Anal. Chem. 1988, 60,706-713
706
Factors Affecting the Sampling of Poly(ethy1enimines) by Electrohydrodynamic Mass Spectrometry Kelsey D. Cook,* John H. Callahan,’ and Victor F. ManZ
Department of Chemistry, University of Tennessee, Knoxville, Tennessee 37996-1600 The effects of polymer structure and solution chemistry on sampling by eiectrohydrodynamlc (EH) mass spectrometry were studied by uslng poly(ethyienimines) (PEIs). I n contrast with results for pdy(ethylene glycols), number average molecular weights (Mn)calculated from desorption ionlzation mass spectra of PEIs are generally too low, with the discrepancy becoming greater with increasing molecular welght. The effects of adding acid and transition-metai ions indlcate that sampHng Mas in EH m a s spectrometry Is In part due to strong solvent-Ion interactions resulting from poor shielding of charge sites in multiply charged adducts, wtth a ConttRxnkn from hydrogen bondlng interactions between polymer and sdvent. Increases in apparent Mnwhen =Is were sampled from mixed giyceroi/water matrices resuit from changes in hydrogen bonding, Ion pairing, and/or mass transport.
The physical and spectroscopic similarity of the various oligomers comprising a typical synthetic polymer sample makes the determination of polymer molecular weight distributions (MWDs) a singularly difficult problem. Because of the key role of the MWD in determining many physical properties (e.g., traces of low MW material can severely embrittle an otherwise strong sample), MWD determination is also a problem of significant importance. Gel permeation chromatography (GPC) has had the greatest success in these applications but is generally a rather low-resolution technique with a fairly narrow dynamic range (1). The mass resolution capabilities of mass spectrometry (MS) are much greater, making it an attractive alternative to GPC. However, successful polymer MS applications (2-1 7) (ref 2 and 3 give recent reviews) have emerged only with the development of “soft” ionization techniques, capable of generating the gas-phase ions essential for MS analysis, without excessive molecular degradation. As new techniques have been reported, virtually all have been tested for applicability to polymer samples. Poly(ethy1ene glycol) (PEG; HO(CH2CH20),H) has been used widely in these tests. In fact, a set of four PEG samples characterized by electrohydrodynamic (EH) MS and endgroup titrations (an ASTM standard method for determinain 1980 (14) has since tion of average molecular weight, been characterized with good accuracy by at least five different desorption ionization techniques in four different laboratories (6, 10, 12,16,17). PEG is a polar but nonionic polymer; each of these MS experiments relied on ion attachment to the polymer (Le., complexation) as a gentle means of ionization. However, despite an expected correlation between polymer chain length and charge capacity, multiply charged adducts were reported only in the EH MS studies, where reduction of mass-to-charge ratios by incorporation of up to four metal ions allowed characterization of samples exceeding the nominal mass range of the instrument employed ( 1 4 , 15). The ability of EH MS to detect such highly charged ions is a manifestation of the exceptional softness of EH sampling (18-21). In the absence of the energetic ion, atom, or photon
an)
’Current address: Naval Research Laboratory, Washington, DC
20375-5000.
‘Current address: Department of nesota, Minneapolis, MN 55455.
Chemistry, University of Min-
probe beams used in other desorption ionization techniques, E H sampling relies exclusively on the action of an applied electric field (ca lo8 V/m) to extract ions from the meniscus of a solution contained in a metal capillary (20). This imparts little or no internal energy to the “field evaporated” ions, resulting in essentially no fragmentation (21)and, evidently, little perturbation of solution equilibria. Thus, EH MS has also been used successfully as a probe of competitive complexation of PEG (22)and a cyclic analogue, 18-crown-6 (23). Despite these attributes, structure-, charge-, and mass-dependent variations in sampling efficiencies have been noted in EH MS studies of other systems (24, 25). We therefore suspected that the absence of molecular-weight-dependent sampling bias that allowed accurate quantitation of PEG MWDs may not persist with other polymers. To test this possibility, studies were undertaken with the nitrogen analogue of PEG, poly(ethy1enimine) (PEI; H2N(CH2CH2NH),H)(26). Preliminary studies found significant evidence of sampling values calculated bias against heavier oligomers of PEI; M,, from EH mass spectra were considerably lower than the manufacturer’s specifications. Similar bias has since been observed in other laboratories using field desorption (FD) ( l l ) , laser desorption (LD) (17),and fast atom bombardment/liquid secondary ion (FAB) (11) MS. The reason for the bias must lie in the structural differences between PEI and PEG. The amine nitrogens in PEI can induce significant hydrogen bonding with solvent and between polymer molecules. The nitrogen atom also represents a potential branching point (27) and provides an opportunity for multiple proton attachment (28) and for complexation with transition metals (29). This study considers in detail how these features of molecular architecture can give rise to sampling bias in EH MS. Because that bias appears to be pervasive in desorption ionization studies of PEI, the conclusions reached should be of interest to anyone seeking quantitative information from these techniques. EXPERIMENTAL SECTION Mass spectra were obtained with a double focusing mass spectrometer (AEI MS-902, equipped with a VG Update electronics console). Design and operation of the EH source have been described elsewhere (18, 19). Spectra were obtained at emitter potentials (VaE)of 8.0-8.2 kV. Vamwas usually optimized to match the electrostatic analyzer (ESA) voltage (avoiding detection of ”fast metastable” ions (20) experiencing in-flight desolvation prior to the ESA) by maximizing the signal intensity of the ion at m / z 207 ([Na + G z ] + where , G represents glycerol). Alternatively, fast metastable ions could be detected selectively by lowering the ESA voltage to pass ions with less than full acceleration energy. The extractor voltage was maintained at about -1.5 kV, and the collector was at ground potential. Spectra were obtained under low-resolution conditions to obtain maximum sensitivity; both source and collector slits were fully open giving resolution of approximately 300-600. MS results reported here represent the average of at least three spectra. Prior to averaging, spectra were normalized (base peak = 100%) to reduce deviations caused by fluctuations in emission intensity. To facilitate comparison of spectra, peak intensities are reported relative to a common standard ([K + Gz]+at m / z 223 when present, or [Na + G4]+ at m / z 391). However, these and other peaks attributable to matrix ions have been deleted from the figures for clarity. For studies of time-dependent be-
0003-2700/88/0360-0706$01.50/0 C 1988 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 60, NO. 7, APRIL 1, 1988 707 havior, five scans were obtained as soon as possible following initiation of emission (typically 5 min passed before emission stabilized and spectra could be obtained). Subsequently, continuous emission could be maintained with glycerol solutions for a period of about 2 h, during which time seta of five spectra were obtained and averaged periodically (requiring roughly 5 min per set). The entire process was then repeated by readvancing the solution to the emitter tip to renew the surface. Sample depletion rates were higher with glycerol/water mixtures, requiring readvancement of the solution as often as every 15-20 min to maintain emission; time dependence of these signals was therefore not characterized. Glycerol spectra were obtained with a 200 wm diameter platinum capillary (Hamilton). Glycerol/water spectra were obtained with 12- or 25-wm tungsten carbide capillary bonding tools (Gaiser or Small Precision Tools). Ion abundances in this study have been estimated from the height of the most intense spectral peak in an isotope cluster, without correction for contributions from the abundance of natural isotopes (mostly 13C).While these contributions can be significant, they are difficult to calculate due to peak overlap at low resolution. Previous studies indicate that their neglect results in underestimation of M,,by only a few percent in the mass range studied here (14). Glycerol solutions were degassed for at least 8 h at low heat (50 "C) under vacuum Torr). Glycerol/water solutions were degassed by at lea@ three successive freeze-pump-thaw cycles. Uncertainties in M , values determined from mass spectra were estimated from the standard deviations of the mean of replicate determinations of normalized oligomer intensities, using standard propagation of errors calculations (30). IR spectra were obtained from chloroform solutions in NaCl cells using a Nicolet 7000 series Fourier transform spectrometer. Electron spin resonance (ESR) spectra were obtained at room temperature from 1mol % solutions (solvent, 100 mol % ) on a Varian E-9 X-band spectrometer operating at 9.45 GHz. Relative viscosities were measured in triplicate by using the falling ball method (31). The 2.4 mm diameter sapphire balls were dropped through 7.0 mL of solution in a 9.9 mm diameter graduated cylinder. Solutions were not degassed and measurements were obtained at room temperature (21 "C). Reagents. The PEI samples were obtained from Polysciences, Inc. Nominal molecular weights were 600 (lot 20221), 1200 (lot 4-1443, except where noted), and 1800 (no lot number specified by manufacturer). Concentrations were calculated based on these nominal MWs. Ethylenediamine (reagent grade) was obtained from Fisher. Diethylenetriamine (DETA, >97%), triethylenetetramine (TETA, >97%), tetraethylenepentamine (TEPA, 85%, practical grade), and pentaethylenehexamine (PEHA, 90%, practical grade) were obtained from Fluka. EH mass spectra of TEPA and PEHA showed that the major impurities were other oligomers. These samples were therefore purified by fractional distillationat reduced pressures. For TEPA, a major fraction was collected at 216220 "C (at 40 Torr) and then redistilled with collection of a fraction at 218 "C. The purity as estimated from the EH mass spectrum improved from 79.0% to 91.6%. PEHA was also redistilled twice at reduced pressure (42 Torr), with the major fraction being collected at 258-260 "C. The purity improved from 73.5% to 84.4%. The purified samples were used in subsequent studies. Glycerol was obtained from Fisher (reagent grade), Sigma (Sigma Grade), or Aldrich (Gold Label). HNO, was obtained from Baker (reagent grade). Except as noted above, all salts were reagent grade (Fisher or Mallinkrodt) and used as received. Supporting electrolyte (KNO,, NaNO,) was added to mass spectral samples to maintain the total ionic strength between 5.0 and 6.0 mol % (glycerol, 100 mol % ) unless otherwise noted.
RESULTS AND DISCUSSION EH MS of PEIs. E H MS spectra of P E I 600,1200, and 1800 were initially obtained from glycerol solutions containing NaN03 supporting electrolyte (Figures 1-3). Major analyte peaks belong to two overlapping series of masses [(nX 43) + 181' (A) and [(n X 43) 401' (B). These can be attributed to attachment of a proton (A) or a sodium ion (B) to oligomers of mass (n X 43) + 17, where n is the degree of polymerization.
+
Figure 1. EH mass spectrum of 1.1 mol % PEI 600 with 5.7 mol % NaNO, in glycerol. Intensities are relative to [Na G,]' at m / z 391 (100%). The first few ions in each series are labeled: [PEI H I + (A); [PEI + Na]' (B).
+
+
Protonation was confirmed by replacement of part of the supporting electrolyte with NaOH, which resulted in a decrease in the relative intensity of (protonated) series A. Replacement of NaNO, with KNOBresulted in the disappearance of series B, confirming sodium ion attachment. No K+-PEI adducts were observed, however, suggesting that interactions between K+ and P E I are weak or that adducts are sampled with low efficiency. These series are consistent with an end-group mass of 17, as expected for termination of the polymerization with NH3 or ethylenediamine. In addition to the two major series of ions, two low-intensity series corresponding to [(nX 43) + H]+ and [(n X 43) + Na]+ were also detected and confirmed by NaOH and KNO, addition. These would result from the absence of an end group on the polymer. Their most likely source is the intramolecular termination of the polymerization reaction (cyclization), as has been reported by Dick and Ham (32). They may also arise from degradation of the polymer in solution (see below). Gas-phase fragmentation or field-induced reactions cannot be ruled out, but there is little evidence that these occur in E H MS (21). Solvated adducts of low molecular weight oligomers were observed, particularly in the spectrum of PEI 600. Peaks at mlz 196 and 288 correspond to the attachment of one and two glycerol molecules, respectively, to the protonated dimer ( m / z 104), while peaks at 218 and 310 correspond to glycerol solvation of the sodiated dimer ( m / z 126). Attachment of a single glycerol molecule each to the protonated and sodiated trimer is also evident (appearing at m / z 239 and 261). While these results represent somewhat more extensive solvation than that observed with (less polar) PEGS (14), the systems are similar in that solvation of larger oligomers is not observed. For PEG, this probably results because long linear polymer chains are able to spiral around metal ions, shielding their charge while exposing primarily nonpolar groups to the solvent. As outlined below, the explanation for PEI may be slightly more complicated, due to the effects of branching and hydrogen bonding. EH mass spectra of PEI 1200 and 1800 (Figures 2 and 3) contain series of doubly charged ions with peaks separated by 21.5 mass units. Three structures are possible: [PEI + 2NaI2+,[PEI + 2HI2+,or [PEI + Na + HI2+. Ions of the first two series (offset by one repeat unit) differ by only 1 dalton (0.5 m / z unit): [PEI 2Na] = 43n + 20; [PEI + 2H] = 43(n + 1)+ 19. There is also potential overlap with solvated singly charged adducts (e.g., Figure 4a), although the low abundance of solvated ions in the spectrum of PEI 600 suggests that this is not a significant interference a t higher mass. Where resolution is sufficient ( m / z