Type Iax Supernovae: Peculiar Explosions

• Type Iax supernovae are a peculiar class of thermonuclear explosions characterized by low ejecta velocities, subluminous peaks, and spectra that never fully transition to a nebular state.
• Detailed photometric and spectroscopic analyses reveal fast rise times, unique light-curve shapes, and a lack of a prominent near-infrared secondary maximum compared to normal SNe Ia.
• Progenitor studies and late-time observations support weak deflagration models that leave behind bound remnants and can contribute iron-rich dust to their host galaxies.

Type Iax supernovae (SNe Iax), also called SN 2002cx-like supernovae, are a distinct class of thermonuclear white-dwarf explosions that are observationally related to, but physically and phenomenologically separate from, normal Type Ia supernovae. They constitute the largest known class of peculiar white-dwarf supernovae, with literature estimates placing their occurrence at $31^{+17}_{-13}$ per 100 SNe Ia and at $\sim 15^{+17}_{-9}\%$ of the SN Ia rate (Foley et al., 2012, Magee et al., 2 Jun 2025). The class is defined by low ejecta velocities, subluminous peaks, an offset relation between luminosity and light-curve shape, the absence of the strong near-infrared secondary maximum characteristic of normal SNe Ia, and late-time spectra that never become fully nebular (Foley et al., 2012, Jha, 2017). A leading interpretation is that many SNe Iax arise from pure or weak deflagrations of near-Chandrasekhar-mass white dwarfs, often leaving behind a bound remnant, although sub-Chandrasekhar helium-ignited channels and additional faint-event channels remain under discussion (Jha, 2017, Lach et al., 2021, Wang et al., 2013).

1. Definition and taxonomic status

The original empirical definition required four observational criteria: no evidence for hydrogen in any spectrum, low photospheric velocity near maximum brightness with $|v| \lesssim 8000~{\rm km~s^{-1}}$, low luminosity for the observed light-curve shape, and spectra similar to SN 2002cx at comparable epochs (Foley et al., 2012). This framing was explicitly designed to separate the class from normal SNe Ia without presupposing a unique explosion mechanism. In that initial study the class had 25 members; by 2017, over fifty members were known (Foley et al., 2012, Jha, 2017).

The peak-luminosity range is unusually broad for a thermonuclear class. Reported values extend from roughly $M_V \simeq -19$ to $M_V \simeq -13$, with commonly quoted ranges of $-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$ mag or $-14 \gtrsim M_V \gtrsim -19$ (Jha, 2017, Foley et al., 2012, Lach et al., 2021). Maximum-light velocities likewise span about $2000$ to $8000~{\rm km\,s^{-1}}$, far below the $\sim 10{,}000~{\rm km\,s^{-1}}$ typical of normal SNe Ia (Foley et al., 2012, Jha, 2017). Ejecta masses are correspondingly low, with estimates of about $\sim 15^{+17}_{-9}\%$0–$\sim 15^{+17}_{-9}\%$1, and the class includes extremely low-mass events such as SN 2008ha (Foley et al., 2014).

Taxonomically, SNe Iax are not merely low-luminosity extensions of the SN Ia sequence. They fall below the normal SN Ia width–luminosity relation by $\sim 15^{+17}_{-9}\%$2 to several magnitudes, and although they exhibit an internal width–luminosity trend, that relation is offset and much noisier than in normal SNe Ia (Jha, 2017, Foley et al., 2012). This separation is reinforced by their host-galaxy demographics, direct progenitor constraints, and late-time spectroscopic behavior.

2. Photometric and spectroscopic phenomenology

Near maximum light, SNe Iax are spectroscopically similar to hot, luminous SN Ia subtypes such as SN 1991T-like and SN 1999aa-like events. Their spectra often show weak Si II absorption, prominent Fe III lines, Fe-group and intermediate-mass elements including Si, S, and Ca, and strong near-UV Fe-group blanketing (Jha, 2017). Near-infrared spectroscopy has identified Co II in the $\sim 15^{+17}_{-9}\%$3 and $\sim 15^{+17}_{-9}\%$4 bands in objects such as SN 2010ae, SN 2012Z, and SN 2014ck, which strongly supports a thermonuclear origin tied to radioactive decay products (Jha, 2017). Carbon is also common: the class-defining study reported that 82–100% of pre-maximum spectra show some indication of carbon absorption (Foley et al., 2012).

Their photometric evolution resembles that of SNe Ia only in outline. Rise times are typically fast, about 10–20 days, and decline rates in $\sim 15^{+17}_{-9}\%$5 and $\sim 15^{+17}_{-9}\%$6 are similar to, but usually faster than, those of normal SNe Ia (Jha, 2017). In contrast, red-band decline rates are relatively slow, and SNe Iax do not show the strong second maximum in the red and near-infrared that characterizes normal SNe Ia (Jha, 2017, Foley et al., 2012). Two objects, SN 2004cs and SN 2007J, show strong He I lines, indicating that helium is present in at least part of the class and strengthening the case for helium-rich progenitor systems (Foley et al., 2012).

The most distinctive spectroscopic property emerges at late times. SNe Iax never become fully nebular; even beyond 200 days they retain a continuum, low-velocity permitted features, and narrow P-Cygni-like lines (Jha, 2017, Foley et al., 2016). The late-time sample separates into narrow-P-Cygni-dominated, broad-forbidden-line-dominated, and intermediate objects. Narrow components have $\sim 15^{+17}_{-9}\%$7, whereas broad forbidden components reach $\sim 15^{+17}_{-9}\%$8 (Foley et al., 2016). Broad forbidden-line strength correlates with peak luminosity and maximum-light ejecta velocity, implying that more luminous, higher-velocity SNe Iax tend to show stronger broad $\sim 15^{+17}_{-9}\%$9 and $|v| \lesssim 8000~{\rm km~s^{-1}}$0 emission at late times (Foley et al., 2016).

SN 2014dt provides an especially detailed illustration of the class. Its spectral time series spans +11 to +562 days after maximum light and remains photospheric throughout, with strong permitted Fe, Ca, and Na lines long after normal SNe Ia would be nebular (Camacho-Neves et al., 2023). TARDIS modeling shows the photospheric velocity falling from about $|v| \lesssim 8000~{\rm km~s^{-1}}$1 at +11 days to $|v| \lesssim 8000~{\rm km~s^{-1}}$2 by +64 days, then slowing through the $|v| \lesssim 8000~{\rm km~s^{-1}}$3–$|v| \lesssim 8000~{\rm km~s^{-1}}$4 regime where the spectrum diverges from normal SN Ia behavior (Camacho-Neves et al., 2023). This velocity interval has been interpreted as a possible physical boundary within the ejecta below which SNe Iax and normal SNe Ia differ most strongly.

3. Host environments and delay-time distribution

SNe Iax overwhelmingly occur in late-type, star-forming galaxies, and only one clear SN Iax has been found in an early-type galaxy, SN 2008ge (Jha, 2017, Takaro et al., 2019). Their explosion sites are generally metal-poor relative to normal SNe Ia and resemble the environments of core-collapse supernovae more closely than those of ordinary thermonuclear events. In a spectroscopic host-galaxy study, the D16 metallicity distribution at explosion sites lay mostly in the range $|v| \lesssim 8000~{\rm km~s^{-1}}$5, with a median of about 8.41 dex, and the Ia–Iax metallicity comparison gave a KS-test probability of $|v| \lesssim 8000~{\rm km~s^{-1}}$6 (Lyman et al., 2017).

These environments are not, however, the most intensely star-forming regions within their hosts. Typical local star-formation rates inferred from H$|v| \lesssim 8000~{\rm km~s^{-1}}$7 are $|v| \lesssim 8000~{\rm km~s^{-1}}$8, and the youngest local stellar populations are usually several $|v| \lesssim 8000~{\rm km~s^{-1}}$9 to $M_V \simeq -19$0 years old (Lyman et al., 2017). This age scale is young relative to much of the SN Ia population but older than the $M_V \simeq -19$1 Myr expected for very massive Wolf–Rayet progenitors.

A direct delay-time analysis using deep Hubble Space Telescope imaging of nine nearby SN Iax host galaxies measured stars within 200 pc of the SN positions, compared the local color–magnitude diagrams with MIST isochrones spanning $M_V \simeq -19$2 to $M_V \simeq -19$3 years, and inferred a class delay-time distribution with median and $M_V \simeq -19$4 confidence interval $M_V \simeq -19$5 years (Takaro et al., 2019). The same study found only a 2% probability that at least one object in the sample had a delay time of 10 Myr or less, and SN 2008ge alone yielded a $M_V \simeq -19$6 minimum age of 18 Myr (Takaro et al., 2019). These results strongly disfavor a Wolf–Rayet-like fallback core-collapse channel for the class as a whole.

Taken together, the environmental evidence places SNe Iax in a prompt but not extremely young regime. This is consistent with thermonuclear white-dwarf progenitors in rapidly evolving binaries, especially systems with helium-star donors, and inconsistent with the very short lifetimes expected for massive Wolf–Rayet stars (Lyman et al., 2017, Takaro et al., 2019).

4. Progenitor systems and direct observational constraints

The most important direct progenitor detection is SN 2012Z. Pre-explosion Hubble imaging revealed a luminous blue source at the future SN position with approximately $M_V \simeq -19$7 and $M_V \simeq -19$8, making it the only white-dwarf supernova for which a pre-explosion progenitor system has been detected (Jha, 2017). The source is consistent with a helium-star companion to an exploding white dwarf, and follow-up imaging showed that the source had not disappeared, strengthening the companion interpretation (Jha, 2017). Detailed binary-evolution calculations later found that a non-degenerate He companion to a massive C/O WD $M_V \simeq -19$9 or to a hybrid C/O/Ne WD could explain SN 2012Z-S1, with the hybrid WD + He-star scenario regarded as more favorable (Liu et al., 2015).

Late-time photometry of SN 2012Z complicated the picture in a productive way. At $M_V \simeq -13$0 days after maximum light the source remained brighter than the pre-explosion system, and the pre-explosion-subtracted excess showed an especially strong F555W component (Schwab et al., 1 Apr 2025). Its color disfavors both a light echo and circumstellar interaction, while the decline rate is consistent with energy deposition from $M_V \simeq -13$1Fe but the luminosity is too high for ejecta-only models. This points to a dense bound remnant or a wind driven by that remnant (Schwab et al., 1 Apr 2025).

SN 2008ha provides a second major constraint. Hubble imaging taken 4.1 years after peak brightness detected a point source, S1, at 0.86 ACS pixels or $M_V \simeq -13$2 from the SN position, corresponding to $M_V \simeq -13$3 (Foley et al., 2014). The source has $M_V \simeq -13$4 mag, and the chance-coincidence probability is low but non-zero: 0.90% within $M_V \simeq -13$5 of any detected star, or 0.54% within $M_V \simeq -13$6 for stars at least as bright as S1 (Foley et al., 2014). If associated with SN 2008ha, S1 is either a luminous bound remnant or the companion star; interpreted as a star, it is consistent with a TP-AGB star of initial mass $M_V \simeq -13$7, and the local population age is $M_V \simeq -13$8 Myr (Foley et al., 2014). Its markedly redder color relative to the SN 2012Z progenitor candidate suggests that the class may contain more than one progenitor configuration.

The helium-donor scenario makes specific post-explosion predictions. Three-dimensional simulations of ejecta–companion interaction in a weak-deflagration SN Iax find that only about $M_V \simeq -13$9 of helium is typically stripped from the donor, close to or below observational upper limits of $-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$0 from late-time spectra (Zeng et al., 2020). This makes the non-detection of late helium lines compatible with a WD + He-star channel. Surviving helium-star donors are predicted to brighten to $-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$1–$-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$2 over a Kelvin–Helmholtz timescale of about $-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$3 yr and later evolve as O-type hot subdwarf stars (Zeng et al., 2022). Surviving main-sequence donors, in a different single-degenerate channel, are expected to expand and reach $-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$4–$-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$5 for $-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$6–$-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$7 yr, potentially becoming visible about 1000 days after explosion (Liu et al., 2020).

5. Explosion physics, bound remnants, and late-time structure

A weak or pure deflagration of a near-Chandrasekhar-mass white dwarf is the leading explanation for bright and intermediate-luminosity SNe Iax. In this picture the flame remains subsonic, burning is incomplete, the ejecta are low mass and low velocity, and the star is often not completely disrupted (Jha, 2017, Lach et al., 2021). The N5def model often used as a reference ejects $-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$8, leaves a bound remnant of $-14.2 \gtrsim M_{V,\mathrm{peak}} \gtrsim -18.9$9, has kinetic energy $-14 \gtrsim M_V \gtrsim -19$0 erg, and synthesizes $-14 \gtrsim M_V \gtrsim -19$1 of $-14 \gtrsim M_V \gtrsim -19$2 (Zeng et al., 2020).

A broader 3D parameter study of single-spot deflagrations in Chandrasekhar-mass CO white dwarfs produced $-14 \gtrsim M_V \gtrsim -19$3Ni masses from $-14 \gtrsim M_V \gtrsim -19$4 to $-14 \gtrsim M_V \gtrsim -19$5, ejecta masses from $-14 \gtrsim M_V \gtrsim -19$6 to $-14 \gtrsim M_V \gtrsim -19$7, and bound-remnant kick velocities from $-14 \gtrsim M_V \gtrsim -19$8 to $-14 \gtrsim M_V \gtrsim -19$9 (Lach et al., 2021). The main characteristics of those models are driven primarily by the $2000$0Ni mass, and the scenario remains a favorable explanation for bright and intermediate-luminosity SNe Iax. However, the same study concluded that the faintest members of the class are not fully reproduced by that sequence alone (Lach et al., 2021).

Late-time observations strongly support remnant survival. Stable $2000$1 emission requires high-density burning and therefore favors a near-Chandrasekhar progenitor (Foley et al., 2016). For SN 2005hk, the blackbody photospheric radius at late times is more than an order of magnitude smaller than the kinematic radius inferred from the observed low photospheric velocity, indicating that the late-time photosphere is not simply homologously expanding ejecta (Foley et al., 2016). A two-component model resolves this discrepancy: broad forbidden lines arise in the SN ejecta, whereas the continuum, narrow forbidden lines, and persistent P-Cygni features arise in an optically thick wind launched from a surviving remnant and powered by radioactive decay (Foley et al., 2016). SN 2014dt independently supports this picture, with a nearly constant photospheric radius of $2000$2–50 AU from about +60 to +420 days and modeling consistent with a weak deflagration plus a quasi-steady-state wind (Camacho-Neves et al., 2023).

Radiative-transfer calculations that explicitly add a luminous central remnant strengthen the same interpretation. Including remnant-powered UVOIR radiation in the 3D Monte Carlo code artis slows the post-maximum decline, especially in red optical and near-infrared bands, and improves the spectra of bright, intermediate, and faint SNe Iax (Callan et al., 2024). The full remnant $2000$3Ni mass predicted by hydrodynamic models works for bright events, but intermediate and faint objects require substantially lower effective remnant power, suggesting that some remnant decay energy drives winds or escapes only after a delay (Callan et al., 2024).

SN 2019muj shows that the late-time inner structure can itself be multi-component. Spectra at about 131 and 480 days after explosion imply a distinct Fe-rich inner component with mass $2000$4, characteristic velocity $2000$5, Ca mass $2000$6, and oxygen mass $2000$7 (Maeda et al., 2022). The preferred interpretation is that radioactive heating expels part of the $2000$8Ni-rich material initially confined in the bound remnant, producing a delayed “second” unbound ejecta component (Maeda et al., 2022). This is a direct late-time realization of the broader failed-deflagration framework.

6. Diversity, alternative channels, and broader implications

Despite the coherence of the remnant-deflagration picture, the class is not spectroscopically or structurally trivial. Abundance tomography of five SNe Iax with TARDIS found that iron-group elements and intermediate-mass elements decrease outward, oxygen can dominate the outer ejecta, and carbon is largely confined to the outer layers (Barna et al., 2018). That stratified structure contradicts the well-mixed abundance profiles predicted by standard pure-deflagration models, although a template abundance profile shifted in velocity reproduced the observed spectra nearly as well as object-specific models, implying that the class may still be a “few-parameter family” with a common physical origin (Barna et al., 2018).

Very early observations introduce a complementary perspective. For a sample of 14 SNe Iax with forced photometry from ATLAS, GOTO, and ZTF, the early flux was modeled as

$2000$9

and the class showed systematically lower rise indices than normal SNe Ia (Magee et al., 2 Jun 2025). SN 2025qe, observed within $8000~{\rm km\,s^{-1}}$0 days of first light, displayed carbon absorption throughout its spectroscopic evolution, including very early spectra only 1.8 and 3.0 days after first light (Magee et al., 2 Jun 2025). Those shallow rise indices and persistent carbon are qualitatively consistent with extended $8000~{\rm km\,s^{-1}}$1Ni distributions and strongly mixed ejecta, similar to pure-deflagration expectations (Magee et al., 2 Jun 2025). This suggests that the apparent stratification inferred from abundance tomography and the strong mixing inferred from early light curves do not yet form a fully settled picture.

Alternative progenitor channels remain active in the literature. Binary population-synthesis studies of helium double detonations in sub-Chandrasekhar CO white dwarfs accreting from non-degenerate helium stars predict Galactic rates of $8000~{\rm km\,s^{-1}}$2 or $8000~{\rm km\,s^{-1}}$3 in a standard model, delay times of $8000~{\rm km\,s^{-1}}$4–$8000~{\rm km\,s^{-1}}$5 Myr or $8000~{\rm km\,s^{-1}}$6–$8000~{\rm km\,s^{-1}}$7 Myr, and brightness ranges from about $8000~{\rm km\,s^{-1}}$8 to $8000~{\rm km\,s^{-1}}$9 mag (Wang et al., 2013, Wang et al., 2013). These rates and delay times are compatible with late-type hosts, but the same studies emphasize uncertainties in helium ignition, shell mass, and the ability of current double-detonation models to reproduce the low velocities of SNe Iax (Wang et al., 2013). Environment studies also leave open stripped-envelope electron-capture supernovae for at least part of the faint and fast SN 2008ha-like subset (Lyman et al., 2017).

The class has further astrophysical implications beyond explosion taxonomy. Chemical-kinetic modeling of dust formation indicates that SNe Iax can condense $\sim 10{,}000~{\rm km\,s^{-1}}$0 to $\sim 10{,}000~{\rm km\,s^{-1}}$1 of dust, with most of the mass forming between 1000 and 2000 days after explosion and with especially abundant Fe-rich silicates such as FeSiO$\sim 10{,}000~{\rm km\,s^{-1}}$2, Fe$\sim 10{,}000~{\rm km\,s^{-1}}$3SiO$\sim 10{,}000~{\rm km\,s^{-1}}$4, and MgFeSiO$\sim 10{,}000~{\rm km\,s^{-1}}$5 (Kumar et al., 19 Nov 2025). Their dust-to-ejecta mass ratio is estimated to be one or two orders of magnitude larger than in normal Type Ia supernovae, making SNe Iax plausible producers of iron-rich silicate dust (Kumar et al., 19 Nov 2025). This suggests that their partial, mixed, low-velocity explosions may matter not only for white-dwarf explosion physics but also for the iron-bearing dust budget of galaxies.