Volume 3 Edition 5

DEAR READERS

Scott Williams has responded to Anne de Paepe’s and my own article from the last issue – REF: P.J. Hendra & A. de Paepe, J. Internet. Vib. Spec.[www.irdg.org/ijvs] 3, 4, 3 (1999). You will remember that we discussed the problems we all have in presenting infrared and Raman data especially presenting it for meetings and discussions within our organisations. I asked you all to reply to a series of questions – so here goes…

  1. Do you agree that a “standard” presentation format would be valuable both in-house and in comparing printed data between labs?
    A standard format would be useful. But, I think a standard format already exists in printed atlases of spectra. (I have given the bibliographic info for several atlases that I think might make useful additions to your Spectroscopists’ Bookshelf.) I think the same standard format should be used everywhere that a standard format is acceptable (see below for non-standard formats). The formats used in several atlases are linear %T vs linear wavenumber, usually from 4000 to 400 wavenumber, with a 2x scale expansion of the fingerprint region below 2000 wavenumbers, with grids having grid pitches of about 100 wavenumbers per 6 to 8 mm below 2000 wavenumber (compared to yours of 100 wavenumbers per 10mm).

    Hummel’s Atlas of Polymer & Plastic Analysis (Spectroscopists’ Bookshelf #2) has used this since at least 1981. The same format was used by the Federation of Societies for Coatings Technology in An Infrared Spectroscopy Atlas for the Coating Industry (Fourth Edition, 1991, 2507 spectra , 100 wavenumber per 6.4mm; Third Edition, 1980, 1400 spectra), and in Sadtler atlases such as The Infrared Atlas of Monomers and Polymers (1980, 2000 spectra, 100 wavenumber per 6.1mm) and Standard Spectra (3-ring binder format, 1972, 100 wavenumber per 8.1mm).

    The Sigma Library of FT-IR Spectra and Aldrich Library of FT-IR Spectra use linear %T vs linear wavenumber, but with wider wavenumber range and three zones of scale expansion: 1x from 4600 to 2200, 2x from 2200 to 1000 (fingerprint region), and 4x from 1000 to 400 (aromatic ring substitution region). This scale change does not cause any significant confusion when comparing spectra with different ranges and expansion. In addition to aromatic ring substitution, it is a useful format for mixtures containing organics and inorganics, such as paints and mineral filled polymers.

    These are very readable formats and can be easily compared to each other and to
    spectra generated by a variety of spectrometers, AS LONG AS ABSCISSA IS LINEAR IN WAVENUMBERS. Trying to compare spectra in wavenumbers to spectra is microns is unpleasant.

    2. Do you like the format proposed in Figure 5? If you have your doubts, please tell us more. If you don’t like our suggestions please send an alternative.
    Since this format is linear in wavenumber and has a decent pitch, it is ok. However, for full range spectra I prefer the format with the 2x scale expansion below 2000 wavenumbers because this makes the fingerprint region easier to read. For close-ups of small wavenumber ranges (usually of a particular band or cluster of bands), no changes in scale expansion should be used, or are needed. By the way, although the printed format of your publication referred to in Fig. 5 may have given a pitch of 10 wavenumber per mm, when I printed your IJVS article from the Web site, my printed copy of your spectra in Fig. 5 had a pitch of 10 wavenumber per 0.63mm (100 wavenumber per 6.3 mm), which is in the range of the printed spectral atlases discussed in Question 1.

    Your format uses Absorbance. This is not common in IR spectral atlases but is quite common in journal articles and any quantitative work. Digital databases use Absorbance and searches are based on Absorbance spectra so Absorbance is becoming much more used. There is not much confusion in comparing Abs and %T
    formats. It is a lot easier to mentally perform the vertical mirror image transformation of Abs to %T than it is to do the microns to wavenumbers transformation. The choice of whether to display as %T or Abs depends on the purpose at hand. Quantitation and search matching require Abs, observation of features just above the baseline seem better in %T.

    I do not think there is too much merit is specifying standard regions for standard spectra. I work in IR spectroscopy of art and art conservation materials and use three different spectrometers with different “standard” accessories each having a different spectral window according to its optical limitations:

empty spectrometer 4000(5000) to 400
with diamond cell 4000 to 400
with ZnSe ATR 4000 to 700
with IR microscope 4000 to 700
with 1m chalcogenide glass fibre optic probe 1900 to 900

I usually plot 9.25 inch (235 mm) wide spectra (no particular reason, that’s just what the software gave me) in landscape format on 8.5 and 11 inch Canadian paper with the wavenumber regions as indicated above. I constantly compare paper copies of spectra collected on one of these instrument – accessory combinations to other paper spectra in our archives collected on all these different instrument./accessories as well as older instruments, and to published spectra in journals and atlases like those those mentioned above. There does not seem to be too much difficulty in doing this as long as all spectra being compared are linear in wavenumbers and have either a grid or many peak positions marked so that interpolation is easy.

I take a small exception to your comment in the caption to Fig. 5. IR spectra would not be displayed on these axes. I think general purpose mid-IR FTIR spectrometry is usually done between 4000 (maybe closer to 5000 these days) and 400 (maybe 200) wavenumbers. The spectral regions that are typically reported or displayed for IR and Raman regions are slightly different: 

                               IR                            4000(4600) to 400 (200)

                             Raman                     3400 to 0 (or is that 140 on your spectrum?)

  1. Where infrared and Raman spectra are to be overprinted together on the same diagram, is it appropriate to use two differently coloured lines?
    Hence – I always try to use two or more differently coloured lines when I plot more than one spectrum on a chart. It does not matter whether the spectra are IR and Raman, or IR and IR, or Raman and Raman. If there is more than one curve they should be different colours. In fact, I think most spectrum plotting (or general curve plotting) software defaults to multiple colours. Whether or not colour printers are available, different colours should always be used for display on colour monitors on computers.

    4. Do you use a colour printer? Are colour printer facilities widely available in your organisation?
    Yes.

    5. Do you prefer having the band cm-1 values printed directly onto the chart or do you prefer a table (remembering how easy it is to estimate the wavenumber value within the format we suggest)?
    To answer this, I think another observation about the printed spectral atlases is important. They all have a grid with %T and wavenumber values labelled in logical evenly spaced integral increments (usually on the 100s), and none have peak positions (wavenumbers) marked directly on the chart, although some do have peak tables in the heading (Aldrich and Sigma). I think a labelled grid with wavenumber lines spaced no more than 1 cm apart on the printed format is essential if the peak positions are not marked. If peak positions are marked, the position should be indicated by a tick mark. Marked peaks are a convenience to indicate peaks of interest but inconvenient to use for interpolation of unmarked peaks, requiring the use of a ruler and awkward math. Also they clutter up the chart when many peaks are marked. On the other hand, when a grid is present, interpolation can be done at a glance, and there is no clutter if the grid is carefully designed to be unobtrusive but still visible.

    If there is a grid, peak values printed on the chart are unnecessary and probably will cause clutter. If there is no grid, then some peak values WITH A TICK MARK AT THE POSITION OF THE INDICATED WAVENUMBER are essential to enable
    interpolation of peak positions for unmarked peaks.

    I think the spectra in the paper, Non-destructive identification of minerals by Raman microscopy by Ray Frost, Theo Kloprogge, and Jolene Schmidt, which preceded yours in the same IJVS demonstrates perfectly the point that a grid is essential. Figure 3 has spectra of calcite, azurite, and malachite, which are three pigments commonly encountered in paintings often all together in the same paint layer. If I used Raman for paintings analysis (I do not, only IR) I would be very interested in the exact wavenumber at which each of these absorbed. I would have to do a lot of measuring and calculating on paper copies of spectra from that figure to get the values. If there was a grid present at say every 100 wavenumber, I could read the peak position values right off the screen of my computer. The same goes for anatase and brookite in Figure 2, and gypsum and anhydrite in Figure 5 (which have bands at very nearly the same values and therefore need very careful interpolation to distinguish).
    R. Scott Williams, Senior Conservation Scientist (Chemist)
    Conservation Processes and Materials Research
    Canadian Conservation Institute
    email: scott_williams@pch.gc.ca

Bob Lancashire can always be relied upon to think about a computer question and give us all a considered view. Here is Bob’s response.

  1. Do you agree that a “standard” presentation format would be valuable both in-house and in comparing printed data between labs?None of our IR machines now have any recorder attached to them and info is passed directly to PC’s. Output is therefore dependent on the software available and your “standard” might be difficult to set up as a template for all users.

It would be feasible to have 2 spectra plotted but setting the 10 cm-1 per mm might be tedious to arrange and keep consistent.

I have never used a split display and prefer one long chart or several charts showing areas of interest. This is so easy to do with present day software. I wonder whether your “standard” will gain much support.

  1. Do you like the format proposed in Figure 5? If you have your doubts, please tell us more. If you don’t like our suggestions please send an alternative.
    The format looks OK except our machines start at 4400 cm-1so we would probably do the changeover at 1800??
  2. Where infrared and Raman spectra are to be overprinted together on the same diagram, is it appropriate to use two differently coloured lines?
    Different colour inks fade at different rates and for journals not as sophisticated as IJVS, is not often possible. Our students automatically use colour while the ink jet cartridges are full!

Hence –

  1. Do you use a colour printer? Are colour printer facilities widely available in your organisation?
    Yes ink jets are cheap to buy although not so cheap to run.
  2. Do you prefer having the band cm-1 values printed directly onto the chart or do you prefer a table (remembering how easy it is to estimate the wavenumber value within the format we suggest)?
    This is an easy option with software packages I suspect. I always print with cm-1 showing, but then I don’t print with Absorbances so maybe not a “real” IR spectroscopist?????

I have seen a shareware program (from Virginia Tech) where the author of the IR program had several options for plotting, including:

  1. a) standard 4000-650 linear scale
    b) 4000-2000 one scale and 2000-650 another scale

I can’t remember any other package that mimicked the old recorders in this way.

Dr. Robert J. Lancashire E-mail mailto:rjlanc@uwimona.edu.jm
Sub-Dean, Technology Management and Development
Department of Chemistry,University of the West Indies, 
http://wwwchem.uwimona.edu.jm:1104/chrl.html

Now a contribution from Ian Wesley in Oz.

Any house style has to be at least reasonably back compatible with existing spectral libraries, especially those which do not exist in some digital format. I suspect that shortly, this will not be an issue as all the major libraries will be available in digital form.

1. Do you agree that a “standard” presentation format would be valuable both in-house and in comparing printed data between labs?
Definitely, but what about the independent publishers of spectra (databases, journals?)

2. Do you like the format proposed in Figure 5? If you have your doubts, please tell us more. If you don’t like our suggestions please send an alternative.
I do prefer the idea of the whole spectrum on a single trace instead of split across two as in Figure 5. To me, it just looks slightly strange, particularly as there is an overlap (2000-1400 appears on both traces). Another thing to consider is the type of software available – any standard format must be able to cope with the lowest common denominator graphic package that an instrument manufacturer chooses to supply and also older packages, which may still be in use. It also seems to me that the style you
propose has the disadvantage that extra steps are need to go from the spectral package to the printout you suggest. I do not know of any software which can produce that type of output. This is the death knell of any proposed standard – everyone must be able to get to the standard format from within their own favourite spectral software without the need for exporting to a graphics/spreadsheet/presentation package.

3. Where infrared and Raman spectra are to be overprinted together on the same diagram, is it appropriate to use two differently coloured lines? Hence – 
Now we get to the crux of the matter! Overprinting of spectra. As we all know, the average computer program can probably produce a dozen or so distinct colours, any many thousands in between. Colour is great if you have a fast printer and plenty of the right colour ink, but its a real pain if you have a laser printer, or worse, there is a scaling problem between the two spectra (eg 1 is 100 times more intense than the other). In this case, stacking the spectra (ie printing each on its own axis on the same page, vertically shifted and autoscaled) solves the problem, but only for a few spectra (I reckon 4 maximum on landscape A4, more if portrait is used). In any case, what’s wrong with a light box!!!!! 

4. Do you use a colour printer? Are colour printer facilities widely available in your organisation?
Only if my network connection is active (which it isn’t at the moment). 

5. Do you prefer having the band cm-1 values printed directly onto the chart or do you prefer a table (remembering how easy it is to estimate the wavenumber value within the format we suggest)? 
I prefer marked peaks – as long as the spectrum is not too complex.

Dr. Ian Wesley
Grain Quality Research Laboratory, CSIRO Plant Industry
email: I.Wesley@pi.csiro.au


Now a response from Gavin Phillips here in the UK.

I read with enthusiasm your article in the recent edition of what can only be described as a superb journal. I have to say that although the need to compare spectra from other labs with that of my own are very seldom, the annoyance of realising that modern software automatically enlarges or reduces spectra scales make internal comparisons difficult enough.

I would therefore add my voice to your call for a standardised approach, if nothing else this would be a good courtesy to show towards our other learned colleagues. However, regarding point 2, the extensive use of computers and printers nowadays means that a good readable spectra can be put onto a single sheet of A4 in Landscape. If there is a need to home in on a particular region or detail, then as suggested a break from the standard would be acceptable. The use of a grid for picking out the cm-1 of peaks is a good idea, but the software I use has only the facility to mark lines every Xcm-1‘s. Making it impossible for me to guarantee a certain number of wavenumbers per unit length. I also suspect many other users of modern software will also have this problem.

Your third proposal, is in my opinion entirely acceptable. I already do this when overlaying IR spectra (not being a user of Raman). I suspect this answer your fourth question.

The final question is one of personal preference. I personally do not label “peaks”. But if a particular peak was being looked at in the spectra of several samples then why not label it? It would depend largely on the number of grid lines used and the colour of them and the label. I am fortunate in being able to change the label or grid line colours.

Gavin Phillips. LRSC BSc(Hons).
EEV Ltd.
Email: Gavin.Phillips@eev.com

And finally a piece from Peter Kauper in Lausanne in Switzerland…

With this e-mail I would like to answer your questions in your feature article “3. Formatting Spectra” in IJVS volume 3 edition 4 section 1 (1999).

I hope I can support your efforts to establish a standard in the “paper” presentation of i.r. and Raman spectra. Hopefully we will find a “Standard Presentation” button implemented in every future’s spectrometer software release. I agree that a paper based spectra library is still a valuable and powerful tool – at least for in-house meetings.

Now the answered questions of your article:
1) YES
2) YES
3) Overprint with different colours is favorable
4) YES
5) Band cm-1 values NOT directly printed onto the chart.

Dr Peter Käuper 
Départment de Chimie, École Polytechnique Fédérale de Lausanne
E-mail: kaeuper@dcmail.epfl.ch


Editor’s Comment

Anne and I were overwhelmed by your response. Experience tells me that when we ask for your response it’s not a good idea to hold your breath – this time your replies flowed in quickly. Please can we have some more? Please!

Quite clearly there is a concern about routine data presentation. Ideally we all want a standard presentation that looks good, covers the whole m.i.r. (or Raman) in one presentation and is adequately long that detailed spectra show up clearly and close multiplets don’t become ‘smudged’. To achieve all this we really need more than the landscape width of A4 and that is why the problem arises in the first place. We can’t re-invent printed charts and flat bed printers so where do we go?

An old – very old – solution is to use a scale split – say 4000-2000 cm-1 at one density, the rest at half. However, we should be careful before we adopt this solution. The reason why this format came about in the first place was that all early infrareds displayed in % transmission vs wavelength and hence people were used to seeing the CH stretch region all squashed up. Further, early infrareds used rock salt prisms and the dispersion of these prisms was so low at short wavelengths that the spectra around 3µ (3000 cm-1) were of dreadful quality. When gratings came along and presentation changed to wavenumbers, manufacturers simply offered the public what they thought the public should buy! i.e. split spectra.

These days with F-Ts there is no loss of performance across the spectra so there is no justification in squashing up one bit of the output just to squeeze the whole lot onto an A4 sheet.

Several people seem to like the idea of a nice logical presentation at say 10cm-1/mm but point out how difficult it is to achieve. We did it by mucking about with the output display – got the wanted cm-1/mm and then memorised the algorithm.

If we decided we wanted it and IJVS duly edicted – the manufacturers would supply it so I don’t think there is a real problem here.

Compatibility with ‘old’ data – yes, we would like to achieve this but the old data exists in a vast range of formats, so perhaps we should start anew. Our main aim was to assist people within their organisations when preparing data for discussions and meetings or for inclusion in internal reports so new data is what normally is involved.

REMEMBER OUR NEW PHONE/FAX NUMBERS WHEN YOU WISH TO CONTRIBUTE TO DEAR READERS. THE EMAIL ADDRESSES REMAIN THE SAME.

THESE NEW NUMBERS APPLY FROM 15TH OCTOBER 1999.

Office Phone: +44 (0) 1703 390935  (answer machine)
Office Fax: +44 (0) 1703 390935
Office Email: ijvs@soton.ac.uk or louise@ijvs.demon.co.uk UNCHANGED
Editor’s Fax +44 (0) 1962 776390 UNCHANGED