Introduction

As a supplement to emission reduction, stratospheric aerosol injection (SAI) has been proposed as a climate intervention that could help to rapidly mitigate some of the most harmful effects of climate change1,2. SAI could potentially help meet the Paris agreement by keeping global warming below 1.5 K, until net-zero greenhouse gas (GHG) emissions are reached and GHG removal techniques are developed and scaled up3,4. The idea developed from observations of the cooling effect of explosive volcanic eruptions. Volcanic analogues are also the reason why research on SAI has so far focused on the injection of gaseous sulfur dioxide (SO2), a major precursor of aqueous sulfuric acid droplets in the stratosphere5,6,7. Due to their natural occurrence, the microphysics, optical properties, and heterogeneous chemistry of stratospheric sulfuric acid aerosols are relatively well-studied and understood5,6,8,9. However, there is evidence that sulfur-based SAI could come along with a number of serious side effects10,11,12,13,14. Some adverse SAI effects could potentially be alleviated by injecting materials such as alumina (Al2O3) or calcite (CaCO3) instead of sulfurous species15,16, as these solid particles absorb less terrestrial infrared radiation and scatter solar radiation better. The improved ability to scatter solar light implies that less stratospheric aerosol loading is needed15,17,18,19,20,21,22, which would reduce the SAI-induced enhancement of diffuse radiation, possibly reducing potential modulation of plant ecology and ecosystem productivity23,24. Most importantly, these materials could also lead to less aerosol-induced heating of the stratosphere15,19, which is known to considerably perturb global atmospheric dynamics and regional precipitation patterns11,12,13,14,25,26,27. Furthermore, the sizable stratospheric ozone loss in the southern hemisphere caused by sulfur-based SAI28,29,30,31 could potentially be mitigated by solid particles with lower surface area densities and chemically more inert properties, limiting the heterogeneous chemistry on their surfaces32,33,34.

Previous attempts to model solid particle SAI have been limited by simplified configurations that, for example, neglect the microphysical and chemical interactions between solid particles and stratospheric composition17,18,19,20 or did not account for interactions between particles, transport and radiation15,35. Here we employ the sectional aerosol-chemistry-climate model (ACCM) SOCOL-AERv2 (SOlar Climate Ozone Links-Atmospheric and Environmental Research Incorporation version 2)7 and expand it to interactively simulate the injection of solid particles, their microphysical interactions and heterogeneous chemistry, as well as their interactions with transport and radiation34 (see Methods). These processes were missing in previous studies and are crucial to realistically simulate the impacts of SAI via solid particles. Using experimentally determined parameters in combination with these novel and unprecedented modeling capabilities, our assessment of the uncertainties, risks and benefits of SAI of solid particles compared to sulfur-based SAI fills a gap highlighted in the World Meteorological Organization’s recent Scientific Assessment of Ozone Depletion: “Comprehensive climate model simulations to quantify these effects [of solid aerosols] have yet to be performed"10.

Results

We performed four simulations in which 5 Mt yr−1 (Megatons per year) of material is injected either as gaseous SO2 or aqueous sulfuric acid aerosol in the accumulation mode (AM − H2SO4, log-normally distributed with rm = 0.095 μm and σ = 1.5) or mono-disperse calcite or alumina particles with a radius of 240 nm, which is close to the optimal radius for backscattering solar radiation19. The injection, e.g., by high-flying airplanes, is assumed to be at 50 hPa (about 20 km), homogeneously distributed within all model grid boxes (i.e., about 325 km × 325 km × 1.5 km resolution) between 30∘N and 30∘S, from where the material is transported all over the globe by the atmospheric circulation.

SAI microphysics and heterogeneous chemistry

In the model, the aerosols undergo nucleation and condensation/evaporation, self-coagulation and coagulation with the natural stratospheric background aerosol, chemical processing and gravitational settling. Concentrations of GHGs and ozone depleting substances (ODS) and sea surface temperatures (SST) are assumed for the years 2020 or 2090 (see Methods).

The model calculates stratospheric particle size distributions and heterogeneous chemical reactions on the particle surfaces fully kinetically (see supplementary Notes 1 and Supplementary Fig. 1). On aqueous H2SO4 droplets the dominant reaction in the tropics and mid-latitudes is

$${{{{\rm{N}}}}}_{2}{{{{\rm{O}}}}}_{5}+{{{{\rm{H}}}}}_{2}{{{\rm{O}}}}{\to }^{{{{\rm{surf}}}}}2{{{{\rm{HNO}}}}}_{3}\,,$$
(1)

which leads to nitrogen deactivation, while in the cold regions of the polar winter vortices or close to the tropical tropopause, heterogeneous chlorine and bromine activation reactions are important36. All of these have major impacts on the ozone layer.

For alumina particles, which can accommodate aqueous sulfuric acid on their surface via condensation of gaseous H2SO4 and coagulation with aqueous H2SO4 droplets, we found a contact angle of about 30∘ 33, leaving part of their surface uncovered. While the heterogeneous chemistry on the H2SO4-covered parts is treated in the same way as on sulfuric acid aerosols, the most important reaction on bare aluminum oxide is

$${{{{\rm{ClONO}}}}}_{2}+{{{\rm{HCl}}}}{\to }^{{{{\rm{surf}}}}}{{{{\rm{Cl}}}}}_{2}+{{{{\rm{HNO}}}}}_{3}\,.$$
(2)

Most probably the uptake of HCl is dissociative, leading to H3O+ and Cl−, and is reduced by competitive HNO3 coadsorption. While this is plausible, it currently remains speculative and is associated with very large uncertainties that may have a critical impact on stratospheric ozone33.

For calcite particles the following heterogeneous reactions are important:

$${{{{\rm{CaCO}}}}}_{3}+2{{{\rm{HCl}}}}{\to }^{{{{\rm{surf}}}}}{{{{\rm{CaCl}}}}}_{2}+{{{{\rm{H}}}}}_{2}{{{\rm{O}}}}+{{{{\rm{CO}}}}}_{2}\,,$$
(3)
$${{{{\rm{CaCO}}}}}_{3}+2{{{{\rm{HNO}}}}}_{3}{\to }^{{{{\rm{surf}}}}}{{{\rm{Ca}}}}{({{{{\rm{NO}}}}}_{3})}_{2}+{{{{\rm{H}}}}}_{2}{{{\rm{O}}}}+{{{{\rm{CO}}}}}_{2}\,,$$
(4)
$${{{{\rm{CaCO}}}}}_{3}+{{{{\rm{H}}}}}_{2}{{{{\rm{SO}}}}}_{4}{\to }^{{{{\rm{surf}}}}}{{{{\rm{CaSO}}}}}_{4}+{{{{\rm{H}}}}}_{2}{{{\rm{O}}}}+{{{{\rm{CO}}}}}_{2}\,.$$
(5)

These reactions not only remove reservoir gases of ozone-destroying radicals, but also convert calcite into other calcium compounds with different radiative and chemical properties. The best estimates for the reactive uptake coefficients associated with Reactions (3–5) are currently \({\gamma }_{{{{\rm{HCl}}}}}=1{0}^{-5},{\gamma }_{{{{{\rm{HNO}}}}}_{3}}=1{0}^{-4}\), and \({\gamma }_{{{{{\rm{H}}}}}_{2}{{{{\rm{SO}}}}}_{4}}=1\)21,34,37. These are realistic but highly uncertain because they must take into account the dependence on the exposure of the particles to stratospheric HCl and HNO3 concentrations34.

Solid particles may reduce some SAI side effects

With SAI, the aim is to induce a negative radiative forcing (RF, in W m−2), reducing the imbalance created by anthropogenic GHG emissions. Therefore, the efficacy of different SAI materials and the reduction of their side effects should be compared with respect to the same negative net all sky top of the atmosphere (ToA) RF. Figure 1 shows the results of four 45-year time slice simulations for gaseous, liquid and solid materials introduced into the stratosphere and normalized to −1 W m−2 (see supplementary Notes 2, supplementary Table 1 and Supplementary Fig. 2 for non-normalized results). The comparison with SO2 injections clearly shows that, in order to offset the same amount of RF at ToA, using solids as injection material can result in reduced burden (Fig. 1b), stratospheric warming (c) and diffuse radiation (d). Per RF, solids might even treat ozone better, but the uncertainties are currently still prohibitive (e). However, these advantages come at a price: the solids require a higher mass injection rate (a) when injected at 240 nm radius, which would drive up the deployment costs. Yet, considering how high the cost of climate change38 will be in itself, these 50% higher deployment costs might be tolerable given that some environmental risks can be reduced compared to using SO2.

Fig. 1: Global mean quantities normalized by RF for 2020 conditions.
figure 1

Calculated injection rates and associated side effects to achieve the same global cooling for four different types of materials that could be used in climate intervention via stratospheric aerosol injection. Global means of (a) stratospheric mass injection rate, (b) stratospheric aerosol burden, (c) stratospheric heating at 50 hPa altitude, annually and zonally averaged between 30∘N and 30∘S, (d) diffuse radiation increase, and (e) change in total ozone column decrease, all normalized to −1 W m−2 total all sky ToA RF (see also supplementary Table 1 and Supplementary Fig. 2). Colored bars show mean results for each of the injected materials: red for gaseous SO2, orange for accumulation mode sulfuric acid aerosols (AM − H2SO4), dark green for alumina (r = 240 nm), and light green for calcite particles (r = 240 nm). Uncertainty bars show the standard deviation of inter-annual variability within the 45 years of time slice simulations. For ozone, solid bars show the results for the most reasonable uptake coefficients (see Section Stratospheric ozone chemistry), dashed bars the highest and lowest conceivable physicochemical sensitivity estimates (see Section Stratospheric ozone chemistry). All results refer to conditions of the year 2020 (see Methods). Quantities normalized to resulting stratospheric aerosol burden or injection rate instead of RF are shown in Supplementary Fig. 3 for 2020 conditions, whereas results for 2090 conditions are shown in Supplementary Fig. 4.

Full size image

In the case of SO2 injections, more than half of the resulting aerosol burden does not need to be lifted to the stratosphere, but is formed in-situ during S(IV) → S(VI) oxidation and water uptake in the stratosphere. This is partly the reason for the higher injection rates required for calcite and alumina particles compared to injection of SO2. The stratospheric residence times for the particles in the different scenarios are 0.76, 0.86, and 0.97 years for alumina, H2SO4 − H2O and calcite particles, respectively under 2020 conditions. The residence time is dependent on gravitational settling velocities, which is a function of the particle densities (i.e., 3.95 g/cm3, 1.6 g/cm3, and 2.71 g/cm3, respectively) and the particle radius, which is 240 nm for the solids, whereas the sulfuric acid aerosol have a effective radius of about 300 nm globally averaged at 50 hPa. Additionally, less self-lofting of solid particles due to reduced stratospheric heating16 contributes to a better injection efficiency per unit of RF for SO2 compared to calcite and alumina particles. The injection rates of sulfur species per RF could be further lowered by a factor of two if elemental sulfur could be injected and oxidized in-situ immediately after release39. However, other solid materials with better scattering properties injected at smaller radius such as diamond particles at 160 nm radius, were shown to require substantially smaller injection rates per RF compared to injecting SO222,40. The injection efficiency per RF of calcite and alumina particles could also be increased by extending the stratospheric residence time of the particles through injection of particles smaller than 240 nm instead22,34. However, using smaller particles would also lead to more efficient agglomeration through coagulation, which in turn reduces the backward to forward scattering ratio of visible light22,34. Moreover, smaller injected particle radii provide greater surface area densities (SAD) for heterogeneous reactions that damage the global ozone layer33.

The large uncertainties in ozone (dashed bars in Fig. 1e) are due to heterogeneous chemistry on solid particles, which is presently still poorly constrained under conditions of stratospheric trace gas concentrations, temperature and relative humidity21,33. The dashed bars in Fig. 1e represent the highest and lowest conceivable physicochemical sensitivity estimates of stratospheric ozone impacts (see Section Stratospheric ozone chemistry). Solid particle SAI would only lead to reduced disturbance of stratospheric ozone compared to injection of SO2 if the presently most reasonable estimates of heterogeneous uptake coefficients can be confirmed experimentally (solid bars in Fig. 1e).

Stratospheric warming and dynamical response

The major advantage of SAI using solid particles compared to sulfuric acid aerosols is reduced stratospheric heating due to less longwave absorption of alumina and especially calcite particles (Fig. 2a, c). For an injection of 5 Mt yr−1, SOCOL-AERv2 shows a warming of the tropical stratosphere, where most of the aerosol resides, of at most 0.8 K for calcite, but up to 2.5 K for SO2. Consequently, the effects of calcite particles on the Brewer-Dobson circulation (BDC) are dwarfed by the impacts of SO2 injections, where the age of air in the middle stratosphere is up to 3 months younger compared to the calcite experiment or the 2020 reference experiment (Fig. 2b). This indicates a faster overturning, and enhanced mixing of tropical and extra-tropical stratospheric air masses which is in agreement with findings of previous studies30. Increased tropical stratospheric heating has also been shown to have major implications for the stratosphere-troposphere coupling and tropospheric circulation11,31,41.

Fig. 2: Changes in stratospheric temperature and age of air caused by SO2 or calcite injections for 2020 boundary conditions.
figure 2

Left column: zonally averaged temperature anomaly compared to the reference run for SO2 (top row) or calcite (bottom row). Right column: corresponding anomaly of the age of air (AoA) in months. Dotted area: statistically significant within the 95% confidence interval (t-test). See Supplementary Fig. 5 for zonal mean wind, H2O and ozone anomalies for 2020 GHG and ODS conditions and Supplementary Fig. 6 for 2090 conditions.

Full size image

According to our model calculations, the aerosol heating would also lead to a warming of the tropical cold point tropopause, which would lead to an increase in stratospheric humidity42. Since H2O is an important GHG, increasing stratospheric H2O mixing ratios would not only alter stratospheric chemistry and composition via an enhanced stratospheric HOx ozone depletion cycle, but also alter the radiative balance of the stratosphere and troposphere, inducing a positive RF41. For injections of 5 Mt yr−1, we found an increase in H2O of 0.75 ppm (+18.0%) in the lower stratosphere for SO2. Most remarkably, alumina and calcite injections lead to a much smaller H2O increase of less than 0.12 ppm and 0.10 ppm, respectively (+3% and +2.5%, Supplementary Fig. 5).

Independent of the injected material, the effects of SAI also depend on the background conditions, which are largely influenced by anthropogenic forcings like GHG and ODS emissions; for example, changes in the general circulation can alter aerosol transport patterns43. This effect becomes clear when comparing age of air (AoA) of the 2090 reference simulation to the 2020 reference simulation, which decreases by about 20–30% (not shown). The decrease in AoA is the result of a faster BDC, stronger and more efficient mixing across latitudes44 as well as faster exchange between troposphere and stratosphere45 in a future warmer climate. The faster BDC effectively decreases the stratospheric residence time of the particles. Therefore, the same scenarios under 2090 conditions result in 5–20% smaller aerosol burdens (supplementary Table 1) compared to 2020 conditions.

Impacts on potential ecological drivers

Besides the effects associated with reduced tropospheric temperatures, potential ecological impacts from SAI can arise, among others, from changes in air humidity, increased UV irradiation, increased diffuse radiation and changes in precipitation patterns46. Changes in diffuse radiation, as manifested in the whitening of the skies47, can considerably affect ecosystems, crop yields and the net primary production rates23,48,49,50, although the effects depend on the specific plant species and ecosystem under consideration, and model formulation. While current studies generally suggest increases in plant productivity from increased diffuse radiation, capping changes in diffuse radiation would reduce the risk of unforeseen impacts of SAI on ecosystems. The increase in diffuse radiation is determined by particle size, refractive index and stratospheric burden. The former two are larger for solid particles than for H2SO4 − H2O, but the burden, whose influence dominates, is nearly halved. This leads to less diffuse radiation from solid particle SAI when normalized to the resulting RF, namely less than 6%/(W m−2), compared to up to 10%/(W m−2) for sulfur-based SAI (see Figs. 1d and 3a, c).

Fig. 3: Impacts on ecosystems through diffuse radiation and deposition changes of different SAI materials.
figure 3

Left column: changes in diffuse radiation (ultraviolet and visible, 250 nm–680 nm) normalized to ToA total sky RF for SO2 (a) and calcite (c) injection scenarios. Right column: the anomaly of the total sulfur-equivalent H2SO4 deposition flux of the SO2 injection scenario compared to the reference scenario (b) as well as the total deposition fluxes (sum of wet and dry deposition) of calcite particles at the surface (d). Dotted areas: statistically significant data in the 95% confidence interval (Student’s test), except for the deposition of calcite particles, for which no significance can be calculated. The scenarios with AM − H2SO4 and alumina injections (not shown) lead to similar deposition and scattered radiation patterns as SO2 and calcite, respectively.

Full size image

One matter of concern might be the interaction with ecosystems via deposition of solid particles at the surface. We argue that the ecological impacts of alumina and calcite deposition are expected to be small, since both materials are naturally abundant and natural dust emissions by orders of magnitude larger51. Similar to SO2 injections, the deposition patterns of solid particles when injected into the stratosphere will be different compared to natural dust injections into the troposphere (Fig. 3b, d)52. The calcite and alumina particle deposition fluxes follow the pattern of global precipitation patterns since more than 98% of the injected alumina and calcite particles are deposited via wet deposition (Fig. 3d). Near-ground concentrations of these particles are, thus, negligible (about 10−8 μg/m3) compared to the PM2.5 concentrations, especially in the densely populated regions53, but the ecological impact of seeding the deposition regions with these materials must still be carefully assessed.

Stratospheric ozone chemistry

An important measure of the environmental side effects of SAI is the ozone column. Changes in ozone have first-order impacts on UV exposure, especially over populated areas like tropics and sub-tropics, and are thus of great societal relevance. Past studies used a large set of assumptions concerning the chemical impacts of solid particles on ozone15,35. Further, they did not attempt to quantify the uncertainty arising from the existing spread in laboratory data. This particularly concerns the efficiency of the uptake of stratospheric acids on the solid particles surface. We provide explicit calculations of the ozone changes arising from such process uncertainty, as discussed next. Using the most reasonable estimates from current knowledge of heterogeneous chemistry on alumina and calcite particles, we simulate a smaller ozone column change compared to the sulfur-based scenarios (thick lines in Fig. 4). The key here is the large uncertainty arising from different assumptions in heterogeneous chemistry, which is highlighted by the shaded areas. The impact on stratospheric chemistry and related uncertainties would decrease sizably compared to present-day conditions, if SAI was done in the future (2090), due to decreasing stratospheric chlorine loading (Fig. 4b).

Fig. 4: Zonally averaged ozone column changes and related uncertainties for injections of 5 Mt of material per year.
figure 4

Injections of SO2 are shown in red, alumina (r = 240 nm) in blue and calcite (r = 240 nm) in green. Thick lines: most reasonable uptake coefficients (Section SAI microphysics and heterogeneous chemistry). Shaded areas show corresponding uncertainties (see Methods): in blue for alumina assuming full coverage by H2SO4 − H2O for the upper and non-dissociative HCl uptake without H2SO4/HNO3 co-adsorption for the lower limit33; in green for calcite assuming γHCl = 10−3 and \({\gamma }_{{{{{\rm{HNO}}}}}_{3}}=1{0}^{-5}\) for the upper and γHCl = 10−5 and \({\gamma }_{{{{{\rm{HNO}}}}}_{3}}=1{0}^{-3}\) for the lower limit; in red for SO2 assuming heterogeneous reaction rate coefficients to be higher or lower than recommended values36 by a factor of 2 for Reaction (1) and by a factor of 5 for the heterogeneous chlorine and bromine activation reactions. a For 2020 and (b) for 2090 boundary conditions. For underlying SADs see Supplementary Fig. 7.

Full size image

Since the interactions of calcite and alumina with stratospheric acids (i.e, HCl, HNO3, and H2SO4(aq)) at stratospheric temperatures (< 220 K), relative humidity (< 5%) and strong UV irradiation are so poorly known, we ran three simulations for each injection scenario (alumina, calcite and SO2), namely for the most reasonable interaction (solid bars in Fig. 1 and thick lines in Fig. 4), as well as for the upper and lower limit estimates of effects on  the ozone column (dashed bars in Fig. 1 and thin lines in Fig. 4).

For the injection of alumina particles, we extrapolate the measured uptake coefficients of ClONO2 in Reaction (2)54 to stratospheric partial pressures of HCl by using a Langmuir-Hinshelwood molecular description of co-adsorption/desorption and reaction33. The maximum ozone loss predicted by the model comes from simulations, in which we assume that HCl does not dissociate, resulting in \({\gamma }_{{{{{\rm{ClONO}}}}}_{2}}=0.019\) under stratospheric conditions. This assumption would cause a sizable globally averaged ozone column destruction of 9.1% for 5 Mt yr−1 injection of calcite particles or 14% per unit of RF. This is about double the historical peak ozone depletion recorded in the 90s due to emissions of ODS10. However, this is inherently uncertain as it is unclear to what extent HCl dissociates at the surface, and furthermore there are no measurements that account for the potential co-adsorption and dissociation of HNO3. The latter would likely decrease the surface coverage of HCl substantially33 and thus, lower the heterogeneous reaction rate of Reaction (2) and ozone depletion (Fig. 4).

For the injection of calcite particles, the uptake of HNO3 and HCl is limited by γ ≲ 10−3, because multiple monolayers of \({{{\rm{Ca}}}}{({{{{\rm{NO}}}}}_{3})}_{2}\) and CaCl2 are formed within a day, which increase the diffusion impedance for Reactions (4) and (3)37 (see supplementary Notes 3 and supplementary Table 2). Therefore, we performed a simulation with γHCl = 10−5 and \({\gamma }_{{{{{\rm{HNO}}}}}_{3}}=1{0}^{-3}\), which represents a lower limit for the ozone column with a mean global depletion of 3.9%, a possible scenario not considered in previous studies. The ozone loss is due to the removal of NOx from the stratosphere and thus a lower availability of NO2 for the deactivation of ClOx via NO2 + ClO + M → ClONO2 + M. Conversely, γHCl = 10−3 and \({\gamma }_{{{{{\rm{HNO}}}}}_{3}}=1{0}^{-5}\) represent an upper limit that leads to a recovery of the ozone hole21,35. Most strikingly, using these values leads to removal of chlorine from the stratosphere, resulting in an increase in the annually averaged ozone column of up to 25% in the south polar region when injecting 5 Mt yr−1 and 4% globally per unit of global RF. However, same as in the case of alumina injections, the assumptions of the extreme scenarios are less likely, as the uptake of HNO3 and HCl on calcite decreases significantly with time due to the sealing of the surface (whereas we assume that γHCl and \({\gamma }_{{{{{\rm{HNO}}}}}_{3}}\) remain constant). Furthermore, the uptake of the two acids is likely to be similar and the effects on ozone would cancel each other out.

For SO2 injections, lower and upper bounds for the ozone column response were estimated by multiplying or dividing the heterogeneous reaction rate of N2O5 hydrolysis (Reaction (1)) by a factor of 2 and heterogeneous reactions of chlorine and bromine activation by a factor of 5, based on uncertainty estimates of measured reaction rates at stratospheric temperatures and trace gas concentrations36,55.

In passing we note that the uncertainty ranges given for the injection of solid particles are based only on Reaction (1) for aluminum oxide and Reactions (3) and (4) for calcite. No experimental data are available for other reactions on these particle types. In addition, since calcite ages chemically in the stratosphere, the largest uncertainties result from changes in surface properties over time56. CaSO4, CaCl2 and \({{{\rm{Ca}}}}{({{{{\rm{NO}}}}}_{3})}_{2}\) as well as their hydrates could harbor different heterogeneous reactions with different reaction rates compared to calcite particles, whereby the heterogeneous chemistry on the calcite particles changes over time as a function of chemical aging. In addition, the heterogeneous chemistry depends on the available SAD and, thus, on particle size. Injecting smaller particles would sizably enhance the resulting effects on the ozone column, as we have recently shown33.

Changes in stratospheric ozone concentrations may also impact stratospheric temperatures and thus the circulation response and resulting RFs. The alumina injection scenario in which HCl does not dissociate (Fig. 1e) shows sizable ozone changes in the lower stratosphere, and a global ozone column depletion of 14%/Wm−2. These ozone changes would cool the lower stratosphere by 1.9 K/Wm−2, completely offsetting the warming of the alumina particles (Fig. 1c). This highlights the importance of interactively coupling heterogeneous chemistry with radiative effects of the solid particles when modeling the SAI response on climate.

Discussion

For the assessment of the risks and benefits of SAI of solid particles it is crucial to employ models with sufficient complexity informed by experimental data and interactivity between the essential model components, which was a limitation in previous studies15,19,35. This was achieved for the first time in this study by using an interactive ACCM incorporating a solid particle microphysics scheme34. We show that the main advantage of SAI by alumina and especially calcite particles is the sizably reduced heating of the stratosphere. This leads to much smaller stratospheric changes, such as acceleration of the BDC, of zonal winds or increasing stratospheric humidity (Supplementary Figs. 6 and 7), which could have important implications for the tropospheric circulation and regional climate11,12,13. It has been shown for SO2 injections that such dynamical side effects could be reduced by using more complex injection strategies31,57, while in our study we only considered tropical injections. Furthermore, the side effects on tropospheric climate and ecosystems are potentially reduced when using solid particles, as they induce smaller changes in diffuse radiation, ozone depletion and UV enhancement compared to SO2 injections, albeit with greater uncertainties.

Even though there is evidence that particle aging processes do not sizably affect the scattering properties of solid particles34, this is another factor which might introduce uncertainty to our results which requires further research. Despite the larger scattering cross sections of alumina and calcite particles compared to sulfuric acid, they require higher mass injection rates to achieve the same RF (see the overview provided in Fig. 5). This is primarily due to the in-situ oxidation of SO2, which results in the biggest part of sulfuric acid aerosol burden already being in the stratosphere. Under conditions for the year 2090 and assuming a high GHG emissions scenario, even larger alumina and calcite particle injections would be required to achieve the same cooling of −1 W m−2 compared to 2020 conditions (see Fig. 5). This might be related to various effects such as warmer surface temperatures, a faster BDC and larger water vapor concentrations in a future warmer climate. This is not as pronounced in the SO2 injection case which points to non-linearity in the response to different materials, which merits further investigation.

Fig. 5: Schematic overview of globally averaged quantities linearly normalized from 5 Mt yr−1 injections to −1 W m−2 total all sky ToA RF.
figure 5

Upper part shows results for 2020 and lower part for 2090 boundary conditions in respect to GHG, sea surface temperatures (SST), sea ice concentrations (SIC) and ODS concentrations when injecting gaseous SO2 (left), alumina (middle) and calcite (right) particles at radius of 240 nm. Shown are required stratospheric aerosol burden in Mt, the injection rate in Mt yr−1, the resulting global average stratospheric residence time of the particles in years, the resulting heating in the tropical stratosphere in K and the resulting ozone column in Dobson units (DU). For comparison the ozone column of the reference scenario, the historical minimum in the years 1992–1995 (modeled by SOCOL-AERv2) as well as the uncertainty ranges for the resulting ozone column (grey areas) are shown (see Section Stratospheric ozone chemistry for details on the uncertainty ranges). For 2090 conditions also the 2090 reference ozone column is shown.

Full size image

Our model calculations with alumina and calcite SAI, when performed on the basis of the current knowledge with the most reasonable assumptions for heterogeneous reaction rates on these materials, exhibit lower ozone depletion rates compared to sulfur-based scenarios. However, due to very limited experimental data on heterogeneous chemistry on solid particles under stratospheric conditions, the modeled ozone response is largely dependent on the underlying assumptions. Therefore, the uncertainties regarding the modeled ozone are considerable, so that both sizable ozone depletion or, in the case of calcite particles, an increase in ozone cannot be entirely ruled out (see Fig. 5). While the impact of sulfur-based SAI on stratospheric chemistry appears to be subject to less uncertainty and therefore be considered potentially safer, SAI via solid particles has potential for substantially reduced side effects in the Earth System. Their greater uncertainties can only be reduced by concerted laboratory efforts.

In addition to the so far discussed risks and benefits of SAI by solid particles, there are a number of other unresolved challenges. First, it is subject to uncertainty that mono-disperse solids can be dispersed into the stratosphere from aircraft in large enough quantities to compete with the relatively simple and fast release of gaseous SO258. Second, modeled injections were continuous over time and equally distributed into the grid boxes of SOCOL-AERv2 which has a horizontal resolution of 325 km × 325 km × 1.5 km at 50 hPa in the equatorial stratosphere, thereby neglecting sub-model-grid-scale microphysical processes occurring within the injection plume of an aircraft. It has been shown that agglomeration could significantly decrease the scattering efficiency of solids22. Third, aerosol-cloud interaction are still highly uncertain and could substantially alter the resulting RF. Calcite particles for example were shown to be good ice nucleation particles, which could increase cirrus cloud coverage and thus weaken the RF from calcite particles56. However, same as for sulfur-based SAI, the cirrus cloud effect of SAI remains subject to large process uncertainty and requires further research.

It is important to highlight that SAI is not a substitute for climate mitigation technologies, which tackle the root cause of climate change, such as getting to global net-zero GHG emissions and scaling up GHG removal methods. Rather, SAI would represent a temporary measure, which could help shaving the peak of global warming and avoid reaching climate tipping points. To inform potential future decision-makers about risks and benefits of SAI, it is important to thoroughly study the associated physical, ethical and social impacts and uncertainties. With our study we show that consideration of SAI of solid particles as an alternative to sulfuric acid particles requires coordinated, collaborative laboratory research to investigate their detailed optical properties as a function of aging in the stratosphere, their ice nucleation properties as well as the heterogeneous chemistry on solid particles at stratospheric temperatures, partial pressures of relevant trace gases, relative humidity and stratospheric UV irradiation. High-quality laboratory data in combination with Earth System modeling would constrain the process uncertainties of SAI of solid particles. Given the large potential of reducing undesirable side-effects on the global environment of this climate intervention strategy, we encourage conducting this research.

Methods

The interactive solid particle microphysics scheme in SOCOL-AERv2

We use the ACCM SOCOL-AERv27, which is based on ECHAM5.4 general circulation model59, coupled to the chemistry module MEZON60 and the aerosol microphysics module AER61,62. SOCOL-AERv2 features a vertical resolution of 39 sigma-pressure levels, extending up to 0.01 hPa (about 80 km altitude), with a horizontal resolution of T42 (2.8∘ × 2.8∘). Dynamics are updated every 15 min, while chemistry calculations occur every 2 h with a 2 min subtimestep for microphysics. Treatment of sulfuric acid aerosols in the model and their various effects have been extensively validated against observations and other models for volcanic cases6,9,63,64, background aerosol layer7,65, and sulfur-based SAI11,32,66,67. A detailed description of the solid particle microphysics module incorporated in SOCOL-AERv2 is available in a previous publication34. The model simulates the injection and advection of solid particles and their microphysical interactions such as settling, agglomeration via self-coagulation, coagulation with sulfuric acid aerosols, condensation of H2SO4 on the particle surface and the settling of the solid particles interactively coupled to the model’s radiation and heterogeneous chemistry schemes34. The model allows for representation of heterogeneous chemistry on the solid particles33. The solid particles are also coupled to the model’s long- and shortwave radiation code via their scattering and absorption cross sections34.

We focus on alumina and calcite particle injection since these are the only potential injection candidates for which some experimental data on heterogeneous chemistry in the stratosphere is available21,54,68,69. The model treats alumina and calcite particles as very different representatives of solid particles concerning their interaction with stratospheric chemistry.

On the one hand, alumina is thought to be largely unreactive and the bulk of the particles (i.e., Al2O3) will most likely not undergo chemical reactions in the stratosphere33, whereas calcite is a base and is thus expected to readily react with most of the acidic molecules available in stratospheric air (i.e., mainly HNO3, HCl, and H2SO4). However, alumina particles still provide surfaces, which allow for adsorption of HCl and HNO3 as well as for the uptake of H2SO4. Thus, the solid particle microphysics model takes into account coagulation of solid particles with H2SO4 − H2O aerosols as well as condensation of gaseous H2SO4 on alumina surfaces. The condensed H2SO4 − H2O is represented on the alumina particles assuming a contact angle of H2SO4 − H2O on alumina particles of about 30∘ 33,34. This leaves parts of the alumina surface uncovered and available for hosting, for example, the heterogeneous reaction HCl + ClONO2 → Cl2 + HNO3, whose rate constant on alumina surfaces has been measured in a study which explored chlorine activation on alumina containing space shuttle exhaust54.

For heterogeneous chemistry on alumina particles, we accounted for Reaction (2) by applying a Langmuir-Hinshelwood description of adsorption and reaction for the uptake coefficient of ClONO2. Experimentally measured uptake coefficients of ClONO254 were extrapolated to stratospheric partial pressures by using the “dissociative HCl" parameterisation derived in a previous study33 and assuming co-adsorption of HNO3. Additionally, for Reactions (1), (6), and (7) we applied the same heterogeneous reaction rates as for sulfuric acid aerosols (Section SAI microphysics and heterogeneous chemistry).

$${{{{\rm{ClONO}}}}}_{2}+{{{{\rm{H}}}}}_{2}{{{\rm{O}}}}{\to }^{{{{\rm{surf}}}}}{{{{\rm{Cl}}}}}_{2}+{{{\rm{HOCl}}}}$$
(6)
$${{{\rm{HOCl}}}}+{{{\rm{HCl}}}}{\to }^{{{{\rm{surf}}}}}{{{{\rm{Cl}}}}}_{2}+{{{{\rm{H}}}}}_{2}{{{\rm{O}}}}$$
(7)

On the other hand, calcite particles are thought to react with HNO3, HCl, and H2SO4 thereby potentially altering the composition of the particles. Therefore, calcite is an example of a material which is expected to undergo chemical aging in the stratosphere, thereby altering its chemical and optical properties over the course of its stratospheric residence time. In case of SAI of calcite, the model thus allows for particle composition changes via uptake of HNO3, HCl, and H2SO4 resulting in particle mixtures of \({{{{\rm{CaCO}}}}}_{3},\,{{{{\rm{CaSO}}}}}_{4},\,{{{\rm{Ca}}}}{({{{{\rm{NO}}}}}_{3})}_{2}\) and CaCl234.

For heterogeneous chemistry on calcite particles we followed previous studies21,37 and applied uptake coefficients of 10−4 for HNO3, 10−5 for HCl on timescales relevant for the stratosphere, even though other studies have measured much larger initial uptake coefficients68,69. However, these uptake coefficients cannot be maintained for relevant timescales due to increased passivation of the surface with reaction products with time37. Conversely, the uptake coefficient for H2SO4 was kept at 1.0 for all calcite simulations, since H2SO4 has a very low equilibrium vapor pressure.

Experimental setup

Each perturbation experiment injected 5 Mt of material per year at 50 hPa (~20 km altitude) equally distributed in all model grid cells between 30∘N and 30∘S at all longitudes with particle radii of 240 nm for alumina and calcite injections, which are the baseline simulations performed in a previous study34. These simulations were extended to 45 years to get a more robust estimate of the effective RF. Each simulation spanned over 50 years, of which the first 5 years served as spin-up to equilibrate stratospheric aerosol burdens. Thus, a total of 5 simulations for the reference run and four perturbation experiments with injections of 5 Mt yr−1, Al2O3, CaCO3, SO2 and direct injections of accumulation mode sulfuric acid aerosols (AM − H2SO4, i.e., log-normally distributed with rm = 0.095 μm and σ = 1.5). The latter scenario assumes that an optimized aerosol size distribution with a mode radius of 0.095 μm can be produced by injecting gaseous H2SO4 into an aircraft plume32,66,70,71. We assumed injection rates of absolute masses and not sulfur equivalent masses since injection of elemental sulfur might not be technically feasible39. In addition to the baseline experiments with timeslice boundary conditions with climatological SST, and SIC, GHG, and ODS concentrations set to year 2020 the five ensembles were also simulated for timeslice 2090 boundary conditions. GHG and ODS concentrations were taken from SSP5-8.572 and73, respectively, while SST and SIC were taken from a 10-year average (2011–2020) of the Hadley dataset74 for 2020 conditions and from the RCP8.5 scenario (2090–2099) simulated by CESM5-CAM1 for 2090 conditions75. Since we are using fixed SST and SIC, it is not possible to calculate the surface cooling effect. We refrain from inferring the potential surface cooling effect by using available estimates of climate sensitivity (quantified as units of surface temperature change per unit of RF from CO2) since the efficacy of SAI as a climate driver might be different from that of CO2 and other GHGs. SAI and GHGs as a climate driver scatter and absorb in very different wavelengths, respectively and the associated climate feedbacks are likely also different.