B
ICEP
/
Keck
XV: The B
ICEP
3 Cosmic Microwave Background Polarimeter and the First
Three-year Data Set
P. A. R. Ade
1
, Z. Ahmed
2
, M. Amiri
3
, D. Barkats
4
, R. Basu Thakur
5
, C. A. Bischoff
6
, D. Beck
2
,
7
, J. J. Bock
5
,
8
, H. Boenish
4
,
E. Bullock
9
, V. Buza
10
, J. R. Cheshire IV
9
, J. Connors
4
, J. Cornelison
4
, M. Crumrine
11
, A. Cukierman
2
,
7
, E. V. Denison
12
,
M. Dierickx
4
, L. Duband
13
, M. Eiben
4
, S. Fatigoni
3
, J. P. Filippini
14
,
15
, S. Fliescher
11
, N. Goeckner-Wald
7
, D. C. Gold
fi
nger
4
,
J. Grayson
7
, P. Grimes
4
, G. Hall
11
, G. Halal
7
, M. Halpern
3
, E. Hand
6
, S. Harrison
4
, S. Henderson
2
, S. R. Hildebrandt
5
,
8
,
G. C. Hilton
12
, J. Hubmayr
12
, H. Hui
5
, K. D. Irwin
2
,
7
,
12
, J. Kang
5
,
7
, K. S. Karkare
4
,
10
, E. Karpel
7
, S. Kefeli
5
,
S. A. Kernasovskiy
7
, J. M. Kovac
4
,
16
, C. L. Kuo
2
,
7
, K. Lau
11
, E. M. Leitch
10
, A. Lennox
14
, K. G. Megerian
8
, L. Minutolo
5
,
L. Moncelsi
5
, Y. Nakato
7
, T. Namikawa
17
, H. T. Nguyen
8
,R.O
’
Brient
5
,
8
, R. W. Ogburn IV
2
,
7
, S. Palladino
6
, T. Prouve
13
,
C. Pryke
9
,
11
, B. Racine
4
,
18
, C. D. Reintsema
12
, S. Richter
4
, A. Schillaci
5
, R. Schwarz
11
, B. L. Schmitt
4
, C. D. Sheehy
19
,
A. Soliman
5
, T. St. Germaine
4
,
16
, B. Steinbach
5
, R. V. Sudiwala
1
, G. P. Teply
5
, K. L. Thompson
2
,
7
, J. E. Tolan
7
, C. Tucker
1
,
A. D. Turner
8
, C. Umiltà
6
,
14
, C. Vergès
4
, A. G. Vieregg
10
,
20
, A. Wandui
5
, A. C. Weber
8
, D. V. Wiebe
3
, J. Willmert
11
,
C. L. Wong
4
,
16
,W.L.K.Wu
2
, H. Yang
7
, K. W. Yoon
2
,
7
, E. Young
2
,
7
,C.Yu
7
, L. Zeng
4
, C. Zhang
5
, and S. Zhang
5
(
B
ICEP
/
Keck
Collaboration
)
1
School of Physics and Astronomy, Cardiff University, Cardiff, CF24 3AA, UK
2
Kavli Institute for Particle Astrophysics and Cosmology, SLAC National Accelerator Laboratory, 2575 Sand Hill Rd, Menlo Park, CA 94025, USA
3
Department of Physics and Astronomy, University of British Columbia, Vancouver, British Columbia, V6T 1Z1, Canada
4
Center for Astrophysics, Harvard & Smithsonian, Cambridge, MA 02138, USA
5
Department of Physics, California Institute of Technology, Pasadena, CA 91125, USA;
hhui@caltech.edu
6
Department of Physics, University of Cincinnati, Cincinnati, OH 45221, USA
7
Department of Physics, Stanford University, Stanford, CA 94305, USA
8
Jet Propulsion Laboratory, Pasadena, CA 91109, USA
9
Minnesota Institute for Astrophysics, University of Minnesota, Minneapolis, MN 55455, USA
10
Kavli Institute for Cosmological Physics, University of Chicago, Chicago, IL 60637, USA
11
School of Physics and Astronomy, University of Minnesota, Minneapolis, MN 55455, USA
12
National Institute of Standards and Technology, Boulder, CO 80305, USA
13
Service des Basses Températures, Commissariat à l
’
Energie Atomique, F-38054 Grenoble, France
14
Department of Physics, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA
15
Department of Astronomy, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA
16
Department of Physics, Harvard University, Cambridge, MA 02138, USA
17
Kavli Institute for Physics and Mathematics of the Universe
(
WPI
)
, UTIAS, The University of Tokyo, Kashiwa, Chiba 277-8583, Japan
18
Aix-Marseille Université, CNRS
/
IN2P3, CPPM, Marseille, France
19
Physics Department, Brookhaven National Laboratory, Upton, NY 11973, USA
20
Department of Physics, Enrico Fermi Institute, University of Chicago, Chicago, IL 60637, USA
Received 2021 September 30; revised 2021 December 22; accepted 2022 January 3; published 2022 March 7
Abstract
We report on the design and performance of the B
ICEP3
instrument and its
fi
rst three-year data set collected from
2016 to 2018. B
ICEP3
is a 52 cm aperture refracting telescope designed to observe the polarization of the cosmic
microwave background
(
CMB
)
on degree angular scales at 95 GHz. It started science observation at the South Pole
in 2016 with 2400 antenna-coupled transition-edge sensor bolometers. The receiver
fi
rst demonstrated new
technologies such as large-diameter alumina optics, Zotefoam infrared
fi
lters, and
fl
ux-activated SQUIDs, allowing
∼
10
×
higher optical throughput compared to the
Keck
design. B
ICEP3
achieved instrument noise equivalent
temperatures of 9.2, 6.8, and 7.1
m
Ks
CMB
and reached Stokes
Q
and
U
map depths of 5.9, 4.4, and 4.4
μ
K arcmin
in 2016, 2017, and 2018, respectively. The combined three-year data set achieved a polarization map depth of
2.8
μ
K arcmin over an effective area of 585 square degrees, which is the deepest CMB polarization map made to
date at 95 GHz.
Uni
fi
ed Astronomy Thesaurus concepts:
Cosmic microwave background radiation
(
322
)
;
Polarimeters
(
1277
)
;
Gravitational waves
(
678
)
;
Cosmic in
fl
ation
(
319
)
1. Introduction
In
fl
ation, a brief period of exponential expansion in the early
universe, was postulated to solve the horizon,
fl
atness, and
monopole problems that arise from the
Λ
CDM
“
standard
model
”
of the universe
(
Brout et al.
1978
; Kazanas
1980
;
Starobinsky
1980
; Guth
1981
; Albrecht & Steinhardt
1982
;
Linde
1982
)
. The perturbations under this paradigm are
adiabatic, nearly Gaussian, and close to scale-invariant, which
are consistent with precise cosmic microwave background
(
CMB
)
observations
(
Planck Collaboration et al.
2020a
)
.
Moreover, many models of in
fl
ation predict the existence of
primordial gravitational waves
(
PGWs
)
, which would leave a
unique degree-scale
B
-mode polarization pattern in the
The Astrophysical Journal,
927:77
(
30pp
)
, 2022 March 1
https:
//
doi.org
/
10.3847
/
1538-4357
/
ac4886
© 2022. The Author
(
s
)
. Published by the American Astronomical Society.
Original content from this work may be used under the terms
of the
Creative Commons Attribution 4.0 licence
. Any further
distribution of this work must maintain attribution to the author
(
s
)
and the title
of the work, journal citation and DOI.
1
CMB
(
Kamionkowski et al.
1997
; Seljak & Zaldarriaga
1997
)
.
If detected, PGWs can serve as a probe of the very early
universe and high-energy physics inaccessible with existing
particle accelerators.
The B
ICEP
/
Keck
experiments are a series of telescopes
designed to search for this degree-scale
B
-mode polarization of
the CMB originating from PGWs. These instruments are
located at the Amundsen
−
Scott South Pole Station in
Antarctica. The
∼
10,000 ft altitude and extreme cold make
the Antarctic plateau one of the driest places on Earth. During
the winter season, the six months of continuous darkness
provides exceptionally low and stable atmospheric 1
/
f
noise,
which allows our telescopes to observe the sky without the
need of an active instrument modulation at these large angular
scales
(
Kuo
2017
)
.
We
fi
rst reported an excess of
B
-mode signal at 150 GHz in
BICEP2 collaboration et al.
(
2014a
)
. In a subsequent joint
analysis with the
Planck
Collaboration, it was found that
polarized emission from dust in our galaxy could account for
most of the signal
(
BICEP2
/
Keck & Planck collaborations
et al.
2015
)
. Dust is currently the dominant foreground
contaminant to CMB polarization measurements, and is most
powerful at high frequencies. Subsequent modeling shows
synchrotron may potentially be another source of foreground
emission at lower frequencies
(
Krachmalnicoff et al.
2018
)
.In
order to probe the physics of the early universe, we need a
dedicated strategy to separate these foregrounds from the
potential faint primordial signal.
The B
ICEP
/
Keck
instruments are small-aperture, compact,
on-axis refracting telescopes, emphasizing high optical
throughput and low optical loading with dedicated calibration
campaigns to control instrument systematics. Five separate
instruments spanning the past two decades have been deployed
to date. B
ICEP
1 operated from 2006 through 2008 with 98
neutron transmutation doped
(
NTD
)
germanium thermistors at
95, 150, and 220 GHz
(
Chiang et al.
2010
; Takahashi et al.
2010
)
.B
ICEP
2 replaced B
ICEP
1 and observed from 2010
through 2012 with 512 planar antenna transition-edge sensors
at 150 GHz
(
BICEP2 collaboration et al.
2014b
)
.
Keck
utilized
the same optical and detector technologies as employed in
B
ICEP
2, comprising
fi
ve independent receivers. It observed at
150 GHz, and later at 95 and 220 GHz, installed in a separate
telescope mount previously used for DASI
(
Leitch et al.
2002
)
and QUaD
(
Ade et al.
2008
)
. It began science observations in
2012, observing until 2019
(
Kernasovskiy et al.
2012
;
Staniszewski et al.
2012
)
.
After B
ICEP
2 was decommissioned at the end of 2012,
B
ICEP3
was installed in the same telescope mount in 2014
November and started scienti
fi
c observation in 2016 with 2400
detectors at 95 GHz. It employed a conceptually similar design
to its predecessor, but with multiple technological improve-
ments allowing an order of magnitude increase in mapping
speed compared to a single
Keck
95 GHz receiver. Bene
fi
ting
from a modular receiver design,
Keck
was gradually adapted
from an all 150 GHz receiver con
fi
guration into a high-
frequency
“
dust telescope,
”
observing at 220 and 270 GHz,
with B
ICEP3
continuing observations at 95 GHz, where fore-
grounds are minimal. In late 2019,
Keck
was decommissioned
and replaced with a new telescope mount
(
Crumrine et al.
2018
)
to accommodate four B
ICEP3
-like receivers that will
form the next phase of the experiment, B
ICEP
A
RRAY
. The
fi
rst
receiver in B
ICEP
A
RRAY
started observation at 30
/
40 GHz in
2020 to probe the low-frequency polarized synchrotron signal.
B
ICEP
A
RRAY
will cover six distinct bands from 30 to
270 GHz when fully deployed. In the meantime, the B
ICEP
A
RRAY
telescope mount carries a mixture of
Keck
and B
ICEP
A
RRAY
receivers, while B
ICEP3
continues to observe. Table
1
shows the B
ICEP
/
Keck
experiments from 2010 to 2020 and
their frequency coverage.
This paper provides an overview of the B
ICEP3
instrument
design and performance with the three-year data set from 2016 to
2018. Figure
1
shows the overall layout of B
ICEP3
as it is installed
at the South Pole. The following sections describe the details of
each of the subcomponents: telescope mount
(
Section
2
)
; optics
(
Section
3
)
; cryostat
(
Section
4
)
; focal plane unit
(
Section
5
)
;
transition-edge sensor bolometers
(
Section
6
)
; and data acquisition
and control system
(
Section
7
)
.
In particular, B
ICEP3
ʼ
s 520 mm diameter aperture is
∼
2
times the size of the
Keck
design. This is realized by the large-
diameter alumina optics shown in Section
3.1
. The increase in
aperture size allowed us to accommodate 2400 detectors in the
focal plane, compared to 288 detectors in the previous
Keck
95 GHz receivers. The new modular focal plane design in
Section
5
allows rapid rework and dramatically reduces risk.
The high number of detectors also requires a mature multi-
plexing readout. B
ICEP3
is the
fi
rst experiment to adapt the new
generation
fl
ux-activated time domain multiplexing system
described in Section
7
. Most CMB experiments utilize low
temperature, superconducting detectors that operate below 1 K.
Rapid development in mechanical compressor cryocoolers
allowed ground-based telescopes to phase out the need of
liquid helium, but the high-pressure helium lines in the system
Table 1
Frequency Coverage in the B
ICEP
/
Keck
Experiment from 2010 to 2020
Receiver
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
B
ICEP2
150 GHz 150 GHz 150 GHz
Keck Rx0
150 GHz 150 GHz 95 GHz
95 GHz
220 GHz
220 GHz
220 GHz
220 GHz
220 GHz
Keck Rx1
150 GHz 150 GHz 150 GHz 220 GHz 220 GHz
220 GHz
220 GHz
[
150 GHz
]
Keck Rx2
150 GHz 150 GHz 95 GHz
95 GHz
220 GHz
220 GHz
220 GHz
220 GHz
220 GHz
Keck Rx3
150 GHz 150 GHz 150 GHz 220 GHz 220 GHz
220 GHz
220 GHz
220 GHz
Keck Rx4
150 GHz 150 GHz 150 GHz 150 GHz 150 GHz
[
270 GHz
]
270 GHz
270 GHz
270 GHz
B
ICEP3
[
95 GHz
]
95 GHz
95 GHz
95 GHz
95 GHz
95 GHz
BA Rx0
30
/
40 GHz
Note.
Brackets in the table indicate an engineering receiver
(
270 GHz
Keck
in 2017 was a prototype of high-frequency focal plane, B
ICEP3
in 2015 only had a partially
fi
lled focal plane, and the 150 GHz
Keck
in 2019 was a demonstration of the
μ
MUX readout; Cukierman et al.
2020
)
, and are not included in science analyses.
Keck
was replaced by B
ICEP
A
RRAY
in 2020. In its
fi
rst season, one slot was
fi
tted with the 30
/
40 GHz B
ICEP
A
RRAY
receiver, and three
Keck
receivers were put back into
the new telescope mount. This paper uses the data collected by B
ICEP3
from 2016 through 2018.
2
The Astrophysical Journal,
927:77
(
30pp
)
, 2022 March 1
Ade et al.
between the telescope and compressor induce signi
fi
cant wear
in a continuous rotating mount. We address this by integrating
a helium rotary joint into the telescope mount system, allowing
for continuous rotation while maintaining a high-pressure seal
and electrical connectivity
(
Section
2
)
.
The achieved performance characteristics of the receiver and
detector properties of B
ICEP3
are presented in Section
8
, the
observing strategy is presented in Section
9
, and in Section
10
we show the
fi
rst three-year data set taken from 2016 to 2018,
reporting its internal consistency validation, sensitivity, and
map depth. The cosmological analysis using
Planck
, Wilkinson
Microwave Anisotropy Probe
(
WMAP
)
, and B
ICEP
/
Keck
observations through the 2018 observing season are presented
in BICEP
/
Keck et al.
(
2021
)
.
2. Telescope Mount, Forebaf
fl
e, and Ground Shield
2.1. Telescope Mount
B
ICEP3
is installed in the Dark Sector Laboratory building,
approximately a kilometer away f
rom the South Pole Station. The
base of the telescope mount is supported by a platform on the
second
fl
oor of the building, with a 2.4 m diameter opening in the
roof for telescope access to the sky
(
Figure
1
)
. The warm indoor
environment of the building is extended beyond the roof level by
a
fl
exible insulating environmental s
hield, so that only the receiver
window is exposed to the Antarctic ambient temperature.
B
ICEP3
uses a steel three-axis mount built by Vertex-RSI.
21
It was originally built for B
ICEP
1 and also housed B
ICEP
2 until
2013. The mount structure was modi
fi
ed in 2014 to
accommodate the larger B
ICEP3
receiver.
The mount moves in azimuth and elevation, with the third
axis rotating about the boresight of the telescope
(
“
deck
”
rotation
)
. The range of motion of the mount is 48
°
–
110
°
in
elevation and 400
°
in azimuth, capable of scanning at speeds of
5
°
s
−
1
in azimuth.
The B
ICEP3
cryostat houses a pulse tube cryocooler that
limits the accessible deck angle to less than a full 360
°
rotation
in the B
ICEP
mount. However, the design still allows the
telescope to scan with two sets of 180
°
opposing deck angles,
offset from each other by 45
°
, retaining an effective set of
observation schedules in order to probe systematic errors, as
shown in Section
9
.
2.2. Helium Rotary Joint
The pulse tube cryogenic cooler comprises two subsystems:
a coldhead installed inside the receiver, and a helium
compressor located in the building, away from the telescope
mount. This pulse tube provides cooling by expanding a high-
pressure helium gas volume, and requires high and low
pressure helium
fl
exible lines to be routed from the compressor,
through the three mount axes
(
azimuth, elevation, and
boresight
)
, to the coldhead in the receiver.
During an observing schedule, movements in elevation and
boresight are intermittent and span a limited range of angles, unlike
the azimuth axis, which scans back and forth continuously in
azimuth with a 130
°
range. To avoid wear on the compressor lines
in the helium line wrap, B
ICEP3
uses a commercial high-pressure
gas rotary joint from DSTI
22
that enables the two pressurized
helium gas lines to pass through the azimuth motion. In this
joint, shown in Figure
2
, one set of lines remains static at the
base of the mount and connects to the helium rotary joint, from
which a second set of lines rotates with the azimuth axis of the
mount. Therefore the azimuth cable carrier only needs to
handle the much more
fl
exible electrical cables.
During the 2015 engineering season, the original design used
a basic two channel rotary joint
(
DSTI model: GP-421
)
to
Figure 1.
The B
ICEP3
telescope in the mount, looking out through the roof of the Dark Sector Laboratory
(
DSL
)
located
∼
1100 m from the geographic South Pole.
The insulating environmental shroud shown in the bottom right photo is hidden in the CAD layout. The three-axis mount previously used in B
ICEP
1 and B
ICEP
2
allows for motion in azimuth, elevation, and boresight rotation. A comoving absorptive forebaf
fl
e extends skyward beyond the cryostat receiver to intercept stray light
outside the designed
fi
eld of view. Additionally, the telescope is surrounded by a stationary re
fl
ective ground shield that redirects off-axis rays to the cold sky.
21
Now General Dynamics Satcom Technologies, Newton, NC 28658,
http:
//
www.gdsatcom.com
/
vertexrsi.php
.
22
Dynamic Sealing Technologies, Inc., Andover, MN 55304,
www.dsti.com
.
3
The Astrophysical Journal,
927:77
(
30pp
)
, 2022 March 1
Ade et al.
connect the compressor
’
s high and low pressure helium
channels at 290 and 90 PSI, respectively. However, helium
gas can permeate materials and gaps much more easily than
other larger gas molecules, and this commercial rotary joint
was not designed speci
fi
cally for helium gas. We found the
overall system lost 3
–
5 PSI of pressure per day, originating in
the dynamic seals of the rotary joint. Such a large leak resulted
in the need to re
fi
ll the compressor system multiple times a
week to maintain optimal pulse tube performance. In addition
to being extremely labor intensive for the telescope operator,
these repeated helium re
fi
lls introduced contamination into the
pulse tube, and eventually degraded the cooling performance.
To remedy the high leak rate, the rotary joint was replaced
with a four channel model
(
DSTI model: GP-441
)
before the
2016 season. The four channels were con
fi
gured such that the
two working high-pressure helium lines would be guarded by
two outer channels, serving as pressurized buffers. Thus, the
dynamic seals between the two inner high-pressure lines would
only
“
sense
”
the small differential pressure to the pressurized
buffers
(
∼
10 PSI
)
instead of the much larger differential to
atmospheric pressure
(
>
100 PSI
)
. This con
fi
guration reduced
the leak rate of the active channels to between
∼
0.1 and 0.5 PSI
per day over an entire season.
23
The reduced helium leak rate
requires less frequent re
fi
lls, and enables optimal pulse tube
performance throughout a full season. The HRJ dynamic seals
receive a complete replacement once per year.
2.3. Ground Shield and Absorptive Baf
fl
e
A warm, absorptive forebaf
fl
e as shown in Figure
1
extends
skyward beyond the cryostat receiver to intercept stray light
outside the designed
fi
eld of view. The forebaf
fl
e is mounted
directly to the receiver and therefore comoving with the axes of
motion of the telescope. The forebaf
fl
e is constructed from a
large aluminum cylinder, 1.3 m in diameter and height, with a
rolled top edge lined by microwave-absorptive Eccosorb HR-
10 foam. Heater tape keeps the forebaf
fl
e a few degrees
above the Antarctic ambient temperature to help avoid snow
accumulation, and a layer of closed-cell polyethylene foam
(
Volara
)
protects the Eccosorb from accumulating moisture.
Based on radiative loading on the detectors observed once the
forebaf
fl
e is installed, the forebaf
fl
e contributes
∼
10% of the
total optical power. The source of this wide-angle response is
likely a combination of scattering and multiple re
fl
ections.
Additionally, the telescope i
s surrounded by a stationary
re
fl
ective ground shield. It is
fi
xedtotheroofofthebuildingtoact
as a second barrier against stray light and signal contamination
from nearby ground sources and reduces the large radiative
gradient between the sky and the ground. The ground shield is
10 m in diameter and 3 m in height, constructed with aluminum
honeycomb panels and steel beams. The combination of the baf
fl
e
and the ground shield is designed such that off-axis rays from the
telescope must diffract at least twice before intercepting the
ground.
2.4. Star Camera
An optical camera is used to determine mount pointing
parameters
(
Section
9.4
)
. It is attached to the side of the
receiver vacuum jacket, and looks up through a hole in the
bottom of the forebaf
fl
e. An optical baf
fl
e reduces stray light
when using the star camera during daylight and twilight
conditions, but is removed for CMB observations. The
telescope is a Newtonian re
fl
ector, with a 10 cm aluminum-
coated
24
objective and a 44 cm focal length. A 700 nm low-
pass edge
fi
lter removes much of the Rayleigh-scattered
sunlight during daylight and twilight. The camera is a CCD
with video readout,
25
and the video-to-digital conversion is
done with a video capture card
26
in one of the control
computers. The
fi
eld of view is approximately 0.8
°
×
0.6
°
. The
CCD is on a linear stage to allow focusing via a remote
controller used by the operator.
3. Optics
3.1. Optical Design
B
ICEP3
utilizes the same concept as previous B
ICEP
/
Keck
receivers, using a compact, on-axis, two-refractor optical
design that provides a wide
fi
eld of view and a telecentric
Figure 2.
Photos of the four channel helium rotary joint
(
HRJ
)
system. Left: two 30
°
bends rotate with the azimuth axis and go on to the receiver through the elevation
and boresight axes. Right: static section with the four connections for the high and low pressure helium and their respective guard channels. In both p
hotos, two ball-
end rod joints act as torque arms to transmit the azimuth rotation to the rotor of the HRJ.
23
The guard channels still have similar leak rate as the two channel design, but
this is acceptable since re
fi
lls for them do not affect the pulse tube performance.
24
Edmund Optics, Inc., Barrington, NJ, USA.
25
Astrovid StellaCam Ex, Adirondack Video Astronomy, 72 Harrison Ave.,
Hudson Falls, New York, USA; the CCD is a Sony ICX248AL B
/
W, Sony
Group Corporation, 1-7-1 Konan Minato-ku, Tokyo, 108-0075 Japan.
26
Sensoray, 7313 SW Tech Center Dr., Tigard, OR, USA.
4
The Astrophysical Journal,
927:77
(
30pp
)
, 2022 March 1
Ade et al.
focal plane. It has a 4 K aperture of 520 mm and beamwidth
given by a Gaussian radius
σ
∼
8.9
′
. The lenses and
fi
lters
operate at cryogenic temperatures inside of the cryostat receiver
to minimize excess in-band photon loading. The HDPE plastic
cryostat window is at ambient temperature. Thermal
fi
lters
mounted behind the window cool radiatively. Table
2
shows B
ICEP3
ʼ
s optical design parameters compared to pre-
vious B
ICEP
/
Keck
receivers.
The ray diagram and full optical chain are shown in Figure
3
.
The radially symmetric optical design allows for well-matched
beams for two idealized orthogonally polarized detectors at the
focal plane.
3.2. Vacuum Window and Membrane
The
fi
rst optical element in the receiver is the vacuum
window. B
ICEP
2
/
Keck
used laminated Zotefoam
27
(
Zotefoam
HD30
)
, but B
ICEP3
instead uses a 31.75 mm thick, 73 cm
diameter high-density polyethylene
(
HDPE
)
window due to the
larger aperture setting more stringent requirements on its
mechanical strength. The surfaces of the HDPE window are
coated with a
λ
/
4 antire
fl
ective
(
AR
)
layer made of Teadit
24RGD
28
(
expanded PTFE sheet
)
. The AR coating adheres to
the window with a thin layer of low-density polyethylene
plastic, melted in a vacuum oven press.
In front of the window is a 22.9
μ
m thick biaxially oriented
polypropylene membrane to protect the window from snow and
create an enclosed space below, which is slightly pressurized
with room-temperature nitrogen gas to evaporate snow that
falls onto the membrane surface.
3.3. Large-diameter 300 K Filters
Inside the receiver and directly behind the vacuum window
is a set of infrared
fi
lters to reduce the thermal loading in the
receiver. These are a stack of 10 thin
fi
lters mounted on a set of
aluminum rings mechanically connected to the room-temper-
ature vacuum jacket. The
fi
lters re
fl
ect or absorb infrared
radiation in stages, and radiatively equilibrate at progressively
lower temperatures to reduce the thermal infrared power into
the cryostat.
The original design used a set of metal-mesh
fi
lters,
composed of 3.5
μ
m Mylar or 6
μ
m polypropylene
/
polyethy-
lene
(
PP
/
PE
)
fi
lm, prealuminized to a 40 nm deposition
thickness and laser ablated to form a grid of metal squares
(
Ahmed et al.
2014
)
. However, we found that the performance
of the metal-mesh
fi
lter depended on the etching process of the
metal on the thin
fi
lm, and minor defects in fabrication
introduced excess in-band scattering from the
fi
lters. The in-
band scattering was slightly polarized, leading to additional
millimeter-wave power on the detectors and associated photon
noise. Furthermore, simulations using high-frequency structure
simulator
(
HFSS
29
)
software indicated of order 0.5% specular
re
fl
ection per layer even without defects.
All of the metal-mesh
fi
lters except for one placed behind the
50 K alumina
fi
lter were replaced in 2017 with a set of ten
3.17 mm thick Zotefoam layers
(
Figure
4
)
. These
fi
lters are
nitrogen-expanded polyethylene foam layers that both scatter
and absorb
/
emit infrared radiation
(
IR
)
isotropically and
therefore act as
fl
oating blocking layers
(
Choi et al.
2013
)
,
while maintaining
>
99% transmission in-band. Using room-
temperature transmission measurements, we estimate an 8%
improvement of in-band transmission compared to the metal-
mesh
fi
lters. Table
3
details the individual
fi
lters used
in B
ICEP3
.
3.4. Alumina Thermal Filter and Optics
Motivated by the larger aperture diameter and faster
f
/
1.6
speed in B
ICEP3
, we developed large-diameter alumina
fi
lters
and their antire
fl
ection coating. Alumina lenses are much
thinner and less aggressively shaped than their HDPE
equivalents owing to the signi
fi
cantly higher index of refraction
at
n
=
3.1. The alumina optics are 21 and 27 mm thick at the
center for the
fi
eld and objective lens, respectively, compared
to
>
67 mm for a comparable HDPE design. Both the lenses
and 50 K
fi
lter are made from 99.6% pure alumina sourced
from CoorsTek.
30
The reduction in thickness and high thermal conductivity of
alumina
(
0.5Wm
−
1
K
−
1
at 4 K
)
enables the optical elements to
cool to base temperatures more rapidly and limits any thermal
gradient across the lenses to less than 1 K from center to edge.
Lab measurements of similar alumina materials indicate low in-
Table 2
Optical Design Parameters for B
ICEP
2 at 150 GHz and B
ICEP3
at 95 GHz
B
ICEP
2
/
B
ICEP
3
Keck
Aperture dia.
264 mm
520 mm
Field of view
15
°
27.4
°
Beamwidth
σ
¢
1
2
¢
8.9
Focal ratio
f
/
2.2
f
/
1.6
Figure 3.
Ray diagram including the elements of the optical chain. The 300 K
metal-mesh
fi
lters were replaced by a stack of 10 Zotefoam
fi
lters in 2017,
which improved both the IR loading on the cryostat and the in-band power
incident on the detectors.
27
Plastazote HD30 from Zotefoams, Inc., Walton, KY 41094, USA,
www.
zotefoams.com
.
28
TEADIT North America, Pasadena, TX, USA.
29
Ansys,
www.ansys.com
.
30
CoorsTek, Golden, CO 80401, USA,
www.coorstek.com
.
5
The Astrophysical Journal,
927:77
(
30pp
)
, 2022 March 1
Ade et al.
band absorption at room temperature that decreases with
temperature
(
Penn et al.
1997
; Inoue et al.
2014
)
. Our own
measurements at room temperature indicated signi
fi
cant
differences between various formulae, and the CoorsTek AD-
996 Si used for the
fi
lter and lenses was the best we tested.
After deployment, we also con
fi
rmed a substantially decreased
loss at 77 K. A single 10 mm thick alumina disk serves as an
absorptive thermal
fi
lter, mounted on the 50 K cryogenic stage.
The high mid-infrared absorption and high thermal conductiv-
ity make alumina a choice material for this application.
The AR coating used for the alumina optics is a mixture of
Stycast 1090 and 2850FT with a homogeneous refractive index of
n
=
1.74. The epoxy is poured and rough molded to 1 mm
thickness on the alumina surface, then either machined
(
lenses
)
or
abrasively ground
(
fl
at
fi
lter
)
to the
fi
nal 0.452 mm thickness. The
thickness of the coating is controlled to less than 25
μ
m tolerance
by referencing precoating surfa
ce measurements of the alumina.
Historically, alumina optics were limited to small sizes
unless accommodation for differential contraction between the
alumina and the epoxy was made. Inoue et al.
(
2014
)
and
Rosen et al.
(
2013
)
put slices through their coatings to allow
cryogenic operation. We adopted a laser-cutting technique using
Laserod,
31
the same commercial laser machining company that
etched the IR blocking metal-mesh
fi
lm
fi
lters described above.
The laser cuts in the AR epoxy are
∼
30
μ
m wide, tuned to
reach the alumina surface, and spaced every 10 mm in a square
grid pattern
(
Figure
5
)
.
3.5. Nylon IR Blocking Filters
Following the same machining and coating approach in
B
ICEP
2
/
Keck
(
BICEP2 collaboration et al.
2014b
)
, two Nylon
IR blocking
fi
lters are placed in the receiver. One is behind the
aperture stop, and the other is behind the
fi
eld lens, above the
focal plane
(
FPU
)
assembly, both at 4 K. Nylon strongly
absorbs far-infrared radiation
(
Halpern et al.
1986
)
and thus
reduces radiated power from 50 K from reaching the 280 mK
focal plane.
3.6. Metal-mesh Low-pass Edge Filters
A set of metal-mesh low-pass edge
fi
lters
(
Ade et al.
2006
)
with a cutoff at 4 cm
−
1
were used to control any out-of-band
response in the detectors. They are made from multiple
polypropylene substrate layers, each coated with copper grids
in different sizes, and hot-pressed together to form a resonant
fi
lter.
Prior to the 2017 season, these
fi
lters were cut into
76
×
76 mm squares and independently mounted onto each
detector module
(
Section
5
)
at 280 mK. We found anomalous
detector spectral responses in the 2016 FTS measurements
described in Section
8.1
. Upon examining the
fi
lters at the end
of the 2016 season, we found the layers had delaminated. It was
determined that the cause of delamination was likely
insuf
fi
cient oven temperature during fabrication. Furthermore,
the cutting of individual, smaller
fi
lters introduced extra stress
on the edge contributing to the delamination.
New
fi
lters were fabricated using a higher oven temperature
in the fusing process. The
fi
lter design was modi
fi
ed to a larger
∼
23
×
15 cm size covering
fi
ve detector modules. This change
Figure 4.
Stack of 10 layers of room-temperature IR
fi
lters installed in B
ICEP3
,
immediately behind the vacuum window. This photo shows the current
con
fi
guration, with each layer composed of 3.17 mm thick HD30 foam, glued
onto a stack of aluminum frames with 3.17 mm spacing. The original design
was a stack of metal-mesh
fi
lters, which was replaced in 2017.
Table 3
Room-temperature IR Filters Installed in B
ICEP3
2016
Square
/
pitch
2017
+
Location
Substrate
(
μ
m
)
Substrate
Behind window
3.5
μ
m Mylar
50
/
80
HD30 foam
(
∼
290 K
)
3.5
μ
m Mylar
40
/
55
HD30 foam
3.5
μ
m Mylar
50
/
80
HD30 foam
3.5
μ
m Mylar
40
/
55
HD30 foam
3.5
μ
m Mylar
90
/
150
HD30 foam
6
μ
mPP
/
PE
40
/
55
HD30 foam
3.5
μ
m Mylar
50
/
80
HD30 foam
3.5
μ
m Mylar
40
/
55
HD30 foam
3.5
μ
m Mylar
50
/
80
HD30 foam
3.5
μ
m Mylar
90
/
150
HD30 foam
Behind 50 K
3.5
μ
m Mylar
90
/
150
3.5
μ
m Mylar
Alumina
fi
lter
Note.
The main stack of 10
fi
lters behind the window are listed in order
beginning with the closest
fi
lter to the window. The metal-mesh
fi
lters were
replaced by Zotefoam in 2017.
Figure 5.
AR coated alumina
fi
lter in B
ICEP3
. The alumina
fi
lter is coated with
a mix of Stycast 1090 and 2850FT. The epoxy is machined to the correct
thickness and laser diced to 1 cm squares to mitigate differential thermal
contraction between alumina and the epoxy.
31
Laserod Technologies LLC, Torrance, CA 90501, USA.
6
The Astrophysical Journal,
927:77
(
30pp
)
, 2022 March 1
Ade et al.
reduced mechanical stress at the
fi
lter edges caused by the
dicing process. Extra spring loaded washers and widened
mounting slots allowed the
fi
lter to slide more freely during
thermal contraction. These modi
fi
cations were done at the end
of the 2016 season and subsequent FTS measurements showed
no evidence of
fi
lter delamination.
3.7. Optical Loading Reduction
The dominant noise source in B
ICEP3
is the photon noise of
the in-band power. For better sensitivity, it is important to
minimize the internal non-sky instrument load.
We measured the upper limit of the total internal loading by
measuring the detector load curves with a
fl
at aluminum sheet
mounted just beyond the cryostat window. The detector beams
in the middle of the focal plane were re
fl
ected back into the
cryostat and trace all the way back to the sub-Kelvin focal
plane, passing through the optical chain twice. The general
detector load curve measurements are described in Section
8.2
.
The detector beams were re
fl
ected and therefore received
power only from within the cryostat. The comoving forebaf
fl
e
loading is measured by differencing of the detector response
with and without the forebaf
fl
e during clear weather while
looking at zenith.
The optical power coupled to the detectors due to each
individual optical element is calcu
lated by using the transmission
properties of the material and
fi
nding the source temperature
distribution. The calculation inc
orporates the cumulative optical
ef
fi
ciency from the detector up to the source, and includes the
emissivity of the source itself. For the 2016 design, a simple
scattering model is used for the room-temperature metal-mesh
fi
lters, in which each
fi
lter isotropically sca
tters a small fraction of
the radiation to wide angles an
d warm surfaces around them.
Table
4
shows the modeled in-band loading estimate for each of
the individual elements and th
e total measured loading. The
agreement between them validates the model assumption.
A signi
fi
cant contributor to the cryostat internal loading was
the scattering of the metal-mesh IR-re
fl
ective
fi
lters before their
replacement in 2017 with the Zotefoam
fi
lters. The reduction of
scattered radiation coupling to the
fi
lters and telescope
forebaf
fl
e results in a decrease of the total instrument loading
by 30%.
Non-sky loading also comes from the room-temperature
HDPE cryostat window, which now dominates the internal
power. We are developing thinner materials that can potentially
replace the window in future seasons
(
Barkats et al.
2018
)
.
4. Cryostat Receiver
4.1. Overview
The cryostat receiver is a compact, cylindrical design that
allows for a large optical path while maintaining sub-Kelvin
focal plane temperatures
(
Figure
6
)
. The fully populated
receiver weighs about 540 kg without the attached electronics
subsystems. The outermost aluminum vacuum jacket is 2.4 m
tall along the optical axis and 73 cm in diameter, excluding the
pulse tube cryocooler extension. It maintains high vacuum for
thermal isolation and is capped at one end by the HDPE plastic
window, as described in Section
3.1
.
The wide-
fi
eld refractor design allows for ground-based
characterization in the optical far
fi
eld. The optical design
further allows the use of a comoving, absorptive forebaf
fl
e
(
Section
2.3
)
that terminates wide-angle responses from the
receiver. Cooling most of the optical elements, including the
internal baf
fl
ing between the lenses, to less than 4 K reduces
the thermal photon noise seen by the detectors, maximizing the
sensitivity of the instrument.
4.2. Cryogenic and Thermal Architecture
Nested within the room-temperature vacuum jacket are the
50 and 4 K stages, each composed of cylindrical aluminum
radiation shields and cooled by the
fi
rst and second stages of
the PT-415 pulse tube cryocooler,
32
which provides continuous
cooling to 35 K at the
“
50 K stage
”
under typical 26 W load
and 3.3 K at
“
4 K stage
”
under 0.5 W load. The stages are
mechanically supported off each other and the vacuum jacket
by low thermal conductivity G-10
fi
berglass. Multilayer
insulation
(
MLI
)
wrapped around radiation shields minimizes
radiative heat transfer between the 300
−
50
−
4 K stages.
A non-continuous, three-stage
(
4
He
/
3
He
/
3
He
)
helium sorp-
tion fridge from Chase Research Cryogenic
33
is heat sunk to the
4 K stage and cools the sub-Kelvin focal plane and supporting
structures. The focal plane and ultra-cold
(
UC 250 mK
)
stage is a
planar copper assembly mounted in a vertical stack on two buffer
stages, the intercooler
3
He
(
IC 350 mK
)
and
4
He
(
2K
)
stages, each
supported and isolated by carbon
fi
ber trusses
(
Figure
7
)
.TheUC
stage cools a 9 mm thick, 46 cm diameter focal plane plate that
supports the detector modules and a thinner secondary plate. These
plates are made from gold-plated, oxygen-free high thermal
conductivity copper. The secondary plate and the focal plane are
separated by seven 5 cm tall stain
less steel blocks that serve as
passive low-pass thermal
fi
lters to dampen thermal
fl
uctuations to
the focal plane. The focal plane and the UC stage are actively
temperature controlled in a feedback loop to 274 mK and 269 mK,
respectively, using neutron transmutation doped
(
NTD
)
germanium
thermometers and a resistive heater. Thermal
fl
uctuations on the
focal plane during CMB observation are controlled to
<
0.1 mK.
Table 4
Per-detector In-band Optical Load
Source
Load
(
pW
)
T
RJ
(
K
)
4K lenses and elements
0.15
1.0
50K alumina
fi
lter
0.12
0.9
Metal-mesh
fi
lters
(
2016
)
0.63
5.2
HD30 foam
fi
lters
(
2017
+
)
0.10
0.8
Window
0.69
5.9
Total cryostat internal
(
2016
)
1.60
13.0
Total cryostat internal
(
2017
+
)
1.10
8.6
Forebaf
fl
e
(
2016
)
0.31
2.7
Forebaf
fl
e
(
2017
+
)
0.14
1.1
Atmosphere
1.10
9.9
CMB
0.12
1.1
Total
(
2016
)
3.13
27
Total
(
2017
+
)
2.46
21
Note.
The total loadings listed in
bold
are direct measurements from detector
load curves, which are in good agreement with the individual modeled optical
elements. A stack of 300 K metal-mesh
fi
lters used in the 2016 season were
replaced by HD30 foam
fi
lters for the 2017 season.
32
Cryomech Inc., Syracuse, NY 13211, USA,
www.cryomech.com
.
33
Chase Research Cryogenics Ltd., Shef
fi
eld, S10 5DL, UK,
www.
chasecryogenics.com
.
7
The Astrophysical Journal,
927:77
(
30pp
)
, 2022 March 1
Ade et al.
4.3. Thermal Performance
The sum of all incident thermal power on the 50 and 4 K
stages determines the temperature pro
fi
le of the elements along
each stage and the base operating temperature of the pulse tube.
The room-temperature HDPE plastic window emits
∼
110 W
of power into the receiver while the pulse tube cryocooler is
rated for less than 40 W on the 50 K stage. We employed two
different types of thermal
fi
lters at 300 K mounted just behind
the cryostat window to reject the majority of the IR load:
(
1
)
a
stack of thin
fi
lm, IR-re
fl
ective, capacitive metal-mesh
fi
lters in
2016; and
(
2
)
a stack of Zotefoam
fi
lters starting in 2017. The
reason for switching the design is discussed in Section
3.3
.
Figure 6.
Cutaway view of the B
ICEP3
cryogenic receiver. The thermal architecture is separated into a two-stage pulse tube cryocooler
(
50 K, 4 K stages
)
and a three-
stage helium sorption fridge
(
2 K, 350 mK, 250 mK stages
)
. All thermal stages are mechanically supported by sets of carbon
fi
ber and G-10
fi
berglass supports. The
focal plane, with 20 detector modules and 2400 detectors, is located at the 250 mK stage, surrounded by multiple layers of RF and magnetic shielding.
Figure 7.
Left: exploded view of the B
ICEP3
sub-Kelvin stages. Each temperature stage is mechanically supported by sets of carbon
fi
ber trusses. Sets of stainless steel
supports connect the two 250 mK copper plates, passively low-pass
fi
ltering thermal
fl
uctuations, and two active temperature control modules maintain thermal
stability over observation cycles. Right, top: the assembled focal plane with 20 detector modules installed into the 250 mK stage without metal-mesh
edge
fi
lters. The
module slot in the lower right is empty due to the capacity of the readout electronics. Right, bottom: a thin aluminized Mylar shroud extends from the to
p of the focal
plane assembly to the bottom of the 4 K plate to close the 4 K Faraday cage.
8
The Astrophysical Journal,
927:77
(
30pp
)
, 2022 March 1
Ade et al.
An alumina
fi
lter is heatsunk to the 50 K stage to provide
absorptive IR
fi
ltering due to Alumina
’
s mid-infrared absorp-
tion and high thermal conductivity. This replaced the
fi
lters
used in previous telescopes. Two additional nylon
fi
lters are
placed in the 4 K stage of the receiver to reduce thermal loading
on the sub-Kelvin focal plane by absorbing infrared radiation.
Table
5
shows the
fi
nal temperature and power deposited onto
each cryogenic stage. Switching from metal-mesh
fi
lters to
Zotefoam
fi
lters in 2017 reduced the thermal loading and
improved the cryogenic hold time of the sub-Kelvin fridge
from
∼
50 to
>
80 hr, with 6 hr of recycling time due to
improvement of the 4 K base temperature, allowing closer to
full condensation of
3
He in the sub-Kelvin fridge. This permits
the continuous three-day observation schedule shown in
Section
9.2
.
4.4. Cryogenic Thermal Monitoring and Control
For general thermometry down to 4 K, we use silicon diode
thermometers
(
Lakeshore
34
DT-670
)
, with thin-
fi
lm resistance
temperature detectors
(
Lakeshore Cernox RTDs
)
on the sub-
Kelvin stages. NTD germanium thermometers are integrated in
the secondary UC stage, copper focal plane, and each detector
module for more sensitive measurements of the temperatures.
The NTDs on the secondary UC stage and focal plane are
packaged with a heater to provide active temperature control on
their respective temperature stages.
Thermal operations are controlled by a custom-built
systemsimilartotheoneusedinB
ICEP
2
/
Keck
.Itcontains
electronic cards used to bias and read out thermometers,
control heaters, and provide temperature control servos, and
is mounted directly to the cryostat vacuum jacket and
interfaces with MicroD
(
MDM
)
connectors
35
at the cryostat.
Signals from the system are routed to the rack-mounted
BLASTbus analog-to-digital converter
(
ADC
)
system
(
Wiebe
2008
)
next to the telescope that generates the AC bias
used for the resistive thermometers and the NTDs, and
demodulates the thermometer signals, which are then digitized
at
∼
100 Hz.
4.5. Radio Frequency Shielding
Several levels of radio frequency
(
RF
)
shielding are designed
into the 4 K stage and sub-Kelvin structures to minimize RF
coupling to the detectors. All cabling inside the cryostat uses
twisted pairs, except for the short lengths of
fl
ex ribbon cable
connecting the detector modules to the focal plane readout
circuit board. These ribbon cables are shielded by the detector
module, copper focal plane module cutout, and the ground
plane of the circuit board that accepts the cable. The 4 K
nonoptics volume is designed as a Faraday enclosure, with all
seams taped with conductive aluminum tape and cabling
passing through inductive-capacitive PI-
fi
ltered connectors.
36
The cage is continued to enclose the stack of sub-Kelvin stages
by wrapping and sealing a single layer of aluminized Mylar
between the 4 K stage and the edge of the focal plane. The
niobium enclosure of each detector module and detector tile
ground plane close the sky side of the Faraday cage. Upon
exiting the cryostat, all of the detector signal lines immediately
interface with a capacitive
fi
ltered connection on the readout
electronics box that is directly mounted on the cryostat.
During the 2015 engineering season, we found an azimuth-
synchronous signal strongly affecting the detectors, largely
common-mode across a large fraction of detectors within each
readout system. This interference showed variation 1000 times
larger than the 50
μ
K CMB temperature variations, causing
“
SQUID jumps
”
because of the strong signals. Our detector
readout scheme works through feedback to maintain linearity in
the SQUID ampli
fi
cation curve
(
Section
7.1
)
, but large current
variations can disrupt the feedback and cause the readout to
jump to a different part of the SQUID curve. We discovered
that this interference signal was caused by RF emission from
the South Pole Station land mobile radio
(
LMR
)
system at
450 MHz, coupling into the cryostat and detectors through the
cryostat window. B
ICEP3
is inherently more susceptible to this
450 MHz signal than
Keck
due to its larger aperture, which has
a cutoff frequency at 340 MHz at the optics cylinder.
Prior to the 2016 season, we applied silver loaded paste
between the detector modules and copper focal plane, so that
reliable electrical conductivity was maintained from the
modules to the focal plane. In 2015, only 9 out of 20 detector
modules were
fi
lled, leaving large gaps at the top of the focal
plane. In 2016, having the full population of 20 detector
modules provided a better RF shielded enclosure. After
implementing the improved internal cryostat shielding, RF
susceptibility in the range of 400
–
500 MHz was reduced by
10 dB. In addition, the LMR antenna was changed to a
directional sector antenna with reduced power output toward
the telescope. Attenuators were also installed to reduce the
overall broadcast power, which was tested to be much more
powerful than necessary to maintain radio communication
across the base. In total, the LMR source power seen at the
telescope was reduced by 35 dB. Azimuth scanning tests
conducted after these changes have shown none of the visible
structure seen in 2015.
4.6. Magnetic Shielding
Earth
’
s magnetic
fi
eld
(
∼
50
μ
T
)
is the most dominant
magnetic environment. While this azimuth-
fi
xed signal is
largely
fi
ltered out during analysis, instrumental magnetic
Table 5
Measured Final Temperature and Thermal Loading on Each Temperature Stage
in B
ICEP3
2016
2017
+
Stages
Temp
/
Load
Temp
/
Load
50 K tube top
58 K
53 K
50 K tube bottom
52 K
49 K
50 K tube loading
19 W
13 W
4 K tube top
4.96 K
4.68 K
4 K tube bottom
4.58 K
4.33 K
4 K tube loading
0.18 W
0.15 W
350 mK stage
354 mK
352 mK
350 mK loading
91
μ
W84
μ
W
250 mK stage
245 mK
244 mK
250 mK loading
3.50
μ
W
3.35
μ
W
Focal Plane
268 mK
268 mK
34
Lake Shore Cryotronics, Westerville, OH 43082, USA,
www.lakeshore.com
.
35
Glenair Inc., Glendale, CA 91201, USA,
www.glenair.com
.
36
Cristek Inc., Anaheim, CA 92807, USA,
www.cristek.com
.
9
The Astrophysical Journal,
927:77
(
30pp
)
, 2022 March 1
Ade et al.