Ultralow-Temperature SI-STM

• Ultralow-temperature SI-STM is a technique that integrates atomically resolved imaging with spectroscopic measurement at millikelvin temperatures to probe quantum phenomena.
• Its design uses dilution refrigeration, advanced vibration isolation, and RF filtering to achieve sub-100 μeV energy resolution and sub-Ångström spatial precision.
• The method enables detailed mapping of superconductivity, magnetism, and electron correlations, offering insights into emergent quantum states and electronic orders.

Ultralow-temperature spectroscopic-imaging scanning tunneling microscopy (ULT SI-STM) is a technique that combines atomically resolved scanning tunneling microscopy with energy-resolved spectroscopy at temperatures well below 1 K, enabling direct measurement of the local electronic density of states (LDOS), collective quantum phenomena, and emergent order in quantum materials. Through integration with dilution refrigeration, advanced vibration isolation, and high-fidelity spectroscopic protocols, ULT SI-STM achieves sub-100 μeV energy resolution and spatial precision at the sub-Ångström scale, facilitating the exploration of low-energy excitations, superconductivity, magnetism, and correlation effects unattainable at elevated temperatures.

1. Fundamental Principles and Theoretical Framework

ULT SI-STM operates by raster-scanning a metallic tip over a sample’s surface while measuring the tunneling current $I(V)$ as a function of voltage $V$ and position $(x, y)$ at cryogenic temperatures, often down to tens of millikelvin (Marz et al., 2010, Misra et al., 2013). The tip-sample bias establishes a tunneling barrier through which electrons quantum-mechanically traverse, with $I(V)$ governed by:

$$I(V) = \frac{4\pi e}{\hbar}|M|^2\rho_t(0)\int_{0}^{eV}\rho_s(\epsilon)\,\mathrm{d}\epsilon,$$

where $|M|^2$ is the tunneling matrix element, $\rho_t$ is the tip DOS, and $\rho_s$ is the sample’s LDOS. Differential conductance, determined by lock-in detection,

$\frac{dI}{dV}(V)\propto |M|^2\rho_t(0)\rho_s(eV),$

serves as a direct proxy for the LDOS at energy $eV$ (Hoffman, 2012, Crespo et al., 2012).

Thermal broadening limits the achievable energy resolution: $\Delta E \approx 3.5\,k_B T$, so operation at $T\lesssim100$ mK yields $\Delta E\lesssim30\,\mu\text{eV}$ (Machida et al., 2018, Roychowdhury et al., 2013). Electronic noise, lock-in modulation amplitude $V_\mathrm{ac}$, and RF filtering further influence the spectroscopic precision.

2. Instrumentation: Cryogenics, Vibration, and Electronic Isolation

2.1 Cryogenic Architecture

The core of ULT SI-STM is the integration of the STM head with ultralow-temperature platforms, primarily dilution refrigerators (base $T\lesssim10$–100 mK) or, for slightly higher temperatures, $^3$He sorption cryostats ($T\sim300$–400 mK) (Marz et al., 2010, Misra et al., 2013, Guan et al., 2018, Salazar et al., 2018). Thermal anchoring of the sample stage and tip assembly to the mixing chamber or $^3$He-pot is accomplished with high-purity OFHC copper conductors, Kapton isolation, and matched-coefficient-of-expansion ceramics such as sapphire or AlN for mechanical rigidity and effective thermalization (Battisti et al., 2018).

Typical cooling strategies implement staged thermal links with silver-, copper-, or gold-plated mechanical connections, yielding temperature gradients $\Delta T$ of less than a few mK across the scanner (Marz et al., 2010, Misra et al., 2013). Base temperatures down to $T_\text{MC}=20$ mK (lattice), electron temperatures $T_\mathrm{eff}=90$–250 mK, and cooling powers up to $100\,\mu$W@100 mK are documented (Misra et al., 2013, Machida et al., 2018, Roychowdhury et al., 2013).

2.2 Vibration Isolation and Mechanical Design

Attaining sub-pm $z$-stability and minimizing vibrational noise necessitate multistage passive and active isolation: heavy concrete plinths ($\sim30$ tons), granite floating slabs, negative-stiffness isolators (e.g., $>$0.5 Hz resonance), and acoustic shielding (Misra et al., 2013, Battisti et al., 2018). STM heads are constructed for maximal stiffness—using high-$E$ materials (sapphire, titanium, gold-plated Macor), compact Pan-style “walker” coarse-approach mechanisms, and explicit finite-element analysis-driven geometry optimization to push body/walker eigenmodes $>10$ kHz (Battisti et al., 2018).

Wiring is strictly thermalized and filtered at each temperature stage, with typical mechanical $z$-noise floor $\sim$0.5 pm/$\sqrt{\text{Hz}}$ and current noise $<10$ fA/$\sqrt{\text{Hz}}$ (Battisti et al., 2018, Roychowdhury et al., 2013).

2.3 Radio-frequency Filtering and Grounding

State-of-the-art RF filtering—$\pi$-filters with corner frequencies $<$10 kHz, copper/bronze-powder lossy filters, and star-point grounding—are mandatory to suppress Johnson and external RF noise, ensure $T_\mathrm{eff}\approx T_\text{MC}$, and prevent instrument heating. Failure to properly filter any scanner, bias, or thermometry line routinely results in $T_\mathrm{eff}$ elevation up to $>200$ mK (Machida et al., 2018, Roychowdhury et al., 2013, Guan et al., 2018).

3. Measurement Protocols and Spectroscopic-Imaging Modes

3.1 dI/dV Mapping

Energy-resolved SI-STM is typically conducted by raster-scanning the tip in a $N\times N$ grid (e.g., $128\times 128$ pixels over $100$–$500$ nm$^2$), recording point spectra $I(V)$ or $dI/dV(V)$ at each position. Modulation voltages $V_\mathrm{ac}$ in the $10$–$100\,\mu$V$_\text{rms}$ range (frequencies $473$–$2000$ Hz) are applied for lock-in detection; dwell times per spectrum are $10$–$100$ ms, balancing energy resolution and mapping speed (Hoffman, 2012, Guan et al., 2018, Narasimha et al., 2024).

Full $dI/dV(x, y; V)$ matrices directly yield spatial maps at any chosen energy slice and enable Fourier analysis (e.g., for quasiparticle interference, vortex imaging).

3.2 Advanced Modes: Spin, Microwave, and Virtual Tunneling

Spin-resolved spectroscopy is enabled by in-situ tip preparation (field emission, functionalization with Fe/Cr) in double-deck sample stages, with detection of $(dI/dV)_A-(dI/dV)_B$ asymmetry across magnetic domains (Guan et al., 2018, Salazar et al., 2018). Calibrated GHz microwave reflectometry is employed to directly access tip-sample junction admittance $Y_\text{jun}=G+i\omega C$, providing local dielectric contrast at $\lesssim0.1$ fF, with spatial resolution $<5$ nm (Wit et al., 2023).

Virtual STM (VSTM) probes buried 2D electron systems via a probe–subject bilayer heterostructure, the upper “probe” 2DES controlled via a scanned, charged metal tip. The tunneling Hamiltonian

$$\hat{H}_T = \int d^2r~[\, t(r)\, \hat{c}_1^\dagger(r) \hat{c}_2(r) + h.c.\, ]$$

with spatially modulated transfer matrix element $t(r)$ enables local spectroscopy of 2DESs with spatial resolution down to $40$ nm and energy resolution $<100\mu$eV at $T=300$ mK (Sciambi et al., 2010).

4. Performance Metrics: Energy and Spatial Resolution, Noise, and Field Integration

4.1 Energy Resolution

ULT SI-STM achieves energy broadening set by convolution of effective electronic temperature and lock-in modulation, yielding $$\Delta E_\text{tot}=\sqrt{(3.5\,k_B T_\mathrm{eff})^2+(eV_\mathrm{mod})^2}$$; in leading dilution refrigerator systems, $\Delta E=16$–$100\mu$eV (Misra et al., 2013, Roychowdhury et al., 2013, Machida et al., 2018). Measurements with superconducting Al or Nb tips over normal metals corroborate effective $T_\mathrm{eff}=90$–400 mK depending on platform and filtering, see Table below:

System	$T_\mathrm{eff}$ (mK)	$\Delta E$ ($\mu$eV)	Field (T)
(Machida et al., 2018)	87–90	26	17.5
(Roychowdhury et al., 2013)	184	16	13.5
(Misra et al., 2013)	250	75	14
(Guan et al., 2018)	400	120	7

4.2 Noise and Stability

Open-loop tunneling current noise is typically $<1$ pA/$\sqrt{\text{Hz}}$ ($<10$ fA/$\sqrt{\text{Hz}}$ in best-in-class setups), with vertical $z$-noise of $<0.5$ pm /$\sqrt{\text{Hz}}$. Spatial drift rates are $<50$ pm/hour, allowing $\gtrsim$18 h continuous mapping (Machida et al., 2018, Battisti et al., 2018, Roychowdhury et al., 2013).

4.3 High Magnetic Field Compatibility

ULT SI-STM is fully compatible with vector superconducting magnets up to $17.5$ T, with field homogeneity $10^{-4}$ over $10$ mm, enabling field-dependent vortex imaging, quantum oscillations, and Zeeman-resolved studies (Machida et al., 2018, Misra et al., 2013, Roychowdhury et al., 2013).

5. Applications: Example Measurements and Methodological Extensions

ULT SI-STM enables mapping of SC gap structure, vortex-lattice formation, Andreev bound states, broken symmetry, and QPI with unprecedented precision (Crespo et al., 2012, Hoffman, 2012, Marz et al., 2010, Kamlapure et al., 2013). Notable measurement protocols include:

• BCS/Dynes fitting of dI/dV spectra at $T\approx30$–350 mK for extraction of $\Delta$, $\Gamma$, and $T_\mathrm{eff}$—e.g., Al$\to$Au yielding $T_\mathrm{eff}=87$ mK (Machida et al., 2018).
• Andreev reflection mapping using superconducting tips for SAS imaging of vortex cores (Crespo et al., 2012).
• Sparse, ML-driven hyperspectral mapping: integration of Bayesian deep learning (deep kernel learning with Gaussian processes) for property-guided sparse sampling allows $\sim100\times$ reduction in acquisition time, with multiscale expansion from mesoscopic to atomic features (Narasimha et al., 2024).
• Spin-resolved SI-STM: in-situ tip functionalization and double-deck sample stages allow atomic-scale mapping of spin textures via domain-contrasted conductance (Salazar et al., 2018, Guan et al., 2018).

6. Specialized Geometries: Virtual STM and Microwave-Enhanced Modes

VSTM provides spectroscopic imaging access to buried 2DESs via a bilayer quantum-well heterostructure. A scanned, charged tip gates the probe 2DES, modulating subband-edge leakage and enabling localized interlayer tunneling. The lock-in-detected dI/dV at each point is essentially a spatial-resolved probe of the LDOS of the subject 2DES at energy $E=eV_\text{tip}$; $40$ nm spatial and $<100\mu$eV energy resolution are achieved, with perturbations to carrier density screened by the probe layer (Sciambi et al., 2010).

Microwave-reflectance STM superimposes GHz signals on the STM bias, with Mach-Zehnder interferometric background suppression and in-situ, tip-retraction–based error calibration (3-error model). Local dielectric permittivity changes are resolved with $<0.1$ fF sensitivity and lateral resolution $<5$ nm, expanding the SI-STM platform into charge dynamics and single-molecule dielectric fingerprinting (Wit et al., 2023).

7. Design Guidelines and Best Practices

Rigorous best practices are distilled across advanced ULT SI-STM implementations (Marz et al., 2010, Misra et al., 2013, Battisti et al., 2018):

• Maximize STM head stiffness and mechanical eigenfrequencies ($>10$ kHz for walker/body).
• Employ high-thermal-conductivity, well-matched materials (sapphire, AlN, gold-plated Cu).
• Integrate multistage RF filtering, all-wiring thermalization, and star-point grounding.
• Implement massive passive isolation and, as needed, negative-stiffness or active damping.
• Validate energy resolution in situ via BCS gap fitting on well-characterized superconductors.
• Monitor drift, noise, and resonant modes in situ; actively tune piezo preload and geometry.
• For advanced modes, incorporate modular tip/sample stages (for field emission, spin polarization, multi-tip, etc.).
• Maintain flexible UHV interface for in-situ tip/sample/film exchange and surface science integration.

This methodological and architectural rigor ensures the reproducibility and quantum-limit performance of ULT SI-STM as a tool for the atomic-scale investigation of correlated electron matter, superconductivity, topological order, and emergent quantum phenomena.