How volatile components catalyze vapor nucleation

See allHide authors and affiliations

Science Advances  13 Jan 2021:
Vol. 7, no. 3, eabd9954
DOI: 10.1126/sciadv.abd9954


Gas phase nucleation is a ubiquitous phenomenon in planetary atmospheres and technical processes, yet our understanding of it is far from complete. In particular, the enhancement of nucleation by the addition of a more volatile, weakly interacting gaseous species to a nucleating vapor has escaped molecular-level experimental investigation. Here, we use a specially designed experiment to directly measure the chemical composition and the concentration of nucleating clusters in various binary CO2-containing vapors. Our analysis suggests that CO2 essentially catalyzes nucleation of the low vapor pressure component through the formation of transient, hetero-molecular clusters and thus provides alternative pathways for nucleation to proceed more efficiently. This work opens up new avenues for the quantitative assessment of nucleation mechanisms involving transient species in multicomponent vapors.


Gas-phase nucleation is the very first step of a vapor’s transition to the condensed phase, taking a critical role in the formation of cloud condensation nuclei (1, 2), various industrial processes (3, 4), and stardust formation (5). It has been observed that the introduction of a second, high vapor pressure gas can markedly increase the nucleation rate of another vapor (6, 7). For instance, sulfuric acid is widely recognized as a major precursor in atmospheric new particle formation (NPF), but its unary nucleation rate is too low to explain the observed NPF rates (8). Realistic estimates of NPF rates must account for the enhancing effect of water, organic compounds, and ammonia/amines on sulfuric acid nucleation (1, 913).

Unravelling the mechanism of the enhancing effect requires the experimental determination of the molecular composition of the clusters initially formed in the nucleation process (nucleating clusters) and the quantitative tracing of their concentration. In general, this is a challenging task (14) because the nucleating clusters are delicate entities—even in mass spectrometric studies, they often escape detection as they are easily altered or even destructed upon ionization and during transport (15, 16). Recent developments in mass spectrometry made it possible to identify at least the relatively strongly bound constituents in the nucleating clusters (17, 18), such as sulfuric acid, amines, or highly oxidized molecules. While this has provided remarkable details about the initial steps of atmospheric NPF (1, 9, 19), the measurements remained blind to the species that exhibit weaker interactions (weakly interacting species) in the nucleating clusters. Because they easily evaporate, weakly bound constituents of nucleating clusters have so far eluded detection. Water is a notable example. Despite its tendency to form hydrogen bonds, it is mostly missing from the mass spectra recorded during NPF events, even though humidity is known to influence nucleation rates (1, 20, 21).

To understand the role of weakly interacting species in nucleation, experimental approaches are required that initiate nucleation under well-controlled conditions, preserve the integrity of the nucleating clusters, and quantify their number concentrations with high accuracy, to enable a data-based kinetic analysis. A new nucleation instrument recently developed in our group is designed to meet these requirements (2224): The postnozzle flow of a Laval expansion creates a nucleation region with uniform temperature and pressure (22, 25, 26), while coupling to mass spectrometry by ultrasoft, single-photon ionization in vacuo minimizes cluster fragmentation upon detection to the lowest level nowadays achievable (27, 28). This instrument is capable of quantifying the number concentrations of weakly bound nucleating clusters and providing time-dependent, molecular-level information on their size and chemical composition. This instrument has already revealed unprecedented molecular-level details on unary nucleation dynamics (2224, 2931), which predestines it for extended investigations of more complex binary/multicomponent nucleation events.

The present study reports molecular-level observations of the nucleation enhancement by weakly interacting species in binary vapors. With the important role of CO2 nucleation in the Martian atmosphere (3234) and the potential use of CO2 for carbon capture and storage (35) in mind, we have selected various model systems containing CO2 as the nucleation-enhancing species in binary vapors. On the basis of a detailed kinetic analysis, we quantitatively discuss the participation of CO2 in the cluster formation process, demonstrating that CO2 essentially functions as a catalyst to enhance nucleation. By covering a broad temperature range, we show that the ability of CO2 to enhance nucleation decreases with increasing temperature. The general implications of our current work are discussed in relation to previous nucleation studies.


Nucleation in the postnozzle flow of a Laval expansion

Figure 1 illustrates the experimental setup (22, 25). Nucleating vapors along with a carrier gas (argon, nitrogen, or mixtures) are supplied to the stagnation volume of a Laval nozzle [red shaded area (3638), at temperature T0 and pressure p0]. As the gas mixture expands through the nozzle, the temperature drops quickly, leading to supersaturation of the nucleating vapors. The unique features of our experiment are the initiation and observation of the nucleation in the uniform postnozzle flow [blue shaded area; this flow uniformity has previously been exploited in studies of chemical kinetics (3743)], and the online retrieval of molecular information about the composition of the molecular clusters. The postnozzle flow acts as a flow reactor without walls, allowing us to study nucleation under uniquely well-controlled conditions (constant velocity VF, flow temperature TF, flow pressure pF, and saturation ratio S; see Materials and Methods for details of flow temperature adjustment), a crucial prerequisite for nucleation experiments because nucleation is exquisitely sensitive to changes in conditions. The mass of the molecular clusters is probed with time-of-flight mass spectrometry directly in vacuo after soft ionization with a vacuum ultraviolet (VUV) laser at a photon energy of 13.8 eV (27, 28, 44, 45). In this way, the detection largely keeps the clusters intact (23, 24), so that their mass provides direct molecular-level information about their chemical composition and their size (number of molecules).

Fig. 1 Illustration of the experimental setup and the nucleation process.

The four inset plots show (from left to right) free vapor molecules in the stagnation volume, the onset of nucleation, a cluster after some growth, and an example mass spectrum of such grown clusters (right). The cluster ions are accelerated by the electrostatic lens and detected by a microchannel plate detector (MCP).

Different parts of the postnozzle flow can be probed by changing the nozzle exit–to–skimmer distance, LNS (Fig. 1). LNS can be modified in steps as small as 1 mm over a maximum length of ~10 cm, corresponding to steps of ~2 μs in the nucleation time t up to a maximum of ~200 μs. The onset of nucleation occurs at short LNS, i.e., short t, where the first molecular clusters appear in the mass spectra (“Nucleation,” Fig. 1). With increasing LNS and thus t, the nucleated clusters continue to grow by vapor condensation (“Growth,” Fig. 1).

Nucleation enhancement by a second vapor component

We start the discussion with an exemplary binary nucleation experiment carried out for a toluene-CO2 mixture at TF = 55 K. While the binary vapor undergoes nucleation and cluster growth (Fig. 2), neither toluene (toluene-only, red trace) nor CO2 can nucleate on their own at the same vapor concentration and conditions. Figure 2A exemplifies this for the mass spectra recorded at the longest nucleation time at LNS = 100 mm (t ~ 200 μs). Even at this point, when the binary mixture (yellow trace) has already nucleated and started to grow, no cluster signal shows up in the corresponding mass spectrum for pure toluene (toluene-only, red trace). Evidently, the presence of CO2, although more volatile than toluene, greatly enhances toluene nucleation.

Fig. 2 An example of nucleation enhancement of toluene by CO2.

(A) Mass spectra for binary toluene-CO2 nucleation as a function of LNS (equivalently, nucleation time t) recorded at TF = 55 K and pF = 40 Pa for a toluene concentration of 1.73 × 1013 cm−3 and a CO2 concentration of 5.55 × 1014 cm−3 (CT4 in table S1). The red mass spectrum is a reference spectrum measured at LNS = 100 mm for the same toluene concentration but without CO2 in the flow (“toluene-only”). (B and C) Zoomed-in segments of the mass spectra for binary nucleation recorded at LNS = 40 and 100 mm, respectively. m/z is the dimensionless mass-to-charge ratio. a.u., arbitrary units.

To understand the mechanism, we examine the composition of the very first clusters formed in the nucleation process of the binary mixture, which sets in at LNS ~30 mm (t ~ 60 μs). Figure 2B shows a zoomed-in segment of the mass spectrum recorded at LNS = 40 mm, i.e., right after the onset of nucleation. We only find cluster signals with mass-to-charge ratios (m/z) that are multiples of the singly ionized toluene molecules (m/z = 92). This means that the only detected nucleating clusters at TF = 55 K are homo-molecular toluene clusters [(Tol)j]. Given, however, the observed strong enhancement of nucleation by the presence of CO2, mixed CO2-containing clusters [(Tol)j(CO2)x>0] must be involved in the nucleation process. Such mixed clusters can indeed be detected at lower temperatures of 31 K (see the “Effect of temperature” section below). So why are they not observed at 55 K? One reason could be that they are losing CO2 (evaporation) on their way to the ionization point, implying a lifetime of <500 μs (the traveling time of the clusters from the skimmer to the ionization point). This could indicate that equilibration through collisions with the carrier gas molecules is too slow at the low flow pressure of our experiments, so that a certain amount of excess energy (i.e., exceeding the thermal energy) remains in the mixed clusters after their formation. The fact that whole series of (Tol)j(CO2)x>0 are detected at lower flow temperatures (see the “Effect of temperature” section below) indicates that such incomplete thermalization is not the only factor. Rather, these observations provide a hint that the mixed clusters play the role of transient species in the nucleation mechanism itself. This provides the basis for the kinetic scheme we propose in the “Kinetic modeling” section.

The above discussion refers to the initial nucleation step. This is followed later on by the growth of the clusters that were formed in the nucleation step. At the later stages of this growth process, CO2 starts to condense on the preformed clusters as evidenced by the mass spectra observed at longer times t. An example is shown in Fig. 2C for a zoomed-in segment of the mass spectrum at LNS = 100 mm (t ~ 200 μs). The m/z difference of 44 between neighboring cluster signals indicates that CO2 condensation is becoming a major driving force of cluster growth at this stage.

Kinetic modeling

To gain further insight into the role of CO2 in the nucleation process, we have performed experiments at the same constant TF = 55 K and pF = 40 Pa as in Fig. 2, but for varying toluene and CO2 concentrations (see conditions in table S1). The number concentrations Nj of clusters (Tol)j(CO2)x (referred to as “j-mers”) are determined as a function of LNS (see Materials and Methods). Conversion of LNS to nucleation times t yields the temporal evolution of Nj (Fig. 3A shows examples for j = 2 to 6). For j = 2 to 6, the mass spectra show predominantly homo-molecular clusters (x = 0), similar to Fig. 2A, while hetero-molecular clusters (x > 0) are not directly observable under these conditions (TF = 55 K).

Fig. 3 Kinetic analysis of binary toluene-CO2 nucleation at constant TF = 55 K and pF = 40 Pa.

(A) Concentration Nj of the j-mers (j = 2 to 6) as a function of the nucleation time t. The toluene and CO2 concentrations in the postnozzle flow are 1.73 × 1013 cm−3 and 5.55 × 1014 cm−3 (CT4 in table S1), respectively. (B) Enhancement factor ηexp, j as a function of the CO2 concentration NCO2 for j = 2 to 6. Dashed lines are quadratic fits to the data points (symbols). (C) Total number of clusters Ntot as a function of t for six different toluene-CO2 concentrations (CT0 to CT5; table S1). The legend shows the concentrations in units of 1013 cm−3. The dashed lines are linear fits to the experimental data points (symbols). (D) Circles: Comparison of the experimental nucleation rates Jexp and the modeled nucleation rates Jmodel (Eq. 2). The straight black line is a reference line with a slope of 1.

From the Nj, we derive experimental rate constants k1, j for the association of toluene monomer with the j-mers (section S2). The ratio to a corresponding gas-kinetic hard sphere collision rate constant, k1, j, hs (section S2), defines the enhancement factor ηexp, j (29)ηexp,j=k1,jk1,j,hs(1)

Its value indicates how much faster the association of a toluene monomer with a j-mer proceeds in the actual experiment (k1, j) compared with hard-sphere collisions [k1, j, hs, calculated within standard gas kinetic theory (46) assuming bulk properties (47)] between toluene monomer and homo-molecular toluene clusters (section S2). By definition, k1, j, hs is independent of the CO2 content, so that variations of ηexp, j with the CO2 concentration show how CO2 affects the toluene monomer-j-mer association. Figure 3B shows the resulting ηexp, j as a function of the CO2 concentration for j = 2 to 6. To observe unary toluene nucleation (no CO2) under the chosen conditions, we increased the toluene concentration 3.5 times compared with Fig. 2 (where no unary toluene nucleation was observed). ηexp, j = 2 − 6 varies between ~2 and 5 for unary toluene nucleation, in qualitative agreement with our previous unary nucleation studies on water, CO2, and propane (29, 31). This range of enhancement factor is likely the result of intermolecular interactions between the colliding toluene monomer and the homo-molecular toluene clusters (48), which are not accounted for by k1, j, hs. With increasing CO2 concentration, ηexp, j systematically increases for all j compared with unary nucleation, by up to a factor of ~4 for the highest CO2 concentration of 8.3 × 1014 cm−3 (Fig. 3B). The dependence of ηexp, j on the CO2 concentration indicates the direct involvement of CO2 in the formation of nucleating toluene clusters, suggesting that hetero-molecular (Tol)j(CO2)x>0 clusters are essential, albeit transient species in the nucleation mechanism. Under the conditions in Fig. 2, where unary toluene nucleation does not occur, transient hetero-molecular clusters must even dominate the nucleation process.

Our previous study revealed dimerization as the rate-limiting step for unary toluene nucleation at 55 K (23). On the basis of this result and the above observations for binary nucleation, we propose the following models (R1 to R3)Tol+Tolα(Tol)2(R1)CO2+Tolβ(Tol)(CO2)(R2)(Tol)(CO2)+(Tol)(CO2)xγ(Tol)2(CO2)y+(xy+1)CO2 with (x1,yx+1)(R3)where α, β, and γ are the rate constants of the respective reactions. R1 accounts for unary toluene nucleation and is ineffective at all our conditions, except for the experiment where we use a high toluene concentration and no CO2 (unary case in Fig. 3; CT0 in table S1). R2 and R3 account for CO2-toluene monomer interaction, postulating the formation of a transient cluster (Tol)(CO2) (R2) that precedes toluene dimerization (R3). The consecutive reactions R2 and R3 represent a CO2-catalyzed toluene dimerization process. Catalytic clustering processes of weakly interacting species were previously hypothesized in free jet expansion experiments, e.g., to rationalize the increased HF dimer population observed in the presence of N2O (49).

The proposed mechanism R1 to R3 is, to some extent, analogous to the chaperon or radical-complex mechanisms for radical recombination, where the direct recombination [analogous to (R1)] competes with a reaction of a radical monomer with a weakly bound preformed radical cluster (5053). The latter is analogous to R3 for x, y = 0 and is typically invoked for radical recombinations of small molecules at low temperatures and higher densities. In our experiments, by contrast, the pressure is extremely low (total pressure, 40 Pa; involved active species, <1 Pa; table S1); with toluene, we already have a relatively large molecule (42 internal degrees of freedom), and the products are only weakly bound. It is thus an intriguing result that such a mechanism turns out to be applicable to vapor nucleation. Our proposed mechanism differs in two aspects from the usual chaperon mechanism. We avoid the assumption of equilibrium for the formation of the chaperon cluster (R2), as equilibration might be too slow at the very low pressure conditions of our experiments. For simplicity, we describe this step with a single forward rate constant β, noting that this is only an effective rate constant. In addition to the usual chaperon mechanism, we also account for the reaction of two chaperon clusters (R3 for x > 0). While this contribution to the overall kinetics at 55 K is minor, it is necessary to account for the detection of a broad range of small CO2-containing mixed clusters at lower temperatures (31 K in Fig. 4B), where at least a notable fraction of toluene is expected to be present in the form of Tol-CO2 (see the “Effect of temperature” section; Fig. 4B).

Fig. 4 Temperature dependence of binary toluene-CO2 nucleation.

The mass spectra are recorded at flow temperatures TF of (A and B) 31 K, (C) 55 K, (D) 73 K, and (E) 84 K at a constant LNS = 100 mm and for constant toluene concentrations. (B) A zoomed-in segment of the mass spectrum in (A) for 0.27% CO2. The color code indicates the composition of the (Tol)j(CO2)x clusters. Detailed experimental conditions are specified in table S2. The indicated temperatures are the postnozzle flow temperatures without CO2 addition, rounded to the nearest integer.

In line with the experimental observations in Figs. 2 and 3, R3 is generally far more efficient than R1 in creating toluene dimers because extra CO2 molecules can accommodate the excess collision energy, either by evaporation or by providing additional degrees of freedom. R3 is therefore assumed to be collision-limited. R1, R2, and R3 are incorporated into a kinetic model (section S3) that is simultaneously fitted to the time-dependent experimental cluster size distributions for all conditions recorded (CT0 to CT5 in table S1). We use β as a free fitting parameter, while α is determined from the unary toluene experiment (CT0), and γ is assumed to be equal to the collision rate constants of the reactants (section S3). Because R3 is not rate determining (γ ≫ β), the model yields the following simple expression for the nucleation rateJmodel=αNTol2+βNTolNCO2(2)with α = 4 × 10−12 cm3 s−1 and β = 1 × 10−12 cm3 s−1 obtained from the fit to the experimental data (fig. S1). NTol and NCO2 are the concentrations of toluene and CO2 monomers, respectively. The first term arises from the bare dimerization, R1, while the second term accounts for the CO2-assisted dimerization, R2 and R3, which is first order in both the toluene and the CO2 concentration. The values of Jmodel are listed in table S1.Alternatively, experimental nucleation rates, Jexp, can be retrieved directly from the experimental spectra by taking the first-order derivative of total cluster concentration Ntot (Materials and Methods) with respect to the nucleation time (24, 54)Jexp=dNtotdt(3)

Figure 3C shows Ntot as a function of the nucleation time t. The dashed lines are linear fits whose slopes correspond to Jexp. The values of Jexp are listed in table S1. Figure 3D compares Jmodel and Jexp on a linear scale illustrating the excellent agreement between the model fit and the experimentally determined rates. Because Eq. 2 describes the nucleation kinetics quantitatively, our simple model of CO2-catalyzed nucleation (R1 to R3) appears to provide a reasonable representation of the underlying mechanism. In accordance with the simple two-parameter expression of Eq. 2, Jexp increases with increasing CO2 concentration at approximately constant toluene concentration. Jexp likewise increases with increasing toluene concentration at constant CO2 concentration. The proposed mechanism R1 to R3 seems to capture the essence of the nucleation process. More advanced direct simulations of the cluster formation processes (26, 55), which could provide deeper insight, would be very challenging for the toluene-CO2 system.

Effect of temperature

Figure 4 illustrates the influence of the temperature on binary nucleation in the temperature range TF = 31 to 84 K. Except for TF = 84 K, the toluene concentration was adjusted so that the onset of the unary toluene nucleation (0% CO2) does just not yet occur at LNS = 100 mm (bottom traces in Figs. 4, A to D). In the case of TF = 84 K, where no enhancement by CO2 is observed (see below), we adjusted the toluene concentration to obtain visible nucleation at LNS = 100 mm (bottom trace in Fig. 4E; see table S2 for the vapor concentrations) to better visualize the trends. The influence of increasing amounts of CO2 on the nucleation is shown in the middle and top traces in Fig. 4. It clearly shows that CO2 enhances toluene nucleation compared with the unary case at all temperatures, except at the highest temperature of 84 K.

As discussed in the context of Figs. 2 and 3, the vast majority of clusters detected in the mass spectra at 55 K at the onset of nucleation are stable homo-molecular (Tol)j clusters. Initially formed (nucleating) hetero-molecular clusters (Tol)j(CO2)x>0 dissipate part of their excess energy through the loss of CO2 (R3). Those clusters that do not lose all CO2 molecules directly after their formation (R3 for y > 0) are stabilized by collisions with the carrier gas. At the relatively low gas density (low pressure limit), the collisional deactivation is apparently not efficient enough to prevent complete loss of CO2 from the clusters on their way to the ionization region. The same holds for the nucleating clusters at the higher temperatures. This behavior changes at the lowest temperature of 31 K (Fig. 4, A and B). In this case, a substantial amount of hetero-molecular (Tol)j(CO2)x>0 clusters is observed in the region of the onset of nucleation (Fig. 4B). Two phenomena contribute to this effect. First, at lower temperature, the reduced collision energy means that less excess energy needs to be dissipated, and collisional deactivation by the carrier gas might also become more effective as the lifetime of the corresponding collision complexes increases. Second, at lower temperature, a larger fraction of toluene will be present in the form of Tol-CO2. With an estimated binding energy on the order of 10 kJ/mol (56), the equilibrium ratio of Tol-CO2 to Tol would change from a small fraction (less than a percent) at 55 K to a large excess of Tol-CO2 at 31 K (see section S4). Even if the equilibrium is not maintained at the very low flow pressure of our experiment, the probability of forming the chaperon cluster described by the effective forward rate constant β (R2) is expected to increase substantially. As a result, the reaction of two chaperon clusters and thus the survival of CO2 in hetero-molecular clusters becomes more likely (R3 with x = 1). By the same token, CO2 becomes less effective in promoting nucleation as the temperature increases from 31 to 73 K. This is evidenced in the experiment by the increasing CO2 concentrations required to enhance nucleation to a similar extent at rising temperatures (see Fig. 4, A, C, and D).

At 84 K (Fig. 4E), the enhancement effect by CO2 is lost altogether, so that the mass spectra hardly change anymore with increasing CO2 concentrations. At TF = 84 K, CO2 starts to behave like an inert carrier gas. Figure 4E demonstrates that both nucleation and growth become independent of the CO2 content. CO2 no longer enhances the nucleation of toluene (compare, e.g., 0% with 3.16% CO2), nor does it anymore contribute to growth by condensation onto the toluene clusters (compare, e.g., 0% with 10.53% CO2). This is in marked contrast to the behavior at the lower temperatures when CO2 not only enhances nucleation but also is mainly responsible for early cluster growth, as revealed by the mass spectra recorded at higher CO2 concentrations in Fig. 4 (A to D) (see also Fig. 2C).

CO2: A more general nucleation enhancer

Is the toluene-CO2 system a special case or does CO2 also act as a nucleation enhancer in other binary systems? To answer this question, we investigated various binary vapors containing CO2 and water, propanol, hexane, and butane, respectively. Figure 5 and fig. S2 show the mass spectra sampled at LNS = 100 mm for three different temperatures (55, 64, and 73 K) at a fixed CO2 percentage of 3.16% of the total flow. For clarity, the mass range m/z < 1000 is shown in Fig. 5, while the spectra for the larger cluster masses (m/z ≥ 1000) are displayed in fig. S2. Before CO2 was introduced into the flow, we adjusted the concentrations of the second compounds so that the onset for unary nucleation is just about to occur right at the chosen distance of LNS = 100 mm (not shown). The fact that the mass spectra after introduction of CO2 show pronounced cluster peaks (Fig. 5 and fig. S2) indicates that CO2 indeed enhances nucleation of all these rather different compounds. As found for toluene-CO2 binary nucleation, an increase in temperature diminishes the enhancement effect. The extent of the temperature effect varies in the different binary mixtures. Butane-CO2 and hexane-CO2 nucleation no longer proceeds at 64 and 73 K, respectively, while water-CO2 and propanol-CO2 nucleation still continues at these higher temperatures. The different intermolecular interactions between CO2 and the various second compounds provide a plausible explanation for this observation. Butane-CO2 and hexane-CO2 mostly interact through weak van der Waals forces, so that the probability of forming the transient hetero-molecular nucleating clusters (and with it the enhancement effect of CO2) decreases more rapidly with rising temperature than in the case of the more strongly bound water-CO2 and propanol-CO2 systems (both induction and van der Waals interactions). As the system with the strongest intermolecular interactions in the series, water-CO2 nucleation is expected to be the least temperature sensitive. The mass spectra in Fig. 5 confirm this expectation. They also reveal that most of the smaller clusters are homo-molecular (i.e., without CO2), which is consistent with the above observations for the binary toluene-CO2 system (Fig. 2): At temperatures at and above ~55 K, the nucleating hetero-molecular clusters responsible for the nucleation enhancement (R2 and R3) lose their CO2 on the time scale of the detection (see the previous section).

Fig. 5 Temperature dependence of the nucleation enhancement by CO2 in several binary vapors (see also fig. S2).

(A) Water-CO2, (B) propanol-CO2, (C) hexane-CO2, and (D) butane-CO2. The mass spectra are recorded at LNS = 100 mm, a fixed CO2 concentration of 3.16%, and a constant flow pressure of pF = 40 Pa. The flow temperatures TF are constant for each row, indicated on the right-hand side of the figure. The mass differences between neighboring peaks are indicated in spectra recorded at 55 K. They correspond to the molecular masses of water, propanol, hexane, and butane, respectively. Detailed experimental conditions are specified in table S3. The temperatures shown for each row are the postnozzle flow temperature without CO2 addition, rounded to the nearest integer.

The temperature trend shown in Fig. 5 is qualitatively consistent with previous binary nucleation studies for hexane-CO2 and water-CO2 in supersonic nozzles by Wyslouzil and co-workers (57, 58). Even though these studies did not provide molecular-level information on nucleating clusters, they show that the addition of CO2 to nucleating water or hexane vapors does not affect the onset of water nucleation at ~200 K or hexane nucleation in the temperature range of 127 to 148 K. This agrees with the trend we observe in Fig. 5: With rising temperature, hexane-CO2 and water-CO2 interactions become increasingly ineffective to initiate nucleation pathways such as R2 and R3.


With molecular-level details, we have shown that transient, hetero-molecular clusters provide the key to explain enhanced nucleation rates in CO2-containing binary vapors. Through the intermediate formation of those clusters, CO2 catalyzes the nucleation of the lower vapor pressure components. This marks an important step in unravelling the mechanisms of vapor nucleation because such weak interactions between a nucleation enhancer (CO2) and another vapor component previously escaped time-resolved, molecular-level observations. Our work also serves as an integral part in understanding the nucleation phenomenon across different temperature regimes: In atmospheric NPF studies, it is established that acid-base reactions and strong hydrogen bonds (6, 59) play an important role in promoting nucleation at ambient temperatures. The present study reveals that this role is taken by weak intermolecular interactions, such as van der Waals and dipole-induced dipole interactions, as temperature decreases. This result has further implications for the parametrization of nucleation rates for multicomponent vapors, as it indicates that with changing temperature, the type and number of species involved in the nucleation process also have to be adjusted properly. The temperature acts as a sensitive control for the nucleation mechanism, turning on and off different nucleation pathways involving different species (1). This could be particularly important in the modeling of atmospheric NPF events involving organics as different organic compounds likely enter the nucleation and particle growth processes at different temperatures (60), depending on their interaction strength with the other vapor components.


Temperature control

To adjust the flow temperature in coarse and fine steps, we use two methods based on gas dynamics. First, for a given carrier gas composition, the degree of gas expansion through the nozzle determines the postnozzle flow temperature. We can therefore change the flow temperature by using Laval nozzles with different physical dimensions. This method is used to change the flow temperature in large steps. Second, for a given Laval nozzle, the heat capacity ratio of the carrier gas influences the degree of cooling. In this study, we use mixtures of nitrogen and argon, which have different heat capacity ratios (1.40 and 1.67, respectively, at room temperature). Raising the nitrogen-to-argon ratio increases the postnozzle flow temperature, which we exploit to fine-tune the flow temperature by changing the carrier gas composition (24, 31). Tables S1 to S3 specify the gas compositions for experiments shown in Figs. 3 to 5, respectively.

To minimize the temperature disturbance of the postnozzle flow upon the addition of CO2, we simultaneously reduced the nitrogen content in the carrier gas by the same amount. Here, we make use of the fact that the heat capacity ratio of CO2 becomes very similar to that of N2 as the temperature drops below 100 K (61). In the most extreme case, i.e., 10% CO2 at 84 K (top trace in Fig. 4E), we estimate that replacing N2 by CO2 increases the temperature by ~1.5 K to TF = 85.1 K. Overall, for the traces presented in Figs. 4 and 5, the flow temperatures for varying CO2 content lie within ±1.1 K of the temperatures specified in the figures.

Determination of cluster number concentration

A reference gaseous species (internal standard) of known number concentration Ns and known photoionization cross section σs is used for the determination of the absolute number concentrations of the clusters (j-mers), Nj. The internal standard is needed to properly account for instrumental variations between experiments, e.g., laser power fluctuations. Nj is derived from the following equation used in our previous work (24, 31)Nj=ITolIsσsj·σTol·Ns(4)

ITol and Is are the ion signals of the (Tol)j (CO2)x clusters and the internal standard, respectively, in the mass spectrum; σTol is the molecular photoionization cross section of toluene. For the unary toluene nucleation experiments, methane is used as an internal standard. For the CO2-containing binary nucleation experiments, CO2 itself serves as the internal standard.

The total cluster concentration Ntot is calculated by the summation of all cluster concentrations from dimer to the largest cluster observed, i.e.Ntot=Σj=2Nj

This definition is consistent with our hypothesis that the formation of dimer is the rate-limiting step of the nucleation process in our experiments.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We are very grateful to E. Chasovskikh from our group for help with the maintenance of the experimental setup and to D. Stapfer and M. Steger from our workshops for technical support. Funding: Financial support was provided by the Swiss National Science Foundation (SNSF project number 200020-172472) and by ETH Zürich. Author contributions: C.L. and R.S. designed the research. C.L., J.K., and M.L. carried out the experiments. C.L., J.K., M.L., and R.S. analyzed the data and discussed the results. C.L. and R.S. wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article