Science magazine cover image

[fmarc]

Hey all,

I'm a PhD student in physics at Hamburg University, Germany.

I was told that my blender image is going to be on the cover of this weeks issue of Science, yay!

I would really like to hear your C&C on it. I'll try to explain some of the physical background below.

Their website is at www.sciencemag.org

Science is one of the two major non-specialized science journals. Nature being the other one. So this image is going to be seen by many scientists worldwide , so it's extremely great for us to be published there, even without the cover page.
The issue is published on the 4., and the publisher put an embargo on all content until then, but I guess I can already show some details of the full image. I'll post the whole thing when I'm allowed to, in the next few days.

The story goes like this:
First, we submit a manuscript for publication, presenting our groundbreaking research.
After a manuscript gets accepted (the hard part), they ask you for a picture that might go on the cover. Out of all submissions (~15 to 30, I guess), they pick their favourite. It's ours, hooray!

Now, about the image...

It's based on data taken with our scanning tunneling microscope (STM), which scans surfaces at very close distances (only a few atom diameters above). This gives us the topography of a surface, in this case it's a stepped platinum crystal with a little cobalt on top. Additionally, we get an "electronic" signal, which tells us a little more about the local characteristics of a material. I used this signal's intensity for the color.

So, this is what you get then:



The blue part on the right and bottom left are two terraces of the stepped surface of the platinum crystal. There is an upwards step right in the center of the image, but you can't directly see the front, except for the colorchange red/blue, because there is the big red thing with the yellow side in front of it. This is some amount of cobalt deposited onto the surface, which has settled in front of a step of the platinum substrate.
This step is one atom high, also the red cobalt stripe is one atom high. You can see that the atomically thin cobalt layer is about as thick as one atomic layer of platinum. The cobalt appears red, because it's giving a different electronic signal.

The yellow-capped hills are single cobalt atoms lying on the surface.

The fine structure on the surface has a corrugation of some picometers (10^-12 m), much less than than atom. I'm amazed by this every single time. :-)

Technical details:
Blender 2.46rc1 for everything.

The source material were a 2000x1000px 16bit heightmap for the topography and a 2000x1000px RGB image of the pseudo-colored electronic signal. A huge fraction of the my work went into getting the colormap CMYK-safe, since science still is mainly a printed journal. I had never done this before and had no experience/established procedure. Originally, we had much brighter colors, but, turns out you can't print them. At least that's what I understood.

Then I took a plane mesh, subdivided it a few times, applied a subsurf modifier to it, to keep it handy during editing but give me the full lateral resolution in the final render. Then I put a displacement modifier with the heightmap texture after that. And of course I used a texture with the colormap for the mesh. I also added some "illustrative elements" which you will only see in the full view, which I post tomorrow when the embargo is lifted.

What do you think?

felix

[PapaSmurf]

I'm thinking candy corn. I don't know why, maybe just because I need a snack and my sugar level is low right now...

Seriously though, amazing stuff. I'd love to know more about the process you went through to convert your raw data to a (mesh?) and then render.

[fmarc]

Right, the surface looks quite edible, but I wouldn't say it looks very tasty. :-)

For the data processing: After measuring the image, we did just one processing step in our analysis software: we fit a plane to the surface and subtract it, to level the data.

The raw data from the microscope usually has quite a steep slope, which would make many things more complicated, i.e. it would mean, that my displaced plane would turn out strongly tilted. I could then rotate it to lie in the XY plane, but we do the plane fitting almost automatically anyway.

Since people here mainly use POVRay, our self-written analysis software knows how to write data in POVRay's heightmap-TGA format. This is basically a container for 16 (24?) bit maps, which is also a valid TGA, but visually not very meaningful, since it spreads its bits over the RGB channels.

On this, I used HF-Lab, an old and very versatile heightfield-editing and conversion tool, to convert this to 16bit PNG for Blender.

The original data were of, for our standards, really high quality in terms of noise etc., so I really didn't have to do any filtering.

Unless you consider rendering in less than the original relolution as filtering. It effectively smoothes the shape of course. I did this because blender would crash when I tried to render at a subsurf level that would give me as many vertices as the map has pixels. One level less was fine though.

[Mats Halldin]

Congrats, this was quite an achievement!

I must admit I don't understand what the image is about ("Probing Singe-atom Magnetization" certainly sounds far beyond my level of knowledge). The render looks like it would require Pixar server capacity and a National Geographic budget to produce.

From your perspective, was Blender more useful than POVray?

Thanks for sharing it
/ Mats

[PapaSmurf]

so, you got your heightmap from the microscope data..what did you map it to in Blender? Displacement? Nor?

What gave you the color information?

[CD38]

Getting a Science cover is tough. They have to be attractive as well as illustrate an important scientific result. Next target for Blender: Nature?

Nice work. Congratulations on getting accepted as well!

[VinceSpäce]

Congratulations!
As far as I understand you use a plane (Subdivided) with a displacement map and a subsurf to smooth that?! And the color map?

greetings from Kiel

[Sammaron]

Pixar level? I mean, its great and scientifically accurate (apparently ) but, it doesn't look like a great achievement in modeling to me... hmmmmm... :P Good thinking though

One atom's diameter away??? How can we have gotten this close to something, and 1). not having the atoms blow each other up (not annhilition and anti-matter, but i thought the EM would keep the microscope away. i thought it was always a myth your hands touched, they were miles apart atomically...?) or 2). not have seen a single atom yet (i just thought we had signals and evidence of them). or have we seen them? Anyhoo nice job on getting cover cool conversion as well

[DenDem]

Hi ! First of all congratulations!!! Science cover page... Hmm I think some jealous students will jump from the windows.
First of all – good job, second – I think that is a problem to subsurf the STM data ….you know… you will loose the data such as atomic steps etc.
From the other side – the shadow of the peaks is kinda problematic, because the structures are “a bit” smaller than the wavelength of the light (try to use “shadeless” option).
But again – great job ( I mean mainly STM achievements).
Can you post please your lab’s homepage (if you have one).

[j0llyr0ger]

It's very inspiring that a blender render will be on the cover of Science.

Maybe this is a dumb question, but since everything is false-color anyway, why not map the cobalt to blue?

[fmarc]

@Mats Halldin: Thanks for the Pixar comment , it was very regular equipment though. But as I wrote earlier, I couldn't subsurf down to full lateral resolution, because Blender would bail out if I tried. I set the mesh to 'Smooth', but you can guess the resolution if you look at the tips:


The render in final quality and 2700x3600px took ~20mins, not very extreme.

The main advantage of POV-Ray in this situation would be, I think, that it has a native heightfield primitive and can deal with the high-resolution map more, um, gracefully. OTOH, I think Blender made it *waaaaay* easier to achieve a more pleasing result quickly*, because I could fiddle with the appearance so directly and intuitively. And I just love the AO shades in the front, and I'm not sure how to achive this in POV.

(*quickly means here that I worked ~1day on the first draft, and maybe another day to reach the final revision. This is really not my field of expertise. I had to go through lots of experiments, any blender pro would be much quicker at this, and better)

@PapaSmurf:
I used a displacement modifier after a subsurf modifier. I tried displacement mapping, but that didn't seem to have any advantage, while the Disp Modifier gives me a nice preview to work with. And I found tweaking the Disp mapping easier with the modifier controls (Midlevel, Strength, Direction) than purely with the texture settings.

Color: Simultaneously with the topography, we measure the differential conductivity (related to the electrical resistance), in this case, they could show that this signal's intensity gave us a clue about the magnetization direction of the single atoms. The signal is rather strong on the stripes, where you see red (let's say "up") and yellow ("down") parts. In this picture, which serves less scientific but more aesthetic purposes, the differences on the atoms are too small to see. But in principle some should be more yellow in the center.

The signal intensity was mapped to a color range, which we adjusted to show the magnetization contrast, as well as the other beautiful rich phenomena, like the blue/black patterns on the crystal. So a map of the diff. conductivity was turned into a color image and became the color channel.

@CD38:
Thanks!!!
Maybe I should state again that I'm not one of the co-authors, and was not really involved in the science, I just share the lab and the office with the authors. And made the image. Scientifically, I work on another project, which is not likely yield a Science publication very soon :-/

@VinceSpäce:
Thanks and moin moin.
I used modifiers for subsurfing to the desired mesh resolution, as well as for the displacement. Mainly to stay flexible during editing. Also, I didn't have to mess with single vertices anyway. The color map, as said above, was an additional measurement signal.

@Sammaron:
I agree. I didn't pay so much attention to previous cover images of Science. But I saw last weeks cover, and found it very good. If this was also done by the researchers themselves, it surely was a much more impressive achievement than what I did, technically and didactically. I honestly don't know how many research groups have professional graphic artists doing stuff for them, not many.

You're absolutely right, the modeling certainly was not problematic here. But I would like to know what you people think about the image as a whole, how the illustration of a scientific issue worked for you, and how it could be improved.

You are right about the electrical force. But it's more like 10 atom diameters away, so that doesn't play a big role. At this distance the force actually is still attractive (Van-Der-Waals). It's definitely not easy to achieve a quality like in this work.

@DenDem:
Thanks!
Well, they shouldn't jump. Science has a cover every week, and there are many more journals. Also, in my opinion, a scientist should try not to derive their self-esteem from the count of their top-journal publications, otherwise they might become huge a**holes or constantly frustrated .

I don't quite follow you about losing atomic steps. The huge step in the image *is* an atomic step. Our raw data has a lateral resolution of ~35pm, much less than the distance between two atoms in the crystal. So even with half or a quarter of the resolution I have a mesh with vertices ~0.1nm apart. The vertical resolution is even higher. Since the colormap is interpolated, you can guess the raw lateral resolution from the colormap.

You're absolutely right about the light. It wasn't my idea There are no shadows in the actual publication, just on the cover.

This work of my colleagues is outstanding and deserves this publicity, but once you're in the journal, there is also a lot of luck involved to get the cover page.

I'd love to explain much more but I want to stay on topic and relevant, and from a Blender perspective this is quite basic stuff, I think.

Our group's website is at www.nanoscience.de.

The Wikipedia entry on scanning tunneling microscopy has a nice picture showing the principle of an STM.

You can find slides and audio of a short talk by Don Eigler, a real pioneer of STM, about this, here.

Also check out these two videos with (a very cool ) Don Eigler explaining how he moved atoms with the STM:
Video 1
Video 2

Thanks for your feedback!
felix

[fmarc]

@j0llyr0ger:
Good idea. I thought of that, but only after we had sent the picture to the publisher. But from my experience, (almost) no scientist would expect or even recognize that.

This is the image as we sent it to the publisher:


Find a highres (1350x1800) version here.

[DenDem]

Excelent work!
Very clear image!
But one more idea: may be you should add some bar (may be 3D) to emphasize the size.... This is the reason I couldn't understand that that is the atomic step
And again - good job.

[fmarc]

Thank you all for so much very positive feedback!! That is so encouraging! I'm still very excited about all this. :-D

Right, scale bars would make it more clear and would be more correct, especially since the z-scale is magnified by ~20 in relation to the lateral scale. In this case, I was unsure, but finally left them out, to make it visually simpler, since it was intended for the cover.

felix

[PapaSmurf]

Recommended to BN.

Q: regarding those rivulet/channels in the surface of the Platinum, where the Cobalt is not, are they the gaps between the atoms then, in other words, showing how the atoms of Platinum are arranged, and is the plateaus of yellow, the platinum layers, about one atom deep?

[fmarc]

Oh, wow.

Seen from the side it's like this:

             +--+-------
             |Co|  Pt   
     +--+----+--+--------
     |Co|    Pt         
+----+--+---------------

Hmm, there's definitely room for improvement in my explanations. :-/

If atoms of one kind form a closely packed layer, like the platinum layers and the cobalt stripes, we can't resolve single atoms, we only see the overall step they form. In general we can, but they didn't do that here. One would need different measurement settings.

If single metal atoms lie on such a surface, we see them as those spikes all over the image.

So the yellow and red areas are closely packed cobalt. The blue part is crystalline platinum, the surface of our substrate crystal. The cobalt stripes run along the step edges of the platinum and are one atom high and maybe up to 40 atoms wide.

The ridges/channels arise from the different spacings of atoms in platinum and cobalt. Imagine having a closely packed layer of balls and then you put a full layer of slightly smaller balls on top of that. They won't fit nicely in the gaps of the lower layer. Well, in some places, the upper lattice might stretch a bit to fit, but it can't do that everywhere. In someplaces the upper spheres lie just on top of the lower ones. In this combination of cobalt on platinum, these areas of mostly-in-gaps and mostly-on-top form a (semi-)regular pattern. The structure you mentioned is what our STM sees of those structural differences.

This is all quite easy if you can see it in a picture. I could draw one for you when I'm back home on monday.

[organic]

I don't know a great deal about physics, but from what's been said it seems this is about as 'photo'real as it gets with current technology( I'm assuming, that as the individual atoms are smaller than the wavelength of light it doesn't make sense to discuss their true 'colour'). So, surely the Gallery...?

The cover image does look good:- http://www.sciencemag.org/

[coodle]

Gratuliere! Ist ja echt geil, dass eure Arbeit in Science veröffentlicht wird und dann auch noch Frontpage!!
----
Interesting research you're doing there. What could it be used for? (I think for almost everything that depends on surface but are there already special demands from the industry?)

[fmarc]

@coodle: Danke!! Ich hab's wie gesagt nicht geschrieben, nur das Bild gemacht. Ich arbeite halt im selben Labor.

Regarding possible uses, well, mass-storage comes to mind. But before you do the math and imagine what you would do with all those terabytes, I should tell you that all experiments were done at 0.3 Kelvin and 10^-11 mbar. That is colder and a better vacuum than in space. Also, our measurement system is 3 by 3 by 3 meters and weighs half a ton. And they looked at ~20 atoms.

In this form, it just doesn't scale up very well :-)
So this is basic research, but as we know that looking at individual atoms this way is possible now, we surely will learn a lot about magnetism on this scale. This will eventually improve certain technologies, but it is too early to say which and how.

[Chris_N]

Fantastic job. Getting a front cover of any mag is a great step, but the cover of Science.. bravo :P

I thought it strange that you used Blender to render raw data.

In college I took a visualisation course. The aim was to take raw data (such as MRI scans or any data scan) and visualise it. It was an independent program of course written in C++ and it took quite a while to write it. I didn't think you could get away with what really is a mock up of the data rather than an actual translation of the data from raw figures to visual presentation. Well done
And there's me thinking that Blender is useless for this stuff as I sat through lectures of visualisation.
Again, lots of congrats.