Electron Solvation and the Unique Liquid Structure of a Mixed‐Amine

Angewandte Communications Chemie

International Edition:DOI:10.1002/anie.201609192 Liquid Phases German Edition:DOI:10.1002/ange.201609192 Electron Solvation and the Unique Liquid Structure of aMixed-Amine

Expanded Metal:The Saturated Li–NH3–MeNH2 System Andrew G. Seel, Helen Swan, Daniel T. Bowron, Jonathan C. Wasse,Thomas Weller, Peter P. Edwards,Christopher A. Howard, and Neal T. Skipper*

Abstract: Metal–amine solutions provideaunique arena in the solutions are electrolytic, whereby the metal valence which to study electrons in solution, and to tune the electron electrons have been ionized into solution and exist as solvated density from the extremes of electrolytic through to true electrons propagating between solvent cavities.[3] Increasing metallic behavior.The existence and structure of anew class of the concentration results in metallization in the liquid phase, concentrated metal-amine liquid, Li–NH3–MeNH2,ispre- which for the Li NH3 system occurs at amere 4mol%metal [2] À sented in which the mixed solvent produces anovel type of (MPM). Interestingly at lower temperatures,below TC = electron solvation and delocalization that is fundamentally 210 K, the point of the Mott-type metal–insulator transition different from either of the constituent systems.NMR, ESR, (MIT) is obscured by apronounced liquid–liquid phase and neutron diffraction allowthe environment of the solvated separation.[2] This illustrates that the localized and delocalized electron and liquid structure to be precisely interrogated. electron states do not readily co-exist, which is dramatically Unexpectedly it was found that the solution is truly homoge- manifested by the fact that the more concentrated metallic neous and metallic.Equally surprising was the observation of solution floats above the dilute electrolytic phase for T< [2,4] strong longer-range order in this mixed solvent system. This is Tc. Above 8MPM the solutions do not exhibit phase despite the heterogeneity of the cation solvation, and it is separation, appearing golden up to the concentration limit of [5] [6,7] concluded that the solvated electron itself acts as astructural 20 MPM, theexpanded metal Li(NH3)4. Theconcentra- template.This is aquite remarkable observation, given that the tion and temperature dependence of this liquid–liquid phase liquid is metallic. separation has been mapped through multi-element NMR spectroscopy.[8] Alkali metals demonstrate an exceptional solubility in NH3, Along with metal concentration, chemical tunability of yielding intensely colored conducting solutions that have the electronic properties of these systems can be achieved by [1,2] fascinated chemists since the time of Sir Humphry Davy. varying the amine.Lithium will also dissolve in MeNH2 to Varying the concentration of metal in these liquids dramat- yield asystem whereby solvated electrons transition to [3,9–12] ically alters the electronic,magnetic,and structural properties ametallic state. TheMIT in Li–MeNH2 occurs at of the solutions,and enables us to experimentally determine 15 MPM, anotably higher concentration than in Li–NH3. the manner in which liquid systems accommodate excess No liquid–liquid phase separation has been detected across electron density.Atalow concentration of metal/electrons, the full concentration range of Li–MeNH2,and the solution remains adeep blue,albeit with ametallic luster. The 1 1 conductivity of liquid Li(MeNH2)4 is 400 WÀ cmÀ (compared 1 1 [*] Dr.A.G.Seel, Prof. P. P. Edwards to 15000 WÀ cmÀ in Li(NH3)4), which is close to Motts Department of Chemistry,Inorganic Chemistry Laboratory minimum conductivity limit, demonstrating that the Li– UniversityofOxford MeNH2 system lies just on the metallic side of the MIT. South Parks Road, Oxford, OX1 3QR (UK) Thediffering metallic properties of Li–NH3 and Li– Dr.H.Swan, Dr.J.C.Wasse, Dr.T.Weller,Dr. C. A. Howard, MeNH2 are also reflected in their distinct liquid struc- Prof. N. T. Skipper [13,14] Department of Physics and Astronomy,University College London tures. All metallic metal amines share the trait of being [13–16] Gower Street, London, WC1E 6BT (UK) highly structured liquids, with distinct M(solv)–M(solv) E-mail:[email protected] correlations.Inboth Li–NH3 and Li–MeNH2 solutions,Liis Dr.H.Swan found to be four-coordinate,which has subsequently been National Nuclear Laboratory,Culham Science Centre found to be the case in the gas phase,[17–19] and through Abingdon, OX14 3DB (UK) computational investigation.[1,19–21] Thevolumetric expansion Dr.D.T.Bowron of these liquid metals across their insulator–metal transition is ISIS Spallation Neutron Source also accompanied by the appearance of void spaces,which is STFC RutherfordAppleton Laboratory presumably linked to the conduction electron density.InLi– Chilton, Didcot, OX11 0QX (UK) NH3 these voids take the form of channels between Li(NH3)4 Supportinginformation for this article can be found under: units,[13] whereas in Li(MeNH ) the voids are spatially http://dx.doi.org/10.1002/anie.201609192. 2 4 isolated.[14] This is concordant with magnetic measurements 2017 The Authors. Published by Wiley-VCH Verlag GmbH & Co. that suggest electronic conduction in Li(MeNH ) is via rapid KGaA. This is an open access article under the terms of the Creative 2 4 Commons AttributionLicense, which permits use, distribution and migration of electrons between these polaronic voids,which [22,23] reproduction in any medium, provided the original work is properly begin to localize in the solid state. cited.

Angew.Chem. Int.Ed. 2017, 56,1561 –1565 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim 1561 Angewandte Communications Chemie

Thequestion then is how amixture of Li(NH3)4 and

Li(MeNH2)4 will behave.Surprisingly,perhaps,these mixed amine solutions have only previously been studied by optical absorption towards the limit of infinite dilution.[24] Herein we

report the first studies of the liquid structure of Li-NH3-

MeNH2 in the concentrated regime at 20 MPM, an empirical

stoichiometry of Li(NH3)2(MeNH2)2.Wenote that after equilibration these solutions appear to the eye as homogenous red/bronze liquids (Figure 1), which upon dilution

7 Figure 1. Topleft: Li NMR for 20 MPM Li(NH3)2(MeNH2)2 Topright: Temperaturedependence of the 7Li Knight shift, Bottom left:Homog- enization of the 20 MPM system showing initial phase separation (left) and homogenous liquid (right). Bottom right:ESR line shape for

20 MPM Li(NH3)2(MeNH2)2.

Figure 2. Top: Experimentaltotal structure factors for 20 MPM 6Li– nat demonstrate apronounced liquid–liquid phase separation ND3–CD3ND2 and Li–ND3–CD3ND2 solutions. Middle:First-order Li whereby the lustrous phase floats above adeep-blue solution. differencestructure factor.Bottom:Corresponding Li partial pair correlation. We have observed this phase separation across avery large concentration range of Li, from 2–19 MPM. As confirmation of the homogeneity at 20 MPM, Figure 1shows that only 7 asingle feature is witnessed in the corresponding Li NMR Table 1: Isotopic substitutions in samples of 20 MPM Li–NH3–MeNH2. spectrum. This has proven to be an extremely sensitive Sample Li NH3 MeNH2 [8] indication of homogeneity in previous metal–amine studies. nat 7 A Li ND3 CD3ND2 Interestingly,the Li peak appears at higher ppm than both nat [8,12,25] B Li NH3 CD3NH2 Li(NH3)4 and Li(MeNH2)4, suggesting ahigher conduc- nat C Li [NH3 :ND3][CD3NH2 :CD3ND2] tion electron spin-density at the Li nucleus (or indeed larger nat D Li ND3 CH3ND2 nat molar spin susceptibility) in the mixed amine than in either of E Li ND3 [CD3ND2 :CH3ND2] 7 nat the parent amines.The temperature dependence of the Li F Li NH3 CH3NH2 [8] nat Knight shift and the Dysonian lineshape of the ESR are also G Li [NH3 :ND3][CD3ND2 :CH3NH2 :CD3NH2 :CH3ND2] 7 indicative of ahomogeneous metallic system.[22,26,27] H Li ND3 CD3ND2 Neutron diffraction is auniquely powerful method for determining the liquid structure of metal amines,allowing for anumber of isotopically distinct samples to be measured the homogeneity of the 20 MPM solution. Thepre-peak is (defined in the Experimental Section below). Figure 2 astructural feature witnessed in solvated-electron systems as presents an example fit to an experimental total structure they transition to the metallic state,and arises owing to strong factor, F(Q), from samples Aand HinTable 1. Also shown is intermediate range ordering in the liquid.[13–15] Integration of

the difference spectrum [F6 (Q) Fnat (Q)] and its Fourier the principal peak in the DG (r)distribution corresponds to Li À Li Li transform, the Li-centered total pair correlation function, the coordination number of Li, and is consistent with 1 DGLi(r). Thesingle principal peak at 1.75 À ,and impor- approximately 4solvent molecules,asreported previously 1 [13,14] tantly the existence of asharp pre-peak at 0.97 À ,attest to for other Li–RNH2 systems.

1562 www.angewandte.org 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Angew.Chem. Int.Ed. 2017, 56,1561 –1565 Angewandte Communications Chemie

To determine the spatial arrangement of lithium and Importantly,the EPSR model also reveals the statistical amine species in the system, the empirical potential structure distribution of ammonia and methylamine molecules in the + refinement (EPSR) technique was invoked to produce [Li(NH3)n(MeNH2)m] tetrahedral complexes (Supporting astructural model that was refined simultaneously to all the Information, Figure S2). This distribution shows that for isotopically unique experimental F(Q)functions (Table 1).[28] around 40%ofcations (n,m)is(2,2), for around 50%itis Figure 3shows the EPSR fits to the experimental data, (1,3) or (3,1), and for the remaining 10%itis(0,4) or (4,0). with the EPSR model fitting well the experimentally obtained Thecation solvation is therefore significantly heterogeneous, partial structure factors.The corresponding real-space distri- and it is all the more surprising that we observe asharp bution of solvent molecules around the lithium ions is given in diffraction pre-peak in our data, indicative of longer-range order in the liquid. In the absence of asingle cationic motif, we must conclude that the solvated electron itself acts as astructural template that drives homogenization. This hypothesis is consistent our observation of asingle 7Li Knight shift (Figure 1), and is truly remarkable given that the system is metallic. Themanner in which excess electrons are accommodated can be determined by examining the liquid structures for void regions:spherical volumes with circa 2.5 radii containing no atoms.Previous studies have demonstrated the existence of these regions in each of the metallic metal amines,with stark differences in their size and orientation, which correlate with their differing metallic properties.Inthe highly con-

ducting Li–NH3 liquid these voids are centered above the

protons of the NH3,forming channels between solvated

Li(NH3)4 units (hence the moniker expanded metal). In the Figure 3. Experimental (black) and modeled (red) partial structure poor metal, Li–MeNH2,the cavities are spatially isolated factors for 20 MPM Li–NH3–MeNH2. from one another,removed from the primary solvation environment of Li and bounded by the methyl groups of

coordinated MeNH2.The void–void distribution function in Figure 4. TheLiNdistance for both NH and MeNH is our mixed amine solution is presented in the Supporting À 3 2 about 2.0 ,inclose agreement with that observed for both Information, Figure S3, and shows weak inter-void correla-

Li(NH3)4 and Li(MeNH2)4.Integration of this first solvation tions beyond around 4 . shell gives an average co-ordination number of 2.01 and 1.78 Figure 5shows the orientation of void regions relative to for Li–MeNH2 and Li–NH3,respectively.The orientational an NH3 and MeNH2 molecule in the 20 MPM Li–NH2– distribution of the coordinated solvent about Li can be MeNH2 system. It is seen that the voids (and conduction extracted from the EPSR model (Figure 4a–d). Only asmall electron density) lie around the circumference of NH3 and distortion from tetrahedral geometry for agiven Li(solv) unit is above the amine group of MeNH2.This structure is distinctly found, with the MeNH2 ligands acting to compress the different to either the Li–NH3 or Li–MeNH2 system, and tetrahedra slightly along the tetrahedral C2 axis.The dipole indicates that the electron density is drawn closer to Li, moment of NH3 is orientated directly towards the cation, around the circumference of the solvating NH3/MeNH2 whereas the Li–MeNH2 angle is about 1098.The relative molecules. orientation of neighboring Li(solv) units is not isotropic,with In conclusion, we have discovered that the mixed amine the vertex (that is,the NH3/MeNH2 molecules) of one system Li(NH3)2(MeNH2)2 forms ahighly structured homo- tetrahedral unit approaching the face and edges of the genous liquid metal. Each cation is tetrahedrally coordinated adjacent tetrahedral unit. by approximately four solvent molecules,but there is

significant heterogeneity in the NH3 :MeNH2 ratio around

Figure 4. Left:Li-centered partial distribution functions for 20 MPM

Li–NH3–MeNH2.Right:Li-centered three-dimensional configurations for radial range r = 1.5–2.5 ,80% probabilities:a)Li N(NH ), b) Li Figure 5. Void regions relative to the NH (left) and MeNH groups À 3 À 3 2 N(MeNH ), c) Li-N-C (MeNH ), d) Li N r =2.5–6.0 . (right) in 20 MPM Li–NH –MeNH . r = 2.4–5.4 . 2 2 À 3 2

Angew.Chem. Int.Ed. 2017, 56,1561 –1565 2017 The Authors. Published by Wiley-VCH VerlagGmbH &Co. KGaA,Weinheim www.angewandte.org 1563 Angewandte Communications Chemie

individual Li+ ions.This lack of asingle cationic motif makes Acknowledgements our observation of strong intermediate-range order in the solutions particularly surprising.Even though the system is Theauthors would like to thank the EPSRC,STFC Ruth- metallic,wetherefore turn to the solvated electron as erford Appleton Laboratory,and the ISIS Pulsed Neutron apotential structural template.Indeed, we find that the and Muon source for access to the SANDALS instrument. We system appears unique amongst current expanded metal– also thank Mark Kibble and Chris Goodway for their aid in amine solutions for the manner in which the excess electron experimental setup. density is accommodated. This is drawn closer to the lithium cations than in either the constituent lithium–ammonia or lithium–methylamine solutions,and we observe only asingle Conflict of interest Knight shift on the lithium nuclei. Mixed-solvent metal– amine solutions therefore present anew set of challenges for Theauthors declare no conflict of interest. our understanding of electron solvation, and the nature of the concomitant phase separation of the localized and delocalized Keywords: amines ·conducting materials ·electron transfer · electronic states. electronic structure ·phase transitions

Howtocite: Angew.Chem. Int. Ed. 2017, 56,1561–1565 Experimental Section Angew.Chem. 2017, 129,1583–1587 H/D and 7Li/natLi isotopic substitution experiments (where nat denotes natural abundance)were performed using the SANDALS [1] E. Zurek, P. P. Edwards,R.Hoffmann, Angew.Chem. Int. Ed. diffractometer at the ISIS Spallation Neutronand Muon Source, 2009, 48,8198 –8232; Angew.Chem. 2009, 121,8344 –8381 and UK.[29] Aclean sample of Li metal was loaded under argon into referencestherein. asealed flat-plate null-coherent scattering Ti/Zr container,with [2] J. C. Thompson, Electrons in Liquid Ammonia,Oxford Univer- sample and wall thicknesses of 1mm. Thecontainer and sample were sity Press,Oxford, 1976. 5 attached to astainlesssteel gas-rig and evacuated to 10À mbar.To [3] K. Maeda, M. T. J. Lodge,J.Harmer,J.H.Freed, P. P. Edwards, prepare asample of 20 MPM Li–NH3–MeNH2,anequimolar ratio of J. Am. Chem. Soc. 2012, 134,9209 –9218. NH3 :MeNH2 were premixed to the exact volume required, and cryo- [4] P. P. Edwards,M.J.Sienko, J. Am. Chem. Soc. 1981, 103,2967 – pumped onto the Li sample at 60 K. This avoided any problem in the 2971. different boiling temperatures of the amines.The sample was then [5] S. Hartweg,A.H.C.West, B. L. Yoder,R.Signorell, Angew. isolated and warmed to 240 K. Thesamples were monitored for Bragg Chem. Int. Ed. 2016, 55,12347 –12350; Angew.Chem. 2016, 128, reflections in the warming process to ensure that the solutions were 12535 –12538. fully homogenized. Data were collected for aperiod of 8h,and [6] R. M. Ibberson, A. J. Fowkes,M.J.Rosseinsky,W.I.F.David, spectra were corrected for background, multiple scattering and P. P. Edwards, Angew.Chem. Int. Ed. 2009, 48,1435 –1438; absorption following standard protocols for SANDALS data. Angew.Chem. 2009, 121,1463 –1466. Themeasured total structure factor for agiven system comprising [7] E. Zurek, X. Wen, R. Hoffmann, J. Am. Chem. Soc. 2011, 133, n chemical species can be expressed as in Eq. (1): 3535 –3547. [8] M. T. J. Lodge,P.Cullen, N. H. Rees,N.Spencer, K. Maeda, J. R. n n F Q a 1 b 1cacbbabb Sab Q 1 1 Harmer,M.O.Jones,P.P.Edwards, J. Phys.Chem. B 2013, 117, ð Þ¼ ¼ ¼ ½ ð ÞÀ ð Þ X X 13322 –13334. where ci denotes the atomic fraction of species i and bi is the bound [9] T. Toma, Y. Nakamura, M. Shimoji, Philos.Mag. 1976, 33,181 –

coherent neutron scattering length for that species. Sab(Q)isthe 187. Faber–Zimanpartial structure factor,and Q the scattering vector.The [10] R. Hagedorn, M. J. Sienko, J. Phys.Chem. 1982, 86,2094 –2097. detailed method for extraction the partial structure factor from [11] P. P. Edwards,J.R.Buntaine,M.J.Sienko, Phys.Rev.B1979, 19, isotopically varying F(Q)values is detailed elsewhere. F(Q)isthe 5835 –5846. Fourier transform of the total pair correlationfunction G(r). The [12] Y. Nakamura, M. Niibe,M.Shimoji, J. Phys.Chem. 1984, 88, specific isotopic substitutions for prepared samples are given in 3755. Table 1. [13] H. Thompson, J. C. Wasse,N.T.Skipper,S.Hayama,D.T. Theempirical potential structure refinement (EPSR) method was Bowron, A. K. Soper, J. Am. Chem. Soc. 2003, 125,2572 –2581. used to model the measured neutron total scattering data.[28] A [14] S. Hayama, J. C. Wasse,A.K.Soper,N.T.Skipper, J. Phys.

structural model comprising 2000 species (Li and molecular NH3, Chem. B 2002, 106,11–14.

MeNH2)issimultaneously refined to each of the isotopically unique [15] J. C. Wasse,S.Hayama, N. T. Skipper,C.J.Benmore,A.K. experimental F(Q)spectra (samples A–H in Table 1), enabling Soper, J. Chem. Phys. 2000, 112,7147. arigorous constraint to the resultant structural model, and extraction [16] J. C. Wasse,C.A.Howard, H. Thompson, N. T. Skipper,R.G.

of atom specific pair-correlationfunctions, Gab(r). Theinter-atom Delaplane, A. Wannberg, J. Chem. Phys. 2004, 121,996. potentials in EPSR are comprised of aLennard-Jones potential and [17] T. E. Salter,V.A.Mikhailov,C.J.Evans,A.M.Ellis, J. Chem. Coulombicterm, and is iterated throughoutthe simulation until Phys. 2006, 125,034302. asuitable convergence to experimental F(Q)spectra is reached.The [18] T. E. Salter,A.M.Ellis, J. Phys.Chem. A 2007, 111,4922. seed potentials used in this work were the same as used previously for [19] T. E. Salter,A.M.Ellis, J. Chem. Phys. 2007, 127,144314. metal–amine liquid structure studies.[12–15] [20] J. Kohanoff,F.Buda, M. Parrinello, Phys.Rev.Lett. 1994, 73, NMR spectra were recorded on aBruker AVIIIHD 400 MHz 3133 –3136.

spectrometer, referenced to LiCl in D2O. ESR was performedatX- [21] T. Sommerfeld, K. M. Dreux, J. Chem. Phys. 2012, 137,244302. band frequencies on aBruker EMXmicro spectrometer. Samples for [22] P. P. Edwards,A.Lusis,M.J.Sienko, J. Chem. Phys. 1980, 72, each were prepared in the manner outlined above,within spectrosil 3103 –3112. quartz tubes that were subsequently sealed. All handling and transfer [23] A. M. Stacy, D. C. Johnson, M. J. Sienko, J. Chem. Phys. 1982, 76, of samples was conducted under cryogenic conditions. 4248 –4254.

1564 www.angewandte.org 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Angew.Chem. Int.Ed. 2017, 56,1561 –1565 Angewandte Communications Chemie

[24] C. M. Stupak, T. R. Tuttle,S.Golden, J. Phys.Chem. 1984, 88, [28] A. K. Soper, Phys.Rev.B2005, 72,104204. 3804 –3810. [29] http://www.isis.stfc.ac.uk/instruments/sandals/. [25] D. Holton, P. P. Edwards,W.McFarlane,B.Wood, J. Am. Chem. Soc. 1983, 105,2104 –2108. [26] A. M. Stacy,P.P.Edwards,M.K.Sienko, J. Solid State Chem. Manuscript received:September 20, 2016 1982, 45,63–70. Revised: November2,2016 [27] P. Damay,M.J.Sienko, J. Phys.Chem. 1975, 79,3000 –3003. Final Article published: January 10, 2017

Angew.Chem. Int.Ed. 2017, 56,1561 –1565 2017 The Authors. Published by Wiley-VCH VerlagGmbH &Co. KGaA,Weinheim www.angewandte.org 1565