© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 3453wileyonlinelibrary.com

CO

MM

UN

ICATIO

N

Thin Film Thermoelectric Metal–Organic Framework with High Seebeck Coeffi cient and Low Thermal Conductivity

Kristopher J. Erickson , François Léonard , Vitalie Stavila , Michael E. Foster , Catalin D. Spataru , Reese E. Jones , Brian M. Foley , Patrick E. Hopkins , Mark D. Allendorf , and A. Alec Talin*

Dr. K. J. Erickson, Dr. F. Léonard, Dr. V. Stavila, Dr. M. E. Foster, Dr. C. D. Spataru, Dr. R. E. Jones, Dr. M. D. Allendorf, Dr. A. A. Talin Sandia National Laboratories Livermore , CA 94551, USA E-mail: aatalin@sandia.gov B. M. Foley, Prof. P. E. Hopkins Department of Mechanical and Aerospace Engineering University of Virginia Charlottesville , VA 22904 , USA

DOI: 10.1002/adma.201501078

creating a new pathway for charge transfer. Using this “Guest@MOF” concept we recently demonstrated that the electrical con-ductivity of the MOF Cu 3 (BTC) 2 (BTC = benzene tricarboxy-late), commonly known as HKUST-1, can be increased by >7 orders of magnitude with the introduction of the guest mol-ecule TCNQ (tetracyanoquinodimethane). [ 8 ] Molecular cluster calculations indicate that TCNQ can bind to two Cu(II) dimer units within the MOF pores, and a partial charge transfer of ≈0.3 e − between the TCNQ and the framework was estimated by infrared and Raman spectroscopic measurements. Tempera-ture dependent conductivity measurements indicated hopping carrier transport with a very low activation energy of ≈0.04 eV.

The creation of an electrical conductor from the insulating Cu 3 (BTC) 2 framework immediately raises the question: could this conducting Guest@MOF material be technologically useful? One application for which the unique combination of properties of conducting MOFs and Guest@MOFs could be advantageous is thermoelectricity. The effi ciency of thermo-electric energy conversion is determined by the dimensionless fi gure of merit ZT :

ZT

S T2σκ

=

(1)

where S is the Seebeck coeffi cient, σ is the electrical conduc-tivity, T is the average temperature, and κ is the thermal con-ductivity. Inorganic, low bandgap semiconductors such as Bi 2 Te 3 have adequate effi ciency ( ZT ≈ 1 at T ≈ 100 °C) for some niche applications, but have not been widely adopted because they can be diffi cult to deposit over complex and/or high sur-face area structures, are not eco-friendly, and are too expensive for many applications. As an alternative, conducting polymers have recently attracted much attention for thermoelectric appli-cations, motivated by their low material cost, ease of process-ability, nontoxicity, and low thermal conductivity. [ 9 ] MOFs share many of the advantages of all-organic polymers, including solu-tion processability and low thermal conductivity. [ 10 ] Additionally, MOFs and Guest@MOF materials offer higher thermal stability (up to ≈400 °C in some cases) and have long-range crystalline order which should improve charge mobility. A potential advan-tage of MOFs and Guest@MOF materials relative to all-organic polymers is their tremendous synthetic and structural versa-tility, which allows both the electronic and geometric structure to be tuned through the choice of metal and ligand. This could solve the long-standing challenge of fi nding stable, high ZT n-type organic semiconductors, as well as promote high charge

Metal–organic frameworks (MOFs) are extended, crystalline compounds consisting of metal ions interconnected by organic ligands, forming scaffolding-like structures that are sometimes referred to as “molecular tinker toys.” [ 1 ] These highly porous materials have attracted considerable attention for potential applications in gas storage, separation, and catalysis, [ 2 ] but their potential use in applications requiring electrical conductivity, such as microelectronics, photovoltaics, and thermoelectrics (TE), has received much less attention because most MOFs are electrical insulators. This results from the low extent of conju-gation in the organic linker groups and poor overlap between their π orbitals and the valence orbitals of the metal ions.

Recently, however, several MOFs with electrical conduc-tivity ranging from ≈10 −7 S cm −1 to ≈10 1 S cm −1 have been demonstrated using two approaches. [ 3–5 ] In the fi rst approach, electrical conductivity is realized by selecting combinations of metal ions and linkers designed to promote metal-ligand charge transfer, or that form framework topologies in which charge transfer between adjacent conjugated groups is enabled. For example, charge delocalization and conductivity along spe-cifi c metal-linker directions can be enhanced by using aromatic bridging ligands that interact covalently with the metal ions. This ‘‘through-bond’’ approach to achieve conductivity includes frameworks with metal ion−thiophenoxide linkages, wherein the sulfur atoms interact with the metal d-orbitals to form a delocalized charge network along the (–M–S–) chain, [ 4 ] Cu−Ni MOFs with redox-active linkers, [ 6 ] and Fe-triazolate frame-works. [ 7 ] Alternatively, conducting MOFs and covalent organic frameworks can be realized by π-stacking redox-active linkers such as pyrene and triphenylene, porphyrine, phthalocyanins, tetrathiafulvalene, and hexaaminotriphenylene. [ 5 ]

A second approach, which transforms an otherwise insu-lating framework into a conducting one, introduces redox-active guest molecules that bridge neighboring metal centers,

Adv. Mater. 2015, 27, 3453–3459

www.advmat.dewww.MaterialsViews.com

3454 wileyonlinelibrary.com © 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

CO

MM

UN

ICATI

ON mobility as a result of the long-range crystalline order inherent

in these materials. Herein we present the fi rst measurement of thermoelectric

behavior by a MOF, using electrically conducting TCNQ@Cu 3 (BTC) 2 thin fi lms deposited by a room-temperature, solu-tion-based method. These measurements reveal a large, posi-tive Seebeck coeffi cient, revealing conclusively that holes are the majority carriers in TCNQ@Cu 3 (BTC) 2 . An explanation for the positive Seebeck coeffi cient is proposed based on ab initio calculations, which predict that the Fermi level is located near the valence band of the MOF. In addition, we used time-dependent thermorefl ectance (TDTR) to measure the thermal conductivity of the fi lms, which is found to be low, possibly due to the presence of disorder, as suggested by molecular dynamics simulations. Our measurements of electrical conduc-tivity, thermal conductivity, and Seebeck coeffi cient, allow us to establish ZT , the thermoelectric fi gure of merit. Our results, combined with the recent advances in developing electrically conducting frameworks and the unprecedented synthetic fl exi-bility and chemical tailorability of Guest@MOFs, suggest a new path for the development of novel thermoelectric materials.

The Cu 3 (BTC) 2 fi lms, deposited by an automated layer-by-layer approach, are polycrystalline with a preferred (111) ori-entation. [ 11 ] The grain size ranges from tens of nanometers to ≈1 µm depending on the fi lm thickness and exact growth con-ditions, as shown in Figure 1 a (see X-ray diffraction (XRD) data in the Supporting Information). A MOF thin-fi lm device par-tially immersed in a TCNQ/methanol solution for infi ltration is shown in Figure 1 b. The deep blue–green color that Cu 3 (BTC) 2 acquires following immersion in the TCNQ/methanol solu-tion (Figure 1 c) is due to several broad absorption bands in the 600–1100 nm wavelength range that arise from ligand-to-metal charge transfer transitions occurring between the guest molecules and Cu 2+ ions in the framework. [ 8,12 ] The infi ltrated (region “A”) and noninfi ltrated (region “C”) portions are sepa-rated by a “boundary” area (region “B” in Figure 1 c,d) that extends along the upper edges of the specimen, due to wicking of the TCNQ/methanol solution along the container ridges that were used to hold the specimen in a vertical position. The TCNQ infi ltration and its interaction with the MOF were further confi rmed using Raman spectroscopy. Raman spectra (Figure 1 e) corresponding to the C≡N vibration (≈2300 cm −1 ) were collected across a portion of the specimen, as indicated by the yellow dots in Figure 1 d. The C≡N stretching frequency in TCNQ salts is sensitive to the degree of charge transfer; [ 13 ] since only two of the four C≡N groups are involved in bridging neighboring Cu 2+ ions in TCNQ@Cu 3 (BTC) 2 , [ 8 ] the C≡N vibra-tion in the infi ltrated region is split into two distinct peaks. Both peaks can still be seen in the transition region (“B”), but the intensity of the red-shifted C≡N vibration is greatly dimin-ished compared to region “A.”

Representative I – V measurements collected in the three regions are shown in Figure 1 f, where it can be seen that only region “A” shows meaningful conductivity. The linear I / V curve in the infi ltrated region is consistent with Ohmic conduction in TCNQ@Cu 3 (BTC) 2 . Contact resistance is small when com-pared with the thin fi lm resistance, as shown by the linear increase in resistance with channel length (see the Supporting Information).

The thermoelectric properties of the conducting portion of the fi lm were measured using Peltier temperature control ele-ments to create a temperature gradient, and an IR camera to measure the temperature gradient ( Figure 2 a,b). Plots of ther-movoltage versus Δ T measured at an average temperature of 40 °C and the Seebeck coeffi cient as a function of the average temperature are shown in Figure 2 c,d, respectively. The posi-tive value of the Seebeck coeffi cient indicates that holes are the majority carrier type for TCNQ@Cu 3 (BTC) 2 ; this is the fi rst direct, unambiguous measurement of carrier type in either a MOF or Guest@MOF material. Moreover, the measured See-beck coeffi cient of 375 µV K −1 at room temperature is compa-rable to that of the best organic [ 9 ] materials and exceeds that of Bi 2 Te 3 , [ 14 ] with the trend in Figure 2 d suggesting that it could reach larger values with increasing temperature.

We also measured the electrical conductivity as a function of temperature between 10 and 40 °C as shown in Figure 2 e. Although the temperature range for this measurement is lim-ited, the increase in conductivity with increasing temperature is consistent with a temperature-activated transport with an energy barrier of 0.052 eV, which is slightly above the 0.041 eV energy barrier we previously reported for similarly prepared fi lms on Si/SiO 2 substrates. [ 8 ] The power factor p S2σ= increases by more than a factor of two over the 30 K temper-ature range, with a value of 0.057 µW m −2 K −1 at room tem-perature. This value is about two orders of magnitude lower than the best organic and inorganic materials, as a result of the lower electrical conductivity of TCNQ@Cu 3 (BTC) 2 .

The positive Seebeck coeffi cient can be rationalized using the results of ab initio calculations, which suggest that the Fermi level in Cu 3 (BTC) 2 moves close to the valence band upon infi ltration with TCNQ molecules. The predicted electronic density of states (DOS) for Cu 3 (BTC) 2 , computed by density functional theory (DFT) using the hybrid functional HSE06, [ 15 ] is shown in Figure 3 . In the case of the uninfi ltrated MOF, the Fermi level ( E F ) lies within the large bandgap (few eV) of the insulating system, far away from the valence and conduction bands of Cu 3 (BTC) 2 . Upon infi ltration with TCNQ, the Fermi level moves to within a few tens of meV from the MOF valence band maximum. This is explained by the fact that the lowest unoccupied molecular orbital (LUMO) levels of TCNQ lie very close to the Cu 3 (BTC) 2 valence band (within ≈50 meV), a con-sequence of the large electron affi nity of TCNQ. This is evident from the calculated partial density of states (PDOS), obtained by projecting the states on the N and Cu atoms, as shown in Figure 3 b. The blue line in the fi gure indicates that the TCNQ LUMO levels are situated just above E F , whereas the Cu 3 (BTC) 2 valence band is located just below E F . The close proximity of the Fermi level to the Cu 3 (BTC) 2 valence band, as well as the fact that the TCNQ molecules are not expected to form an ordered structure with long-range order (which would lead to the formation of a new distinct band) implies hole transport in the MOF, explaining the positive Seebeck coeffi cient.

The thermal conductivity of the TCNQ@Cu 3 (BTC) 2 MOF sample was determined via TDTR measurements at a pump-frequency of 1.034 MHz. The data were fi t to a multilayer, cylindrically symmetric thermal diffusion model that accounts for pulse accumulation from the 80 MHz laser source and fur-ther modulation of the pump-path at pump frequency f . [ 16–18 ]

Adv. Mater. 2015, 27, 3453–3459

www.advmat.dewww.MaterialsViews.com

3455wileyonlinelibrary.com© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

CO

MM

UN

ICATIO

N

(20 different TDTR scans were performed to ensure repeat-ability in our measurements.) As discussed in detail in the Supporting Information, at a pump modulation frequency of f = 1.034 MHz we have pronounced sensitivity to the thermal conductivity of the TCNQ@Cu 3 (BTC) 2 , compared to other thermal resistances in our sample, [ 19 ] including the volumetric heat capacity of the TCNQ@Cu 3 (BTC) 2 , C MOF . This is exempli-fi ed in the inset of Figure 4 a, which shows typical TDTR data from the TCNQ@Cu 3 (BTC) 2 MOF sample, along with the best fi t of the thermal model using the thermal conductivity of the MOF, κ MOF , as the only fi tting parameter (all assumptions and

parameters used in the fi ts, along with thermal sensitivities, are discussed in the Supporting Information). Along with this best fi t, we show thermal model predictions that result from per-turbing either κ MOF or C MOF by 10%. Clearly, the measurement sensitivity to κ MOF is much greater than C MOF , indicating that our TDTR measurements are dominated by the thermal con-ductivity of the TCNQ@Cu 3 (BTC) 2 . As shown in Figure 4 a, the TDTR data yield a thermal conductivity that is relatively con-stant in the temperature range from 10 to 40 °C, with a room-temperature value of 0.27 ± 0.04 W m −1 K −1 for the TCNQ@Cu 3 (BTC) 2 fi lm grown on a Si/SiO 2 substrate. This thermal

Adv. Mater. 2015, 27, 3453–3459

www.advmat.dewww.MaterialsViews.com

-1.0

-0.5

0.0

0.5

1.0

Cu

rren

t (A

)

-10 -5 0 5 10Bias (V)

"A" "B" (x10

5)

"C" (x105)

500 nm

Raman

500 µm

. . . . . . .

‘B’

‘A’

Inte

ns

ity

(a

.u.)

2280224022002160Raman shift (cm

-1)

Region A Region B Region C

(a) (b)

‘A’

‘B’

‘C’

(c) (d)

(e) (f)

Figure 1. Electrically conductive MOF through TCNQ infi ltration. a) Cross-section SEM image of an infi ltrated MOF thin fi lm. b) Thin fi lm device on a quartz substrate in the TCNQ/methanol solution for infi ltration. c) Device after removal from the solution and drying. The infi ltrated regions are blue due to the interaction of the TCNQ with the MOF. Regions labeled A, B, and C correspond to infi ltrated, transition, and uninfi ltrated regions, respectively. Pairs of electrical contacts can be seen in the image, with probes contacting two of them for electrical measurements. d) Higher resolution image showing the electrodes in more detail. e) Raman spectra measured along the yellow dotted line in panel (d). f) Current–voltage characteristics of devices in regions A, B, and C. Meaningful conductivity is only observed in the infi ltrated region.

3456 wileyonlinelibrary.com © 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

CO

MM

UN

ICATI

ON

conductivity is comparable to the best thermoelectric poly-mers, [ 9,20 ] an order of magnitude smaller than Bi 2 Te 3 .

To understand the factors that dominate this low thermal conductivity, we used molecular dynamics simulations together with the Green–Kubo method to calculate the thermal conduc-tivity of crystalline HKUST-1 and TCNQ@Cu 3 (BTC) 2 . We fi rst benchmarked our approach by calculating the thermal conduc-tivity of crystalline MOF-5, which was previously measured [ 10 ] and modeled by molecular dynamics. [ 21,22 ] As discussed in the Supporting Information, we fi nd excellent agreement between our calculated thermal conductivity ( κ = 0.25 ± 0.03 W m −1 K −1 ) and these previous experimental and numerical results, both of which yielded a value of κ = 0.31 ± 0.02 W m −1 K −1 . Extending our computational approach to Cu 3 (BTC) 2 and TCNQ@Cu 3 (BTC) 2 , assuming on average one TCNQ molecule per pore (the experimentally observed concentration), we obtained a thermal conductivity of 0.58 W m −1 K −1 for the un-infi ltrated crystalline MOF and 3.84 ± 0.27 W m −1 K −1 for the infi ltrated crystalline MOF, as shown in Figure 4 b. The addition of TCNQ signifi cantly increases the thermal conductivity, which is, at least in part, due to the increased density of phonons in the low-frequency/high group-velocity range (see inset in Figure 4 b). Since the calculated values of the thermal conductivity for Cu 3 (BTC) 2 and TCNQ@Cu 3 (BTC) 2 , which are expected to sim-ulate a single crystal rather than a polycrystalline material, are both larger than our experimental value, it is likely that disorder

in the form of grain boundaries, molecules in the pores, or other phonon scatterers, signifi cantly decreases the thermal conductivity.

Inserting the experimental values we measured for S , σ (=0.45 S m −1 ), and κ in Equation ( 1) , we obtain ZT 7 10 5≈ × − at 298 K, with an increase to ZT 1.5 10 4≈ × − at 313 K, as shown in Figure 5 . These values are well below those of the best inorganic thermoelectric materials and the maximum ZT of ≈0.4 reported for the conducting polymer poly(3,4-ethylene-dioxythiophene) (PEDOT) doped with poly(styrenesulphonate) (PSS) [ 23,24 ] with the values S 70 V K , 101 5σ≈ μ ≈− S m −1 , andκ ≈ − −0.24W m K1 1 . While the MOF Seebeck coeffi cient is more than fi ve times larger than PDOT:PSS and the thermal conductivity comparable, it is clear that the MOF electrical conductivity is the main reason for the reduced performance. We expect that a signifi cant improvement can be achieved by reducing the number of grain boundaries in the material. This should also increase the thermal conductivity, however, as suggested by our molecular dynamics calculations; the competition between these two effects needs to be explored. Further increases in σ , however, will require improved elec-tronic coupling between the MOF metal centers. This may be achieved by exploring other guest molecules that will form a more covalent bond with the metal ions [ 12 ] or with MOFs that are inherently conducting. [ 4 ] We note, for example, that a MOF based on hexaaminotri phenylene linkers coordinates to

Adv. Mater. 2015, 27, 3453–3459

www.advmat.dewww.MaterialsViews.com

Figure 2. Thermoelectric characterization of the MOF thin fi lm devices. a) Schematic of the measurement. Two Peltier coolers are used to heat and cool the two sides of the substrate and an IR camera is used to measure the temperature profi le. The voltage generated between two of the metal contacts is measured using probes and a multimeter. b) Example IR image taken during one of the measurements. The electrical contacts appear cold in the image because of their different emissivity compared with the MOF. c) Measured voltage as a function of the applied temperature difference. The See-beck coeffi cient is extracted from the linear fi t. d) Measured Seebeck coeffi cient as a function of the average temperature. The dashed line is a linear fi t. e) Conductivity as a function of temperature. The dashed line is a linear fi t that could represent a thermally activated process with activation energy of 50 meV. f) Power factor as a function of temperature obtained from the Seebeck coeffi cient and the electrical conductivity. The dashed line is a linear fi t.

3457wileyonlinelibrary.com© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

CO

MM

UN

ICATIO

N

Ni(II) ions with an electrical conductivity of 4000 S m −1 was recently reported. [ 4 ] Although this electrical conductivity is still ≈50× lower than PDOT:PSS, it shows a path towards signifi -cant improvements in MOF electrical conductivity. The thermal conductivity and Seebeck coeffi cient of these conducting MOFs will have to be determined to assess their potential for TE applications.

In conclusion, we present the fi rst experimental observation of a thermoelectric effect in a MOF thin fi lm. A large and posi-tive Seebeck coeffi cient is obtained at room temperature, which can be explained by hole transport in the MOF valence band. We also described the fi rst thermal conductivity measurements in this system, with low values most likely due to the presence of disorder. Our results introduce MOFs as a new class of ther-moelectric materials, having properties that can be tailored and optimized through judicious use of the extensive synthetic toolbox available for this diverse and expanding category of supramolecular materials.

Experimental Section Growth and Characterization of MOF Thin Films : A Cu 3 (BTC) 2 fi lm with

200 nm nominal thickness was grown from the liquid phase on a fused quartz substrate 19 × 19 mm 2 , 0.55 mm thick according to a previously described procedure. [ 11 ] The quartz substrate had prepatterned Au pads with dimensions of 200 × 500 µm at spacings of 85, 110, 135, and 160 µm, which provided bottom electrical contacts to the MOF fi lm. Grazing incidence XRD measurements and scanning electron microscopy (SEM) imaging (see Figure S1, Supporting Information) indicated that the Cu 3 (BTC) 2 fi lms are polycrystalline with preferred orientation along the (111) direction. Prior to infi ltration, the residual solvent and precursors were removed by heating the specimen at 180 °C in vacuum (≈10 −6 Torr) for 1 h. After activation, the sample was immediately transferred to a glass bottle partially fi lled with a saturated solution of TCNQ in methanol. The specimen was held vertically in the bottle by side ridges,

Adv. Mater. 2015, 27, 3453–3459

www.advmat.dewww.MaterialsViews.com

Figure 3. Electronic density of states (DOS) calculated ab initio. a) For Cu 3 (BTC) 2 the Fermi level is in the electronic gap. b) For TCNQ@Cu 3 (BTC) 2 the Fermi level is near the MOF valence band. The blue lines indicate the partial density of states (PDOS) obtained by projecting the wavefunctions on the TCNQ N atoms, while the red lines indicate the PDOS obtained by a similar projection on the MOF Cu atoms. A broad-ening of 50 meV was used for the energy levels.

Figure 4. Measured and calculated thermal conductivity. a) TDTR data for TCNQ@Cu 3 (BTC) 2 (solid squares) as a function of temperature. The inset shows example TDTR data using a pump modulation frequency f = 1.034 MHz along with the best fi t of the thermal model using the thermal conductivity of the MOF, κ MOF , as the only fi tting parameter (solid black line). Along with this best fi t, we show example thermal model calculations when perturbing either κ MOF or C MOF by 10% (Δ κ MOF – black dashed line and Δ C MOF – red dotted line, respectively). The meas-urement sensitivity to κ MOF is much greater than C MOF , indicating that our TDTR measurements are dominated by the thermal conductivity of the TCNQ@Cu 3 (BTC) 2 . b) Data from molecular dynamics simulations show the time dependence of the integral of the heat fl ux autocorrelation, resulting in thermal conductivity values for Cu 3 (BTC) 2 and Cu 3 (BTC) 2 with 2 TCNQs per pore. The inset shows the phonon density of states for the two systems.

3458 wileyonlinelibrary.com © 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

CO

MM

UN

ICATI

ON

Adv. Mater. 2015, 27, 3453–3459

www.advmat.dewww.MaterialsViews.com

such that only roughly the lower half was immersed in the methanol/TCNQ solution, as shown in Figure 1 a. The specimen was removed after ≈24 h, rinsed with methanol, and dried under a stream of N 2 .

Raman Characterization : Raman spectra were collected using a system assembled in house, which uses a Princeton Instruments spectrograph and camera, a 532 nm laser excitation source focused into a ≈2 µm diameter spot, and ≈0.5 mW of laser power at the specimen. Several consecutive spectra collected at the same location remained essentially unchanged indicating that the laser was not modifying the MOF fi lm.

Electrical and Thermoelectric Measurements : Electrical conductivity was measured under DC bias using a computer-controlled voltage/current source with measurement capabilities (National Instruments) and a current amplifi er. To determine the Seebeck coeffi cient, the specimen was placed on a custom stage consisting of two Peltier heaters/coolers (TEC1-12710 TEC), which were used to establish a thermal gradient. The thermovoltage was then measured using a precision voltage meter (Agilent) while the local temperature was determined using an Inframeterics 760 IR thermal imaging camera. A schematic of the setup and a thermal image of the specimen are shown in Figure 2 a,b, respectively. The temperature was collected at positions 200 µm above and below the Au pads to avoid the effect of the Au pads (which have low emissivity Au), and the average of these two temperatures was recorded. The emissivity of the MOF fi lm was experimentally determined to be 0.80.

Thermal Conductivity Measurements : Thermal conductivity was measured using the TDTR technique on a 200 nm-thick TCNQ@Cu 3 (BTC) 2 MOF fi lm that was deposited, using a method similar to that discussed above, onto a p-Si wafer coated with 100 nm of thermal oxide and having patterned Au pads for conductivity measurements. In addition, 200 nm thick, 0.75 mm diameter Al dots were deposited on top of the TCNQ@Cu 3 (BTC) 2 MOF fi lm and used for the TDTR measurements, as shown in Figure 4 . Details of the TDTR technique and measurements can be found in the Supporting Information.

Thermal Conductivity Calculations : To create a model for molecular dynamics, we used DFT-relaxed super-cells of the MOF together with anharmonic covalent bonds, short-range Lennard–Jones and long-range Coulomb interactions. [ 25,26 ] After thermalizing the system to 300 K, the Green–Kubo method was used to calculate the thermal conductivity based on the time correlation of the fl uctuations in the equilibrium heat fl ux (see the Supporting Information for details).

Electronic Structure Calculations : We treated Cu 3 (BTC) 2 and TCNQ@Cu 3 (BTC) 2 with standard and hybrid DFT using periodic boundary conditions with a lattice constant of 2.63 nm, which matches the experimentally determined value for Cu 3 (BTC) 2 . The calculations included spin-polarization and were carried out using the VASP

code, [ 27 ] with Gamma-point-only sampling of the Brillouin zone. In the uninfi ltrated MOF case we used the experimentally deduced atomic structure of Cu 3 (BTC) 2 . As a representative case for the TCNQ-infi ltrated MOF, we considered a confi guration of two TCNQ molecules per half of the MOF large pores, each of them coordinated with two Cu dimers, and we included two water molecules for each of the remaining uncoordinated Cu dimers. In this case atomic forces were relaxed within LDA+U [ 28 ] (LDA: local density approximation) to better than 0.05 eV Å −1 accuracy. DOS calculations included exchange-correlation effects via the hybrid DFT HSE06 functional.

Supporting Information Supporting Information is available from the Wiley Online Library or from the author.

Acknowledgements The authors gratefully acknowledge Bernice Mills for providing the thermal imaging camera and help with refl ectivity measurements. B.M.F. and P.E.H. appreciate support from the United States Army Research Offi ce (W911NF-13–1–0378). This work was supported by the Sandia Laboratory Directed Research and Development (LDRD) Program. A.A.T. acknowledges partial support for the device fabrication work and writing of the manuscript by the Science of Precision Multifunctional Nanostructures for Electrical Energy Storage (NEES), an Energy Frontier Research Center funded by the U.S. DOE, Offi ce of Science, Offi ce of Basic Energy Sciences under award DESC0001160. Sandia National Laboratories is a multiprogram laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security Administration under Contract No. DE-AC04–94AL85000.Note: Units for conductivity on page 3456 were corrected on June 5, 2015, after initial publication online.

Received: March 4, 2015 Revised: April 4, 2015

Published online: April 28, 2015

[1] M. P. Suh , H. J. Park , T. K. Prasad , D.-W. Lim , Chem. Rev. 2011 , 112 , 782 .

[2] J. Lee , O. K. Farha , J. Roberts , K. A. Scheidt , S. T. Nguyen , J. T. Hupp , Chem. Soc. Rev. 2009 , 38 , 1450 .

[3] D. M. D’Alessandro , J. R. R. Kanga , J. S. Caddy , Aust. J. Chem. 2011 , 64 , 718 .

[4] L. Sun , T. Miyakai , S. Seki , M. Dinca , J. Am. Chem. Soc. 2013 , 135 , 8185 . [5] D. Sheberla , L. Sun , M. A. Blood-Forsythe , S. Er , C. R. Wade ,

C. K. Brozek , A. Aspuru-Guzik , M. Dinca , J. Am. Chem. Soc. 2014 , 136 , 8859 .

[6] Y. Kobayashi , B. Jacobs , M. D. Allendorf , J. R. Long , Chem. Mater. 2010 , 22 , 4120 .

[7] F. Gandara , F. J. Uribe-Romo , D. K. Britt , H. Furukawa , L. Lei , R. Cheng , X. F. Duan , M. O'Keeffe , O. M. Yaghi , Chem. Eur. J. 2012 , 18 , 10595 .

[8] A. A. Talin , A. Centrone , A. C. Ford , M. E. Foster , V. Stavila , P. Haney , R. A. Kinney , V. Szalai , F. El Gabaly , H. P. Yoon , F. Léonard , M. D. Allendorf , Science 2014 , 343 , 66 .

[9] M. Culebras , C. Gómez , A. Cantarero , Materials 2014 , 7 , 6701 . [10] B. L. Huang , Z. Ni , A. Millward , A. J. H. McGaughey , C. Uher ,

M. Kaviany , O. Yaghi , Int. J. Heat Mass Transfer 2007 , 50 , 405 . [11] V. Stavila , J. Volponi , A. M. Katzenmeyer , M. C. Dixon ,

M. D. Allendorf , Chem. Sci. 2012 , 3 , 1531 .

Figure 5. Measured ZT factor as a function of temperature for TCNQ@Cu 3 (BTC) 2 .

3459wileyonlinelibrary.com© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

CO

MM

UN

ICATIO

N

Adv. Mater. 2015, 27, 3453–3459

www.advmat.dewww.MaterialsViews.com

[12] M. D. Allendorf , M. E. Foster , F. Léonard , V. Stavila , P. L. Feng , F. P. Doty , K. Leong , E. Y. Ma , S. R. Johnston , A. A. Talin , J. Phys. Chem. Lett. 2015 , 6 , 1182 .

[13] J. S. Chappell , A. N. Bloch , W. A. Bryden , M. Maxfi eld , T. O. Poehler , D. O. Cowan , J. Am. Chem. Soc. 1981 , 103 , 2442 .

[14] N. Lowhorn , W. Wong-Ng , Z. Q. Lu , E. Thomas , M. Otani , M. Green , N. Dilley , J. Sharp , T. N. Tran , Appl. Phys. A 2009 , 96 , 511 .

[15] J. Heyd , G. E. Scuseria , M. Ernzerhof , J. Chem. Phys. 2003 , 118 , 8207 .

[16] P. E. Hopkins , J. R. Serrano , L. M. Phinney , S. P. Kearney , T. W. Grasser , C. T. Harris , J. Heat Transfer 2010 , 132 , 081302 .

[17] A. J. Schmidt , X. Chen , G. Chen , Rev. Sci. Instrum. 2008 , 79 , 114902 . [18] D. G. Cahill , Rev. Sci. Instrum. 2004 , 75 , 5119 . [19] J. Liu , J. Zhu , M. Tian , X. Gu , A. Schmidt , R. Yang , Rev. Sci. Instrum.

2013 , 84 , 104904 . [20] J. C. Duda , P. E. Hopkins , Y. Shen , M. C. Gupta , Appl. Phys. Lett.

2013 , 102 , 251912 .

[21] B. L. Huang , A. J. H. McGaughey , M. Kaviany , Int. J. Heat Mass Transfer 2007 , 50 , 393 .

[22] L. P. Han , M. Budge , P. A. Greaney , Comput. Mater. Sci. 2014 , 94 , 292 . [23] G. H. Kim , L. Shao , K. Zhang , K. P. Pipe , Nat. Mater. 2013 , 12 , 719 . [24] O. Bubnova , Z. U. Khan , H. Wang , S. Braun , D. R. Evans ,

M. Fabretto , P. Hojati-Talemi , D. Dagnelund , J. B. Arlin , Y. H. Geerts , S. Desbief , D. W. Breiby , J. W. Andreasen , R. Lazzaroni , W. M. M. Chen , I. Zozoulenko , M. Fahlman , P. J. Murphy , M. Berggren , X. Crispin , Nat. Mater. 2014 , 13 , 190 .

[25] W. D. Cornell , P. Cieplak , C. I. Bayly , I. R. Gould , K. M. Merz , D. M. Ferguson , D. C. Spellmeyer , T. Fox , J. W. Caldwell , P. A. Kollman , J. Am. Chem. Soc. 1995 , 117 , 5179 .

[26] L. Zhao , Q. Yang , Q. Ma , C. Zhong , J. Mi , D. Liu , J. Mol. Model. 2011 , 17 , 227 .

[27] G. Kresse , J. Furthmuller , Phys. Rev. B 1996 , 54 , 11169 . [28] S. L. Dudarev , G. A. Botton , S. Y. Savrasov , C. J. Humphreys ,

A. P. Sutton , Phys. Rev. B 1998 , 57 , 1505 .