Thulium-Doped Argon Spectroscopy

• Thulium-doped argon is a cryogenic matrix system where Tm atoms retain a near-1140 nm inner-shell transition with minimal matrix shift.
• The system leverages matrix isolation to preserve free-atom spectral features, while crystal-field effects and hyperfine splitting yield diverse, narrow spectral components.
• Controlled annealing and site engineering in Tm:Ar enhance its potential for surface sensing, nanoscale magnetometry, and quantum photonics.

Thulium-doped argon denotes a cryogenic rare-gas matrix system in which neutral thulium atoms are incorporated into solid argon and retain an unusually narrow inner-shell optical transition near $1140$ nm associated with the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$ fine-structure manifold of the $4f^{13}6s^2$ ground configuration. Across successive experiments, the system has emerged as a practical platform for matrix-isolated atomic spectroscopy, because solid argon can be grown on arbitrary substrates or directly on an optical-fiber tip, while the thulium $f\!\!-\!f$ transition remains only weakly shifted from the free-atom wavelength and can be resolved into narrow site- and sublevel-dependent features rather than being catastrophically broadened by the host. The resulting combination of narrow optical structure, trapping-site sensitivity, and direct Zeeman resolvability has made Tm:Ar relevant to surface sensing, nanoscale magnetometry, and the broader search for dense solid-state ensembles of nearly identical optical emitters (Gaire et al., 2019).

1. Host system, scope, and physical significance

Thulium-doped argon is a matrix-isolation system rather than a conventional substitutionally doped crystal in the oxide or fluoride sense. In the experiments reported to date, thulium atoms are introduced during cryogenic growth of solid argon and occupy metastable local trapping environments inside the rare-gas solid. The host is chemically inert and only weakly perturbative, yet it is sufficiently structured to lift degeneracies, produce crystal-field splittings, alter selection rules, and expose trapping-site heterogeneity through high-resolution optical spectroscopy (Marfey et al., 10 Jul 2025).

The distinctive physical reason the system is spectroscopically favorable is that the optically active $4f$ electrons are shielded by filled outer $5s$ and $5p$ shells. The supplied studies identify this shielding as the main reason that the $1140$ nm transition is much less perturbed by the matrix than ordinary valence-electron transitions, allowing narrow spectra in a solid environment. This suggests that Tm:Ar occupies an intermediate regime between isolated-atom spectroscopy and solid-state defect spectroscopy: the transition remains atom-like enough to preserve recognizable fine-structure and hyperfine signatures, but the matrix is strong enough to generate site structure, anisotropy, and field sensitivity (Gaire et al., 2023).

A recurrent misconception is to conflate Tm:Ar as a condensed-matter host system with thulium spectroscopy in an argon discharge. The hollow-cathode work on neutral thulium in a Tm–Ar plasma is explicitly not a study of thulium incorporated into solid argon; rather, argon there serves as an inert discharge gas and calibration source in visible Fourier-transform spectroscopy. That plasma literature is relevant for line identification and hyperfine constants of Tm I, but not for the host-matrix phenomenology of thulium-doped solid argon (Parlatan et al., 2022).

2. Sample preparation and experimental architectures

The initial matrix-isolation experiments employed a closed-cycle helium cryostat with a sapphire window of $19$ mm diameter as the growth substrate, mounted in a nickel-plated copper holder thermally anchored to the cryostat second stage with an indium gasket. Argon entered through a $1/16"$ OD stainless steel tube with flow controlled in the range ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$0–${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$1 standard cm${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$2/min. The growth strategy was to first deposit an argon film and then begin thulium incorporation by laser ablation of a metallic Tm target while deposition continued. Thulium ablation used a frequency-doubled Q-switched Nd:YAG laser at ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$3 nm, with ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$4–${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$5 ns pulse width, ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$6 Hz repetition rate, and ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$7–${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$8 mJ pulse energy, focused by an ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$9 mm lens to an approximately $4f^{13}6s^2$0m diameter spot, corresponding to a fluence of about $4f^{13}6s^2$1 J/cm$4f^{13}6s^2$2 per pulse. Under optimized conditions, a crystal about $4f^{13}6s^2$3m thick was deposited in one hour, and the Tm concentration was estimated to be on the order of several parts per thousand (Gaire et al., 2019).

A later, higher-sensitivity geometry condensed the rare-gas crystal directly onto the tip of a multimode optical fiber rather than onto a separate window. The fiber had a $4f^{13}6s^2$4m core diameter and $4f^{13}6s^2$5, and both excitation and fluorescence collection occurred through the same fiber. For argon growth, an initial $4f^{13}6s^2$6–$4f^{13}6s^2$7 s undoped layer was deposited first to monitor growth and protect the fiber tip; subsequent Tm incorporation again occurred by laser ablation during continued argon deposition. Thickness was monitored by Fabry–Perot oscillations, and the inferred deposition rate was about $4f^{13}6s^2$8–$4f^{13}6s^2$9 nm/s, with total thicknesses of roughly $f\!\!-\!f$0–$f\!\!-\!f$1m after $f\!\!-\!f$2–$f\!\!-\!f$3 min of doped growth. The temperatures reported at the end of growth were $f\!\!-\!f$4 K at the cold head and $f\!\!-\!f$5 K at the lower part of the clamp, with the explicit caution that the actual sample could be warmer (Gaire et al., 2023).

The most detailed site-resolved study retained the optical-fiber-tip architecture and reported argon condensation while the holder was kept at $f\!\!-\!f$6 K, with a measured growth rate of $f\!\!-\!f$7 nm/s by reflectometry using a $f\!\!-\!f$8 nm diode laser, over a total growth time of $f\!\!-\!f$9 min. Thulium was supplied continuously by ablation of a rotating Tm metal target using a pulsed $4f$0 nm Nd:YAG laser. After growth, samples were annealed by resistive heating up to $4f$1 K for short intervals; this improved spectral quality but risked sublimation or detachment. Once kept cold, a sample remained stable for months (Marfey et al., 10 Jul 2025).

Two practical features recur across the literature. First, the ablation source is not guaranteed to be purely atomic at the moment of generation; one study explicitly states that “atoms and possibly some clusters of atoms are generated,” although the spectroscopy is interpreted primarily in terms of atomic thulium in a low-concentration impurity regime because the narrow $4f$2 nm features correspond to known atomic Tm levels and the line pattern is reproducible across growths (Gaire et al., 2019). Second, annealing is a major control parameter in argon. Spectra change substantially after warming and re-cooling, indicating redistribution among trapping environments, and line narrowing improves markedly after stronger anneals (Gaire et al., 2023).

3. Electronic structure and the $4f$3 nm transition

The central optical feature of thulium-doped argon is the near-infrared magnetic-dipole transition between the fine-structure levels of the $4f$4 ground configuration. The lower level is $4f$5, the upper level is $4f$6, and the papers repeatedly identify the observed emission as the $4f$7 line near $4f$8 nm (Gaire et al., 2019).

In free neutral Tm, this is a single vacuum transition, quoted in the most recent study as $4f$9 nm. In solid argon, the observed site centers are at $5s$0 nm for both major trapping sites, so the net matrix shift is very small. Earlier work already emphasized the same point qualitatively, stating that the argon-hosted infrared transition is “otherwise unshifted from the vacuum wavelength of 1140 nm,” even though the visible transitions are substantially broader and more shifted (Marfey et al., 10 Jul 2025).

The host, however, removes the free-atom rotational symmetry. In a crystal-field picture, the $5s$1 and $5s$2 manifolds split into Kramers doublets. One high-resolution analysis notes that a single trapping site could in principle yield up to $5s$3 optical transitions between the four Kramers doublets of $5s$4 and the three Kramers doublets of $5s$5. Because the observed argon spectrum contains more than a dozen components overall, more than one trapping-site class must be present (Gaire et al., 2023).

Visible Tm fluorescence is also observed in argon, but it is spectroscopically less clean. Early laser-induced fluorescence measurements showed broad emission features in the $5s$6–$5s$7 nm range and an additional weak feature at $5s$8 nm. The authors of that study explicitly noted that specific visible-level assignments are difficult in argon because the relevant levels “could be mixed by interaction with the host.” Later work under $5s$9 nm excitation reported visible features mostly matching known thulium transitions in the $5p$0–$5p$1 nm range, plus an unknown line near $5p$2 nm (Gaire et al., 2019).

4. Spectroscopic phenomenology: from sub-nanometer structure to sub-GHz lines

The first direct evidence for narrow infrared structure in Tm:Ar came from monochromator-based fluorescence spectroscopy under excitation at $5p$3 nm. In argon, the $5p$4 nm band was fit by three Gaussian peaks at $5p$5 nm, $5p$6 nm, and $5p$7 nm, with the strongest component at $5p$8 nm. The monochromator-limited width at $5p$9 nm was $1140$0 nm FWHM, so each fitted component was resolution-limited and therefore narrower than $1140$1 nm. The same study concluded that the free-atom line becomes a three-component pattern in argon, with a central feature, a clear red satellite, and a likely weaker blue-side feature (Gaire et al., 2019).

Subsequent high-resolution absorption and emission spectroscopy resolved this coarse triplet into a dense manifold of narrow peaks. In the region near $1140$2 nm, four peaks were identified with separations of approximately $1140$3–$1140$4 GHz, equivalent to about $1140$5–$1140$6 nm. Fits to temperature-dependent scans in that window used three Lorentzians plus background and yielded a minimum linewidth of about $1140$7 GHz FWHM at the lowest temperature reached. The broad stitched argon spectrum contained more than a dozen apparent components, and fluorescence-excitation at one wavelength produced emission at multiple distinct wavelengths within the same manifold, demonstrating internal connectivity among lines rather than a set of completely independent resonances (Gaire et al., 2023).

The most refined measurements pushed the linewidth substantially lower. After brief anneals above $1140$8 K, the narrowest absorption lines reached $1140$9 pm, corresponding to $19$0 MHz. In the same work, the full fluorescence spectrum contained at least $19$1 sharp features, and pump–probe saturation spectroscopy separated them into at least two internally correlated subsets associated with two stable trapping sites, Site I and Site II. This progression from a three-feature unresolved envelope to $19$2 MHz lines is one of the key developments in the subject, because it enabled direct optical resolution of hyperfine and Zeeman structure in the matrix-isolated ensemble (Marfey et al., 10 Jul 2025).

The interpretation of linewidth also changed as instrumental resolution improved. The $19$3 measurements could only establish that the intrinsic linewidth remained unresolved and that its determination was “an open question” (Gaire et al., 2019). The $19$4 work argued that the observed $19$5 GHz line is likely mostly homogeneous, not predominantly inhomogeneous, on the basis of temperature dependence and pump–probe behavior. The $19$6 analysis retained that view and attributed the remaining width to a combination of homogeneous relaxation and residual matrix disorder, while suggesting that lower temperature and improved annealing should further reduce the homogeneous and inhomogeneous contributions, respectively (Gaire et al., 2023).

5. Trapping sites, crystal fields, hyperfine structure, and selection rules

The origin of the argon multiplet structure was initially discussed in terms of crystal-field effects versus multiple trapping sites. Early work noted that crystal-field splitting is an attractive explanation because the matrix breaks the full rotational symmetry of the free atom, but also acknowledged that multiple trapping sites with small relative shifts could account for similar phenomenology. The reproducibility of the line pattern across growths made either explanation plausible at that stage (Gaire et al., 2019).

Higher-resolution spectroscopy established that both ingredients are required. The strongest evidence for multiple sites is the substantial spectral rearrangement upon annealing, which indicates that thermal treatment redistributes the population among metastable trapping environments. The strongest evidence for sublevels within a site comes from emission spectroscopy: exciting one wavelength produces emission at several others in the same $19$7 nm manifold, including both red- and blue-shifted fluorescence, which implies internal level structure within at least some common site classes (Gaire et al., 2023).

The most detailed site model identifies at least two stable trapping sites, Site I and Site II, each producing $19$8 observable lines. This is consistent with complete lifting of the $19$9 and $1/16"$0 degeneracies into Kramers doublets, with one transition absent or strongly forbidden and identified as $1/16"$1. For Site I, the inferred crystal-field levels are $1/16"$2, $1/16"$3, $1/16"$4, and $1/16"$5 in $1/16"$6, together with $1/16"$7, $1/16"$8, and $1/16"$9 in ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$00. For Site II, the inferred levels are ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$01, ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$02, ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$03, and ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$04 in ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$05, together with ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$06, ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$07, and ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$08 in ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$09 (Marfey et al., 10 Jul 2025).

These level schemes are described by a nearly axial Stevens-operator crystal field,

${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$10

specialized primarily to

${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$11

The fitted axial parameters are ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$12 for Site I and ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$13 for Site II. The sign reversal corresponds to opposite energetic ordering of the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$14 doublets in the two sites. The paper concludes from the fit quality that both sites have a nearly axial crystal field and that the trapping-site symmetry must be lower than tetrahedral, although no unique microscopic site geometry is assigned (Marfey et al., 10 Jul 2025).

Hyperfine structure adds a further layer. Because the relevant isotope is ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$15 with ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$16, the hyperfine interaction is relatively simple, and the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$17 study provisionally suggested that the local four-line cluster near ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$18 nm is likely hyperfine-related, since the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$19–${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$20 GHz spacings are comparable to the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$21 GHz ground-state hyperfine splitting of free thulium (Gaire et al., 2023). The ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$22 study then resolved much of the hyperfine structure for Site I and showed excellent agreement with no-free-parameter calculations based on known atomic hyperfine constants plus the axial crystal-field assignments, thereby validating the site-resolved level scheme (Marfey et al., 10 Jul 2025).

Selection rules in argon are not those of a purely free-space M1 transition. In free thulium, the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$23 line is magnetic dipole. In solid argon, however, the observed spectra contain allowed ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$24 transitions, and line-strength analysis shows that pure M1 selection rules are insufficient. The ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$25 paper compares three channels—M1, forced-dipole (FD), and pseudoquadrupole (PQ)—and concludes that a combination of M1 and PQ transitions best explains the observed selection rules, while FD remains an alternative or additional contribution. Quantitatively, fits gave either ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$26 M1 plus ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$27 FD with fitted sample temperature ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$28 K, or ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$29 M1 plus ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$30 PQ with fitted sample temperature ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$31 K, with the latter favored qualitatively by the selection rules. This is the basis for the description of the host-modified transition as “magnetic-electric” (Marfey et al., 10 Jul 2025).

6. Relaxation, annealing, Zeeman spectroscopy, and applications

The metastable ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$32 lifetime in argon was first measured by chopping the excitation beam and fitting the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$33 nm fluorescence decay to

${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$34

That work reported ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$35 ms, shorter than the corresponding values in neon, solid helium, and a near-free-atom optical-lattice environment. The authors speculated that a non-centrosymmetric trapping site could induce a small electric-dipole matrix element and thereby enhance the radiative rate (Gaire et al., 2019).

Site-resolved lifetime measurements later refined the picture. The ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$36 study reported a radiative lifetime of ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$37 ms for Site I and ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$38 ms for Site II. It further noted that the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$39 ms lifetime measured for the upper clock level in an optical lattice would be reduced in argon to about ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$40 ms by ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$41 scaling of the M1 spontaneous-emission rate using the refractive index ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$42 of solid argon at ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$43 K, whereas the observed ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$44 ms suggests additional FD or PQ decay channels. This supports the broader conclusion that the host modifies not only the level energies but also the transition character (Marfey et al., 10 Jul 2025).

Annealing is one of the most consequential materials variables in Tm:Ar. In the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$45 experiments, argon absorption spectra changed substantially after annealing, and one sample annealed to ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$46 K displayed reproducible subsequent temperature dependence (Gaire et al., 2023). In the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$47 work, brief anneals above ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$48 K were associated with linewidths as narrow as ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$49 MHz (Marfey et al., 10 Jul 2025). By contrast, prolonged illumination can degrade optical performance. Earlier work reported bleaching over several hours, a visible bleached spot where the dye laser hit the sample, no recovery under white-light exposure or simple beam blocking, and only partial fluorescence recovery upon annealing. The mechanism remained unresolved (Gaire et al., 2019).

The Zeeman effect is directly resolvable in Tm:Ar at small fields. The most detailed measurements studied millitesla fields, especially around ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$50 mT, and used polarized fluorescence spectra to distinguish subensembles with different orientations of the local crystal axis relative to the applied field. Defects whose local axis is parallel to the field exhibit predominantly linear Zeeman splitting, whereas other orientations can show more quadratic behavior. This orientation selectivity was modeled with Monte Carlo simulations for ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$51 randomly oriented defect axes and used to connect amplitude patterns with transition multipolarity (Marfey et al., 10 Jul 2025).

These properties make Tm:Ar relevant to sensing and quantum photonics. The earliest study already emphasized argon and neon as “prime targets for surface sensing applications,” because the host can be straightforwardly grown on arbitrary substrates at cryogenic temperatures and the ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$52 nm line remains very narrow on the scale of the apparatus (Gaire et al., 2019). The ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$53 work reframed this in terms of dense solid-state ensembles of nearly identical optical emitters or scatterers, arguing that the inhomogeneity is well below the GHz scale and that many embedded atoms may therefore behave as though they share nearly the same resonance frequency (Gaire et al., 2023). The ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$54 paper then demonstrated an all-optical DC magnetometry protocol based on a two-color fluorescence ratio using the frequencies ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$55 nm and ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$56 nm, with an estimated sensitivity of ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$57 near ${}^2F_{7/2}\leftrightarrow{}^2F_{5/2}$58 mT. The same work explicitly states that millitesla fields can already be detected optically in matrix-isolated atoms without microwave or RF fields (Marfey et al., 10 Jul 2025).

Several open questions remain. The exact microscopic identity of the trapping sites is still not established; the broader multiplet is not fully decomposed into site classes and sublevel transitions; the degree to which the linewidth can approach the transform-limited scale remains unknown; and bleaching, site transfer, and growth-dependent disorder are not yet quantitatively modeled. A plausible implication is that future progress will depend less on discovering the existence of narrow transitions—which is already established—and more on mastering crystal growth, annealing, and site selectivity so that a single site class, or even a single axis class, can be isolated reproducibly (Marfey et al., 10 Jul 2025).