---
title: Partial Lyman Limit Systems (pLLSs)
url: https://www.emergentmind.com/topics/partial-lyman-limit-systems-pllss
type: topic
---

# Partial Lyman Limit Systems (pLLSs)

Partial Lyman Limit Systems (pLLSs) are quasar absorption-line systems that occupy the neutral-hydrogen column-density interval immediately below the classical Lyman-limit threshold, generally with $16.0 \lesssim \log N_{\mathrm{H\,I}} < 17.2$, while classical Lyman Limit Systems (LLSs) occupy $17.2 \le \log N_{\mathrm{H\,I}} < 19.0$; some low-redshift surveys adopt $16.1$ or $16.2$ as the lower pLLS boundary, but the physical role is the same across these definitions [1612.00458]. They are partially optically thick at the Lyman limit, with $\tau_{\mathrm{LL}}$ somewhat below unity, and they lie at the interface between the highly ionized Ly$\alpha$ forest and the denser gas traced by sub-damped Ly$\alpha$ systems and damped Ly$\alpha$ absorbers. Because they are selected by H I opacity yet require substantial ionization corrections, pLLSs have become a principal observational tracer of the cool, dense, predominantly ionized circumgalactic medium (CGM), of dense intergalactic structures, and of the metal-poor gas reservoirs that are widely interpreted as evidence for cold accretion onto galaxies [1612.00458].

## 1. Definition and placement within the H I absorber hierarchy

In the literature summarized here, pLLSs are defined by their neutral hydrogen column density rather than by prior metallicity or galaxy information. Following Lehner et al. (2013), pLLSs have $10^{16}\,\mathrm{cm}^{-2} < N_{\mathrm{H\,I}} < 10^{17.2}\,\mathrm{cm}^{-2}$, while LLSs span $10^{17.2}\,\mathrm{cm}^{-2} \le N_{\mathrm{H\,I}} < 10^{19}\,\mathrm{cm}^{-2}$; Wotta et al. adopt $10^{16.1} \le N(\mathrm{H\,I}) < 10^{17.2}\,\mathrm{cm}^{-2}$ for low-$z$ pLLSs, and the COS CGM Compendium uses $16.2 < \log N(\mathrm{H\,I}) < 17.2$ [1612.00458; 1608.02584; 1811.10654]. In practice, pLLSs are “partially” optically thick, with $\tau_{\mathrm{LL}}$ below unity, whereas LLSs satisfy $\tau_{\mathrm{LL}} \ge 1$ [1612.00458].

On a cosmological density scale, their $N_{\mathrm{H\,I}}$ values and implied total hydrogen densities place them between the diffuse intergalactic medium (IGM) and the denser, more neutral absorbers at higher column density. The review literature describes pLLSs and LLSs as occupying the interface between the Ly$\alpha$ forest and the virialized gas traced by SLLSs and DLAs, with implied densities $\log n_{\mathrm{H}} \approx -4$ to $-1\,\mathrm{cm}^{-3}$ via the Schaye (2001) prescription [1612.00458]. At low redshift, pLLSs probe cool gas with $T \sim 10^4\,\mathrm{K}$, $n_{\mathrm{H}} \sim 10^{-2.5}\,\mathrm{cm}^{-3}$, and baryonic overdensities $\delta_m \sim 10^2$–$10^3$, characteristic of galaxy halos [1608.02584].

This intermediate position in column density is astrophysically consequential. Higher-$N_{\mathrm{H\,I}}$ absorbers at $z<1$, especially SLLSs and DLAs, show mostly $[\mathrm{X}/\mathrm{H}] > -1.4$ and unimodal metallicity distributions peaked around $-0.5$ to $0.0$, whereas pLLSs and LLSs probe a distinct regime in which low-metallicity and metal-rich gas coexist over a broad dynamic range [2204.13229].

## 2. Observational identification and measurement of $N_{\mathrm{H\,I}}$

Large, unbiased pLLS samples became feasible through two observational developments. At low redshift, the installation of COS on HST enabled direct measurements of $\tau_{\mathrm{LL}}$ in G140L mode and metal-line spectroscopy with the higher-resolution G130M/G160M gratings; at high redshift, Keck/HIRES and VLT/UVES archives, together with surveys such as KODIAQ and HD-LLS, provided high-S/N spectra of the full Lyman series and weak metal lines [1612.00458]. These instrumental developments turned pLLSs from isolated case studies into statistically tractable absorber classes.

Two complementary methods are used to determine $N_{\mathrm{H\,I}}$. One is direct measurement of the Lyman-limit optical depth,
$$
\tau_{\mathrm{LL}} = N_{\mathrm{H\,I}}\,\sigma_{\mathrm{LL}},
$$
with $\sigma_{\mathrm{LL}} \approx 6.3 \times 10^{-18}\,\mathrm{cm}^2$ at $912\,\text{\AA}$; the other is curve-of-growth or Voigt-profile fitting of higher-order Lyman-series lines [1608.02584; 1710.03232]. In the HST/COS ultraviolet survey of 102 AGN sight lines, pLLSs were recognized through the “Lyman-comb” of higher-order Lyman lines and, when $\log N_{\mathrm{H\,I}} \gtrsim 16.2$, by the continuum decrement at the redshifted Lyman edge. Combining line-series fitting and Lyman-limit measurements yielded column densities accurate to $\Delta(\log N_{\mathrm{H\,I}}) \lesssim 0.05\,\mathrm{dex}$ over $16.2 \le \log N_{\mathrm{H\,I}} \le 17.9$ [1710.03232].

Incidence measurements place pLLSs among the common strong H I absorbers at low redshift. Over a total pathlength $\Delta z = 31.94$ at $0.24 \le z \le 0.84$, 54 pLLSs yielded $\langle dN/dz \rangle_{\mathrm{pLLS}} = 1.69 \pm 0.23$ at $\langle z \rangle \approx 0.39$, while the maximum-likelihood column-density distribution was fitted as
$$
f(N,z) = C_0 N^{-\beta}(1+z)^\gamma
$$
with $\beta = 1.48 \pm 0.05$ and $\gamma = 1.14^{+0.88}_{-0.89}$ for $15 \le \log N_{\mathrm{H\,I}} \le 17.5$ [1710.03232]. Interpreting this incidence in terms of halo cross-sections gives characteristic gaseous halo radii of $R \approx 100$–$150\,h^{-1}\,\mathrm{kpc}$ for $0.3$–$3L^\ast$ galaxies, consistent with direct CGM measurements [1710.03232].

Beyond absorber statistics, pLLSs are also significant contributors to far-UV opacity. Integrating over $15 \le \log N_{\mathrm{H\,I}} \le 17$ yields
$$
d\tau_{\mathrm{eff}}/dz \approx 0.444(1+z)^{1.14},
$$
implying mean LyC optical depths $\tau_{\mathrm{eff}} \approx 0.3$–$0.5$ toward sources at $z = 1$–$2$ [1710.03232].

## 3. Ionization modeling and abundance inference

pLLSs are highly ionized systems, with $\mathrm{H\,II} \gtrsim 90\%$ in low-redshift samples, so metallicity determinations require explicit ionization corrections [2204.13229]. The standard metallicity definition is
$$
[\mathrm{X}/\mathrm{H}] \equiv \log_{10}(X/H) - \log_{10}(X/H)_\odot,
$$
and the relevant control parameter in photoionization modeling is
$$
U \equiv \frac{n_\gamma}{n_{\mathrm{H}}},
$$
or equivalently $U = \Phi/(c\,n_{\mathrm{H}})$ in the COS CGM Compendium formalism [2204.13229; 1811.10654].

Most modern analyses use Cloudy grids for a uniform plane-parallel slab photoionized by an extragalactic UV background. The review literature summarizes modeling with Haardt–Madau UV backgrounds, including HM05 and HM12; BASIC used Cloudy plane-parallel slabs with the HM96 background and Bayesian MCMC; CCC employed Bayesian inference with emcee over grids in $\log N(\mathrm{H\,I})$, redshift, $[\mathrm{X}/\mathrm{H}]$, $\log n_{\mathrm{H}}$, and $[\mathrm{C}/\alpha]$ [1612.00458; 2204.13229; 1811.10654]. Where ionic constraints are sparse, CCC imposed Gaussian priors on $\log U$ and, in carbon-only cases, on $[\mathrm{C}/\alpha]$ [1811.10654].

Low-resolution abundance methods have also been developed for large samples. Wotta et al. used an empirically derived $\langle \log U \rangle = -3.1 \pm 0.6$ distribution together with Cloudy v13.03 models to compute ionization-correction factors for Mg II and showed that this approach reproduces metallicities to $\lesssim 0.3\,\mathrm{dex}$ when tested against 22 previously modeled pLLSs/LLSs [1608.02584]. In BASIC, typical statistical errors on $[\mathrm{X}/\mathrm{H}]$ were $\sim 0.1$–$0.2\,\mathrm{dex}$, while systematic uncertainties of $\sim 0.2$–$0.4\,\mathrm{dex}$ associated with the UV background shape were explicitly recognized [2204.13229].

The inference machinery is not merely procedural; it controls the physical interpretation of pLLSs. By coupling Cloudy-derived $U$ and $[\mathrm{X}/\mathrm{H}]$ with observed $N_{\mathrm{H\,I}}$, one can infer absorber sizes and total hydrogen columns through $N_{\mathrm{H}} = N_{\mathrm{H\,I}}/f_{\mathrm{H\,I}}(U)$ [1612.00458]. A central result of subsequent high-S/N work is that single-phase solutions can be misleading: pLLSs can contain kinematically aligned low- and high-ionization gas that would masquerade as a single cloud if only integrated columns were modeled [2110.13933].

## 4. Metallicity distributions and their evolution with redshift

The most prominent empirical result for pLLSs is the metallicity distribution function (MDF). At $z<1$, low-redshift H I-selected samples established a strongly bimodal MDF for pLLSs. In Wotta et al., the combined sample of 44 pLLSs showed a low-metallicity peak at $\langle [\mathrm{X}/\mathrm{H}] \rangle = -1.87 \pm 0.11$ and a high-metallicity peak at $\langle [\mathrm{X}/\mathrm{H}] \rangle = -0.32 \pm 0.07$, with a trough near $[\mathrm{X}/\mathrm{H}] \approx -1$; the low-metallicity peak comprised $57\%$ of the pLLSs [1608.02584]. CCC, using 82 pLLSs at $z<1$, confirmed this pattern with peaks at $[\mathrm{X}/\mathrm{H}]_{\mathrm{low}} = -1.72 \pm 0.09$ and $[\mathrm{X}/\mathrm{H}]_{\mathrm{high}} = -0.38 \pm 0.06$, a deep dip near $[\mathrm{X}/\mathrm{H}] \simeq -1.0$, and a very metal-poor fraction $[\mathrm{X}/\mathrm{H}] < -1.4$ of $41$–$59\%$ [1811.10654].

The statistical evidence for bimodality is strong. In the low-redshift samples, Gaussian-mixture modeling rejects unimodality at $>99.9\%$ and the dip test at $\gtrsim95\%$ in Wotta et al.; the review summarizes GMM and dip-test rejections of unimodality at $>99\%$ confidence for a sample of $\sim 50$ pLLSs+LLSs [1608.02584; 1612.00458]. BASIC recast the low-$z$ bimodality in a simpler two-branch form, with a low-metallicity branch at $[\mathrm{X}/\mathrm{H}] \le -1.4$ and a high-metallicity branch at $[\mathrm{X}/\mathrm{H}] \gtrsim -1.4$, containing approximately $40$–$50\%$ and $\sim 50$–$60\%$ of the pLLSs, respectively [2204.13229].

At high redshift, the MDF changes qualitatively. In the KODIAQ Z sample of 31 pLLSs+LLSs at $2.3<z<3.3$, the metallicity distribution is unimodal, with survival-analysis mean $\langle [\mathrm{X}/\mathrm{H}] \rangle = -2.00 \pm 0.17$, dispersion $\sigma \approx 0.8\,\mathrm{dex}$, median $-2.10$, and range from $-3.35$ to $+0.20$ [1608.02588]. The review describes a similarly selected sample at $z \approx 2.3$–$3.3$ as broad and unimodal, centered at $\langle [\mathrm{X}/\mathrm{H}] \rangle \approx -2.0$, with little evidence for a dip and with the bulk of systems in the interval $[-2.5,-1.0]$ [1612.00458]. Very metal-poor gas remains common: $25$–$41\%$ of the high-$z$ pLLS+LLS sample has $[\mathrm{X}/\mathrm{H}] \le -2.4$, while the fraction with $[\mathrm{X}/\mathrm{H}] < -3$ is only $3$–$18\%$, implying that the gas is rarely pristine [1608.02588].

Relative to higher-column absorbers, pLLSs remain distinctive. At $z<1$, SLLSs and DLAs have metallicities largely confined to $-1 \lesssim [\mathrm{X}/\mathrm{H}] \lesssim 0$ and contain almost no absorbers below $[\mathrm{X}/\mathrm{H}] < -1.4$; even LLSs show a lower fraction of very metal-poor systems than pLLSs [1608.02584; 1811.10654]. This column-density dependence is one of the main reasons pLLSs are regarded as especially sensitive tracers of metal-poor halo and interface gas [2204.13229].

## 5. Galaxy associations, CGM locations, and group-scale environments

Low-redshift pLLSs were initially established as CGM absorbers through absorber-galaxy associations at modest impact parameter. A widely cited example is the $z \approx 0.3$ pLLS toward PG 1630+377, which has $[\mathrm{X}/\mathrm{H}] \approx -1.7$, lies only $37\,\mathrm{kpc}$ in projection from a $0.3L_\ast$ galaxy with near-solar ISM abundance, and has inferred physical conditions $n_{\mathrm{H}} \sim 10^{-3.5}\,\mathrm{cm}^{-3}$, $T \sim 2\times 10^4\,\mathrm{K}$, and $N_{\mathrm{H}} \approx 10^{19.5}\,\mathrm{cm}^{-2}$ [1612.00458]. The review further notes that most $z<1$ pLLSs have impact parameters $\rho \approx 100$–$200\,\mathrm{kpc}$ and that their covering fraction within $R_{\mathrm{vir}} \sim 200\,\mathrm{kpc}$ is $f_c \sim 50$–$70\%$ [1612.00458].

Subsequent galaxy-survey work complicated the earlier picture without overturning its CGM relevance. BASIC targeted 36 H I-selected pLLSs/LLSs at $z \lesssim 1$ in 23 QSO fields and, in 11 IFU fields containing 19 pLLSs, identified 23 distinct associated galaxies, with an average of one associated galaxy per absorber [2204.13229]. The adopted association criteria were $\rho/R_{\mathrm{vir}} \le 1.5$ and $|\Delta v|/v_{\mathrm{esc}} \le 1.5$ [2204.13229]. Yet seven absorbers, all with $<10\%$ solar metallicities, had no associated galaxies with $\log M_\star \gtrsim 9.0$ within those cuts. The probability that a metal-poor system has a host galaxy with $\log M_\star \ge 9.0$ within $\rho/R_{\mathrm{vir}} \le 1.5$ is $0.39^{+0.16}_{-0.15}$, compared with $0.78^{+0.10}_{-0.13}$ for higher-metallicity absorbers [2204.13229].

BASIC therefore argued for a dual low-metallicity population. Approximately $40\%$ of the metal-poor pLLSs appear associated with galaxy CGM, likely representing cold accretion onto $\gtrsim 10^9\,M_\odot$ halos, while another population with no detected host within $R_{\mathrm{vir}}$ or $\log M_\star \gtrsim 9$ probably traces dense IGM or filamentary gas with overdensities $\delta \sim 10$–$10^2$ [2204.13229]. Using absorbers without identified galaxies, BASIC estimated an unweighted geometric-mean metallicity of the overdense IGM at $z<1$ of $\langle [\mathrm{X}/\mathrm{H}] \rangle_{\mathrm{geom}} \lesssim -2.1$ [2204.13229].

pLLSs also arise in richer environments. The $z=0.83718$ pLLS toward HE1003+0149 has $\log N(\mathrm{H\,I}) = 16.87 \pm 0.10$ from the Lyman-limit break and resides in a field where MUSE found three dwarf galaxies with stellar masses $M_\ast \sim 10^8$–$10^9\,M_\odot$ at $\rho/R_{\mathrm{vir}} \approx 1.8$–$3.0$, while wider spectroscopy identified 21 additional galaxies within projected separation $\le 5\,\mathrm{Mpc}$ and $|\Delta v| \le 1000\,\mathrm{km\,s}^{-1}$, consistent with a loose galaxy group [2105.04841]. The absorber was interpreted as tracing a multiphase, cool, photoionized intragroup medium [2105.04841].

## 6. Multiphase structure, O VI, and chemical inhomogeneity

Detailed component-level analyses show that pLLSs are not generally single-phase clouds. In CUBS, two pLLSs at $z=0.4723$ and $z=0.5413$ with $N_{\mathrm{H\,I}} \approx (1$–$3)\times 10^{16}\,\mathrm{cm}^{-2}$ were resolved into multiple kinematic components spanning five ionization stages, from Mg II, C II, O II, and Si II through C III, Si III, N III, O III, O IV, N IV, and O VI [2110.13933]. Their inferred densities span more than two decades, from $\log n_{\mathrm{H}}/\mathrm{cm}^{-3} \approx -2.2$ to $<-4.3$, while $\alpha$-element abundances vary from $[\alpha/\mathrm{H}] \approx -0.8$ to near solar [2110.13933].

The distinctive point is that velocity alignment does not imply phase uniformity. In the CUBS systems, single-phase photoionization models could not reproduce the observed O II, O III, and O IV ratios, and two-phase decompositions were required even within individual velocity components [2110.13933]. If unresolved, these multiphase blends would masquerade as single clouds with ambiguous densities, metallicities, and sizes. The inferred environments were correspondingly diverse: one absorber lies $55\,\mathrm{kpc}$ from a star-forming galaxy of $M_\star \approx 2\times10^{10}\,M_\odot$, whereas the other lies in an overdense environment of 11 galaxies within $750\,\mathrm{kpc}$, including a luminous red galaxy of $M_\star \approx 2\times10^{11}\,M_\odot$ at $375\,\mathrm{kpc}$ [2110.13933].

O VI provides a systematic diagnostic of this multiphase structure. In CCC V, the robust pLLS sample at $0.14 \lesssim z \lesssim 0.73$ contained 26 pLLSs, with a $2\sigma$ O VI completeness limit of $\log N(\mathrm{O\,VI}) \simeq 13.6$ [2403.02374]. The total O VI detection rate was $16/26 \rightarrow 61 \pm 9\%$, with nearly identical frequencies for low-metallicity and high-metallicity pLLSs: $5/8 \rightarrow 61^{+15}_{-16}\%$ for $[\mathrm{X}/\mathrm{H}] \le -1.4$ and $11/18 \rightarrow 60 \pm 11\%$ for $[\mathrm{X}/\mathrm{H}] > -1.4$ [2403.02374]. Systems without O VI are consistent with a single cool photoionized phase, whereas systems with O VI require a distinct lower-density or collisionally ionized phase at $T \sim 10^5$–$10^6\,\mathrm{K}$ because single-phase models underpredict $N(\mathrm{O\,VI})$ by factors $\gtrsim 10$ [2403.02374].

The O VI phenomenology is itself metallicity-dependent. In detected pLLSs, the mean $\log N(\mathrm{O\,VI})$ is $13.67 \pm 0.03$ for low-metallicity systems and $14.20 \pm 0.02$ for high-metallicity systems; the strongest absorbers with $\log N(\mathrm{O\,VI}) \gtrsim 14$ are nearly always associated with high-metallicity pLLSs and with broad O VI profiles characteristic of the CGM of star-forming galaxies [2403.02374]. By contrast, weak O VI occurs over the full metallicity range and is interpreted as tracing extended CGM or IGM, while pLLSs without O VI are likely to originate in the IGM [2403.02374].

A conceptually similar multiphase decomposition was required for the $z \approx 0.8$ intragroup pLLS toward HE1003+0149. There, low-ionization models reproduced O II and O III but underproduced O IV and O V in all three velocity components, so a separate higher-ionization phase was necessary, with predicted H I columns $1$–$3$ dex lower than those of the low-ionization phase [2105.04841].

## 7. Physical interpretation, simulations, and direct size constraints

The dominant interpretation of metal-poor pLLSs is that they trace gas accretion, although the precise location can vary between CGM halos, dense IGM structures, and group environments. The review literature states that the low metallicity of cold gaseous streams is the key discriminator relative to metal-rich outflows or recycled gas and argues that metal-poor pLLSs/LLSs are among the best observational evidence for cold, metal-poor gas accretion onto galaxies [1612.00458]. At $z>2$, simulations predict cold streams with $T < 10^{5.5}\,\mathrm{K}$, metallicities $[\mathrm{X}/\mathrm{H}] \sim -2$ to $-1.5$, densities $\log n_{\mathrm{H}} \approx -4$ to $-2\,\mathrm{cm}^{-3}$, and ionization parameters $\log U \sim -3$ to $-2$, often yielding $N_{\mathrm{H\,I}}$ in the pLLS/LLS range [1612.00458].

The mass budget implied by the low-redshift pLLS population is non-negligible. Using characteristic sizes $L \approx N_{\mathrm{H}}/n_{\mathrm{H}}$ of a few to $10\,\mathrm{kpc}$, the review estimates gas masses per absorber of $M_{\mathrm{gas}} \sim 10^7$–$10^8\,M_\odot$ and an aggregate cool CGM reservoir of $M_{\mathrm{CGM,cool}} \sim 10^{10}$–$10^{11}\,M_\odot$ after integrating over the observed $dN/dz \sim 1$ at $z<1$ [1612.00458]. For fiducial inflow conditions, a continuous infall rate of order $\dot{M}_{\mathrm{in}} \simeq 1$–$10\,M_\odot\,\mathrm{yr}^{-1}$ may be sufficient to fuel ongoing star formation in $L_\ast$ galaxies over Gyr timescales [1612.00458].

Comparisons with simulations remain mixed. CCC found that FIRE zoom simulations predict a nearly unimodal, relatively metal-rich PDF and strongly underproduce absorbers with $[\mathrm{X}/\mathrm{H}] < -1.5$, while EAGLE HiRes yields a broader PDF with more low-metallicity gas but a stronger redshift evolution than observed and similar pLLS/LLS PDF shapes not seen in the data [1811.10654]. The dedicated EAGLE metallicity study likewise reported broad, continuous pLLS metallicity distributions and explicitly did not recover the observed trough near $0.1\,Z_\odot$; at $z\sim0$–$1$, EAGLE finds $63\% \pm 5\%$ of pLLSs below $0.1\,Z_\odot$ and $37\% \pm 5\%$ above, in broad agreement with the observed overall fraction above and below $0.1\,Z_\odot$ but not with the strong low-$z$ bimodality [1712.03988]. This suggests that current feedback and CGM-resolution prescriptions may overmix metals and underresolve low-metallicity structures [1811.10654].

Recent tomography has added direct geometrical information that was previously inferred only from ionization modeling. Using the multiply imaged Sunburst Arc at $z \approx 2.37$, HST/WFC3 UVIS G280 spectra identified two pLLSs at $z=2.18916$ and $z=2.15420$ with $\log N_{\mathrm{H\,I}} = 16.59 \pm 0.19$ and $16.66 \pm 0.29$, respectively, and found consistent H I column densities across $\lesssim 2\,\mathrm{kpc}$ [2507.06443]. The reported de-lensed cloud-length measurements are $R_{\mathrm{de-lensed}} \lesssim 1.3\,\mathrm{kpc}$ and $\lesssim 1.7\,\mathrm{kpc}$, with average H I masses of $\approx 10^3\,M_\odot$ and a total range of $10^2$–$10^4\,M_\odot$ [2507.06443]. Two of the absorbers show strong C IV and are therefore likely in the CGM of foreground galaxies, while the metal-free system is most consistent with an IGM location [2507.06443].

Taken together, the observational record defines pLLSs as a transition population in column density but not as a single physical class. They include metal-rich recycled or outflowing halo gas, very metal-poor CGM clouds consistent with cold accretion, dense IGM or filamentary structures without luminous hosts, and cool intragroup media. Their low-redshift metallicity bimodality, high-redshift unimodal metal-poor distribution, frequent multiphase structure, and increasingly direct environmental constraints make them one of the most stringent empirical probes of baryon cycling across the IGM–CGM interface [1612.00458].

Source: https://www.emergentmind.com/topics/partial-lyman-limit-systems-pllss