Anal. Chem. 1998, 70, 843-850
Determination of Block Length Distributions of Poly(oxypropylene) and Poly(oxyethylene) Block Copolymers by MALDI-FTICR Mass Spectrometry Gerard J. van Rooij, Marc C. Duursma, Chris G. de Koster,† Ron M. A. Heeren,* and Jaap J. Boon
FOM Institute for Atomic and Molecular Physics, Kruislaan 407, 1098 SJ Amsterdam, The Netherlands P. J. Wijnand Schuyl and Erik R. E. van der Hage‡
Unilever Research Laboratorium Vlaardingen, Olivier van Noortlaan 120, 3133 AT Vlaardingen, The Netherlands
Matrix-assisted laser desorption/ionization (MALDI) was performed on an external ion source Fourier transform ion cyclotron resonance mass spectrometer (FTICR-MS) to analyze the block length distributions of triblock polymers of poly(oxypropylene) and poly(oxyethylene). The first series of results presented demonstrate that the apparent molecular weight distributions are distorted. This distortion is induced by the flight-time-induced mass discrimination inherent in the experimental technique, the variation of isotopic patterns over the measured mass range, and the overlap of peaks in the spectrum. Subsequently, a method for the treatment of molecular weight distributions measured by MALDI on an external ion source FTICR-MS is developed to yield the actual molecular weight distribution and, from that, the individual block length distributions. For the first time, detailed and accurate molecular weight data were obtained on a complex sample using this methodology, which independently validates the data provided by the manufacturer. The experimentally verified random coupling hypothesis proves the validity of the methodology. Water-soluble triblock polymers of poly(oxypropylene) and poly(oxyethylene), the so-called Pluronics, find widespread application as nonionic surface-active agents. Poly(oxyalkylene) block copolymers, having surfactant properties, have been utilized as lubricants, dispersants, antistatic agents, foam control agents, and solubilizers and in numerous other applications in the areas of pharmaceuticals, cleaning agents, foods, and personal care products.1,2 This diversity in applications results from the various possible structural arrangements of poly(oxyalkylene) block copolymers. In these materials, ethylene oxide (EO) blocks provide hydrophilicity, and propylene oxide (PO) blocks provide * Address correspondence and reprint requests to Dr. R. M. A. Heeren, Kruislaan 407, 1098 SJ Amsterdam, The Netherlands. Fax: (+31) 20 6684106. † Present address: DSM Research, P.O. Box 18, 6160 MD Geleen, The Netherlands. ‡ Present address: Diosynt BV, P.O. Box 20, 5340 BH Oss, The Netherlands. (1) Priorr, R. In Surfactants in Consumer Products; Falbe, J., Ed.; SpringerVerlag: Heidelberg, 1987; pp 5-22. (2) Ryan, A. J.; Stanford, J. L. In Comprehensive Polymer Science; Allen, G., Bevington, J. C., Eds.; Pergamon Press: Oxford, UK, 1989; pp 427-455. S0003-2700(97)00609-4 CCC: $15.00 Published on Web 01/27/1998
© 1998 American Chemical Society
the hydrophobicity necessary for surfactancy. The chemical composition distribution can be modified to achieve the desired surfactant performance. Within this context, it is necessary to obtain a complete description of the molecular composition distribution of the individual blocks in the copolymer in order to understand the structure-function relationships of surfactants with respect to physical, rheological, and mechanical properties. It is important to have detailed knowledge about the sequence and distributions of EO and PO blocks, the degree of polymerization and block size, the identity and structure of particular end groups (initiator and terminator type), and impurities. Detailed reviews of analytical approaches for studying poly(oxyalkylene) nonionic surfactants have been compiled by Kalinoski,3 Chu and Zhou,4,5 and Schmitt.6 These methods include liquid chromatography, infrared spectroscopy, raman spectroscopy, viscosimetry, calorimetry, and NMR spectroscopy. In general, these methods yield only an average distribution of specific structural features and are not suitable for the characterization of the individual components of molecular weight distributions. In recent years, it has been demonstrated that mass spectrometric methods in combination with soft ionization techniques, which minimize fragmentation during ionization and thus produce (pseudo-) molecular ions, can provide this information.7 Within this range of techniques, especially matrix assisted laser desorption/ionization (MALDI) mass spectrometry has proven its value in providing accurate and detailed molecular weight data on high-molecularweight materials.8,9 Since its introduction by Tanaka and coworkers10 and by Karas and Hillenkamp,11 the technique has (3) Kalinoski, H. T. In Nonionic Surfactants Polyoxyalkylene Block Copolymers; Nace,V. M., Ed.; Marcel Dekker: New York, 1996; pp 31-66. (4) Chu, B. Langmuir 1995, 11, 414. (5) Chu, B.; Zhou, Z. In Nonionic Surfactants Polyoxyalkylene Block Copolymers; Nace, V. M., Ed.; Marcel Dekker: New York, 1996; pp 67-143. (6) Schmitt, T. M., Ed. Nonionic Surfactants; Surfactant Science Series 40; Marcel Dekker: New York, 1992. (7) Nuwaysir, L. M.; Wilkins, C. L.; Simonsick, W. J., Jr. J. Am. Soc. Mass Spectrom. 1990, 1, 66-71. (8) Belu, A. M.; DeSimone, J. M.; Linton, R. M.; Lange, G. W.; Friedman, R. M. J. Am. Soc. Mass Spectrom. 1996, 7, 11-24. (9) Bahr, U.; Deppe, A.; Karas, M.; Hillenkamp, F.; Giessmann, U. Anal. Chem. 1992, 64, 2866-2869. (10) Tanaka, K.; Waki, H.; Ido, Y.; Akita, S.; Yoshida, Y.; Yoshida, T. Rapid Commun. Mass Spectrom. 1988, 2, 151-153.
Analytical Chemistry, Vol. 70, No. 5, March 1, 1998 843
evolved as one of the most successful volatilization and ionization methods for a wide variety of molecules, such as peptides, proteins, oligosaccharides, and synthetic polymers. Several reviews illustrate the continuously growing scope of MALDI applications.8,12-14 In many cases, the MALDI ionization technique is coupled to relatively inexpensive time-of-flight (TOF) mass spectrometers. These systems provide a large mass range over which the samples can be analyzed, with high sensitivity, a mass accuracy of typically 0.1%, and oligomeric mass resolution.15 For example, MALDITOF has been successfully used for obtaining molecular weights of segments of synthetic copolymers and weight averages of the molecular weight distributions,16 and for obtaining the detailed description of the individual components of triblock copolymers of polystyrene-block-poly(R-methylstyrene) by Wilczek-Vera et al.17 The recent introduction of delayed extraction has extended the performance of MALDI-TOF systems to isotopic resolved measurements for masses up to ∼4000 amu with mass accuracies of ∼15 ppm.18 However, for copolymers exhibiting frequent overlap of peaks corresponding to different copolymer compositions, even higher mass resolution is required in order to differentiate between them. In these cases, the mass resolution of MALDI-TOF systems (even with delayed extraction) is insufficient, and the use of Fourier transform ion cyclotron resonance (FTICR) mass spectrometry becomes imperative. Using the standard resolution in broadband mode of this instrument, isotopically resolved, singly charged polymers in the mass range up to m/z ∼4500 with mass accuracies better than 20 ppm are measured routinely.19-23 Changing to the heterodyne mode (highresolution mode) increases performance easily to mass resolutions >1 000 000 and mass accuracies 2 were labeled and used for further analysis. The composition of these copolymer molecules is indicated for the series of poly(oxypropylene) homopolymers in the figure and in the expanded mass scale for all monoisotopic peaks. Here the first number refers to the number of EO units present in the copolymer (nEO), and the second number refers to the number of PO units present (nPO). In this way, 130 monoisotopic peaks were identified, leading to 130 data objects, which we refer to with the following notation (following the notation used in the paper of Wilczek-Vera et al.17): peak no. 1 l k
m/z (m/z)1 l (m/z)k
nEO
nPO
EO
PO
n1 l nkEO
rel intens
n1 l nkPO
I1exp l Ikexp
(in this nkEO, nkPO data format, m/z is increasing for increasing k). The distribution of both components of the copolymer is visualized in Figure 2 by plotting the measured intensities as a function of both nEO and nPO in a contour map. This plot indicates that a strong coupling exists between the molecular weight distributions of the individual parts because there is a significant shift of the center of the distribution in the PO part of the copolymer as a function of the number of EO units. However, for this type of block copolymer, one would expect that the polymerization process has followed the random coupling hypothesis (i.e., no correlation between the molecular weights of the individual parts). Therefore, the observed asymmetric distribution would indicate that the results are strongly influenced by the flight-time-induced mass discrimination inherent in the experimental technique, as mentioned in the introduction. 846
Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
Figure 2. Distribution of monomer units as a function of the individual monomer units in the polymer. The data plotted are extracted from the spectrum in Figure 1.
Due to the distance between the ion source and the ICR cell and the pulsed nature of both the ionization and the analysis event, the measured molecular weight distribution is a convolution of the real molecular weight distribution of the polymer sample and the flight time distribution arising from the ion transport. The trapping delay between the ionization and analysis events determines which segment of the flight time distribution is analyzed. In the recent literature, two different methods of compensation for flight-time-induced (either gated deceleration times or trapping times) mass-selective trapping have been proposed. The first one is an integral method, based on the averaging of time domain transients,28 and the second one is a projection method, based on the superimposition of spectra.32 Here, we will follow the methodology described in ref 32, where it was demonstrated that this method leads to the most accurate descriptions of the actual molecular weight distribution. According to that methodology, several mass spectra must be acquired at several different trapping delays, which are then projected (not summed) on the same intensity axis to obtain the correct distribution. In an experimentally ideal situation, where the decrease of the trapping delay between two successive experiments is very small, this is equivalent to optimizing the trapping delay for each peak in the spectrum. Of course, this method is allowed only if the total ion production in the MALDI process remains constant over the successive experiments. Therefore, multiple laser shots (in the measurements presented, typically 250 laser shots) have been averaged to minimize the shot-to-shot variation in the MALDI process. In addition, the “spot-to-spot” variation was eliminated by recording all the spectra from the same spot. A possible decrease in the ion yield was monitored by taking reference spectra (with the trapping delay at 900 µs) in the course of the series of measurements. Following these procedures, a series of 10 experiments has been performed, in which the trapping time was varied from 600 to 1000 µs and from 1050 to 1450 µs in steps of 100 µs. Plotting the resulting mass spectra on a common intensity scale yields the molecular weight distribution shown in Figure 3. It can be seen that the molecular weight distribution in Figure 3 has become significantly broader in comparison to that resulting from the single measurement in Figure 1. Especially the lower mass range was affected by the flight time effect. This
Figure 3. Overall MALDI-FTICR-MS spectrum of poly(oxyethylene)- poly(oxypropylene). The molecular weight distribution was constructed by projecting all the measured spectra in Figure 3 on the same intensity axis.
explains the shift of the center of the distributions in the PO part toward higher numbers for lower numbers of EO units in Figure 2. After all, discrimination of lower masses means discrimination of the lower end of the distributions in the PO part, which becomes less apparent for larger EO lengths. Before processing the spectrum in Figure 3 as described above to reveal the individual block length distributions, two additional phenomena, affecting the measured molecular weight distribution, have to be taken into account. First, the contribution of the naturally occurring isotopes in the molecular ion changes over the mass range of interest. For example, 62% of the sodiumcationized homopolymer with 13 PO units at m/z 796 will be monoisotopic, whereas, for the homopolymer with 25 PO units at m/z 1492, this will be the case for only 40% of the molecular ions. Taking the monoisotopic peaks as the relative abundance of the molecular ions will evidently induce significant errors and, therefore, will distort the molecular weight distribution. The second effect can be illustrated with the peak at m/z 971.6 in the expanded mass scale in Figure 1, where it is seen that no distinction can be made between the second isotopic peak of the homopolymer with nEO ) 0 and nPO ) 16 (monoisotopic peak at m/z 969.6) and the monoisotopic peak of the copolymer with with nEO ) 4 and nPO ) 13 at m/z 971.6. Because all the components of the molecular weight distribution of the Pluronic sample differ by an even number of mass units, and because contributions of isotopic peaks of polymers with a monoisotopic molecular weight less than or equal to (m/z)k - 4 may be neglected as these can be estimated to be smaller than 1% in the molecular weight range considered here, the intensities of the peaks in the spectrum at (m/z)k which satisfy
Figure 4. High-resolution (heterodyne) MALDI-FTICR-MS spectrum of poly(ethylene oxide)-poly(oxypropylene) around m/z 970. The expansion of the mass scale shows that the second isotopic peak of the (0,16) copolymer and the monoisotopic peak of the (4,13) copolymer are separated, which cannot be accomplished in broadband experiments.
(m/z)k - (m/z)k-1 = 2
(2)
have to be corrected for the contribution from the molecular ions with the monoisotopic peak at (m/z)k-1. To illustrate the importance of this correction, let us consider the two homopolymers used in the previous example. In the case of the 13-mer, 8% of the molecular ions will contain two 13C isotopes (or 13C + 17O or 217O or 18O); for the 25-mer this will be 17%. On the basis of the the intensities in the mass spectrum in Figure 3, it was determined that the contribution of the 213C isotope of the 13-mer to the peak corresponding to the monoisotopic (nEO,nPO) ) (4,10) was ∼50%; for the 213C isotope of the 25-mer and for the (nEO,nPO) ) (4,22), this was determined to be ∼10%. This second effect is further illustrated by making use of the high mass resolution of the instrument in the heterodyne mode. The mass range m/z 9301015 has been recorded in this high-resolution mode, resulting in the spectrum shown in Figure 4. The inset reveals that the resolution (m/dm)50% ) 135 000 (at m/z 669.7) is sufficient to separate the two peaks around m/z 971.7. In principle, it would be possible to record the complete mass spectrum at high resolution, which would overcome the problem of peak overlap. However, this would take approximately 20 experiments for each different trapping delay (i.e., a total of 200 experiments) to cover the entire molecular weight distribution, which can practically not be correlated due to inevitable variations in ion yields. Therefore, we propose to apply a correction method instead. To check this method, we compare the data from the high-resolution spectrum, in Figure 4 with those from the broadband spectrum in Figure 3. From the high-resolution spectrum, we determine from the peaks at m/z 969 and 971.65 that the ratio between the abundance of the monoisotopic (nEO,nPO) ) (4,13) copolymer and the abundance Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
847
of the monoisotopic (nEO,nPO) ) (0,16) copolymer would be 0.19. Determination of this ratio from the spectrum in Figure 3 yielded the value 0.36. After correcting the intensities for the expected overlap (according to the method which is proposed below), this value became 0.18, which is in excellent agreement with the result from the high-resolution measurement, and thus confirms the validity of the correction method. This example also illustrates that the precision of the correction process for the individual intensities is on the order of 5% (of course, for very low S/N ratios, this value might increase significantly). By analytical expression of the correction for the change of the isotopic patterns over the mass range and the overlap of peaks, it was possible to automate the calculation of the corrections of the measured relative intensities. The first step in the derivation of this expression is the calculation of the probability that the poly(ethylene oxide-propylene oxide) molecule is monoisotopic. Standard probability theory35 yields, for a copolymer consisting of nkEO EO units and nkPO PO units, i.e., [HO-(C2H4O)nkEO(C3H6O)nkPO-H], the following probability that its mass equals (m/z)k (i.e., it is monoisotopic): EO+3n PO k
P(nkEO,nkPO,(m/z)k) ) A2nk
EO+n PO+1 k
(12C)Ank
(16O) (3)
where Ab(mX) denotes the bth power of the natural abundance of the isotope mX. If there is no polymer present in the spectrum with a monoisotopic molecular weight (m/z)k -2, then the corrected abundance Ikcorr for this component of the molecular weight distribution is simply given by
Ikcorr )
Ikexp
(4)
P(nkEO,nkPO,(m/z)k)
the spectrum. Subsequently, it calculates the corrected intensities for all the components of the molecular weight distribution using eq 1. Finally, it calculates the number and weight averages of the number of monomers, the polydispersity for both monomer units, and the number and weight averages of the molecular weight distribution of the copolymer sample. The effect of the corrections can be quantified by calculating the number- and the weight-average molecular weights (Mn and Mw respectively) and the polydispersity factor (Mw/Mn) for the copolymers with nEO ) 4 for the uncorrected data, for the data corrected only for the isotope effect, and for the corrected data. The resulting values are listed in Table 1. It should be mentioned here that all peak intensities were determined by integration over the peaks, because otherwise the correction for the peak overlap would no longer hold. For example, if the overlapping peaks do not entirely coincide, then peak broadening will occur, and the use of peak heights will underestimate the actual intensities. It can be seen from the table that, for the sample under investigation, the change induced by each correction is about the same. The total percentile change in the values listed in the table is 6-8%. Of course, these differences in molecular weight do not lead to any measurable differences in physical properties. However, the corrections appear to be valuable in the evaluation of the second step of the synthesis of the block polymer (where the propylene block is sandwiched between ethylene blocks). In order to characterize the entire sample under investigation, we calculated the usual molecular weight averages Mn (number average), Mw (weight average), and the polydispersity factor Mw/ Mn. In addition, some averages describing the chemical composition of the block copolymer in terms of the individual segments are introduced:
(1) the average number of units for segment X where Ikexp refers to the measured intensity of the peak at (m/z)k. If there is a peak at (m/z)k - 2, the probability has to be calculated that also the mass of the copolymer described by nk-1EO and nk-PO equals (m/z)k. This probability is equal to the sum of the individual probabilities that this molecule contains two 13C isotopes, one 13C and one 17O isotope, or one 18O isotope (the chance of having two 17O isotopes present is negligibly small). The calculation of this probability P(nk-EO,nk-1PO,(m/z)k) is analogous to the one in eq 3. Using these probabilities, the corrected abundance becomes corr
Ik
)
1 EO
PO
P(nk ,nk ,(m/z)k)
(
Ik
exp
)
To aid in the interpretation of the copolymer spectra and the data processing of the large data sets, a small computer program was developed. This program needs as input the elemental composition of the two different monomeric units of the copolymer in order to determine the composition of the copolymer for each peak in (35) Kreyszig, E. Advanced Engineering Mathematics; John Wiley & Sons, Inc.: New York, 1993; Chapter 23.
Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
∑n
x k
Ikexp
k
n j nx )
∑
(6) Ikexp
k
(2) the weight-average number of units for segment X
∑(n n j wx )
x 2 k
) Ikexp
k
∑
(7) nkx Ikexp
k
-
Ik-1exp P(nk-1EO,nk-1PO,(m/z)k) (5) P(nk-1EO,nk-1PO,(m/z)k-1)
848
(X ) EO, PO) in the block copolymer
(3) the polydispersity factor for segment X, nwx/nnx, (4) the chemical composition of segment X jxx )
Mxn j nx M h nx
(8)
The values of these quantities were calculated from the data extracted from the spectrum in Figure 3 before as well as after correction and are listed in Table 2. The characteristic numbers according to the manufacturer are nj nPO ) 16.4 ()950/58.04) and jxEO ) 0.2. When these numbers are compared with the values presented in Table 2 (nj nPO ) 15.57
Table 1. Mn, Mw, and n Determined from the Individual Molecular Weight Distribution of the Copolymers with Four Ethylene Oxide Units
uncorrected corrected for isotope effect fully corrected
Mn
Mw
n ) Mw /Mn
1047 1083 1125
1104 1136 1177
1.055 1.049 1.047
Table 2. Properties of the Pluronic L31 Sample Calculated from the Spectrum in Figure 3 uncorrected EO block n j nx n j wx n j wx/n j nx jxx M hn M hw M h w/M hn
uncorrected
PO block
4.53 8.51 1.88 0.19 1098.4 1175.9 1.07
EO block
14.78 15.83 1.07 0.81
PO block
4.11 8.14 1.98 0.17 1125.7 1200.6 1.07
15.57 16.55 1.06 0.83
and jxEO ) 0.17 after correction), it can be seen that our results derived from MALDI-FTICR-MS confirm the manufacturer’s specifications with greater accuracy. The individual block length distributions are again visualized in contour maps. Figure 5a shows the data after flight time compensation, and Figure 5b shows the fully corrected data. Comparing the two contour maps, one sees that, after the full correction, the discontinuity in the EO distribution around nEO ) 4 (a local maximum) has disappeared. This discontinuity arose from the peak overlap of the second isotopic peaks with nEO ) 0 and the monoisotopic peaks with nEO ) 4. The contour map in Figure 5b clearly indicates that the random coupling hypothesis holds for this type of copolymers, as the distribution in the PO units remains the same for different EO segments. Moreover, the fact that the experiment verifies the random coupling hypothesis also indicates that the correction of MALDI-FTICRMS spectra following the methods described above leads to the actual molecular weight distribution. This can be further verified by calculating the experimental marginal probabilities for the two components of the copolymer.17,35 For the PO units, the marginal probability ΓexpPO is defined by
ΓexpPO(njPO) )
∑Γ
EO
exp(ni
,njPO)
Figure 5. Contour maps of the overall distribution of monomer units as a function of the individual monomer units in the polymer, both before (A) and after correction (B) for the isotope effect and peak overlap. The data plotted are determined from the reconstructed molecular weight distribution in Figure 4.
(9)
i
where Γexp(niEO,njPO) is the experimental distribution function, which is equal to the measured intensity for a given niEO and njPO. According to this formula, all the measured intensities for a given njPO are summed. Similarly, the marginal probability for the EO units ΓexpEO is defined. If the random coupling hypothesis holds, then the experimental distribution function can be represented as the product of the two marginal distributions:35
Γexp(niEO,njPO) ) ΓexpEO(niEO)ΓexpPO(njPO)
(10)
In order to test eq 10, we assumed the distributions in the PO units for each number of EO units to follow the Gaussian function.
Figure 6. Test of the random coupling hypothesis for the Pluronic L31 copolymer. The measured distribution in the propylene oxide units is fitted with a Gaussian function for different numbers of ethylene oxide units. The center and width of the fitted Gaussian functions are plotted as a function of the number of ethylene oxide units.
If we further consider the heights of the (fitted) Gaussian functions to be representative for the distribution in the EO units, we can rewrite eq 10 as
(
ΓexpEO(niEO)ΓexpPO(njPO) ) A(niEO) exp
)
njPO - ncPO w
Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
(11) 849
If eq 11 holds, it follows that the fitted PO distributions have equal centers ncPO and equal widths w. This is evaluated in Figure 6, where the results of the individual fits are plotted as a function of nEO. It can be seen that the width w shows a slight dependence to on nEO. This is most probably a dependence on the S/N ratio instead of a dependence on nEO. At lower S/N ratios (
