Ashes or Breath: Residue vs. Active Process

• Ashes or Breath is a multifaceted concept contrasting inert residue with dynamic activity, spanning astrophysics, physiology, and ethical design.
• It is operationalized in studies ranging from neutron-star burst analyses and vocal breath segmentation (with 89% boundary accuracy) to immersive Mixed Reality games.
• Interdisciplinary applications in wireless sensing, aerosol communication, and biometric authentication underscore its role in advancing both technical and societal research.

“Ashes or Breath” designates a contrast that several recent arXiv literatures use in markedly different but structurally related ways. In one use, it is the title of a Mixed Reality game in which a player in a museum fire must save either a living cat or a priceless cultural artifact (Sun et al., 18 Aug 2025). In others, the phrase is a conceptual shorthand for the difference between inert residue and active process: vocal breath sounds as contactless physiological data (Yadav et al., 2023), exhaled breath as an information-bearing aerosol or wireless sensing signal (Khalid et al., 2018, Tewes et al., 2022), protostellar outflows as the earliest “breath” of star formation in the ASHES survey (Lin et al., 19 Jul 2025), and neutron-star burst ashes as the compositional products of thermonuclear burning, sometimes carried outward in a radiative wind (Zhen et al., 16 Feb 2025, Zand et al., 2010). The phrase therefore does not name a single doctrine. It names a recurrent opposition between what remains and what moves, between residue and emission, and between legacy and living process.

1. Conceptual scope

The cited literature uses “ashes” in at least three distinct senses. In the neutron-star literature, “ashes” are the nuclear products that remain on, and are later buried into, the neutron star surface after a Type I X-ray burst (Zhen et al., 16 Feb 2025). In the superexpansion-burst literature, the same term denotes freshly synthesized heavy-element products of helium burning, predominantly iron-peak nuclei, that can be exposed in a radiative wind (Zand et al., 2010). In the Mixed Reality game “Ashes or Breath,” the contrast is ethical rather than material: the player must choose between preserving a living cat and preserving the Mona Lisa during a museum fire (Sun et al., 18 Aug 2025).

“Breath” is equally heterogeneous, but it is consistently tied to ongoing activity. In vocal breath sound analysis, it is a measurable acoustic signal recorded at the mouth and segmented into breath phases and boundaries (Yadav et al., 2023). In wireless sensing, it is a periodic thoracic motion that becomes a phase change in Channel State Information at 5 GHz (Tewes et al., 2022). In aerosol transmission, it is a stream of volatile organic compounds and biological entities that can be modeled as a communication channel (Khalid et al., 2018). In the ASHES star-formation survey, “breath” is used metaphorically for protostellar outflows that inject momentum, energy, shocks, and chemical change into a cold clump (Lin et al., 19 Jul 2025).

A common misconception is that the contrast is always binary. The neutron-star wind results instead show “ashes in the breath”: heavy-element ashes are inferred from absorption edges in the outflow itself (Zand et al., 2010). Another misconception is that ASHES denotes relic or post-activity matter. In that literature, ASHES is the “ALMA Survey of 70 $μ$m Dark High-mass Clumps in Early Stages,” and the targeted clumps are explicitly described as “not the ‘ashes’ of past activity” but as very early cluster seeds (Lin et al., 19 Jul 2025). This suggests that the phrase is best understood as a family resemblance across domains rather than a fixed technical term.

2. Breath as segmented and sensed physiology

In “An unsupervised segmentation of vocal breath sounds,” breath is treated as a quasi-periodic acoustic process whose boundaries can be inferred without expert labels (Yadav et al., 2023). The work focuses on vocal breath sounds (VBS), that is, breath sounds recorded at the mouth. They are “easy and contactless to record” relative to tracheal and lung breath sounds, and they are used because they can support downstream discrimination between asthmatic and healthy subjects. The proposed method exploits the periodic nature of breath signal energy and applies dynamic programming with prior information of the number of breath phases $P$ and breath phase duration $d$ to find boundaries. On 367 breath boundaries from 60 subjects, with 307 breaths, the reported performance is $M = 89\%$, $I = 13\%$, $D = 11\%$, and $S = 79\%$, and classification performance using estimated boundaries is comparable with ground-truth boundaries (Yadav et al., 2023).

In “IRS-enabled Breath Tracking with Colocated Commodity WiFi Transceivers,” breath becomes a radio-sensed displacement signal (Tewes et al., 2022). The physical mechanism is a path-length change caused by chest motion, summarized as

$$\Delta \phi(t) \approx \frac{2\pi}{\lambda}\Delta d(t),$$

with $\lambda \approx 60$ mm in the 5 GHz band. The study uses a 256-element IRS prototype, commodity WiFi hardware, Channel State Information from the 5 GHz band, and wired local-oscillator synchronization. The system is designed to suppress self-interference in the analog domain so that small breathing-induced phase variations can be recovered. The reported case study shows breath tracking “regardless of the position in an indoor environment in a room-level range,” and the system captures both breath frequency and breathing patterns (Tewes et al., 2022).

Taken together, these works treat breath not as background noise but as structured signal. In one case the structure is temporal segmentation into inspiration and expiration phases; in the other it is a slow periodic phase modulation in CSI. This suggests that breath is technically useful precisely because it is rhythmic, constrained, and difficult to replace by arbitrary heuristics.

3. Breath as communication, authentication, and forgery cue

“Communication Through Breath: Aerosol Transmission” formalizes exhaled breath as a communication source (Khalid et al., 2018). The paper states that a systematic review identified 872 VOCs in the breath of healthy humans, and it emphasizes that breath also carries biological entities and pathogenic aerosols produced by sneezing, coughing, talking, and normal breathing. In this framework, the transmitter is the human body, the signal is the time-space pattern of concentrations of targeted molecules or particles, and the message is encoded biological information. The channel is modeled by advection and diffusion, including Gaussian plume and Gaussian puff approximations, and the receiver is a biosensing system that may use electrostatic precipitation and silicon nanowire field-effect transistors (Khalid et al., 2018).

In “Personalized breath based biometric authentication with wearable multimodality,” breath is a biometric rather than a broadcast message (Bui et al., 2021). The hardware combines an acoustic sensor near the nose with an accelerometer and gyroscope on the chest, and the dataset comprises 20 volunteers performing normal, deep, and strong breathing across multiple sessions. Two multimodal models are evaluated: CNN-LSTM and TCN. For identification, the multimodal CNN-LSTM reaches $97.06\% \pm 1.24$ on normal breathing, and for verification its EER is $P$0 in Scenario 1 and $P$1 in Scenario 2. The multimodal TCN reaches $P$2 identification accuracy on normal breathing, with EER $P$3 in Scenario 1 and $P$4 in Scenario 2 (Bui et al., 2021). The study therefore treats breath as a persistent multimodal signature anchored in nose sound features and chest movement.

In “Every Breath You Don’t Take: Deepfake Speech Detection Using Breath,” breath becomes a cue for authenticity (Layton et al., 2024). The central hypothesis is that current speech deepfake generation techniques do not sufficiently incorporate breaths. The method trains a breath detector and reduces each article to average breaths per minute, average breath duration, and average spacing between breaths. On in-the-wild news audio, the breath-based discriminator attains 1.0 AUPRC and 0.0 EER on test data across 33.6 hours of audio, whereas the compared SSL-wav2vec model attains 0.72 AUPRC and 0.99 EER on the same test set (Layton et al., 2024). The paper’s technical claim is not that all future deepfakes will fail in this way; it is that breath is currently a strong high-level discriminator because natural speech remains physiologically constrained.

These three lines of work converge on a single point: breath can function as message, credential, and liveness marker. A plausible implication is that breath is especially useful when a system must discriminate between physically grounded processes and signals that only reproduce surface form.

4. Breath in moving air

The transport literature makes breath a fluid-dynamical object. “Turbulent dispersion of breath by the wind” models exhaled air and its CO$P$5 as a passive scalar carried by a turbulent flow, and it emphasizes the distinction between a short-distance ballistic-like regime and Taylor’s diffusive-like regime at larger scales (Poydenot et al., 2021). Using a single correlation time model for velocity fluctuations, the radial spread is written as

$P$6

For $P$7, $P$8 grows linearly with $P$9, yielding an inverse-squared decay of concentration along the axis. The paper reports that CO$d$0 concentration emitted when breathing typically decays as the inverse squared distance to the mouth, and field experiments in corridors and shopping malls are fit by

$d$1

with fitted values $d$2 m and $d$3 for the dry-ice source, and $d$4 m and $d$5 for human breathing in a fan-induced wind (Poydenot et al., 2021).

The aerosol-communication literature supplies the complementary systems view (Khalid et al., 2018). There the concentration field obeys an advection–diffusion equation, and Gaussian puff models are used for a single breath, cough, or sneeze. The transmitter is mobile and non-cooperative, multiple transmitters create source interference, and the main theoretical difficulty is that synchronization and parameter estimation are much less controlled than in conventional communication systems (Khalid et al., 2018).

These works correct a simple diffusion intuition. Breath does not immediately become a homogeneous background. It first forms a coherent, turbulent, downwind structure whose dilution depends on mean flow, turbulent fluctuations, and source geometry. This is why the same physical process can support both public-health risk models and communication-system abstractions.

5. Ashes and breath in neutron-star bursts

In the neutron-star literature, ashes are explicitly defined and quantified. “Systematic study of the composition of Type I X-ray burst ashes” uses 1D MESA simulations with a 304-isotope reaction network to study how neutron-star structure changes the composition of Type I X-ray burst ashes (Zhen et al., 16 Feb 2025). The paper defines the mean mass number of the ashes as

$d$6

and analyzes its relation to burst strength

$d$7

The reported result is that heavier-element mass fractions increase with increasing surface gravity, yielding a higher $d$8 except for higher mass neutron stars where envelope temperature and recurrence time compete. Heavier nuclei synthesis is inversely correlated with base heating $d$9 and metallicity $M = 89\%$0, and the new reaction rates considered have only minimal effects on burst ashes. In hydrogen-rich X-ray binary systems, nuclei heavier than $M = 89\%$1 are fertile produced with larger neutron-star mass, smaller neutron-star radius, smaller base heating, and smaller metallicity (Zhen et al., 16 Feb 2025).

“Evidence of heavy-element ashes in thermonuclear X-ray bursts with photospheric superexpansion” shows that ashes can become visible in the burst outflow itself (Zand et al., 2010). The paper surveys 32 superexpansion bursts and finds that at least 31 are from candidate ultracompact X-ray binaries. In 2 bursts it detects strong absorption edges during the expansion phase, with edge energies and depths consistent with the H-like or He-like edge of iron-peak elements with abundances greater than 100 times solar. The superexpansion phase lasts only a few seconds and is always followed by a moderate expansion phase with $M = 89\%$2 km, while the duration of moderate expansion is of order the burst decay time $M = 89\%$3 (Zand et al., 2010).

A common misconception is that “ashes” and “breath” are opposites in this setting. The superexpansion results indicate the opposite. The radiative “breath” of the burst is compositionally altered by the ashes, and the ashes are inferred precisely through the spectral properties of the outflow. This suggests that, in neutron-star burst physics, the contrast resolves into coupling rather than exclusion.

6. ASHES and the earliest outflows of massive star formation

In the star-formation literature, ASHES is an acronym rather than a residue. The “ALMA Survey of 70 $M = 89\%$4m Dark High-mass Clumps in Early Stages (ASHES)” targets 51 dense, cold clumps that are dark at 70 $M = 89\%$5m, with ages $M = 89\%$6 yr, temperatures $M = 89\%$7 K, and densities $M = 89\%$8 (Lin et al., 19 Jul 2025). These are described as the earliest phases of massive star-forming cluster seeds. At a typical resolution of $M = 89\%$9 au, the project probes fragmentation into cores, infall onto protostars, protostellar outflows and shocks, and early chemical changes in dense gas (Lin et al., 19 Jul 2025).

Paper XII reports an “unanchored forked stream” in G34.74-0.12 (Lin et al., 19 Jul 2025). The main outflow is driven by a $I = 13\%$0 core and exhibits a fork-shaped stream in its red-shifted lobe traced by CO (2-1), SiO (5-4), and H$I = 13\%$1CO (3$I = 13\%$2-2$I = 13\%$3). The abstract reports a momentum of $I = 13\%$4, an energy of $I = 13\%$5, and a dynamical timescale of $I = 13\%$6 yr. The forked point does not coincide with any dust continuum core above $I = 13\%$7, while enhanced relative abundances of SiO, H$I = 13\%$8CO, and CH$I = 13\%$9OH with respect to CO, together with increased temperature at the forked point, indicate a collisional origin. CO (2-1) also traces three other outflows in the region, with masses 0.40, 0.02, and 0.15 $D = 11\%$0 and momenta 5.2, 0.2, and 1.8 $D = 11\%$1 km s$D = 11\%$2 (Lin et al., 19 Jul 2025).

The paper explicitly frames the result on the “breath” side of the contrast. Protostellar outflows are treated as the first energetic injection into a clump that is still dark at 70 $D = 11\%$3m. The significance is not only kinematic. The forked point R2 reaches $D = 11\%$4 K and shows chemical enrichment associated with shocks, so the outflow’s “breath” is also a mechanism of thermal and chemical transformation (Lin et al., 19 Jul 2025).

7. Mixed Reality, moral dilemma, and the life–legacy opposition

In HCI, “Ashes or Breath” names a values-centered Mixed Reality game rather than a physical process (Sun et al., 18 Aug 2025). The experience runs on MR-HMDs, specifically Meta Quest 3 / 3s, using passthrough mixed reality and hand tracking. In the Museum Scene, the system scans the real room, anchors a virtual Mona Lisa to a wall, and places a virtual cat on the floor. A fire breaks out, and the player must physically pick up either the cat or the painting and run to the door. There is no score, no explicit correct answer, and the decision is irreversible within that run (Sun et al., 18 Aug 2025).

The second phase is the Rewind Room, in which four memory bubbles present immediate aftermath, values reflection, personal emotional impact, and broader societal consequences (Sun et al., 18 Aug 2025). The bubbles frequently emphasize the path not taken. After exploring them, the player can either end the experience or return to the museum and choose again from the opposite perspective. The design goal is to stage ambiguity through embodied interaction and spatial immersion rather than abstract thought experiment.

The preliminary evaluation includes 12 participants, ages 21–35, from digital design, computer science, and the general public (Sun et al., 18 Aug 2025). Reported means include 4.5 for “I would consider the game worth playing,” 4.5 for “I felt emotionally immersed when the museum caught fire,” 4.1 for “Interacting with virtual objects using my real hand felt immersive,” and 4.1 for “The game helped me understand the ‘other side’ of the dilemma.” Five participants initially saved the cat, five initially saved the painting, all participants made at least two decisions, two changed their initial choice after reflection, and seven reported better understanding the alternative perspective (Sun et al., 18 Aug 2025). The paper also states clear limitations: the sample size is small, cultural variability is not fully explored, the looping structure may dilute original decision weight, and no long-term follow-up was done.

The MR usage is therefore distinct from the signal-processing and astrophysical usages, but the underlying contrast is consistent. “Breath” marks a living being with immediate moral salience; “ashes” marks cultural legacy under threat from destruction. The paper does not resolve the conflict. It stages it, then returns the unresolved tension to the participant.

Across these literatures, “Ashes or Breath” consistently organizes questions of residue, activity, and value. In respiratory signal processing and wireless sensing, breath is a periodic, measurable signature. In aerosol physics, it is a turbulent scalar with near-field structure and far-field dilution. In biometrics and deepfake detection, it is an authenticity cue. In neutron-star astrophysics, ashes record nucleosynthesis while burst outflows can expose those ashes. In the ASHES survey, early massive star formation is already dynamically active rather than quiescent. In Mixed Reality ethics, the contrast becomes an explicit choice between living life and cultural legacy. The phrase therefore functions less as a single concept than as a recurring analytical opposition whose meaning depends on whether the domain is physiology, communication, astrophysics, or moral design.