Origin of stress and enhanced carrier transport in solution-cast organic semiconductor films

Deposition of organic small-molecules from solution is a versatile low cost route for the fabrication of organic electronic devices. Recently, there has been a tremendous improvement in the performance of solution-processed organic field effect transistors (OFETs). Average charge carrier mobility of 2.1 cm² V⁻¹ s⁻¹ for 6,13-bis(tri-isopropyl-silylethynyl)-pentacene (TIPS-pentacene) and 16.4 cm² V⁻¹ s⁻¹ for 2,7-dioctyl[1]benzothieno[3,2-b][1]benzothiophene have been reported.^1,2 Currently, there is no clear consensus on what basic approaches are most likely to lead to further increases. However, several avenues have proven to be promising, such as control over molecular packing at the stage of chemical synthesis^3,4 and utilizing nearly single crystalline thin films.^2,5 Additional approaches such as engineering the molecular arrangement at the grain boundaries,⁶ or even applying mechanical strain to flexible OFETs,^7,8 have been suggested as methods to improve carrier transport properties. There is a lack of knowledge of the ordering mechanisms that take place during the processing and their impact on the final film properties. For example, thin film deposition at elevated substrate temperature has been correlated with improved device performance,^1,5 and this improvement has been attributed to an increase of the carrier mobility by means of establishing compressive lattice strain conditions.¹ However, the mechanisms that lead to strain and improved mobility are not well understood.

Examples of solution based thin film deposition techniques include drop-casting,^9,10 spin coating,^11–13 ink-jet printing,¹⁴ dip coating,¹⁵ zone casting,¹⁶ hollow capillary writing,¹⁷ solution shearing,¹⁸ and flow coating.¹⁹ All of these processes have been adapted in some variation as “laterally directed” deposition methods, which we define as crystallization that advances across the substrate surface at a liquid solution/solid film boundary along a controllable direction. In the hollow capillary writing process, film deposition is carried out by allowing the solution held in a rectangular capillary to make contact with the substrate, followed by lateral translation of the substrate at a constant writing speed. This process serves as a simple and controllable deposition method.^5,17

Two different growth regimes are obtained as the writing speed is varied. For writing speeds below a critical speed, in the convective regime, the liquid solution transforms into a solid form at a well defined contact line.²⁰ Oriented nearly single crystal films can be produced because the crystallization is seeded by the previously formed crystalline film as it emerges from the meniscus. At higher speeds, the contact line becomes less defined and crystallization occurs via nucleation and coalescence from a supersaturated solution, often leading to a spherulitic-type grain morphology. This regime is known as the Landau-Levich-Derjaguin (LLD) regime.^21,22 In addition to the variation in grain structure, film thickness varies as a power law with the writing speed, where the exponents are −1 and 2/3 in the convective and LLD regimes, respectively.²³ Note that the film thickness increases with writing speed in the LLD regime because solution is pulled out of the capillary by viscous forces that becomes dominant as the speed increases.

Here we focus on the course of ordering during the process of solution-deposition of small molecule organic semiconductor thin films via hollow capillary writing. We have investigated four distinct stages of ordering: (i) in the first stage, laterally directed film deposition occurs with the transformation of the liquid solution into an intermediate solid form. In both the growth regimes, convective and LLD, this transformation takes places very rapidly at a sharp boundary on the order of less than a millisecond, i.e., faster than observable time scales in our video microscopy results. Solvent evaporation causes a lowering of the surface temperature resulting in a thermal gradient between the drying film and the substrate. Films are nearly strain free at this stage, as we infer from the lack of buckling in all cases; (ii) on a timescale of 3–4 s the intermediate form gradually turns into a fully ordered solid state, which we observe through in situ x-ray diffraction and polarized optical microscopy; (iii) when the substrate is held at $60 °C$⁠, mechanical buckling and delamination of the film are observed as the solid film returns to thermal equilibrium. This phenomenon occurs when the strain energy in the film exceeds the adhesive energy at the film/substrate interface.^24,25 This effect is not observed for deposition on unheated substrates; (iv) during cooling of the heated samples from $60 °C$ to room temperature, the buckling direction is observed to rotate by $90°$ indicating a reversal in the sign of stress. When buckling is observed, it leads to the formation of cracks and other mechanical defects in the film, which are found to be highly detrimental to the transport properties of the films.

A strong variation in charge carrier mobility is observed as the writing speed is varied when films are deposited on heated substrates. As the writing speed is decreased in the convective regime, films become thicker and can exceed the critical thickness d_c for delamination via buckling. Since the film thickness varies systematically with both the writing speed and the solution concentration, both process variables affect the film quality. We find that cracks are eliminated when the solution concentration is reduced from 1.0 to 0.1 wt. %, even at the slowest writing speed employed in our experiment (0.1 mm/s); this leads to a fivefold enhancement of average charge carrier mobility. Based on these observations, we propose that the large variation in mobility with the writing speed in the convective regime is primarily due to the elimination of strain-induced defects. Although large enhancement of thin film mobility has been attributed to lattice strain alone for a similar deposition process,¹ we find that strain-induced enhancement of the intrinsic mobility plays a much smaller role.

TIPS pentacene $>$99%, purchased from Sigma Aldrich was used without further purification. Heavily doped n-type (100) silicon wafers with 300 nm thermally grown silicon oxide layer were used as substrates for OFET fabrication and x-ray measurements. Phenyltriethoxy Silane (PTS) was purchased from Sigma Aldrich and was used as received, for the substrate treatment to improve the wettability and reduce surface charge traps.¹⁸ 

The PTS surface was prepared by immersing the wafers in 3.0 wt. % PTS solution in Toluene for 15 h at $90 °C$ followed by sonication in toluene, rinsing in toluene, acetone, and isopropyl alcohol. For room temperature samples, prior to the PTS treatment, the substrates were cleaned with piranha solution (7:3 mixture of H₂SO₄ and H₂O₂) for 25 min, rinsed with deionized water, acetone, isopropyl alcohol, and toluene.

The deposition technique used in this study is a direct-write method using a hollow rectangular capillary with size of 0.5 × 5.0 mm² I.D. (Wale apparatus Co.4905-100), as depicted in Fig. 1(a).⁵ A solution with a concentration of 0.1–3.0 wt. % in toluene is held in the capillary by capillary forces. The substrate is mounted on a computer-controlled linear translation stage (Newport, M-VP-25XA) equipped with a thermoelectric module (Merck Technologies) for temperature control. Film deposition is accomplished by lowering the capillary with a manual translation stage to make solution contact with the surface and then laterally translating the substrate at a controlled rate. An optical microscope (Olympus BXFM) equipped with a video camera was used to capture the deposition process in real time. The thickness measurements were performed with an optical reflectometer (Angstrom Technologies).

The specular x-ray diffraction via synchrotron radiation source is carried out at the X-21 beam line, National Synchrotron Light Source (NSLS) at Brookhaven National Laboratory. The monochromatic x-ray beam of 9 keV (⁠$λ=1.38$ Å) is incident on the sample and the scattered signal is collected on a 487 × 195 pixel Pilatus detector with a pixel size of 172 μm. The 2D images are translated into reciprocal space maps. Q_z and Q_y are the out of plane and in-plane components of the scattering vector Q, respectively. The hollow capillary stage and the XYZ substrate translation stage equipped with the sample heater are mounted together on the sample support stage of a Huber 4-circle diffractometer. The trailing edge of the capillary is positioned $≈$1 mm away from the x-ray beam to avoid any scattering from the glass. The footprint of the x-ray beam on the film is 0.6 mm along the writing direction and 5 mm along the beam direction. The integration time used during the real time x-ray experiment is 0.25 s and the rate at which x-ray frames are taken is 0.37 s.

A heavily doped n-type (100) Si wafer with 300 nm thermal oxide layer is used as the gate contact in the bottom gate-top contact OFET geometry. After the semiconductor film is deposited on the substrate, gold contacts are evaporated using a shadow mask to form source and drain electrodes. The transistors are fabricated with channel length of 50 μm and width of 1000 μm. All field effect transistors are fabricated in ambient condition and tested under N₂ with exposure to light. The electrical measurements are carried out on a microprobe stage (Cascade M150) with Keithley 2636 dual source meter units. For each sample, 8–10 transistor measurements are performed.

The field effect mobility is measured at saturation by a transfer characteristic measurement. The gate bias is swept from 60 V to −60 V and back to 60 V when the drain-source voltage is fixed at −60 V. The drain current is given by

where I_d, W, L, V_g, V_t, and C_i are the drain current, channel width, channel length, gate voltage, threshold voltage, and capacitance per unit area (in our experiments, C_i = 10.0 nF cm⁻²). The saturation mobility $μsat$ is extracted from the slope of the $Id$ vs. V_g transfer curve. The output characteristics are measured by sweeping V_sd from 60 V to −60 V and increasing V_g from 0 V to −60 V in steps of 10 V.

To monitor the ordering, we performed synchrotron based x-ray diffraction and polarized optical microscopy during the deposition. Fig. 1(a) shows the film writing process. In the LLD regime, the speed at which the film crystallizes is less than the writing speed. The speed of crystallization at room temperature was found to be 1.8 mm/s and was highly reproducible from sample to sample. Fig. 1(b) shows several video frames, showing the evolution of the crystalline front just after the stage has stopped. Here, the non-crystalline solution appears dark and polarization contrast develops as the super-saturated solution crystallizes. This shows the formation of the intermediate solid form of the film. The boundary between the liquid and solid is found to be sharper than the resolution of the images, about 3 μm.

Fig. 1(a) also illustrates the geometry of the x-ray experiment, which was performed separately from the microscopy experiments. Note that it takes $≈0.3$ s for the crystallization front to sweep through the region illuminated by the x-ray beam (0.6 mm). Fig. 1(c) shows the (001) integrated intensity as a function of time during the deposition. As the deposition starts, there is an initial increase in the (001) intensity. This increase is from the leading edge of the film, which has crystallized before the stage starts to move. The Bragg intensity drops abruptly after the leading edge sweeps through the x-ray beam because the freshly coated film is still in a liquid state since the speed of the translation stage is in the LLD regime. From the video microscopy, we know that the optical contrast is fully developed in the film at the time indicated by the inset. It is interesting to note that the (001) integrated intensity has only reached 5% of the final intensity at this time. This indicates that the intermediate solid form has limited long range ordering. The Bragg intensity takes $≈3$ s after the stage is stopped to fully develop. No change in the position of the (001) Bragg reflection is observed as it develops, indicating that the interlayer spacing changes less than $±1$%. Additional detail suggesting a gradual propagation front of the long range ordering along the writing direction is shown in Supplementary Figure 2.²³ 

The limited long range ordering revealed by the in situ x-ray measurements combined with the presence of optical polarization contrast implies that the film passes through a transient liquid-crystalline like state with significant orientational order and weak positional order. We favor a picture of an intermediate liquid-crystalline like state templating the final crystalline state over one with growing crystalline domains because the video microscopy shows that the grain structure is fully developed at a stage when the reflected x-ray intensity is very low. The amount of solvent remaining in the intermediate state is evidently very small, since the interlayer spacing changes by less than 1%. Although the mechanical properties of the intermediate state are not well known, we infer that it is nearly strain-free due to the absence of buckling. Note that a lyotropic liquid-crystalline phase has also been proposed for poly(3-hexythiophene) during drop-casting based on polarized Raman spectroscopy measurements.²⁶ As we discuss below, we also infer that evaporation of solvent causes lowering of the film surface temperature relative to the substrate due to the high evaporation rate of the solvent.

The transition between the intermediate liquid-crystalline like state and the final crystalline state, which exhibits higher-order reflections (see Supplementary Figure 7),²³ should be observable by simultaneous observation of both the small and wide angle scattering. Unfortunately, we needed to use a small-format area detector to achieve a high frame-rate read-out so could only measure one region at a time. An obvious improvement for future runs is to acquire a second fast area detector to simultaneously detect the small and wide angle scattering regions. Simultaneous detection is required due to the kinetic nature of the process.

We will now discuss the video microscopy results shown in Figs. 2 and 3. The intermediate form has been observed in optical microscopy under some conditions. These results also reveal buckling during the writing process and cracking when the sample temperature is ramped down. Fig. 2(a) shows a video microscope frame viewed with crossed polarizers during the deposition of the film at room temperature in the convective regime. We observe that the emerging film has a narrow darkened region extending $≈300 μm$ from the contact line. The final bright appearance of the film is formed in 3–4 s. This is consistent with the timescale of the Bragg peak development observed in the real-time x-ray diffraction measurement.

Fig. 2(c) shows the (001) Bragg reflection measured under the same deposition condition. The specular reflection, which is mainly from the substrate, was offset by 0.1° to separate it from the film Bragg peak. It was helpful to separate them during the real-time experiment in order to monitor potential misalignment of the substrate. The Bragg peak is accompanied by a faint arc that represents a small percentage of misoriented grains which are slightly tilted away from the surface normal.

Fig. 2(b) is a video microscope frame taken in bright-field (i.e., without crossed polarizers) during deposition at $60 °C$ in the convective regime. It shows that for deposition at $60 °C$⁠, the darker shaded region becomes much more pronounced. This is due to the increase in the thickness of the film at $60 °C$ relative to room temperature, combined with weakening of the optical reflectivity when the film is in its intermediate state. As the optical contrast develops, buckling in the solid film also starts to appear. In this particular film, buckling is fully developed by about 12 s after the film emerges from the contact line. Fig. 2(d) shows that for this deposition condition the (001) Bragg reflection is accompanied by a much stronger arc. The increased film thickness at $60 °C$ also reduces the intensity of the specular reflection. The strong arc in the Q_y direction appears from local tilting of the film. In addition to the arc, streaks appearing along the Q_z direction are observed. The streaks in the out-of-plane direction are interpreted as a crystal truncation effect due to the buckled surfaces of the film that are physically delaminated from the substrate.

Delamination via buckling suggests the formation of compressive strain. This is interpreted as being due to a modest temperature increase during thermal equilibration after the solvent evaporation has subsided, and suggests that little or no viscous relaxation of the film occurs. Evaporative cooling lowers the temperature of the intermediate state as it develops into a solid form, causing a lateral temperature gradient. The film solidifies completely before reaching thermal equilibrium with the substrate, imparting significant strain in the film after thermal equilibrium has been established. Buckling in the film is induced only in cases where the elastic energy due to compressive strain exceeds the adhesion energy of the film.^24,25 Since the elastic strain energy is dependent on the thickness of the film, this implies a critical thickness for delamination d_c. Note that the thickness of the film varies with the writing speed and solution concentration. As we discuss below, buckling in the films can generally be eliminated either by increasing the writing speed or by decreasing the solution concentration.

Fig. 3 shows the final stage of the process as it takes place during the cooling of the heated sample from $60°C$ to room temperature. Fig. 3(b) shows that as the sample temperature is slowly ramped down the buckles disappear. This indicates that the film has reached a nearly strain free state, and provides an estimate of the temperature excursion that must have occurred during the solidification of the film (⁠$ΔT≈4°C$⁠). Upon further cooling, new buckles appear with their direction rotated by $90°$ relative to the as-deposited film, as exhibited in Fig. 3(c). This is interpreted as being either buckles or wrinkles from a combination of two effects: (i) at this stage, the strain is tensile along the writing direction, forming tension-induced wrinkling; (ii) TIPS-pentacene has an unusual negative thermal expansion coefficient approximately perpendicular to the writing direction in our experiment, which may produce buckling due to compression in this direction. Fig. 3(d) shows that as the temperature of the sample is further reduced to room temperature, the inbuilt stress is released through the formation of a large number of cracks and defects in the film. Most of the cracks are consistent with failure due to tensile stress.

The observation that temperature gradients and cooling from the deposition temperature are the origins of stress leads to quantitative estimates for the stress at each stage of the process. Chen et al. have measured lattice d-spacings as a function of temperature for TIPS-pentacene crystals.²⁷ From their data, we deduce a coefficient of thermal expansion of $≈1.5×10−4$ K⁻¹ for the (⁠$101¯$⁠) d-spacing, and a negative coefficient for the (011) d-spacing with a magnitude about half as large. Taking the (⁠$101¯$⁠) coefficient as an approximate value for our films along the writing direction, we estimate the strain for a temperature excursion of $ΔT=4°C$⁠, to be $≈0.05$%. That the films buckle under relatively little compressive stress underscores that the adhesive energy between the film and the substrate surface is very weak. Upon cooling from $60°$C to room temperature, the film is subjected to a much larger tensile strain along the writing direction, which can exceed 1.0% if the film does not fracture. In-plane strain of this magnitude can be measured in thin films by x-ray diffraction, as shown in Supplementary Figure 7.²³ This model also predicts that after cooling to room temperature the films will be under compressive stress in the direction perpendicular to the writing direction due to the negative thermal expansion coefficient in this direction. Compressive stress in a similar film deposition method on heated substrates has been correlated with dramatically improved carrier mobility.¹ Below, we examine the links between process variables and saturation mobility in field effect transistors.

We present mobility data for deposition at $60 °C$⁠, for speeds covering both the convective and the LLD regimes. Fig. 4(a) shows a polarized optical image of a thin film deposited in the convective regime under conditions where no cracks are formed, and Fig. 4(b) shows the corresponding transfer characteristics for a field effect transistor. Figs. 5(a)–5(c) show the OFET average saturation mobility (⁠$μsat$⁠) versus writing speed for several concentrations. For each concentration, we have observed that the mobility decreases as the writing speed increases in the LLD regime, for deposition speeds $>1.0$ mm/s. This behavior has previously been understood as being due to a reduction in the size of grains and a corresponding increase of grain boundary density as the writing speed is increased in the LLD regime.⁵ The grain structure for each concentration and speed represented in Figs. 5(a)–5(c) is documented in Supplementary Figures 3–5.²³ 

A significant variation of the carrier mobility is observed in the convective regime both as a function of writing speed and solution concentration. In Figs. 5(b) and 5(c), reduced mobilities are observed for speeds below 0.5 mm/s, which is correlated with cracking observed in the corresponding optical images. In contrast, Fig. 5(a) shows little variation down to 0.1 mm/s; due to the lower solution concentration the films are not cracked, as shown in Fig. 5(d). These studies confirm the critical thickness model discussed above for buckling, and we are able to establish $dc≈100$ nm for buckling/cracking. They also suggest a causal relationship between film thickness, mechanical failure of the films, and degraded mobility. Similar correlations between cracking and mobility versus writing speed and concentration are observed for transistors fabricated from thin films deposited at $90 °C$ (Supplementary Figure 6).²³ 

We now consider the important question of whether the large (factor of 5) variation in mobility in the convective regime is mainly caused by enhancement of mobility by compressive strain. We argue that gross mechanical failure of the films as we have observed introduces defects that limit mobility, while enhancement of the intrinsic mobility is a comparatively smaller effect. Although the spacing of cracks appears to be somewhat larger than the transistor channel length employed in our experiments, additional microscopic defects may be introduced that are difficult to observe directly. Buckling induced delamination is also expected to introduce dislocations at the film/dielectric interface. Therefore, our model is plausible based on the observations presented above.

In order to test these ideas, we have optimized thin film deposition at three temperatures (room temperature, $60°C$⁠, and $90°C$⁠) in the convective regime. Table I summarizes the mobility and corresponding strain values measured by x-ray diffraction. Only the data for speeds where no cracking or buckling is observed were included in order to average over the largest possible set of reliable measurements. For example, the average mobility for $60°C$ was obtained from the data shown in Fig. 5(a), where the entire convective regime up to 1 mm/s were included in the average. The x-ray diffraction results were also averaged over the same range of writing speeds.

The result show very little variation in the mobility with process temperature. On the other hand, the strain is seen to vary considerably. In particular, the (011) lattice constant is compressed for deposition at $60 °C$ and $90 °C$ relative to room temperature. The compressive strain varies systematically with temperature, and substantially confirms our model based on lattice contraction/expansion during cooling to room temperature. Our data show no more than a modest mobility increase up to 2.5% compressive strain. The variation in the (101) lattice constant is not as systematic with temperature, and suggest that some strain relaxation occurs at $90°C$ even in the absence of noticeable cracking. A detailed plot of the d-spacing as a function of writing speed and temperature is available in Supplementary Figure 8.²³ 

Given the apparent lack of a large strain effect on the mobility, we look to the literature for relevant examples. External hydrostatic pressure effects on single crystalline rubrene thin films suggest a 20% increase in the mobility for applied pressure of 600 MPa.²⁸ Uniaxial strain applied by bending of flexible substrates produced an 8% increase in mobility for 1% compressive strain in pentacene.⁸ However, this effect was attributed to improve hopping at grain boundaries rather than an enhancement of the intrinsic mobility. Overall, there seems to be little support for large effects on the intrinsic mobility due to a few percent applied stress.

We note that our mobility values are comparable to but slightly less than corresponding results in the literature for TIPS-pentacene. The average mobility reported by Jang et al. for transistors made from dip-coated TIPS-pentacene films at room temperature is 1.5 cm² V⁻¹ s⁻¹.¹⁵ Also, an average mobility of 2.1 cm² V⁻¹ s⁻¹ was reported for a solution sheared film deposited at $90 °C$⁠. Differences in results between groups are most likely due to the variation in defects and interface trap states, which are difficult to eliminate.

In conclusion, the origin of stress and strain is observed to be related to thermal expansion effects. We argue that the “giant” variation in mobility in the convective regime for deposition on heated substrates is mainly due to defects introduced through buckling and cracking of thin films, and we show that this effect can be eliminated by controlling film thickness. In this work, we also demonstrate the great utility of the combined real-time optical and x-ray diffraction methods for the study of ordering processes for solution-deposited organic thin films, and the potential importance of real-time x-ray scattering for characterization of intermediate phases. Future experiments will simultaneously measure the evolution of the small and wide angle X-ray scattering to unambiguously detect the transition from liquid crystalline like to crystalline order. We also plan to use microbeam X-ray diffraction to probe the intermediate phase as it forms in the convective regime.

The authors thank: Steve Lamarra and Christie Nelson for their assistance with the beamline instrumentation and setup for the synchrotron x-ray experiments at the NSLS; Arthur Woll and Thomas Howard for assistance during the x-ray diffraction measurements conducted at the Cornell High Energy Synchrotron Source (CHESS); Michael Ghebreab for assisting with PTS treatment on Si/SiO₂ wafers. IC, PVC, and RLH were supported by Grant No. DE-FG02-07ER46380 from the U.S. Department of Energy (DOE), Office of Basic Energy Sciences, Division of Materials Sciences and Engineering. The pen writer instrumentation development was supported by the National Science Foundation Major Research Instrumentation program under Grant No. DMR-0722451. CMS, YY, and RC were supported in part by Grant No. DE-FG02-06ER46273 from the U.S. DOE, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering. Use of the NSLS was supported by the U.S. DOE, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-98CH10886. Use of the CHESS was supported by the NSF and the National Institutes of Health/National Institute of General Medical Sciences under NSF award DMR-0936384.