Effects of polymer grain boundary passivation on organic–inorganic hybrid perovskite field-effect transistors

Organic–inorganic hybrid perovskites (OIHPs), including the archetypical CH₃NH₃PbI₃, which is often referred to as MAPbI₃ or MAPI, are promising novel semiconductors with excellent optoelectronic properties such as clean band structure, large charge carrier diffusion lengths, and low effective mass of charge carriers.¹ Most of the associated research to date has been mainly focused on the photo absorption and emission properties of the OIHP for the application in high-performance optoelectronic devices, including solar cells and light emitting diodes (LEDs), in which the photovoltaic power conversion efficiency (PCE) is now approaching 25%,^2,3 and more than 20% LED external quantum efficiency (EQE) has been achieved.^4,5

However, the fabrication of high-performance thin-film field-effect transistors (FETs) based on OHIPs, especially MAPI, remains highly challenging, ever since one of the first studies on three-dimensional MAPI FETs in 2015.⁶ One of the major concerns rises from the ion migration phenomenon in OIHPs, where various mobile ions, including I⁻, Br⁻, methylammonium ion (MA⁺), formamidinium ion (FA⁺), or even Pb²⁺, are created as lattice defects and migrate under the electrical bias. While such an ion migration is generally perceived to contribute to a fast degradation of the OIHP structure,^7–9 it also screens the gate electric field in the OIHP FET, which not only reduces the extent of gate-indued modulation of the charge carrier concentration and field-effect mobilities but also causes a strong hysteresis in the FET output transfer characteristics during the gate operation.^6,10,11 In addition, the reported effects of ion migration on the charge transport properties in the OIHP FET are often inconsistent in the literature.¹² For instance, both n- and p-type transport modes have been observed in the OIHP FET despite nominally identical OIHP materials and fabrication conditions used.¹² Paulus et al. have suggested that even a small difference in the OIHP composition and stoichiometry originating from inadvertent processing variations could affect the type of major ionic species, resulting in an inconsistent observation of the dominant charge transport mode (i.e., either n- or p-type) in the FET.^12,13

Recent reports show that these undesirable impacts of ion migration on OIHP FETs can be mitigated by either optimizing device operation conditions and engineering FET structures or modifying the composition of OIHPs; for example, the ion migration in the OIHP FET can be suppressed by: (a) operating the FET at low temperatures below which the ion motions are inactive;^6,14 (b) engineering the interfaces existing in the FET;¹⁵ (c) using a high-quality, single-crystalline OIHP without grain boundaries (GBs);¹⁶ (d) using OIHPs with mixed cations such as those in which FA, Cs, or Rb are mixed with MA;¹⁰ (e) replacing three-dimensional OIHPs with lower dimensional metal halide compounds,^17,18 etc. However, these approaches, which require either low-temperature operation or delicate material processing, are, in general, either impractical or complicated, limiting wider applications of OIHP FETs.

Meanwhile, it is well known that GBs, which are omnipresent in solution-processed polycrystalline OIHPs, act not only as sites attracting moisture and oxygen but also as channels facilitating ion diffusions.^19–21 In hybrid perovskite solar cells (PSCs), it has been shown that various types of additives, such as organic and inorganic small molecules and long-chain polymers, could be incorporated into the intergranular regions within the OIHP layer to passivate the GB and improve the performance and stability of PSCs.^7,22–24 For example, a polymer additive, polycaprolactone (PCL), was incorporated into the MAPI PSC by our group recently,²³ where monolayers of stretched PCL polymer chains were passivating the perovskite GB area via the interaction between electron donating carbonyl (C=O) groups on the polymer backbone and electron accepting Pb²⁺ of the perovskite; this resulted in an impeded moisture ingression and, more importantly, a suppressed ion migration. Specifically, the MAPI PSC with PCL GB passivation exhibited reduced out-migration of I⁻ ions to the hole transport layer (HTL), electrodes, and their interfaces, largely preventing the degradation of the HTL and electrodes and contributing to the improved device stability. Such successes of polymer additives in controlling the ion migration in PSCs hint that the same approach may also be utilized to control the ion migration in OIHP FETs and, eventually, improve their performance. However, few reports are available to date, investigating such potential impacts of polymer additives on OIHP FETs.

In this study, we examine the influence of a polymer additive, PCL, and the associated GB passivation on the device characteristics of MAPI-based FETs. Particularly, we compare the charge transport properties of the pristine and PCL-incorporated MAPI FET devices to probe the effects of the PCL concentration in MAPI on the associated FET device characteristics, including device on-off ratio, charge mobilities, majority carrier type (i.e., n- or p- type), and FET current hysteresis. The observed FET transport behaviors are also correlated with the microstructure of PCL-incorporated MAPI and its ion migration activation energy. We find that the PCL passivation of MAPI GBs can mitigate the ion migration issue in the MAPI FET, resulting in an improved control over the charge transport behavior and the hysteresis in the FET transfer characteristics along a minor improvement in the charge mobility.

The FET devices based on PCL-incorporated MAPI were fabricated by spin-casting the MAPI layer on top of the SiO₂ (300 nm; dry-thermal oxide)/p⁺⁺ Si substrate with lithographically patterned source and drain Ti/Au electrodes (10 nm/90 nm) with 80 μm channel length and 2 mm channel width, where SiO₂ and Si act as a gate dielectric and a bottom gate electrode, respectively [Fig. 1(a)]. The deposition of the PCL-incorporated MAPI layer (∼320 nm thick) on the base FET substrate was achieved by a conventional solution process typically used for fabricating PSCs, which combines spin-casting of PCL-containing MAPI precursor solution, anti-solvent treatment, and thermal annealing at 100 °C for 10 min, all in an inert environment²³ (see the supplementary material for more details of the device fabrication procedure).

PCL has both polar and non-polar groups within the chain, and the polar carbonyl groups are strongly adsorbed to the MAPI grain surfaces via the interaction with Pb²⁺,²³ thus being incorporated into the MAPI GBs when the MAPI crystals grow during the thermal annealing step. Previously, we have shown that the width of the GBs in MAPI was ∼1.5–3.5 nm, while the radius of gyration (R_g) of PCL polymer chains was ∼1.77 nm.²³ Given that 2R_g is slightly larger than the GB width, 1–2 monolayers of PCL chains would occupy the MAPI GBs with the GB-confined polymer chains taking a distorted (i.e., stretched) molecular conformation, effectively passivating the MAPI GB [Fig. 1(b)].^23,25,26

The GB incorporation of PCL in the MAPI FET was confirmed by examining the surface morphology of PCL-added MAPI thin films on SiO₂/Si FET substrates by scanning electron microscopy (SEM) and atomic force microscopy (AFM); Figures 2(a)–2(d) show the SEM micrographs of the MAPI layers with 0, 1, 2, and 4 mg/ml of incorporated PCL concentrations, respectively. The MAPI films exhibited compact polycrystalline morphology without visible pinholes regardless of the PCL concentration, but the surface grain size was decreasing as the PCL concentration increased, which is consistent with what has been previously observed in the PCL-incorporated MAPI PSC.²³ The statistical analysis of grain sizes indicates that the average, surface grain size decreased from 184 to 122 nm with a narrower size distribution as the concentration of incorporated PCL increased from 0 to 4 mg/ml [Fig. 2(e)].

AFM surface topography and corresponding lateral force (friction) micrographs obtained from the PCL-incorporated MAPI film further confirmed the presence of PCL at the MAPI GB region [Figs. 2(f) and 2(g)]; the height and friction force line profiles extracted from the AFM micrographs [Fig. 2(h); corresponding to the regions marked by blue lines in Figs. 2(f) and 2(g)] clearly revealed that the surface depressions in the topographical images (i.e., GB locations) were spatially matching the mechanically softer regions in the friction images [Fig. 2(g); brighter regions under the color scheme]. Given that the crystalline MAPI grains are mechanically harder than PCL polymer chains, the data confirm the localization of flexible PCL polymer chains in the MAPI GB region and the resulting PCL GB passivation.

The FET transport characteristics of PCL-incorporated MAPI were obtained by measuring the FET drain current (I_D) during double (forward and reverse) sweeps of sequential increment (i.e., without setting back to zero as in the pulsing mode measurement) in the gate voltage (V_G) from –60 to 60 V at a constant drain-source bias (V_DS) of 60 V at room temperature (300 K) in vacuum (pressure < 0.1 mTorr) (Fig. 3). It should be noted that for each PCL concentration variation, transfer characteristics were measured for at least three devices, and the representative device data are presented in Fig. 3. The corresponding gate leakage current plots associated with these FETs can be found in Sec. S1 of the supplementary material.

In order to estimate the evolution of hysteresis of the PCL-added MAPI FETs, we define V_MR as the average of V_G corresponding to the max and min I_D during the reverse sweep and V_MF as V_G corresponding to the same I_D value during the forward sweep [Fig. 4(a)]. The hysteresis (ΔV_M) for each dual sweep of the transfer curves can then be defined as difference between V_MF and V_MR. As depicted in Fig. 4(b), increasing the PCL concentration leads to decreased extent of the hysteresis in FET transfer characteristics. The least hysteresis is attained by the highest PCL concentration tested (4 mg/ml), as evidenced by the closely overlapping forward- and reverse-direction FET transfer curves, accompanied by evolution of seemingly ambipolar transport [Fig. 3(d)], which we discuss later.

Meanwhile, we find that there is no obvious correlation between the field-effect mobilities (μ_FE) of the MAPI FET and the incorporated PCL concentration [Fig. 4(c)]. The μ_FE values of the MAPI FETs were estimated using the expression, $μFE=LVDSWCi∂ID∂VG$⁠, where L represents the channel length, W is the channel width, and $Ci$ is the insulator capacitance per unit area,^27–29 and they were all in the range of ∼10⁻³–10⁻⁴ cm²V⁻¹s⁻¹ with the highest value observed at the PCL concentration of 1 mg/ml (∼2.5 × 10⁻³ cm²V⁻¹s⁻¹), which is slightly lower than the values reported previously in MAPI FETs.³⁰ This lower μ_FE may have resulted from the fact that no passivating top dielectric layer was used in the current study, while the active MAPI layer was also thick (∼320 nm), which can cause a large back-channel current as well as a limited gate control over the channel region due to the back-gate, back-contact FET device configuration we used. It should be noted that the PCL GB passivation approach may only suppress the ion migration but not eliminate it. It may be possible to further reduce the ionic screening effects by adopting the device architecture having better electrostatic gate control on the charge transport.

It is noteworthy that the FET transfer curve of the pristine MAPI exhibits typical n-type characteristics, where the I_D increases with increasing V_G (positive direction) beyond V_ON, while a constant off-state I_D is observed for V_G < V_ON [Fig. 3(a)]. Surprisingly, passivating the MAPI GBs with PCL leads to the emergence of the seemingly p-type transfer characteristics—as V_G increases during the forward sweep, I_D decreases initially, as if a p-channel is being turned off, which is followed by an upturn of I_D as V_G further increases (i.e., onset of n-type behavior) [Figs. 3(b) and 3(c)]. This may be a sign of ambipolar FET transfer characteristics, which has been observed in a previous study on MAPI FETs.³¹ The possible activation of p-type behavior and ambipolar FET transfer characteristics becomes most pronounced at the highest PCL concentration, 4 mg/ml; the MAPI FET displays transfer characteristics marked by the coexistence of n-type and p-type branches [Fig. 3(d)], similar to the ambipolar behaviors identified in MAPI FETs in previous studies.^15,16 In the meantime, the device on-off ratio [Fig. 4(d)], defined as the difference between the highest I_D and the lowest I_D during the reverse sweeps (from 60 to −60 V), slightly increases to ∼50 in the champion device (average 37.3 ± 9.2) with a low PCL concentration (1 mg/ml), as compared with ∼40 in the champion device (average 35.7 ± 3.6) of the pristine MAPI FET. However, as the PCL concentration further increases, the on-off ratio starts decreasing, reducing to ∼5 (average 5.3 ± 0.1) at the 4 mg/ml PCL concentration, although the likelihood of ambipolar transfer characteristics is the most prominent at that concentration.

Previous reports suggest that I vacancies, generated by actively migrating negatively charged I^- ions, lead to the apparent dominance of the n-type transport mode in MAPI FETs.^32,33 Our recent MAPI PSC study has shown that the ion migration, especially that of I^-, could be suppressed by the PCL GB passivation.²³ Therefore, the appearance of the p-type transport mode with increasing concentration of PCL in the MAPI FET is expected to be driven by the suppressed I⁻ ions migration and reduced I vacancy populations in MAPI by the PCL GB passivation. The accompanying, drastically reduced hysteresis of MAPI FET transfer characteristics by the addition of PCL (e.g., 4 mg/ml PCL concentration) also further supports the notion of suppressed ion migration. Generally, many FET devices based on nanomaterials, including nanowires,³⁴ nanotubes,³⁵ and two-dimensional materials,³⁶ often display the hysteresis in transfer characteristics, which, in principle, is caused by a V_G-dependent charge redistribution within the active FET channel, typically via charge traps in the FET device active layer (e.g., intrinsic defects such as vacancies) and interfaces.^34,36 It has also been shown that mobile ions could induce the hysteresis as identified in carbon nanotube FETs.³⁵ The observed large hysteresis in the pristine MAPI is, thus, expected to originate from both vacancies and mobile ions, which are prevalent in the pristine MAPI. The fact that such a hysteresis in the pristine MAPI FET was greatly reduced by incorporating PCL again provides a strong evidence that the passivation of the MAPI GB by PCL suppressed the ion migration and, likely, the population of ionic vacancies in the MAPI FET.

Finally, we directly confirm the suppressed ion migration in the MAPI FET by experimentally measuring the ion migration activation energy (E_a) via the temperature-dependent measurement of electrical conductance for the pristine and PCL-incorporated (2 mg/ml) MAPI films in the temperature range from 180 to 350 K. We used a two-terminal, lateral-contact device structure [Fig. 5(a)], in which a large channel length (50 μm) helps render the ionic transport more prominent over the electronic transport at high temperatures.³⁷ This method and setup have been widely used to measure the E_a of ion migration in different types of perovskite films.^22,37–42 In this experiment, E_a was estimated by fitting the measured temperature-dependent conductance to the Nernst–Einstein equation: $σ=σ0Texp −EakT$⁠, where σ, σ₀, T, and k denote measured conductance, temperature-independent pre-factor, temperature, and Boltzmann constant, respectively.^37,43 The Nernst–Einstein plots [i.e., ln(σT) vs 1/T] of the pristine and PCL-incorporated MAPI clearly display two distinctive linear regimes, where the low and high temperature regions represent the charge transports dominated by electrons and ions, respectively [Figs. 5(b) and 5(c)]. E_a of ion migration, estimated from the high-temperature region slope, was 0.35 eV for the pristine MAPI, comparable to what has been reported for polycrystalline OIHPs,^{7,22,23,37,41,42} but was markedly increasing to 0.51 eV for the PCL-incorporated MAPI, accompanied by the increased onset temperature for ionic conduction from 263 to 290 K, which was determined from the intersection temperature between the two linear regions in the Nernst–Einstein plots [Figs. 5(b) and 5(c)]. These trends provide direct experimental evidence for the suppressed ion migration in the MAPI FET upon the PCL passivation of MAPI GBs. These results confirm the positive effects of PCL GB passivation on suppression of ion migration in the perovskite. According to our previous study, the PCL polymer can only partially wet the MAPI perovskite (the contact angle at ∼26 °C),²³ which results in not only pushing the polymer phase into GB channels during the crystal growth but also reducing the mobile ions from leaching out of the polymer-passivated GBs.

In this study, we introduced the polymer GB passivation strategy for controlling the undesirable ion migration issues in the MAPI FET. The polymer additive used in the current work, PCL, was shown to be clearly passivating the MAPI GBs, resulting in a drastically reduced hysteresis in the FET transfer characteristics, compared with the pristine MAPI. Furthermore, with incorporation of PCL, the FET device on-off ratio and μ_FE improved up to ∼50 and ∼3 × 10⁻³ cm²V⁻¹s⁻¹, respectively, for 1 mg/ml PCL added device, compared to ∼40 and ∼1 × 10⁻³ cm²V⁻¹s⁻¹, respectively, for the pristine MAPI FET. Especially, at the highest PCL concentration (4 mg/ml), the MAPI FET displayed not only the smallest hysteresis in transfer characteristics but also plausible emergence of the ambipolar transport behavior, unlike the predominant n-type transport mode seen in the pristine MAPI FETs. We attributed these trends to the suppression of I⁻ ion migration and passivation of the GB associated defects by PCL, as confirmed by an increased, measured E_a of ion migration in MAPI upon the incorporation of PCL in this study, as well as the enhanced photoluminescence emission intensity with longer decay lifetime shown in our previous study.²³ Generally, the ambipolar transport in FETs is one of the key factors for designing logic circuits with low-power dissipation and good noise margins, as in complementary metal-oxide semiconductor (CMOS) devices.⁴⁴ The current results represent one of the first studies that indicates control of the ambipolar transport behavior in the MAPI FET by addressing the ion migration in MAPI via the polymer-additive-based passivation of GBs. When combined with further optimized FET device structures and materials (e.g., thinner MAPI active layer, implementation of high-k dielectric top gate, etc.), the demonstrated polymer-additive-based GB passivation strategy should be further applicable to the development of other high-performance OIHP FETs and devices.

See the supplementary material for the detailed material preparation and device fabrication procedures.

The research was carried out at the Center for Functional Nanomaterials (CFN), Brookhaven National Laboratory (BNL), which was supported by the U. S. Department of Energy, Office of Basic Energy Sciences, under Contract No. DESC0012704. This work was also supported by Department of Navy Award No. N00014-20-1-2858 issued by the Office of Naval Research.