A Kinetic Study of the Autoxidative Formation of VOCs, Including Formaldehyde, Acetaldehyde and Acrolein from Polyurethane Soft Foams

Abstract

The oxidation of flexible polyurethane (PUR) foams significantly impacts product durability, vehicle indoor air quality, and volatile organic compound (VOC) emissions. This study investigates oxidation kinetics and VOC emissions (65–155 °C) from foams with indices between 70 and 115 (molar ratio of NCO to NCO-reactive groups × 100), where a higher index represents greater hard segment (methylene diphenyl diisocyanate) and lower soft segment (polyether polyol) content. Using a flow-through setup with PTFE chambers and Tenax thermodesorption tubes and dinitrophenylhydrazine (DNPH) cartridges, VOCs from initial analyte loading, hydroperoxide degradation, and autoxidation were distinguished, providing robust kinetic data unaffected by diffusion interference. A higher index accelerated soft segment degradation, increasing oxidation rates and VOC emissions. The activation energy of 1,2-propanediol-1-acetate-2-formate increased from 87 kJ/mol in low-index to 108 kJ/mol in high-index formulations. VOC emissions from high-index foams were tripled for acetaldehyde during long-term aging at 65 °C. While most emissions followed Arrhenius behavior, formaldehyde and acrolein deviated above 100 °C, with higher hard-segment content extending their Arrhenius range. These findings link PUR composition to degradation behavior and emissions, enabling formulation improvements. The results advance methods for evaluating raw material contributions and the performance of antioxidants under realistic aging conditions.

1. Introduction

Polyurethane (PUR) is synthesized using polyols and isocyanates as primary raw materials [1]. One of polyurethane’s characteristics is the tunability of its physical properties, which can be precisely controlled by modifying the raw materials [2,3,4,5,6,7,8]. This tunability allows the production of a vast selection of different PUR products including thermoplastics, coatings, sealants and many types of foams.

One significant challenge faced by the polyurethane foam-producing industry is the material’s susceptibility to oxidation, which leads to the formation of volatile organic compounds (VOCs) emitted from open-celled foams, such as those used in automotive seating or dashboards. Emissions in general originate from contaminants dissolved in the raw materials or are products of the autoxidation of starting materials of the polymer or the polymer itself. PURs based on polyether polyols are particularly susceptible to oxidative degradation through hydroperoxide intermediates [9,10,11,12,13,14,15,16,17]. Eventually these intermediates can decompose into VOCs, including formaldehyde, acetaldehyde, acrolein, and propanal [18]. The related potential risks have garnered attention from researchers and regulatory bodies [19,20,21,22,23,24,25]. Given these concerns, VOC analysis has become essential for evaluating their sources and potential impacts on human health and indoor air quality.

VOC analysis plays a critical role in assessing indoor air quality and evaluating the toxicological impacts of emitted VOCs [26,27,28,29,30,31,32,33]. This analysis is pivotal in regulatory and consumer health efforts to address VOC sources that contribute to poor air quality or potential health risks. VOC analysis is not limited to indoor air quality assessments; it has also been applied to studying emissions from building materials and daily-use consumer products. For example, initial emission potentials of materials like particle boards have been extensively studied [34,35,36], while investigations into products such as mattresses and textiles have evaluated their long-term and qualitative VOC emissions [37,38,39,40,41]. Due to the extended periods humans are exposed to these products, emissions from PUR mattresses and similar products have been studied under various atmospheric conditions to determine time-dependent behavior [40,41]. Similarly, consumer products in proximity to respiratory systems, such as filtering respirator masks, have been investigated for their VOC profiles, including initial emissions, time-dependent decline, and associated toxicological harm potential [42].

In parallel, substantial effort has been devoted to improving VOC sampling methods and addressing analyte-specific challenges. Advancements in sampling strategies focus on optimizing methodologies to adapt to diverse conditions and enhance robustness and reliability. For example, studies have investigated derivatization efficiencies for aldehydes [43,44] and the efficacy of different adsorbents [45]. In particular, VOC method development is often focused on specific target analytes, such formaldehyde. Results of formaldehyde emission testing typically depend heavily on the method used for its assessment. Consequently, formaldehyde emission measurements are often influenced more by the testing methodology than by the material itself [34,36,46,47,48]. Assigning a precise concentration to formaldehyde emissions presents a significant challenge due to its high reactivity, which can result in partially irreversible sorption that diminishes both formation and emission results [49]. Additionally, formaldehyde emissions are strongly influenced by the relative humidity of the atmosphere into which it is released, further complicating accurate quantification [50,51,52,53]. While formaldehyde illustrates the significant testing and analytical challenges in VOC research, similar issues arise for many other analytes, highlighting the broader necessity of overcoming these obstacles to develop robust and standardized methods across diverse compounds.

Efforts in material optimization have shown options to mitigate VOC emissions during the service life of products. For instance, the use of antioxidants has demonstrated a reduction in emissions [54]. Additionally, research into polyurethane chemistry has revealed that the type and concentration of catalysts used during synthesis can serve as a source of certain aldehyde emissions [55]. Understanding these methodological challenges is essential for the development of methods for reliable emission testing that enables researchers to understand the influence of various parameters on sample emissions.

A comprehensive understanding of the formation and emission kinetics of these VOCs is crucial for accurate exposure modeling [56,57]. Formation rates of aldehydes and other VOCs are influenced by temperature [58,59,60,61,62,63], sample age [18], polymer structure [64], and potentially other parameters. In addition, emission rates also depend heavily on analyte transport processes, such as diffusion through the polymer matrix and foam cell structures, as well as adsorption at interfaces, adding further complexity to quantification attempts [65].

Designing experiments to study these processes is therefore challenging [34,35,36,65,66,67,68,69,70]. The inherent discrepancy between analyte formation rates and emission rates is confounded by the varying initial levels of contaminants (often derived from synthesis byproducts or adsorption from the atmosphere) and by the thermal history of the material. During manufacturing, the quasi-adiabatic conditions in the foam dictate high core temperatures, influenced by the reaction enthalpy, mold temperature, and the foam’s insulating properties. These core temperatures can range from approximately 70 °C in automotive panel cushioning to 120 °C in seating foams and up to 160 °C in slabstock foam systems. These elevated and spatially inhomogeneous temperatures subject the material to significant thermal stress, leading to inhomogeneous distribution of contaminants due to evaporation and condensation, as well as localized thermally degraded regions within the polymer matrix. All these complexities have raised concerns about the accuracy of industry-standard aging tests in predicting long-term formaldehyde and VOC release under real-world conditions.

There is a notable lack of academic research investigating the underlying chemical mechanisms of aldehyde and VOC formation in PUR foams. Most prior investigations have focused on cumulative VOC emissions or initial analyte loading and their time-dependent emission behavior, leaving the fundamental autoxidation mechanisms of PUR largely unexplored [38]. Temperature-dependent emission data is only put into context with emissions, not distinguished from analyte formation. The question of whether findings from polyether autoxidation can be directly extrapolated to polyether-based PUR remains unanswered. The structural complexity of PUR, with its interplay of hard and soft segments, may fundamentally alter oxidation pathways, impact product distributions, and result in distinct rates of VOC formation [71]. Research investigating VOC emissions typically only observes transitory and declining emission rates, not the long-term continuous emission potential of a material. For example, automotive seating’s long-term foam emission should be mostly free of its initial loading, however, the emission rate of volatile compounds continuously formed through the material’s autoxidative degradation is unknown yet potentially toxicologically relevant.

To address this gap, this study investigates VOC formation rates and oxidation kinetics in five methylene diphenyl diisocyanate (MDI)-based PUR foam formulations, with indices (denoting the molar ratio of NCO to NCO-reactive groups × 100) spanning the range of 70 to 115. By employing a novel flow-through setup, this study isolates VOC sources, quantifying emissions from initial loadings, hydroperoxide decomposition, and autoxidation. Temperature-dependent emissions between 65 °C and 155 °C are analyzed to obtain Arrhenius parameters, providing valuable insights into the oxidative stability of different formulations. This temperature range was selected to encompass typical production core temperatures and testing temperatures used in industry standards for polyurethane aging and VOC emission studies, enabling transferability of insights between this study and current industrial practices. The experimental design bridges the limitations of standard methods, such as VDA278 (VDA is the German Automotive Manufacturers Association (Verband der Automobilindustrie e. V.)) [72], by enabling mechanistic investigations of VOC formation under controlled test conditions. Additionally, long-term aging studies at 65 °C capture aldehyde formation trends over three months, offering a reliable framework for understanding PUR degradation.

The experimental setup has been used by the authors before to reproducibly determine aldehyde and VOC emissions from PUR [17,18]. Initial loadings can be stripped and hydroperoxides decomposed before sampling. Diffusion-limiting features of foams are overcome by forced convection. The approach here employs representative sample masses to facilitate a comprehensive examination of VOC emission in various atmospheres over a wide temperature range.

Building on previous foundational research and understanding of PUR autoxidation and the associated formation and emission of VOCs, the present study adds essential kinetic data in the form of activation energies and pre-exponential factors derived from linear Arrhenius fits. The aim is to provide a robust scientific basis for researchers and manufacturers working with PUR materials worldwide.

2. Materials and Methods

2.1. Polyurethane Soft Foams

Foams were synthesized by Covestro Deutschland AG (Leverkusen, Germany). The two components were pressurized to 150 bar by piston pumps and mixed through nozzles in a high-pressure mixing head [73]. The resulting isocyanate-in-polyol emulsion flows laminarly into a 5 L bucket and after resting (cream time) rises freely. At the time of maximum rise (rise time) the cell walls yield and the internal pressure in the foam is released. The foam is open-celled. The calculated adiabatic rise in temperature in the core of the foam is between 82 °C (index 70) and 111 °C (index 100).

The foams’ cream times were 6 s. The string times were between 35 s and 39 s. The rise times were between 36 s and 42 s. Longer times were observed for higher-index foams.

Cylinders with 14 mm diameter were punched out of each foam sample and used for computed tomography (CT) analysis (Figure 1, Figure 2 and Figure 3). Samples were scanned with an acceleration voltage of 50 kV, a current of 100 mA, a binning of 2048 by 2048, no filter and a spatial resolution of 8 to 9 µm per voxel. Scanning time was 2.5 h each.

2.2. Polyethers

The polyol formulation is a product of Covestro Deutschland AG. Its base polyalkylene oxide block copolymers and the blend including the additives were produced in Dormagen, Germany.

The polyether polyols are produced batch-wise through anionic ring-opening polymerization and have the following block structure:

(C₆H₁₄O₆)-((O-CH₂CH(CH₃)-)_b(O-CH₂-CH₂)_a-OH)₆

These polyether polyols contain a certain amount of allyl-terminated polymer chains, as propylene oxide rearranges to allyl alcohol during anionic ring-opening polymerization. Allyl alcohol then acts as an additional starter for polymerization.

CH₃-C₂H₃O → CH₂ = CH-CH₂OH → CH₂ = CH-CH₂-(O-CH₂CH(CH₃)-)_b(O-CH₂-CH₂)_a-OH

Allyl-terminated polyethers can rearrange into propenyl-terminated polyethers under strongly basic conditions [74].

CH₃-CH = CH-(O-CH₂CH(CH₃)-)_b(O-CH₂-CH₂)_a-OH

Acidic workup in industrial production eliminates most of the propenyl groups. There may be residual quantities or yet unidentified organocatalytic conditions that lead to further allyl-propenyl rearrangements during foaming or in foams. These may eventually lead to propanal emissions.

2.3. Polyisocyanate

The isocyanate component consisted of a blend of monomeric MDI and oligomeric homologs, with an isocyanate content of 33.1% blended by Covestro Deutschland AG, Krefeld, Germany.

The isocyanate employed had the following structure:

O=C=N-Ar-(CH₂-Ar-NCO)_c-CH₂-Ar-N=C=O (c ≈ 0.2)

The polymer resulting from the polyaddition of these compounds is a segmented polyurethane (Figure 4).

The recipe corresponds to a previously published one [17], with two modifications: triethylenediamine has been replaced by the reaction product of N,N-dimethylamino-3-propylamine with two equivalents of propylene oxide, and the glycerol-started polyether with 0.62 mol OH/kg has been replaced by a similar block-co-polyether with 0.50 mol OH/kg. The dimethylamino groups from the catalysts are known to contain and oxidatively yield formaldehyde [55,75].

A/B is the mass ratio of polyol blend A to isocyanate blend B, PO is propylene oxide as a monomer in the polyether in A, EO is ethylene oxide as a monomer in the polyether in A. At index 100 the water, if consumed quantitatively in the reaction with isocyanate, consumes 90% of the NCO groups to yield urea. Weight loss (69–81 g/kg) caused by the reaction of isocyanate with water to CO₂ was calculated and the weight of the components was adjusted. The effect of buoyancy (2.4%) was also accounted for in the apparent density given in

Table 1. Mixing ratio of compounds used to synthesize foams of varying index (mol NCO/mol NCO-reactive × 100).

Index	A/B	Hard Segment			Soft Segment		Hard Segment	Dimethylamino Group	Foam Core Density
(mol NCO/mol NCO-Reactive × 100)	g/g	Mol NCO/kg Foam	Mol CH2/kg (A + B)	Mol Urea/kg Foam	Mol PO/kg Polymer	Mol EO/kg Polymer		mmol/kg	kg/m3
70	1.93	2.9	1.7	1.9	9.6	2.9	29%	38.0	71.2
85	1.59	3.3	1.6	2.3	9.0	2.8	34%	35.5	68.6
90	1.51	3.5	1.6	2.5	8.8	2.7	35%	35.0	65.3
100	1.35	3.7	1.5	2.7	8.4	2.6	38%	33.4	54.6
115	1.17	4.0	1.4	3.0	7.9	2.4	42%	31.5	59.7

. The changes in density show the utilization of water for foaming: if there is not sufficient isocyanate not all the water is transformed into CO₂ and the density rises accordingly.

In ideal stoichiometry 85 mol-% of all isocyanate groups are converted to urea groups by reaction with water. Then, 12 mol-% will react to primary urethane aryl-NH-CO-O-CH₂- and 3 mol-% will be secondary urethane aryl-NH-CO-O-CHMe-. Primary hydroxyl groups react 4 times faster than secondary ones. The dangling ends of the polyethers will accumulate secondary OH groups.

Polyurethanes with indices below 100 are covalently adaptable networks: excess hydroxyl groups will lead to “vitrimeric” behavior through transesterification of the carbamate group. Excess isocyanate may react with urea to biuret or with urethane bonds to allophanate bonds. If this is not possible because of the mobility of the polymer they will eventually react with ambient moisture to amino groups that can act as aldehyde scavengers or can be easily oxidized to colored quinoid structures. Based on visual inspection of the samples, there was very little oxidation at the time of the experiments (Figure 5).

2.4. Chemical Compounds

The solvents used were water, deionized by Sartorius arium pro, Acetonitrile LiChroSolv Reagent Grade, and Tetrahydrofuran HiPerSolv ChromaNorm.

Commercial sources and synthetic pathways for the analytes investigated in this study are available in [18] (C₁- to C₃-aldehydes) and [17] (compounds analyzed by thermal desorption gas chromatography mass spectrometry (TD-GC-MS)).

2.5. Sample Chamber and Sample Preparation

A polytetrafluoroethylene (PTFE) block measuring 70 mm by 70 mm by 220 mm underwent milling processes to carve out a cavity resembling the shape of a sarcophagus. Subsequently, a second PTFE block was milled to fabricate a fitting lid. The resulting closed sampling chamber features a cavity of approximately 500 mL accommodating foam samples measuring 50 mm by 50 mm by 200 mm. To facilitate sealing, a ¼ inch diameter hole was drilled into each of the two short ends of the chamber, into which sockets were inserted to allow for the insertion of sealing gaskets composed of fluoroelastomer material. A total of five chambers of this design were produced for this experiment.

Sample foams were created by cutting rectangular cuboids from the center of the provided bucket foams using a ceramic serrated knife (

Table 2. Sample index and masses used in the two different investigations.

Investigation	Index	Sample Mass (g)
Long-term	70	27.5
Long-term	115	27.3
Kinetics	70	35.6
Kinetics	85	34.3
Kinetics	90	32.6
Kinetics	100	27.3
Kinetics	115	29.9

). Sample dimensions were chosen to be 10% wider than the inner dimensions of the sample chamber so that a tight fit of the samples was achieved to ensure plug flow through the sample rather than gas flowing around the sample.

2.6. Thermo-Oxidation Methodology

The experimental setup for investigating foam sample conditioning and aging comprises a PTFE chamber with internal dimensions of 50 mm by 50 mm by 200 mm, a gas supply line equipped with mass flow control, an oven for precise temperature regulation, a gas purification cartridge, and a sampling protocol enabling time-dependent analyses. PTFE was selected to minimize surface interactions between the chamber walls and analytes. The gas supply is employed to purge samples continuously with purified pressurized air.

This continual replacement of the atmosphere within the foam induces a shift in the adsorption equilibrium of VOCs toward the gaseous phase. The removal of adsorbed analytes diminishes surface concentration, prompting an augmentation in analyte diffusion toward the polymer surface. Owing to the chamber’s design (Figure 6) and the cellular structure of the sample (Figure 7), a plug flow is established, ensuring a uniform gas velocity throughout the sample’s cross-section [76].

Plug flow precludes back-mixing of emitted substances within the sample, thereby enhancing sample purging efficiency. The VOCs emitted from the sample are convectively transported outside the oven and sampled on dinitrophenylhydrazine (DNPH) cartridges attached to the emission testing chamber exhaust port.

A laboratory oven, Binder Model FD 115, was used to maintain precise temperature control throughout the experiments (Figure 8).

Flow control was managed utilizing a Buerkert (Ingelfingen, Germany) single-phase primary switched power supply, coupled with a CM22-0-10 V potentiometer by COBI ELECTRONIC (TME Germany GmbH, Leipzig, Germany), and Mass Flow Controllers of type 8741, also by Buerkert (Ingelfingen, Germany). The gas supplies underwent filtration via an Agilent gas purifier cartridge, BIG HYDROCARBON TRAP Model BHT-4, effectively reducing hydrocarbon levels to below 15 ppb. PTFE tubing was used for supplying gas flow to the sample chambers and directing emissions from within the oven to the exterior for sampling purposes.

2.7. High-Pressure Liquid Chromatography (HPLC)

A VWR Elite LaChrom with an L-2300 column oven, L-2200 autosampler, L-2130 solvent pump, and L-2455 DAD detector (Darmstadt, Germany) was used to analyze hydrazone derivatives of emitted carbonyls. A 10 mm RP18e column guard and two Chromolith Performance RP18e 100 mm (Merck KGaA, Darmstadt, Germany) were used as analytical columns.

A blank value measurement of chambers not filled with foam was measured, and no background values of formaldehyde, acetaldehyde, propanal, or acrolein were found.

The analytical method and calibrations were conducted as described in [18].

2.8. Thermal Desorption Gas Chromatography Mass Spectrometry (TD-GC-MS)

The thermal desorption unit (PerkinElmer TurboMatrix 350 ATD, Rodgau, Germany) was set to −30 °C for cryofocusing with a valve temperature of 250 °C, a temperature of 280 °C for the ten-minute tube desorption, and a transfer line temperature of 200 °C. The column flow was set to 1 mL/min. The inlet and outlet splits were set so that 5% of the total sample mass was injected into the GC system.

Following the measurement, the used tubes were conditioned using a TC-20 by Markes International, Offenbach am Main, Germany. The tubes were heated to 280 °C for 2 h in a constant nitrogen flow.

A GCMS-QP2020 by Shimadzu Deutschland GmbH, Duisburg, Germany, with a PerkinElmer TurboMatrix350 (Rodgau, Germany) was employed for the analysis of the samples. A Shimadzu SH-Rtx-200MS (Duisburg, Germany) crossbond trifluoropropyl-methylpolysiloxane GC column with an inner diameter of 0.25 mm and a length of 30 m was used for the chromatographic separation of the analytes.

The oven temperature was held at 30 °C for 5 min and then heated to 120 °C with a heat rate of 2 °C min⁻¹, followed by a heat rate of 5 °C min⁻¹ to the final temperature of 240 °C. The total program time was 74 min.

The ion source temperature was set to 200 °C, and the interface temperature was held at 250 °C. The MS started at 1.7 min with a scan speed of 10,000 and an event time of 70 ms, and it scanned the mass range from 19 to 500 m/z.

Qualitative analysis of the VOCs was based on reference standards or proceeded through comparison of mass spectra with NIST 05, NIST 05s, NIST08, and NIST08s databases using Shimadzu’s LabSolutions GCMS solution version 4.45.

Calibrations were conducted as described in [17].

2.9. Initial Unloading and Hydroperoxide Unloading

To investigate the initial loading of a sample with ketones, the initial loading of unreleased carbonyls (hydroperoxides), and the long-term oxidation reaction at a constant temperature, two foams were cut and subjected to thermal non-oxidative and thermal oxidative treatment. The initial loading of the samples was investigated by flushing the foams with nitrogen at room temperature and sampling the complete nitrogen exhaust with DNPH cartridges over 24 h. Then, the cartridges were replaced with new cartridges, and the foams were heated to 120 °C while maintaining nitrogen flow. This heating was performed to degrade all hydroperoxides in the polymer and transform them into carbonyl compounds for emission and sampling. After 24 h, the oven was set to 65 °C, and after the temperature equilibrated, the cartridges were replaced again, and the carrier gas was changed to air.

2.10. Long-Term Investigations

The samples used in the initial unloading and hydroperoxide unloading experiment were then investigated over three months in an air stream at 65 °C. Sampling was conducted over 24 h each time. Sample taking ended on the following days: 1, 2, 3, 4, 5, 10, 12, 17, 20, 25, 32, 40, 46, 54, 60, 88.

2.11. Index- and Temperature-Dependent Emissions

New samples with five different mixing ratios of polyether polyol and polyisocyanate were purged at 65 °C for two days in a continuous nitrogen flow of 200 mL/min. The emissions on both days were tested by 24 h sample collection with DNPH cartridges. On the third day, nitrogen was replaced with air, and sampling was started for the kinetic study.

After each sampling duration, the oven’s temperature setting was increased by 15 K, and sampling did not begin until the end of a four-hour waiting period to allow thermal equilibration of the whole setup. The sampling times were decreased to account for increased formation rates (

Table 3. Sampling parameters used in the HPLC kinetics study.

Temperature (°C)	Sampling Duration for HPLC Analysis (Min/h)	Sampling Duration for TD-GC-MS Analysis (Min/h)
65	1475/24.58	60/1
80	1248/20.8	60/1
95	300/5	20/0.33
110	60/1	30/0.5
125	46/0.76	20/0.33
140	20/0.33	5/0.08
155	10/0.16	0.5/0.008

). Emission rates were calculated based on the number of analytes released after thermal equilibration till the end of the individual sampling duration and are therefore only representative of transitory initial oxidation rates at the respective temperature.

2.12. Arrhenius Graphs of Emission Rates

For the analysis of the DNPH derivatives, the integrated peak areas were used to calculate extract concentrations. These were multiplied with the extraction volume to calculate the whole mass of the formed hydrazones. This mass was then used to calculate the molar amount of the hydrazones, which is equal to the molar amount of the derivatized aldehydes. These molar amounts were divided by the sampling time and the sample mass to generate mass-dependent emission rates. These temperature-dependent emission rates were used to plot Arrhenius graphs. The logarithm of the calculated emission rate was plotted over the inverse temperature in 1/K.

Analysis of the TD-GC-MS analysis was conducted analogously without the volumetric concentration calculation.

2.13. Index Dependence of the Kinetics of Oxidative Degradation

During the investigations, some measurements were disturbed by DNPH depletion. Undisturbed values were selected for the calculation of activation energies and pre-exponential factors. Aldehyde emission values observed with DNPH depletion can only be below the “true” value and were only used for regression when correcting the linear regression line towards higher values and not lower values.

The TD-GC-MS results are not prone to such issues and were therefore used as measured.

2.14. Generative AI Statement

Generative artificial intelligence has been used in this paper to generate Python 3.12.10 scripts for data visualization.

3. Results

While this study is subject to certain limitations, including a low sample number and a non-ideal derivatization scheme, these factors were recognized and addressed as comprehensively as possible within the constraints of the study’s methodology. These limitations may affect the exact results of these kinetic investigations; however, careful analysis was employed to minimize their impact and ensure the robustness of our conclusions.

The reproducibility of VOC measurements has been thoroughly examined in previous publications using TD-GC-MS [17] and HPLC [18] methodologies. Based on the findings from those studies, differences in sample emissions observed in this study were only considered and discussed if they exceeded the expected measurement variability or confidence band associated with the applied techniques.

3.1. Results of Initial Loading and Hydroperoxide Loading

The initial loading of polyurethane samples interferes with the analysis of oxidation processes. Therefore, in the first twenty-four-hour sampling, the foams were purged of volatile physically adsorbed and absorbed analytes. In this first step most aldehydes chemically absorbed in the form of hydroperoxides remained within the polymer backbone due to their stability at room temperature. In the second 24 h sampling the hydroperoxide depot is decomposed at 120 °C under nitrogen. The emission rates based on new oxidation shown for the 24 h sampling period are measured under oxidizing conditions in air at 65 °C. Normative tests only observe the adsorption–desorption equilibrium situation that corresponds to the total of the first 48 h and focus on the depots in the foam that largely depend on the foam’s history. The second period of 120 °C will also lead to equilibrium microphase separation if this had not been achieved after foaming.

Day 1 represents the physically absorbed and adsorbed aldehydes, day 2 the hydroperoxide decomposition. The new oxidation is started on day 3 and the rates obtained are predominantly compound formation rates from new oxidation.

On day 1, purging the foam with N₂ at ambient temperature, only formaldehyde and propanal depots are identified (Figure 9). On day 2, purging with N₂ at 120 °C, the relative composition is very different. Propanal is decreased from 13% of the total for all four aldehydes (day 1) to 3% on day 3. The ratio of the three other aldehydes changes remarkedly.

• The relative composition of initial depot (day 1) is independent of the index (72:19:9 (8:2:1) for formaldehyde:acetaldehyde:acrolein) but the total is 50% higher for the high index. The purging is nearly quantitative for formaldehyde but not for the other three aldehydes (Figure 9 below).
• On day 2 the hydroperoxide depot breaks down mainly into acetaldehyde: the ratio is 4:67:29 (1:17:7) if there is an excess of OH groups and 5:81:14 (1:16:3) for an excess of NCO groups. The emissions of formaldehyde and acetaldehyde hardly depend on the index, but a high index suppresses the formation of acrolein by 62%.
• Once oxygen is offered formaldehyde is formed again and the ratio changes to 16:55:29 (2:7:4) and 24:59:16 (3:7:2). Again, the high index disfavors acrolein, this time at the expense of formaldehyde.

Emissions of physically adsorbed and absorbed formaldehyde are roughly half of the hydroperoxide-related amount (Figure 9). The oxidative formation rates of formaldehyde (day 3) are three times as high in the sample with 42% hard segment and no free OH groups compared to the one with 20% free OH groups.

There is hardly any acetaldehyde adsorbed and absorbed in the foam compared to the amounts released by the decomposition of the hydroperoxide depot (Figure 9). Acetaldehyde does not form a hydrate comparable to formaldehyde. The amount of acetaldehyde formed in the first 24 h of oxidation at 65 °C is less than a quarter of the amount released from the hydroperoxide depot: hydroperoxides will form on day 3 but at 65 °C the decomposition is much slower than at 120 °C. The index does not cause a difference in the hydroperoxide decomposition at 120 °C on day 2 but the high-index foam formed at 65 °C under air yields twice as much acetaldehyde as the low-index foam.

Most of the acrolein is likely formed through hydroperoxides on the allyl end groups. The low-index foam releases nearly three times as much acrolein as the high-index one during hydroperoxide decomposition on day 2 (Figure 9). During day 3 the higher-index foam emits 20% more acrolein than the lower-index foam.

Propanal emissions show a similar pattern to acetaldehyde but lower by one order of magnitude (Figure 9). Again, emissions under oxidative conditions correlate with the hard-segment content.

3.2. Results of Long-Term Investigations

After the initial three-day investigations, the same foams were monitored for three months at 65 °C in a constant air stream, testing emissions on selected days with a sampling time of 24 h each.

The graph illustrates the emission rates of formaldehyde (in mol/kg·s) from two polyurethane foams characterized by different indices, 70 and 115, over three months (Figure 10). The aldehydes stem from the alkylenoxide AO and formaldehyde additionally from the dimethylamino groups of the catalysts. We also referenced the data to this basis. All commercial polyethers contain ~1 mmol/kg Irganox 1076 antioxidant as a basis. Some of the initial concentration will have been exhausted at the time of the testing. It was not possible to quantify the residual active part before the trial.

• The first three days are shown in Figure 9 but not included in Figure 10. They reflect the foams’ initial loading rather than their long-term behavior which Figure 10 focusses on.
• All emissions peak after 17–20 days (induction time for hydroperoxide formation and depletion of residual phenolic antioxidant) and decline from there. The index dependence of formaldehyde emissions vanished after 30 days. Acetaldehyde emissions show index dependence over the full time of three months. Emissions of acrolein and propionic aldehyde do not show index dependence after the peak.
• Foam with index 115 demonstrates a sharp initial increase in formaldehyde emission rate, peaking at around 1.75 × 10⁻¹⁰ mol CH₂O/kg·s (14 pmol/mol_AO·s) on day 17, before experiencing a notable decline to stabilization near 1 × 10⁻¹⁰ mol/kg·s (8 pmol/mol_AO·s) after 30 days.
• Foam with index 70 exhibits a more gradual increase initially, reaches a maximum rate below 0.92 × 10⁻¹⁰ mol/kg·s (9 pmol/mol_AO·s) on day 20, followed by a gradual decline to around 0.75 × 10⁻¹⁰ mol/kg·s (7 pmol/mol_AO·s). Over time, the emission rate from both foams shows convergence towards a similar value of 0.65 × 10⁻¹⁰ mol/kg·s (6 and 5 pmol/mol_AO·s).

Acetaldehyde emissions from high- and low-index foams decline in parallel and do not converge.

• High-index foam starts with emissions of 6.5 × 10⁻¹¹ mol/kg·s (6 pmol/mol_AO·s). As Figure 10 shows, this may still be related to some residual depot. These decline over the first 30 days, stabilizing near 3 × 10⁻¹¹ mol/kg·s (3 pmol/mol_AO·s). During the decline, the emission rate does show a local maximum on day 17.
• Low-index foam starts with an emission rate of 2.7 × 10⁻¹¹ mol/kg·s (2 pmol/mol_AO·s). The rate increases to ~3.7 × 10⁻¹¹ mol/kg·s (3 pmol/mol_AO·s) and gradually decreases to below 0.7 × 10⁻¹¹ mol/kg·s (0.6 pmol/mol_AO·s).

For acrolein, both foams’ emission rates initially decline steeply, then increase to a maximum of 2 × 10⁻¹¹ mol/kg·s (2 pmol/mol_AO·s). The low-index foam reaches the maximum on day 17, three days before the high-index foam. Emissions converge and decline to 10% of the initial values.

Propanal emissions increase to a maximum of approximately 5 × 10⁻¹² mol/kg·s (0.4–0.5 pmol/mol_AO·s) on day 17 and then steadily decrease below 1 × 10⁻¹² mol/kg·s.

Figure 11 shows a comparison of the emissions of all four aldehydes in a plot with a log-scaled emission axis. The higher-index foam always emits higher amounts of aldehydes. The data indicate close similarity and convergence in emissions patterns with the exception of acetaldehyde.

3.3. Results of the Index- and Temperature-Dependent Emissions

To obtain data on emissions from compounds formed by autoxidation, rather than those adsorbed in the polymer matrix before treatment, all foams were heated to 65 °C in a continuous nitrogen stream (200 mL/min). Notably, all aldehyde emissions decreased from the first 24 h sampling to the second. For this experiment, samples were not purged at 120 °C in nitrogen to avoid thermal degradation from occurring before oxidative treatment. However, this does not guarantee full degradation of hydroperoxides before running the tests. Therefore, a second purging was conducted to evaluate the amount of potentially remaining hydroperoxides.

Following the purging, the nitrogen carrier gas was replaced with air, and a third 24 h sampling period commenced. The quantity of analytes formed oxidatively during this period was approximately equivalent to the initial loading and the degradation products of existing hydroperoxides. Due to the amount of oxidatively formed compounds not being a loading, again, a secondary axis giving the emission rate was introduced to the data plots.

The purging seems to be close to quantitative within 24 h for formaldehyde at 65 °C (Figure 12). It looks like the foam of index 85 holds the highest amount of adsorbed and absorbed formaldehyde while indices 70 and 115 show the lowest amounts. Hydroperoxide decomposition also does either seem to be quantitative within 24 h at 65 °C or not to be occurring to an observable degree. Oxidative degradation releases as much or more formaldehyde than the purging steps did.

Acetaldehyde emissions decrease strongly within the second 24 h. During oxidative treatment, the analyte emissions are higher than during the prior two days. The lowest emission is observed for the sample of index 90. Emissions increased in the order of 90, 85, 100, 70, 115.

Propanal behaves in nearly the same manner even though emissions are roughly only 10% as high as the acetaldehyde emissions (Figure 12). The index dependence is nearly the same as for acetaldehyde. Propanal is also reformed at 65 °C under oxidative conditions. The current opinion, that propanal is generated by hydrolysis of propenyl end groups, needs to be supplemented with some additional mechanism.

The emissions of acrolein do not seem to show index dependence when purging or at 65 °C under oxidative degradation conditions. During oxidation, the emissions double compared to the second day of purging but do not reach the original day 1 levels.

After 24 h of sampling at 65 °C, the oven temperature was increased to 80 °C, followed by a four-hour thermal equilibration to ensure uniform temperature distribution. This was followed by a 21 h sampling period. Subsequently, a similar thermal equilibration was conducted at 95 °C, leading to a 5 h sampling period. Additional temperature increments were applied: 1 h at 110 °C, 46 min at 125 °C, 20 min at 140 °C, and 10 min at 155 °C, all with four-hour equilibration periods and continuous airflow.

In Figure 13, all results of the kinetic study are shown. Those measurements marked with a red upward arrow show DNPH depletion. Residual DNPH is required to ensure quantitative hydrazone derivative formation of the aldehydes and carbonyls, therefore, results displayed with an upward arrow underquantify the respective analyte.

The formaldehyde emission rates at elevated temperatures increase with increasing index (Figure 13). There is a pronounced drop in emissions between 140 °C and the 155 °C measurements, probably through consecutive or parallel reactions following the compounds’ formation.

In general emissions from oxidation increase with the index starting at 110 °C. There are exceptions in acetaldehyde and propanal explained by the DNPH depletion that occurred during sampling. Formation of formaldehyde and acrolein hydrazones at the expense of acetaldehyde and propanal hydrazones has been observed before. The hypothesis of a hydrazone metathesis has been formulated before [18]. Propanal and acetaldehyde emissions show the same trend over index, with the same exceptions at 125 °C and 140 °C. The emission rates of acrolein correlate with a higher index at temperatures of 125 °C and above. If acrolein hydrazone is formed through metathesis on the DNPH cartridges, high values are observed even in DNPH-depleted cartridges.

3.4. Emission Rates Show Arrhenius Behavior

Arrhenius plots show the natural logarithm of the compound molar emission rates over the inverse temperature. Figure 14 shows linear behavior of the logarithm of the emission rates over the full temperature range for acetaldehyde and propanal. In the case of formaldehyde and acrolein, the linear range correlates with the index. The highest compound emission values are reached by the foam samples with the highest hard-segment content. Data points deviating from Arrhenius-like behavior were excluded from the linear fitting as marked by the data points’ edge color. Values determined on cartridges with depleted DNPH are marked with triangles. The calculated emission value in these cases can only be equal or below the “real” value. Depleted measurements that would change the Arrhenius plot to lower slopes were therefore excluded from the linear fit. Only those points with black edges were included in the fitting.

Logarithmic formaldehyde emissions increase linearly with 1/T for index 70, index 85, and index 90 up to 110 °C. For index 100 the linearity extends to 125 °C and for index 115 up to 140 °C. Minimum emissions are found at the lowest index, i.e., the maximum excess of NCO-reactive groups. The emissions at the higher end of the temperature range increase with the index. At low temperatures, there is no obvious order.

Again, the pattern of acrolein emissions resembles the one of formaldehyde. They also gradually deviate from the Arrhenius-like behavior with a lower index already at lower temperatures. Again, emissions of higher-index foams are generally higher than for lower-index foams. The correlation is most prominent at high temperatures while not observed at lower temperatures.

Aldehydes are only part of the degradation process. They are the products of chain-scission reactions occurring during autoxidation that also led to a multitude of other VOCs. These include diols, formates, acetates, dioxanes, dioxolanes, ethers, carbonyls, amines, and amides. All these are not residues from production of the raw materials or production process but are formed in the oxidation process. In Figure 15 the Arrhenius plots of ethanediol, its mono- and diformate, and of 1,4-dioxane for all five sample foams are shown. All Arrhenius plots are linear in the investigated range.

Figure 16 shows the plots for 1,2-propanediol, hydroxyacetone, for hydroxyacetoneacetate and 1,2-propanediol-1-formate. Again, most show linear behavior. Just like ethanediol, 1,2-propanediol deviates most from linearity. For hydroxyacetone, hydroxyacetoneacetate, and 1,2-propanediol-1-formate a clear correlation of emissions and index is visible.

Figure 17 gives the Arrhenius plots of 1,2-propanediol-2-formate, propenyloxypropanol, 1,2-propanediol-1-acetate, and 1,2-propanediol-1-acetate-2-formate. All show the index-dependent trend. Of these four, emissions of 1,2-propanediol-1-acetate-2-formate are highest. The spread of emissions between low and high indices is also the highest for this analyte.

Figure 18 plots emissions of 2,4-dimethyl-1,3-dioxolane, dimethylallylamine, pyridine and 4-methylmorpholine. For the dioxolane, just like the acetates and formates, they correlate with the index. This is not the case for nitrogen-containing compounds. Especially at higher temperatures the highest-index foam emitted the least of them. Dimethylallylamine was only observed at 125 °C, 140 °C and 155 °C—as expected for the Cope reaction of amine oxides. This illustrates that the amine catalysts are part of the oxidation processes in the foam.

Dimethylformamide is emitted most by the high-index foam (Figure 19). The situation is not as clear for dimethylacetamide and toluidine for which there is not a clear trend. Aniline might stem from the residual phenyl isocyanate which is bound in the hard segment and liberated only above 80 °C in the lowest-index foam. The foams with index 85 and 90 show aniline starting at 110 °C. The higher-index foams do not exhibit aniline emissions at all.

The emissions of benzoxazole and methylbenzoxazole are generally low. For benzoxazole there is a clear correlation of emissions with index (Figure 20). There is no clear trend for methylbenzoxazole.

3.5. Results of Index Dependence of Kinetics of Oxidative Degradation

Pre-exponential factors and activation energies are calculated from the linear ranges of the Arrhenius graphs (Figure 21 and Figure 22). The analytes from left to right are ordered in the following categories: low-molecular-weight aldehydes, scission products of the EO segment, scission products of the PO segment, nitrogen-containing compounds and aromatic nitrogen-containing compounds. Activation energies and pre-exponential factors increase with index in most cases. Emissions of 1,2-propanediol, dimethylallylamine, pyridine, 4-methylmorpholine, dimethylformamide and methylbenzoxazole deviate from this behavior.

The activation energies increase from propanal over formaldehyde and acetaldehyde to acrolein. The same order holds for the pre-exponential factor, except for an outlier of index 100 in acrolein (9.7) that is below the value of index 100 in acetaldehyde (10.6). This is attributed to the inclusion of values at a temperature at which rates for acrolein do not follow the linear correlation.

The calculated apparent activation energies shown in Figure 21 are shown in

Table 4. Compound Activation Energies E_A in kJ/mol of all investigated compounds for all foam samples and the linear dependency E_A = A × [OH]_calc + B using the calculated excess of NCO-reactive groups [OH] 1.24 mol/kg (index 70), 0.58 mol/kg (index 85), 0.39 mol/kg (index 90), and 0 for index 100. Most curve shapes indicate an even better fit if it accounted for the residual unreacted NCO-reactive groups, best with [OH]_index100 = 0.2 (equivalent to 5% unreacted NCO-reactive groups). The slope has been used to sort the table.

Analyte	Activation Energy (kJ/mol)
	Index 70	Index 85	Index 90	Index 100	Index 115	A	B	R2
EA with inverse correlation to [OH] (NCO-reactive groups look like inhibitors)
Aniline (linear fit excluding index 100 for lack of data)	97.6	144.8	146.2			−60.6	174.3	0.96
1,2-Propanediol-1-acetate-2-formate	87.1	99.9	103.7	107.3	108.1	−16.7	108.8	0.98
Acetaldehyde	75.8	88.6	91.1	95.5	95.5	−16.1	96.7	0.98
1,2-Propanediol-2-monoformate	72.2	85.3	87.0	91.3	95.3	−15.6	92.6	0.97
Hydroxyacetone	73.1	83.6	85.9	88.5	91.5	−12.8	89.8	0.96
Hydroxyacetoneacetate	84.6	94.4	96.3	100.1	104.5	−12.6	100.8	0.99
Ethanediol diformate	62.2	70.5	70.9	76.0	79.9	−10.9	75.9	0.98
1,2-Propanediol-1-monoformate	69.6	77.9	79.9	82.7	88.1	−10.8	83.5	0.98
1,2-Propanediol-1-acetate	78.5	87.5	87.8	90.3	92.9	−9.6	91.4	0.94
Propanal	68.3	75.0	77.4	78.8	78.4	−8.8	79.7	0.96
Formaldehyde (in brackets after omitting Index 100)	72.9	75.0	78.0	83.0	88.8	−7.9 (−13.6)	81.6 (83.1)	0.87 (1.00)
Dimethylformamide (maximum at NCO/OH = 1)	84.3	90.2	91.1	93.0	88.6	−7.1	93.6	0.97
Dimethylacetamide	42.0	43.4	43.8	50.9	58.6	−6.4	48.6	0.69
Propenyloxypropanol	70.7	73.3	74.4	75.3	78.3	−3.8	75.5	0.98
Toluidine	54.7	56.8	57.6	58.6	59.8	−3.2	58.7	1.00
EA with correlation to [OH]
Pyridine (minimum at NCO/OH = 1)	94.0	80.9	77.1	64.1	66.9	+23.4	66.1	0.98
4-Methylmorpholine	115.7	98.2	96.4	90.4	89.1	+20.6	88.8	0.97
EA does not correlate to [OH]
N,N-Dimethylallylamine	77.8	78.2	78.7	73.5	71.6
Benzoxazole	80.9	80.1	91.3	93.0	97.1
Methylbenzoxazole (regression without index 70)	73.6	75.4	73.8	69.8	72.1	+9.7	69.9	1.00
Acrolein (regression without index 100)	72.0	95.2	98.6	95.1	102.9	32.2	112.3	0.99
1,4-Dioxane (regression without index 100)	92.0	103.0	107.8	97.9	102.6	−18.0	114.2	0.99
2,4-Dimethyl-1,3-dioxolane (regression without index 100)	93.9	100.9	103.4	105.3	107.7	11.0	107.5	1.00
1,2-Propanediol	78.3	78.8	73.5	82.3	81.6
Ethanediol	79.2	82.0	86.7	81.1	81.5
Ethanediol monoformate (regression without index 70)	118.6	106.0	112.4	130.3	113.0	−42.4	129.9	0.99

including their linear dependency on the excess of NCO-reactive groups.

The calculated pre-exponential factors shown in Figure 22 are listed in

Table 5. Pre-Exponential Factor in (ln(mol/(kg·s))) of all investigated compounds for all foam samples. The linear dependency y = A’ × [OH]_calc+ B’ is using the calculated excess of NCO-reactive groups [OH] 1.24 mol/kg (index 70), 0.58 mol/kg (index 85), 0.39 mol/kg (index 90), and 0 for index 100. Most curve shapes indicate an even better fit if it accounted for the residual unreacted NCO-reactive groups, best with [OH]_index100 = 0.2 (equivalent to 5% unreacted NCO-reactive groups).

Analyte	Pre-Exponential Factor (ln (mol/(kg·s)))
	Index 70	Index 85	Index 90	Index 100	Index 115	A’	B’	R2
Inverse correlation to [OH]
Acetaldehyde	3.89	8.15	8.95	10.55	10.63	−5.4	10.9	0.98
Hydroxyacetoneacetate	4.01	7.6	8.26	9.82	11.51	−4.7	10.0	0.99
Hydroxyacetone	1.82	5.36	6.11	7.07	8.03	−4.3	7.5	0.97
1,2-Propanediol-2-monoformate	−3.13	1.7	2.27	3.89	5.35	−4.3	7.5	0.97
Propenyloxypropanol	−1.54	−0.4	0.05	0.45	1.39	−4.3	7.5	0.97
1,2-Propanediol-1-acetate	2.33	5.55	5.65	6.65	7.59	−4.3	7.5	0.97
1,2-Propanediol-1-acetate-2-formate	5.43	9.99	11.22	12.79	13.57	−4.3	7.5	0.97
Ethanediol diformate	−3.82	−0.81	−0.62	1.23	2.56	−4.0	1.2	0.99
1,2-Propanediol-1-monoformate	−2.74	0.05	0.97	2.12	4.0	−4.0	2.3	0.99
2,4-Dimethyl-1,3-dioxolan	3.83	6.53	7.35	8.35	9.49	−3.7	8.6	0.99
Formaldehyde	2.58	4.42	5.37	7.21	8.96	−3.7	6.9	0.98
Propanal	−0.9	1.43	2.22	2.87	2.94	−3.1	3.1	0.98
Dimethylformamide	4.6	6.84	7.11	7.78	6.36	−2.6	8.0	0.96
Toluidine	−8.1	−7.38	−7.1	−6.7	−6.44	−1.1	−6.7	1.00
Correlation to [OH]
Pyridine	3.96	0.35	−0.84	−4.72	−4.04	6.8	−4.1	0.97
4-Methylmorpholine	14.2	8.94	8.5	6.61	6.17	6.2	6.2	0.97
Partial inverse correlation to [OH]
Ethanediol monoformate (linear fit excluding index 70)	13.89	10.76	12.82	18.39	13.22	−13.3	18.3	1.00
Aniline (linear fit excluding index 100 for lack of data)	5.91	18.42	18.62			−15.9	26.1	0.96
Acrolein (linear fit excluding index 100)	1.41	9.61	10.63	9.45	11.93	−11.2	15.5	0.99
1,4-Dioxane (linear fit excluding index 100)	3.17	6.55	8.01	5.14	6.87	−5.5	10.0	0.99
Partial correlation to [OH]
N,N-Dimethylallylamine (linear fit excluding index 70)	2.08	2.1	2.23	0.58	−0.06	2.8	0.7	0.85
No reliable correlation to [OH]
Ethanediol	2.67	3.15	5.0	3.29	3.14
1,2-Propanediol	2.79	2.82	0.82	4.16	3.77
Dimethylacetamide	−10.03	−9.21	−9.11	−6.88	−4.98	−2.3	−7.5	0.79
Benzoxazole	−0.33	−0.16	3.01	3.7	5.21	−3.4	3.5	0.72
Methylbenzoxazole	−3.65	−2.95	−3.5	−4.73	−4.04

including their linear dependency on the excess of NCO-reactive groups.

4. Discussion

The autoxidation of polyurethanes yields numerous analytes, and not all formation pathways are understood. Previously, we have shown various possible pathways for three of the four aldehydes [18]. Acrolein can be released by oxidation of allyl groups in the polyol or by aldol condensation and polyurethanes are known to act as organocatalysts for carbonyl reactions [77].

A potential mechanism for the formation of propanal is given in Figure 23.

In normative VOC analysis without prior purging, propanal emissions are explained by the isomerization of propylene oxide used in polyether polyol synthesis. During autoxidation under dry conditions, there is no such straightforward explanation.

One potential mechanism involves the formation of a terminal radical methylene oxide during chain scission. This methylene oxide would then be eliminated from the backbone as formaldehyde and the resulting methine radical could be saturated by an intramolecular six-ring transition state hydrogen abstraction. The so-formed propyl ether macroradical can then split off a propyloxy radical that then eliminates a hydrogen radical to form the propanal (Figure 23). Instead of the hydrogen abstraction from the methyl group, the same mechanism is possible by abstraction from the methylene group. The macroradical would form a vinyl-ether instead of the allyl-ether then.

The observed formates and acetates are explained by the chain-scission reactions during hydroperoxide degradation.

The degradation of polyethylene oxide segments of the soft segment leads to the formation of formaldehyde, formates and primary alcohols. While in theory possible, we did not observe the formation of terminal aldehydes (Figure 24).

Due to the asymmetry of the repeating unit of the polypropylene oxide segments, the autoxidative degradation mechanisms are more complex. Hydroperoxide formation and degradation follow the same steps as in polyethylene oxide, however, in PPO, both secondary and tertiary hydroperoxides can be formed and the degradation products of C-C or C-O cleavages are not the same in either instance. The degradation of secondary hydroperoxides in PPO leads to the formation of terminal primary and secondary alcohols, acetaldehyde, formaldehyde and formates. Again, while in theory possible, the terminal aldehydes formed through C-O cleavage of the original macroradical has not been observed (Figure 25).

The degradation of tertiary hydroperoxides in PPO leads to the formation of terminal primary alcohols and acetates. In contrast to the unobserved terminal aldehydes, a ketone that is formed through an analogous mechanism is one of the major degradation products observed (hydroxy acetone acetate) (Figure 26).

The mechanisms shown in Figure 24, Figure 25 and Figure 26 are not exhaustive. More detailed descriptions of possible autoxidative degradation mechanisms of polyether polyols are provided in [18,78,79,80,81,82,83].

The formation of low-molecular-weight species can be explained by repeated chain-scission reactions or intramolecular initiation reactions leading to the formation of hydroperoxides in consecutive repeating units of the polyethers.

The emission of dimethylallylamine can be explained by oxidation of the amine catalyst and subsequent Cope elimination [84] (Figure 27).

This mechanism is supported by the observation of N,N-dimethylhydroxylamine in the TD-GC-MS analysis. The observation is important because the amine catalysts turn out to be hydroperoxide scavengers. This is a reaction pathway for hydroperoxides in PUR not available in pure polyethers (Figure 28).

The formation of dimethylformamide and -acetamide seems to be a reaction product of the hydroperoxidation of the methylene groups in the dimethylpropylamine moiety of the catalyst (Figure 29).

Aniline and toluidine are degradation products of the hard segment of the polyurethane. Aminophenols are known contaminants in the production of MDI and could be precursors to benzoxazole and methylbenzoxazole.

At this point we cannot provide a mechanism for the formation of pyridine and 4-methylmorpholine, the only two remaining products with a clear correlation of activation energy and NCO-reactive groups.

4.1. Discussion of Initial Unloading and Hydroperoxide Unloading

The results of the initial unloading, hydroperoxide unloading, and oxidation at 65 °C experiment have implications for the interpretation of conventional VOC analysis results. When these emissions are not tested separately, their respective source cannot reliably be identified. Therefore, the VOC analysis of PUR samples is not reflective of the material’s properties but of the specimen’s history, the combined former aging, hydroperoxide loading and oxidation during testing.

Carbonyls show starkly different behavior. Formaldehyde is purged in comparatively high amounts at room temperature due to a substantial loading in the sample. Emissions of acetaldehyde, acrolein and propanal are nearly absent at room temperature. Formaldehyde adheres to the foam matrix, and the other aldehydes readily evaporate off between production and testing. These aldehydes are mainly released when hydroperoxide depots decompose.

The difference between formaldehyde and the other aldehydes may be explained in part by the reactivity of formaldehyde.

• Formaldehyde can reversibly form hydrates that adsorb to the hydrogen bond network of the hard segment or the hydrophilic EO blocks.
• Aromatic amine groups are formed in the foaming process and most react with isocyanate to urea hard segments. Approximately 10% of the isocyanate functional groups within the MDI oligomer are in central positions. They are largely inaccessible for reaction with polyol, particularly in later phases of the polymerization. However, reactions with water to amines are possible with a conversion to amines. These will react with aldehydes to imines. Formaldehyde should be the most reactive [85].
• The reaction of urea with formaldehyde to methylol urea is well-known and industrially used. The reaction is less efficient with other aldehydes.

Formaldehyde from raw materials or released during decomposition of hydroperoxides in the raw materials during polymerization may be reversibly stored as hydrate, imine or methylol group in the hard segment (Figure 30 and Figure 31).

Other aldehydes (such as acetaldehyde, propanal) may not be adsorbed as efficiently and if formed may accumulate in the skin of the foam due to the heat gradient between the hot foam core and the colder mold surface. The skin was trimmed and discarded and only the core of the foams was used for the experiments.

Compared to the amount of hydroperoxides decomposing into acetaldehyde, the initial acetaldehyde depot is small. This can be attributed to the formation of stable hydroperoxides within the polypropylene oxide soft segment or to low adsorption and absorption of acetaldehyde. A comparable behavior is observed for acrolein whose emission during hydroperoxide degradation is lower and the difference between lower and higher indices is more prominent. Propanal is also mostly emitted during hydroperoxide degradation. Adsorbed and autoxidatively formed propanal is emitted in comparable amounts.

4.2. Discussion of Long-Term Investigations

All aldehyde emissions decline over time. There is a one-week initiation phase followed by a rapid increase after which a quasi-steady state is reached, which slowly diminishes. These results agree with the general autoxidative scheme: an induction period follow by an auto-accelerating reaction leads to a drastic increase in radical concentration over time. With increased radical concentration, chain termination reactions become more likely and counteract chain initiation reactions. These two mechanisms determine the all-over radical concentration and maintain the steady state of autoxidation and emissions.

High-index foam releases more formaldehyde and acetaldehyde than the low-index foam. The ratio of formaldehyde to acetaldehyde in low-index foam increases throughout the experiment while the high-index foam rises to a three to one ratio that is then maintained (Figure 32).

This is an interesting result as it is unexplained what mechanisms lead to a continuously increasing ratio of formaldehyde to acetaldehyde over three months only in the case of the low-index foam. Polypropylene oxide (PPO) forms both acetaldehyde and formaldehyde during its oxidative degradation while polyethylene oxide (PEO) only forms formaldehyde. There is 30–40 mmol of dimethylamino groups per kg of polymer that needs to be considered as a source of formaldehyde and as sacrificial scavenger for hydroperoxides. Hydroxylamines above 100 °C may add to the complexity.

The secondary CH group in PPO is much more susceptible to oxygen attack than the primary one in PEO. The assumption is that the oxidative degradation starts in the PPO segment or on the secondary end groups. The secondary OH groups accumulate with incomplete conversion. The radicals in the PPO segment propagate intramolecularly before they start attacking the PEO segment later which causes a shift in emission ratio. The increased degradation of the PEO segment increases the ratio of formaldehyde to acetaldehyde over time.

In the higher-index foam, degradation of the two different polyether segments is not as time-dependent. Potentially, the increased hard-segment content is decisive in the initiation reaction and initiates reactions equally intermolecularly in both polyether segments. Therefore, radicals formed in the PPO segment are selectively oxidizing the PPO segment, while radicals formed in the hard segment are unselective in oxidizing the total soft segment (Figure 33).

Alternatively, the formation of further acetaldehyde is inhibited in the higher-index foam as the substrate is depleted quicker due to lower abundance of polyether polyols and higher degradation rates.

The integrated total emissions of these four aldehydes over 90 days was approximately 1 mmol/kg. This is ~0.01% of all ether bonds. Even over such a long exposure, we expect a constant excess of alkylene oxide units without a significant change in ether and end group concentrations. However, the ratio of emitted aldehyde to chain scissions is unknown and VOC emission might only reflect a fraction of all chain-scission events.

The long-term oxidation of the two foams leads to approximately 1 mmol of aldehydes per kg of foam being released within three months. The foams used in these investigations contain more than 10.000 mmol of alkylene oxides per kg. The sample substrate can be assumed to remain mostly unchanged. However, there is a change in the molar ratio of emissions observed for the index 70 foam. An explanation for the change in emission ratio without contemplating the influence of a change in substrate is challenging.

These results are important to discuss the results of the temperature-dependent emission investigations. It seems like running the investigations on the samples in intervals while ramping up the temperature in between each measurement allows one to investigate a polymer with the smallest possible changes in the substrate composition.

Further, the long-term investigations show a factor of approximately two between the highest and the lowest formaldehyde and acetaldehyde emission rates after the initiation phase. While this time dependence is a considerable factor when investigating a foam at one set temperature, the impact on the kinetics investigation is small as a factor of two hardly influences the logarithmic scale of the Arrhenius plots.

4.3. Discussion of Index- and Temperature-Dependent Emissions

The substantial variation in emission rates of the generated carbonyl compounds presents a challenge in determining an optimal sampling duration for temperature and foam samples. By leveraging the now-characterized temperature-dependent emission behavior, a more precise sampling duration can be identified for each sample in every temperature condition.

All investigated foam indices showed a fully reticulated cell structure (Figure 1). Figure 2 and Figure 3 show that the polymer strut geometry is highly comparable between the samples. There is a deviation in cell size for the lowest-index foam, however, the cell size should not have a major effect on polymer chemistry.

Polyurethanes show increased oxygen permeability with increased soft-segment content [86]. As the observed oxidation rates here do not increase with increased soft-segment content, it is assumed that oxygen permeation of the polymer struts is not limiting the oxidation rate. Therefore, we assume diffusion-limited oxidation (DLO) to be negligible as it would lead to decreased oxidation of high-index foams and increased oxidation of low-index foams.

The lowest-index foam sample emitted the least amounts of formaldehyde compared to the other indices over the whole temperature range. It also shows the strongest deviation from linearity in the Arrhenius plot. It is noticeable that this is the only foam that shows markedly more acetaldehyde than formaldehyde emissions over the whole temperature range. At temperatures above 110 °C there is no strong increase in formaldehyde emissions observed. The higher-index foams show a wider temperature range for linear behavior. The highest-index foam shows linear behavior up to 140 °C. Due to DNPH depletion, it is unclear whether the linearity extends to the study’s full temperature range.

At increased temperatures, the highest-index foams are generally emitting the highest number of aldehydes. This result is surprising as aldehydes are formed from the polyether polyol segment of the polyurethane which is at lower concentrations in high-index foams. In contrast to this, a higher concentration of polyether polyols leads to higher reaction rates in liquid systems.

There are several processes to consider:

(1) 

Inhibition of autoxidation through hydroxyl groups

Polyurethane soft foams of low indices exhibit an excess of hydroxyl groups in the polyurethane synthesis. These excess hydroxyl groups remain in the polymer matrix after synthesis and participate in the complex reaction scheme involved in polyurethane autoxidation. Increased concentrations of hydroxyl groups in polyether polyols, such as polyethylene glycol and polypropylene glycols, are known to decrease the rate of autoxidation through hydroperoxide stabilization [10].

(2) 

Increased hydroxyl concentration decreasing kinetic chain length

During autoxidation, hydroperoxyl radicals can abstract the alpha-hydrogens of hydroxy groups that were in excess in the synthesis of a low-index foam. Alpha-hydroxyl peroxyl radicals are then formed after oxygen addition, which readily form HO₂ radicals. These radicals react with the chain-carrying peroxyl radicals and form oxygen and hydroperoxides [87], therefore terminating a chain reaction.

(3) 

Increased antioxidative activity of hard segment through aromatic amines

During synthesis, the lower-index foams react with an excess of hydroxyl groups and water molecules. This leads to the formation of terminal aromatic amines as the amines formed through hydrolysis do not find additional isocyanates to form urea groups. These aromatic amines can inhibit autoxidation through hydrogen donation to reactive peroxyl species.

(4) 

Higher concentration of chain initiation sites

MDI and its carbamates form hydroperoxides on the methylene groups between the aromatic rings of the molecule. Aromatic amines may be active as redox catalysts: the quinoid form is calculated to be a good oxidant for terminal alcohol groups. As higher indices lead to higher VOC formation rates, the initiation reaction may take place on the MDI’s methylene bridge. This leads to a higher number of radical chain initiation reactions and therefore increased autoxidative degradation. This would also explain the increased pre-exponential factors for higher-index foams.

(5) 

Low bond dissociation energy of MDI methylene group hydrogen

Commonly, in the autoxidation of polymers, the abstraction of hydrogen radicals by peroxyradicals is the rate-determining step. This process is promoted by carbon–hydrogen bonds with low bond dissociation energies. The bond dissociation energy of an ether’s alpha hydrogen is between 93 kcal/mol (tertiary carbon) and 96 kcal/mol (secondary carbon). The bond dissociation energy of a dibenzylic C-H bond is 82 kcal/mol and that of an allylic ether C-H bond is 74–78 kcal/mol. While the allylic ether C-H bond energy is the lowest in the PUR system, its impact can be assumed to be exhausted in the short term due to its low abundance. However, the increased concentration of dibenzylic C-H bonds through a higher hard-segment content might be an explanation for the increased autoxidative degradation and emission [88,89,90].

(6) 

Impact of the amino catalysts

Amino groups react with hydroperoxides to amine oxides. This reaction lowers the concentration of radicals in the autoxidation scheme. As lower-index foams carry a higher amine catalyst load, the effect might be more pronounced there.

There is further research required to identify the most influential mechanism impacting autoxidation in PURs of varying indices in real foams.

4.4. Discussion of Arrhenius Graphs of Emission Rates

As detailed in the long-term emissions section, the time-dependent variability of emission rates introduces a factor of approximately two between the highest and lowest values observed during the post-initiation phase. While this variability is significant when analyzing emissions at a single temperature over longer time scales, its impact on kinetics investigations remains small, as the logarithmic transformation used in constructing Arrhenius plots greatly diminishes the influence of such variations. This variability is an inherent consequence of the experimental approach chosen to minimize the effects of degradation. Specifically, we deliberately opted to measure emissions from a single foam sample as close to its fresh state as possible during sequential temperature increases. This was done to avoid relying on separate foam samples that would require weeks to months to fully reach steady-state behavior which would have introduced additional variability due to batch-to-batch differences or extended exposure to environmental conditions. Thus, the choice to measure a single foam at multiple temperatures in a time-limited sequence represents a trade-off between two experimental constraints, ensuring the results remain as representative and consistent as possible for reliable kinetic analysis.

The depletion of the DNPH in some of the measurements is inconvenient. However, in an earlier study, it was shown that DNPH consumption does not necessarily mean that no further hydrazones can be formed [18]. The hydrazone for formaldehyde and acrolein seems to keep forming by consuming the acetaldehyde hydrazone. Therefore, the emission rate data of formaldehyde and acrolein might be close to the actual rate, while other hydrazones might have been suppressed in those cases where DNPH was depleted.

The temperature dependency of the emission rate is highly important for the validity of accelerated aging tests. The data shows that tests at 65 °C (VDA276, ISO 12219-4 [91]), at 80 °C (VDA270), and at 90 °C (VDA278 VOC) are within the linear range of the Arrhenius plot. A temperature of 120 °C (VDA278 FOG) is in the linear range for acetaldehyde and propanal but not for formaldehyde and acrolein if there are still significant concentrations of hydroxyl groups present (index below 100).

For formaldehyde, the linearity within the Arrhenius plot correlates to the index. With increased index the linear range expands to 140 °C. For index 70 the range is limited to 110 °C. Lower-index foam allows for other reactions to take place compared to the higher-index foams. Those reactions would either immediately consume formaldehyde or hinder reaction pathways that lead to the formation of formaldehyde in low-index foams and at higher temperatures. Another option is that formaldehyde is oxidized but this should occur independent of the index and is therefore probably not a main driver of the differences in our observations.

4.5. Discussion of Index Dependence of Kinetics of Oxidative Degradation

As stated above, due to the homogeneity of the cell structure of the foam samples, we assume all differences in emission behavior to stem from chemical differences rather than cell structure differences.

Lower activation energies and pre-exponential factors mean that at low temperatures low-index foams release more aldehydes compared to high-index foams. At higher temperatures, high-index foams emit more VOCs. This could imply that formulations can be designed for specific testing temperatures. If VOC emission was correlated with polymer chain degradation, the same principle could be applied: low-index foams for use at higher temperatures and high-index foams for use at lower temperatures.

There are still free hydroxyl groups even at index 100. Linear regression of E_A and the pre-exponential factor indicate a proportion of 5 mol-% as, with increasing polymer viscosity, the reaction is diffusion-limited and molded parts cool down quickly and the polymer structure is kinetically frozen. At index 115 no free NCO-reactive groups should remain. Most of the excess isocyanate at index 115 is crosslinking the hard segment by biuret and allophanate bonds. If the formation of these groups is sterically not possible, isocyanates will react with ambient humidity to amines and CO₂. Allophanate and biuret bonds open reversibly to urethane and free isocyanate at temperatures of 80 °C to 140 °C [92]. In the case of formaldehyde and acrolein, this seems to play a role: the activation energy increases from index 100 to 115, while in the case of propanal and acetaldehyde, it does not.

In the range up to index 100, if there are free hydroxyl end groups, activation energy correlates with the calculated excess of NCO-reactive groups. Just the activation energy for acrolein seems to be independent of the index or excess NCO-reactive groups [OH/NCO-1]. The same pattern is observed for the pre-exponential factor A, which is correlated to the activation entropy in transition state theory. Again, the pre-exponential factor for acrolein is independent of the index and excess free hydroxyl groups (Figure 34).

The Arrhenius plots of the compounds analyzed by TD-GC-MS all show a positive correlation between emissions and index for all VOCs except for the amines dimethylallylamine, pyridine, 4-methylmorpholine, aniline and toluidine. All compounds formed from the autoxidation of the soft segment are increasingly emitted with increased hard-segment content or with decreasing content of free hydroxyl groups (Figure 15, Figure 16, Figure 17, Figure 18, Figure 19 and Figure 20).

The activation energies are shown in Table 4 and the pre-exponential factors in Table 5. The R² values of the linear regression of the activation energies and the pre-exponential factors are given in Figure 35: nearly all R² values are above 0.6 with the majority of values above 0.9. A correlation of the oxidative breakdown mechanism with the foam’s composition is apparent. The R² values approximate 1 when it is assumed that five percent of the NCO reactive groups have not reacted due to steric hinderance [93].

As the determination of ethanediol and 1,2-propanediol via TD-GC-MS analysis can be challenging, it is unclear whether or not the low R² values are caused by artifacts of the analysis or by chemical mechanisms independent of the index.

The compound activation energy comprises several steps such as the transition of oxygen from the gas phase to the polymer, the diffusion of oxygen in the polymer, different autoxidation reactions, different breakdown steps, the diffusion of the molecular fragments in the polymer and their transfer from the polymer into the gas phase. The compound activation energies E_A differ by a factor of 3. The mass transfer processes from materials (without asymmetric diffusion and without reaction) have been modeled before [94,95]. Despite the variety of processes involved, the linear correlation of both the activation energy and the pre-exponential factor with the content of NCO-reactive groups [OH] or with the hard-segment content is surprising and indicates that the rate-determining step is similar for many VOCs. This observation aligns well with the assumption that the hydrogen abstraction by peroxyl radicals is the rate-determining step in the autoxidation process. This process is then accelerated by a higher abundance of labile hydrogens in high-index foams.

4.6. Comparison to Industrial Results

In industrial settings, it is commonly observed that foams with a lower index tend to perform poorly in VOC analysis compared to their higher-index counterparts. Interestingly, this research indicates that emissions correlate positively with index.

To resolve this apparent contradiction, it is essential to consider the origins of the emissions in question. This study focused specifically on the autoxidation of the polyurethane matrix. All initial loading had been removed. Industrial analyses, such as those conducted using VDA 278, involve a composite assessment of emissions from contamination, oxidation products from polyether polyols, and oxidation products from polyurethane. Our data show that most emissions measured in industrial environments originate from the polyether polyols. Therefore, these VOCs are highly soluble in polyether (confirmed by calculations on a COSMO RS level) and the depot of VOCs in PUR foams correlates with the amount of polyether in the polymer. Formaldehyde may be a different case because it is more easily stored in the hard segment and the microphase separation may influence this storage. The initial concentration of propanal depends on the workup and therefore may vary from site to site and lot to lot; a multi-site–multi-lot comparison was beyond the scope of this study.

5. Conclusions

The new method for testing porous materials like PUR flexible foams shows insights into chemical processes eliminating the effects of the specimens’ history and the effects of physical transport in the material by using convective analyte transfer through the open-cell foam structure rather than relying on diffusion. This allows focusing on the oxidation chemistry in the polymer.

This research highlights the significant impact of sampling conditions and sample preparation on VOC analytical results, particularly due to depots of adsorbed, absorbed, chemisorbed, and oxidatively formed aldehydes (Figure 9 and Figure 12). These findings emphasize the need for clear handling specifications in emissions testing, as recommended by Society of Automotive Engineers (SAE) specifications J2989 and J3233 [96].

Over three months at 65 °C, the higher-index foam consistently emits higher amounts of formaldehyde, acetaldehyde, acrolein, and propionic aldehyde (Figure 10 and Figure 11). While formaldehyde and acetaldehyde reach a quasi-steady emission plateau after one month, acrolein and propionic aldehyde exhibit continuously decreasing emissions after an initial spike. Aldehyde emission behavior of polyurethane foam samples is not a static value but depends, in addition to the aforementioned factors, on the oxidative history of the sample.

Aldehyde emission rates increase over temperature and in two cases (acrolein and formaldehyde) consecutive and parallel reactions lead to decreasing emissions at temperatures from 110 °C to 140 °C (typical temperatures for the VDA278 FOG test). While VOC emissions primarily originate from the polyether soft segment, this study and previous work demonstrate that hard-segment concentration plays a critical role in influencing oxidative reactions and VOC emission rates. Although few VOCs stem directly from the hard-segment molecules, their relevance to oxidative degradation chemistry remains significant and warrants further investigation (Figure 13).

For eighteen VOCs the autoxidative degradation kinetics show linear Arrhenius plots (Figure 15, Figure 16, Figure 17, Figure 18, Figure 19 and Figure 20). Activation energies and pre-exponential factors correlate linearly with the foam index, the hard-segment content or the content of free hydroxyl groups (Figure 21 and Figure 22).

The kinetics of the autoxidation of polyurethane foams offers chemistry that is not limited to the autoxidative behavior of polyether polyols alone. Index-dependent activation energies and pre-exponential factors, time-dependency of different processes, adsorption and chemisorption and the oxidation of amine catalysts play a role. This work shows that, while VOC emissions stemming from molecules in the hard segment are low, it is very relevant for the chemistry of oxidative degradation and emissions of polyurethane although the mode of action is still open for discussion.

These findings underscore the importance of careful sample preparation and characterization, providing a basis for improving normative tests (e.g., VIAQ standards). By advancing the understanding of VOC emission and formation, this work paves the way for future research and product enhancement.

The findings of this study open several avenues for future research into polyurethane oxidation and VOC emissions. First, while the role of the hard segment in influencing oxidative degradation has been established, its specific mode of action and underlying mechanisms remain unclear and merit deeper investigation. Determining whether hard-segment interactions accelerate soft-segment oxidation through catalytic processes or other pathways could provide insights into formulation optimization. Additionally, the effects of different catalysts and additives, such as antioxidants, on VOC formation kinetics and emission mitigation require further study to develop tailored strategies for reducing long-term emissions. Beyond formaldehyde and the aldehydes analyzed in this work, future studies might extend long-term investigations to other hazardous VOCs or products of oxidative decomposition that are less commonly monitored. Moreover, expanding the experimental conditions, such as broader temperature ranges or humidity-controlled environments, could enhance the applicability of the findings to diverse real-world conditions. Finally, integrating computational modeling with experimental data could help predict VOC emissions from polyurethane materials under varying scenarios, aiding in the development of predictive tools for material design and regulatory compliance.