Peak overlap detection by multi detector methods

The most fre q u e n tly used d etecto r in chrom atography is th e u ltrav io let d etecto r. Most of th ese d e tecto rs o p erate a t fixed w avelength and

m easure th e absorbance of th e eluent. O ther more advanced UV d etecto rs can d etect a t more th an one w avelength - both dual and m ultiwavelength a re available, th e most sophisticated being th o se capable of scanning over a spectrum of w avelengths. These d etecto rs a re known as photodiode a rra y s . UV d etecto rs a re relativ ely in sen sitiv e to flow ra te or tem p eratu re changes b u t th e choice of mobile phase is re s tric te d as th ey must be tra n s p a re n t a t th e w avelength a t which th e d etecto r is operating ( Hamilton and Sewell 1977 ).

Another ty p e of d etecto r is th e re fra c tiv e index d etecto r. These work using a differential technique in which th e sample’s re fra c tiv e index is compared with th a t of th e eluent. Any su b sta n c e which has a re fra c tiv e index significantly d ifferen t from th e elu en t may be d etected b u t th e method is v e ry sen sitiv e to changes in elu en t and so is u n su itab le for g rad ie n t elution unless solvents a re chosen with identical re fra c tiv e indices. I t is also v e ry tem p eratu re d ep en d en t ( E n g elh ard t 1979 ). Compounds may be d etected using fluorescence. The p ro cess stream containing th e su b stan ce to be d etected is f ir s t excited by UV radiation and monitored for emission a t an o th er w avelength. Fluorescence detection is v e ry specific b u t fluorescence may be su p p re sse d or quenched by th e p resen ce of contam inants. The se v ere effect th a t su ch contam inants have on th is detection system means th a t carefu l selection of column conditions must be c arried out.

1.3.5.1 Difference and Ratio Methods

More th an one d etecto r may be used fo r detection. This has th e ad v an tag e of providing more information about th e sep aratio n eg. if th e sep aratio n contains one su b stan ce which flu o resces th en th e d eg ree of overlap may be m easured by comparing th e chromatogram obtained from UV with th a t from fluorescence.

Drouen, Billiet, and de Galan (1984) have used th e ab so rb an ce ra tio a t two d ifferen t w avelengths to provide more information about overlapping peaks. They defined th e ratio, RAT, in th e following way ;

RAT = (A1/A2) when A2 > Al > th resh o ld RAT = 2 - (A2/A1) when A1 > A2 > th resh o ld RAT = 0.0 when A1 and A2 < th resh o ld

RAT = -0.1 when A1 < th resh o ld and A2 > th resh o ld or when A2 < th resh o ld and A1 > th resh o ld .

Where A1 and A2 a re th e ab so rb an ces m easured a t th e two d ifferen t w avelengths a t any given point in th e chromatogram.

Initially th ey teste d th e ratio method using simulated fully-resolved peaks. They found th a t th e ratio produced block shaped resp o n ses when th e ratio was plotted ag ain st th e elution time. They also showed th e ratiogram provided information when th e simulated peaks were overlapped. When th e two peaks were fully resolved two blocks were shown on th e ratiogram co rresponding to each of th e peaks. Once peaks were overlapped th e blocks combined with a ste p change from th e fir s t value to th e second. As th e d eg ree of overlap increased th e change from th e f ir s t to th e second became smoother un til th e two peaks coincided exactly when a single block o ccu rred . These re s u lts were fo r p e rfe ct data ie. no noise, or baseline offset, or time delay between signals. Baseline offsets may occur because of incompletely co rrected background signal and may be variable or constant. They have th e effect of d isto rtin g th e blocks into asym metrical shapes. This effect can be red u ced by increasing th e th resh o ld value b u t th is has th e effect of making th e p u rity check of th e peak more difficult. Any d isto rtio n of th e ratiogram caused by time delays may be removed by a d ju stin g th e d etecto rs so th a t th ey a re m easuring a t th e same w avelength and determ ining th e time delay between th e two by comparing chrom atograms. This time delay can th en be tak en into account. Tailing may also d isto rt th e ratiogram and lead to m isin terp retatio n s eg. more components ap p ear th a n a re actually p resen t. Again c le are r data may be gained by in creasin g th e th resh o ld value b u t th is red u c es th e amount of information available. The ability of th is method to identify overlapping peaks also depends on th e difference between th e RAT of th e two d ifferen t peaks. The g re a te r th e difference th e sim pler th e recognition of overlap. Drouen, Billiet and de Galan (1984) found th a t if th e difference in th e RAT of two components was less th an 0.05 th en th e peak sep aratio n must be g re a te r th an 2 o f o r overlap to be identified. This is still b e tte r th an single w avelength detection. The

selection of w avelengths for m ultichannel monitoring is im portant and it should be made to gain as much information as possible. Mathematical tech n iq u es ( key se t facto r analysis (KSFA) ) have been developed to determ ine th e optimum w avelength se ts ( Warren, Bidlingmeyer and Delaney, 1987a and Warren, Bidlingmeyer, and Delaney, 1987b ).

Drouen, Billiet, and de Galan (1985) extended th e ir previous work on dual w avelength d etecto rs to m ultiwavelength d etecto rs such as diode a rra y d e tecto rs which allow rap id scanning and a more detailed analysis of overlapped peaks to be obtained by m easuring th e ab so rb an ce spectrum of th e eluent. This in tu r n gives more a cc u ra te analysis and identification of sev erely overlapping components. However with proteins, since th e ir absorbance spectrum s a re often v e ry similar, diode a rra y method would be essen tial ( C arr e t al, 1988 ).

A facto r analysis method has been developed for th e resolution of unresolved peaks ( Sakema, e t al, 1990 ). Using th is method estim ates can be made of th e elution profiles of components eluting a t th e risin g or tailing edge of a peak. With unresolved th re e component peaks, peak a re a of th e th re e components were determ ined to within 10%. Methods have been developed ( Marr, Seaton, Clark and Fell, 1990 ) using modern m utliw avelength d e tecto rs for examining th e homogeneity, id en tity and p u rity of peaks, based on work for stopped flow conditions ( Ostojic, 1974 ).

Krstulovic, Rosie and Brown (1976) used th e ratio of peak a re a s to id en tify chrom atographic peaks which were fully resolved b u t p u rity was not considered as th e method was p u rely for identification p u rp o ses. O ther methods have been devised fo r analysis of chrom atogram s m easured on two w avelengths. Li and A rrington (1979) used th e difference betw een chrom atogram s for th e identification of components as well as q uantification of components in overlapped peaks. They were able to find a relatio n sh ip between th e amount of su b sta n c e p re s e n t and th e d ep th of th e valley caused by th e su b tra ctio n of th e chromatograms.

1.3.5.2 D erivative Methods

G rushka and Monacelli (1972) developed a method of analysis u sing th e second deriv ativ e. They te ste d th is method on overlapped simulated Gaussian peaks and on real chrom atographic data. The method relies on mathematical knowledge ab o u t th e Gaussian function. The second

d eriv ativ e of th e function has two maxima and one minimum which a re positioned a t ( th e maxima ) and x=x^ ( th e minimum ). Where x^ is th e Gaussian peak’s c e n tre of mass and o is th e sta n d ard deviation. The ratio s of th e maxima and minimum a re in d ep en d en t of th e peak height. The ratio s of th e second d eriv ativ e maxima and minimums were defined as follows:

R| maximum on fro n t of peak / minimum Rg maximum on tail of peak /minimum

Rj maximum on fro n t of peak / maximum on tail of peak

A g rap h of R^ and Rg was obtained for v arious peak h eig h t ratio s of overlapping peaks onto which co n to u rs of resolution were draw n. Once c u rv e s such as th ese a re produced it is th en possible to id en tify peaks if th e peak width a re known since th e second d eriv ativ e ratio s will give information about th e peak h eig h t and resolution. This method has th e ad v an tag e of not being too d ep en d en t upon th e exact sh ap e of th e chrom atographic peaks b u t it is n e ce ssa ry th a t th e data is of a good and rep ro d u cib le quality and so may not be of g eneral use. The main disadvantage of th is method is th a t calibration sep aratio n s must be c arried out to d etect overlapped peaks and th e extension of th is method to multicomponent system s would re q u ire a larg e database of information about overlap of each peak with each o th er peak, eg. a combination of four peaks would re q u ire information about six perm utations of overlapping peaks. Measurement of ab so rb an ce ratio s would re q u ire much less information in comparison - th e value of th e ratio for each p u re peak which could be gained from one sep aratio n with all peaks fully resolved. A nother disad v an tag e is th a t a num erical d ifferen tiatio n m ust be c arried out which may not be a cc u ra te depending upon th e quality of th e data. D ifferentiation also causes magnification of noise in th e signal. The limits of th is method have been found to be d ep en d en t on th e peak h eig h t ratio and th e relativ e sep aratio n ( G rushka and Israeli, 1990 ).

1.3.5.3 Methods suitable for general use

Ideally th e method chosen for ju d g in g th e quality of th e sep aratio n should be g eneral enough to be applied to sep aratio n s based on th e same mechanism and on o th er mechanisms. Fluorescence detection is heavily d ep en d en t upon p ro p erties of th e components within th e m ixture to be sep ara ted and th e re fo re will not be considered. I t is also su b je c t to

quenching. The most v e rsatile method ap p ea rs to be measurement of ultraviolet absorbance a t two or more w avelengths as th is allows a number of d ifferen t methods of analysis:

(i) ratio s of area

(ii) ratio s of absorbance (iii) slope analysis

(iv) difference analysis

I t also allows th e concentration of th e solution in th e d etector flow cell to be determ ined using th e Beer - Lambert law ( see section 2.3.9 ). The difference in ab so rb an ce sp e c tra between d ifferen t p ro tein s is generally not th a t g re a t and so a photodiode a rra y would be req u ire d , to g eth er with superim position tech n iq u es. ( C arr e t al, 1988 and Frank, Braat and Buine, 1987 )

A combination of th ese methods could be used to gain as much information about th e chromatogram as possible. For example ratio s of absorbance and a re as could be used for peak identification and difference analysis for quantification.

In addition to being g eneral in term s of sep aratio n and mechanism any method should also be applicable to both overload and analytical conditions. Of th e methods of analysis stu d ied only th e fractional overlap param eter Q ( Guiochon 1990 ) is applicable to non-G aussian peaks.