Dual-Excitation Photon Absorption Remote Sensing

• Dual-Excitation PARS is an all‐optical, non-contact technique that uses two excitation wavelengths to capture both radiative and non‐radiative absorption signals.
• It enables virtual histology and multi-stain synthesis by integrating contrast from endogenous chromophores without the need for exogenous labels.
• Recent systems interlace pulses and deploy advanced detection schemes to achieve high resolution, fast imaging speeds, and quantitative biomolecular mapping.

Dual-excitation Photon Absorption Remote Sensing (PARS) denotes a family of all-optical, non-contact absorption microscopies in which two excitation wavelengths are used to interrogate endogenous chromophores, while a co-focused probe and, in later systems, additional radiative channels recover complementary contrast without acoustic coupling or exogenous labels. The lineage begins with reflection-mode Photoacoustic Remote Sensing for histology-like imaging of unstained tissue (Ecclestone et al., 2020) and broadens into second-generation and “total-absorption” implementations that jointly capture radiative and non-radiative relaxation, enabling virtual hematoxylin-and-eosin (H&E), multi-stain synthesis, biomolecular characterization, and functional oximetry (Ecclestone et al., 2021, Ecclestone et al., 25 Jun 2025, Tweel et al., 5 Sep 2025).

1. Historical development and scope

The earliest dual-excitation histology-oriented PARS implementation extended reflection-mode PARS by combining a tunable excitation laser at 250 nm and 420 nm with a fast 266 nm source. In that system, 250 nm targeted the strong ultraviolet absorption peak of DNA for “hematoxylin-like” nuclear contrast, 420 nm targeted cytochromes and other cytoplasmic proteins for eosin-like contrast, and 266 nm provided rapid preview imaging with nuclear emphasis (Ecclestone et al., 2020). The same work framed PARS as an all-optical, non-contact, reflection-mode technique that visualizes endogenous optical absorption contrast in thick, unprocessed tissues.

A related but distinct precursor to later dual-excitation systems was the single-acquisition dual-contrast PARS microscope, which used 266 nm ultraviolet excitation for nuclear absorption contrast while leveraging the 1310 nm detection path for concurrent scattering contrast analogous to eosin staining. That architecture demonstrated label-free histology-like imaging in formalin-fixed paraffin-embedded (FFPE) blocks and frozen sections, and established that histology-relevant nuclear and extra-nuclear information could be recovered in one pass without chemical dyes (Ecclestone et al., 2021).

Second-generation PARS then broadened from a principally non-radiative photoacoustic readout to a total-absorption framework. The TA-PARS implementation used 266 nm and 515 nm excitation together with a 405 nm probe to record scattering, non-radiative, and radiative signals simultaneously, and introduced the quantum efficiency ratio (QER) as an absorption-derived metric for distinguishing biomolecular response classes (Ecclestone et al., 2021). A later formulation explicitly renamed the modality “Photon Absorption Remote Sensing,” emphasizing simultaneous collection of radiative and non-radiative de-excitation channels and using 266 nm and 532 nm excitation to generate a 6-dimensional per-pixel signature for statistical biomolecule mapping (Ecclestone et al., 25 Jun 2025).

The most recent dual-excitation histopathology-oriented platform adds long-wave ultraviolet A (UVA) 355 nm excitation to the established 266 nm source. In that system, the two wavelengths are interlaced pulse-by-pulse, allowing concurrent acquisition of complementary radiative and non-radiative contrasts without compromising imaging speed, and supporting virtual synthesis of H&E, Masson’s trichrome, periodic acid-Schiff (PAS), and Jones’ silver stains from a single scan (Tweel et al., 5 Sep 2025).

2. Governing physics and signal channels

The foundational absorption model in PARS follows Beer–Lambert attenuation,

$I(x,\lambda)=I_0 e^{-\mu_a(\lambda)x},$

where $\mu_a(\lambda)$ is the absorption coefficient at wavelength $\lambda$. In the conventional photoacoustic description, the initial pressure rise is

$p_0=\Gamma \mu_a F,$

with $\Gamma$ the Grüneisen parameter and $F$ the local fluence. Reflection-mode PARS does not detect transmitted ultrasound; instead, the transient thermoelastic perturbation changes local refractive index and modulates a co-focused probe beam. In the 2020 dual-excitation brain-tissue system, the detection sensitivity was expressed as

$$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$

linking the measured back-reflected modulation directly to absorbed optical energy (Ecclestone et al., 2020).

Later PARS systems formalized the separation of radiative and non-radiative relaxation. In TA-PARS, for chromophore concentration $N$, total absorption cross-section $\sigma_a$, and excitation intensity $I_0$, the radiative and non-radiative absorption coefficients are

$\mu_a(\lambda)$0

with measured signals satisfying

$\mu_a(\lambda)$1

That work defined

$\mu_a(\lambda)$2

after normalization by detector responsivity, excitation pulse energy, and probe power (Ecclestone et al., 2021).

A different QER definition appears in the later Photon Absorption Remote Sensing framework, where PARS simultaneously measures radiative amplitude $\mu_a(\lambda)$3, non-radiative energy $\mu_a(\lambda)$4, and non-radiative decay rate $\mu_a(\lambda)$5. There,

$\mu_a(\lambda)$6

so the reported “QER” is not uniform across the literature (Ecclestone et al., 25 Jun 2025). This distinction is methodologically important because papers using the same acronym do not necessarily encode the same physical normalization or dynamic range.

For ophthalmic functional imaging, PARS amplitudes at two wavelengths $\mu_a(\lambda)$7 were modeled as

$\mu_a(\lambda)$8

allowing inversion for $\mu_a(\lambda)$9, $\lambda$0, and oxygen saturation

$\lambda$1

In that implementation, PARS served as an absorption modality co-registered with OCT structural scattering contrast (Hosseinaee et al., 2021).

3. Optical architectures and excitation schemes

Reported dual-excitation PARS systems differ primarily in wavelength pair, readout channels, and target application. The principal configurations reported to date are summarized below.

Paper	Excitation scheme	Principal contrast or use
(Ecclestone et al., 2020)	250 nm, 420 nm, plus fast 266 nm preview	DNA, cytoplasm, H&E-like brain histology
(Ecclestone et al., 2021)	266 nm and 515 nm with 405 nm probe	Radiative, non-radiative, scattering, QER
(Ecclestone et al., 25 Jun 2025)	266 nm and 532 nm with 405 nm probe	6-D biomolecular fingerprinting and unmixing
(Tweel et al., 5 Sep 2025)	Interlaced 266 nm and 355 nm with 405 nm scattering probe	Whole-slide virtual multi-staining
(Hosseinaee et al., 2021)	Fiber-SRS-generated 532 nm and 558 nm selection	Hemoglobin oximetry in vivo

The 2020 reflection-mode histology system used a 1310 nm superluminescent diode as a continuous-wave detection beam, a 15× reflective objective with $\lambda$2, galvanometric scanning over fields up to $\lambda$3, and 14-bit digitization of the probe modulation. The excitation subsystem combined a tunable 1 kHz, 3 ns pulsed laser at 250 nm or 420 nm with a 20 kHz, 0.6 ns 266 nm source (Ecclestone et al., 2020).

The TA-PARS platform used 515 nm visible pulses generated as the second harmonic of a 1030 nm, 2 ps pulsed fiber laser, a 266 nm 400 ps pulsed diode laser, and a 405 nm continuous-wave probe beam co-focused through a single $\lambda$4 UV-corrected objective. Back-reflected 405 nm light was directed to an avalanche photodiode for non-radiative contrast and baseline scattering, while spectrally separated radiative emissions from the 515 nm and 266 nm excitation channels were sent to two additional photodiodes (Ecclestone et al., 2021).

The 2025 whole-slide multi-stain system implemented the first PARS use of long-wave UVA 355 nm together with 266 nm. A delay generator alternately triggered both 50 kHz lasers so that pulses were interlaced, the stage moved at $\lambda$5, alternate-wavelength pulses were separated by 125 nm, same-wavelength pulses by 250 nm, and a 200 MS/s digitizer captured $\lambda$6 traces from each detector per pulse. That system simultaneously collected 266 nm non-radiative and radiative signals, 355 nm non-radiative and radiative signals, the 405 nm scattering baseline, and power references (Tweel et al., 5 Sep 2025).

The biomolecular PARS system described in 2025 used 266 nm and residual 532 nm excitation with a 405 nm probe and transmission-mode collection through a second high-NA visible objective. Each pump pulse produced a radiative signal, a non-radiative time trace, and a per-pulse reference, yielding a 6-dimensional feature vector $\lambda$7 at each of the two excitation wavelengths (Ecclestone et al., 25 Jun 2025).

4. Histopathology, virtual H&E, and virtual multi-staining

The most developed application area for dual-excitation PARS is label-free histopathology. In the 2020 brain-tissue study, 250 nm excitation yielded hematoxylin-like nuclear contrast and 420 nm excitation yielded eosin-like cytoplasmic contrast; false-color mixing of the two generated a composite closely resembling standard pink-and-purple H&E. A fast 266 nm mode provided real-time preview of necrotic and viable tumor regions, microvascular proliferations, and internuclear spacing, while 1 mm² could be imaged in 1.5 min at 266 nm or in 44 min for the full dual-wavelength tunable scan. The reported lateral resolution was $\lambda$8 at 250 nm, $\lambda$9 at 420 nm, and $p_0=\Gamma \mu_a F,$0 at 266 nm (Ecclestone et al., 2020).

A related single-acquisition dual-contrast PARS system, although not dual-excitation, clarified how much of the H&E analogue can be produced from intrinsic absorption plus scattering alone. With 266 nm excitation and 1310 nm scattering detection, it reported absorption-channel lateral resolution $p_0=\Gamma \mu_a F,$1, scattering-channel lateral resolution $p_0=\Gamma \mu_a F,$2, axial resolutions of $p_0=\Gamma \mu_a F,$3 and $p_0=\Gamma \mu_a F,$4, 50 kHz A-line rate, and a typical $p_0=\Gamma \mu_a F,$5 frame acquired in $p_0=\Gamma \mu_a F,$6 minutes. Combined H&E-emulated images achieved $p_0=\Gamma \mu_a F,$7 and $p_0=\Gamma \mu_a F,$8 against brightfield H&E, with one-to-one correspondence of nuclear size, shape, and spacing (Ecclestone et al., 2021).

TA-PARS further expanded the histological analogy by measuring radiative and non-radiative responses in the same scan. Its non-radiative channel highlighted nuclear and strongly absorbing structures, the radiative channel visualized extranuclear matrix including collagen, elastin, and heme proteins, and QER maps separated high-yield chromophores from low-yield tissues. The reported lateral resolution was approximately 350 nm, axial resolution approximately $p_0=\Gamma \mu_a F,$9, excitation pulse energies as low as 400 pJ, and probe powers down to $\Gamma$0. The work reported $\Gamma$1 for dye QER correlation (Ecclestone et al., 2021).

The 266/355 nm interlaced whole-slide system moved from physical H&E emulation to learned virtual staining. Using a RegGAN framework with a ResNet-based encoder-decoder generator, a 70×70 PatchGAN discriminator, and a ResUNet registration network, it generated virtual H&E, Masson’s trichrome, PAS, and Jones’ silver across diverse human and murine tissues. Quantitatively, on 256×256 test patches ($\Gamma$2), the dual-excitation model achieved $\Gamma$3 and $\Gamma$4, compared with $\Gamma$5 and $\Gamma$6 for 266 nm only, and $\Gamma$7 and $\Gamma$8 for 355 nm only. In a masked evaluation by three board-certified pathologists using 20 whole-slide image regions, mean diagnostic quality was $\Gamma$9 for chemical stains and $F$0 for virtual stains, and image origin could not be reliably distinguished (Tweel et al., 5 Sep 2025).

An important distinction follows from these studies. In some PARS papers, “virtual H&E” means direct wavelength-selective false-coloring of measured endogenous absorption and scattering channels; in others, it means an image-translation output trained against chemically stained whole-slide images. The two usages are related but not identical, and they impose different constraints on registration, validation, and failure modes (Ecclestone et al., 2020, Tweel et al., 5 Sep 2025).

5. Functional and biomolecular imaging beyond histology

Dual-excitation PARS is not restricted to histology-like rendering. In ophthalmic imaging, a multimodal PARS/OCT system used a 532 nm ytterbium-doped fiber laser to generate stable Stimulated Raman Scattering peaks in a temperature-controlled single-mode fiber, then selected two wavelengths—commonly 532 nm and 558 nm—for hemoglobin oxygenation imaging. The system acquired 500×500 pixel en-face rasters over $F$1 in approximately 2.5 s per wavelength, automatically co-registered with swept-source OCT because both modalities shared scan mirrors and telecentric optics. Reported metrics included in vitro PARS absorption contrast lateral resolution of approximately $F$2, in vivo ocular vasculature resolution of $F$3, OCT axial resolution of approximately $F$4, PARS absorption SNR of $F$5, and the ability to resolve arterial saturation of approximately $F$6 versus venous or capillary saturation of approximately $F$7 around the iris (Hosseinaee et al., 2021).

The 2025 Photon Absorption Remote Sensing formulation generalized dual-excitation PARS into a biomolecular mapping framework. Each pixel was represented by a 6-dimensional vector comprising radiative amplitude, non-radiative energy, and non-radiative decay rate at 266 nm and 532 nm. Endmembers were extracted by a Gaussian mixture model, and abundance maps were obtained by non-negative least squares under non-negativity and sum-to-one constraints. In murine brain FFPE sections, the reported five endmembers were nuclei, gray matter, white matter, red blood cells, and paraffin, and the resulting abundance maps were described as structurally indistinguishable from chemical stains and deep-learning virtual stains (Ecclestone et al., 25 Jun 2025).

TA-PARS occupies an intermediate position between virtual histology and quantitative biomolecule analysis. Because it simultaneously measures scattering, radiative, and non-radiative absorption responses, it can visualize a broad range of endogenous and exogenous chromophores and has been proposed for in vivo functional imaging of metabolic cofactors such as NADPH and flavins, as well as for multiplexed assays using dyes of known quantum yield (Ecclestone et al., 2021). A plausible implication is that dual-excitation PARS is evolving from a stain-replacement technology into a higher-dimensional absorption phenotyping platform.

6. Performance limits, misconceptions, and open technical questions

Across the literature, dual-excitation PARS combines high lateral resolution with strong wavelength-dependent trade-offs in speed, field of view, and penetration depth. Histology-oriented UV systems reported submicrometer to approximately $F$8 lateral resolution: $F$9 at 250 nm and $$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$0 at 420 nm in the 2020 reflection-mode brain system, approximately 350 nm in TA-PARS, approximately 500 nm at both 266 nm and 355 nm in whole-slide dual-excitation PARS, and $$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$1 non-radiative resolution in the 266/532 nm biomolecular platform (Ecclestone et al., 2020, Ecclestone et al., 2021, Tweel et al., 5 Sep 2025, Ecclestone et al., 25 Jun 2025).

A common misconception is that dual-excitation alone removes the depth limitation of optical histology. The reported systems indicate otherwise. In dual-contrast PARS, UV absorption was restricted to surface nuclear layers below $$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$2 depth, while the scattering channel extended to approximately $$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$3 into FFPE blocks (Ecclestone et al., 2021). In the 266/355 nm virtual multi-stain platform, effective imaging was confined to $$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$4 section thickness due to strong UV absorption (Tweel et al., 5 Sep 2025). The 266/532 nm Photon Absorption Remote Sensing system was demonstrated on $$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$5 FFPE sections in transmission mode (Ecclestone et al., 25 Jun 2025). By contrast, the earlier reflection-mode photoacoustic PARS literature reported penetration through at least 1 mm of unsectioned tissue in other organs, indicating that depth performance is strongly architecture- and wavelength-dependent rather than an intrinsic consequence of “dual excitation” itself (Ecclestone et al., 2020).

Throughput also varies markedly with implementation. The 2020 tunable-laser system required 22 minutes per wavelength for a $$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$6 mechanically scanned image at 900 nm pixels, though its 266 nm optical galvo mode captured $$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$7 at 100,000 points in 5 s (Ecclestone et al., 2020). The 2021 dual-contrast system achieved up to 50,000 pixels/s per channel and was described as approximately 100× faster than prior 1 kHz tunable-laser PARS systems (Ecclestone et al., 2021). The interlaced 266/355 nm platform reported $$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$8 per contrast channel with no penalty in imaging speed due to wavelength interlacing, though with a slight $$S \propto \left(\frac{dR}{dn}\right)\Delta n \propto \left(\frac{dR}{dn}\right)\left(\frac{\beta T_0}{\rho C_p}\right)\mu_a F,$$9 SNR reduction per channel versus monochromatic scanning (Tweel et al., 5 Sep 2025).

Current limitations are correspondingly diverse. Histology-oriented systems remain sensitive to stage synchronization, baseline correction, and UV overexposure management (Tweel et al., 5 Sep 2025). The single-acquisition dual-contrast architecture depends on refractive-index mismatch with paraffin for strong scattering contrast, so fresh or unembedded samples yield poor background reference (Ecclestone et al., 2021). In biomolecular PARS, endmembers may shift between tissue types or preparations, fluorescence spectra and lifetimes are not yet measured, and the non-radiative trace is dominated by thermal rather than purely acoustic contributions within the recorded window (Ecclestone et al., 25 Jun 2025). In ophthalmic PARS, excitation wavelengths are limited to discrete Stimulated Raman peaks, motion artifacts impose frame-rate constraints, and fluence calibration in vivo remains difficult because vessel geometry, melanin background, tear film, and corneal curvature all perturb attenuation and scattering (Hosseinaee et al., 2021).

The main open direction, as the literature presents it, is expansion of contrast dimensionality without sacrificing acquisition speed. Reported proposals include additional wavelengths such as 266/355/405 nm “trifurcation,” deeper-learning models for 3D alignment, galvo or resonant scanning for real-time imaging, shorter-wavelength scattering probes for improved resolution, and extension to reflection-mode thick-tissue biomolecular mapping (Tweel et al., 5 Sep 2025, Ecclestone et al., 2021, Ecclestone et al., 25 Jun 2025). This suggests that dual-excitation PARS is best understood not as a single fixed microscope design, but as a modular absorption-imaging framework whose diagnostic value depends on the chosen wavelength pair, the partition of radiative versus non-radiative detection, and the downstream reconstruction or inference strategy.