WO2001094897A1

WO2001094897A1 - Calibration and use of spectroscopic apparatus

Info

Publication number: WO2001094897A1
Application number: PCT/GB2001/002442
Authority: WO
Inventors: Robert Bennett; Brian John Edward Smith; Yan Yong Yang; Ian Mac Bell
Original assignee: Renishaw Plc
Priority date: 2000-06-02
Filing date: 2001-06-01
Publication date: 2001-12-13
Also published as: GB0013408D0

Abstract

Scattered light from a sample is dispersed at a diffraction grating (12) and focused onto a detector (16) comprising an array of individual detector units (n). The dispersion across each individual detector unit (k(n)) is calculated for a range of angles (α) of the diffraction grating (12). The intensity at each individual detector unit (I(n)) is corrected by dividing it by the dispersion at that detector unit (k(n)) for the same angle $(g(a)) of the diffraction grating (12). The collected spectral data is thus independent of dispersion.

Description

CALIBRATION AND USE OF SPECTROSCOPIC APPARATUS

This invention relates to the calibration of spectroscopic apparatus. It also relates to methods of use of such apparatus when calibrated.

Our US Patent No. 5,689,333 describes a known spectroscopic apparatus, particularly intended for Raman spectroscopy. A sample is illuminated by a laser beam. The resulting Raman spectrum is dispersed by a diffraction grating across a multi-channel detector such as a charge coupled device (CCD) .

Prior to use, it is normal for such apparatus to be calibrated. For example, it is necessary to know the wavenumber k corresponding to each position x across the detector. This may be referred to as an X-axis calibration. Additionally, for each given wavenumber, an intensity calibration may be performed using a standard white light source to correct for the response of the instrument. This can be termed a Y-axis calibration (since spectra measured by the apparatus are commonly displayed in the form of a graph with intensity on the Y axis and wavenumber on the X axis) .

US Patent No. 5,689,333 also describes methods by which a wide spectral range may be measured at high spectral resolution. The patent discusses a prior art step-and- repeat method, in which one part of a wide spectrum is exposed onto the detector, with wide dispersion giving high spectral resolution. The data for this partial spectrum is stored in a computer. Next, the diffraction grating is indexed to a new rotary position, so that another part of the spectrum is exposed onto the detector, and its data stored in the computer. This is repeated until all the data for the full spectrum has been acquired. The data is joined or "stitched" together by subsequent computer processing in order to display the full spectrum. However, in practice, it is found that there are intensity mismatches between the separate blocks of data acquired from the separate parts of the spectrum, making it difficult to stitch the data blocks together. There are also problems in the step-and-repeat method if the optical system causes vignetting (distortion at the edges of the detector) since the full width of the detector is used.

To solve this problem, US Patent No. 5,689,333 describes an "extended scanning" technique in which the grating is rotated to cause relative traversal of the dispersed spectrum across the detector. Synchronously with this traversal, the accumulating data is transferred from one element of the detector to the next. The data continues to accumulate during the relative traversal. This method gives good results in the apparatus described in US Patent No. 5,689,333. However, it has been found that the method cannot be used in apparatus with a significantly shorter focal length and lower dispersion than that described, because the dispersion is then significantly different at one end of the detector compared to the other.

Preferred embodiments of the present invention will now be described by way of example, with reference to the accompanying drawings, wherein:

Fig 1 is a schematic diagram of the layout of a spectroscopic apparatus, Fig 2 is a diagrammatic representation of a multichannel detector used in the apparatus of Fig 1,

Fig 3 is a flow chart of a method performed in a computer in the apparatus of Fig 1 ; and Fig 4 is a diagrammatic representation of a photodiode array used in the apparatus of Fig 1.

In Fig 1, a polychromatic input beam 10 is to be analysed spectroscopically. This is performed by a diffraction grating 12 which disperses the incoming beam into its component wavenumbers . A lens 14 focuses the dispersed spectrum onto a multi-channel detector 16 such as a CCD. The output of the CCD is acquired and subsequently processed by a computer 18. The grating 12 may be rotated through an angle c. about an axis normal to the plane of Fig 1, in order to bring different parts of a widely dispersed spectrum into register with the detector 16.

The wavenumber k at a position x on the detector, for a given angle a , is given to the first order by the equation:

where : d is the focal length of the lens in mm x is in mm and equals the number of pixels from the centre of the detector times the size of a pixel g is the grating constant in grooves/mm m is the grating diffraction order k is wavenumber of light in absolute cm^"1. φ is the included angle at the grating of incoming light to the centre of the CCD α is the angle of deviation of the grating. It is measured from zero where the grating reflects light directly to the center of the CCD i.e. the zero order.

In the following, the width of a single pixel of the CCD may be considered as a detector unit. However, sometimes several pixels may be binned together, and the combined width of these binned pixels may then be taken as a single detector unit. Alternatively, a multi-channel detector other than a CCD may be used, such as a linear photodiode array, and one element of such an array may then be considered as a detector unit .

An X axis calibration may be performed on the apparatus, making use of the above equation (1) and of a measurement upon a reference sample having peaks at known wavenumbers. This calibration should be performed across the range of angle o. of the grating.

The X axis calibration may be used during subsequent processing of spectral data in the computer 18, in order to produce graphical outputs of spectra which are evenly spaced in wavenumber k, taking into account the fact that the acquired data is taken from detector units which are not evenly spaced in wavenumber k.

Figs 2 and 4 show sections of a detector 16 which comprises a CCD 30 in Fig 2 and a photodiode array 32 in Fig 4. The individual photodiodes of the photodiode array are separated by dead spaces 36. Referring to both Figs 2 and 4, the position x₀ represents the centre of the detector. For a detector unit n, Xi and x respectively represent the distances of each side of the detector unit from the central position. The wavenumbers k(xι) and k(x₂) for the positions Xj. and x₂ are derived (for different angles α.) from the X axis calibration performed in conjunction with equation (1) above. Then the local dispersion for this particular detector unit n, at a given angle c-, is

Δk(n) = k(xα) - k(x₂) (2)

This dispersion calculation is carried out in the computer 18 for each detector unit and each angle a used.

Fig 3 is a flow chart showing a data acquisition method performed in the computer 18 when taking a spectrum from a sample to be analysed. The method is performed for each detector unit n, starting with the first

(n=l) . In a step 20, the measured intensity data I (n) for that detector unit is selected. In step 22, the local dispersion Δk(n) for that detector unit is calculated (or, if desired, may be read from a previously calculated look-up table) . This local dispersion is derived as discussed above, depending upon the angle o! as well as the position x of the detector unit.

Next, in step 24, the intensity I (n) is corrected, by dividing it by the dispersion Δk(n) for the particular detector unit concerned. In step 26, the value of I (n) as corrected by the division in step 24 is stored. As indicated at 28, the loop is then repeated for all other detector units n.

The result is that all the acquired intensity values, for each detector unit n, have been corrected by division by the local dispersion at the particular detector unit concerned. This is a notable difference between the present method and the prior art . Traditionally, a spectrum is interpreted as intensity per detector unit (e.g. intensity per pixel) at each data point. In practice, however, the intensity per detector unit varies with wavenumber and with the angle of diffraction, because of dispersion (amongst other effects) . The present method therefore uses intensity per unit of wavenumber, rather than per detector unit.

The above method is particularly advantageous when using the known step-and-repeat method to acquire a wide spectrum with high spectral resolution, for example in apparatus where the extended scanning method of US Patent No. 5,689,333 cannot be used for the reasons discussed above.

The present inventors have realised that one of the problems in stitching together partial spectra in the step-and-repeat method is as follows. The conventional step-and-repeat method assumes that, for a given wavenumber, the dispersion is constant. In fact, this is not so, because the dispersion also depends upon the angle of diffraction at the grating 12, which in turn depends on the angle α shown in Fig 1. Thus, in one partial spectrum, a given wavenumber k may occur at the right-hand end of the detector, at an angle a__.. In the adjacent partial spectrum, the same wavenumber k will occur at the left-hand end of the detector, at an angle 0.₂. Since the local dispersion is not the same in these two measurements, this is a cause of the mis- match between the two partial spectra. Dividing the intensity by the local dispersion, as in step 24 of Fig 3, corrects this mis-match. This is particularly so in the preferred method, where the local dispersion is derived separately for a range of angles a . The correction produces the intensity per unit of wavenumber instead of the intensity per detector unit.

While it is possible in conventional apparatus for the local dispersion Δk(n) to be calculated, this is normally only performed in order to determine the spectral resolution at a particular part of the measured spectrum (e.g. in the region of an interesting peak) . It is not conventional to calculate the dispersion Δk(n) for every unit detector, nor to use this local dispersion information to correct the measured intensity data, nor to derive separate local dispersion values for a range of angles c..

In the conventional step-and-repeat method, the usual method to hide the mis-matches when the partial spectra are stitched together is by averaging the data from a few overlapping detector units from the adjacent partial spectra. This is an artificial manipulation of the data which is less accurate than the method described above.

A further refinement of the above method according to the invention will now be described. As already mentioned, a conventional Y axis calibration (calibration for the instrument response) makes use of a standard white light source to provide the incoming beam 10 in Fig 1. Conventionally, the instrument response is determined by measuring the intensity output of the instrument for each wavenumber k, and comparing this with the known pre-calibrated output of the light source for that wavenumber. This provides a correction value for the wavenumber concerned. A table of such correction values, for all wavenumbers, is stored in the computer 18 and used to correct the measured intensity values at the corresponding wavenumbers during the analysis of a sample.

In the preferred method of the present invention, account is taken of the inventors' realisation that the instrument response varies not only with the detector unit giving the output value, but also with the angle c. of the diffraction grating. The efficiency of the grating varies with the angle . ; and also a change of angle . means that a given wavenumber k is now measured by a different detector unit which may have a different sensitivity.

Thus, the calibration with the standard white light source is performed not only for each wavenumber k, but also it is repeated for a plurality of different grating angles o.. For example, the detector may be exposed to the spectrum of the standard source at a first angle of the grating, and the correction values for each wavenumber k stored in a first row of a two-dimensional matrix in the computer 18. Next, the grating is rotated to another angle Q!₂, corresponding to a shift of (say) 15 or 20 wavenumbers at the centre of the detector 16. Again, correction values are recorded for each wavenumber k, and stored in a second row of the matrix. The process is repeated at angles c-₃, 0.₄, etc, each representing a change of 15 or 20 wavenumbers, and each producing a set of correction values stored in further rows of the matrix. The procedure is repeated to cover the whole range of angles c which may be used.

Thus, one dimension of the matrix represents correction values for the intensity at given wavenumbers on given detector units (e.g. CCD pixels) . The other dimension of the matrix represents correction values corresponding to the grating angle at which a measurement might be taken.

During an actual measurement on a sample, each data point in the measured spectrum is corrected as follows. Knowing which unit detector n (or which wavenumber) and which angle c- the data point was taken at, the four correction values which surround it in the matrix are taken. These represent detector units n (or wavenumbers) and angles c- immediately above and below the measured data point. From these four correction values, the correct correction value to be used on the current data point is calculated by interpolation, and applied to the data point. The data point is thus corrected for the system response, taking account not only of the detector unit n and wavenumber, but also the angle c- at which that data point was measured.

Referring to Fig 3, the correction just described may be applied to the intensity values I (n) either before they are divided by the local dispersion Δk(n) in step 24. Alternatively it may be applied after this step, when the value I (n) is stored (step 26) in which case the correction must also be divided by the local dispersion Δk(n) before being applied to the stored intensity value I (n) .

In the above description, reference has been made to particular grating angles c-, both in the calculation of the local dispersions Δk(n) and in relation to the correction values stored in the matrix. However, it is equally possible if more convenient to refer to the corresponding wavenumber k₀ at the centre x₀ of the detector, which results from the particular angle at. A corresponding equation to equation (1) may be derived, and the values in the matrix may referenced in terms of values of k₀.

The methods described above, between them, provide a number of advantages .

Firstly, they enable accurate stitching together of partial spectra in a step-and-repeat method, by solving the problem of mis-match at the joins between the partial spectra.

Secondly, it is possible to accurately compare data obtained at different parts of the detector (different detector units) . For example, an intensity obtained at a given wavenumber k on one detector element may be accurately compared with an intensity obtained for the same wavenumber k on a different detector unit (at a different angle c.) because the method described provides calibration values for different angles a . Comparisons between different wavenumbers k are also more accurate .

Thirdly, if there is a vignetting problem at the edges of the detector, the calibration nevertheless compensates for this .

Fourthly, if a standard light source is used to calibrate two separate instruments, data acquired from the two instruments can be accurately compared.

Claims

1. The method of collecting spectral data from a sample comprising: illuminating the sample so as to produce a spectrum of scattered light; dispersing the scattered light into its component wavenumbers by a dispersive means; and detecting the dispersed light at a detector comprising an array of individual detector units; characterised in that the spectral data is made independent of dispersion across individual detector units by correcting the light intensity at each of a plurality of said individual detector units by compensating said intensity for dispersion at the respective individual detector unit.

2. A method of collecting spectral data according to claim 1 wherein said compensation comprises dividing the intensity by dispersion at the respective individual detector unit .

3. A method of collecting spectral data according to claim 1 or 2 , wherein the dispersive means may be rotated to bring different parts of the dispersed spectrum into register with the detector and wherein the value of the dispersion at an individual detector unit used to correct the intensity at said individual detector is derived using the same position of the dispersive means as when intensity at said individual detector unit is detected.

4. A method of collecting spectral data according to claim 3 wherein calibration of instrument response using white light is performed for each wavenumber at a plurality of different angles of the dispersive means.

5. A method of collecting spectral data according to claim 4 wherein the data is corrected by interpolation of correction values relating to the detector units and angles of the dispersive means immediately above and below the measured data point .

6. A method of collecting spectral data according to any preceding claim wherein the dispersive means is a diffraction grating.

7. A method of collecting spectral data according to any preceding claim wherein the detector is a charge- coupled device.

8. A method of collecting spectral data according to any preceding claim wherein the detector is a linear photodiode array.

9. A method of collecting spectral data from a sample according to any preceding claim wherein each individual detector unit comprises several detector units binned together.