When Friday morning of SPIE week rolls around, it feels like my brain is completely full. Even the half-day of conference left seems too long. In the Optical Lithography conference, Friday morning is traditionally the “tool” session, and the first tree talks were by Nikon, Canon and ASML giving their roadmap status reports. While the topics were interesting, I found myself fascinated instead with a different lesson they were teaching me: How to Lie with Graphs. Much of the data was of the sort to show how some parameter was either very high (uptime), or very low (overlay error), or very stable (immersion fluid temperature). To “enhance” the desired impression of high, low, or unchanging, the range used for the y-axes of graphs can be properly manipulated. So, if defect densities range from 0.05 to 0.11 per square centimeter, make the graph go from 0 to 0.3. Thus, all the number seem low. For uptime, with numbers ranging from 85% to 95%, make the graph got from 0 to 100 so that all the results seem high (you can also use a bar chart so it is not as obvious that the bottom 80% of the graph is unused). But my favorite is the CD uniformity wafer contour plot. If the data has a mean of 40 nm and a three sigma of 2 nm, make the contour range go from 32 nm to 48 nm so that only two or three contour colors are actually used, making the wafer data seem smooth as silk. Common guys – you should know better than that.
I am glad that I stay through most of the morning, though, because I saw my favorite paper of the conference. Lieve Van Look of Imec gave a great talk on matching scanners to enable their use for a given OPC’ed mask. A tremendous amount of work was shown, with clear analysis and well supported conclusions. Good work.
No mention of the 3 imprint sessions. Did you have a chance to sit in on any of these? Curious to see your impression especially since a large number of industrial companies presented data.
I totally agree with you about Lieve’s paper; as the work unfurled, I became progressively more impressed. This was an elegant implementation of things that I was trying to make happen almost a decade (!) ago.
I once had to do one of those Fri morning presentations – glad I don’t have to do em anymore – it feels like being between the silicon rock and hard place!
Having been in the industry so long (almost as long as Chris), I am impressed more by the numbers (which were usually given), rather than the scale of the plots – often, data were reported in nm to 2 decimal places – arguable how significant those last two digits are, but that is where the action is, now – and will be one factor in whether optical (which includes EUV photons, as pointed out by Winfred Kaiser) will stand or fall against nanoimprint.
I agree about Lieve’s paper, but it was somewhat similar to what all the tool vendors are doing, now – plus we have to also design and provide the knobs for lithographers to use and write papers about!
All toolmaker engineers and scientists put a huge amount of work (most of which goes unpublished, other than in Patents) to arrive at a few papers that management and marketing departments (and graphic designers) have the final say on – so I wouldn’t use the word "Lie" – but I am biased – maybe this could be called toolmaker/lithographer bias 🙂