screen allows the user to turn the image into an experimental interface, coordinating with and amplifying the multi-directional affordance of the tip-surface interaction, or into what Daston and Galison call the “image-as-tool” (414). For example, in one of the experiments mentioned above, Jaklevic used images created by the STM to monitor an experiment. As interfaces, the STM images Jaklevic produced to observe the behavior of gold became events through which the researchers conducted the experiment. The involvement of Jaklevic and other STM users with the image (and the data) was intensified by incorporating the physical actions of the experimenters into repeated scanning and image constructions in order to conduct the experiment. The “Quantum Corral” image (Figure 4) is also a good example of how STM use becomes an event, as Eigler and other researchers created the nano-corral structure in order to conduct experiments on the electron standing waves they “caught” inside (Crommie, Lutz, and Eigler 218–220). The ability to use the STM image as an experimental space—to build structures (as in the case of the corral) or to observe events (as in Jaklevic’s experiment)—is part of what makes the STM a significant visualization technology. As Gimzewski et al. comment in the journal Surface Science: “Scanning probe microscopies (SPMs) . . . , and in particular scanning tunneling microscopy (STM) . . . , have revolutionized the real-space imaging of molecules, providing a detailed understanding of the ways in which they interact with each other and with adsorbents” (101). Gimzewski et al.’s observation highlights the focus on using the STM as a scientific tool for understanding interactions.
The STM encourages further interaction with the data through the image, as researchers engage in understanding the significance of the data. Amann and Knorr Cetina explain the mode of practice that involves assessing significance as one in which scientific visuals “act as a basis for sequences of practice rather than observation at a glance. They [visuals] are subjected to extensive visual exegeses, rendering practices which attempt to achieve the work of seeing what the data consist of” (90). GUI characteristics suit STM images well to this interpretive task, thus extending users’ interaction with the images. As one scientist I interviewed explained,
It’s [the experiment is] very much like putting something on a surface, seeing what it does, and trying to figure out exactly how it’s behaving, and there’s a lot of control experiments that you do between changing biases, changing tunneling currents, changing processing of the sample, to confirm like exactly what’s happening, how you’re looking at it.33
The ubiquity of the GUI, and the ability of researchers to use the same GUI to continue to interact with the image and explore and arrange data, while also making sense of the data and then producing images as evidence, all allow the user to go back and forth between Amann and Knorr Cetina’s observational, interpretive, and evidence-producing modes of practice. While microscope users have almost always manipulated samples being viewed to produce images, the GUI intensifies and structures the involvement of the STM user in all stages of image production (Keller, “Biological” 110). The STM user’s involvement creates a different relation to the image than if, for example, the user positioned the sample in an electron microscope and created an image, because the sample is destroyed in the process of viewing through an electron microscope, making only a limited amount of interaction is possible.
The STM user’s involvement in image processing also encourages users to engage in continuing dynamics with the image and atoms; not only is it possible to engage with the sample again, but it is also possible to engage with the image again: The electronic screen affords the possibility of “refreshing” the image, just as the fact that the sample is not destroyed in STM scanning affords the possibility of “refreshing” the data through another raster scan. The GUI amplifies the effect of dynamic, manipulable images, data, and atoms. GUI dynamics thus structure interactions that invite multiple encounters, increasing a sense of immersion that heightens the feeling of engagement, as Rafaeli and Sudweeks explain. The affordance of the STM that allows users to repeatedly interact with samples through the image also reinforces engagement with the data—also perhaps suggesting the individuality of atoms through that repeated interaction.
Emphasis on manipulation as opposed to observation alone enters discourses about how the STM and its images can be used. For example, in an article describing imaging with the scanning tunneling microscope that appeared six years after Binnig and Rohrer won the Nobel Prize for developing the STM, IBM physicist John Foster writes, “After imaging a molecule, the next step is to do something to it” (26). Discourses about imaging functions also emphasize manipulation. More recently, a 2006 National Academy of Sciences Board on Chemical Sciences and Technology report on challenges and possibilities for chemical imaging acknowledges expanded uses of images that allow scientists to “do something to” what they see (Board 14).34 Among important challenges for imaging, the report lists “understand[ing] and control[ling] complex chemical structures and processes”; “understanding and controlling self-assembly”; and “understanding and controlling complex biological processes” (Board 22–25). While the Board’s report does not focus entirely on the STM, the report includes the STM as one of the visualization tools that can help researchers meet these challenges (114–21).
Producing STM Images: Image Processing and the STM User’s Role
The interactive dynamics of the STM extend to the STM user’s interpretation of the data, as well as the production of images as evidence through the tools of image processing. A key point of image processing, as John Russ explains in his Image Processing Handbook, is that “image processing, like food processing or word processing, does not reduce the amount of data present but simply rearranges it” (xiii). Russ’s mention of arrangement in relation to image processing highlights one of the affordances of the GUI that becomes significant in structuring STM dynamics. While researchers using non-digital imaging processes may plot a graph from numbers or photograph experimental results, non-digital imaging processes limit how much researchers can change the graph or photo after production without also changing the data. In contrast, the process of digital image production associated with the STM allows researchers to be involved longer with the image during and after data collection. Prolonged involvement includes interaction with the data during what Amann and Knorr Cetina articulate as the transformation of data into images for publication that function as a “way of visually reproducing the sense of ‘what was seen’” (114).
Like other digital imaging practices, the image production processes of the STM also incorporate the user into GUI practices that are contingent on interaction. Arranging information in visual form, and in the form of pixels, allows the STM user to continue the imaging process for far longer than a developer’s involvement with optical film, extending the time the researcher participates in the imaging process. Michael Lynch’s explanation of an extended imaging process in his study of the digital image productions of astronomers provides a sense of what also occurs with the STM images: “The real-time work of digital image processing involves a play at the keyboard, where images on the monitor are continuously recomposed by changing the palette, using touch-screen routines, plugging in parameters, and trying out different software manipulations” (“Laboratory Space” 72). The extended time that digital imaging processes require allows researchers to continue interacting with the data through the image and with different imaging techniques to develop images that contain experimental evidence.
To prepare an image for publication, researchers interact with the image to “clean it up,” often by filtering the data. In an article that appeared soon after Nature published Eigler and Schweizer’s images, in the IBM Journal of Research and Development, E. P. Stoll explains that raw data needs to be processed further due to interference, or “noise”35 (following Shannon’s division of information received into two categories, signal and noise), including noise that creates stripes “visible in nearly every real STM picture” (69). Therefore, some manipulation of the image, what Stoll calls “picture processing,” tends to occur (87). Some STM researchers do not present filtered images in journal articles; however, “cleaning up” data is a common scientific visualization practice (Brodbeck, Mazza, and Lalanne 31).36 Various ethnographers have discussed data processing as part of scientific image production. For
