1. Introduction

The recent development of a new vapor-deposited material for third-order nonlinear optics [1] has demonstrated a novel approach to the problem of translating high nonlinear optical properties of individual molecules to a solid state material that has the high optical quality, in addition to the optical nonlinearity, that is required for the realization of applications based on the third-order susceptibility in an integrated optics setting. This novel approach consists of using optimized small molecules that have a large third-order polarizability with respect to their size (a large specific third-order polarizability [2, 3]) to create dense, single-component molecular assemblies where the molecular nonlinearities are not diluted [1].

The molecule used for the first dense supramolecular assembly with both a large third-order susceptibility and a high optical quality was DDMEBT, which stands for “2-[4-(dimethylamino)phenyl]-3-([4(dimethylamino)phenyl]-ethynyl)buta-1,3-diene-1,1,4,4-tetracarbonitrile” [1]. This molecule was presented as molecule 2 in [4] and molecule 11 in [5]. It is part of a larger family of 1,1,4,4-tetracyanobuta-1,3-diene (TCBD) based molecules [4–6].

The success of the DDMEBT compound in forming thin films with high optical quality [1] that could be integrated with the silicon-photonics platform to create all-optical switching devices [7–11, 23], and the role played by DDMEBT’s non-planar structure in achieving such a good morphology of the solid-state material [1], motivated us to investigate the effects of a variation of the molecular end-groups around the TCBD core on both the molecular nonlinearity and on the supramolecular assemblies that can be produced. The different compounds studied in this work all involve new donor groups grafted onto the molecular structure in place of the dimethylanilino (DMA) groups that characterize DDMEBT. The nonlinearity of DDMEBT and similar small molecules is enhanced by this donor-acceptor substitution around their compact conjugated system [2, 3, 12].

2. Molecular properties

Figure 1 shows the compounds studied in this work. The synthesis of 1 through 5 was originally described in [13] and discussed in [14]. Molecules 6 and 7 had their synthesis described in [15]. These molecules are non-planar, characterized by a TCBD core on which two donor groups are attached. Their basic structure can be discussed based on the DDMEBT molecule (Fig. 1a). Here, one of the donors, attached to the triple-bond, shares its conjugated system with that bond and two of the CN groups, creating a molecular subunit that is similar to another molecule with strong nonlinear optical response that we studied earlier, TDMEE (1,1,2-tricyano-2-[(4-dimethylaminophenyl)ethynyl]ethene) [1, 2, 16] (Fig. 1(i)). It is this subunit that is responsible for a strong nonlinear optical response. Thus, even though the DDMEBT molecule has a broken conjugation, its nonlinearity remains relatively large. The non-planarity is described by the angle between the dicyanovinyl groups which was determined to be −96.64° by X-ray data from molecular crystals of DDMEBT [4, 5]. This extension into three almost orthogonal directions, while not optimal from the point of view of extended conjugation alone, enables the efficient translation of molecular nonlinearity to the solid state, and should also raise the rotational average of the third-order polarizability when compared to more one-dimensional molecules.

 

Fig. 1 (a–h) Chemical structure of a series of similar compounds based on substitutions on the TCBD backbone; (a) The DDMEBT molecule, which has two dimethylanilino (DMA) groups. (b) to (f): Substitution of different donor groups for the methyl groups of DDMEBT, isopropylmethylamino, diisopropylamino, isopropylamino, butylamino, respectively. (f) Substitution of an hexahydropyridoquinolinyl (HPQ) group. (g) and (h): Two positional isomers obtained by substituting a ferrocene group. (i) The TDMEE molecule.

Download Full Size | PDF

Molecules 1 through 5 use small variations of the original donor group from DDMEBT. The last two, 6 and 7, represent a more dramatic change: they are positional isomers obtained by substitution of one ferrocenyl group — which also has strong electron donating characteristics — for each of the two DMA groups.

All the molecules were first studied in CH₂Cl₂ solutions. The third-order susceptibility of solutions consisting of low concentrations (< 2 wt. %) depends on the concentration C of the solution as ${\chi }_{1111}^{(3)}(C)={\chi }_{1111}^{(3)}(0)+f^4\rho N_A{\gamma }_{\mathit{rot}}C/M$, where f = (n² + 2)/3 is a Lorentz local field factor, n = 1.42 is the refractive index of CH₂Cl₂, N_A is Avogadro’s number, M is the molar mass, and C is the concentration expressed as mass of the solute divided by mass of the solution. γ_rot is the effective scalar third-order polarizability that is obtained from a rotational average of the corresponding molecular tensor [17]. We determined concentration-dependent values of ${\chi }_{1111}^{(3)}(C)$ by degenerate four wave mixing (DFWM) using 1 ps pulses at 1500 nm obtained from a TOPAS traveling wave optical parametric amplifier system from Light Conversion, pumped by a Clark-MXR amplified laser. Absolute values were derived from a reference measurement that established the ${\chi }_{1111}^{(3)}(0)$ for a 1 mm cell filled with the pure solvent to be 6 ± 1 times larger than for a 1 mm thick fused silica sample, for which we used a third-order susceptibility of 1.9 × 10⁻²²m²V⁻². This reference value for ${\chi }_{\mathit{fs}}^{(3)}(-\omega ,-\omega ,\omega ,\omega )$ at 1500nm was chosen as an informed weighted average of the values given in [25–28] taking into account the different wavelengths at which these measurements were taken, as well as the different nonlinear process as in the case of [25]. All values are within 10% of each other. All of the molecules tested here have their longest wavelength absorption peak at less then half of 1500 nm, and their absorption at 750 nm is negligible, which means that two-photon absorption has no relevant contribution at the wavelength where our DFWM measurements are performed, and that the third-order polarizabilities are real-valued.

The results are presented in Table 1. The table also gives new values of the third-order polarizabilities of DDMEBT and TDMEE that have been newly determined and that also take into account a calibration error which caused the third-order polarizabilities reported in [2] and [4] to be too large by a factor 1.5.

Table 1. Properties of the Molecules of Fig. 1^a

View Table | View all tables in this article

Table 1 lists the experimental values for the rotational averages of the third-order polarizabilities (γ_rot), as well as the corresponding specific (γ̃) and intrinsic (γ_I) third-order polarizabilities. These quantities are defined by dividing the experimental values by the molecular mass and by the highest possible limit for the third-order polarizability, respectively. Their use is motivated by the fact that it is desirable to maximize not just γ_rot, but also the nonlinear susceptibility χ⁽³⁾ when the molecules are assembled into a bulk material. This requires other metrics to evaluate the potential of a molecule. The specific third-order polarizability is a simple way to normalize the third-order polarizability to the size of the molecule and obtain a quantity that can be related to the potential of obtaining a large third-order susceptibility in a dense bulk material [2]. The intrinsic third-order polarizability is a measure of how close a molecule is to its fundamental limit [18], and can be calculated in terms of the maximum off-resonant third-order polarizability that could be theoretically obtained given the number of conjugated π electrons and energy-gap between ground and first excited state. Here, we obtain it by dividing the experimental value of γ_rot by the fundamental limit for the absolute value of the third-order polarizability of a centrosymmetric system with the same number of conjugated electrons N_π and same excited state energy E₀₁. This quantitative is a positive number representative of the fundamental limit in general that gives a generally useable value for the intrinsic third order polarizability. In the Si system of units, it is given by [19]

To better discuss the results in Table 1, we also present quantum-chemical calculations of the Highest Occupied Molecular Orbital (HOMO) and the Lowest Unoccupied Molecular Orbital (LUMO) of four of these molecules. The calculations were performed by B3LYP level DFT with a 6–31G(d,p) basis set. A typical result is again exemplified by DDMEBT (see Fig. 2), where HOMO and LUMO tend to be localized on that part of the molecule that corresponds to the “TDMEE subunit” mentioned above: The HOMO has more electron density on the DMA donor group to the left of the figure, which is separated by a triple bond from the cyano groups where most of the LUMO gravitates. This is consistent with the concept that it is this molecular subunit containing the triple bond that is most responsible for the nonlinear optical response.

 

Fig. 2 For each molecule, HOMO is on top, and LUMO is on bottom. Two different isosurface values were chosen, 0.005 on the left to show the general shape of the orbitals and 0.05 on the right to emphasize the respective locations of HOMO and LUMO on the molecule.

Download Full Size | PDF

In Table 1, the intrinsic and specific third-order polarizabilities can be interpreted as a relative measure of efficiency. As measured, DDMEBT is approximately half as efficient compared to TDMEE, both in terms of its use of π electrons (γ_k), as well as in terms of its nonlinearity compared to its size (γ̃). Clearly, the additional DMA group in DDMEBT adds mass and volume to the molecule, and it also decreases the intrinsic third-order polarizability because we count the conjugated electrons on this group despite the absence of conjugation (caused by the nonplanar structure) between this DMA group and the other one. See the next section for the advantages of this non-planar structure.

The first slight variations on the DDMEBT structure, from 1 to 5, do not strongly affect the nonlinearity of the molecule, unless one methyl group is missing from the donor complex (i.e when the anilino nitrogen is only monoalkylated) as is the case of 3 and 4. While the long chains of 4 may contribute to γ_rot, 3 has both a blue shifted absorption and half the third-order polarizability of the other molecules, with correspondingly low specific and intrinsic values.

As to the ferroecene substituted compounds (6 and 7), the “TDMEE subunit” (DMA donor and CN acceptors) in 7 seems to bring with it a larger nonlinear optical efficiency by maximizing the conjugation path, keeping the bonds co-planar, and thus providing a stronger contribution to γ. The dimethylanilinoacetyleno group give a larger nonlinearity when compared to the ferrocenylacetyleno group. Despite the fact that 6 has an apparently lower HOMO-LUMO gap caused by the stronger ferrocene-based donor attached to the triple bond, its nonlinearity is not correspondingly enhanced, resulting in a low value for the intrinsic third-order polarizability.

A comparison of the HOMO and LUMO geometries in Fig. 2 shows that the main electronic transition responsible for the lowest energy excitation in these molecules is localized in the molecular branch that contains the triple bond. It is therefore to be expected that the nonlinearity in this family of molecule is sensitive to the donor that is attached to this particular branch.

The conclusion of this section is that variation of the donor groups does not have a dramatic influence on these TCBD-based molecules. From the point of view of molecular nonlinearity alone, TDMEE is the best, while the DDMEBT compound is the best among the TCBDs. However, the planar structure of TDMEE leads to the formation of micro crystals in the molecular-beam-deposited thin films [1]. While single-crystal TDMEE films are advantageous for some applications [20], random crystallization in thick films impedes the realization of optical applications because of light scattering. While the non-planar structure of DDMEBT renders it less efficient in terms of optical nonlinearity, the same structure also leads to high-quality homogenous films upon molecular beam deposition [1], as will be reviewed in the next section.

3. Supramolecular assemblies and bulk properties

The DDMEBT molecule has been used to create, by molecular beam deposition [1], featureless films that have been successfully integrated with the silicon photonics platform to create all-optical switching and frequency conversion devices [7–11, 23].

We fabricated DDMEBT thin films by organic molecular beam deposition in a high vacuum at an average pressure of 10⁻⁶ Torr, sublimation temperature of 120°C, and deposition rate of the order of nm/min. The DDMEBT thin films appear to the eye as highly reflective and dark. The resulting high optical quality films are transparent starting from around 700 nm. A transmission spectrum (obtained with a Perkin Elmer Lambda-9 Spectrophotometer) of a 1215 nm thick film is shown in Fig. 3(a), showing the clear oscillations due to multiple reflections inside the film. The regularity of the oscillations and the fact that the value of the transmission peaks match the transmission of the naked substrate are an indication of the high optical quality and homogeneity of this film. The transmission can be fitted very well by a model that takes into account multiple reflections at the film’s surfaces, the onset of bulk absorption around 700 nm, and the refractive index dispersion of the DDMEBT material (represented by the solid white line in the middle of the data points in the figure). The refractive index dispersion obtained from this fit is reliable in the range between ∼ 700 nm and ∼ 2500 nm, where it can be described by a Sellmeier formula of the type $n^2-1=S_1{\left(1-{\lambda }_1^2/{\lambda }^2\right)}^{-1}+S_2$ with S₁ = 0.77±0.1, λ₁ = 580±5 nm, and S₂ = 1.335 ± 0.01.

 

Fig. 3 Transmission spectra of DDMEBT films. (a) Transmission of a 1250 nm film (data points in the background) and a theoretical fit (white curve in the foreground) including multiple reflections and the refractive index dispersion of the material. (b) Changes in the transmission spectrum of a 430 nm thick film after several annealing steps in air. Note that there is no change up to 100° for 30 minutes.

Download Full Size | PDF

Next, the robustness of DDMEBT thin films was tested by controlled heating at several different temperatures for different time intervals. The modifications caused in the transmission spectrum by subjecting a 430 nm thick film to subsequent annealing periods at increasing temperatures are shown in Fig. 3(b). Before annealing, the transmission spectrum was measured and a perfect fit (solid curve in the figure) was obtained using the same parameters as were used for the thick film in Fig. 3(b), which is a further proof of the stability and reproducibility of the bulk DDMEBT material obtained by molecular beam deposition. The data points give the spectra after successive annealing steps. The first important result is that heating at 100°C for 30 minutes did not significantly affect the film. The first signs of damage could only be seen when treating the film at 125°C and higher, which is not surprising given the fact that sublimation starts to occur at these temperatures. This is seen by the modifications of the absorption edge in the film, which can be understood by the formation of sub-micrometer pinholes which increase transmission near 500nm and increase scattering in the region between ∼ 700 nm and ∼ 1000 nm in a way roughly compatible with the λ⁴ expectation due to standard scattering theory. The average thickness of the film, as reflected by the position of the transmission maximum around 1500 nm did not change significantly because of the annealing. The DDMEBT molecule itself has a reported decomposition temperature of 451°C with weight loss observed starting at 385°C (See [5], supporting information).

Elemental analysis pure molecular powders and clean substrates guarantee that the DDMEBT films will remain stable and unchanged for years, with no visible degradation spots. This was confirmed with various DDMEBT thin films deposited on glass substrates and stored without any particular protection. Figure 4 (left) shows transmission spectra for various thicknesses of a DDMEBT thin film taken right after fabrication (in July 2008) and again 2 years later. Comparison of data taken two years apart does not show any relevant changes in the optical properties. Slight shifts in the data are caused by thickness differences of less than 10 nm due to slight changes in the positions where the transmission spectra have been taken on films with a thickness gradient. The important result is no systematic change in thickness and no observation of increased scattering after 2 years. The same film was also characterized by atomic force microscopy (AFM). Figure 4 shows that the average roughness of the sample was initially of the order of a few nanometers over distances of the order of micrometers, and that this surface structure and roughness did not significantly change in two years. Both optical and AFM measurements confirm a long shelf-life of DDMEBT thin films stored without any particular protection.

 

Fig. 4 Transmission spectra and AFM surface reliefs of a DDMEBT thin film taken right after fabrication and two years later. The transmission spectra were taken at various positions in a film with a thickness gradient. Small deviations are likely due to errors in positioning the film in the exact same position between subsequent measurements. The most recent AFM data is the plot at the bottom; the larger high frequency noise in this plot is instrumental.

Download Full Size | PDF

The losses of DDMEBT planar waveguides fabricated on glass substrates were determined via a prism coupling [21,22] experiment at 1550 nm. A typical plot of the intensity of a guided wave mode in a DDMEBT thin film as a function of distance is shown in Fig. 5. The fluctuations are caused by scattering centers such as dust on the surface or other defects. The data was fit to the valleys of the fluctuations and the optical loss was determined to be 7 ± 2 dB/cm, which is enough to efficiently exploit nonlinear optical interactions in devices shorter than 1 cm.

 

Fig. 5 LEFT: Prism coupling loss measurement in a DDMEBT thin film. The peak close to zero distance is light scattering from the prism. RIGHT: Third harmonic generation (THG) Maker Fringe measurement for two DDMEBT thicknesses and a fused silica sample.

Download Full Size | PDF

As an example, a Silicon-on-oxide (SOH) device based on a slot waveguide covered by the DDMEBT supramolecular assembly had a large nonlinear parameter, γ = 100W⁻¹m⁻¹, and only needed to be 4 mm long [7]. For a more complete discussion of this FOM and comparison of this parameter to other configurations, see [9]. This waveguide was used to demonstrate demultiplexing of a data stream of 170.8 Gbit/s to 42.7 Gbit/s [7]. Similar DDMEBT-based SOH devices were also used for several schemes of all-optical wavelength conversion [7–11, 23]. The propagation loss of a slot waveguide was measured to be 12 dB/cm when covered with silica glass and 15 dB/cm when covered with the DDMEBT material [7]. In such a device, losses are also influenced by the surface quality of the silicon interface, and therefore it is not possible to extract precise values for the loss caused by DDMEBT absorption in the bulk, but this measurement confirms that the DDMEBT cover layer has a low loss compatible with the prism coupling results.

Finally, a third-harmonic generation (THG) Maker Fringe [24] measurement was used to determine the third-order susceptibility. An off-resonant value was determined by choosing a fundamental wavelength of 2.1μm (obtained from the same laser system presented earlier in the experimental section). We obtained the third-order susceptibility of the glass substrates and the films deposited on them by comparing the Maker Fringe pattern to that for THG in fused silica, which is described by a third-order susceptibility ${\chi }_{1111}^{(3)}=1.6\times {10}^{-22}{\text{m}}^2{\text{V}}^{-2}$ according to [25]. Multiple measurements on several films delivered very reproducible and stable results. Figure 5 shows measurements of films with different thickness and of a fused silica reference. From the Maker Fringe data, the THG third order susceptibility for DDMEBT is ${\chi }_{1111}^{(3)}=1.0\pm 0.3\times {10}^{-19}{\text{m}}^2{\text{V}}^{-2}$, which is 540 times that of fused silica. This is the same order of magnitude obtained in earlier DFWM measurements [1], which is to be expected because all measurement took place at off-resonant wavelengths and vibrational contribution to DFWM, which are possible when using 1 ps pulses, are not likely to be very large compared to the electronic polarizabilities.

All molecules in Fig. 1 with the exception of TDMEE share the same non-planar structure that allows DDMEBT to produce high quality thin films. Molecular beam deposition experiments were performed using molecules 5, 6, and 7. All films were then characterized by transmission spectroscopy to determine the linear refractive index, and by THG Maker Fringe measurements to determine the third-order susceptibility using the same experiments as have been outlined above for DDMEBT. The results are presented in Table 2. The table gives both the bulk nonlinearity that can be predicted from the measured third-order polarizabiltiy and the experimental result. The predicted χ⁽³⁾ are calculated by using the polarizabilities γ_rot measured in solution, and the bulk mass density ρ determined by the crystallographic data. DDMEBT has the largest values for both the predicted and actual values of the χ⁽³⁾, and both values agree very well with each other. This suggests that intermolecular interactions in the assemblies are weak enough that the third-order susceptibilities can be well estimated from the molecular properties, and also that the DDMEBT molecular assemblies achieve a relatively high density not too dissimilar to that determined in crystalline samples. In addition, the intermolecular interaction was weak enough to allow the formation of homogenous assemblies without detectable formation of microcyrstals. The same is roughly true also for the other compounds that we studied. In particular, molecule 7 delivered results similar to those of DDMEBT, with the bulkier ferrocene group not negatively affecting the quality of the films. The other films had some larger deviations between prediction and experiment but were still within a factor of two. Given the errors involved in using the densities extracted from crystal-phase data and all other experimental errors we caution against overinterpreting these results. Still, we observe that films made with molecule 6 showed a weaker third-order susceptibility despite a similar potential than molecule 7, maybe because it was more difficult for this molecule to maintain a high density in the film. Also, molecule 5 had both a smaller refractive index and a small third-order susceptibility, which is also likely to be caused by a smaller density in the film.

Table 2. Third-Order Polarizabilities γ_rot of the Molecules Used for Molecular Beam Deposition and Properties of the Resulting Molecular Assembly

View Table | View all tables in this article

In summary, the use of different end-groups did not impede the formation of a high-quality supramolecular assembly, proving that the DDMEBT results [1] were not one-of-a-kind. We also found that for these molecules it is possible to obtain a good idea of the bulk properties of the corresponding dense supramolecular assembly by simple estimations based on solution data, an indication of weak intermolecular effects. This ties together a complete picture of molecules with good qualities (like γ̃) self-assembling into a material with aggregate qualities that stem from the constituent particles.

4. Conclusion

We have reviewed several small organic molecules which feature large third-order nonlinear optical polarizabilities with respect to their size. One molecule in particular (DDMEBT) forms, by molecular beam deposition, a solid-state material with low loss, large nonlinearity, and the ability to mediate light-light interaction. We have found the DDMEBT thin films to be quite resistant to high temperature treatment and to the passage of time, with basically unchanged thin film properties over two years in an unprotected environment. In addition, we have shown that variations in the end-groups of DDMEBT do not have a strong influence on the quality of the thin films that can be obtained by molecular beam deposition. Within the TCBD family that we investigated, the individual molecular properties are a good predictor for the properties of the aggregate solid state material. While DDMEBT remains the best molecule to date for dense supramolecular assemblies for third-order nonlinear optics, we have shown that an entire class of non-planar TCBD molecules are capable of demonstrating similar linear and nonlinear optical characteristics to those that make DDMEBT so suitable for applications.