BICEP3: a 95 GHz refracting telescope for degree-scale CMB
polarization
Z. Ahmed
†a,b
, M. Amiri
c
, S. J. Benton
d
, J. J. Bock
e,f
, R. Bowens-Rubin
g
, I. Buder
g
,
E. Bullock
h
, J. Connors
g
, J. P. Filippini
e
, J. A. Grayson
a,b
, M. Halpern
c
, G. C. Hilton
i
,
V. V. Hristov
e
, H. Hui
e
, K. D. Irwin
a,b
, J. Kang
a,b
, K. S. Karkare
g
, E. Karpel
a
, J. M. Kovac
g
,
C. L. Kuo
a,b
, C. B. Netterfield
d
, H. T. Nguyen
f
, R. O’Brient
e
, R. W. Ogburn IV
a,b
, C. Pryke
h
C. D. Reintsema
i
, S. Richter
g
, K. L. Thompson
a,b
, A. D. Turner
f
, A. G. Vieregg
j
,
W. L. K. Wu
a,b
, K. W. Yoon
a,b
a
Department of Physics, Stanford University, Stanford, CA 94305, USA
b
Kavli Institute for Particle Astrophysics and Cosmology, SLAC National Accelerator
Laboratory, Menlo Park, CA 94025, USA
c
Department of Physics and Astronomy, University of British Columbia, Vancouver, BC,
Canada
d
Department of Physics, University of Toronto, Toronto, ON, Canada
e
Division of Physics, Mathematics, & Astronomy, California Institute of Technology,
Pasadena, CA 91125, USA
f
Jet Propulsion Laboratory, Pasadena, CA 91109, USA
g
Harvard-Smithsonian Center for Astrophysics, Cambridge, MA 02138, USA
h
School of Physics & Astronomy, University of Minnesota, Minneapolis, MN 55455, USA
i
National Institute of Standards and Technology, Boulder, CO 80305, USA
j
Department of Physics, University of Chicago, Chicago, IL 60637, USA
ABSTRACT
Bicep3 is a 550 mm-aperture refracting telescope for polarimetry of radiation in the cosmic microwave back-
ground at 95 GHz. It adopts the methodology of Bicep1, Bicep2 and the Keck Array experiments — it pos-
sesses sufficient resolution to search for signatures of the inflation-induced cosmic gravitational-wave background
while utilizing a compact design for ease of construction and to facilitate the characterization and mitigation
of systematics. However, Bicep3 represents a significant breakthrough in per-receiver sensitivity, with a focal
plane area 5× larger than a Bicep2/Keck Array receiver and faster optics (f/1.6 vs. f/2.4). Large-aperture
infrared-reflective metal-mesh filters and infrared-absorptive cold alumina filters and lenses were developed and
implemented for its optics. The camera consists of 1280 dual-polarization pixels; each is a pair of orthogonal
antenna arrays coupled to transition-edge sensor bolometers and read out by multiplexed SQUIDs. Upon de-
ployment at the South Pole during the 2014-15 season, Bicep3 will have survey speed comparable to Keck Array
150 GHz (2013), and will significantly enhance spectral separation of primordial B-mode power from that of
possible galactic dust contamination in the Bicep2 observation patch.
Keywords: Inflation, Gravitational Waves, Cosmic Microwave Background, Polarization, BICEP, Keck Array
1. INTRODUCTION
Standard ΛCDM cosmology provides a successful and self-consistent framework that explains experimental data
sampling various cosmic epochs – light element abundances match those predicted by Big Bang Nucleosynthesis,
1
relic radiation from the last photon-electron scattering is observed as the Cosmic Microwave Background (CMB)
matching theory,
2, 3
observations of large-scale structure in the later history of the Universe appear to arise from
†
Corresponding Author: Z. Ahmed, 382 Via Pueblo Mall, Stanford, CA 94305. Email: zeesh at stanford dot edu
Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy VII,
edited by Wayne S. Holland, Jonas Zmuidzinas, Proc. of SPIE Vol. 9153, 91531N · © 2014
SPIE · CCC code: 0277-786X/14/$18 · doi: 10.1117/12.2057224
Proc. of SPIE Vol. 9153 91531N-1
the expected primordial power spectrum.
4
However, standard ΛCDM cosmology fails to explain the homogeneity
and isotropy of the observable Universe, as well as the observed geometric flatness of the Universe. Inflation,
an exponential expansion of space-time at the earliest epochs, resolves these issues and provides a mechanism
to stretch primordial quantum fluctuations in the early Universe to seed structure we observe today.
5–7
The
paradigm of Inflation already enjoys circumstantial evidence, most prominently by the observation of a power law
spectrum of perturbations and a measured departure from exact scale-invariance of perturbations.
3
However,
inflation models also generically predict a stochastic gravitational wave background generated by magnification of
quantum perturbations of the gravitational field.
8–10
The ratio of the amplitude of the tensor perturbations that
generate primordial gravitational waves to the amplitude of scalar perturbations that seed structure is called ‘r’,
and is used to characterize inflationary models as well as the energy scale of inflation. Primordial gravitational
waves are expected to interact with the cosmic microwave background at the surface of last scattering and
imprint a parity-odd linear polarization at degree angular scales.
11, 12
A detection of this primordial ‘B-mode’
polarization would provide direct evidence for inflation and gravitational waves, and would provide the first
direct hints of the quantum nature of gravity.
However, B-mode polarization is also generated by other mechanisms. Lensing of the parity-even or ‘E-mode’
component of CMB polarization by large-scale structure
13
is one such source. This component is quantifiable
and can be subtracted from spectra and maps.
14
A second source of contamination of B-modes is dust in our own
galaxy, along the line of sight to the CMB. Removal of this contaminant requires multi-color observations to enable
spectral discrimination between CMB signal and galactic foregrounds.
15
The level of challenge presented by these
contaminants to the measurement of primordial B-mode signal depends on the amplitude of the contaminants
relative to the amplitude of the primordial signal.
Recently, Bicep2 made the first detection of B-mode power on degree angular scales at 150 GHz.
16
The data
fit well a model that includes ΛCDM, expected B-mode power from lensing and r = 0.2, suggesting a detection of
primordial gravitational waves generated by inflation. However, a lack of sufficiently constraining spectral data
from the Bicep/Keck collaboration or other publicly available datasets permits the high value of r suggested by
the fit to be lowered by an uncertain amount to accommodate galactic dust contamination. The path forward to
deconvolving the foregrounds from the primordial signal lies in observations of the CMB at multiple frequencies.
Several existing and planned experiments are on track to do this in the next few years. Two receivers of the
Keck Array have begun re-observing the Bicep2 sky patch at 95 GHz.
17
Also, the Planck collaboration intends
to release data products and analysis results from its multi-frequency polarization data within the year. The
data from these experiments might have the sensitivity required to resolve the foreground uncertainties in the
Bicep2 results.
In parallel, the Bicep/Keck collaboration has continued R&D towards even more sensitive instruments to
cross check the Bicep2 results and measure CMB B-mode power at degree scales at higher significance. These
efforts have resulted in a new 95 GHz instrument called Bicep3. Bicep3 presents a breakthrough in CMB
polarimetry throughput and sensitivity for refracting telescopes. This single instrument doubles the traditional
Bicep2/Keck Array aperture and combines the detector count of five Bicep2-like receivers or the entire Keck
Array. This has been achieved by implementing modular detector-array packaging, improved infrared filtering
for large clear aperture, and fast alumina optics with a novel implementation of anti-reflection coating. These
proceedings discuss the design of Bicep3.
2. DETECTORS, READOUT & CAMERA
Bicep3 uses polarization-sensitive millimeter-wave detector technology developed at JPL/Caltech, and also
used in Bicep2/Keck Array.
18, 19
Each camera pixel combines two orthogonal slot-antenna networks, band-
defining filters, absorbers, and transition-edge sensor (TES) bolometers on a single silicon substrate, as shown
in Figure 1. Single-moded, diffraction-limited beams are obtained without the need for feed horns by in-phase
summation of power collected by antennae in a corporate feed network. The feed network is designed to minimize
beam pointing differences between the two polarizations for a pixel and uses gaussian-tapered illumination to
suppress side lobes.
20
The microwave power collected in a single polarization feed network is deposited on a
gold absorber coupled to a TES bolometer. To control the bolometer conductance to the thermal bath, the
absorber and bolometer sit separately on a silicon nitride island suspended from the substrate. The silicon tiles
Proc. of SPIE Vol. 9153 91531N-2
Figure 1. Left: An array of Bicep3 dual-polarization pixels under a microscope. Each square unit is a pixel and consists
of two 12x12 network of slot antennas, one each for an orthogonal polarization. Right: A transition edge sensor (TES)
island under a microscope. Each dual-polarization pixel has two TES islands, one for each slot-antenna network. Power
is introduced through the long microwave strip entering on the right side of the island and absorbed in an Au absorber
(meander). The incoming power changes the resistance of the Ti TES on the left side of the island.
are held at ∼ 280 mK and the titanium TESs self-heat to their superconducting transition temperature in the
range 470–530 mK. Fluctuations in incoming power alter the TES resistance, which in turn changes the current
through the bias circuit. An inductor in series with the TES couples the changes in current to superconducting
quantum interference devices (SQUIDs). Bicep3’s SQUIDs, designated MUX11d, were developed and fabricated
at NIST, Boulder.
21
The SQUIDs are arranged in time-domain multiplexing arrays at cold temperatures. Room
temperature electronic complete a fast negative-feedback loop to cancel magnetic flux changes in the series
inductors. The room temperature electronics, called Multichannel Electronics (MCE)
22
were developed at the
University of British Columbia and are implemented in several other experiments.
Depending on the desired frequency, ∼30–150 dual-polarization pixels can be patterned onto single 3”×
3” Si tiles via photolithography. Such detector-array tiles can be mass produced. Bicep3’s focal plane can
accommodate 20 detector-array tiles, packaged into ‘plug-and-play’ modules as shown in Figure 2. Each 95 GHz
module for Bicep3 has 64 dual-polarization pixels (128 bolometers) for a total of 1280 pixels (2560 bolometers).
A module consists of a detector-array tile and its supporting cold electronics, such that hand-affixed cryogenic
flex ribbon cable and mechanical fasteners are the only required connections between the module and the focal
plane heat sink. The module housings are fabricated using niobium to provide a superconducting magnetic shield
around the SQUIDs at operating temperatures. The detector-array modules present a significant simplification
from the Bicep2/Keck Array focal plane design; detector-array tiles of those experiments were directly wire
bonded to the focal plane and were challenging to modify or repair — a scheme nearly unviable for a focal plane
with 20 detector-array tiles.
3. RECEIVER DESIGN
The Bicep3 receiver is housed in a custom-designed vacuum cryostat 2.4 m tall and 0.73 m in diameter. It has
been designed to fit into the BICEP mount at the South Pole. In addition to azimuth and elevation rotation the
mount has rotation about its optical axis for instrumental systematics control. The Bicep3 receiver design is
modeled on Bicep1
23
and Bicep2.
24
A CAD cross-section is displayed in Figure 3. The receiver was designed
and constructed at Stanford University and SLAC. The vacuum jacket is sectioned into three lengthwise segments
for ease of access and assembly/disassembly, enabling rapid test and operation cycles. The top section houses
an HDPE vacuum window, a stack of reflective metal-mesh infrared shaders, an infrared-absorptive alumina
filter at 50 K, alumina lenses at 4 K, and a nylon filter also at 4 K. The middle section houses the sub-kelvin
focal plane with the associated thermo-mechanical structure. The lower section houses the cryogenic system and
cold electronics to support the focal plane, as well as ports to warm electronics. Inside the vacuum jacket, the
cryostat is radially partitioned by 4 K and 50 K radiation shields. The 4 K and 50 K volumes are nested using
Proc. of SPIE Vol. 9153 91531N-3
Figure 2. Top Left: A Bicep3 detector-array tile undergoing assembly into a module. Top Right: A Bicep3 module without
its niobium enclosure. The detectors face downward in this picture, exposing the circuit boards, SQUID multiplexing
chips, and wire bonds. A standard connector on top of the PCB provides electrical connections to the outside. Bottom:
Three modules installed in the Bicep3 receiver for testing.
Proc. of SPIE Vol. 9153 91531N-4
Series
SQUID
Arrays 4K
Chase He -3
Sorption Fridge
Filters and
connectors
for readout
HDPE Window
Reflective Metal -
mesh filters
Alumina filter 50K
Alumina Objective
Lens 4K
Nylon filter 4K
HR -10 baffled lens tube
Alumina Field Lens 4K
Detector modules
250 mK
PT-415
cryocooler
Figure 3. Cross section of Bicep3 receiver showing various components.
Proc. of SPIE Vol. 9153 91531N-5
