The susceptibility of microphysical properties of young contrails to changes in aircraft soot emissions is studied with a microphysical plume model. Liquid plume and ambient particles compete with exhaust soot particles for the formation of contrail ice particles, assuming that soot particles are activated into water droplets prior to homogeneous freezing. Soot controls ice formation in contrails for high number emission indices including the range of current global fleet values. A fivefold reduction of soot emissions from average levels of 5 × 1014 − 1015 (kg-fuel)−1 approximately halves the initial contrail visible optical depth. Further soot reduction reverses this trend at temperatures well below the formation threshold temperature unless emissions of sulfur and organics are cut substantially. Whether and to which degree reductions in soot emissions help mitigate the contrail climate impact depends on subsequent aircraft wake vortex processing of contrails and their development into contrail cirrus.
 The meteorological conditions describing the formation of aircraft contrails are determined by the moisture and heat budgets of aircraft exhaust plumes that mix with drier and colder ambient air. Thermodynamic considerations suffice to predict contrail formation threshold temperatures T*, typically below 228 K at air pressures below 300 hPa in an ice-saturated atmosphere for modern jet aircraft [Schumann, 1996]. A detailed kinetic treatment of plume turbulence, microphysical and chemical properties of plume particles, and phase transitions is required to describe the number, size, and optical properties of ice particles forming in contrails [Kärcher, 1996].
 The environmental challenges of aviation operations are well recognized both within the atmospheric sciences [Sausen et al., 2005] and aeronautics [Green, 2003] communities. As contrails substantially contribute to aviation-induced climate change [Forster et al., 2007] and demand for air transport is expected to increase strongly [International Civil Aviation Organization, 2007], operational and technological options to mitigate contrail formation are under scrutiny, none of which appear to be sufficiently well evaluated as to their efficacy [Noppel and Singh, 2007; Gierens et al., 2008]. Technological options to suppress contrail formation include modifications of jet engine emission properties.
 Two plume particle types, affected by emissions of water vapor (H2O), charged molecular clusters (chemi-ions), and condensable sulfur and hydrocarbons (organics), serve as the main contrail ice particle precursors. Ultrafine liquid particles (mean diameters of few nm) mainly forming on chemi-ions in the plume prior to contrails [Yu and Turco, 1997; Kärcher et al., 2000] and larger soot particles (mean sizes of few tens of nm) forming during kerosene fuel combustion [Anderson et al., 1998; Petzold et al., 1999]. The latter control contrail ice formation in threshold conditions [Kärcher et al., 1995, 1996]. In the absence of contrails, aircraft soot particles may also trigger or modify cirrus cloudiness by acting as ice-forming nuclei away from the source aircraft; this subject is examined elsewhere [Kärcher et al., 2007].
 The simulations have been carried out with a comprehensive parcel model of jet plume aerosol and ice microphysics [Yu and Turco, 1998]. It accounts for electrical charge effects on liquid particle nucleation and growth. Soot, liquid (without soot cores), and ice particles are discretized over a fixed size grid and can take variable radii in each size bin. Their chemical components are tracked individually.
 Turbulent plume mixing with ambient air cools the exhaust plume, entrains ambient particles and H2O, and dilutes plume constituent mixing ratios. Uniform mixing is assumed across the plume cross section, which we regard as a reasonable approach to represent average conditions. Liquid plume and ambient particles consist of water and sulfuric acid (H2SO4). Coagulation and condensation lead to an aqueous H2SO4 coating around the emitted soot particles, turning them into internally mixed particles. Organic emissions have been omitted because little is known about their chemical nature. However, the mass fraction of condensable organics may become comparable to H2SO4 in cruise conditions, increasing the size of ultrafine liquid plume particles [Kärcher et al., 2000]. We point to uncertainties introduced by this omission in sections 3 and 4.
 Ice particles in visible contrails mainly form by homogeneous freezing of water-activated soot and liquid (plume and ambient) particles, an assumption that is consistent with available measurements [Kärcher et al., 1998a, Figure 3]. This holds even though aqueous H2SO4 and coated soot particles might as well freeze homogeneously prior to water activation. Allowing soot particles to nucleate ice heterogeneously prior to water saturation (comparable to the homogeneous freezing of liquid solutions) would lead to a slightly earlier onset of contrail formation but would not significantly affect our results. More effective ice nucleation of soot particles closer to plume ice saturation contradicts in situ observations. Ice particles formed from these two particle types are tracked individually in the model and compete for available H2O during depositional growth.
 The simulations have been initialized as follows: ambient air pressure and relative humidity over ice 230 hPa and 110%, respectively; plume dilution history for a large aircraft in cruise according to Yu and Turco ; average fuel sulfur (S) content 0.2 g/kg; S-to-H2SO4 conversion efficiency 3% and chemi-ion emission index 2 × 1017 (kg-fuel)−1 consistent with airborne measurements [Kärcher et al., 2000]; bimodal lognormal soot particle size distribution (PSD) with scaled concentrations (fixed ratio of 2615:1 of small versus large mode) and mean diameters D (geometric width σ) 25 nm (1.55) and 150 nm (1.65) estimated from in situ data [Petzold et al., 1997]; lognormal ambient PSD with total number density 600 cm−3, D = 30 nm, and σ = 2.3, representing average upper tropospheric cases. Changing these parameters within reasonable ranges, adding organic emissions, and including applications to smaller aircraft would not significantly alter our results for medium to high soot emission levels.
3. Results and Discussion
3.1. Formation Conditions
 The evolution of temperature and relative humidities over supercooled water (RH) and over ice (RHI) in the jet plume are shown in Figure 1 assuming warm (Figure 1a), average (Figure 1b), and cold (Figure 1c) ambient temperatures (T) and a soot emission index EIs = 1015 (kg-fuel)−1. In the threshold case at T* ≈ T = 223 K, RH only slightly surpasses water saturation when ice particles start to form. In colder conditions, the plume cooling rates are higher due to faster mixing, leading to a steeper rise of RH and RHI with higher maxima, and an earlier onset of contrail formation. As more ice particles form with decreasing T (see below), RHI becomes more rapidly reduced to saturation despite continuous entrainment of slightly supersaturated air.
 To obtain our main results presented in Figure 2, EIs has been used as a free parameter covering a wide range of values. We report the total number densities ni of contrail ice particles versus EIs at a plume age t = 1 s shortly after ice formation (arrows in Figure 1) (Figures 2a–2c) and their PSDs for selected EIs at t = 5 s at the end of the jet regime after which aircraft wake vortex processes control the subsequent microphysical evolution (Figures 2d–2f).
3.2. Soot-Rich Regime
 In the soot-rich regime for EIs ≥ 1015 (kg-fuel)−1, ice particles mainly form by homogeneous freezing of water around the soot cores (compare solid and dashed curves in Figures 2a–2c). For EIs = 1015 (kg-fuel)−1, the fraction of soot particles forming ice (not shown but reflected in the ni-values) increases with decreasing T, because freezing rate coefficients rise steeply, achieving values 0.09 (223 K), 0.98 (218 K), and 1 (213 K). The frozen soot fraction changes very rapidly near T*. The ni-values at t = 1 s increase from ∼5000 cm−3 (223 K) to ∼50,000 cm−3 (213 K). These values are 1.5–2.5 times higher at the time of contrail formation because of plume dilution, respectively.
 Furthermore, at a given T, ni rises with increasing EIs as mixed particles are turned into ice before RHI decreases. The decrease in RHI is mainly forced by dilution (223 K) or by deposition of H2O on growing particles (218 K). In the latter case, ni tends to saturate with increasing EIs indicating that the rate of depletion of supersaturation exceeds the freezing rate. At 213 K, despite an even faster decrease of RHI, ni rises again strictly in proportion to EIs because freezing rates dominate over the water vapor depletion rates.
 The contribution of liquid (soot-free) particles to contrail formation is negligible (dot-dashed curves in Figures 2a–2c). Concentrations of ice particles forming on entrained ambient particles reach 200-300 cm−3, depending on T. The plume supersaturation ratios are too small to activate the ultrafine liquid plume particles into water droplets for subsequent homogeneous freezing.
 The PSDs at t = 5 s (Figures 2d–2f) reflect the partitioning of contrail ice between liquid and mixed particles as discussed above. The mean diameters Di of ice particles from ambient liquid particles (dashed) are slightly larger than those from the more abundant soot-containing particles (solid) since the former are initially larger and therefore more efficient in freezing. The ice PSDs contain particles with Di ≈ 3 μm (223 K) and ∼1.5 μm (218 K and 213 K).
3.3. Soot-Poor Regime
 In the soot-poor regime, most contrail ice originates from liquid particles without soot cores. Figures 2a–2c show that this regime is realized for EIs ≤ 1013 (kg-fuel)−1 in the cases 223 K and 213 K (for different reasons) but only below 1012 (kg-fuel)−1 at 218 K.
 In threshold conditions, total ni is determined by the number of entrained ambient particles available for freezing at the time of maximum supersaturation. Somewhat more particles freeze (∼350 cm−3) than in the soot-rich regime because soot particles acting as condensational sinks for H2O are lacking. At lower T, ni increases and becomes controlled by freezing of the liquid particles that nucleated and grew in the plume prior to ice formation. Only a small fraction of these copious, nm-sized particles can be activated. This fraction increases sharply between 218 K and 213 K, resulting in very high ni similar to levels in the soot-rich regime.
 Less abundant engine metallic particle emissions, from burning of traces of metal compounds contained in kerosene or from engine erosion, would contribute to contrail ice [Kärcher et al., 1998a] but would not be noticeable in Figures 2a–2c. Uptake in aqueous plume particles of nitric acid formed via rapid oxidation of emitted nitrogen oxides [Kärcher, 1996] and emissions of condensable organics from unburned fuel hydrocarbons [Yu et al., 1999] would further increase the fraction of freezing liquid plume particles by increasing their size and activation probability. Therefore, our ni-values in the soot-poor regime are considered lower-limit estimates, especially near and below 218 K.
 The simulated ice PSDs (Figures 2d–2f) confirm that ice particles originating from liquid particles are more abundant than those from soot. Furthermore, Di is larger since the total number of particles formed is smaller and they share the same amount of available H2O. At 223 K, both liquid and mixed ice modes have Di ≈ 5.5 μm. At 218 K, these two modes begin to split up into a smaller liquid (∼4.2 μm) and larger mixed (∼4.8 μm) mode. At 213 K, this bimodality is even more distinct, the liquid mode itself being separated into a first mode (∼3 μm) originating from freezing of ambient particles and a second mode (∼2 μm) from plume particles.
3.4. Intermediate Regime
 Total ni decreases from the high values in the soot-rich regime to the comparatively low levels in the soot-poor regime (Figures 2a–2c). The crossover at which ice formation from liquid and mixed particles becomes equal shifts from 1014 (kg-fuel)−1 at 223 K to 1013 (kg-fuel)−1 at 218 K and back at 213 K. In the latter case, ni takes a minimum near 1014 (kg-fuel)−1, minimizing the susceptibility of ni as a function of EIs. As the simulated ni-values in the soot-poor regime are lower-limit estimates, this minimum would already develop near 218 K had we included condensable organic emissions and nitric acid in the simulations. The mean diameters of contrail ice particles vary in the range 2.5–4.5 μm from low to high T.
 At each T, decreasing EIs from 1015 to 1014 (kg-fuel)−1 causes a tenfold decrease in ni and roughly a twofold increase in Di; including changes in the scattering efficiency of visible light, this reduces the contrail optical depth by less than 30–40% at t = 5 s.
4. Summary and Implications
 Salient points of previous model results [Kärcher et al., 1998a] obtained for fixed soot emissions have been confirmed by an independent model approach, building confidence in our ability to model the microphysical processes controlling persistent contrail formation. The present study, representative for extratropical upper tropospheric flight levels, confirms that (1) it is not possible to reduce the occurrence frequency of, or entirely avoid, contrails by changing or eliminating soot emissions; (2) total ni increases and Di decreases with decreasing T; (3) at ambient temperatures below but near T*, a large fraction of unactivated soot particles remains as interstitial aerosol in the contrail.
 Furthermore, we have shown that the majority of contrail ice particles originates from aircraft soot particles for EIs > 1–2 × 1014 (kg-fuel)−1, including the range ∼3 × 1014 − 2 × 1015 (kg-fuel)−1 from the current fleet [Petzold et al., 1999]. Total ni is not sensitive to H2SO4 and chemi-ion emissions and increases (Di decreases) with increasing EIs. Detailed knowledge of soot PSDs and ice nucleation properties is only needed to predict visible contrail formation accurately near but slightly below T* when the plume barely overshoots water saturation, or to predict (initially) subvisible contrail formation near but slightly above T* when the plume becomes ice- but not water-supersaturated.
 For EIs ≤ 1012 − 1013 (kg-fuel)−1 depending on ambient temperature, contrail ice formation is dominated by ambient and liquid plume particles, with T determining their relative contributions. These contributions can only be predicted with confidence if their PSDs and ice nucleation properties are known. Total ni decreases to ambient levels and Di increases with decreasing EIs, but only for T > 218 K as plume cooling rates and hence supersaturation are insufficient to activate and freeze the small acidic liquid plume particles. In colder conditions, the situation reverses, and predictions of ni depend on chemi-ion and condensable sulfur, nitric acid, and organics levels available at emission. Changes in concentrations of condensable species result in changes in the size of liquid particles and hence their water activation properties [Kärcher, 1996; Yu and Turco, 1997].
 Our work relates to technological efforts aiming at reducing soot emissions using jet fuel additives [Liscinski et al., 2001; Bae and Avedisian, 2004; Montgomery et al., 2005]. Reducing EIs from current average levels of ∼5 × 1014 − 1015 (kg-fuel)−1 by a factor 5 would lower ni by approximately the same factor and less than double Di. This result highlights the potential to alter nascent contrail properties. Further reduction of EIs would push ni down to ambient levels in warm (T > 218 K) conditions, especially near T*. This is the largest possible effect, achievable only for very efficient removal (>90%) of soot particle emissions, or when using non-sooting fuel.
 However, in the soot-poor regime, no further reduction of ni, even the opposite effect, is realized at average and cold temperatures because ultrafine liquid plume particles then act as ice forming agents despite their small sizes. With today's use of kerosene, a concomitant removal of fuel sulfur would not alter this conclusion, as those soot-free particles still form by ion-induced nucleation and condensation of emitted organic vapors according to analyses of in situ observations [Kärcher et al., 1998a, 1998b; Yu et al., 1998]. Any organic compound present in soot-reducing fuel additives might increase the size of liquid plume particles and thus heighten their role in contrail ice formation. This implies that it might be more beneficial to reduce EIs rather than to entirely remove soot emissions.
 Our simulations suggest a significant potential to lower the initial contrail visible optical depth. Subsequent microphysical processing of ice particles after the jet regime might enhance or alleviate such changes. Therefore, it remains open if and to which degree the differences in contrail properties resulting from reductions in soot emissions propagate into the further evolution of contrails and affect their radiative properties. This issue, however, is crucial for assessing the potential of mitigating the contrail climate impact based on reductions in soot emissions.
 Future atmospheric research into persistent contrails should investigate their development into cirrus as a function of the factors controlling contrail ice formation. Similar efforts are needed to study contrail cirrus induced by aerodynamic effects [Kärcher et al., 2008] or forming from alternative fuels such a liquid hydrogen [Ström and Gierens, 2002].
 This study was performed within the DLR-project “Climate-compatible air transport system” (CATS). F.Y. acknowledges funding support from NSF under grant 0618124. We acknowledge U. Schumann for motivating this work and D. Baumgardner and K. Gierens for fruitful discussions.