Comments

Comments from Nenad Vukmirovic, IPB, Serbia email: nenad.vukmirovic@ipb.ac.rs URL: www.scl.rs/nenad
This is a detailed and very interesting document that summarizes the current role of HPC in science and the expected achievement in the next 5-10 years. I really enjoyed reading this document.

I have several optional comments coming from my own field of expertise (simulations in materials science and nanoscience):

1. Section 5.2.1.1 discusses the state of the art and challenges in hard matter and 5.2.1.2 in soft matter. Traditionally, theoretical techniques in hard matter are based on quantum mechanics (for example: density functional theory) and describe very short length scales. On the other hand, techniques in soft matter are based on statistical physics and describe longer length scales. A challenge that, I believe, will be to some extent tackled by the new generation of computer systems is: "unification of hard and soft matter". One may envision a multiscale approach for the simulation of soft matter which links several length scales for the short ones where the laws of quantum mechanics apply and one used the methods of hard matter to long ones where the laws of statistical physics rule.

2. The beginning of section 5.3 gives a nice review of simulation methodologies that have been used in materials science in the last 10-20 years. I would add to that list the (semi)empirical methods for electronic structure calculations which enable large system calculations with atomistic details (for a non-exhaustive list of these, see for example this CECAM workshop: http://www.cecam.org/workshop-470.html). I believe that these methods will be essential in realizing some of the visions from this document, such as "full device simulations from first-principles" (page 96).

3. As a general comment, I would add that for the effective use of new computer systems, their expected properties should be disseminated to scientists as early as possible. Scientists who develop methodologies for the simulation of certain physical systems and properties might invent completely different (and better) approaches if they have a certain computer architecture in mind.

4. In page 90, ab-initio quantum device simulations are identified as a challenge. As one example of envisaged simulations, it is mentioned that picosecond time-domain simulations of 1,000,000 atom systems will be possible. I believe that this will indeed be the case. However, I don't think that 1ps simulation is sufficient for the simulation of device characteristics (such as current-voltage curve). This is especially the case in materials that contain a certain degree of disorder, such as organic electronic materials. For a full device simulation, either a different approach (not the brute force time-domain) and/or a multiscale procedure will be necessary.

Comments from Kersti Hermansson, Department of Chemistry - Angstrom, Uppsala University; Theoretical Chemistry, The Royal Institute of Technology, Stockholm. Email: kersti.hermansson@kemi.uu.se
I have read through the document "Update of the scientific case for future provision 2012-2020", and particularly Section 5 (Materials Science,  Chemistry and Nanoscience). The document is impressive in many ways, and it is clear that much work has gone into writing it. Many good and strong points are rasied. However, there are a few aspects that particularly bother me.

1) It is important to make funding agencies aware of the importance of HPC though a document of this sort so it will hopefully have good impact. At the same time it is important to be clear that PRACE deals with infrastructure and is not a Research council. I believe that if these missions are kept apart, the impact will be greater.

Take pp. 88-89 for example (Chemistry). Two challenges are listed: Quantum chemistry and photochemistry. These are two different types of challenge (the first being more of a methodological one, according to the way in which it is presented; the second one is an application area). This is OK. However, as interesting as photochemistry may be, it can probably not be claimed to be THE application area to be singled out as THE most important in Chemistry for the next 8 years.

Moreover, HPC-dependent chemistry in too narrowly defined in Section 5, in my opinion. Computational/theoretical chemistry is much broader than quantum chemistry. For example, the entire area of force-field based simulations for chemical problems has been omitted as far as I can see. Other missing important areas were mentioned in some other comments on the web.

These example and others like it, lessen the strength of this document.

However, this can fairly easily be fixed. Someone should go through the document with this issue in view, to assure that it is clear that PRACE knows that PRACE deals with infrastructure. Noone should get the impression that PRACE is trying to write the entire roadmap for all of the HPC-dependent science in Europe for the next decade. This can become a very valuable document then.

2) Point 8 on p. 6 reads: "8. The primary objectives of this update to the scientific case are to identify the scientific areas for which PRACE is an important Research Infrastructure and the key challenges within these areas. . . ."

The sentence would be more accurate if the word "the" was deleted from the phrases "identify the scientific areas" and "the key challenges". The reasons were discussed under point 1.

Sincerely, Kersti Hermansson. Department of Chemistry - Angstrom, Uppsala University; Theoretical Chemistry, The Royal Institute of Technology, Stockholm.