C.
Ruiz
a,
I.
Arrechea-Marcos
b,
A.
Benito-Hernández
a,
E.
Gutierrez-Puebla
a,
M. A.
Monge
a,
J. T.
López Navarrete
b,
M. C.
Ruiz Delgado
b,
R. Ponce
Ortiz
b and
Berta
Gómez-Lor
*a
aMaterials Science Factory, Instituto de Ciencia de Materiales de Madrid, CSIC, C/Sor Juana Inés de la Cruz 3, Cantoblanco, Madrid 28049, Spain. E-mail: bgl@icmm.csic.es
bDepartamento de Química Física, Facultad de Ciencias, Universidad de Málaga, Campus de Teatinos, 29071, Málaga, Spain
First published on 28th November 2017
Three crystalline N-trialkyltriindoles in which the length of the alkyl chains attached to the nitrogen has been enlarged from a methyl to a butyl and to a hexyl group have been investigated in the search for semiconducting triindole easy to process from solution. We have found that the number of carbon atoms of the N-alkyl chains has a significant impact on how these molecules interact with each other in the bulk materials and it strongly influences the final morphology of the crystals, which grow as long crystalline wires, cubes or microbelts. Single crystal analysis allows us to recognize the contribution of several cooperative CH–π interactions to guide the self-assembly of these types of molecules. In addition, alkyl chain engineering allows triindole derivatives processable for solution, which render OFETs showing field effect mobilities of 0.03 cm2 V−1 s−1 and 6 × 10−3 cm2 V−1 s−1 when vapor deposited and drop casted, respectively. The results of this study represent a step forward towards the rational control of the supramolecular arrangement of this high performance semiconducting platform, this being a key fact for designing efficient solution-processed organic semiconductors.
Electronic devices involve charge transport as their main operation process and charge transport is highly dependent on electronic coupling between neighbouring units, which in turn depend on the intermolecular order.4–7 In fact, highest mobilities have been determined in perfectly ordered crystalline materials. Unfortunately, organic single crystals are fragile and difficult to process, and their incorporation into devices often requires manual selection and handling of individual crystals. Alternatively, crystalline semiconductors have been incorporated into devices as films prepared by vaporization and condensation onto the desired substrate. However, the need of an evaporation step eliminates the easy-to-process advantages of organic electronics. At this stage, solution processability is an indispensable requirement for large area low cost mass production.
There is therefore, much interest in functional molecules able to self-organize through the action of intermolecular interactions, providing bulk ordered materials upon solution evaporation, which requires a deep knowledge of the non-covalent interactions existing among the molecules. This is usually a challenging task as intermolecular interactions involving π-conjugated systems are weak in nature and act in a cooperative way but often in opposite directions complicating their understanding.8–11 Single crystal analysis provides invaluable information on the nature of intermolecular interactions among neighbouring units and offers an excellent opportunity to investigate the molecular arrangement dependent properties.
In this context we became interested in the electron-rich 10,15-dihydro-5H-diindolo[3,2-a:3′,2′-c]carbazole (triindole) as a self-assembling p-type organic semiconductor. This molecule has been shown to have a strong tendency to self-assemble into one-dimensional superstructures both in solution12 and in the bulk.13 Triindole-based single crystalline or liquid crystalline materials have been found to exhibit very high hole mobilities since they combine the good intrinsic charge transport properties with a high tendency to organize in stacked supramolecular arrangements providing a path for charge carriers. In fact, triindole-based liquid crystals present hole mobilities above 2 cm2 V−1 s−1 qualifying triindoles among the highest hole mobility semiconducting mesogens.14 Crystalline triindoles have shown mobility values of up to 0.4 cm2 V−1 s−1 as determined by SCLC on single crystals of N-trimethyltriindole13 and up to 3 × 10−2 cm2 V−1 s−1 when this compound was incorporated as a sublimed thin film in a field effect transistor.15 However, efficient triindole-based transistors processed from solution are still lacking.16,17
We approach here the study of a family of triindole semiconductors functionalized with different lengths of N-alkyl chains in order to shed light on the effect that these chains exert on the organization preferences and electronic properties of N-alkyl triindoles towards the development of solution processed OFETs. Attaching alkyl chains to improve solution-processability of semiconducting molecules is a common design strategy, however the effect of the alkyl chains in their self-assembly is not always evident.5 The investigation of the crystal structures of different N-trialkylated single crystalline triindoles demonstrates that the size of the alkyl chains has an important effect on their tendency to self-assemble, and points towards cooperative CH–π interactions in the origin of the different superstructures observed. The influence of the diverse alkyl chains on the electrical performance and processability of these materials have been evaluated by incorporating them into OFET devices both by evaporation and solution processes. DFT calculations help us to rationalize the role played by the N-alkyl chains in the electronic properties by analysing the changes in the main molecular charge-transport parameters as a function of the length of the alkyl chain (Fig. 1).
Fig. 2 Comparison of the FT-Raman spectra normalized with respect to the 1607 cm−1 band, recorded for 2, 3 and 4. Inset: Zoom of the 1650–1500 cm−1 region. |
The images show that the size of the alkyl chains has a dramatic influence on the crystalline characteristics of these compounds: thus, while compounds 2 and 4 show a preference to grow into anisotropic shapes forming long crystalline wires or microbelts, the tributyl derivative (3) grow as cubes.
In order to gain insight into the origin of these different crystalline characteristics and macroscopic morphologies, we have obtained crystals with sufficient sizes and quality for X-ray single crystal analysis of 3 and 4 by slow evaporation of 1:1 CH2Cl2:CH3CN solutions of the corresponding compounds. The crystal structure of 2 has been previously reported by us.13 Simulated powder diffraction patterns of the bulk crystals of 2–4 match well with those obtained for the aggregates deposited on substrates confirming that they correspond to the same crystal phase (see Fig. S5–S7, ESI†). An analysis of the crystal packing shows that compounds 2 and 4 crystallize as columnar assemblies, favouring the one-dimensional macroscopic growth; in sharp contrast, the tributyl derivative 3 crystallizes as dimers that pack in a herringbone-like array resulting in no preferred growth direction.
To understand the origin of these different packing tendencies we have traced how molecules interact with each other by establishing close contacts in the structures.
In the crystals, of 2, molecules pack forming stacks along the crystallographic c-axis (Fig. 4) which is coincident with the longest dimension of the crystal. In the structure of 2, each methyl group is involved in CH–π interactions with the aromatic rings that lie above and below it (contact distances: 2.935 Å and 2.985 Å). Fig. 5a depicts the unique interactions of 2, but considering the hexagonal symmetry of the crystals, all the methyl groups are involved in the stabilization of the column.
Fig. 4 Different crystal packings shown by (a) 2 view along (a and b) 3 view along (a, and c) 4 view along b. |
It should be noted that although these types of interactions are individually quite weak, their effects are additive and therefore their influence can be really powerful.21,22 This network of interactions explains the one-dimensional growth of these molecules. It has long been accepted that the growth rate of a facet is directly related to its attachment energy, which is considered as the energy necessary to split a slide of a facet. In this case cooperative interactions can be observed along the [001] direction crossing the most energetic plane.23 No interaction can be observed which would justify the growth direction of other facets of the crystal.24,25
By elongating the chain, we increase the number of potential CH-donors resulting in the columnar slipped arrangement found in the crystallographic packing of 4. In fact, as shown in Fig. 5b, in this derivative, one of the alkyl chains of the molecule interacts with different 5 and 6 membered rings of the heptacyclic platform below it (contact distances: 2.820 Å, 2.810 Å, 2.713 Å and 2.774 Å) while another chain is involved in several CH–π interactions with the molecule located above it (contact distances: 2.599 Å and 3.049 Å). Again no interactions can be traced between different stacks, which explains the anisotropic growth of these structures.
Molecules of 3 arrange in the crystal as dimers that pack in a herringbone-like array. The different packing tendency exhibited by 3 can also be ascribed to the cooperation of a number of CH–π interactions. An analysis of the close contacts in its crystal packing evidences that each molecule interacts with the other component of the dimer (contact distances: 2.621 Å and 2.981 Å) but also with molecules of two adjacent dimers (contact distances: 2.714 Å) placed in planes above and below. These results in a 3D network of interactions which would explain the cube-like morphology observed (Fig. 6).
To this end, 3 and 4-based field effect transistors with a vacuum-deposited active layer in standard bottom gate-top contact architecture were fabricated. Prior to device fabrication the active films were characterized using X-ray diffraction techniques. Simulated powder diffraction patterns of the bulk crystals of N-tributyltriindole (3) and N-trihexyltriindole (4) match well with those obtained for the thermal sublimed thin films, as determined using grazing incidence X-ray diffraction (GIXRD). Fig. 7 shows the comparison between the simulated power diffraction data and the GIXRD spectra of sublimed films of 4 on bare Si/SiO2 substrates, chosen as a reference. Note that only two peaks are observed in the experimental spectrum, corresponding to the ones indexed as (100) and (300) in the simulated spectrum, which indicates a highly aligned crystalline film with the columns parallel to the surface.
Curiously, only the devices fabricated with N-hexyltriindole 4 were active, despite its less conjugated skeleton as evidenced using Raman spectroscopy. This highlights the importance of a favourable supramolecular arrangement for achieving efficient charge transport. Note that the alkyl chain length not only affects the molecular structure but also significantly influences crystal packing as demonstrated in the previous section.
Fig. 8 shows the transfer and output characteristics of 4-based devices. Charge transport evaluation was carried out via analysis of the OFET current–voltage response in the saturation regime with the hypothesis of conventional transistor theory, following eqn:
These conditions were chosen after evaluating different temperatures and surface treatments that, as can be observed in Table 1, induce important differences in the extracted electrical data. Thus, higher preheating temperature results into two orders of magnitude lower mobility while devices deposited on HMDS substrates were inactive and those devices deposited on bare Si/SiO2 substrates presented hole mobilities one order of magnitude lower.
Subst. treatment | Substrate temp. (°C) | μ (cm2 V−1 s−1) | V T (V) | I ON/IOFF |
---|---|---|---|---|
OTS | 80 | 2.8 × 10−4 | −27 | 8 × 102 |
60 | 3.0 × 10−2 | −34 | 2 × 104 | |
25 | 1.7 × 10−2 | −6 | 2 × 104 | |
HMDS | 80 | N.A. | N.A. | N.A. |
60 | N.A. | N.A. | N.A. | |
25 | N.A. | N.A. | N.A. | |
No treatment | 80 | N.A. | N.A. | N.A. |
60 | 2.4 × 10−3 | −21 | 2 × 104 | |
25 | 4.6 × 10−5 | −2 | 7 × 101 |
To understand the differences observed in the mobility values obtained under different conditions, the effects of temperature and surface treatment on thin film crystallinity and morphology were evaluated with the help of AFM. Thermal treatments were performed at temperatures significantly below the melting points (see the corresponding differencial scanning calorimetry traces in Fig. S8, ESI†), thus temperature induced differences expected to derive from morphological changes. As shown in Fig. 9 for OTS-treated substrates, belt-like structures similar to those found in single crystals are recorded for 60 °C temperature, while on bare and HMDS-treated substrates, round-like grains are formed, which are less efficient for charge transport. Note that the non active or poorly performing films, either deposited on HMDS-treated substrates and/or deposited at high temperatures, show a remarkable roughness of 0.5–2 μm s, which probably contributes to the low performances.
In an attempt to give a step forward to the practical applications of these materials, OFETs were also fabricated by drop-casting ∼2.5 mg ml−1 solutions of 2–4 in chloroform. Again only devices prepared with semiconductor 4 as an active layer presented the field effect behaviour. Curiously in spite of the good results previously obtained for 2 in evaporation processed OFETs, solution processed devices prepared with this compound were inactive, probably due to the formation of poor quality films; AFM images indicate the appearance of non well-interconnected crystalline wires, probably hindering efficient charge transport between wires in the device channel (Fig. S7, ESI†). In contrast, AFM images of solution processed 4 samples show quite homogeneous films, highlighting the remarkable processability modulation by increasing the N-alkyl chain length from methyl to hexyl groups.
Prior to device fabrication, the active films were again characterized using X-ray diffraction techniques. The GIXRD spectrum recorded for film 4 rendering the best performance in OFETs shows, as in sublimed films, two peaks corresponding to the ones indexed as (100) and (300) in the simulated spectrum, indicating comparable crystalline phases (Fig. S8, ESI†).
Hole mobilities of up to 6.4 × 10−3 cm2 V−1 s−1 with a threshold voltage of ∼−31 V were obtained for a 4-based OFET annealed at 80 °C, with a bottom gate-top contact architecture, and using HMDS-treated Si/SiO2 substrates as gate/dielectric layers (see Table 2). Note that despite the fact that this value is around one order of magnitude smaller than the values recorded for vacuum-deposited films, to the best of our knowledge this is the highest value reported so far for solution processed triindole-based semiconductors in organic field effect transistors.
Subst. treatment | Annealing temp. (°C) | μ (cm2 V−1 s−1) | V T (V) | I ON/IOFF |
---|---|---|---|---|
OTS | 120 | 4.5 × 10−4 | −32 | 1 × 104 |
80 | 1.5 × 10−3 | −37 | 2 × 103 | |
60 | 1.7 × 10−3 | −38 | 4 × 104 | |
HMDS | 80 | 6.4 × 10−3 | −31 | 3 × 103 |
60 | 2.6 × 10−3 | −12 | 5 × 102 |
In order to rationalize the effect of the length of the N-alkyl chains on the charge transport properties of triindoles we have theoretically predicted the fundamental charge transport parameters, such as intramolecular reorganization energies (λh), electronic coupling (or transfer integrals) and mobilities. As expected, the structural reorganization needed to accommodate a charge is not affected by the length of the N-alkyl groups and very similar λh values are found for 2–4 (∼235 meV). However, significant differences are found when comparing the transfer integrals for hole transport (th) for dimers extracted from the experimental single crystal data of 2–4. Our results show: (i) larger th values for 2 (−41 meV), which is driven by efficient HOMO–HOMO overlap due to its intermolecular columnar organization. (ii) Slightly lower th values (−28 meV) for dimers along the columns of 4, with the slipped columnar arrangement and the larger number of CH–π interactions likely decreasing wave function overlap. (iii) Moderate th values (−23 meV) between cofacial dimers of 3 extracted from the crystal, with th values that drop to −3 meV between molecules of two adjacent dimers in a herringbone-like configuration. This result underlines the strong impact of the intermolecular packing on the transport properties. Considering that the calculated reorganization energies in triindoles 2–4 are much larger than the electronic coupling, charge localization takes place and transport is expected to occur via a charge-hopping mechanism. We calculated the intrinsic hole mobilities by assuming the charge hopping regime in the context of the semiclassical Marcus theory.26 Mobilities of 0.140 and 0.067 cm2 V−1 s−1 were estimated for 2 and 4 (note the good agreement between experimental and theoretical values for 4, being 0.03 and 0.067 cm2 V−1 s−1 respectively).
A colorless crystal of 4 showing well defined faces was mounted on a Bruker four circle kappa-diffractometer equipped with a Cu INCOATED microsource, operated at 30 W power (45 kV, 0.60 mA) to generate CuKα radiation (λ = 1.54178 Å), and a Bruker VANTEC 500 area detector (microgap technology). Diffraction data were collected exploring over a hemisphere of the reciprocal space in a combination of φ and ω scans to reach a resolution of 0.86 Å, using a Bruker APEX2 software suite (each exposure of 40 s covered 1o in ω or φ). Unit cell dimensions were determined for the least-squares fit of reflections with I > 20σ. A semi-empirical absorption and scale correction based on equivalent reflection was carried out using SADABS APEX. The structures were solved using direct methods. The final cycles of refinement were carried out by full-matrix least-squares analyses with anisotropic thermal parameters of all non-hydrogen atoms. The hydrogen atoms were fixed at their calculated positions using distance and angle constraints. All calculations were performed using SMART27 and APEX228 software for data collection and for data reduction; SHELXS29 and SHELXL30 to resolve and refine the structures using Olex2.31 CCDC 1021612 and 1446933 for 3 and 4 respectively.†
The HOMO transfer integrals (tHomo,Homo, referred to in the text as th) for the nearest-neighbour pairs of molecules taken out of the 2–4 crystal structures were calculated at the B3LYP/6-31G(d,p) level, according to the approach described by Valeev et al.39 with the corresponding matrix elements evaluated with Gaussian 09. Note that the coupling values depend on the functional used and generally increase with the increasing percentage of Hartree–Fock exchange in the functional.40 For molecule 2, where a two-fold degenerate HOMO is observed on isolated molecules at the geometry of the crystal, the HOMO−1 transfer integrals (tHomo−1) were also calculated in order to recalculate average th values as [(tHomo,Homo2+ tHomo−1,Homo−12 + tHomo−1,Homo2 + tHomo,Homo−12)/2]1/2. However, no difference is found between the average th values and the previously calculated tHomo,Homo values.
Footnote |
† Electronic supplementary information (ESI) available: Cyclic voltamograms and UV spectra of 2–3, DFT theoretical eigenvectors for the Raman bands and GIXRD and AFM of solution processed films, explanation to checkCIF alerts. CCDC 1021612(3) and 1446933(4). For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/c7tc03866f |
This journal is © The Royal Society of Chemistry 2018 |