---
title: Magnetic Microbolometer (MMB) Overview
url: https://www.emergentmind.com/topics/magnetic-microbolometer-mmb
type: topic
---

# Magnetic Microbolometer (MMB) Overview

A magnetic microbolometer (MMB) is a cryogenic bolometer concept derived from metallic magnetic calorimeter (MMC) technology in which absorbed power is transduced thermomagnetically rather than through a resistive thermometer or a superconducting transition. In the formulation proposed for cosmic microwave background (CMB) instrumentation, an MMB uses a paramagnetic temperature sensor such as Au:Er or, in closely related magnetic thermal-detector literature, Ag:Er; the sensor is biased by a magnetic field, and its temperature-dependent magnetization is converted into a flux signal that is read out by a SQUID. The concept has been advanced as an alternative cryogenic detector technology for CMB polarization measurements because it combines a broad and smooth responsivity dependence with temperature, the absence of Joule dissipation in the sensing principle, and compatibility with scalable SQUID-based multiplexing [2209.06088]. A first dedicated microwave SQUID multiplexer developed specifically for MMBs was reported in 2025, together with a proof-of-principle magnetization measurement on an MMB prototype, establishing a practical readout path but not yet a full science-grade bolometric array [2509.23507].

## 1. Definition, lineage, and detector class

An MMB is a **microbolometer** rather than a **microcalorimeter**: it is intended for absorbed power or radiation power measurement, whereas an MMC is usually used for event-by-event calorimetry. The two detector classes nevertheless share the same basic magnetic thermal-sensing chain. In both cases, absorbed energy or power changes the detector temperature; a magnetic thermometer converts that temperature change into a change in magnetization; a superconducting pickup structure converts that magnetization change into magnetic flux; and a SQUID-based readout measures the flux [2209.06088].

The MMB was explicitly proposed for CMB polarization surveys as an adaptation of state-of-the-art MMC technology. The rationale given for that adaptation is that MMC-style magnetic thermometry offers a non-dissipative sensing concept, straightforward calibration, and higher dynamic range because the responsivity varies broadly and smoothly with temperature rather than being tied to a narrow superconducting transition [2209.06088]. In the 2025 microwave-readout work, the MMB is described as “a recent novel cryogenic bolometer type” based on the “well-established magnetic microcalorimeter (MMC) technology,” with the central operating principle stated as exploitation of “the magnetic properties of paramagnetic materials at sub-kelvin temperatures” [2509.23507].

This lineage is significant because it places MMBs within the broader family of SQUID-read cryogenic thermal detectors while separating them from transition-edge sensors (TESs) and microwave kinetic inductance detectors (MKIDs). Relative to TES bolometers, the MMB appeal is explicitly framed in terms of high dynamic range, low intrinsic noise, and the absence of Joule dissipation in the sensing chain [2209.06088; 2509.23507]. Relative to canonical MMCs, the main difference is the target observable: continuous or quasi-continuous optical loading rather than discrete quanta.

## 2. Thermomagnetic sensing principle and device physics

The defining physics of the MMB is thermomagnetic transduction. A paramagnetic sensor material is placed in a magnetic bias field, and at sub-kelvin temperature its magnetization depends strongly on temperature. In the dedicated MMB prototype reported in 2025, the sensor material is explicitly **Au:Er**. Absorbed power changes the sensor temperature; the Au:Er magnetization changes; and the resulting change in magnetic flux is coupled into the SQUID input circuit [2509.23507].

For the CMB-oriented formulation, the sensor is modeled as a dilute alloy such as **Au:Er** or **Ag:Er**, with effective spin $\tilde{S}=1/2$ and $g=6.8$ below about $1\,\mathrm{K}$. The magnetic thermodynamic quantities are written as [2209.06088]
$$
C_{\text{S}} = \frac{N_{\text{s}}}{V k_{\text{B}}T^{2}}\Big\{\langle E^{2}\rangle - \langle E\rangle^{2}\Big\},
$$
$$
M = - \frac{N_{\text{s}}}{V}\Big\langle \frac{\partial E}{\partial B}\Big\rangle,
$$
and
$$
\frac{\partial M}{\partial T} = \frac{N_{\text{s}}}{V k_{\text{B}}T^{2}}\Big\{\Big\langle E \frac{\partial E}{\partial B}\Big\rangle - \langle E\rangle \Big\langle \frac{\partial E}{\partial B}\Big\rangle\Big\}.
$$
The free-electron contribution to the heat capacity is included as
$$
C_{\text{e}} = \gamma \cdot T,
$$
with, for gold,
$$
\gamma = 6.9\times10^{-4}\ \text{J}\text{mol}^{-1}\text{K}^{-2}.
$$

The paper proposing MMBs for CMB measurements factorizes the responsivity as [2209.06088]
$$
\mathfrak{R}_{\text{MMB}}=\frac{\partial\Phi_{SQ}}{\partial P}=\frac{\partial \Phi_{SQ}}{\partial \Phi}\cdot\frac{\partial \Phi}{\partial T}\cdot\frac{\partial T}{\partial P}.
$$
The temperature-to-flux conversion of the pickup structure is written as
$$
\frac{\partial \Phi}{\partial T} = \bigintss_V
\frac{|\vec{B}(\vec{r})|}{I_{\text{field}}}
\frac{\partial M}{\partial T}\Bigr|_{|\vec{B}(\vec{r})|} d^3r,
$$
which makes explicit that transduction depends on the field distribution, the sensor volume, and the local thermomagnetic susceptibility.

In the 2025 prototype, the detector geometry is described as “two identical gradiometric meander-shaped coils: one used as a load inductor and the other as a pickup coil. The latter is coated with the paramagnetic sensor alloy Au:Er” [2509.23507]. The experimentally emphasized observable is the flux–temperature curve $\Phi(T)$, which functions as the direct detector transduction variable in that work.

Closely related MMC literature supplies important engineering context. In high-resolution X-ray MMCs with integrated dc-SQUID readout, **Ag:Er** sensors, meander-shaped Nb structures, persistent magnetic biasing, and flux readout are combined in a closely analogous magnetic thermal-detector chain [2310.08698]. This suggests that many of the limiting mechanisms in MMCs—absorber thermalization, athermal phonon escape, thermal contamination from dissipative SQUID elements, and magnetic coupling geometry—are directly relevant to MMB design, even though the intended measurement mode differs.

## 3. Thermal model, responsivity, and noise

The CMB-oriented MMB formulation uses a two-subsystem thermal model consisting of a conduction-electron system with heat capacity $C_e$ and temperature $T_e$, and a spin system with heat capacity $C_s$ and temperature $T_s$, coupled to a bath at temperature $T_b$ through $G_{\text{bath}}$ and to one another through $G_{\text{se}}$ [2209.06088]. Standard bolometer relations are used as the starting point,
$$
\Delta T = \frac{P}{G_{\text{bath}}},
\qquad
\tau = \frac{C_{\text{abs}}}{G_{\text{bath}}},
$$
but the full small-signal thermalization function is two-body:
$$
\frac{\partial T}{\partial P} = \frac{1}{G_{\text{bath}}} \cdot \left[ 1-\frac{C_{\text{e}} C_{\text{s}}}{G_{\text{bath}}G_{\text{se}}}\cdot \omega^2 + \frac{C_{\text{s}}(G_{\text{bath}} + G_{\text{se}}) + C_{\text{e}} G_{\text{se}}}{G_{\text{bath}}G_{\text{se}}}\cdot i\omega\right]^{-1}.
$$
The papers emphasize that thermalization is the slowest process and therefore sets the detector bandwidth [2209.06088].

For scanning CMB instruments, the time constant is constrained by beam distortion:
$$
\tau \leq \frac{\Theta_{\text{beam}}}{2\pi \Dot{\Theta}}.
$$
In the representative $150\,\mathrm{GHz}$ case study, the assumed parameters are a fractional bandwidth of $25\%$, aperture $0.3\,\mathrm{m}$, beam FWHM $\Theta_{\text{beam}} = 0.39^\circ$, scan speed $\dot{\Theta} = 1\,\text{deg/sec}$, and average optical power $P_{\text{opt}} = 6\,\text{pW}$. From these assumptions the paper derives
$$
\tau \leq 62\,\text{ms},
\qquad
f_{\text{BW}} \geq 2.57\,\text{Hz},
\qquad
\text{NEP}_{\gamma} = 55.79 \,\text{aW}/\sqrt{\text{Hz}}.
$$
The stated design goal is then $\text{NEP}_{\text{det}} < \text{NEP}_{\gamma}$ while preserving the time-constant requirement [2209.06088].

The detector-noise model includes thermal fluctuation noise, magnetic Johnson noise of the metallic paramagnetic sensor, erbium excess $1/f$ noise, and SQUID flux noise:
$$
\text{NEP}^2_{\text{det}} =
\frac{S_{\text{th}}}{\big|\frac{\partial T}{\partial P}\big|^2} + \frac{S_{J,\Phi}}{\big|\frac{\partial T}{\partial P}\big|^2\big|\frac{\partial \Phi}{\partial T}\big|^2} + \frac{S_{\Phi , Er}}{\big|\frac{\partial T}{\partial P}\big|^2\big|\frac{\partial \Phi}{\partial T}\big|^2} + \frac{S_{\Phi_s}}{\big|\frac{\partial T}{\partial P}\big|^2\big|\frac{\partial \Phi}{\partial T}\big|^2\big|\frac{\partial \Phi_s}{\partial \Phi}\big|^2}.
$$
The bath phonon thermal-fluctuation term is
$$
S_{\text{bath}} = 4 \gamma k_B T^2 G_{\text{bath}},
$$
with
$$
\gamma = \frac{\beta+1}{2\beta+1}\cdot\frac{1-(T_{\text{bath}}/T)^{2\beta+1}}{1-(T_{\text{bath}}/T)^{\beta+1}},
$$
and $\beta = 3$ for phonon transport, while the electron-spin exchange term is
$$
S_{\text{se}} = 4 k_B T^2 G_{\text{se}}.
$$
The magnetic Johnson term is written as
$$
S_{\text{J},\Phi} = \mathfrak{K} \cdot \sigma \cdot k_\mathrm{B}T,
$$
and the erbium excess noise as
$$
S_{\Phi , Er} = N_{Er} \frac{|B(\bold{r})|^2}{I_f^2} S_m(f),
\qquad
S_m(f) = 0.12\mu_B^2 \,/\, f^{\eta},
\qquad
\eta \approx 0.8 - 1.
$$
A recurrent point in the literature is that the phrase “lack of Johnson noise associated with the detector readout” does **not** mean the total absence of Johnson-related noise: the CMB study explicitly retains **magnetic Johnson noise** intrinsic to the metallic sensor [2209.06088].

The proof-of-concept optimization trends are specific. Increasing sensor thickness raises responsivity initially and then reaches a plateau around **600–1500 nm**, while decreasing bandwidth because of added heat capacity. Increasing leg length lowers $G_{\text{bath}}$ and initially improves responsivity, but for leg lengths above about **800 \,\mu\text{m}** the detector NEP is not further reduced. Increasing erbium concentration improves signal response approximately linearly but also increases spin heat capacity and erbium $1/f$ excess noise, with low-frequency NEP curves beginning to converge above about **1000 ppm**. Increasing the magnetic bias current $I_{\text{field}}$ is identified as especially favorable because responsivity increases roughly linearly without a significant bandwidth penalty [2209.06088].

## 4. Readout architectures and multiplexed implementations

The natural readout of an MMB is SQUID-based magnetic-flux readout. In the CMB proof-of-concept, the pickup-coil to SQUID transfer is approximated by [2209.06088]
$$
\frac{\partial\Phi_{\text{SQ}}}{\partial\Phi} \approx \frac{k\sqrt{L_{\text{in}}L_{\text{s}}}}{L_{\text{m}}+L_{\text{stray}}+L_{\text{in}}},
$$
where $L_m$ is the pickup coil inductance, $L_{\text{in}}$ the SQUID input coil inductance, $L_s$ the SQUID loop inductance, $L_{\text{stray}}$ parasitic inductance, and $k$ the coupling factor. The representative case study uses a magnetic bias current of **100 mA**, SQUID self-inductance **50 pH**, SQUID input inductance **2 nH**, and coupling factor **0.57**, with the sensor area chosen to match pickup-coil inductance to SQUID input inductance [2209.06088].

The most consequential systems development to date is the adaptation of the **microwave SQUID multiplexer** ($\mu$MUX) to MMBs. In generic form, the $\mu$MUX is a GHz-frequency frequency-division multiplexing architecture consisting of “a single transmission feedline capacitively coupled to $N$ microwave resonators, each with a well-defined and unique characteristic frequency,” with each resonator inductively coupled to a non-hysteretic rf-SQUID and all rf-SQUIDs coupled to a common modulation line for flux-ramp modulation [2509.23507]. The readout chain described for the dedicated MMB implementation runs from detector pickup circuit to room-temperature digital demodulation, with a **ZCU216-based software-defined radio (SDR)** used for tone generation, transmission, and demodulation. The cryogenic readout chip sat on the mixing chamber at **$T_\mathrm{MXC}=100\,\mathrm{mK}$**, the input line had **40 dB attenuation**, and the output line used a **Low Noise Factory LNF-LNC4\_8C HEMT** at **4 K** together with a **Mini-Circuits ZX60-83LN12** room-temperature LNA, giving approximately **$62\,\mathrm{dB}$** total gain [2509.23507].

The MMB-specific architectural modification is explicit: “Unlike the TES-SQUID scheme, here the rf-SQUID input inductance $L_\mathrm{IN}$ and the MMB coil $L_\mathrm{DET}$ form a closed superconducting loop” [2509.23507]. Because this loop partially screens the resonator-SQUID magnetic coupling, the effective SQUID–resonator coupling is reduced relative to the nominal design value by a screening term involving mutual inductances and the total closed-loop inductance $L_\mathrm{IN}+L_\mathrm{DET}$. The paper states that this effect leads to a predicted **$\sim 12\%$** deviation between nominal and effective parameters [2509.23507]. This is one of the main technical distinctions between MMB-$\mu$MUX and TES-$\mu$MUX design.

The dedicated 16-channel MMB multiplexer was fabricated at **KIT IMS** using a niobium trilayer and multilayer thin-film process, beginning with **Nb/Al-AlOx/Nb \((100\,\mathrm{nm}/7\,\mathrm{nm}/100\,\mathrm{nm})\)** Josephson junctions, **ICP-RIE** patterning of the feedline and quarter-wave resonators, two sputtered **$\mathrm{SiO_2}$** insulation layers, **AuPd** RL low-pass filters in the detector input circuit, and a final **$450\,\mathrm{nm}$** sputtered Nb layer [2509.23507]. The paper focuses on a **16-channel** chip within a family of **8-, 16-, and 32-channel** devices spanning the **$4{-}8\,\mathrm{GHz}$** electronics band. The 16-channel design parameters were: resonator bandwidth **$BW_\mathrm{res} = 200\,\mathrm{kHz}$**, channel frequencies **$f_0 = 5.744{-}6.256\,\mathrm{GHz}$**, channel spacing **$\Delta f_0 = 32\,\mathrm{MHz}$**, resonance shift **$\Delta f_r = 200\,\mathrm{kHz}$**, SQUID-input coupling **$M_\mathrm{SI} = 316\,\mathrm{pH}$**, SQUID-resonator coupling **$M_\mathrm{ST} = 2.58{-}2.34\,\mathrm{pH}$**, and SQUID-modulation coupling **$M_\mathrm{SM} = 10\,\mathrm{pH}$** [2509.23507].

Each rf-SQUID used a four-lobe gradiometric loop with a **$3.5 \times 3.5\,\mu\mathrm{m}^2$** Josephson junction. The self-screening parameter was designed as
$$
\beta_\mathrm{L} = \frac{2\pi I_\mathrm{C} L_\mathrm{S}}{\Phi_0} = 0.5,
$$
ensuring non-hysteretic behavior, and the resonator shift parameter
$$
\eta = \frac{\Delta f_\mathrm{r}}{BW_\mathrm{res}}
$$
was tuned to the target **$\eta = 1$** at an equivalent RF drive power of approximately
$$
\Phi_\mathrm{RF} \approx 0.3\,\Phi_0.
$$
The narrow **$200\,\mathrm{kHz}$** bandwidth was chosen to preserve a path to “**1000 channels within the $4{-}8\,\mathrm{GHz}$ band**” [2509.23507].

Adjacent μMUX literature is relevant because it demonstrates the maturity of the warm and cryogenic infrastructure even where MMBs themselves are not the detector under test. A full-scale SDR-based μMUX system for MMCs was reported in 2025 as capable of handling **up to 400 channels**, operating reliably across the entire **4–8 GHz** band, and achieving mean flux-ramp-demodulated white flux noise of **$(1.4 \pm 0.2)\,\mu\Phi_0/\sqrt{\mathrm{Hz}}$** [2509.07671]. This suggests that the principal remaining MMB-specific challenges lie at the detector–SQUID interface and in detector-referred performance, not in the existence of a general-purpose wideband μMUX/SDR platform.

## 5. Experimental demonstrations and current performance status

The dedicated MMB-$\mu$MUX demonstrator produced the expected microwave transmission structure. The measured $|S_{21}|$ showed **15 working resonances out of 16**, with “**one missing channel due to defects produced during the fabrication process**,” and the measured characteristic frequencies were shifted by approximately **1.25%** relative to design while still remaining inside the available electronics band [2509.23507]. The maximum resonator frequency shift as a function of RF tone power was found to be “**consistent with the expected RF power dependence**,” which the authors interpret as confirmation that the $\eta = 1$ design strategy was appropriate [2509.23507].

The central detector demonstration was a magnetization measurement rather than a full optical bolometric measurement. **Two MMB prototype devices** were connected to two channels of the characterized $\mu$MUX. A probe RF tone was applied at a frequency $f_\mathrm{exc}$ above the maximum resonance-shifted value, and the tone amplitudes were recorded while sweeping the cryostat temperature from **$500\,\mathrm{mK}$** down to **$20\,\mathrm{mK}$** for increasing values of the magnetic bias current $I_F$ [2509.23507]. After correction for the temperature-dependent resonator frequency shift attributed to **Gao et al.**, the response showed the expected periodic SQUID modulation versus temperature. The paper gives a concrete interpretation: successive maxima and minima correspond to changes of **$0.5\,\Phi_0$** in the flux coupled into the input coil by the paramagnetic sensor, and the extracted **$\Phi(T)$** curves were “**consistent with the expected simulated values in [Geria 2023]**” [2509.23507].

What has therefore been demonstrated experimentally is precise but limited in scope. The readout can directly probe the thermomagnetic behavior of the Au:Er sensor, remains stable at **100 mK**, and behaves compatibly with the expected operating power and modulation targets. However, the same paper explicitly does **not** report final detector NEP, crosstalk, integrated flux noise, demodulated white noise level, linearized dynamic range under flux ramp, detector bandwidth in final form, or a many-pixel demonstration with active MMBs. Its conclusion states: “**Future work will focus on detailed noise characterization and the realization of a fully integrated readout chain with functional MMB arrays**” [2509.23507].

The earlier CMB proof-of-concept remains simulation-based rather than experimental. Under its representative assumptions, the antenna-coupled design with suspended-leg length **$700\,\mu\text{m}$** and $G_{\text{bath}} = 162.8\,\text{pW/K}$ yields a computed time constant **$\tau = 10.6\,\text{ms}$**, while the absorber-coupled design with a **Pd mesh** absorber, four legs each **$200\,\mu\text{m}$** long, leg cross section **$30\,\mu\text{m}^2$**, $G_{\text{bath}} = 277.3\,\text{pW/K}$, and added absorber heat capacity **$3\,\text{pJ/K}$** gives **$\tau = 16.7\,\text{ms}$**; both satisfy the beam-smearing limit of **$62\,\text{ms}$** and achieve detector NEP below the background NEP over a relevant bandwidth [2209.06088].

Related MMC work provides a benchmark for what magnetic thermal detectors can achieve when the detector architecture is more mature. An integrated dc-SQUID MMC with **Ag:Er** paramagnetic sensors, a **tetrapod absorber geometry**, and membrane isolation of the shunt section reached an energy resolution of **$\Delta E_\mathrm{FWHM} =1.25(18)eV$** with **two overhanging Au absorbers** of **$150\mu$m x $150\mu$m** and thickness **$3\mu$m** [2310.08698]. This does not constitute an MMB result, but it indicates that the basic magnetic-thermometer/SQUID paradigm can sustain state-of-the-art performance when thermal engineering and readout integration are solved.

## 6. Relation to other detector technologies, misconceptions, and open problems

The most direct comparison class for MMBs is the TES bolometer. TESs and MKIDs are described as the predominant cryogenic detector technologies in CMB instruments searching for primordial B-modes, whereas the MMB is proposed as an additional option [2209.06088]. The MMB differs from a TES in its transduction mechanism: it is not based on the steep temperature-dependent resistance of the normal-to-superconducting transition, and its appeal is therefore framed in terms of broad and smooth responsivity, high dynamic range, and no Joule dissipation in the sensing principle [2209.06088; 2509.23507]. A common misconception is that the MMB is merely a magnetically shielded or magnetically biased TES; the literature does not support that interpretation. It is instead a magnetic thermal detector derived from MMC technology.

A second misconception concerns noise. The CMB study states that MMBs lack “Johnson noise associated with the detector readout,” but it also explicitly includes **magnetic Johnson noise** generated by the metallic paramagnetic sensor [2209.06088]. The correct reading is therefore narrower: the absence refers to a TES-like resistive readout path, not to the elimination of all Johnson-related fluctuations.

A third issue is the distinction between MMBs and MMCs. The readout and thermometer physics can be nearly identical, and the 2025 MMB readout paper explicitly presents the MMB as derived from MMC technology [2509.23507]. Yet the detector objective remains different: bolometric power sensing for CMB applications versus discrete-event calorimetry in spectroscopy. This difference matters for dynamic range, thermal-link design, optical loading, and detector-referred figures of merit such as NEP.

Open problems remain concentrated in four areas. The first is **low-frequency erbium excess $1/f$ noise**, identified in the CMB study as the dominant low-frequency detector-side limitation and not yet microscopically understood [2209.06088]. The second is **full detector-readout integration**, since the dedicated μMUX work demonstrates magnetization readout and $\Phi(T)$ recovery but not final bolometric sensitivity or full-array operation [2509.23507]. The third is **fabrication yield and scaling**, already reflected by the missing resonance channel in the 16-channel demonstrator and by the stated goal of eventual $\sim 1000$-channel packing within **4–8 GHz** [2509.23507]. The fourth is **packaging and magnetic-field control**. Adjacent TES work has shown that a local superconducting groundplane can suppress unwanted external and self-generated magnetic fields, with a shielding factor of at least **$\sim 75$**, but an MMB cannot simply inherit such a strategy because its intended signal is itself magnetic [2208.10775]. This suggests that magnetic shielding in MMB systems must be selective and geometry-aware.

Adjacent detector concepts also clarify the technological neighborhood. The **Mutual Inductance Sensing SQUID (MISS)** is a superconducting microcalorimeter using a temperature-dependent mutual inductance rather than paramagnetic magnetization, and is best treated as adjacent rather than canonical MMB/MMC literature [2602.07917]. Its relevance is architectural: it confirms continued activity in SQUID-coupled cryogenic thermal detectors that seek alternatives to TES and conventional MMC sensor physics.

The present state of the field is therefore that MMBs are no longer only a theoretical extrapolation from MMCs. They now have a defined CMB-oriented detector model, explicit responsivity and noise formalism, and a first MMB-specific microwave multiplexed readout with cryogenic validation and measured $\Phi(T)$ curves on real Au:Er prototypes [2209.06088; 2509.23507]. At the same time, the literature remains careful: the enabling readout path has been demonstrated more clearly than the final bolometric system performance.

Source: https://www.emergentmind.com/topics/magnetic-microbolometer-mmb