q/I 53
AnIntroduction
toMulti-Component
PeakPurity
MarcV. Gorenstein,Ph.D.and JeanneB. Li,Ph.D.
_'
One of theclassic dilemmasof chromatography
when usingsingle-wavelengthabsorbance detec-
tots, is that coeluted compoundsmay appear as a
singlepeak. Photodiodearray (PDA}-baseddetec-
rorsadd a new dimensionto therecord
rants.PDA-baseddetectorscan detectionof coelu-
absorbance
acrossthe completeUV/Vis spectrumat each time
point during an HPLCseparation.Within a given
peak, the comparisonof absorbance spectracan
revealthe presenceof coelutants.Visualcomparison
of absorbance spectracan be augmentedwith the
useof peak purity algorithmsthat allow the experi-
menterto detectcoelution at very low levels,
Thisarticle describestheMulti-ComponentPeak
Purity_ (MCPP)algorithm recentlydeveloped by a
team of scientistsat Waters Corporation. TheMCPP
algorithm can determinethe numberof spectrally
distinctcompoundswithin a coeluted peak.
1 _
1 " _4
_0.1
0 sampling interval,greatly enhancingthe ability to
EluUon Time
FigureI: Simulatedsinglechannel
chromatogram of fourcompounds
©
MCPP is implementedin theMillennium
2010/2020 ChromatographyManager, Rev.2.1
and the
greater, as an extensionto sensitiveSpectral
Contrast peak purity algorithm.
TM
Forthe sakeof thisarticle, firstwe will review coelu-
tion detectionwith a single-wavelengthdetector,
and secondwith a PDA-baseddetector;finally we
will discussthe useof theMulti-ComponentPeak
Purityalgorithm.
l
r-,Chromato"ra'hv
--
I111
- :,
II
Using
.11
_in,,le iivii _11
yvavelengln Detectors
As seen in thesimulationof a chromatogramin
Figure 1, thesolid trace was generatedusing a sin-
die wavelengthdetector. Thesamplecontainsfour
compounds.Compound 1 is baseline-resolved,
while compounds2, 3, and 4 coelute to form a
singlefused peak. Thesimulatedchromatogram
demonstratesthat the profile of a fusedpeak can
appear the sameas the profile of a baseline-
resolvedcompound. A singlewavelengthdetector
will missthe coelutantsunder the second peak.
PDA-baseddetectorson the other hand, can record
absorbancesacrossthe entireUV/Vis spectrumdur-
ing a separation.Using the Waters 996 PDA, a
completeabsorbance spectrumis recordedat each
detect coelution.Following are data demonstrating
how absorbance spectraassociatedwith a particu-
lar peak are displayed.
j Comparative
Absorbance
Spectra
Analysis
Theplots in Figure2 overlay spectraassociatedwith
a chromatographicpeak. Bothplots are of the same
peak, representingthecoelution of threecom-
pounds.The lower panelsdisplay the chromato-
graphic profile; theupper panelsdisplay spectra
from the peak (notethat thewavelengthaxis isverti-
cal.) Timesof the spectraare indicated by the verti-
cal lineson the lower panels.
Theleft plot displaysabsorbancespectra recorded at ThreeSpectra FourSpectra
threetime points--the peak apex and the inflection
points.Thesethreespectraoverlay exactly, suggest
ing spectralhomogeneitywithin the peak.

Butthesespectrado nol paint a complete picture of

thisseparation; it is importantIo compare spectra
from all parts of a peak in order to identify tailing or
frontingcoelution.The right plot overlaysfour spectra
from the samepeak. Thisadditional spectrumwas
chosenby the peak purityalgorithm at the Maximum
ImpuritySpectrum.
To be more specific: if a peak is in fact pure, with no
coelution, the shapeof the absorbance spectra
would be uniformacrossthewhole peak. However,
becausethe software is able to identify a spectrum
that,when overlaid with the apex spectrum,is clearly
differentin shape, the peak is spectrallyheteroge-
neous,indicating the presenceof coelution. Given
that the spectraacrossthe peak are non-homoge-
neous,the systemidentifiestheMaximum Impurity
Spectrumi.e. the spectrumwhose shape differsmost
significantlyfrom the apex spectrum.
Figure
2: Thepeakdisplayed
above(left andright)usingtheSpectrum
IndexPlotisa coelution

PeakPurityPlots of threecompounds.
Equipment: Waters996 PDADetector, 717 PlusAutosampler,
616 SolventDeliverySystem,
The PeakPurityPlotcompareseach spectrumin an Millennium 2010 Chromatography Workstation
integratedpeak to the spectrumat the peak apex. Chemistry:Nova-Pak ®C18, 3.9 x 75 mm,Water70%, acetonitrile
30%, phosphoricacid0.015%
This is accomplishedthroughthe useof vector Flowrate: !.0 mL/min
analysis.Eachspectralshape is representedas a PDAacquisition:! 95-350 am, 1.2amspectralresolution
vector direction,while concentrationin absorbance Compounds: Ethyl-para-amino
benzoate,ethylparabea,
acetophenone.
unitsis representedas the lengthof the vector. The Note:Thecompounds abovewereselected forthesakeofthisexampleduetothefactthatthey
PurityPlotplots the shapedifferencebetween the coeluteas a singlepeakundertheaforementionedchromatographicconditions.
apex spectrumand theother spectra.

Foreach integratedpeak, the Spectral Contrast

algorithmalso computesthe maximumexpected Coeluted
Peak PurePeak
shapedifferencebe_een each spectrumand the cII
apex due .tonoiseand solventeffects,assumingthe Fil_ J_diL
._l.iunu.l:l_lp
peak is spectrallypure. Thesethresholdvaluesare [-_Phv,
ncg_K,
ll'(uLIIIilll IJlI_ _._-"IU MiilUl_lU _ I ul I
representedon the PurityPlotby a line labeled
JlI En,,_,'n--_0EMh,o,,,o
Puuli_
Ilh.,.i.
ufl / r'_'_llmlt_;:_'P?PlaP'lmlln
F:laa"l-"l
"Noise and Solvent."This line rises near the peak _
tails, reflectingthe fact that baselinenoise changes
the shapeof Iow-absorbance spectra.

ThePurityPloton the left in Figure3 is of thetriple PuII y ..a

coelution.Thepurity trace risesabove the noiseand
solventtrace, indicating coelution.The plot on the
right is of a pure peak. Notice that the purity trace
remainsbelow the noiseand solventtrace at all
times,confirmingthe peak's spectralpurity.

continueson nextpage

Figure3: Thepeakin thewindowontheleft iscoelated.Thepeakontherightis

spectrallypure.

0
AnIntroduction
toMulti-Component
PeakPurity
continuedfrom ,oreviouspage

Multi-Component
PeakPurity Foreach peak, MCPPmakes_: "o"0,-,_:sses
Pass1 shown in Figure4 co_oaes :re :ce'., s_ec-
Thoughthe traces in a peak puriV plot can distin-
trum to all other peak spectra _ass i s :re same as
guish be_een o pure and o coeluted peak, these
the peak purity testand determines,.,,r-errer"me_,eak
tracescannot tell us how many componentsare containscoelution As before. Passi snov,.s"_'arthe
present After comparing spectraloverlays, then
peak containsmore thanone comoe_,,'a
examining the purity plot. we now investigate
the triDle coeluted peak with Multi-Component Ale

Peak
Pur,ty. MCPPAmysts
Theadditional passesprovide moredetailed infer-
nation: Pass2 determineswhether the peak con-
_File Edit O_O_ptions Help tains two or more spectralJydistinct compoundsby
-Purity Results computing a new purity trace and a new noise +
Rcl_ltliull Ti.._; 2.205 Mi.ultr_ P_uk;1 u[ 1 solventtrace, usingthe apex and first maximum
Peak Limit_: 2.055 to 2.588 min Baseline-to-Ba._eline impurityspectrumas references.If the new purity
Pass ,it1 Purity Angle: 1.004 Purity'Threshold: 0.108 trace risessignificantly above the noise-solvent
trace, then the peak contains morethan two cam-
Poak PurityPaseo_Table pounds. If the new purity trace remainsbelow the
noise+solventtrace, the peak contains two, or
fewer compounds.

For Pass3 the systemrepeats the process, using

the apex and the Maximum ImpuritySpectra from
Pass 1 and Pass2 as references.The set,are
automatically repeats this processfor all four pass-
es and displays the resultsfor easy comparison.
The table in Figure5 summarizesthe logic of
MCPP for each passwhile Figure6 shows the
actual plots for passes2 and 3. For Pass2, the
purity trace ties above the noise+solventtrace,
,_ _ while for pass 3, the purity trace lies below the
noise+solventtrace. Thusthe peak contains
three compounds.

'_ 2.26 :2.410 2,.'_'

Mlxg_

It Pass
Number Asksif the peak
containsmore than If the is
answer Then

Figure4: TheMulti-C0mp0nent 1 1 compound? Yes. Go to Pass2

PeakPurityPI0t(firstpass) No. Stop: 1 Compound.
showsthepuritytraceandthe 2 2 compounds? Yes. Go to Pass3.
location
of thefirstMaximum No. Stop:2 Compounds.
ImpuritySpectra,
labeledM. 3 3 compounds? Yes. Go to Pass4.
Thepeakiscoeluted. No. Stop: 3 Compounds.
4 4 compounds? Yes. Stop: More than 4.
No. Stop: 4 Compounds.

Figure5: Thistablesummarizes
the logicof
MCPPfor eachpass.
 _!
,_:&
Figure6: Passes
2 and3 forthetriple
.-_1 coelutedpeak.Eachpassidentifies
a new
Maximum Impurity
Spectra.ThePurity
i_%_!t Platfor Pass2 indicates
thatthereare
"-,,, morethantwocompounds present.The
Pass3 PurityPlotshowsthatthree
L *_ : ........
,,, _ ,_*._"'
_ "'_'_,',,._._
_ spectraaccount
for all peakspectra.

Factorsfor Success Oe-tor Waters 996

Peakpuri_and MCPP are notabsolutetechniques Parameters Optimization Specifications
for assessingchemical purity, but ratherspectral Optical & Spectral High 1.2 nm optical
purity.To be detected, a coeluted compound resolution resolution
musthave: Baseline noise at Low ± 1.5 x 10-5 AU
• Absorbance--within the detector'swavelength all wavelengths
range. Absorbance High 10 mm pathlength
• Differencein spectralshape--be_een the signal
othercoelutors. Band broadening Low 8 pL flow cell
• Chromatographic resolution-- between the
othercoelutors.

Optimumdetection of low-level coelutantsis Figure7:Thesuccess of MCPPdepends

ultimately
on
achieved when the maximumabsorbance of the thequalityof dataprovided
bythedetector.
majorcompound is near 0.5 AU, and a sensitive,
linear array-basedabsorbance detector, like the
Waters 996 PDA, is used. Thedetectorparameters
that mostaffect performanceshould be optimized as
detailed in Figure7.

Conclusions
Photodiodearray-based detectorscan determinethe
presenceof coelutorshidden undera chromato-
graphic peak. Thisis accomplished by comparing
all absorbancespectrato the apex spectrumwith a
peak purity plot.

The introductionof Multi-ComponentPeakPurity

takesspectralanalysisof PDAdata a step further,
providing experimenterswith the meansto determine
not only the presence/absenceof coelutors,but
also the numberof spectrallydistinct compounds
beneatha peak. The combinationof a PDAand the
MCPPalgorithmsrepresentsa powerfuladdition to
the chromatographer'sanalytical tool kit.