Breaking Open the Black Box

I was sitting at my desk in lab, taking a break from looking through one too many Excel data files, when I received this short email from my sister. “On a scale of amazingly to insanely, how awesome is this?” The link she included could have led to any number of web pages: pictures of a double-chocolate devil’s food cake from her favorite food blog, event details to a Mumford and Sons concert she would be attending, or maybe even one of the videos of goats yelling like humans. Instead what popped up was the latest Science paper, released just hours earlier.

This came as no big surprise since my sister and I both lead careers in chemistry research. What was a surprise, however, were the pictures that stared back at me from my screen:  the first-ever crystal-clear images of a molecule before and after a reaction. They looked just like the wire-framed representations I had been drawing since high school chemistry, yet these were the real thing. Not only that, but Berkeley Golden Bears were behind this work.

Professor Mike Crommie of the physics department and Professor Felix Fischer of the chemistry department collaborated to produce the images. Although visually striking, it is the science behind how they were produced that will make the most lasting impact. Fischer and Crommie used non-contact atomic force microscopy (nc-AFM), a technique capable of imaging molecules at the nano-scale. First developed by a research team at IBM in Zurich, the ability to produce images of molecules caused a ripple of excitement among scientists and non-scientists alike in 2009. The Berkeley duo took it a step further and used the technique as a way to determine the products and mechanism of surface reactions, something that has, to date, been concealed frustratingly inside the chemical “black box”.

The starting reactant, left, with two of the major products, right after the reaction. The scanning tunneling microscopy (STM) images above provide little information about the identity of the molecules, while the non-contact atomic force microscopy (nc-AFM) images below reveal the details of their structure. Credit: Felix Fisher

Many areas of chemistry research boil down to one simple equation: A + B → C + D. A and B are the starting materials, while C is the desired product(s) and D may be any undesired side-product(s). C can be a new drug for Parkinson’s, biodiesel produced from algae, or the latest nano-material to be used in electronic devices of the future. Researchers need to understand how changing their As and Bs, reaction conditions (temperature, pressure, etc.), and countless other factors will affect their Cs. They also need to understand how to tune their process to limit the production of D. More often than not, the process is not quite optimized, and the products are hidden behind a black box, their identities unknown. One of the most important questions then becomes: What are we making and how can we make it better?

Characterization techniques used to answer such a question are abundant for solution-based chemical reactions. Once restricted to a surface, however, the options are limited. Fischer and Crommie ran into this limitation and decided to break open the black box.

Using chemistry to make nano-scale computer chips

For the Fischer lab, their desired product (C) is highly ordered carbon-structured nano-materials, particularly derivatives of graphene. Graphene is a two-dimensional sheet of carbon atoms organized into a honeycomb lattice. While it can theoretically take any shape or size along a flat plane, it is only one atom thick. These graphilic nano-structures have high strength, flexibility, and efficiency in transmitting charge, making them an alluring material for electronics. Electrons in graphene move faster than in silicon, and the hope is that it could one day replace this commonly-used material to make smaller, faster, and better transistors, the building block of electronic devices.

The Fischer group is focused on the development of efficient, reproducible methods for making graphene structures of varying shapes and sizes.

“If you start cutting out structures from these graphene sheets, they have inherent electronic properties that are defined by the shape you cut out,” explains Fischer.

As with most materials, limiting them to the nanometer size scale yields many unique electronic properties. This is known as the quantum confinement effect. Within this size regime, small changes in size and shape change a structure’s properties significantly. Learning what shapes and sizes are required to yield a given property of interest is part of what the Fischer group hopes to understand. Nano-sized ribbons made of graphene, for instance, could one day be used as nano-scale wires for smaller and smaller electronic devices.

The trick, however, is making reproducible graphene structures. Most groups in the field go from big to small, taking larger structures and chiseling them down with lasers to the shape and size they want at the nano-scale. Resolution becomes the issue here, since harvesting the desired quantum confinement effects requires atomically precise structures. It may be possible to make a structure once, but to reproduce the exact same structure with the exact same properties again is challenging, as lasers are inherently limited by their resolution. As Fischer explains, “One atom in the wrong place or the wrong atom at any place will basically destroy the properties of the material.”

The Fischer group has a more chemical approach to the problem, and has decided to work from the bottom-up instead. Their strategy is to begin with simple molecules deposited on a surface, which will ultimately undergo several chemical reactions to rearrange and link together into a final, more intricate graphitic product. Fischer likens piecing these starting molecules together to building with Legos. The hope is that by beginning with the same reactant(s) and applying the same conditions, the reproducibility of chemical reactions will kick in to yield identical materials with identical properties. There is still a lot of work to be done before this dream can become a reality, however, as much of this chemistry has been largely unexplored; the field of graphitic nanostructures only recently took off in 2008. Fischer sees this as an advantage.

“It’s a lucky break when you can start your career at a time where one of these new fields just starts taking off,” says Fischer.

One project, three UC Berkeley labs

An integral member of the team, Professor Crommie of the physics department supplied the imaging resources and know-how to acquire the striking images of Gorman’s molecules. Credit: Berkeley Lab - Roy Kaltschmidt

An integral member of the team, Professor Crommie of the physics department supplied the imaging resources and know-how to acquire the striking images of Gorman’s molecules. Credit: Berkeley Lab – Roy Kaltschmidt

Patrick Gorman joined the Fischer group as a graduate student less than three years ago. At that point the Fischer group technically didn’t even exist yet since Professor Fischer was new to Berkeley himself.

“When I came to Berkeley, I wasn’t sure what to look for,” admits Gorman. “I didn’t even know Felix was a professor yet.”

He did, however, know that he wanted to be involved in energy-related materials research, and was thinking of making the jump to organic synthesis after having spent his undergraduate research career at Notre Dame working primarily on inorganic chemistry.

As a new professor, Fischer was grateful that high-caliber students such as Gorman are common at Berkeley, and that they were willing to help him build up a lab from scratch.

“Really good students tend to gravitate to younger groups here, which is a little bit surprising,” says Fischer. “I admire that, honestly: someone joining a young faculty and taking all the risks associated with it.”

Gorman’s project is to generate graphene nano-structures directly on the insulating surface to be used in electronics, thus avoiding the traditional requirement for a metal surface to catalyze the formation reaction. To achieve this, they are taking advantage of a reaction discovered at UC Berkeley in 1972 by Professor Bob Bergman’s group in the chemistry department. Known as Bergman cyclization, the reaction begins with a linear carbon chain composed of alternating single, double, and triple bonds which rearrange to form a cyclic six-membered ring, the building block of graphene. It is advantageous in that it only requires heat, not a catalytic metal surface. Thus Fischer’s group can use non-reactive insulating surfaces to produce their sought-after sheets of graphene via Bergman cyclization.

“The [mechanism] we’re working with is not surface dependent, it’s just thermally dependent,” says Gorman. In fact, most of the work with Bergman cyclization had been done in solution, without the use of a surface at all. This raised several questions, then, of what reactant(s) to start with, how this well-known solution chemistry would behave when sequestered to a surface, and what the final products would be.

“As chemists, we can build reaction mechanisms in solution where everything is flying around more or less freely, but on a surface it becomes much more complicated and the energies of what a plausible reaction pathway is and what is an implausible reaction pathway sometimes switch,” explains Fischer. He knew the group would need some help. So once Gorman had synthesized their starting material to for the Bergman cyclization reaction, Fischer turned to the Crommie group in the physics department to determine what products it would turn into, and crack into their black box chemical reaction.

Sometimes it takes more than one attempt

Members of the Crommie group are masters of scanning tunneling microscopy (STM), which they use regularly to probe the electronic and magnetic properties of various nano-structures, including graphene. STM measures the available energy levels of electrons in a material by scanning a conducting tip across the surface of a molecule. As the tip comes near the electrons of the molecule, electrons tunnel between the two due to the voltage difference. This tunneling current is measured by STM, which is relayed to the user as an image and can provide information about the higher electronic energy levels of the molecule.

The plan was to use Crommie’s instruments and know-how to place the starting material on a surface, cool it to -452 degrees F to immobilize it, collect an STM image, apply heat to induce a reaction, and then determine the products using STM after cooling it back down to -452 degrees F. Gorman wanted to see what his starting material had become in order to understand how he needed to modify it, the ultimate goal being to synthesize a molecule that can predictably form the larger carbon-based nano-structures they desired.

“Predictability: this is what we’re after right now,” says Gorman.

Although the STM images collected by Yen-Chia Chen of the Crommie group revealed that Gorman’s reactant had changed, the final products remained veiled behind the fuzzy images. The group was disappointed in what they saw.

Sebastian Wickenberg, a graduate student in the Crommie group who worked with Gorman, explains that the capability of STM is simply limited by what the instrument measures.

“You get a lot of tunneling current when you tunnel into the electronic orbitals of a molecule, which are extended and not very localized,” he says. Since this tunneling current extends beyond the bounds of the atoms and their bonds, the resulting image is ambiguous. “It’s just a blob.”

“We got a variety of structures and we didn’t know what we were looking at,” says Gorman. “So then the project became, ‘Well, what are we looking at?’”

Gorman spent months trying to decipher the identity of his “blobs” by synthesizing the molecules they predicted his starting material would cyclize to, depositing them on a surface, and imaging them with STM. Nothing matched. Frustration kicked in.

“What we expected and what we saw were unambiguously different structures,” recalls Gorman.

Traditional characterization techniques (nuclear mass spectroscopy, infrared spectroscopy, mass spectroscopy) were impractical in this case, and are better suited for samples in solution with greater yields, or reactions that result in changes in mass. In this case Gorman was dealing with a simple reorganization reaction where mass remained constant. In addition, these techniques are all subject to ensemble averaging, where results provide averaged information of the bulk product, and not the exact structure of individual product molecules. This becomes an issue when one reaction yields several products, which can all get muddled together.

“We knew we had to come up with a better characterization technique,” says Gorman. “That’s when non-contact AFM kicked in.”

Instead of tunneling current, non-contact AFM measures electron charge density. The tip, which is functionalized with a CO molecule to provide superior resolution, oscillates at a certain frequency. As it scans across the surface of the molecule, this frequency changes due to repulsion with the electrons of the molecule.

non-contactAFM

Click to enlarge. Credits: Lawrence Berkeley National Laboratory; Felix Fisher

“We’re actually dipping the electrons of our tip into the electrons of the molecule, more or less,” explains Wickenberg admiringly, clearly in awe of the sub-atomic level of precision of his instrument. “Since they’re occupying the same space, they repel.” Repulsion occurs when particles of the same charge, in this case negatively-charged electrons, come into close contact. This causes a shift in oscillation frequency of the tip, which is measured by a laser to produce an image of amazing clarity. The CO molecule is sensitive enough to provide such atomic resolution.

Wickenberg’s enthusiasm about the technique is obvious. What else can you expect from someone who has been reading about particle accelerators and the like in his dad’s science magazines since he was a kid?

Prior to Gorman’s project, the Crommie lab had never used non-contact AFM. The resources were there, but the need wasn’t. The groups were thrown into unfamiliar territory. Wickenberg says it took a while to figure out the right parameters, but once the system had been tuned properly, they could consistently take clear images.

“We didn’t quite know what to expect until we saw the pictures,” says Gorman, recalling the day he received the images from the Crommie group. Once they saw them, however, they knew: “This is something special.” No one before had ever captured images of molecules before and after a reaction.

From Science to Twitter

With the non-contact AFM images in hand, Gorman could now identify without a doubt what the molecular structures of his mystery blobs were. The images themselves, however, produced the biggest stir.

“We can actually tell where the triple bond is, where the carbon atoms are. You can even hazily see the hydrogen atoms on the periphery,” raves Fischer. “This is a very powerful method.”

Not only were the Crommie and Fischer labs blown away by the images, but soon so was the rest of the world. After their work debuted in Science, the media went abuzz with excitement: bloggers blogged, tweeters tweeted, and Facebook statuses were updated, broadcasting UC Berkeley’s latest research accomplishment. It’s not every day that scientific research captivates the general public.

“One of the things I like the most is that it resonates with non-scientists since everyone has seen a diagram at some point in high school,” says Wickenberg. “It’s really bringing that science into the living room in a way that’s very easy to understand. The result is a very gut-level reaction.” Gorman agrees that the accessibility of their work has been one of the most rewarding outcomes. “That’s just really cool to me: the fact that people who know nothing about chemistry can look at this and understand what it is. A picture is so tangible. It’s a visceral connection to science.”

Peeking inside the black box

Of course these were not just pretty pictures. These were Gorman’s answer to the mystery of what his starting material had become. Non-contact AFM revealed that there were two major products, together making up about 80% of the product pool and the remainder consisting of a variety of minor products. Fischer was happy to see this. “We were expecting a zoo of molecules on the surface,” he explains. “The fact that we actually see a preference for two certain pathways was a big relief because it makes it possible to tailor the structure so that you can shove it down one pathway over another.”

Had there been hundreds of different products, the Fischer lab’s goal of making predictable carbon nano-structures would have become a more daunting task. Gorman says the next steps involve understanding the thermodynamics of the surface-bound Bergman cyclization process in order to minimize the number of possible cyclization routes and produce a single, extended product. One way to do this? Make the starting molecule smaller and simpler. Following the work published in Science, the group has now successfully made conductive graphene chains on surfaces.

Crommie and Fischer’s images have also demonstrated the key role of the surface. In solution, Bergman cyclization occurs at about 180 degrees Celcius. On a surface, it occurred at about 90 degrees Celcius. Additionally, the products generated were unlike those expected in solution. These effects are due to surface stabilization of the radicals generated in the reactant molecule.

Fischer expects their technique to be very useful for understanding surface reactions, particularly in the field of heterogeneous catalysis, which he deems to be “largely black box processes”.  In heterogeneous catalysis, the role of surfaces, often metals, are studied as catalysts for many environmentally and industrially relevant reactions. By sequestering them to a surface, reactions become faster, more efficient, and more selective to a given product than if they were conducted in a free-floating solution. This is because the surface can facilitate the bond making and breaking processes of a reaction, and decrease the mobility of the reactants, increasing the likelihood that they will come together to react. The use of rhodium, platinum, or palladium in catalytic converters to remove toxic carbon monoxide and nitrogen oxides from automobile exhaust is one common example.

Currently, research in this field relies on indirect methods such as infrared (IR) spectroscopy to determine how a reaction plays out on a surface. IR spectroscopy uses infrared light to excite the electrons of molecules such as CO and NO that are adsorbed on a surface. Detectors measure the amount of light adsorbed at varying wavelengths of light, which can indicate how the molecules vare bound to the surface and with what strength. Reaction conditions, however, can be difficult to reproduce. Wickenberg agrees with Fischer that these indirect techniques make it challenging to determine surface reaction mechanisms.

“It’s much more satisfying to get a picture and just move on.” This is where Wickenberg believes non-contact AFM can potentially become a valuable asset.

One example of a black box reaction is the Fischer-Tropsch (FT) process, which converts a gaseous mixture of CO and H2 derived from coal or biomass into liquid hydrocarbons that can be used as fuels. Typical catalysts include cobalt, iron, and ruthenium. FT has been widely employed in industry, and has been the subject of countless research papers for decades. Nonetheless, the exact mechanism that transforms reactants into products continues to be hotly debated.

“There are mechanistic models that explain what happens, but no one has proven or agrees that that is what actually happens,” says Fischer. “No one has clearly seen which positions of the catalyst are active.”

Such knowledge would help researchers tailor catalysts to provide fuels more cleanly, quickly, and with greater yields. The application of Fischer and Crommie’s technique towards heterogeneous catalysis is still a distant dream, however.

Click to enlarge. Credit: Felix Fisher

Click to enlarge. Credit: Felix Fisher

There’s always more work to do

“The biggest problem with the technique is that it really only works well if the molecules are two-dimensional,” says Wickenberg.

Once molecules are no longer flat, it becomes much more difficult to develop clear images since areas of electron density start stacking on top of each other and become difficult to de-convolute. Given that most heterogeneous reactions of interest will likely involve three-dimensional molecules, this is a critical problem to overcome. Wickenberg is skeptical of finding a solution, but admits that it also used to be difficult to imagine that a machine with sub-atomic control such as non-contact AFM would have existed as well.

Fischer sees another major issue

“The temporal resolution of this is still crazy,” he confesses. While it takes 20 minutes to yield one non-contact AFM image, a surface reaction can be over in a fraction of a second. Luckily, instruments are being developed with high end laser optics that can trigger a reaction on a femtosecond (10-15 second) time-scale, cool the surface quickly down to -452 degrees Fahrenheit for imaging, and lastly use AFM or STM to determine what transition states are present on the surface. By taking images at varying time points researchers can create a flip-book to determine the complete chemical story.

If scientists hope to image a reaction in-situ as it is occurring, there are even greater hurdles to overcome. In-situ imaging would require being able to take clear images at reaction conditions, which are often well above 100 degrees Celsius. Higher temperatures mean molecules are moving around more, and would require higher scan rates to capture cleanly.

“Right now the scanning speed is limited by the signal to noise ratio,” says Wickenberg. Improving the signal to noise ratio would be key to developing an instrument that can image a reaction at elevated temperatures. To what temperature scientists would be able to go is debatable.

“There’s always more noise at higher temperatures,” explains Wickenberg. “There may be a limit you can’t reach.”

For the moment, however, Fischer and Crommie are simply hoping to improve the technique for their own purposes, which is to characterize their extended, two-dimensional carbon nano-structures. Fischer hopes that it will one day become user-friendly enough that any graduate student can use it, just like a nuclear magnetic resonance (NMR) facility. NMR is currently one of the most popular methods of identifying the structure of reaction products, but is mostly limited to reactions in solution, and requires a larger concentration of material than the surface products Fischer is dealing with.

“As a chemist, the moment I saw those pictures my first thought was: ‘Yeah! Screw NMR!’,” recalls Fischer, laughingly. He acknowledges, however, that “it’s not that easy.”

He hopes it will one day become just that easy, and believes the field will benefit from moving towards a technique that requires mere fractions of what is currently required, much like biochemistry. Will this become a reality in the next 30 to 40 years? Fischer won’t take any bets yet.

“There are lot of plain physical limitations to this technology,” he says. “Most of this can still be considered a dream.”

The Berkeley way: working as a team

What is certain, however, is that the Crommie and Fischer labs will continue to work together on the issue. Not only because Gorman says that a joint grant now means they’re “stuck with each other like family,” but also because the groups each bring something unique and essential to the mix.

When Fischer started at Berkeley two years ago, other members of the department urged him to get in contact with Crommie about the possibility of working together. Their goals fell in line and the labs became natural partners.

“We mutually benefit from each other, which I think creates the most productive environment,” says Fischer. “I can rely on his expertise of analytical techniques, and he can benefit from the diversity of structures I can deliver him.”

Fischer finds this collaborative nature to be one of the biggest perks of being a professor at UC Berkeley. Nowhere else has he experienced the same encouragement towards interdisciplinary collaborations, which he believes is essential towards solving some of the most challenging research questions.

“We probably can’t succeed if we lock ourselves up in our little ivory tower in chemistry and focus on what we’re good at: synthesizing molecules,” admits Fischer. “I think the really hot research is happening at the interfaces, whether it’s chemistry and biology, chemistry and physics, chemistry and astronomy, et cetera.”

Wickenberg feels he has learned a lot from working with chemists. Given the same problem, physicists and chemists may interpret it in different ways and end up with two different, possibly incomplete, solutions. Together they can fill in the holes.

“When you have that nice synergy, you can make some interesting things happen.”

Given the attention and hoopla this chemistry-physics duo has conjured with their groundbreaking images, they certainly made some very interesting things happen. Of course there were several other key players that led to their success, including the IBM research group who first developed the non-contact AFM technique in Zurich, research from the Bergman lab at Berkeley, and maybe even a middle school science teacher from New Jersey.

As soon as Gorman’s Science paper debuted, he made sure that one person in particular would see his work: Mrs. Stickle. As his middle school science teacher, Mrs. Stickle represents Gorman’s introduction to the world of science. He has loved it ever since.

“I had to email her and tell her I did this,” laughs Gorman. “She just said she was so proud of me.”

Leave a Reply

1 comment

  1. loanemu

    I really like this article. I think it’s full of useful and smart information and valid points with which I agree for the most part.