Thermoelectric Properties of Hybrid Organic–Inorganic Superlattices

Apr 30, 2012 - Guru P. Neupane , Krishna P. Dhakal , EunHei Cho , Bong-Gi Kim , Seongchu Lim , Jubok Lee , Changwon Seo , Young Bum Kim , Min Su Kim ...
1 downloads 0 Views 2MB Size
Article pubs.acs.org/JPCC

Thermoelectric Properties of Hybrid Organic−Inorganic Superlattices Jesús Carrete,† Natalio Mingo,*,‡ Guangjun Tian,§ Hans Ågren,§ Alexander Baev,∥ and Paras N. Prasad∥ †

Departamento de Física da Materia Condensada, Facultade de Física, Universidade de Santiago de Compostela, E-15782 Santiago de Compostela, Spain ‡ LITEN, CEA-Grenoble, 17 rue des Martyrs, BP166, 38054 Grenoble, France § Department of Theoretical Chemistry & Biology, School of Biotechnology, Royal Institute of Technology, SE-106 91 Stockholm, Sweden ∥ Institute for Lasers, Photonics and Biophotonics, Department of Chemistry, State University of New York, Buffalo, New York 14260-3000, United States ABSTRACT: We theoretically evaluate the thermoelectric transport coefficients of hybrid thiophene/SiGe superlattices and the effect of their chemical tuning via phenyl groups. Owing to the interplay between alloy scattering and phonon transmission at the molecular layers, very low thermal conductivities under 1 W/(m K) and values of ZT more than twice as large as those of bulk SiGe can be attained. These results highlight exciting possibilities of organic−inorganic hybrid systems, as compared to traditional inorganic thermoelectrics.



INTRODUCTION The rational control of thermoelectric transport has been the subject of growing interest in recent years. A major obstacle to viable thermoelectric conversion is the inability to independently manipulate the thermal and electric properties of traditional semiconductor materials. In traditional inorganic thermoelectrics the density and mobility of charge carriers are interlinked with thermal transport and thermopower. This complicates the optimization of thermoelectric performance as enhanced electrical conductivity, σ, is typically associated with increased thermal conductivity, κ, thereby making it difficult to increase the dimensionless thermoelectric figure of merit, ZT = (σS2T)/κ. On the organic side, conducting polymers have also been explored for thermoelectric applications owing to their relatively low thermal conductivity (typically an order of magnitude less than inorganic materials). Nonetheless, the Seebeck coefficient values so obtained also tend to be an order of magnitude lower than for inorganic materials, and maintaining high electrical conductivities through robust doping schemes can be challenging.1,2 An emerging class of materials that has not been investigated for thermoelectric applications is that of hybrid organic/ inorganic structures (HOI).3 By HOI we refer to a bulk material, typically a superlattice or a nanocomposite, composed of alternating organic and inorganic layers or domains of nanometer size. A particular case of this is a superlattice where inorganic layers of the same thickness are separated by selfassembled monolayers of organic molecules. Such materials have been synthesized in several instances,4−9 and their application as semiconducting channels in thin-film field-effect transistors,10,11 photoconductive devices,12 and diodes13 has © 2012 American Chemical Society

been envisaged. However, research on thermoelectric properties of hybrid systems has so far been limited to one single molecular interface,14−17 so the properties of a hypothetical bulk HOI superlattice are not known, either theoretically or experimentally. Intuitively, one would expect that the thermal conductivity of the HOI superlattice may be extremely low, because phonons will be strongly backscattered by the soft organic interfaces. In addition, organic chemistry techniques might allow the electronic properties of the interface to be finetuned so as to achieve a high enough thermoelectric power factor, resulting in an overall high ZT for the bulk HOI. In this paper we investigate this question, concluding that HOI materials can indeed display very low thermal conductivities and enhanced ZT and that they can be adjusted by organic chemistry techniques. We also find that using a semiconductor alloy as the inorganic component is crucial to obtain a high ZT in the HOI structures. The particular systems we have investigated are made of periodically repeated layers of monocrystalline Si or SiGe alloy, separated by self-assembled monolayers (SAM) of trithiophene (see Figure 1). We are aware that it may not be easy to synthesize a HOI superlattice containing SiGe, since the high temperatures typically required to grow the semiconductor may destroy the organic part. Reference 4 presents various alternative synthesis approaches already used in other systems. For example, it would be possible to roll up an epitaxial layer of SiGe coated with the organic SAM, in the way already Received: March 15, 2012 Revised: April 30, 2012 Published: April 30, 2012 10881

dx.doi.org/10.1021/jp3025039 | J. Phys. Chem. C 2012, 116, 10881−10886

The Journal of Physical Chemistry C

Article

extended molecule was calculated with density functional theory (DFT). With such a setup, we were able to accurately describe both the bulk electrodes and the extended molecule. In the framework of elastic-scattering theory, the transition probability can be calculated as T = ∑i|T(Ei)|2. Here T(Ei) is the energy-dependent transition matrix element, which is defined as25 T (Ei) =

∑ ∑ VJSVDK ∑ J

K

η

⟨J |η⟩⟨η|K ⟩ zi − εη

(1)

where J and K run over all atomic sites and VJS (VDK) denotes the coupling energy between the atomic site J (K) and the left (right) electrode. The coupling energies, which are the key parameters in computing the transition matrix elements, can be calculated analytically with the electronic structure obtained from quantum-chemical calculations.26 The product of the two overlap matrix elements ⟨J|η⟩⟨η|K⟩ represents the delocalization of the molecular orbital |η⟩. εη is the eigenvalue of orbital |η⟩. The overlap elements and the molecular orbitals can also be obtained from DFT calculations. The complex variable z is defined as zi = Ei + iΓi, where Ei is the energy at which the scattering process is observed. h/Γi is the escape rate which is determined by Fermi’s golden rule.26 The lattice thermal conductivity can be expressed generically as an integral over phonon frequencies:

Figure 1. Scheme of a trithiophene molecule, with (bottom) or without (top) a lateral phenyl group, inserted between two hydrogenpassivated semiconductor atoms.

demonstrated for other hybrid HOIs in refs 7−9. Substrate transfer techniques may also be an option to circumvent this problem. SAMs of related molecules, such as ω-thiophenefunctionalized n-alkyltrichlorosilane, allyl phenyl thiophene ether, and thiophene-terminated alkyl chains, have been experimentally grown on Si substrates,.18−20 Our choice of SiGe is based on the ease of calculation, but the physical phenomena identified by our study are likely to take place in any generic HOI superlattice that uses a thermoelectric alloy as the inorganic component. From an experimental point of view, it may be simpler to employ other alloys such as BiSbTeSe, which are compatible with low temperature growth and wet chemistry techniques.

κl = (L + l)



dfBE dT

ℏω ;(ω)

dω 2π

(2)

where ; is the transmission function per unit area, f BE is the Bose−Einstein distribution, and L + l is the period length. The transmission function of the layered system can be approximately evaluated from the separate transmission functions of the interface and the bulk semiconductor parts.29−31 The phonon transmission of the molecular interface, ; i, was calculated using the three-region formula from the nonequilibrium Green’s function formalism.32 This method is adapted to a two-terminal setup, with the molecule inserted between two infinite semiconductor leads. In our case, these leads were either pure silicon or pure germanium, and the results were then averaged to approximate the transmission expected in the real situation, in which the molecule is attached to layers of a SiGe alloy. All intermolecular and moleculesemiconductor force constants were taken from our ab initio results, whereas interactions between semiconductor atoms were modeled by a Stillinger−Weber potential.33,34 Ab initio constants where corrected so as to enforce the appropriate translational and rotational symmetries.35 The Green’s function of the infinite leads was calculated by decimation,36,37 a wellknown renormalization-group approach. The transmission corresponding to the Si or SiGe semiconductor segment is an integral over the Brillouin zone (BZ)



METHODOLOGY Density functional theory calculations at the B3LYP level with the 6-31G basis set were performed to get the optimized geometries and electronic structures of SiH3-trithiophene-SiH3, GeH3-trithiophene-GeH3, SiH3-phenyl-trithiophene-SiH3, and GeH3-phenyl-trithiophene-GeH3 molecules. The force constants, that is, the second derivatives of the energy with respect to displacement of the atoms, were calculated on the optimized geometry at the same level of theory. All of the quantumchemical calculations were performed with GAUSSIAN 09.21 Our choice of basis set was based on previous experience and investigations on the matter, in some of the authors’ laboratory22 as well as elsewhere,23 indicating that compact basis sets that describe valence-like orbitals well are both costeffective and accurate for molecular electron conductivity calculations. In fact, large basis sets with diffuse functions may introduce worse results in terms of “ghost transmission”; see ref 23. The electronic transmission probabilities across the molecular interface were calculated with the QCME program24 which is based on the general Green’s function formalism.25,26 QCME has been successfully applied to describe the electronic transport properties of various molecular junction systems.22,27,28 In our calculations, we divided the molecular junction into three parts: source, extended molecule, and drain. The source and drain electrodes were described by an effective mass approximation, while the electronic structure of the

;b =

1 ∑ (2π ) (L + l) α 2

∫BZ vq(z⃗ ,α) λq ⃗ ,αδ(ω − ωq ⃗ ,α)dq ⃗

(3)

where q⃗ is the phonon wave vector, α is the phonon branch, λq⃗,α is the mode's relaxation length, vq⃗ ⃗,α is its group velocity, and ωq⃗,α is its frequency. The calculation is simplified by the use of an approximate linear phonon dispersion with parametrized relaxation rates.38 The total transmission for the period is given by29−31 ;(ω) = [; −b 1(ω) + ; −i 1(ω) − 1]−1 10882

dx.doi.org/10.1021/jp3025039 | J. Phys. Chem. C 2012, 116, 10881−10886

The Journal of Physical Chemistry C

Article

power factor P = σS2, which happens close to the edge of the conduction band and yields carrier densities n ∼ 1020 cm−3. A superlattice whose semiconducting part is made of pure silicon proved to be a poor choice for thermoelectric applications. Therefore, the discussion hereafter will concentrate on the maximally disordered, optimally doped SiGe:SAM HOI.

As regards charge carrier behavior, our model for SiGe alloys comprised a single valence band and two conduction bands (L, with six equivalent minima, and X, with four). The required band parameters where taken from refs 39 and 40. For each band, an energy-independent relaxation time was assumed with two contributions, bulk and alloy scattering; see ref 40 for details about its parametrization and comparison with experimental data. Under these hypotheses, the electric conductivity of SiGe (σ), its Seebeck coefficient (S), and the contribution of charge carriers to its thermal conductivity at zero electric field (κ0) can be calculated as:41 2e 2 h

σ=

⎛ dfFD ⎞ ⎟Π(x)dx dx ⎠

∫ ⎜⎝−

S = kB

2kB 2T ℏ

(4)

dx

dfFD dx

κ0 =

RESULTS AND DISCUSSION The left panels of Figure 2 show the electronic transmission of a trithiophene molecule, with or without a lateral phenyl group,

dfFD

( )xΠ(x)dx ∫ (− )Π(x)dx

∫ −



⎛ dfFD ⎞ 2 ⎟x Π(x)dx dx ⎠

∫ ⎜⎝−

(5)

(6)

−1

where x = (kBT) (ε − μ), kB and h are Boltzmann’s and Planck’s constant, respectively, e is the electronic charge, f FD is the Fermi−Dirac distribution, and Π is defined as Π(x) = Figure 2. Electronic (left) and phononic (right) transmissions for a trithiophene molecule (top, in red) and a phenyl-trithiophene molecule (center, in green) between SiGe leads, along with the difference between the results without and with the lateral phenyl group (bottom, in blue). In all cases, transmissions were calculated for pure Si and pure Ge leads and averaged to yield these results.

8 [ ∑ Mjτj Θ(kBTx + μ − ε0, j) 3π ℏ j ∈ cb m*j (kBTx + μ − ε0, j)3 +τval Θ(ε0,val − kBTx − μ) m0*(ε0,val − kBTx − μ)3 ] (7)

placed between Si0.5Ge0.5 leads, for an energy range around the Fermi level. Those transmissions were obtained by averaging our ab initio results for pure Si and pure Ge. As expected, the transmission profile consists of narrow peaks centered around the electronic states of the tiophene oligomer, and the modifications introduced by the side phenyl ring are relatively minor. In the same figure, the right panels show the phononic transmissions of the same systems. In the frequency range most relevant for thermal transport (ω ≲ 40 THz) the side ring introduces several vibrational modes and changes the overall transmission profile in a comparatively more significant fashion. To compute the transport properties of a Si0.5Ge0.5/ trithiophene superlattice, these data must be used in conjunction with the approximations to the phononic and electronic behaviors of SiGe alloys described in the previous section. While the width l = 2 nm of a molecular monolayer is fixed, the superlattice period l + L can be adjusted by a suitable choice of the width of the semiconductor part (L), subject only to technological constraints. Therefore, different transport regimes can be achieved. The long mean free path of the aforementioned long-wavelength phonons in SiGe42 provides adequate ground for treating phonon transport as quasiballistic, that is, for combining the scattering probabilities of the semiconductor and the molecular monolayer into a joint transmission and using the latter in the calculations of the phenomenological coefficients. On the other hand, an equivalent combination rule cannot be expected to hold for electrons in superlattices with a period beyond 5 nm,

where Mj is the number of equivalent minima in each band, τ is the appropriate relaxation time, m* is the corresponding effective mass, ε0 is the energy of the band edge, cb stands for conduction bands, and Θ is a step function. By comparing these equations to those from the Landauer−Buttiker formalism,32 it can be seen that they are also applicable to a bidimensional array of quasi-one-dimensional conducting elements such as a molecular monolayer if Π is replaced with ρSL; e, that is, the product of the electronic transmission of each element (; e), the surface density of elements (ρS) and the width of the layer (L). Thus, Π can be thought of as a bulk analogue of the transmission; knowledge of Π for a given material is sufficient to calculate its thermoelectric power factor P = σS2 and, if the phononic contribution to its thermal conductivity can be estimated, also its dimensionless thermoelectric figure of merit ZT = (σS2T)/(κe + κphonons). In this last equation, κe denotes the charge carrier contribution to the thermal conductivity at zero electric current, which is related to κ0 through κe = κ0 − PT. The percentage of germanium in the Si1−xGex alloy, x, plays a crucial role in the modulation of the thermoelectric properties of the superlattice. Compared to pure silicon, an alloy with significant disorder scatters both electrons and phonons more efficiently. To evaluate the net effect of this parameter, calculations were run for the two most extreme cases: x = 0 (no germanium) and x = 0.5 (maximum disorder). The position of the Fermi level was chosen so as to maximize the 10883

dx.doi.org/10.1021/jp3025039 | J. Phys. Chem. C 2012, 116, 10881−10886

The Journal of Physical Chemistry C

Article

by the main panel of Figure 3, which contains a comparison between Si0.5Ge0.5/trithiophene and Si/trithiophene superlattices with the same period length and shows differences of up to 1 order of magnitude. Electronic transport is also somewhat hindered by the thiophene molecules, as reflected in Figure 4, which shows the

approximately, since this is the order of the electronic mean free path.40 Thus, for longer periods a hypothesis of ohmic (or diffusive) transport is more appropriate, whereby the transport coefficients derived from the transmission probabilities of the charge carriers (electric conductivity, electronic contribution to the thermal conductivity, and Seebeck coefficient) are calculated separately for each kind of layer in the superlattice and then combined using rules analogous to those from ordinary concentrated-parameter circuit theory: electric and thermal resistances are composed by summing, and the joint Seebeck coefficient of two thermoelectric elements in series is a weighted sum of both Seebeck coefficients, each weight being equal to the ratio of the thermal resistance of the element to the total thermal resistance (both of them evaluated at zero electric field). In either case, ohmic or ballistic, coherency between neighboring semiconductor layers is broken by the organic molecules inserted between them; therefore, different periods of the superlattice behave incoherently, a phenomenon known to lead to low thermal conductivities in similar systems.43 Figure 3 (inset) shows the total thermal conductivity at room temperature (always taken as 300 K for the purposes of this

Figure 4. Room-temperature power factor of a Si0.5Ge0.5/trithiophene superlattice as a function of its period length, using ohmic and ballistic composition rules for the electronic calculations, with or without a side phenyl group.

power factor P = σS2 of the superlattice under study for optimal doping. The presence or absence of the phenyl group has little effect, as expected from the similarity of the electronic transmission profiles in both cases. The attainable values of P are rather modest, 5−10 times lower than those of bulk SiGe.42 Fortunately, the reduction in P is not enough to compensate for the extremely low thermal conductivity of the superlattice, so the dimensionless thermoelectric figure of merit ZT = PT/κ (Figure 5) still improves by about a factor of 2 the optimal Figure 3. Comparison between the room-temperature thermal conductivities of Si0.5Ge0.5/trithiophene and Si/trithiophene superlattices, using ohmic composition rules, as a function of their period length. Inset: results for Si0.5Ge0.5/trithiophene, with κ in linear scale to highlight the effect a phenyl side group.

article) of a Si0.5Ge0.5/trithiophene superlattice for optimal doping, which includes the contributions from phonons and charge carriers, as a function of its period. There is very little difference between the results obtained under an assumption of quasiballistic electron transport (which have been omitted for clarity) and those afforded by ohmic combination rules, due to the predominance of the phononic contribution. It can be seen, accordingly, that a phenyl ring bonded to the oligomer can make a significant difference through the addition of conduction channels for phonons and the modification of the transmission profile of the already existing modes. In any case, the computed total thermal conductivity is extremely small: by way of comparison, it is 2 orders of magnitude lower that the value for bulk silicon, about one-half of the thermal conductivity experimentally reported for some of the most promising thermoelectric Si nanostructures,44 and one tenth of the value for bulk Si0.5Ge0.5.45 The molecular layers play a crucial role in scattering low-frequency phonons which are relatively unaffected by alloy scattering due to their being in the Rayleigh limit when compared to the characteristic length of alloy disorder. The importance of alloy scattering is highlighted

Figure 5. Room-temperature dimensionless figure of merit of a Si0.5Ge0.5/trithiophene superlattice as a function of its period length, using ohmic (left) and ballistic (right) composition rules for the electronic calculations.

theoretical value for SiGe.42 Figure 5 further shows that practically useful values of ZT can be achieved in both the ballistic and the ohmic coupling regimes. Although it seems unlikely that the ballistic picture afford good results for l + L as large as 10 nm, even if it only holds up to a fifth of that value it would still yield ZT ≃ 0.23. At the other extreme, a very long period (∼0.2−0.3 μm) maximizes the figure of merit in the ohmic regime, with a similar optimum value. A third interesting conclusion which can be drawn from Figure 5 is that thermoelectric properties are easily tunable by 10884

dx.doi.org/10.1021/jp3025039 | J. Phys. Chem. C 2012, 116, 10881−10886

The Journal of Physical Chemistry C

Article

(5) Xing, L.-L.; Li, D.-P.; Hu, S.-X.; Jing, H.-Y.; Fu, H.; Mai, Z.-H.; Li, M. J. Am. Chem. Soc. 2006, 128, 1749. (6) Yoon, K.-H.; Han, K.-S.; Sung, M.-M. Nanoscale Res. Lett. 2012, 7, 71. (7) Deneke, C.; Zschieschang, U.; Klauk, H.; Schmidt, O. G. Appl. Phys. Lett. 2006, 89, 263110. (8) Deneke, C.; Schumann, J.; Engelhard, R.; Thomas, J.; Sigle, W.; Zschieschang, U.; Klauk, H.; Chuvilin, A.; Schmidt, O. G. Phys. Status Solidi C 2008, 5, 2704. (9) Bufon, C. C. B.; Espinoza, J. D. A.; Thurmer, D. J.; Bauer, M.; Deneke, C.; Zschieschang, U.; Klauk, H.; Schmidt, O. G. Nano Lett. 2011, 11, 3727. (10) Mitzi, D.; Chondroudis, K.; Kagan, C. IBM Res. Dev. 2001, 45, 29. (11) Kagan, C. R.; Mitzi, D. B.; Dimitrakopoulos, C. D. Science 1999, 286, 945. (12) Carrasco-Orozco, M. A.; Stirner, T.; O'Neill, M.; Ellis, C.; Dong, D.; Kelly, R.; Piepenbrock, M. O.; Kelly, S. M. Phys. Rev. B 2007, 75, 035207. (13) Park, Y.; Han, K. S.; Lee, B. H.; Cho, S.; Lee, K. H.; Im, S.; Sung, M. M. Org. Electron. 2010, 12, 348. (14) See, K. C.; Feser, J. P.; Chen, C. E.; Majumdar, A.; Urban, J. J.; Segalman, R. A. Nano Lett. 2010, 10, 4664. (15) Malen, J. A.; Doak, P.; Baheti, K.; Tilley, T. D.; Majumdar, A.; Segalman, R. A. Nano Lett. 2009, 9, 3406. (16) Ke, S.-H.; Yang, W.; Curtarolo, S.; Baranger, H. U. Nano Lett. 2009, 9, 1011. (17) Yan, L.; Shao, M.; Wang, H.; Dudis, D.; Urbas, A.; Hu, B. Adv. Mater. 2011, 23, 4120. (18) Appelhans, D.; Ferse, D.; Adler, H.; Plieth, W.; Fikus, A.; Grundke, K.; Schmitt, F.; Bayer, T.; Adolphi, B. Colloids Surf., A 2000, 161, 203. (19) Sathyapalan, A.; Ng, S.; Lohani, A.; Ong, T.; Chen, H.; Zhang, S.; Lam, Y.; Mhaisalkar, S. Thin Solid Films 2008, 516, 5645. (20) Fabre, B.; Lopinski, G. P.; Wayner, D. D. M. Chem. Commun. 2002, 2002, 2904. (21) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G. A.; Nakatsuji, H.; Caricato, M.; Li, X.; Hratchian, H. P.; Izmaylov, A. F. et al. Gaussian 09 Revision A.02; Gaussian Inc.: Wallingford, CT, 2009. (22) Hu, W.; Jiang, J.; Nakashima, H.; Luo, Y.; Kashimura, Y.; Chen, K.-Q.; Shuai, Z.; Furukawa, K.; Lu, W.; Liu, Y.; Zhu, D.; Torimitsu, K. Phys. Rev. Lett. 2006, 96, 027801. (23) Herrmann, C.; Solomon, G. C.; Subotnik, J. E.; Mujica, V.; Ratner, M. A. J. Chem. Phys. 2010, 132, 024103. (24) Jiang, J.; Luo, Y. QCME-V1.0 (Quantum Chemistry for Molecular Electronics); Royal Institute of Technology: Sweden, 2005. (25) Wang, C.-K.; Fu, Y.; Luo, Y. Phys. Chem. Chem. Phys. 2001, 3, 5017. (26) Jiang, J.; Kula, M.; Luo, Y. J. Chem. Phys. 2006, 124, 034708. (27) Su, W. Y.; Jiang, J.; Lu, W.; Luo, Y. Nano Lett. 2006, 6, 2091. (28) Cao, H.; Jiang, J.; Ma, J.; Luo, Y. J. Am. Chem. Soc. 2008, 130, 6674. (29) Markussen, T.; Rurali, R.; Jauho, A.-P.; Brandbyge, M. Phys. Rev. Lett. 2007, 99, 076803. (30) Savić, I.; Mingo, N.; Stewart, D. A. Phys. Rev. Lett. 2008, 101, 165502. (31) Carrete, J.; Gallego, L. J.; Varela, L. M.; Mingo, N. Phys. Rev. B 2011, 84, 075403. (32) Datta, S. Quantum transport: atom to transistor, 2nd ed.; Cambridge University: New York, 2005. (33) Stillinger, F. H.; Weber, T. A. Phys. Rev. B 1985, 31, 5262. (34) Jian, Z.; Kaiming, Z.; Xide, X. Phys. Rev. B 1990, 41, 12915. (35) Mingo, N.; Stewart, D. A.; Broido, D. A.; Srivastava, D. Phys. Rev. B 2008, 77, 033418. (36) Guinea, F.; Tejedor, C.; Flores, F.; Louis, E. Phys. Rev. B 1983, 4397, 4397.

chemical methods. The phenyl group marks a significant 20% difference in the optimum ZT in the ohmic regime and a huge departure of almost 100% for ballistic coupling, by virtue of affecting the phononic and electronic transmission profiles to very different degrees. The almost infinite flexibility offered by organic molecules for substitution with different residues, different choices of the length of the oligomer, and so forth, therefore appear as a promising avenue for improving ZT even further. At the current stage of technological development, these methods are possibly more economical and easily scalable to industrial sizes than other alternative routes dependent on nanoscale material engineering.44



CONCLUSIONS In summary, we used ab initio DFT calculations, the nonequilibrium Green's function formalism, and transport theory based on the Boltzmann transport equation to study the thermoelectric transport behavior of a superlattice composed of self-assembled monolayers of a conducting polymer alternated with semiconducting layers of either pure silicon or a SiGe alloy. Our calculations reveal that, whereas when the inorganic part consists of pure silicon the thermoelectric behavior of the resulting structure is rather poor, an alloy with a percentage of Si around 50%, leading to a maximal configurational disorder, offers promising results. In particular, as regards electric transport, we found that the organic layers significantly reduce the conductivity with respect to that of bulk SiGe. Nevertheless, since we also detected that thermal conductivity is reduced by a much larger factor (resulting in values in the order of magnitude from 0.1 to 1 W/(m K)) the dimensionless thermoelectric figure of merit (ZT) of this kind of hybrid structure can improve upon that of SiGe alone by a factor of more than two. We achieved optimal values of ZT around 0.25 at room temperature (∼2.5 times larger than for bulk SiGe), which could make this kind of material competitive for practical thermoelectric applications if, as expected, they are easier to engineer and manufacture than other, more sophisticated nanostructures. Moreover, we showed that the addition of a phenyl ring to each polymer alters the value of ZT by about 20%, which suggests that the thermoelectric properties of these systems can be easily tailored and improved by chemical methods.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS J.C. thanks the Spanish Ministry of Education for an FPU grant. We are grateful to Ali Shakouri for helpful comments. We thank B. Hu, A. K.-Y. Jen, A. J. Epstein, J. Heremans, J. Goldberger, S. Banerjee, T. S. Fisher, D. Srivastava, and M. Muhammed, for discussions at the initial stages of this work.



REFERENCES

(1) Leclerc, M.; Najari, A. Nat. Mater. 2011, 10, 409. (2) Dubey, N.; Leclerc, M. J. Polym. Sci., Part B: Polym. Phys. 2011, 49, 467. (3) Sanchez, C. J. Mater. Chem. 2005, 15, 3557. (4) Mitzi, D. B. Chem. Mater. 2001, 13, 3283. 10885

dx.doi.org/10.1021/jp3025039 | J. Phys. Chem. C 2012, 116, 10881−10886

The Journal of Physical Chemistry C

Article

(37) Zhang, W.; Mingo, N.; Fisher, T. Numer. Heat Transfer, Part B 2007, 51, 333. (38) Mingo, N. Phys. Rev. B 2003, 68, 113308. (39) Krishnamurthy, S.; Sher, A.; Chen, A. Appl. Phys. Lett. 1985, 47, 160. (40) Slack, G. A.; Hussain, M. A. J. Appl. Phys. 1991, 47, 2694. (41) Mahan, G. D.; Sofo, J. O. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 7436. (42) Vining, C. B. J. Appl. Phys. 1991, 69, 331. (43) Chiritescu, C.; David, G.; Cahill, N. N.; Johnson, D.; Bodapati, A.; Keblinski, P.; Zschack, P. Science 2007, 315, 351. (44) Hochbaum, A. I.; Chen, R.; Diaz Delgado, R.; Liang, W.; Garnett, E. C.; Najarian, M.; Majumdar, A.; Yang, P. Nature 2008, 451, 163. (45) Schaffler, F. In Properties of Advanced Semiconductor Materials GaN, AlN, InN, BN, SiC, SiGe; Levinshtein, M. E., Rumyantsev, S. L., Shur, M. S., Eds.; Wiley: New York, 2001; p 149.

10886

dx.doi.org/10.1021/jp3025039 | J. Phys. Chem. C 2012, 116, 10881−10886