Analytical and Bioanalytical Chemistry

, Volume 395, Issue 2, pp 377–386 | Cite as

Absolute integrated intensities of vapor-phase hydrogen peroxide (H2O2) in the mid-infrared at atmospheric pressure

  • Timothy J. Johnson
  • Robert L. Sams
  • Sarah D. Burton
  • Thomas A. Blake
Original Paper


We report quantitative infrared spectra of vapor-phase hydrogen peroxide (H2O2) with all spectra pressure-broadened to atmospheric pressure. The data were generated by injecting a concentrated solution (83%) of H2O2 into a gently heated disseminator and diluting it with pure N2 carrier gas. The water vapor lines were quantitatively subtracted from the resulting spectra to yield the spectrum of pure H2O2. The results for the ν6 band strength (including hot bands) compare favorably with the results of Klee et al. (J Mol. Spectrosc. 195:154, 1999) as well as with the HITRAN values. The present results are 433 and 467 cm-2 atm−1 (±8 and ±3% as measured at 298 and 323 K, respectively, and reduced to 296 K) for the band strength, matching well the value reported by Klee et al. (S = 467 cm−2 atm−1 at 296 K) for the integrated band. The ν1 + ν5 near-infrared band between 6,900 and 7,200 cm−1 has an integrated intensity S = 26.3 cm−2 atm−1, larger than previously reported values. Other infrared and near-infrared bands and their potential for atmospheric monitoring are discussed.


Infrared Fourier transform infrared Quantitative Band strengths Hydrogen peroxide 


The Pacific Northwest National Laboratory (PNNL) continues to build a database of quantitative vapor-phase infrared (IR) data for ambient monitoring: The data are from gases pressure-broadened to 760 Torr, recorded at moderately high resolution (0.1 cm−1), and with a broad spectral bandwidth so as to be optimized for tropospheric monitoring, either passive or active [1, 2, 3]. The signal-to-noise ratio of the spectrometer system is optimized for the long-wave IR, 800–1,300 cm−1, but each spectrum has a spectral range of 600 cm−1 or below to 6,500 cm−1 or above. The data are recorded as a series of single measurements, with each burden having the analyte pressurized with pure N2 to 760 Torr, and then averaged to form a composite [1]. The spectra are recorded at 0.1 cm−1 resolution so as to resolve all possible vibrational band spectral features, e.g., the Q branches of polyatomic molecules that may have Lorentzian halfwidths of 1 cm−1 or less.

During the construction of the IR database, it was recognized that H2O2 has recently garnered great interest amongst the spectroscopic community for two reasons. First, it is well known that H2O2 and other organic peroxides are powerful oxidizers. They have long been used, for example, as dilute solutions in medicine and dentistry, e.g., as oral antiseptics. But as a strong oxidant, H2O2 can sustain radical reactions, and concentrated peroxide solutions therefore have the potential to form powerful explosives [4]. Second, it has more recently been recognized that H2O2 plays a very significant role in global atmospheric chemistry, as a stratospheric reservoir species for HOx [5, 6], but also more recently elevated concentrations have been associated with biomass burning [7, 8]. In the stratosphere, H2O2 is well established as a reservoir molecule for both the hydroxyl and hydroperoxyl radicals (HO and HO2 = HOx) as has been discussed by several authors [5, 6, 9]. This is seen in the following two reactions:
$$ {\text{H}}_2 {\text{O}}_2 + {\text{hν}} \to {\text{OH}} + {\text{OH}} $$
$$ {\text{OH}} + {\text{H}}_2 {\text{O}}_2 \to {\text{H}}_2 {\text{O}} + {\text{HO}}_2 $$
where reaction 2 exceeds reaction 1 consuming OH. In terms of source production, H2O2 comes about predominantly in the stratosphere via the HO2 self-reaction [10, 11]:
$$ {\text{HO}}_2 + {\text{HO}}_2 \to {\text{H}}_2 {\text{O}}_2 + {\text{O}}_2 $$
Certainly in the aqueous phase, but also in the vapor phase, peroxide can decompose to water and oxygen [4], catalyzed by any of several metals or higher temperatures:
$$ {\text{H}}_2 {\text{O}}_2 \to {\text{H}}_2 {\text{O + ½O}}_2 $$

In terms of biomass burning, Rinsland et al. [8] have very recently detected elevated H2O2 mixing ratios as high as 1.7 ppbv (1 ppbv = 1  × 109 per unit volume) where meteorological back trajectories associated the air mass with a young biomass burning plume. Previously Lee et al. [12] had shown that H2O2 mixing ratios can be as high as approximately 10 ppbv in biomass burning plumes and correlate well with CO and other known biomass burning species. It is also known that H2O2 is an important atmospheric oxidant (e.g., sulfur compounds), but is vulnerable to both dry and wet deposition owing its high water solubility. von Kuhlmann et al. [9] have shown that vapor-phase H2O2 can also be regenerated via various photochemical pathways, and has a tropospheric lifetime, τ, on the order of days, but there is large variability owing to deposition. Typical background concentrations in nonurban atmospheres have been measured by multiple techniques and have been found to typically range from approximately 0.2 to 2.0 ppbv using techniques including diode laser absorption spectroscopy, enzymatic fluorescence, and luminol chemiluminescence [7, 13, 14]. With the exception of these few works, relatively few researchers have studied H2O2 in the vapor phase; quantitative reference data, in particular, are quite sparse. H2O2 is in fact found in the HITRAN [15] vapor-phase IR spectral database, but there are only data for the strong ν6 band from 1,170 to 1,380 cm-1. Realizing that broadband IR spectroscopy is one of the few techniques available for remote sensing in an open-path sensor configuration [3], and also that many of the lead salt and quantum cascade IR laser methods [16, 17] offer good detectivity for in situ sensing such as for explosives detection or smog chamber experiments, the goal of the present work is to provide quantitative band strengths for as many H2O2 bands as possible, particularly with data resolved to 0.1 cm−1 for atmospheric sensing. We thus describe our methods used to generate the H2O2 vapor-phase mixtures, and how the quantitative spectra are then derived by subtracting water lines. The resulting band strength data are compared with band strengths available in the literature, which are essentially only for the strong ν6 band. However, we also discuss other H2O2 band strengths and line strengths in the IR and near-IR (NIR) and their potential use for atmospheric or industrial monitoring.


One of the prerequisites of data collection for the PNNL gas-phase IR database is that the chemicals be in a relatively pure state. This sometimes means purification via distillation, or more often “drying” chemicals such as ethers or ketones by removing the H2O impurities via adsorption onto CaSO4. For some species there are still trace amounts of other gases, most notably H2O vapor or CO2 gas, that must be subtracted from each of the individual pressure-pathlength burdens; typically 10 to 12 spectra are measured and the exact partial pressures of the analyte and the impurity are calculated for each burden. The composite spectrum has all impurities removed in this fashion. While tedious, impurity spectral subtraction presents no major difficulties so long as (1) one is confident there are no IR-transparent impurities, e.g., H2, N2, Ar . . . and (2) that any impurities are at sufficiently low concentration that their absorption features are not optically saturated; that is, they stay within the linear domain of Beer’s law behavior [1]. Otherwise, subtracting a given impurity ro-vibrational feature may result in removal of that optical feature, but may result in undersubtraction or oversubtraction of a different feature for the same species.

H2O2 preparation

For H2O2 clearly the “impurity” of concern is water. Our experience had shown that for IR spectral subtraction to be effective, it was necessary to achieve H2O2 concentrations of more than 50%. As 50% solutions are the strongest concentrations easily available owing to shipping regulations, and because 90% solutions were formerly commercially available with still-manageable hazard levels, it was deemed desirable to start with a 50% solution and obtain concentrations in the 70–90% range. That is, for this study achieving concentrations in this range was sufficient enrichment to avoid water line subtraction problems, yet deemed reasonably safe. Obtaining still higher concentrations is technically difficult and increases the chances of accidental detonation. Moreover, at room temperature the solutions are not stable, decomposing to H2O and O2 unless they are refrigerated. Fractional crystallization, whereby the water freezes at 0.00°C and the supernatant is rich in H2O2, was one possibility for increasing the H2O2 concentration, but is only successful at achieving concentrations up to approximately 55%. Classical distillation under a vacuum was deemed more appropriate: A commercial roto-vacuum device was used with an aspirator to achieve vacuum.

The 50% solution of H2O2 was obtained from Sigma-Aldrich, CAS no. 7722-84-1. To start, a Fourier transform (FT)-Raman spectral purity check of the stock liquid was collected over the range from 50 to 3,600 cm−1 Stokes shift and inspected for impurities using a previously described instrument [18]. The FT−Raman spectra have frequency accuracies shown to be better than 0.5 cm−1 or less, and inspection of the data showed only H2O2 and H2O bands. Prior to use, all components of the roto-vacuum device were soaked for hours (typically 24 h) in the 50% solution to scavenge the glassware (and metal) parts for trace organic residues. In a similar fashion, the syringes used to flow the solution into the disseminator T-piece were rinsed with 50% solution and soaked for at least 2 h to scour for organics. The distillation was conducted by placing approximately 20–25 ml of solution in the distillation flask and gradually warming the bath water over many minutes and eventually obtaining a temperature of 341 K. At this temperature, under an aspirator vacuum, the neck of the flask was filled with condensate and small amounts were collected in the condensation flask. The solution was distilled from about 50% by weight to between 82.8% for the 298 K spectrum and 82.5% for the 323 K spectrum. The density of the solution was obtained by drawing 500 µl into a Hamilton 500-µl syringe and weighing the amount on a balance with a precision of 0.001 g. The concentration was calculated from the measured density at room temperature (297 K) by using the formulae generated by Easton et al. [19]. A typical measured density was 1.352 g/ml, which corresponds to a solution that is 82.8% by weight. The weight percent concentrations were converted to mole percent using the molecular masses of H2O2, 34.0147 g/mol, and water, 18.0153 g/mol. The corresponding mole percent of H2O2 for the 298 K solution was 71.8% and for the 323 K batch was 70.7%.

Dissemination system and spectrometer

For the present experiments, rather than use a static cell configuration, the IR spectra were recorded using a previously described disseminator flow system [20]. This is an IR long-path cell coupled to a liquid vaporizer whereby the liquid analyte is quantitatively delivered via a syringe pump into a stream of ultra-high-purity N2 carrier gas, using a specially constructed heated vaporization piece [20]. The measurements were made in a customized White cell with the optical path set to 8.05 m (±0.5%). The cell has a circulating liquid jacket, which can provide more precise temperature control for the gas. The gas temperature is measured by placing a NIST-traceable temperature probe with an absolute accuracy of better than ±0.01 K directly into the gas, adjacent to the cell mirrors. One advantage of the flow system compared with a static cell for the H2O2 measurements is that the time to vaporize the pure liquid is minimized. As soon as the analyte leaves the syringe tip, it is flash-vaporized by the disseminator block (held at 44 or 56°C for the two sets of measurements) and diluted to just a few parts per million in the ultra-high-purity N2 carrier gas. Although there are additional wall contacts in the homogenization process, these are minimized, and all flow components were passivated with 50% solution prior to measurement, further reducing decomposition. For the concentrated solutions, however, it was noted that with time small bubbles developed inside the syringe owing to the spontaneous reaction/decomposition leading to the formation of H2O liquid and O2 gas (reaction 4). Although the O2 gas does enter the gas stream, its contribution to the ballast gas is negligible, and we still have a quantitative value for the number density of H2O2. It is assumed that the total number of H2O2 and H2O molecules leaving the syringe tip per unit time is the same as the total number flowing through the White cell per unit time, i.e., that there is a stoichiometric conversion of H2O2 to H2O according to reaction 4. [While H2O2 can also be photolyzed in the (upper) atmosphere via reaction 1, the effect is negligible in the dark gas cell.] The number density of H2O molecules is calculated from the IR spectrum and is subtracted from the total to yield the number of H2O2 absorbers.

A Bruker IFS 66v/S vacuum spectrometer [1] was used over the 520-7,500-cm−1 range with an external mercury cadmium telluride detector for the White cell. The spectrometer hardware characteristics have been previously documented [1, 21, 22, 23]. These papers also contain details of measurement parameters, as well as modifications to redress sundry artifacts, including ghosting, “warm aperture” [23], and detector nonlinearity [24, 25] phenomena.

Data reduction

The PNNL method typically measures 10 to 12 separate burdens ranging over approximately 2 orders of magnitude, with systematic errors in absorbance of approximately 7% for well-behaved molecules. For problematic species such as H2O2, the values can be higher. To account for any of several known nonlinearity phenomena in the fit at each wavelength channel, the individual burdens are also weighted according to T2 (where the transmittance T=I/Io). The multiple measurement with weighted data approach has several advantages in that the signal-to-noise ratio is enhanced, especially where the high-burden measurements bring out a better signal-to-noise ratio for weak bands, and also for the strong bands, where the weighting scheme brings out a better fidelity to account for Beer’s law saturation or detector nonlinearity effects. The statistical analysis is also useful at discerning chemical impurities, since their IR signatures typically do not scale with the fit. The present data were reduced in the same manner [1]. For the individual burdens, the water mixing ratios were quantified using the PNNL IR database. These vapor-phase-measured values were used to correct for the loss/conversion of H2O2 during the elution process as described above. The molar concentration for 298 K was measured to be 62.05% and that for 323 K was measured to be 60.75%. The molar concentrations represent a relative loss of approximately 14% (absolute loss approximately 10%) of H2O2 from the time of the original density measurement to the time the vapor-phase mixing ratio value was determined. Using the adjusted molar concentrations, we scaled the H2O2 spectra by 1.612 for 298 K and 1.646 for 323 K, i.e., the water features were removed by spectral subtraction to derive the H2O2 spectra presented in this paper.


Figure 1 presents our broadband IR spectrum of H2O2 from 600 to 4,300 cm−1. The H2O2 spectrum was first reported more than a half century ago by Giguère [26], and we use many of his assignments. We note, however, that the present data are quantitative with decadic absorbance values and the ordinate corresponds to an optical depth of 1 ppm H2O2 through a path of 1 m, with the number density adjusted to 296 K and 1 atm of N2 total pressure. The reported spectra, both for 298 and for 323 K, represent the weighted average of ten individual H2O2 measurements for each temperature. We have labeled the positions of the six vibrational fundamentals bands [26], where the torsional mode ν4 is beyond the range of the current spectrometer [27], but its first overtone 2ν4 is seen at the red end of the spectrum. (Note that for 2ν4 and 3ν4 the levels indicate the τ = 1 level only; see the paper by Camy-Peyret et al. [27].) Also labeled are the ν5 + ν4, ν1 + ν4, ν2 + ν6, as well as the ν4 + ν6 combination bands, all of which are active in the C2 (actually C2h double group) point group [28]. Redington et al. [29] showed that the ground-state structure has the hydrogens neither cis nor trans, but rather with a dihedral angle of 119.8o, with a cis barrier that is several times higher than the trans barrier. However, it was established that to a good approximation one can consider H2O2 as a symmetric rotor to separate the rotational levels into overall and internal rotational coordinates, and then consider the asymmetry perturbation for the overall levels. Essentially all of the bands show a systematic doubling [26], with the torsional barrier so small that the O-H groups are nearly freely rotating about the O-O bond at room temperature. The ν4 torsional mode has large rotational spacings owing to the relatively small moment of inertia for the axis through the plane of the molecule. The ν5 and ν1 bands (OH antisymmetric, symmetric stretches, respectively) overlap near 3,614 cm−1. The ν2 + ν6 band is relatively strong for a combination band, and was assigned by Redington et al. [29], displaying a prominent RQ0 subband near 2,658 cm−1 as discussed below.
Fig. 1

Broadband infrared spectrum of H2O2 vapor. The y-axis is quantitative and the absorbance corresponds to an optical depth of 1 ppm H2O2 through an optical path of 1 m. Band assignments are discussed in the text

It is also seen in Fig. 1 that the ν6 band, which has been used for remote sensing, is indeed the strongest band in the spectrum. Our results indicate that the integrated band intensities are 433 cm−2 atm−1 for the 298 K data and 467 cm−2 atm−1 for the 323 K data, where Napierian logarithm units are used. These are both in good agreement with the HITRAN values as discussed below [15]. The random error estimates on these values are approximately ±8% for the 298 K data and ±3% for the 323 K data. Although the integrated band strength should normally be invariant with temperature, we put more credence in our 323 K data, not because of the better agreement with the HITRAN values [15], but because of the smaller error bars and the greatly reduced adhesion/sticking phenomena at 298 K that can be understood as follows. Empirically, it is well known that as well as dissociating easily, H2O2 is a “sticky” molecule, easily sorbing or adhering to experimental surfaces [26], similar to other species, such as HNO3 or SOCl2 [30, 31]. For the 298 K data, the experiment forces one to decrease the temperature of all components downstream from the disseminator device, including the gas cell. For the 298 K data there is thus an inherently much greater adhesion to system surfaces. This in turn means that even after many minutes of flowing, the system will not have fully equilibrated. Moreover, after a high-burden measurement, it can take many minutes, or even hours, for residual H2O2 to fully desorb from the cooler surfaces, thus affecting the ensuing measurement through contamination. For the 323 K measurements, on the other hand, the exponential vapor-pressure dependence with temperature greatly reduces all of these effects, allowing for more rapid data acquisition and more reliable values at significantly higher concentrations (factors of approximately 3–5 times), which in turn greatly increases the signal-to-noise ratio and decreases all sorption or desorption “memory” effects.

The experimental results from the present measurements for the absolute intensity of the ν6 band are plotted versus the HITRAN values in Fig. 2. As seen, the agreement is excellent, with the difference plotted as the lower trace. For this plot the HITIRAN values [15] were plotted with the line width set to 0.1 cm−1 and the optical depth set to 1 ppm-m so as to correspond to the PNNL values. The resulting calculated HITRAN spectrum has been vertically offset for clarity. We note that the HITRAN data are in fact from the HITRAN2004 update: Those band intensities are scaled to the 1991 work of May [32], but the original compilation of lines was actually taken from an earlier work by Hillman et al. [33], which did not include all the torsional levels, and this partially accounts for the imperfect fit of some of the line profiles.
Fig. 2

Comparison of the ν6 band of H2O2 vapor from the HITRAN database (top trace) and from the Pacific Northwest National Laboratory (PNNL) experimental values (middle trace). Both plots correspond to an optical depth of 1 ppm m at 0.1-cm−1 resolution. The HITRAN spectrum [15] has been vertically offset for clarity. The bottom trace is the difference between the two spectra

In addition to the HITRAN2004 data, there have also been more recent high-resolution measurements of the ν6 band, as made by Klee et al. [34], which give very similar values. The Klee et al. values are slated to be in the HITRAN2008 release, and yield values similar to both the HITRAN2004 data and the present work. In that work Klee et al. derived a ν6 integrated band intensity for the natural isotopologue in the 1,170–1,380-cm−1 region of Sν6 (296 K) = 458 cm−2 atm−1. They further pointed out that if their results were modified to take into account the contributions from all hot bands as well as different isotopologues (as the present measurements inherently include), the amended value is STOT(296 K) = 467 cm−2 atm−1, with an estimated uncertainty of ±10% for the band strength. We note that the values reported by Klee et al. for ν6 were based on a calibration method that measured two bands in the same spectrum: In the same measurement they recorded not only the ν6 band, but also the rotational-torsional lines of the R branch of the 2ν4 band in the 370–700-cm−1 domain. They then used the absolute dipole moment measured by the Stark effect along with the rotational modes in the 2ν4 band to calculate the absolute line intensities and thus number densities, and applied these number densities to the ν6 band to calculate the line strengths.

As opposed to the technique of Klee et al., our method is a flow method based on quantitative liquid volume dissemination and assumes (1) complete vaporization of the analyte in N2, a safe assumption at these low mixing ratios, and (2) any H2O2 loss in the flow is converted stoichiometrically to H2O according to reaction 4. Mass balance is preserved and the water vapor is then quantitatively subtracted. We note that, irrespective of method, any quantitative H2O2 measurements are enigmatic in that any attempt to reduce the effects of adhesion or impurities by increasing the temperature or H2O2 concentrations, respectively, exponentially increases the rate of the decomposition and thus increases the impurities! Any analysis must either account for such inevitable H2O2 transformations or have an indirect method of calibration, such as that used by Klee et al. [34].

The optical densities reported here are absolute, and we estimate a random error on the order of ±8% for the 298 K data and ±3% for the 323 K data. For “well-behaved” molecules [20], which H2O2 is not, the estimated systematic errors are ±7%. That the band strength agrees so well with both the HITRAN [15] as well as the Klee et al. [34] values when they were derived by independent methods is satisfying. We also note while the agreement with HITRAN data [15] is good, i.e., the Klee and May [32] values, the present value for Sν6(296 K) of 467.0 cm−2 atm−1 is substantially higher than either of two earlier values, namely, the value of 200 cm−2 atm−1 of Niki et al. [35] and the value of 346 cm−2 atm−1 of Valero et al. [36]. While at lower resolution than, e.g., the HITRAN data, the present broadband IR spectrum does allow for the measurement of strengths of several different H2O2 bands. The integrated band intensities of two fundamentals, two combination bands, as well as the region containing the ν5 and ν1 O-H fundamentals are presented in Table 1. While base 10 logarithms were used to calculate the values of the data in the PNNL database, Napierian logarithms were used to calculate the S values of the band integrals values at 296 K. While the two values are nominally independent of temperature for each band [31], it is seen that the integrated intensities are systematically about 7% smaller for the 298 K values. More certainty is ascribed to the 323 K values for the reasons discussed above. Most of these band strengths are reported for the first time.
Table 1

Band strength measurements (reduced to 296 K) for the stronger fundamental and combination bands of vapor-phase H2O2


Region (cm−1)

S (cm−2 atm−1)

298 K

323 K













ν5, ν1 region




ν1 + ν5




For the O-H stretching fundamentals ν5 and ν1 the bands overlap and are of comparable magnitude and thus the 3,300–3,800-cm−1 integral represents both modes

In addition to the mid-IR band strengths discussed above, the present work also reports the quantitative NIR spectra and band strengths of vapor-phase H2O2. In the last decade, the NIR region has emerged as a spectral domain of great utility and importance. Though NIR cross sections are typically 1–2 orders of magnitude weaker than the mid-IR fundamentals, the enhanced sensitivity and robustness of NIR components, the availability of optical fiber couplings, and the power of chemometric methods such as partial least squares has led to a renaissance in NIR applications. H2O2 is no exception: H2O2 vapors are now routinely used in the medical and pharmaceutical industries as a sterilization technique, and NIR methods are used to quantify the H2O2 vapor concentrations [37, 38]. Clearly other applications are possible using such NIR methods. However, many of the NIR (dispersive and FT) systems use low resolution, so there is significant spectral overlap, which in turn requires extensive calibrations to account for the interferents. For H2O2 this means calibration to account for water (and other) vapors. A further problem is almost a complete lack of quantitative reference data, meaning that each system must develop its own calibration data or training methods [37, 38]. Such calibrations are usually not portable because they are specific to the spectral bandwidth and resolution of the system, as well as the chemical list developed for the training set.

The data shown in Fig. 3, however, are independent of the system and the instrumental linewidth (0.1 cm−1) is narrower than any of the observed pressure-broadened features at these wavelengths. The plot represents 1 ppm-m of H2O2 and 100 ppm m of H2O. The signal-to-noise ratio for these NIR data is only average because this region is at the absolute limit of coverage for the present spectrometer configuration; furthermore, the resolution is quite high. A better signal-to-noise ratio can be had in this region by using alternate optical components as was done for our CH2I2 studies [39], but the signal-to-noise ratio is sufficient for this work as the region is “for free.” Close inspection of Fig. 3 shows that the feature is a parallel-type band and was assigned as the ν1 + ν5 combination band by Giguère [26]. That work also showed its doublet character of the band, the doublet indicating a tunneling between the double minimum of the ν4 torsional mode. Also plotted as the top trace in Fig. 3 is the water vapor spectrum from the PNNL database, but amplified to 100 ppm m. This shows that trace H2O2 monitoring in the NIR is feasible even with water concentrations approaching ambient. It is seen that some of the peroxide P-branch doublet lines, especially for high temperatures, are relatively free of water interference, and that several of these lines, e.g., 6,965.0, 6,976.8, or 6,990.2 cm−1, could be used for atmospheric or industrial peroxide monitoring. There exist different types of lasers at these wavelengths (such as the 1.35-µm telecommunication lasers) that would be well suited to such applications [16]. When the spectra are deresolved to resolutions used in most monitoring applications, inspection of the two spectra shows that even at low resolution, shifting to longer wavelengths, e.g., from 1,410 to 1,425 nm (from 7,092 to 7,018 cm−1) reduces the water absorbance values close to zero, obviating the lengthy calibration methods to compensate for the water signal [37, 38]. Table 1 also includes the band strength (26.3 cm−2 atm−1 ±10%) for the ν1 + ν5 band. In addition to the relative freedom from interference, the present result indicates further utility for this band since this value is more than double the only previously reported estimate of 12 ± 2 cm−2 atm−1 from Hagen and Sanders [40], although that was estimated from data in the paper of Adams et al. [37]. Using theoretical chemistry self-consistent-field models at the 4–31G level, Rogers [41] predicted a band intensity for the ν1 + ν5 band of 86.0 cm−2 atm−1, clearly higher than the value measured here or predicted by Kjaergaard et al. [42]. As the red end of the P-branch is almost free from water O-H as well as organic C-H stretch overtones, these lines could be useful for dedicated peroxide detection, provided the spectral resolution is sufficient. A current need is for H2O2 process monitoring in piping [43].
Fig. 3

Near-infrared spectrum of the ν1 + ν5 combination band of H2O2 (bottom trace). The bottom trace corresponds to an optical depth of 1 ppm m with 0.1-cm−1 spectral resolution. Also plotted (top trace) is the spectrum of water vapor at 100 ppm m, also from the PNNL database

For laser or FTIR monitoring at resolutions better than approximately 0.5 cm−1 in applications where high concentrations (approximately more than 1 ppm) might be expected, Fig. 4 shows an alternative H2O2 detection possibility, namely, the v2 + v6 combination band with a pronounced feature at 2,658.62 cm−1. Other workers [26] have assigned this as the v2 + ν3 + ν4 band. However, Redington et al. [29] reported the ν2 + ν6 assignment in their paper on the H2O2 internal rotation. The ν2 + ν6 band shows no resolvable doubling at high resolution, but the PQ2, RQ1, and RQ0. subbands are clearly resolved. The strong feature at 2,658.62 cm−1 has been assigned to RQ0, although it was not resolved in the present pressure-broadened spectra. Figure 4 indicates it could be of significant utility for in situ monitoring purposes. While ν5, ν1, and the R branch of the ν6 bands are obscured by the fundamental modes of water, this 2,658.6-cm−1 subband is in a relatively uncongested region.
Fig. 4

The PNNL H2O2 vapor infrared spectrum (black) of the ν2 + ν6 combination band with an optical depth of 1 ppm m. Also plotted from the HITRAN [15] database (blue) is the spectrum of water vapor through 1 m, corresponding to 296 K and a 10% relative humidity

In addition to the H2O2 spectrum for a mixing ratio of 1 ppm, the 298 K, 760 Torr spectra with a 10% relative humidity data are also plotted in Fig. 4. A value of 1 ppm is far greater than expected for ambient concentrations, namely, more than 2 orders of magnitude greater than what might be found in clean air background levels for H2O2. However, there may be environments where H2O2 is being monitored at high concentrations, such as smog chambers, process piping, or perhaps in biomass burning scenarios [8]. Aside from weak water lines, this spectral region is relatively free of bands for common IR vapor-phase interferents such as CO, CO2, SO2, CH4, and NOx, although HNO3 does have a band in this region. We note that the RQ0 2,658-cm−1 feature has a linewidth that is only 0.40 cm−1 wide, offering the possibilities of not only extractive, but also open path in situ monitoring, also suggested by Rogers [41]. Open path spectroscopic sensing is of great advantage for species such as H2O2 or HNO3, where the effects of any inlet system are always problematic. Contemporary IR laser systems, either lead salt or quantum cascade, have good detection limits on the order of 5 × 10−6 optical density for approximately 1-min averaging times, and even approximately 5 × 10−7 optical density for the best systems [44, 45]. If we assume that such a laser system can achieve such limits with a 200-m-long path cell length, and noting that the feature near 2,658 cm−1 has a differential cross section of 3.16 × 105 (ppm m)−1, this corresponds to an optimistic detection limit of 790 ppt for a good system or 79 ppt for the best systems.

The current commonly used lines [8, 13] for H2O2 monitoring in the IR region are in the P-branch of ν6 and have differential cross sections (at atmospheric pressure) approximately 4 times stronger than the 2,658-cm−1 feature. However, while 4 times weaker, the present line offers the possibility of more sensitive InSb or other detectors, as well as more powerful lasers near 3.7 μm. As Fig. 4 shows, however, interference from water could be a concern and would be best aimed at systems where the humidity is artificially or naturally low, e.g. upper atmosphere or a reaction chamber. We are optimistic that these lines as well as some of those lines in the ν1 + ν5 combination band in the NIR region are of sufficient strength that when used with quantum cascade or diode laser systems they will be useful for open path monitoring, an important concern for a reactive and easily decomposed species such as H2O2. Many such lead salt and quantum cascade laser systems have been used throughout the IR region (λ = 2–20 µm) [46]. Several NIR laser systems have been deployed using techniques such as frequency modulation [47] or cavity ring-down methods [48]. In recent years, the utility of such methods has increased owing to extended tuning ranges of the lasers, allowing them to quickly scan across such a pressure-broadened line.


Although H2O2 plays an important role in atmospheric chemistry, there exist relatively few data for vapor-phase quantification, including in the IR region. There are H2O2 reference data in the HITRAN database [15], but unfortunately only for the strong ν6 band from 1,200 to 1,350 cm−1. These data originated from the quantitative work of May [32], as well as the more recent work of Klee et al. [34], and are some of the few experimental spectra with which we can compare the present results. The present system started with a concentrated (83%) solution of H2O2 and measurements were made in a flow system whereby the H2O2 was dispensed from a syringe and instantly diluted with N2 gas. There are several advantages of such a system in that prepassivation reduces decomposition losses, as does the analyte being instantly diluted to the parts per million level in the N2 carrier gas. Although there are clearly losses (forming water), there is a steady-state stability over time, as we saw for extended periods using the 50% stock solution. One disadvantage of such a system is the many surfaces of the disseminator and gas cell that result in longer equilibration times (i.e., more adhesion or decomposition). But a real advantage of the present method lies in the fact that our H2O2 quantitation system can easily account for all impurities and decomposition products. This is true for any system, even if a drying agent (e.g., P2O5) is used: One must account for H2O2 transformations, and such losses are exacerbated by metals or high temperatures. In our case, quantifying the water vapor concentrations (reaction 4) is straightforward with IR spectroscopy. The alternative is to use a secondary in situ method to quantify the number of H2O2 absorbers in the beam.


The results presented here are broadband, and for the ν6 band agree quite well with both the HITRAN and the Klee et al. values. The good agreement for the ν6 band integrals gives us equal confidence in the values for other bands as they are derived from the same measurements. The new values for the NIR ν1 + ν5 band suggest great utility for both broadband and NIR laser monitoring methods, particularly using the resolved lines of the P-branch. The strong feature at 2,658.6 cm−1, the RQ0 subband of the ν2 + ν4 combination band, looks attractive for open-path monitoring. For this compound open-path monitoring using such data offers a distinct advantage over any extractive measurement owing to the inherent decomposition of the analyte.

Broadband quantitative IR and NIR spectra of vapor-phase H2O2 have been reported for the first time. Results for the strong ν6 integrated band intensities agree very well with those of previous work. We have reported several new band integrals for the first time, as part of the quantitative data for both 298 and 323 K. The disseminator method works very well for such reactive species, but in the case of H2O2 vapor relies on quantitative subtraction of the inevitable H2O degradation product. Such data are clearly useful in many fields, such as medicine, explosives sniffing, biomass burning, and general atmospheric and physical chemistry.



We thank Jean-Michel Régimbal of John Abbott College in Sainte-Anne-de-Bellevue in Montreal for helpful advice. PNNL is operated for the US Department of Energy by the Battelle Memorial Institute under contract DE-AC06-76RLO 1830. This work was supported by the Strategic Environmental Research and Development Program (SERDP) sustainable infrastructure program. The work was also supported by the DOE NA-22 program and we thank both sponsors for their support. The experiments were performed at the W.R. Wiley Environmental Molecular Sciences Laboratory, a national scientific user facility sponsored by DOE's Office of Biological and Environmental Research and located at PNNL.


  1. 1.
    Sharpe SW, Johnson TJ, Sams RL, Chu PM, Rhoderick GC, Johnson PA (2004) Appl Spectrosc 58:1452–1461CrossRefGoogle Scholar
  2. 2.
    Rinsland CP, Malathy-Devi V, Blake TA, Sams RL, Sharpe SW, Chiou L (2008) J Quant Spectrosc Radiat Transf 109:2511–2522CrossRefGoogle Scholar
  3. 3.
    Johnson TJ, Roberts BA, Kelly JF (2004) Appl Opt 43:638–650CrossRefGoogle Scholar
  4. 4.
    Kirk RE, Othmer DF, Kroschwitz JI, Howe-Grant M (eds) (2005) Kirk-Othmer encyclopedia of chemical technology, vol 13. Wiley, New YorkGoogle Scholar
  5. 5.
    Davies DD (1974) Can J Chem 52:1405–1414CrossRefGoogle Scholar
  6. 6.
    Connell PS, Wuebbles DJ, Change JS (1985) J Geophys Res 90:10726–10732CrossRefGoogle Scholar
  7. 7.
    Snow JA, Heikes BG, Shen H, O’Sullivan DW, Fried A, Walega J (2007) J Geophys Res. doi:10.1029/2006JD00746
  8. 8.
    Rinsland CP, Coheur PF, Herbin H, Clerbaux C, Boone C, Bernath P, Chiou LS (2007) J Quant Spectrosc Radiat Transf 107:340–348CrossRefGoogle Scholar
  9. 9.
    von Kuhlmann R, Lawrence MG, Crutzen PJ, Rasch PJ (2003) J Geophys Res 108(D23):4729CrossRefGoogle Scholar
  10. 10.
    Ruey-Rong L, Gorse RA, Sauer MC, Sheffield G (1979) J Phys Chem 83:1803–1804Google Scholar
  11. 11.
    Takacs GA, Howard CJ (1984) J Phys Chem 88:2110–2116CrossRefGoogle Scholar
  12. 12.
    Lee M, Heikes BG, Jacob DJ, Sachse G, Anderson B (1997) J Geophys Res D 102:1301–1309CrossRefGoogle Scholar
  13. 13.
    Kleindienst TE, Shepson PB, Hodges DN, Nero CM, Arnts RR, Dasgupta PK, Hwang H, Kok GL, Lind JA, Lazrus AL, Mackay GI, Mayne LK, Schiff HI (1988) Environ Sci Technol 22:53–61CrossRefGoogle Scholar
  14. 14.
    Staffelbach TA, Kok GL, Heikes BG, McCuilly B, Mackay GI, Karecki DR, Schiff HI (1996) J Geophys Res 101:33–66CrossRefGoogle Scholar
  15. 15.
    Rothman LS, Jacquemart D, Barbe A et al (2005) J Quant Spectrosc Radiat Transf 96:139–204CrossRefGoogle Scholar
  16. 16.
    Johnson TJ, Wienhold FG, Burrows JP, Harris GW (1991) Appl Opt 30:407–413CrossRefGoogle Scholar
  17. 17.
    Sharpe SW, Kelly JF, Hartman JS, Gmachl C, Capasso F, Sivco DL, Baillargeon JN, Cho AY (1998) Opt Lett 23:1396–1398CrossRefGoogle Scholar
  18. 18.
    Williams SD, Johnson TJ, Gibbons TP, Kitchens CL (2007) Theor Chem Acc 117:283–290CrossRefGoogle Scholar
  19. 19.
    Easton MF, Mitchell AG, Wynne-Jones WFK (1952) Trans Faraday Soc 48:796–801CrossRefGoogle Scholar
  20. 20.
    Johnson TJ, Sharpe SW, Covert MA (2006) Rev Sci Instrum 77:094103. Erratum in Johnson TJ, Sharpe SW, Covert MA (2007) Rev Sci Instrum 78:019902Google Scholar
  21. 21.
    Foster NS, Thompson SE, Valentine NB, Amonette JE, Johnson TJ (2004) Appl Spectrosc 58:203–211CrossRefGoogle Scholar
  22. 22.
    Johnson TJ, Valentine NB, Sharpe SW (2005) Chem Phys Lett 403:152–155CrossRefGoogle Scholar
  23. 23.
    Johnson TJ, Sams RL, Blake TA, Sharpe SW, Chu PM (2002) Appl Opt 41:2831–2839CrossRefGoogle Scholar
  24. 24.
    Birk M, Hausmann D, Wagner G, Johns JW (1996) Appl Opt 35:2971–2985CrossRefGoogle Scholar
  25. 25.
    Chase DB (1984) Appl Spectrosc 38:491–494CrossRefGoogle Scholar
  26. 26.
    Giguère PA (1950) J Chem Phys 18:88–92CrossRefGoogle Scholar
  27. 27.
    Camy-Peyret C, Flaud J-M, Johns JWC, Noël M (1992) J Mol Spectrosc 155:84–104CrossRefGoogle Scholar
  28. 28.
    Hougen JT (1984) Can J Phys 62:1392–1402Google Scholar
  29. 29.
    Redington RL, Olson WB, Cross PC (1962) J Chem Phys 36:1311–1326CrossRefGoogle Scholar
  30. 30.
    Chackerian C, Sharpe SW, Blake TA (2003) J Quant Spectrosc Radiat Transf 82:429–441CrossRefGoogle Scholar
  31. 31.
    Johnson TJ, Disselkamp RS, Su Y-F, Fellows RJ, Alexander ML, Driver CL (2003) J Phys Chem A 107:6183–6190CrossRefGoogle Scholar
  32. 32.
    May RD (1991) J Quant Spectrosc Radiat Transfer 45:267–272CrossRefGoogle Scholar
  33. 33.
    Hillman JJ, Jennings DE, Olson WB, Goldman A (1986) J Mol Spectrosc 117:46–59CrossRefGoogle Scholar
  34. 34.
    Klee S, Winnewisser M, Perrin A, Flaud J-M (1999) J Mol Spectrosc 195:154–161CrossRefGoogle Scholar
  35. 35.
    Niki H, Maker PD, Savage CM, Breitenbach LP (1980) Chem Phys Lett 73:43–46CrossRefGoogle Scholar
  36. 36.
    Valero FPJ, Goorvitch FS, Bonomo FS, Boese RW (1981) Appl Opt 20:4097–4101CrossRefGoogle Scholar
  37. 37.
    Adams D, Brown GP, Fritz C, Todd TR (1998) Pharm Eng 18:1–11CrossRefGoogle Scholar
  38. 38.
    Corveleyn S, Vandenbossche GMR, Remon JP (1997) Pharm Res 14:294–298CrossRefGoogle Scholar
  39. 39.
    Johnson TJ, Masiello T, Sharpe SW (2006) Atmos Chem Phys 6:2581–2591Google Scholar
  40. 40.
    Hagen CL, Sanders ST (2007) Meas Sci Technol 18:1992–1998CrossRefGoogle Scholar
  41. 41.
    Rogers JD (1984) J Phys Chem 88:526–530CrossRefGoogle Scholar
  42. 42.
    Kjaergaard HG, Goddard JD, Henry BR (1991) J Chem Phys 95:5556–5564CrossRefGoogle Scholar
  43. 43.
    Wehrum WL (1993) Process Saf Prog 12:199–202CrossRefGoogle Scholar
  44. 44.
    Richter D, Fried A, Wert BP, Walega JG, Tittel FK (2002) Appl Phys B 75:281–288CrossRefGoogle Scholar
  45. 45.
    Wert BP, Fried A, Rauenbuehler S, Walega J, Henry B (2003) J Geophys Res 108(12):4350–4362CrossRefGoogle Scholar
  46. 46.
    Werle PW, Mazzinghi P, D’Amato F, De Rosa M, Maurer K, Slemr F (2004) Spectrochim Acta A 60:1685–1705CrossRefGoogle Scholar
  47. 47.
    Johnson TJ, Wienhold FG, Burrows JP, Harris GW, Burkhard H (1991) J Phys Chem 95:6499–6502CrossRefGoogle Scholar
  48. 48.
    Bitter M, Ball SM, Povey IM, Jones RL (2005) Atmos Chem Phys 5:2547–2560CrossRefGoogle Scholar

Copyright information

© US Government 2009

Authors and Affiliations

  • Timothy J. Johnson
    • 1
  • Robert L. Sams
    • 1
  • Sarah D. Burton
    • 1
  • Thomas A. Blake
    • 1
  1. 1.Pacific Northwest National LaboratoryRichlandUSA

Personalised recommendations