Validation of SEC/GPC Software M. Gray, P. KiIz*, G. Reinhold* PSS-USA, 13531Cedar Creek Lane, Silver Spring, MD 20904 *PSS Polymer Standards Service GmbH, POB 3368, D-55023 Mainz Poster presented at the International GPC Symposium, San Diego, 1996
Introduction In many laboratories certification (ISO 9000) and accreditation (EN 45000) plays a dominant role. GLP and GMP regulations also require validated analytical procedures. The ISO/EN/DIN SEC/GPC standard confirmity is important, but cannot replace complete and systematic software validation. Due to the complextity of todays software systems, it is very difficult for laboratory managers to proof to the authorities that their data acquisition systems handle analytical data precisely and adaquately.
Method The following software validation methods are used at PSS to proof and document the proper functionality of the PSS WINGPC software! traditional software validation procedures! world-new PSS AUTOVAL validation process
Method Pro Con traditional each step analyzed time consuming directed debugging data flow dependent not tranferable to user no installation validation independend validation for each test step PSS verifies on-site installation degugging difficult AUTOVAL repeated on-site user tests complete integral analysis tests data flow paths tests software interactions indicates computer problems incorporation into system suitibility user training option easy testing of modified routines
The PSS AUTOVAL Validation Method 1. Step Creation of raw data (c(v), P(V), I (V) and M(V)) h from theoretical molar mass distributions (with well-known molecular parameters) 2. Step Import of validation raw data into the PSS WINGPC Software. Test of all software features on-the-fly 3. Step Comparison of the theoretical molar mass distribution with the software-calculated results (numerical values and distribution curves). The results must be identical (taking numerical errors into account) for the test to be successful.
Results Validation of WINGPC Calculations with Direct Calibration Run Conditions and Data Processing Parameters Polymer Type Eluent Temperature Calibration K [ml/g] a Polystyrene THF 25 C Polystyrene 0.01363 0.714 W(log M) 1.2 1.1 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 10 3 10 4 10 5 10 6 10 7 g/mol Mol mass Fig. 1: Molar mass distribution using conventional calibration as calculated by PSS WINGPC (the theoretical dsitribution is not shown, since no differences are visible) PSS WinGPC scientific V 3.0 b-36 M [D] n theoret. calculated error [%] 150000 150300 0.200 M w [D] 300000 300070 0.023 M z [D] 450000 450010 0.002 D 2 1.997 0.150 M v [D] 280869 280230 0.228 [η] [ml/g] 105.68 105.69 0.009 c [g/l] 4.343 4.343 0.000 Very good agreement of all parameters between theory and software testing Minor deviations caused by limited number of data-points (52).
Testing GPC Viscometry Setups (Example: Diff. Viscometer) Run Conditions and Data Processing Parameters Polymer Type Eluent Temperatur Univ. Calibration K [ml/g] a e Polystyrene THF 25 C PMMA 0.0129 0.688 η sp [η] R g g 0.0225 ml/g nm Viscosity 0.0200 0.0175 0.0150 0.0125 0.0100 0.0075 0.0050 0.0025 0.0000 10 2 10 1 10 2 10 1 1.06 10 0 1.14 1.12 1.10 1.08 1.04 1.02 10 4 10 5 10 6 g/mol molecular mass UE Fig. 2: Molar mass dependance of spezific and intrinsic viscosity, radius-of-gyration (R g) and branching (g ) in GPC-Viscometry calculations in PSS WINGPC PSS WinGPC scientific V 3 0 b-36 theoret. calculated error [%] M n [D] 150000 150110 0.073 M w [D] 300000 298230 0.590 M z [D] 450000 447220 0.618 D 2 1.987 0.650 M v [D] 278997 278510 0.175 [η] [ml/g] 105.68 105.69 0.001 K [ml/g] 0.01298 0.0137 0.514 a 0.688 0.714 0.000 c [g/l] 4.343 4.343 0.000
Testing GPC Light-Scattering Setups (example: LALLS) Run Conditions and Data Processing Parameters Polymer Type Eluent Temperature λ [nm] h [grd] n dn/dc [ml/g] 0 Polystyrene THF 25 C 633 6 1.403 0.187 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 10 6 10 5 10 4 10 5 10 6 g/mol molecular mass BE Fig. 3: Light scattering intensity and molar mass by LALLS detection, and comparison of M w(ls) vs. fitted (interpolated) molar mass in the LS-module of PSS WINGPC; straight line indicates perfect fit of measured M w(ls) PSS WinGPC scientific V 3.0 b-36 theoret. calculated error [%] M n [D] 150000 150700 0.467 M w [D] 300000 300240 0.080 M z [D] 450000 450220 0.049 D 2 1.992 0.400 M v [D] 278997 278560 0.157 c [g/l] 4.343 4.343 0.000 Very good agreement of all parameters between theory and software testing Minor deviations caused by limited number of data-points (52).
Testing Combined GPC-Viscosity-LS Configurations (example shown: diff. viscometer with RALLS) Run Conditions and Data Processing Parameters Polymer Type Eluent Temp. Calibration λ [nm] h [grd] n dn/dc K [ml/g] a 0 Polystyrene THF 25 C PMMA 633 6 1.403 0.187 0.01298 0.688 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 10 6 10 5 10 4 10 5 10 6 g/mo molecular mass UE Fig. 4: Molar mass distribution using GPC with viscometry and RALLS detection; measured molar masses with dyssymetry correction (linear) and without ( ) using the triple detector calculations in PSS WINGPC Software PSS WinGPC scientific V 3.0 b-36 theoret. Visco-LS error [%] M n [D] 150000 150680 0.453 M w [D] 300000 299760 0.080 M z [D] 450000 447410 0.576 D 2 1.989 0.550 M v [D] PMMA 278997 278510 0.175 [η] [ml/g] 105.68 105.76 0.076 K [ml/g] 0.01363 0.0137 0.514 a 0.714 0.714 0.000 c [g/l] 4.343 4.343 0.000 Very good agreement of all parameters Minor deviations by limited number of data (52) Expected M LS deviation at about 100000 D Accurate M with viscometer correction
Influence of Detector Noise on GPC Calculations 5% statistical noise on all raw data in worst case scenario results obtained without smoothing and despiking in software LALLS RALLS IV RI 0.250 225 0.0200 0.009 0.225 0.008 200 0.0175 0.200 0.0150 0.007 175 0.175 0.006 0.0125 150 0.005 0.150 0.0100 125 0.004 0.125 0.0075 100 0.003 0.100 0.0050 75 0.002 0.0025 0.075 50 0.001 0.0000 0.050 25 0.000-0.0025 0.025-0.001 0-0.0050 Voltage [V] 10 12 14 16 18 20 Elution volume [ml] LALLS RALLS IV RI PSS WinGPC scientific V 3.0 b-36 Fig. 5: LALLS, RALLS, Viscometry and RI raw data shown with baselines. All signals have a 5% random noise (S/N: 20:1) to test the robustness of PSS WINGPC algorithms; baselines are set as shown at 9.4 ml and 20.3 ml; integration limits can be set separately (see further explanation in text)
theoretical conventional universal LALLS LS-Visko M n [D] 150000 131630 150200 169120 151040 M w [D] 300000 303310 306260 299670 294420 M z [D] 450000 477670 478210 427110 415290 D 2 2.304 2.039 1.772 1.949 M v [D] (PS) 280869 282240 285670 284030 276170 [η] [ml/g] 105.68 106.79 106.86 105.90 104.59 K [ml/g] 0.01363 n/a 0.01457 n/a 0.01295 a 0.714 n/a 0.708 n/a 0.719 c [g/l] 4.343 4.327 4.276 4.284 4.296 conventional calibration: underestimated M n, overestimated M z (about 10%) universal calibration: larger molar mass deviations good bulk-iv agreement Mark-Houwink coefficient surprisingly close to reality light scattering: big discrepancies: M n clearly overestimated polydispersity clearly underestimated improvement of the results: integration-limits set to significant data (S/N>3). This corresponds to a reduction of the calculated injected mass of about 1%, but improves precision dramatically
Summary All PSS WINGPC software packages and modules are validated and conform to the GPC/SEC standards ISO/CD 13885 and DIN 55672. With the efficient PSS AUTOVAL test procedure, the GPC-user has! more security in the selection of GPC software packages! better control of GPC calculation methods! time-savings for otherwise costly software validation! the security of correct data reduction.! the option of testing future or in-house calculation procedures! ideal for training users for optimzed baseline settings, smoothing etc The PSS AUTOVAL validation:! is independent on any preconditions! applicable for all software-algorithms! tests the complete flow of data! can be used the in-house as required