Plastic semiconductors are thin films of long conjugated polymer chains. They conduct better when the chains are well ordered, and one reliable way to order them is to cool the film slowly through liquid-crystalline phases. These are intermediate states that are fluid but already partly aligned, so the chains can slide into place before the film solidifies. Sabury et al. showed that a polymer cooled this way gives transistors with an unusually narrow spread of performance. The narrowing has always been credited to better molecular packing. There is a second explanation that nobody has measured. The film sits on a rigid substrate that shrinks by a different amount on cooling, so it is stretched unevenly once it solidifies. In crystals, uneven strain spreads out the local electronic energy levels and blurs spectra, an effect called inhomogeneous broadening. Franke et al. showed this directly for donor spins in silicon clamped to sapphire.
This note proposes to build a quantitative "strain budget" for these films. The plan has three steps. First, measure how strongly a film's optical transition energy shifts per unit strain. Second, measure how widely strain varies inside the film. Third, multiply the two to predict how much of the film's total energetic disorder comes from strain. The prediction is then checked with a direct intervention: float the same film off its substrate. This releases the clamping without changing the molecular packing. The liquid-crystal route enters as a specific hypothesis. A film does not store any mismatch strain while it is fluid, so the temperature of its last fluid-to-solid transition should control how much strain it carries at room temperature.
Core Claim
A film only accumulates substrate-mismatch strain after it stops flowing. The mean residual strain should therefore scale as
$$\bar{\varepsilon} \approx \Delta\alpha\,(T_s - T_{use}),$$
where $\Delta\alpha$ is the difference in thermal expansion coefficient between film and substrate, $T_s$ is the temperature at which the last mesophase solidifies, and $T_{use}$ is the measurement temperature. The form comes from linear thermal contraction under full clamping below $T_s$. The contribution of strain variations to energetic disorder is
$$\sigma_{E,strain} = |D|\,\sigma_\varepsilon, \qquad \sigma_{E,total}^2 = \sigma_{E,strain}^2 + \sigma_{E,other}^2.$$
Here $D$ is the deformation potential, the shift of the lowest optical transition per unit strain. $\sigma_\varepsilon$ is the width of the local strain distribution, and $\sigma_{E,other}$ collects everything else, including packing disorder and chemical traps. We assume the two sources are independent and roughly Gaussian.
The strain-onset temperature argument makes a concrete prediction. TPT-TT stays smectic down to 48 °C. It should therefore clamp only over about 25 K of cooling, while a non-liquid-crystalline polymer that vitrifies near its annealing temperature clamps over roughly 150–200 K. The claim is falsifiable in both directions. If $\sigma_{E,strain}$ turns out to be a small fraction of $\sigma_{E,total}$, the field learns that strain is irrelevant in these soft films and that the packing explanation stands. That is also a useful result.
Relation to Prior Work
Three papers motivate the question. Franke et al. produce strain by cooling silicon bonded to sapphire to 8 K. They show that spatial variation of this strain broadens nuclear resonances enough to hide a roughly 10 kHz quadrupole structure. Tuniz et al. bend kagome crystals to shift a charge-density-wave mode from 1.34 to 1.39 THz. They note that bending gradients smear the response. Both treat strain inhomogeneity as a fixed nuisance in rigid crystals. We borrow their logic of calibrating a response per unit strain and then attributing linewidth to a strain distribution, and apply it to soft films whose strain history can be designed through phase behavior.
Sabury et al. and Makala et al. supply the materials and the observation to be explained. Sabury et al. report a nematic phase from 218 to 107 °C and a smectic A phase from 107 to 48 °C in TPT-TT, with tight mobility distributions. Makala et al. contrast edge-on TPT-T, whose mobility rises from about 1.8×10⁻² to 1.5×10⁻¹ cm²/Vs, with helical TPT-2T, which ends up broadly distributed around 3.3×10⁻² cm²/Vs. Neither paper measures residual strain, so the packing and strain explanations are confounded in both.
Two papers constrain the design because they offer rival explanations for disorder. Nikolka et al. show that much of the disorder in polymer diodes comes from water trapped in voids, and that additives shrink tail states to about $k_BT$. Any strain claim must survive controlling for this. Rinehart et al. show that ordinary melt annealing already raises lamellar coherence by 280–400%. Ordering therefore does not require a liquid-crystal pathway, which makes their polymers a natural non-liquid-crystal comparison.
Mohammadi et al. coat films onto an ionic-liquid-infused substrate and reach 2.72 cm²V⁻¹s⁻¹. We reuse their fluid interface as a near-zero-clamping substrate condition. The rejected alternative intersection deserves a brief note. The obvious idea of using a liquid-crystal layer to buffer silicon-on-sapphire strain fails physically, because every mesophase is frozen solid long before 8 K.
Research Program
The work has three parts, all run on the same set of polymers. TPT-TT and TPT-T represent thermotropic liquid-crystal templating. TPT-2T represents a helical, poorly ordered outcome. A regio-asymmetric PAcDOT from Rinehart et al. serves as a melt-annealed polymer without a thermotropic mesophase path.
The first part calibrates $D$. Thin films are annealed on thick polyimide foils and bent over mandrels of known radius. Because the film is much thinner than the foil, the strain across the film is uniform to good approximation. This avoids the through-volume gradients Tuniz et al. encountered. At each bend radius we record the shift of the 0-0 absorption peak and the change in lamellar and π-stacking spacings measured by grazing-incidence X-ray diffraction (GIXD). The diffraction spacings convert bend radius into actual lattice strain.
The second part measures $\sigma_\varepsilon$ and $\sigma_{E,total}$. Microstrain is taken from GIXD peak widths across several diffraction orders. Plotting width against order, as in a Williamson-Hall analysis, separates broadening from finite crystal size (constant across orders) from broadening from strain variation (growing with order). Total disorder is taken from the absorption Urbach energy, which measures the width of the tail below the absorption edge, and from the 0-0 absorption linewidth. The spread of mobility across at least 30 transistors per condition gives a device-level measure.
The third part intervenes on strain while holding packing fixed. Four manipulations do this. Films can be floated off the substrate onto water and re-picked onto a transfer grid, which releases clamping. The cooling rate through the mesophases can be varied. Films can be annealed on substrates of different expansion coefficient: SiO₂/Si, soda-lime glass, polyimide, and the ionic-liquid template. Finally, a film can be annealed on one substrate and laminated onto another at room temperature, which changes the substrate without adding thermal mismatch.
Baselines
The naive reference point is the as-cast film of each polymer on SiO₂/Si. It shows how much disorder exists before any ordering step, and the proposed decomposition must explain the gap between this film and the annealed films.
The strongest existing method is liquid-crystal templating exactly as published, on the published substrate. Its reported benchmark is the transistor hole mobility of (1.4 ± 0.1)×10⁻² cm²V⁻¹s⁻¹ with 100% device yield for TPT-TT in Sabury et al., together with the annealed TPT-T mobility of about 1.5×10⁻¹ cm²/Vs in Makala et al.. We rerun this protocol in our own hands on identical wafers, with the same device geometry and the same number of devices. We report its microstrain and Urbach energy for the first time. We found no reported result on Lacuna for GIXD microstrain or Urbach energy in liquid-crystal-templated films, so the comparison is internal.
The substrate-engineering baseline is the ionic-liquid dynamic template of Mohammadi et al.. It reports mobility up to 2.72 cm²V⁻¹s⁻¹, dichroic ratios of 10.1 (optical absorption) and 22.6 (GIXD), and no strain data. We run it with the same polymers and a matched thermal history.
The chemical baseline is the additive treatment of Nikolka et al., which reports trap-free diode mobility up to 0.2 cm²/Vs and tail widths comparable to $k_BT$. We apply their additive to every condition in a crossed design.
The ablation that isolates the key ingredient is the float-off film. Its packing is unchanged, as verified by identical GIXD peak positions and crystal-size terms, but its clamping is released. The proposal beats the packing-only explanation under two conditions. First, the float-off film's reduction in Urbach energy and linewidth must match the predicted $|D|\,\Delta\sigma_\varepsilon$ within measurement error. Second, $\bar\varepsilon$ must scale with $T_s - T_{use}$ across the cooling-rate and substrate series.
Why not just…?
Why not just anneal slowly without any liquid-crystal phase? Rinehart et al. show that plain melt annealing already gives large coherence gains, so ordering alone does not need a mesophase. Our claim is narrower: the mesophase lowers the temperature at which clamping begins. The PAcDOT comparison, annealed and cooled at the same rate, tests this directly. If it reaches the same $\sigma_\varepsilon$ as TPT-TT, the strain-onset hypothesis is wrong.
Why not just use a matched or fluid substrate? This is a real threat. If annealing on the ionic-liquid template or a matched-expansion substrate removes strain broadening for every polymer, the liquid-crystal pathway adds nothing on the strain axis, and we will report exactly that. The substrate series combined with the lamination control answers the question. The strain budget itself remains useful either way, because it tells device builders whether substrate choice matters at all.
Why not just add molecules that fill voids? Nikolka et al. show that traps can dominate diode disorder. In the crossed design, the additive should reduce $\sigma_{E,other}$ and leave the strain term unchanged, while float-off should do the opposite. If float-off does nothing once the additive is present, strain is negligible and the proposal's central premise fails.
Why not assume strain is negligible in soft films and skip the work? No paper in the file measures it, so the assumption is untested. The bending calibration settles the question cheaply: a small $D$ multiplied by the measured $\sigma_\varepsilon$ would give a quantitative null result.
Decisive Checks
Three outcomes decide the proposal. The first is whether the calibrated $|D|\,\sigma_\varepsilon$ accounts for a measurable fraction of the Urbach energy in films annealed on SiO₂/Si, and whether that fraction disappears on float-off with no change in packing. The second is whether mean strain follows $\Delta\alpha\,(T_s - T_{use})$ across substrates and quench points, with TPT-TT showing the smallest value because of its 48 °C solidification. The third is whether the spread of transistor mobility tracks $\sigma_\varepsilon$ after the additive has removed trap effects.
TPT-2T serves as an internal stress test. Its broad mobility distribution should be explained by packing, meaning a large $\sigma_{E,other}$, rather than by strain. If strain instead appears to explain TPT-2T as well, the decomposition is probably misattributing packing disorder to strain and must be revisited.
References
- Non-covalent planarizing interactions yield highly ordered and thermotropic liquid crystalline conjugated polymers
- Quadrupolar effects on nuclear spins of neutral arsenic donors in silicon
- Strain-Induced Enhancement of the Charge Density Wave in the Kagome Metal <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline"><mml:mrow><mml:mi mathvariant="normal">S</mml:mi><mml:mi mathvariant="normal">c</mml:mi><mml:msub><mml:mrow><mml:mi mathvariant="normal">V</mml:mi></mml:mrow><mml:mrow><mml:mn mathvariant="normal">6</mml:mn></mml:mrow></mml:msub><mml:msub><mml:mrow><mml:mi>Sn</mml:mi></mml:mrow><mml:mrow><mml:mn>6</mml:mn></mml:mrow></mml:msub></mml:mrow></mml:math>
- Non-Covalent Interactions and Helical Packing in Thiophene-Phenylene Copolymers: Tuning Solid-State Ordering and Charge Transport for Organic Field-Effect Transistors
- High-mobility, trap-free charge transport in conjugated polymer diodes
- Impact of Side-Chain Polarity and Symmetry on the Structure and Properties of Acyclic Dioxythiophene Polymers
- Dynamic-template-directed multiscale assembly for large-area coating of highly-aligned conjugated polymer thin films