Development of TES-microcalorimeter arrays and Frequency Domain Multiplexed read-out Henk Hoevers

Netherlands Institute for Space Research

Development of

TES-microcalorimeter arraysand

Frequency Domain Multiplexed read-out

Henk HoeversDivision Sensor Research and Technology


Research Facilities for TES-microcalorimeter array and FDM development

Two Kelvinox 100 dilution fridges • 1 + 2 SQUID channels• moveable slit to position X-ray beam (micron resolution)• data-acquisition facilities: RT-, IV-curves, analog and digital pulse processing• X-ray sources: 5.9 keV and 1.5/2.2/3.0/3.3/5.9 keV

In-house clean room -> short turn-around times for detector research• class 10-100• processing on 4” wafers, sputter deposition, thermal evaporation, • spinners, mask aligners, wire bonders, inspection equipment

Staff• Physicists: 6 fte (senior) scientists• Electronics: 2.5 fte (senior) design engineers• Support staff: 2.5 fte (mechanical, electronical, lab assistant)• Clean room staff: 3 full-time persons


Funding of TES-microcalorimeter array and FDM development

SRON staff from NWO income (Dutch Organization for Scientific Research)

ESA TRP contract ‘Cryogenic Imaging Spectrometer’ (XEUS – NFI2)• SRON

TES-based arrays (thin film processing, testing), prime contractor• MESA+, Twente University, the Netherlands

micromachining development• VTT Automation Technology, Espoo, Finland

SQUID development • Space Research Center, Leicester, UK

detector packaging, pulse processing• University of Jyvaskyla, Finland

material parameters at sub-Kelvin temperatures• Metorex, Espoo, Finland

electrical crosstalk simulations (dense wiring)

SRON and partners will tender on X-10 (expected ESA TRP on array read-out)


Set-up of the talk

TES microcalorimeter array development

Status of Frequency Domain Multiplexing

Outlook area: energy resolution


Microcalorimeter array development for XEUS; approach

1) development of 5 x 5 pixel array with XEUS specification 2) address/investigate scalability from 5 x 5 array to 32 x 32 array

Production: fabrication of prototype 5 x 5 arrays following two routes• bulk micromachining• surface micromachining

Performance characterization of 5 x 5 pixel arrays• R(T), I(V) curves and their reproducibility• noise and energy resolution

Detector (re)design (5 x 5 -> 32 x 32) uses• performance (and understanding) of 5 x 5 arrays• additional measurements of all relevant low-temperature material parameters• development of a Finite Element Model of the 5 x 5 array (thermal design)

The Finite Element Model is also suited for performance analysis (time dependentpulse modeling) and investigation of other pixel sizes and/or geometries


Bulk and surface micromachining (SRON-MESA); the 5 x 5 arrays

TiAu TES (100 mK)and Cu absorber

on slottedSiN membrane

240 μm


R(T) and I(V) curves

Bulk MMThree pixels in the same array

Bulk MMThree pixels in different arrays/chips

Surface MMThree pixels in the same array


Bulk micromachining; pulse performance/energy resolution

Effective time constant: τeff = 300 to 400 μs• This is 3 - 4 times lower than expected (and what was designed for)• Thermal conductance of SiN/Si(110) is 3 - 4 times lower than of SiN/Si(100)

Measured energy resolution ∆E = 6 - 7 eV at 6 keV• Resolution not understood; from the measured noise 4 - 5 eV is expected• Note: the best single pixels have ∆E = 3.9 eV and τeff = 85 μs


In progress: Bi-absorber arrays (7 μm thick with Cu bottom/thermalisation layer)

Mushroom shaped absorbers: thermal evaporation and lift-off

Problems encountered for XEUS sized pixels (240 x 240 μm2) in 5 x 5 arrays:lift-off edges, particle-like anomalies, μ-cracks in Bi

5 x 5 arrayBi absorbersHat: 240 x 240 μm2

Stem: 100 x 100 μm2

Single pixelBi-absorberHat: 160 x 160 μm2

Stem: 100 x 100 μm2


Advanced detector design through Finite Element Modelling (2D, 3D)

Basic layout of a sensor pixel Measured and modelled thermal transport

• el-ph coupling in TES and absorber• Kapitza coupling between TES - SiN• conductance silicon nitride membranes• conductance Si support beams• thermal coupling of Si chip to heat bath

5 x 5 pixel array 32 x 32 pixel array

Material parameters and Finite Element Modeling


Thermal coupling of detector pixels to heat bath• Bare Si chip Tchip = 187 mK TSi beam = 200 mK

• Cu coating back-side Si-chip Tchip = 41 mK TSi beam = 140 mK

• Cu coating on Si beams and chip Tchip = 41 mK TSi beam = 47 mK

Finite Element Model of the XEUS array (1000 pixels of 10 pW each)

Improvement of coupling to heat bath and a small thermal gradient in Si beam: proper heat bath

Improvement of coupling chip to bath

Future XEUS 32 x 32 pixel array

SiN

Si-beamBeam with Cu

side view Si(110) beam

Si-beam with Cu

SiN

Uncoated Si


Array development - summary

• Array production based on bulk and surface micromachining• The pixel to pixel performance in BMM and SMM arrays is quite good (R(T), I(V))

• Thermal conductance of SiN on Si(110) is lower than expected -> τeff too high To be measured: pixels with redesigned thermal support • Improvement of ΔE from 6 - 7 eV to values smaller than 5 eV needed Working on reduced sensitivity of set-up for EMI Working on more fundamental issues of the energy resolution

• Development of large mushroom-shaped absorbers is critical and has high priority

• The XEUS detector chip and pixels can be adequately cooled• Detailed Finite Element Model is available for thermal design, all relevant low-temperature material parameters measured


Frequency Domain Multiplexing - Motivation for multiplexed read-out

Thermal aspects related to the read-out and biasing of one pixel• Power dissipation: 10 pW/pixel• Power dissipation; 100pW/shunt resistor• Power dissipation: 1 nW/SQUID current amplifier• Heat input through wiring (5 at minimum, 4 twisted-wire pairs preferred)

Available cooling power XEUS ADR: 5.5 μWh @ 35 mK • 32 x 32 pixel array without multiplexing: only ~4 hours of operation!


Frequency Domain Multiplexing

AC-biasing of TES microcalorimeter• TES acts as AM-modulator• LC noise blocking filters• One SQUID per column


Need for loop gain

Large dynamic range required (current pulse vs current noise)DR = 8.106 = Φ0/(2ΦSQUID) (1+LFLL); low-noise VTT SQUID: LFLL > 1.4

Common impedance (SQUID input coil) leads to cross talk (from f1 to f2) CT = 0.001 = [Lc/(1+LFLL )/L]2; LFLL > 10

The SQUID is an non-linear component -> mixing products (from f1 to f2) CT = 0.001 requires LFLL > 15


Limitations of the achievable loop gain

• Phase rotation due to cable delay imposes: tdelay.foperation < 0.11/LF;

For an optimized cryostat with 20 cm distance between SQUID andwarm electronics (t delay = 3 ns)

• Phase rotation of the amplifier (simulation performed for a 200 MHz amplifier with one pole zero compensation)

Standard FLLCombining 32 channels (with 100 kHz seperation) requires 3 MHz• LFLL ~ 10 @ 3 MHz It requires very close packing and the available low loop gain is low; it implies an appreciable fraction of mixing products

Baseband feedbackThe carrier is deterministic and carries no information (use it in the feedback)In principle, only the signal bandwidth (200 kHz) is relevant; it allows for high loop gains• LFLL ~ 200 @ 200 kHz


Standard FLL

Pro: Con:FLL proven concept - A-linearity in SQUID introduces crosstalk at 0.2% level

- Common impedance leads to crosstalk at 0.4% level- Bandwidth limited to ~3 MHz, 32 close packed channels


Pro: Con:• L > 200 results in: Complex demux/mux electronics A-linearity in SQUID: crosstalk at < 0.01% level Idle-current cancellation not required Available bandwidth is up to ~10 MHz; allows for well-spaced carriers -> cross-talk due to common impedance is no longer a problem

Baseband feedback


Experimental status Frequency Domain Multiplexing (biasing of detectors)

Biasing of microcalorimeterAC-bias measurements at 500 kHz to study potential switch-off behavior

Microcalorimeter can be biased over the whole transition provided that the there is a small impedance in the biasing circuit <-> low dielectric loss in C

Set-up for 250 kHz FLL operational and optimizedElectronic resolution of FLL electronics, bias sources and mixers/de-mixers, and SQUID for detector biased in normal state is at present 2 eV

Tests with a TES microcalorimeter with 5 eV resolution @ 6 keV (DC) AC-bias experiment at 50 kHz with 6.5 eV @ 6 keV

Baseline noise 5 eV (~sensor dominated resolution)AC-bias experiment at 500 kHz with 8.8 eV @ 6 keV

SQUID back-action noise (LSQUID) limits resolution of 8-9 eVAC-bias experiment at 250 kHz with 7 eV @ 6 keV

Baseline noise 5 - 6 eV (sensor dominated resolution)


Status Frequency Domain Multiplexing (noise blocking filters)

LC filters required with Q ~2500 (depends on carrier frequency)

Washer-type superconducting coilsMinimize dimensions (reduce intercoil cross-talk)

Superconducting capacitorsHigh dielectric constant (small components)Low dielectric losses (tan δ < 0.001; introduces resistance)

Si3N4 (VTT): Q = 2800 (compatibility of Si3N4 process)Al2O3 (SRON): Q = 300 (limited by critical current)

Test structures: 2.4 to 240 nF; size up to 4 x 4 mm2 Low leakage, R ~ 1 mΩ

L = 100 nH

500 m


Outlook area: energy resolution

ΔE = 2.35 ξ [kBT2C]1/2

Microcalorimeter physics Low heat capacity absorbers

What motivates the improvement of ΔE?• there is still very limited margin on ΔE with respect to the specification for XEUS and other applications• the development of large area pixels with a good ΔE requires that ξ and/or C are as low as possible (and under control)


Outlook – summary

Detector physics – improvement of energy resolution (design: 1 – 1.5 eV) Typical energy resolution (measured) 4 - 5 eV• Possible improvement pulse filtering factor 1.5 – 2 (Fixsen method)• Possible improvement by ITFN reduction factor 1.5 – 2 (TES with lower RN)Device testing (single pixels and arrays)

Large area pixelsLow C materials: Bi, Sn Compatibility with large arrays?Design optimization, thermalization issues: ΔE and ΔE_xy (position dependence)ProductionDevice testing (single pixels and arrays), E < 6 keV

Absorber development

Documents

Development of TES-microcalorimeter arrays and Frequency Domain Multiplexed read-out Henk Hoevers