Low-Temperature Degradation (LTD)

• Low-temperature degradation is a phenomenon where materials continue to degrade at low temperatures due to mechanisms like charge trapping, moisture-induced phase transformation, and impact ionization.
• Studies across systems such as phase-change memory, ceramics, 2D transistors, diamond nanopillars, and PVC reveal that degradation is governed by defect kinetics, interface effects, and electronic state dynamics rather than simple thermal processes.
• Mitigation strategies for LTD include trap-state engineering, boundary chemistry stabilization, and protective passivation, tailored to the specific material and degradation mechanism.

Low-temperature degradation (LTD) is a domain-dependent term used for degradation phenomena that remain operative under temperatures low relative to the characteristic thermal scale of a material, device, or environment. In the works considered here, LTD encompasses cryogenic resistance drift in amorphous Ge$_2$Sb$_2$Te$_5$ phase-change memory, hydrothermal aging in 3 mol% yttria-stabilized zirconia (YSZ), high-field failure of monolayer MoS$_2$ transistors at 77 K and 150 K, optical degradation of near-surface NV centers in diamond nanopillars at $\sim 6$ K in high vacuum, and sub-glass-transition thermal deterioration of unplasticized PVC at $60\text{--}80^\circ$C (Khan et al., 2019, Garg et al., 27 Aug 2025, Ansh et al., 2020, Kumar et al., 30 Apr 2026, Saad et al., 20 Mar 2025). Across these systems, low temperature does not eliminate degradation; rather, the dominant kinetics are shifted toward trap occupation, vacancy exchange, impact ionization, surface chemistry, or humidity-coupled dehydrochlorination.

1. Terminological scope and shared features

LTD is not a single mechanism. In amorphous GST phase-change memory, it denotes continued resistance drift down to 125 K. In YSZ ceramics, it denotes moisture-induced, irreversible tetragonal-to-monoclinic transformation that begins at the surface and progresses inward. In CVD-grown monolayer MoS$_2$ transistors, it denotes severe material and device degradation under high electric field at low temperature. In diamond nanopillars, it denotes laser-exposure-induced loss of NV optical quality under high vacuum and low temperature. In unplasticized PVC, it denotes thermal degradation below the glass transition, with temperatures chosen at $60^\circ$C, $70^\circ$C, and $80^\circ$C for a material with $_2$0C (Khan et al., 2019, Garg et al., 27 Aug 2025, Ansh et al., 2020, Kumar et al., 30 Apr 2026, Saad et al., 20 Mar 2025).

A common feature is that degradation persists when simple thermal freezing-out would be expected to suppress it. In GST, drift coefficients remain in the $_2$1 range between 125 K and 300 K. In MoS$_2$2, lowering temperature can intensify failure because weak electron-phonon scattering enables hot-electron damage. In YSZ, hydrothermal access to defect-rich boundaries dominates despite moderate temperatures. In NV-diamond systems, low temperature reduces but does not remove laser-induced carbon formation on oxygen-terminated surfaces. In PVC, radical signatures are still detected at $_2$3C. Taken together, these studies suggest that LTD is often governed by defect, interface, or electronic-state kinetics rather than by bulk thermally activated structural relaxation alone.

2. Electronic LTD in amorphous Ge$_2$4Sb$_2$5Te$_2$6 phase-change memory

The GST study examined amorphous phase-change memory line cells fabricated using 90 nm technology over the temperature range 300 K to 125 K. Cells were amorphized by a 500 ns electrical pulse, and resistance was measured using low-voltage sweeps to minimize disturbance. Photoexcitation was introduced with a white LED with peak $_2$7 nm inside a cryogenic chamber, allowing direct comparison of resistance drift in dark and under illumination (Khan et al., 2019).

Resistance drift followed the established power law

$_2$8

with drift coefficients extracted from bi-log plots of $_2$9 versus $_5$0. In darkness, $_5$1 at 125 K and $_5$2 at 300 K, a narrow range comparable to or slightly lower than commonly reported room-temperature values of $_5$3. At 150 K, illumination reduced the drift coefficient from $_5$4 in dark to $_5$5 under illumination, i.e. about a factor of 2 suppression.

The photoresponse had two timescales. LED on/off cycling produced a sudden resistance decrease or increase associated with direct photogeneration of carriers, followed by a much slower convergence toward a stable resistance on the order of minutes, with a response of $_5$6 minutes at 150 K attributed to trap-related processes. Full stabilization could require $_5$7 minutes in the $_5$8 K range. The persistence of drift down to 125 K, together with its suppression under illumination, implicates charge traps within the amorphous phase as the predominant low-temperature mechanism. The paper accordingly argues that resistance drift in amorphous phase-change materials is predominantly an electronic process, not one dominated by thermally activated atomic rearrangement.

This has direct reliability implications. Resistance drift remains a limiting factor for multibit-per-cell PCM because intermediate resistance levels do not stabilize even at cryogenic temperature. The reported photoexcitation response suggests possible stabilization routes based on trap occupation management, including optical or electrical carrier injection.

3. Hydrothermal LTD in yttria-stabilized zirconia ceramics

In 3 mol% YSZ, LTD is described as a moisture-induced, irreversible tetragonal-to-monoclinic ($_5$9) transformation that begins at the material surface and progresses inward, producing microcracking and performance loss. The study combined machine learning-guided hydrothermal aging with multiscale characterization to resolve a two-stage degradation mechanism. Approximately 100 literature data points were compiled to relate aging conditions to monoclinic phase content, eight regression models were trained, and Gaussian Process Regression was selected because it showed $_2$0 with low overfitting based on RMSE. The ML output guided aging conditions such as 132 $_2$1C, several bar pressure, and 0–60 hrs exposure (Garg et al., 27 Aug 2025).

Characterization combined tapping-mode AFM, standard and grazing-incidence XRD, and atom probe tomography. The monoclinic fraction was computed using Toraya’s equation,

$_2$2

while oxygen-vacancy distributions were assessed through the pseudo-coordination number, defined as the number of O neighbors for each Y atom in 8 nearest neighbors. Low PCN corresponded to vacancy-rich regions.

The degradation proceeded in two stages. Stage 1, from 0 to 30 hrs, showed rapid surface relief development: AFM roughness $_2$3 increased from 116 nm to 308 nm at 30 hrs, and grain-size distributions broadened. Stage 2, from 30 to 60 hrs, showed partial refinement and redistribution of the surface relief: $_2$4 dropped to $_2$5 nm at 60 hrs, while monoclinic content rose, especially near the surface, with a clear depth gradient in GI-XRD. The interpretation given is that apparent early coarsening reflects volume expansion and shear at transformation fronts, while later evolution reflects continued subsurface transformation.

The critical mechanistic result concerns boundary chemistry. Triple-junction grain boundaries (TJGBs) were identified as degradation hotspots because they combine high energy, chemical disorder, and high connectivity for water ingress. Vacancy percentages at TJGBs fell from 38% in the unaged state to $_2$6 at 30 and 60 hrs, indicating preferential vacancy annihilation. By contrast, Al/Si-rich boundaries retained vacancies in the 31–39% range. Grain-boundary thickness also evolved, with Y-rich grain boundaries decreasing from 5.3 nm to 3.7 nm with aging. The paper therefore argues that grain boundary chemistry, rather than grain size alone, is the key LTD driver in this regime, and recommends boundary engineering, reduced TJGB density, and stabilizing dopants such as Al/Si as mitigation strategies.

4. High-field LTD in CVD-grown monolayer MoS$_2$7 transistors

For CVD-grown monolayer MoS$_2$8 transistors, LTD refers to temporal material and performance degradation under electrical stress, particularly under high electric field at low temperature. The study distinguishes a low-field regime, $_2$9 MV/cm, from a high-field regime, $\sim 6$0 MV/cm, and reports that LTD is most pronounced in the high-field regime at 77 K and 150 K (Ansh et al., 2020).

In the low-field regime, prolonged stress caused a monotonic decrease in current until thermal equilibrium was established, with kinetics described as

$\sim 6$1

At 300 K, current decayed over $\sim 6$2 s. The temperature dependence was linked to phonon populations,

$\sim 6$3

so that higher temperature increased phonon scattering and altered the decay time constant.

The high-field regime behaved differently. At low temperature, suppressed electron-phonon scattering allowed electrons to gain higher energies over longer distances, and impact ionization became significant. The reported signature was an abrupt current increase followed by catastrophic collapse after a short period of operation of $\sim 6$4 s, with current dropping by orders of magnitude to an extremely low pA range. This failure mode was attributed to lattice damage and amorphization of the channel.

Post-stress electrical signatures included a dramatic 3–4 orders of magnitude increase in OFF-state current, degraded gate control, negative threshold-voltage shift, and ON-state current degradation. Under some conditions the channel became highly conductive and metal-like, while at low temperature complete loss of transistor action could occur. Microscopy and spectroscopy supported this interpretation: KPFM no longer resolved grain boundaries after high-field low-temperature stress, Raman peaks were suppressed, broadened, or lost, PL disappeared at some spots with dramatic FWHM broadening, and SEM showed a continuous channel, implying amorphization rather than macroscopic fracture. The paper also identifies S-vacancy migration as contributing to localized low-resistance regions and gate-control loss. A central implication is that low-temperature operation is not intrinsically protective for 2D FETs; safe operating regimes must explicitly exclude prolonged high-field low-temperature stress.

5. Optical and surface-chemical LTD in diamond nanopillar NV centers

In diamond nanopillars containing near-surface NV centers, LTD refers to degradation under optical illumination in vacuum, especially at low temperature. The study used electronic-grade diamond chips implanted with 6 keV $\sim 6$5N$\sim 6$6 ions to form NV centers at $\sim 6$7 nm depth, followed by nanopillar fabrication with 300 nm diameter and 1 $\sim 6$8m height, with an average of $\sim 6$9 NV per pillar. Oxygen-terminated nanopillars were compared with structures coated by $60\text{--}80^\circ$0 nm amorphous alumina deposited by ALD. Measurements were performed in high vacuum down to the $60\text{--}80^\circ$1 mbar range and at $60\text{--}80^\circ$2 K under 522 nm continuous-wave laser exposure (Kumar et al., 30 Apr 2026).

The key metric for optical quality was the single-photon purity $60\text{--}80^\circ$3. In oxygen-terminated nanopillars, laser exposure in vacuum caused substantial reduction in single-photon purity and increased fluorescence background. At room temperature in vacuum, 0.25 mW exposure produced an average reduction in $60\text{--}80^\circ$4 across five nanopillars within $60\text{--}80^\circ$5 hours. At 6 K and 1.8 mW, all NV centers in oxygen-terminated pillars lost single-photon purity, with no remaining $60\text{--}80^\circ$6, together with increased background and decreased brightness. Degradation was linear with laser power, and stronger at low temperature when normalized per unit power, although overall laser-induced carbon formation was lower at 6 K than at room temperature.

The spectroscopic evidence pointed to two coupled mechanisms. First, laser exposure in vacuum produced non-diamond carbon features in Raman spectra, denoted laser-induced carbon (LIC), with increased broad fluorescence between 550 and 900 nm. Second, PL spectra showed a decrease in the NV$60\text{--}80^\circ$7 zero-phonon line and increased NV$60\text{--}80^\circ$8 zero-phonon line, indicating charge-state conversion associated with surface modification or reconstruction. The LIC contribution could be largely reversed by subsequent laser exposure in air, consistent with an oxygen-dependent surface process.

Alumina passivation strongly altered the LTD response. Under identical exposure conditions, alumina-coated nanopillars showed no significant degradation of single-photon purity or extra PL background at or below PL saturation power, including 1.8 mW at 6 K. At high power some gradual degradation could still appear, but much more slowly than for oxygen termination. After exposure at 1.8 mW and 6 K, the fraction of nanopillars remaining single-photon emitters was 0% for oxygen-terminated pillars and 60% ($60\text{--}80^\circ$9) for alumina-capped pillars. The coating left $_2$0 unaffected and improved $_2$1, likely by reducing surface noise. The proposed mechanism is that conformal alumina physically blocks adsorbate attachment and polymerization and suppresses surface reconstruction, thereby stabilizing the NV$_2$2 charge state and preserving single-photon emission in harsh cryogenic-vacuum environments.

6. Sub-glass-transition LTD in unplasticized PVC

In unplasticized PVC, LTD was studied as accelerated thermal degradation below the glass transition. The material was a rigid, transparent, 0.2 mm PVC foil stabilized with organotin, with Irganox 1076 used as stabilizer and no detectable plasticizers. Aging was performed at $_2$3C, $_2$4C, and $_2$5C, with low-RH and high-RH regimes and durations up to 22 weeks; the chosen temperatures are below the reported $_2$6C (Saad et al., 20 Mar 2025).

The principal chemical reaction was dehydrochlorination,

$_2$7

This generated conjugated polyene sequences responsible for yellowing. FTIR and both $_2$8H and $_2$9C NMR did not show significant changes up to 20 weeks at $60^\circ$0C, indicating that the concentration of new moieties remained low. By contrast, Raman spectroscopy under resonant 514 nm excitation and UV-Vis spectroscopy were sensitive to early degradation. Polyene bands at $60^\circ$1 cm$60^\circ$2 and $60^\circ$3 cm$60^\circ$4 appeared after $60^\circ$5 weeks at $60^\circ$6C and $60^\circ$7 weeks at $60^\circ$8C, especially at higher RH. UV-Vis showed a bathochromic shift from pristine absorption at $60^\circ$9 nm to an absorption maximum at 419 nm, corresponding to $70^\circ$0 conjugated C=C units, and up to $70^\circ$1 units at the highest exposure. The summary further states that only 0.003–0.05% of bonds in the visible range, with $70^\circ$2 mass loss, are sufficient for perceptible color change.

Mechanistically, the study contests a simple ionic-only picture for low-temperature PVC degradation. EPR detected weak to moderate organic radical signals with $70^\circ$3 after $70^\circ$4 weeks at $70^\circ$5C and in samples degraded at $70^\circ$6C under 60% RH, showing that a radical pathway cannot be excluded even at these temperatures. XPS at $70^\circ$7C and 60% RH showed a significant decrease in organically bound Cl, consistent with HCl elimination and polyene formation; ionic Cl was also detected under low RH. High RH accelerated degradation, plausibly by increasing HCl mobility and matrix polarity.

Despite the chemical changes and strong yellowing, the macroscopic mechanical response remained modest. SEC showed no significant change in molecular weight or distribution, and DMA found only a very slight increase in storage modulus, with a statistically insignificant effect at lower temperatures or lower RH. The paper therefore places the earliest and most sensitive manifestation of LTD in the optical domain rather than in mechanical embrittlement: aesthetic degradation precedes measurable structural failure.

7. Comparative mechanisms, misconceptions, and mitigation

The assembled literature shows that LTD is best treated as a family of low-temperature or moderate-temperature degradation pathways rather than as a single materials concept. In GST, the operative variables are charge traps and carrier occupation. In YSZ, water access, oxygen-vacancy annihilation, and grain-boundary chemistry dominate. In MoS$70^\circ$8, weak phonon scattering at low temperature promotes hot-electron impact ionization and amorphization. In diamond nanopillars, vacuum-enabled surface chemistry and charge-state instability degrade NV optical performance. In PVC, humidity-assisted dehydrochlorination and radical formation drive yellowing and chemical change (Khan et al., 2019, Garg et al., 27 Aug 2025, Ansh et al., 2020, Kumar et al., 30 Apr 2026, Saad et al., 20 Mar 2025).

Several recurring misconceptions are directly challenged. Cryogenic operation does not necessarily freeze degradation: GST resistance drift continues to 125 K, and MoS$70^\circ$9 degradation is most severe in the high-field low-temperature regime. Grain size alone is not a sufficient LTD descriptor for YSZ, because the cited study identifies TJGB chemistry and connectivity as the dominant susceptibility factor. Reduced LIC formation at 6 K in diamond does not imply stable NV operation without passivation, since oxygen-terminated nanopillars still lose single-photon purity under high-intensity illumination. In PVC, the conventional expectation that radical pathways require higher temperatures is weakened by direct EPR evidence of radicals at $80^\circ$0C.

The mitigation strategies proposed in these works are correspondingly system-specific. For GST, the data suggest trap-state engineering and photoexcitation or carrier-injection-based preconditioning. For YSZ, the recommended path is boundary engineering: reducing TJGB density and stabilizing boundary chemistry, including Al/Si segregation effects. For MoS$80^\circ$1, reliability depends on restricting electric field, gate bias, and stress history, together with grain-boundary, passivation, and electrostatic design. For NV-diamond probes, thin ALD alumina is presented as a viable passivation route for stable cryogenic-vacuum operation. For PVC, low temperature and low RH remain the central preservation variables, while colorimetry, UV-Vis, resonance Raman, and EPR provide a sensitive diagnostic hierarchy. A plausible implication is that LTD mitigation is most effective when it targets the active defect population and interface chemistry of a given system, rather than temperature alone.