=ax[:,0])
comparison_plot(2, theta_low, theta_high, npoints, ax=ax[:,1])
ax[2,1].set_ylim([-1.1, +1.1])
fig.tight_layout()
Results for analytical solutions:
∆x1: 4.0000000000000014e-07 -4.0000000000000014e-07 m
∆x2: 5.000000000000001e-07 -5.000000000000001e-07 m
∆θ: 9.999999999999993e-07 -9.999999999999993e-07 m
Results for finesse solutions:
Intermodal_1: 0.14758329930279382 -1.273240787814041e-07 pi
Intermodal_2: -3.1834763941235237e-07 -1.5915492491370065e-07 pi
∆x1: 4.000003905247471e-07 -4.0000039052474715e-07 m
∆x2: 4.999999428915171e-07 -4.999999428915171e-07 m
∆θ1: 9.999955236676984e-07 -9.999955236676976e-07 m
∆θ2: 4.876993207838182e-12 -4.8769932079878946e-12 m
Results for analytical solutions:
∆x1: 5.000000000000001e-07 -5.000000000000001e-07 m
∆x2: 5.000000000000001e-07 -5.000000000000001e-07 m
∆θ: 0.0 0.0 m
Results for finesse solutions:
Intermodal_1: 1.5161488574404476e-07 -1.591551193134631e-07 pi
Intermodal_2: 0.8524165339644525 -1.5915515721722656e-07 pi
∆x1: 5.000005536163827e-07 -5.000005536163827e-07 m
∆x2: 5.000006726945674e-07 -5.000006726945675e-07 m
∆θ1: 1.1907818465421746e-12 -1.1907818474034485e-12 m
∆θ2: 1.118034161606098e-06 -1.1180341616060985e-06 m
In each plot the mirror angle is varied by from -1 μrad to +1 μrad, the left columns correspond to the M1 being rotated and the right columns correspond to M2 being rotated.
Note
Notice the wavefront tilt is only equal to the geometric cavity axis tilt at the beam waist and so these are only compared at M1. If the cavity does not have a beam waist at one of the Finesse nodes, it is not possible to directly compare these parameters.
As shown in the plots, the modal and geometric models agree up to pico-meter precision. The remaining error arises because the geometric model does not take into account the finite finesse of the optical resonator—which is fully modeled in the Finesse modal model. We can see the effect of this by reducing the mirror reflectivity.
kat.M2.xbeta = theta_high
kat.M1.xbeta = 0
kat.parse(
f"""
# Actions
xaxis(
parameter=M1.T,
mode=log,
start=1e-7,
stop=1,
steps={npoints},
pre_step=none,
post_step=none)
"""
)
(
intermodal_phase_1,
intermodal_phase_2,
dx_finesse_1,
dx_finesse_2,
dtheta_1,
dtheta_2,
xaxis
) = process_sim(kat.run())
delta_x1, delta_x2, delta_theta = geometric(2, 0, kat.M2.xbeta, npoints)
fig, ax = plt.subplots(figsize=(8,8/3))
## Plot phases
ax.loglog(xaxis, 1e6*np.abs(delta_x1[-1] - dx_finesse_1), '-', label='Plane Mirror (M1)')
ax.loglog(xaxis, 1e6*np.abs(delta_x2[-1] - dx_finesse_2), '--', label='Hemispherical Mirror (M2)')
ax.set_ylabel('Difference in predicted\n spot position [um]')
ax.set_xlabel('Mirror Transmissivity')
ax.set_title(f'Mirror M2 Tuning {1e6*kat.M2.xbeta:.0f}urad')
ax.legend()
Results for finesse solutions:
Intermodal_1: 0.0 -1.5915511931348423e-07 pi
Intermodal_2: 0.8524168204449615 -1.5915522883739875e-07 pi
∆x1: 0.0 -5.000005536164491e-07 m
∆x2: 0.0 -5.000008976959743e-07 m
∆θ1: 1.1174658828196761e-06 -3.4407969937731483e-12 m
∆θ2: 1.7335501779427744e-06 0.0 m
Results for analytical solutions:
∆x1: 0.0 -5.000000000000001e-07 m
∆x2: 0.0 -5.000000000000001e-07 m
∆θ: 0.0 0.0 m
<matplotlib.legend.Legend at 0x7f26ded02c40>
As shown, when the mirror transmission is very low, the spot position is accurately predicted up to a noise floor. This noise floor could either be numerical noise, or, the limitations of the approximations used in the modal model. When the mirrors are infinitely transmissive the geometric model fails completely, predicting a spot shift where none should occur.