Atomic-Resolution Transmission Electron Microscopic Movies for

Article pubs.acs.org/accounts

Atomic-Resolution Transmission Electron Microscopic Movies for Study of Organic Molecules, Assemblies, and Reactions: The First 10 Years of Development Published as part of the Accounts of Chemical Research special issue “Direct Visualization of Chemical and SelfAssembly Processes with Transmission Electron Microscopy”. Eiichi Nakamura* Department of Chemistry, The University of Tokyo, Bunkyo-ku, Tokyo 113-0033, Japan S Supporting Information *

CONSPECTUS: A molecule is a quantum mechanical entity. “Watching motions and reactions of a molecule with our eyes” has therefore been a dream of chemists for a century. This dream has come true with the aid of the movies of atomicresolution transmission electron microscopic (AR-TEM) molecular images through real-time observation of dynamic motions of single organic molecules (denoted hereafter as single-molecule atomic-resolution real-time (SMART) TEM imaging). Since 2007, we have reported movies of a variety of single organic molecules, organometallic molecules, and their assemblies, which are rotating, stretching, and reacting. Like movies in the theater, the atomic-resolution molecular movies provide us information on the 3-D structures of the molecules and also their time evolution. The success of the SMART-TEM imaging crucially depends on the development of “chemical fishhooks” with which fish (organic molecules) in solution can be captured on a single-walled carbon nanotube (CNT, serving as a “fishing rod”). The captured molecules are connected to a slowly vibrating CNT, and their motions are displayed on a monitor in real time. A “fishing line” connecting the fish and the rod may be a σ-bond, a van der Waals force, or other weak connections. Here, the molecule/CNT system behaves as a coupled oscillator, where the low-frequency anisotropic vibration of the CNT is transmitted to the molecules via the weak chemical connections that act as an energy filter. Interpretation of the observed motions of the molecules at atomic resolution needs us to consider the quantum mechanical nature of electrons as well as bond rotation, letting us deviate from the conventional statistical world of chemistry. What new horizons can we explore? We have so far carried out conformational studies of individual molecules, assigning anti or gauche conformations to each C−C bond in conformers that we saw. We can also determine the structures of van der Waals assemblies of organic molecules, thereby providing mechanistic insights into crystal formationphenomena of general significance in science, engineering, and our daily life. Whereas many of the single organic molecules in a vacuum seen by SMART-TEM are sufficiently long-lived for detailed studies, molecules with low ionization potentials (8 eV) than that of CNT (5 eV) remain stable for a time long enough for observation at 60−120 kV acceleration voltage, as they are not oxidized by the CNT radical cation. Alternatively, the reaction may have taken place via an excited state of a molecule produced by energy transfer from CNT possessing excess energy provided by the electron beam. SMART-TEM imaging is a simple approach to the study of the structures and reactions of molecules and their assemblies and will serve as a gateway to the research and education of the science connecting the quantum mechanical world and the real world.

1. INTRODUCTION

boundary behave probabilistically and form nonequilibrium and nonperiodic molecular systems. Such systems are difficult to analyze by conventional tools based on the data averaged over molecules or time and often both (Figure 1). This Account

Regardless how precisely designed, molecules do not function as we wish. This is an everyday problem for chemists stemming from the lack of fundamental understanding of the molecular systems at the nanoscale that connect the molecular world and the real-world chemistry made of the Avogadro number of molecules. Molecules in nanoscale liquid phase, solid, and © 2017 American Chemical Society

Received: February 6, 2017 Published: May 8, 2017 1281

DOI: 10.1021/acs.accounts.7b00076 Acc. Chem. Res. 2017, 50, 1281−1292

Article

Accounts of Chemical Research

Figure 2 illustrates two frames out of a 43-s long movie of a dialkylcarborane 1 in a single-walled carbon nanotube (CNT). Does it not look like an eel caught in an “eel trap”? The carborane group was installed as a marker to uniquely identify the specimen molecules. Each frame of the movie was recorded with an exposure time of 0.5 s on a charge-coupled device or more recently a complementary metal oxide semiconductor (CMOS) device that captures the molecular motions as they happen (i.e., real time). Some images were blurred because the molecule moved during the 0.5-s exposure time. This single-molecule atomic-resolution real-time TEM (SMART-TEM) movie provided the first demonstration of the ability of AR-TEM to decipher the structure, motions, and reactions of a single organic molecule, an ultimate microanalysis. The information obtained from the movie during the lapse of time is of the same dimensional quality as black and-white movies in theaters; that is, a series of the two-dimensional pictures reveals the threedimensional features of the actors and actresses and the time evolution of their behavior, a pseudo-four-dimensional quality. It is amusing to note in passing that sculptures are threedimensional but lack time information. We have so far filmed movies of hydrocarbons,2 amide,3 biotin,4 van der Waals molecular clusters,5 chemical reactions,6 and metal catalysis.7 Some of these movies were collected and published recently.8,9 It is important to note at this juncture that the CNT underwent frequent and seemingly random thermal vibration of sub-nanometer magnitude like a waving pole, but these events were eliminated visually from the movies by normalization for the position of the CNT so that we can focus on the molecular motions. The motions of CNTs are an important element of the SMART-TEM studies because they cause motions of the attached molecules that occurred randomly or probabilistically.

Figure 1. AR-TEM movies for studies of time evolution and inhomogeneity of molecular systems.

describes our effort to establish the molecular movie images obtained by atomic-resolution transmission electron microscopy (AR-TEM) as an analytical tool to address this bottleneck of research through real-time observation of dynamic motions and chemical reactions of individual molecules and molecular assemblies. AR-TEM has so far been largely alien to the molecular science community because of the belief widely shared by the TEM community that molecules decompose rapidly under TEM conditions. In contradiction to this conventional wisdom, we have clarified in the past 10 years that the TEM molecular movies provide information not only on the static structures of organic and inorganic molecules but also on their dynamic motions and reactivity at atomic resolution, giving us a glimpse of the quantum mechanical world of molecular systems. In February 2007, we reported the AR-TEM movies of the conformational change of a single hydrocarbon molecule.1

Figure 2. SMART-TEM images of a wobbling ortho-carborane 1 bearing two C22H45 chains in a CNT of 1.2 nm diameter (thermal vibration of CNT is normalized). Hydrogen, white; boron, pink; carbon, gray. Adapted with permission from ref 1. Copyright 2007 The American Association for the Advancement of Science. See SI for a movie. 1282


Article


Figure 3. Metal atom on CNH. (a) A model of CNH bearing one ferrocene and five phenyl groups. Iron atom in red. (b) Scanning electron microscopic image of CNH aggregates (circle). (c) TEM image of a single Gd atom attached to an oxidized end of CNH ((5−10) × 104 electrons/(nm2·s) at 120 kV). The arrow indicates the motion of the metal atom during imaging. Adapted with permission from ref 16. Copyright 2004 National Academy of Sciences.

Figure 4. Fishhook concept in SMART-TEM technology and an AR-TEM instrument (JEM-ARM200F, JEOL).

foundation of single atom imaging using TEM13 and later discovered CNTs.14 Even Iijima, however, could not find any trace of the structure, and told me, “Organic molecules are known to be unstable under the TEM conditions, and it is natural that we could not see them.”15 His remark was burned in my mind. Then, we set up a less ambitious project, imaging of a Gd(III) ion bonded via COO−/Gd3+ ionic interactions to the oxidized tip of a CNH (Figure 3c); in 2002, we obtained TEM movies of Gd atoms moving slowly without any sign of the loss of the COO−/ Gd3+ linkage.16 The Gd atoms were identified by electron energy loss spectroscopy at nanometer resolution. The clear (i.e., no blurring) image of the Gd atom in Figure 3c is particularly notable, meaning that the atom stands still during the imaging time of 0.5 s. With only these data in hand, I started the Nakamura Functional Carbon Cluster ERATO project in 2004,17 which allowed me to fulfill my dream through a combination of the synthetic expertise of myself and Dr. H. Isobe, and electron microscopic expertise of Dr. K. Suenaga.18 We then shared the same concerns with the skeptical reviewers as to the radiation damage19,20 and the rapid molecular motions and built a TEM instrument equipped with a helium-cooled sample stage and an aberration corrector (JEM-2100FC). While waiting for com-

Retrospectively, the referees’ comments on our 2007 publication1 provided valuable guidelines for our later studies. One pro: “The ability to image conformations of individual small molecules is the “holy grail” of microscopy, and the authors present a convincing case that they have managed to do so.” Cons: “It is not clear what new horizons such imaging techniques will allow us to explore,” “Why does this technique work at all?”, and “The TEM data are heroic and very new. However, the interpretation of those incredible images is not convincing at best and dead wrong at worst.” For the past 10 years, we have striven to answer these important questions. In this Account, I describe how I came across the idea of SMART-TEM imaging, the new horizons to be explored, and some considerations on the chemistry of TEM imaging of organic molecules.

2. BACKGROUND In the 1990s, my interest in in silico analysis of metal catalysis10 led me to dream about watching chemical reactions of real molecules instead of computational virtual images. In 2000, I asked Prof. S. Iijima to take a look at an organometallic architecture11 built on the tip of a carbon nanohorn (CNH, a seaurchin-like aggregate of tapered CNTs; Figure 3a and 3b).12 Erwin Müller in the 1950s and Albert Crew in the 1970s developed methods to observe atoms,15 and Iijima laid the 1283


Article

Accounts of Chemical Research 3.2. Covalent Fishhook

pletion of this instrument, we found a miracle that led us down the unbeaten track at the interface of chemistry and electron microscopy. Initial trials suggested that single organic molecules are stable and can be seen even at room temperature (Figure 2).

The amine groups on amino CNHs25,5 serve as a chemical fishhook (Figure 6, bottom left). This substance has a formula of C70,000,000N100,000H200,000−C170,000,000N300,000H600,000 and a diameter of 50−100 nm isolable as a black powder by filtration. We can fix a molecule (e.g., 3) to amino CNH using standard amide forming reactions (Figure 6, bottom left), which allows us to deal with molecules of any size attach to the CNT surface. Figure 6 illustrates how one starts from the millimeter-sized TEM grid down by a step of 10 or 20 times to arrive at the quantum mechanical world of a tetracyclic aromatic molecule 4.

3. OUR METHODOLOGY: CHEMICAL FISHHOOK AND EEL TRAP To address our goal, we first needed to capture the molecules of our interest in solution and bring them into the nanoscale

3.3. Ionic, Hydrogen-Bonding, Coordination, And Supramolecular Fishhooks

An “ionic hook” is illustrated by the simplest example in Figure 3c, where the Gd(III) ion was trapped by CNH carboxylate anions in methanol and brought onto a TEM grid. A great variety of combinations of cations and anions are available for the design of ionic fishhooks. We can also consider “supramolecular hooks” utilizing van der Waals forces,5 hydrogen bonds, and a variety of weak interactions. Figure 5. An alkyl fullerene molecule (2) attached at a broken end of a CNT with its fullerene head partially encapsulated in CNT and its alkyl tail exposed in a vacuum. Adapted with permission from ref 24. Copyright 2008 Nature Publishing Group.

4. PROBABILISTIC NATURE OF SINGLE MOLECULAR IMAGING Interference is a fundamental feature that characterizes the wave behavior of electrons,26 and therefore passage of electrons near a molecule produces an interference image of the molecule, which is processed though a lens system to obtained a magnified image on a pixel detector (Figure 7a; for example, the de Broglie wavelength of an electron (λ) is 3.35 pm at 120 kV, while the C− C bond length of benzene is 140 pm). A probabilistic behavior is another feature of electrons as particles, making the image interpretation a complex issue. We first need to consider the time scale of the TEM imaging. Because the 120 kV voltage accelerates electrons to 59% of the speed of light, an electron passes by a carbon atom in 10−18 s, and hence each electron carries an attosecond time scale information on the location of the atoms in a molecule. However, each electron hitting the detector produces only one dot on a pixel, and therefore, to obtain an interference image, we need many more electrons that collectively form the molecular image on the detector. This probabilistic behavior of electrons was elegantly demonstrated by Tonomura in the

viewing field of TEM. For this molecular fishing, we conceived an “eel trap” and a “fishhook” using CNTs as a “fishing rod” (Figure 4); CNTs and their tapered variant CNHs are so far the best fishing rods because of their robustness and near TEM transparency and functionalizability at molecular precision.21 Their conductivity prevents sample charging. A single graphene sheet may also serve as a viable substrate, although it may be a little too flexible for sample manipulation.22 3.1. van der Waals “Eel Trap”

As previously reported for [60]fullerene encapsulation,23 and for trapping of a hydrocarbon 1 (Figure 2), we can trap molecules in a CNT. However, this method is only useful for solvophobic small molecules that have an affinity to the CNT interior. An interesting view of an alkyl fullerene (C10H21(C60H), 2) hanging on an open end of a CNT is shown in Figure 5.

Figure 6. An illustration of the sampling of a Y-shaped single molecule 3 on a nanocarbon substrate (4). Adapted with permission from ref 5. Copyright 2012 Nature Publishing Group. See SI for a movie. 1284


Article


Figure 7. Quantum mechanical nature of molecular imaging. (a) Interference image formation by fast electrons (120 kV) that pass near a 1 nm sized molecule in 10−18 s with an interval of 10 μm (dose of 1 × 105 electrons/(s·nm2)) and exhibit wave/particle behavior. The aromatic C−C bond length in 4 is ca. 140 pm, and the de Broglie wavelength of an electron (λ) is 3.35 pm at 120 kV. (b) Schematic bond rotation events occurring probabilistically.

1980s.26 We note here on the lack of our observation of the particle behavior of electrons under SMART-TEM conditions; that is, none of the flexible molecules bonded on the surface of a CNT were blown away into a direction opposite to the direction of the e-beam. We hence conclude that there occurs no direct kinetic energy transfer from electrons behaving as particles, as analyzed at the resolution used in the present study. To illustrate further, let us focus on a rotating molecule 4 in Figure 6 and 7b, which has a projected area of ca. 1 nm2. Consider that we obtain the 1 nm2 image on a detector consisting of 50 × 50 pixels, we would need at least 2.5 × 103 electrons/nm2 to obtain a good molecular image. If an electron gun (e-gun) sends out 2.5 × 105 electrons/(s·nm2) (a typical flux that we use), we can form 100 molecular images/s. If we send 10 times more electrons/s (i.e., 2.5 × 106 electrons/(s·nm2)) and use a state-ofthe-art CMOS sensor, we can obtain 1000 molecular images/s or one image/ms. Thus, we can obtain higher time resolution by the use of both a larger electron flux and a fast camera to reach up to millisecond resolution. We must note that σ-bond rotation is also inherently probabilistic, occurring with seeming randomness but following quantum mechanical principles (Figure 7b; that is, no continuous rotation as chemistry students may naively expect). The frequency of rotation events also depends on the frequency of the energy supply from the vibrating CNT substrate, which is also probabilistic at the AR level of analysis described in this Account.

Figure 9. Organic solid vs single molecules in vacuum. (a) E-beam irradiation causes chain events in solid. (b) In vacuum, no chain reactions occur.

In 2010, we received the following comment from a reviewer: “The electron dose is much higher than the accepted electron dose (a dose of 103 electrons/nm2), which completely destroys molecules.” The reviewer is correct. We showed that solid fluorocarbon 5 began decomposition from a dose of 800 K, undergoes thermally forbidden [2 + 2] cycloaddition33 under TEM observation in a CNT (Figure 15a,b,c; 0.9), which is followed by extensive C−C bond reorganization (Figure 15a,b,c; 3.0−10.4). We followed the progress of these reactions and identified several intermediates. The most notable structure is a Cs symmetric [2 + 2] dimer (Figure 15a; 3.0) formed first, an isomer of the D2h-symmetric dimer synthesized in bulk solid (Figure 15d).34 The D2h dimer looks highly symmetrical while the Cs dimer unsymmetrical as seen from the direction shown in Figure 15d. The apparent dissymmetry, seen in the product at 0.9 in Figure 15a,b, therefore indicates that the TEM conditions produced the Cs symmetric [2 + 2] dimer. Hole catalysis35 is a well-established reaction mechanism, and a [4 + 2] cycloaddition catalyzed by an aminium radical is typical (Figure 16). Here, an aminium radical reversibly generates a diene cation radical, which adds rapidly to another diene molecule.36 Given that the e-beam frequently ionizes π-electronrich CNTs (ionization potential, IP = 5.0 eV), we suggest that the fullerene (6.2 eV) dimerization is catalyzed by a CNT radical cation as driven by the release of the intrinsic strain of the [60]fullerene skeleton. Reaction pathways via excited states of CNT and the substrate molecules may also be possible as discussed in section 8.3 or go parallel to the radical cation pathways as summarized in Figure 17.

Figure 13. Rapidly moving biotin derivative 8 at 80 kV. (a) Structure of 8. (b) Cartoon illustrating 8 on a waving CNH. (c) TEM images. The numbers refer to the frame numbers of the movie. (d) Potential surface for the conformational change of 8 as deduced from the TEM images and MD calculations. Adapted with permission from ref 27. Copyright 2015 American Chemical Society. See SI for a movie.

8.2. Fe- and Ru-Catalyzed Reaction of [60]Fullerene

Metallocenes fused to fullerene, called bucky ferrocene (Fe(C60Me5)Cp; 10)37 and bucky ruthenocene (Ru(C60Me5)Cp; 11),38 provide an interesting showpiece of the catalytic effects of d-block transition metals. Upon irradiation with an electron dose as small as 5 × 104 electrons/nm2, the metal atom starts to etch the [60]fullerene sphere, and with a dose of 1.6 × 106 electrons/ nm2 forms an entity similar to [70]fullerene with the incorporation of the cyclopentadiene and five methyl groups (Figure 18).7 This observation has a direct relevance to the Fischer−Tropsch reaction and metal-catalyzed CNT and graphene synthesis.39 The high reactivity of 10 and 11 stands

oligo(ethylene oxide) OEO group. The conformational change between 49 and 50 was caused by N−CCNH bond rotation (θ1 in Figure 13a), and the next change after 84 is due to cis to trans isomerization of the amide bond (θ2). Overall, the observed conformational changes took place along an energetic downhill (Figure 13d). For the analysis of such complex conformational behavior, we are developing a computational cross-correlation image analytical tool.28 1287


Article


Figure 14. A van der Waals cluster of two molecules of 9 attached to 4. ChemDraw illustration, 3-D model, and one frame of the movie. Scale bar 1 nm. Adapted with permission from ref 5. Copyright 2012 Nature Publishing Group. See SI for a movie.

Figure 15. Dimerization of [60]fullerene. (a) Original TEM images. The numbers refer to time in seconds. (b) Binary-formatted images. (c) Models. Scale bar = 1 nm. Adapted with permission from ref 6. Copyright 2010 Nature Publishing Group. See SI for a movie. (d) Two dimerization pathways illustrating Cs and D2h products.

in contrast to the stability of [60]fullerene up to a dose of 107 electrons/nm2. It is reasonable to expect that the e-beam activates the iron atom via electron and energy transfer to cause the changes of spin and oxidation states, hence vastly accelerating chemical reactions.40 f-Block lanthanide metals are intrinsically catalytically inactive (Figure 19). Er@C82 decomposed only after extensive electron irradiation (e.g., 107 electrons/nm2 dose).6 Parenthetically, the Er atom (Z = 68) seldom rotated in the fullerene cage during several minutes of observation at 298 K, despite the small 2−3 kcal/mol energy barrier of rotation determined in solution.41 Apparently, isotropic thermal energy does not excite an

Figure 16. Aminium radical catalyzed [4 + 2] cycloaddition.

Figure 17. Fullerene dimerization catalyzed by CNT cation. 1288


Article


120 kV).6 On the other hand, perfluoroalkane, hydrocarbon, amide, urethane, sulfide, ether, aromatic compounds, and van der Waals molecular clusters are stable up to a total dose of ca. 5 × 107 electrons/nm2 at 120 kV, where the CNT substrate starts to decompose.8 A tertiary amine reacts with CNT to form a quaternary ammonium ion under conditions not well-defined in the literature.42 With these data and the hole catalysis in mind, we suggest that the reactivity of organic molecules may depend on their IP relative to that of CNT (IP = 5.0 eV). As summarized in Figure 20 (bottom graph), the molecules with IP similar to that of CNT are reactive (red box). In contrast, the molecules with IP larger than 8 eV (blue box) are stable, the IP difference of 3 eV making their oxidation by CNT radical cation hardly possible. Noticing the parallelism between IP values and HOMO/ LUMO gaps as shown in the top graph in Figure 20, we may also consider a possibility that the excessive energy provided from the e-beam to the CNT may be transferred to the molecules nearby, which bring the molecules to their excited states of the molecules to cause chemical reactions, light emission, or thermal relaxation. Given the proximity of the CNT and the molecules, such energy transfer may occur via the Dexter mechanism involving electron exchange, and therefore it may not be readily apparent at this time which of the radical cation path or the energy transfer path operates in individual cases that we encounter. The IP argument also suggests that the CNT can stabilize higher-IP molecules against ionization by electron donation, even if the latter may be occasionally ionized by direct interaction with the e-beam. It is possible, with hindsight, that TEM has long provided images of hydrocarbon molecules, which have thus far been considered merely as contaminants.

9. CONCLUSION AND OUTLOOK

Figure 18. Conversion of bucky ferrocene 10 to [70]fullerene-like molecule. The star may be either radical cation or excited species. (a) Reaction scheme. (b−e) TEM images: (b) 3.7 × 104, (c) 2.2 × 105, (d) 3.0 × 105, and (e) 1.6 × 106 electrons/nm2 at a 293 K sample stage. Adapted with permission from ref 7. Copyright 2011 American Chemical Society. See SI for a movie.

This Account has summarized the development of the SMARTTEM imaging technique, which has several unique features. Specimen molecules are captured chemically on a CNT and analyzed using AR-TEM, where the CNT conveys low-frequency vibrations to the molecules and also prevents sample charging. While low-IP molecules may undergo chemical reactions triggered by the CNT radical cation, common organic molecules with high IP values are stable. The speed of the accelerated electrons being extremely fast, each electron that passes by the molecules carries attosecond-level molecular information, but the probabilistic behavior of both electrons and bond rotations makes the image interpretation a somewhat complex issue. “What new horizons can we explore?” Goals in the near future include kinetic studies of chemical transformationsa gateway into the quantum mechanical world of molecular science. Further upscale applications will include imaging of individual small proteins and their motions (Figure 21). The capability of handling a mixture of molecules for imaging is a significant feature of this new analytical method, suggesting a possibility of structural analysis of compounds of natural origins without extensive purification. Atomic-resolution imaging of molecules and molecular assemblies is a simple and ultimately bottom-up approach to the study of the nanoscale entity and will serve as a gateway to the research and education of the science connecting the quantum mechanical world and the real world.

Figure 19. Er atoms standing still in the cage of Er@C82 in a CNT. Adapted with permission from ref 6. Copyright 2010 Nature Publishing Group. See SI for a movie.

anisotropic translation motion of the Er atom in the fullerene cage. 8.3. Why Do Some Compounds Decompose and Some Do Not?

Some compounds are stable and some unstable under the SMART-TEM conditions. We found 10 and 11 extremely reactive (

Atomic-Resolution Transmission Electron Microscopic Movies for

Recommend Documents