Embedding Pd into SnO2 drastically enhances gas sensing

Combustion aerosol processes can uniquely embed noble metals into semiconducting particles. Here, monocrystalline SnO2 particles embedded with Pd and/or PdOx were made by flame spray pyrolysis (FSP) of appropriate precursors through microexplosions by droplet-to-particle conversion as the crystal size was proportional to the cube root of precursor solution concentration, C. These particles were air-annealed and leached with nitric acid for removal of metallic Pd from their surface. The SnO2 crystal size varied from 11 to 24 nm and was in close agreement with the primary particle size determined by nitrogen adsorption. The embedded fraction of Pd ranged from about 30 to 80% of the nominal Pd-content. This was achieved by judiciously varying the C, Pd content and the ratio of precursor solution to dispersion oxygen flowrates during FSP. The response of sensors made by doctor blading films of such particles to 1 ppm of acetone and CO was evaluated at 350 °C and 50% relative humidity. Embedding Pd/PdOx into SnO2 significantly increased the sensor response: 2–6 times over that of pure or conventionally-made Pd-containing SnO2 sensors at low nominal Pd-contents (0.2 mol%). For higher ones (i.e. 1 mol% Pd), the sensor response was enhanced by up to two orders of magnitude. This is attributed to Pd atoms in the SnO2 lattice near the particle surface and/or Pd/PdOx clusters acting as nanoelectrodes into SnO2 films and altering their transducing properties as shown by high resolution electron microscopy, XPS and baseline resistance measurements of pure and Pd-embedded SnO2 sensing films.


Introduction
Gas sensors are used widely for detection of drunk driving and monitoring compliance of automotive emissions as well as to prevent gas explosions and poisoning in mines.Due to their low cost and portability, they are most promising for monitoring air quality outdoors (e.g.NO 2 ) 1 or indoors (e.g.formaldehyde), 2 food quality (e.g.methanol in alcoholic beverages), 3 for medical diagnostics (e.g.diabetes or inammatory markers such as NO for asthma) 4 or monitoring fat-burning from breath acetone. 5hat links these applications is the requirement to selectively measure very low, sometimes at few ppb, concentrations of specic analytes.Therefore, there is a clear need for highly sensitive materials driving the search of new chemical compositions and novel material structures (or morphologies).
Standard chemiresistive gas sensors consist of a functional lm of semiconducting nanoparticles, deposited between a pair of electrodes.When that lm is exposed to analyte gas it serves as the catalyst for the reaction between adsorbed oxygen species and analyte on the lm surface. 6This reaction alters the concentration of charge carriers in the lm and can be measured as a change of its resistance.Therefore, a good sensor, among other features, should give the highest possible response, i.e. a change in resistance in relation to the baseline in clean air.
Tin oxide is the most common gas sensor.Doping it with noble metals (NM) is frequently used to enhance its sensitivity. 7n contrast to most methods for preparation of SnO 2 with NM (impregnation, 8 reverse micelles, 9 ball mill mixing 10 and photodeposition 11 ) in which SnO 2 particles are only externally decorated with NM clusters, ame spray pyrolysis (FSP)

Nanoscale Advances
PAPER uniquely allows placing NM on the surface and inside the sensing nanoparticles. 12Recently it was shown that solely embedding palladium (Pd) or PdO into SnO 2 increased signicantly the sensor response to CO, acetone and ethanol over that from pure and Pd-decorated SnO 2 made, for example, by Pd photodeposition. 13ere, this discovery is explored systematically by varying the embedded Pd-content and SnO 2 crystal size by closely controlling the FSP synthesis of such particles.Palladium is removed from the surface of SnO 2 by leaching and the effect of embedded Pd fraction on sensor response to 1 ppm of inorganic (CO) and organic vapors (acetone) at 350 °C and 50% relative humidity is investigated focusing on the sensitivity of such materials as their selectivity can be enhanced readily by lters. 14otential mechanisms for the observed drastic increase of sensor response to both analytes due to Pd embedding are assessed revealing the role of Pd near the SnO 2 surface and Pd/ PdOx domains into its bulk.
As our focus is on understanding the partitioning of Pd between the surface and interior of SnO 2 and the subsequent impact on gas sensing, we benchmarked the sensitivity, the most basic property of gas sensors, at a commonly employed temperature, humidity and concentration of acetone and CO.Other sensor characteristics such as selectivity, stability and response and recovery times have not been investigated as they largely follow those of pure and Pd-containing SnO 2 sensors.The effect of sensing temperature, humidity and analyte concentration on the performance of SnO 2 -based sensors has been documented extensively for acetone 13,[15][16][17][18] and CO 11,16,[19][20][21][22] as summarized in Table S1 in the ESI.†

Synthesis of sensing nanoparticles
Pure and Pd-doped SnO 2 particles were prepared by FSP. 13 A precursor solution of tin(II)-ethylhexanoate (Aldrich, purity >92.5%) and palladium(II)-acetylacetonate (Aldrich, purity $99%) in xylene (Aldrich, purity >96%) with a total metal ion (or precursor solution) concentration, C, of 0.1, 0.5, 1.0 or 1.5 M was used.Palladium was added at the concentration of 0, 0.1, 0.2, 0.5, 1 and 3 mol% of total metal content.This solution was fed to the capillary of a concentric spray nozzle at 1, 5 and 9 ml min −1 by a syringe pump and dispersed by an annular oxygen ow of 2, 5 or 8 l min −1 .A premixed methane/oxygen (1.3/3.2 l min −1 ) ame ignited and supported the spray ame.A circular sheath oxygen ow of 5 l min −1 was used around that ame to assure complete combustion of the precursor.The FSP inlet conditions for all experiments are listed in Table S2 in the ESI.† Using a vacuum pump, the powder was collected (Busch, Seco SV 1025C) on a glass-ber lter (GF6 Albet-Hahnemuehle, 257 mm diameter) 50 cm above the burner.Product powders were scraped off the lter surface with a spatula and sieved with 0.25 mm stainless steel sieve to remove lter fragments.A portion of all powders was annealed in air at 500 °C for 5 hours (Carbolite Gero, Gero 30-3000 °C).Both as-prepared and annealed powders were kept for further characterization.

Removal of surface Pd
First, 150 mg of annealed powder were placed in a quartz reactor tube and secured with glass wool on both ends.The powder was then reduced by passing 100 ml min −1 of 5% H 2 in Ar for 30 min at 150 °C.Then 100 mg of such reduced powder was placed inside a round bottom ask where 85 ml of ultrapure water (Merck Millipore) and 15 ml of nitric acid (65% HNO 3 in water, purity $99.99%) were added.Then the ask was heated up to 60 °C in an oil bath and stirred under reux for 4 h at 500 rpm to dissolve the palladium on the surface of SnO 2 particles. 13Then the ask contents were transferred to falcon tubes and the particles were separated from solution by centrifugation (7200 of relative centrifugal force for 10 min), washed thrice with ultrapure water and dried in a vacuum oven (SalvisLAB, Vacucenter) at 50 °C and 50 mbar for 12 h.The resulting leached solution was analyzed with inductively coupled plasma-optical emission spectroscopy (ICP-OES, Agilent 5110) to determine the amount of leached Pd.The ICP-OES was calibrated using standards for Sn and Pd in HNO 3 solutions (inorganic ventures).The surface palladium fraction was dened as the ratio of the measured amount of leached palladium and the nominal Pd content in the powder.The SnO 2 powders with Pd content 0, 0.1, 0.3 and 0.5 mol% were completely digested by adding 3 ml concentrated HNO 3 (69%) and 500 mL HF (40%) and heating to 240 °C for 10 min in a pressurized microwave (UltraClave, MLS GmbH).Digestion was performed both on annealed and leached powders to determine the total and embedded Pd content.Aer digestion, samples were lled up to 50 ml with ultrapure water.The concentrations of Sn and Pd ions then were measured by ICP-OES (Agilent 5110) as the above leached solutions.

Particle characterization
Particle phase composition and crystal size, d, were obtained by X-ray diffraction (XRD, Bruker D2 Phaser) at 40 kV and 40 mA at 2q (Cu Ka) of 20-70°with a step size of 0.1°and scanning speed of 0.0972°s −1 .The patterns and crystal size were determined by Rietveld fundamental parameter renement with the Topaz 4.2 soware (Bruker) using the characteristic parameters of cassiterite SnO 2 (PDF 41-1445) and Pd (PDF 01-1310) and PdO (PDF 43-1024).
The specic surface area (SSA) and surface equivalent particle diameter (d BET ) of the powders were measured by 5point N 2 adsorption at 77 K (Micromeritics Tristar II Plus) using the Brunauer-Emmett-Teller method.Powder samples were degassed under N 2 at 150 °C for 1.5 h prior the adsorption measurement.The d BET was calculated using the densities of SnO 2 (6.95 g cm −3 ) and PdO (8.3 g cm −3 ).
Specimens for transmission electron microscopy were prepared by dispersing a spatula tip of dry powder in 500 mL of high purity ethanol (>99.9%,LiChrosolv, Supelco) in an agate mortar.A volume of 10 mL of supernatant from the mortar was then drawn through a lacey carbon TEM grid (EM resolutions) using a tissue.Then, the grid was washed thrice with ultrapure water and air dried.High-angle annular dark-eld scanning transmission electron microscopy (HAADF-STEM) images and energy-dispersive X-ray (EDXS) elemental maps were then obtained from these specimens using a FEI Talos (F200X, Super-X EDS, 4 detector conguration, FEI, Hillsboro, OR, USA).
X-ray Photoelectron Spectroscopy (XPS) measurements were conducted on samples of as-prepared, annealed and leached powders containing 1 and 3 mol% of Pd, as for that with 0.2 mol% Pd, the signal to noise ratio was too low for analysis.Powders were compressed into separate indium foil pieces and subjected to analysis using a Quantum 2000 XPS system (Physical Electronics).Detailed descriptions of these measurements can be found in the ESI.† Literature binding energies for relevant Pd species along with respective average, min/max and standard deviation values that were used to constrain the curve tting model are shown in Table S3, ESI.†

Sensor assembly and measurements
Powders were mixed with 1,2-propanediol (Sigma-Aldrich, purity >> 99%) to form viscous and homogeneous pastes.The pastes were doctor bladed 13 with a razor on Al 2 O 3 sensor substrates (Electronic Design Center, Case Western Reserve University, Electrode type #103) having two interdigitated Pt electrodes (sputtered, 350 mm width and spacing) on the front and a Pt heater on the back.The sides of the substrates were secured with 15 mm thick aluminum foil to control the thickness of the spread lm.The sensors were dried at room temperature for 4 hours, followed by 16 hours at 80 °C.Finally, the sensors were annealed for 30 minutes at 300 °C with 10 °C min −1 ramping to improve mechanical stability of the lm and remove any remains of organics used during paste preparation.An illustration depicting the sensor, outlining the measurement principle of lm resistance (R) and demonstrating the doctor-blading procedure can be found in Fig. S1  and S2, respectively, within the ESI.† The sensor was placed in a closed chamber and heated to 350 °C by DC through its substrate Pt heater, adjusted by linear regression for temperature and the resistance was determined individually for each sensor using thermocamera (Fluke, Ti110).A synthetic air stream of 1 l min −1 and 50% relative humidity (RH) was continuously passed through the chamber with the sensor until a stable baseline resistance (R air ) was achieved.The humidity was tuned by mixing a stream of dry synthetic air with an equal stream of air saturated with water vapor in an inline water-lled bubbler.To measure sensor response to analytes, acetone or carbon monoxide was admixed to the air stream from calibrated gas standards in an appropriate amount to result in a concentration of 1 ppm.The sensor exposure to analyte lasted for, at least, 10 minutes to ensure that resistance has reached a nal value (R analyte ).The interval between two consecutive pulses was, at least, 20 minutes to ensure the resistance returned to the baseline.The sensor response (S) is: as SnO 2 is an n-type semiconductor and both acetone and carbon monoxide are reducing gases. 23The sensor response and recovery times were dened as the time needed for the sensor resistance to reach 90% of the nal change upon analyte exposure or 90% of recovery aer the dosing of the analyte has been discontinued.

Results and discussion
SnO 2 crystal size control In contrast, the Pd content had no effect on the as-prepared SnO 2 crystals that increased from 5.6-5.8nm to 13.4-14.2nm with C increasing from 0.1 to 1.5 M. In contrast, increasing the Pd content hindered the growth of annealed SnO 2 by the solute drag effect 24 as shown also in Fig. S3 in the ESI † in agreement with Pineau et al. 13 Given that the particle size as determined by N 2 adsorption (Fig. S3, ESI, † squares) was close to the crystal size (Fig. S3, ESI, † circles) at all Pd contents for both as-prepared and annealed particles, monocrystalline SnO 2 particles were made here.Most interestingly, the size of all as-prepared particles followed a power law (Fig. 1, dotted line) with C, d = 12.2C 1/3 , indicating that particle size directly relates to precursor solution concentration, C, and consequently to formation from single droplets by evaporation-precipitation. 25 Most likely, such droplets were formed by microexplosions. 26So SnO 2 formation takes place largely by droplet-to-particle conversion at the highly oxidic conditions of spray combustion.
Increasing the ratio of FSP precursor solution to dispersion O 2 owrate, P/D, increases the precursor aerosol concentration and ame temperature (Fig. S4 in the ESI †) that increase SnO 2

Paper
Nanoscale Advances crystal size (Fig. S5, ESI †) as seen in Fig. 1 with increasing C.However, the increased ame temperature at high P/D results in slightly larger SnO 2 crystals, i.e. 12 nm at P/D = 5/5 (Fig. 1) and 15 nm at P/D = 5/2 for pure SnO 2 (Fig. S5, ESI †).It should be noted that the SnO 2 crystal size did not change much due to reduction with H 2 , leaching and subsequent washing and drying.

Palladium distribution and oxidation state
Metallic Pd can be leached from the SnO 2 surface as shown for photodeposited Pd onto SnO 2 , where more than 99% of the nominal Pd content was removed. 13For ame-made particles, the question arises regarding the partition of Pd between the particle's surface and interior, as well as its oxidation state.All annealed Pd/SnO 2 particles were leached with HNO 3 to remove Pd from the SnO 2 surface and determine the Pd surface fraction from the Pd ion concentration in the leachate and subsequently the embedded fraction of the total Pd content.To conrm results from HNO 3 leaching, Pd/SnO 2 powders with Pd content ranging from 0.1 to 0.5% of Pd were completely digested using HF.Digestion was performed before and aer leaching to determine the total Pd and the embedded Pd content.Solutions remaining aer HF digestion were diluted and the concentrations of Sn and Pd ions were measured by ICP-OES.Juxtaposition of the results of HNO 3 leaching and HF digestion experiments is shown in Fig. S6 in the ESI.† The sum of Pd determined in the leachates and that in the residual material aer leaching was close to 100% of the nominal Pd content in the SnO 2 powder for all Pd-contents.
To investigate the impact of FSP process conditions on the embedded content of Pd, one FSP process variable (precursor concentration, C, precursor solution/dispersion O 2 owrate ratio, P/D, or Pd content, Pd%) was varied while the other two were kept constant.For example, the fraction of embedded Pd into SnO 2 from total (nominal) Pd contents of 0.1, 0.2, 0.5, 1 and 3 mol% produced at C = 0.5 M and P/D = 5/5 is shown as a function of SnO 2 crystal size in Fig. 2 (squares).Increasing the total Pd content, drastically decreases the embedded Pd fraction from 73 to 35% as the actual amount of embedded Pd increases into SnO 2 since the total Pd content in SnO 2 increases (Fig. S7 in the ESI †).
Increasing the C from 0.1 to 1.5 (Fig. 2, triangles) and P/D from 1/5 to 5/2 (circles) at 0.2 mol% nominal Pd content hardly altered the mole fraction of embedded Pd in SnO 2 (Fig. S7 in the ESI †), even though the embedded fraction decreased from 76 to 56% of the nominal value.The same increase of C (Fig. 2, diamonds) and P/D (stars) at higher Pd content, 1 mol%, decreased the embedded fraction of Pd from 60 to 35% as both FSP variables increase SnO 2 crystal size while inversely affect the ame temperature (Fig. S4 in the ESI †) indicating its limited signicance at the employed conditions.As SnO 2 particles are formed by droplet-to-powder conversion (Fig. 1), the precursor solution concentration, C, largely determines their size.Precursor solutions of increasing C and/or P/D increase the metal ion concentration and form larger SnO 2 crystals (Fig. 1 and S7 in the ESI †).As particle formation takes place within microdroplets 26 (Fig. 1), Pd and Sn are well mixed there and embedding of Pd species into the SnO 2 matrix takes place.
To visualize the change in content and dispersion of Pd in and/or on the particles at different Pd doping levels, STEM-EDXS mapping was used.Fig. 3 presents EDXS elemental maps of annealed SnO 2 doped with various amounts of Pd (a) before and (b) aer leaching with nitric acid.The change of particle Pd content can be observed on the basis of the respective image EDXS spectrum shown in the lower panel of Fig. 3a.As the Pd content increases from 0 to 3%, the relative height of the characteristic Pd-L a peak at 2.839 keV increases as well.Even though there is no Pd peak in the EDXS spectrum of pure SnO 2 (0 mol% Pd, Fig. 3a), it still appears on the elemental map above.The Pd map in this case is only based on unspecic Bremsstrahlung X-ray counts, generated once material is within the path of the electron beam.The exact spectrum for the different Pd map areas for pure SnO 2 in Fig. S8cd in the ESI † proves that there is no detectable Pd in this sample.Before leaching, pure SnO 2 (0% Pd in Fig. 3a) showed the characteristic Sn L l peak at 3.045 keV and was free from Pd L a X-ray counts at 2.839 keV.Using a high beam current of 8 nA, Pd down to 0.2 mol% was detected at 2.839 keV of the EDX spectrum.At the lowest Pd-contents of 0.2 & 0.5 mol%, virtually no Pd clusters were visible while still a distinct Pd peak could be seen at 2.839 keV.This suggests that Pd is distributed inside and/or on the surface of SnO 2atomically dispersed or in Pd clusters close to or below z2 nm in diameter (LOD).
In the elemental map for SnO 2 containing 1 mol% Pd before leaching (Fig. 3a), single Pd clusters larger than 2 nm can be seen.Also for 3 mol% Pd before leaching (Fig. 3a) even more Pdrich spots with a diameter from a few to a dozen nm can be observed.These observations are in line with Koziej et al. for impregnated 27 Pd and Gschwend et al. for photodeposited 11 Pd onto SnO 2 .Using K-edge XANES & EXAFS, for example, Koziej et al. 27 found that at low Pd-contents, e.g.0.2 mol%, Pd is nely dispersed and does not form any clusters which only become apparent at high Pd concentrations (3 mol%).
Aer leaching, no Pd clusters were detected for 0.2 mol% Pd (as expected) and hardly any for 1 mol% Pd (Fig. 3b).When comparing the latter to its initial sample, we go from some to virtually no clusters, indicating that most of the Pd clusters were actually on the SnO 2 particle surface.However, when 3 mol% Pd/SnO 2 was leached, still some Pd clusters were visible.Their number and average size were smaller than before leaching, but still single clusters up to 10 nm in diameter were present.Such clusters could be embedded in SnO 2 which would protect them from the HNO 3 leaching.Another possibility is that such visible Pd-rich regions are not clusters, but only supercially Pdenriched SnO 2 areas that were in contact with larger Pd clusters prior to their leaching.Such enrichment might be caused by annealing-induced solid-state diffusion of Pd into SnO 2 and/ or solid-state reaction 13 forming PdSnO x.This is supported by the fact that the fraction of leachable Pd decreases by 20-40% aer annealing. 13learly very small amounts of Pd can be incorporated into SnO 2 by FSP that could form solid solutions.Such atomic Pd doping should occur only to a limited extent as the SnO 2 Nanoscale Advances Paper diffraction patterns were not affected by Pd at all Pd contents employed here (Fig. S9 and S10 in the ESI †).However, with increased Pd loading, Pd forms a separate phase (i.e.Pd/PdO x ) that cannot be detected by XRD.This phase resulted in domains and clusters into and onto the SnO 2 proportional in size and number to the nominal Pd content as shown in Fig. 3a.Larger clusters are not easily embedded in SnO 2. As a result, more Pd would be exposed to leaching, decreasing the embedded Pd fraction at higher nominal Pd loadings (i.e.only 35% is embedded at the 3 mol% nominal Pd loading, Fig. 2).
To more closely investigate the oxidation states of Pd, X-ray Photoelectron Spectroscopy (XPS) analysis was performed (Fig. S11 †).Due to the limited depth of analysis (typically z1-5 nm), 28,29 this technique provides valuable information about the surface and near-surface regions rather than the bulk of the SnO 2 powder.Detailed spectra with deconvolution into individual oxidation states are presented in the ESI, Fig. S12 and  S13.† They revealed the dominating presence of oxidized forms (PdO x ) of palladium as well as metallic Pd 0 in all the as-prepared and annealed (unleached and leached) samples (Fig. S14 in the ESI †).
The measurements showed PdO x : Pd 0 ratios of z1.1 and z0.7 for particles containing 1 and 3 mol% of Pd, respectively, for both as-prepared and annealed samples (Fig. S15, ESI †).These are consistent with Deligiannakis et al. 30 who found a PdO x : Pd 0 of 0.8 for FSP-made Pd/TiO 2 .As the boiling point of metallic palladium is 2963 °C and the decomposition temperature of palladium oxide is around 800 °C, 31 palladium clusters formed in high temperatures regions of the ame will possess a metallic character.However, as these clusters reach temperatures of around 700 °C away from the ame, they undergo oxidation 32 but only at the surface layer which has access to oxygen.The Pd clusters core and portion incorporated into SnO 2 , lacking access to oxygen, would remain in their metallic form as conrmed by the XPS spectra.
Aer leaching however, the PdO x : Pd 0 ratio increased dramatically to 2.2 and 2.8 for 1 mol% and 3 mol% initial Pd, respectively (Fig. S15, ESI †) suggesting a lower content in metallic Pd aer leaching.In this case, it was possible to remove all Pd from the surface, suggesting surface only Pd can be fully reduced to Pd 0 aer the reduction process, as only Pd 0 is soluble in HNO 3 . 11However, in this study, Pd was introduced already in the ame and thus mixed into the entirety (inside and outside) of the SnO 2 particles.Therefore, here, it is possible thatas suggested from XPSa portion of metallic Pd can remain shielded beneath a protective PdO x /SnO 2 surface.
In a rst line, the XPS results demonstrate that the particle surface is dominated by oxidized Pd species.However, they also suggest the presence of metallic Pd, most likely near the surface.Whether this is true and valid also for deeper layers of

Sensing performance
To investigate the impact of embedded Pd on the sensing of organics (acetone) and inorganics (CO), the response of SnO 2 sensors containing (a) embedded and surface Pd, (b) only embedded and (c) only surface Pd was compared.Fig. 4 shows the sensor response to 1 ppm of acetone at 350 °C in 50% RH by pure (green diamonds) and 0.2 mol% Pd-doped SnO 2 (an optimal Pd content 33,34 for gas sensors) as a function of SnO 2 crystal size made at various P/D at C = 1 M (circles in Fig. 2).For all sensors, increasing the P/D and the resulting SnO 2 crystal size from 13 to 24 nm (Fig. S5 in the ESI †) hardly affects their response consistent with Güntner et al. 12 Pure SnO 2

Nanoscale Advances Paper
sensors had a stronger response (green diamonds) than those containing only surface Pd (gold star) 11 but comparable to those having both surface and embedded Pd (blue triangles).However, when surface Pd was removed from the SnO 2 particles, their response (red circles) increases by 2 to 6 times over pure (diamonds) and Pd-containing SnO 2 (triangles) that had 50% more Pd, regardless of SnO 2 crystal size.These observations are attributed to the detrimental effect of surface Pd (Fig. 4, gold star) and the enhancement of sensing provided by embedded Pd, as depicted in Fig. 4 (red circles), which is fully manifested only aer removal of surface Pd.If the sensing reaction takes place mostly locally at the surface of noble metal clusters, no electron transfer to the conduction band of SnO 2 will occur and thus no change in the conductivity of the sensing lm. 35Therefore, it is commonly observed that sensors exhibit a maximum response when there is only a small amount of dopant present, as sensitivity decreases with increasing dopant content.For SnO 2 containing both surface and internal palladium (Fig. 4, triangles), it is plausible that their effects mutually negate each other.
When the same sensors were tested towards 1 ppm of CO at the same conditions (Fig. S16 in the ESI †), the response for all sensors again was hardly affected by SnO 2 crystal size 12 from 13 to 23 nm.The response of the pure SnO 2 sensors (green diamonds) is lower or comparable to that containing both fractions of Pd (blue triangles).Again, when the surface Pd is removed, the sensor response (red circles) increases 2-6 times over that of pure SnO 2 (green diamonds) and that containing both surface and embedded Pd (blue triangles).
To better understand this enhancement, sensors with high nominal palladium loading of 1% (or 0.6 to 0.4 mol% actual embedded Pd, Fig. S7 in the ESI †) made at increasing precursor solution concentrations C (and subsequently particle size from 11 to 18 nm, Fig. 1) were tested towards acetone at the same conditions as above (Fig. 5).Sensors containing only surface Pd (gold star) or both surface and embedded Pd (blue triangles) showed hardly any response to 1 ppm of acetone and well below that of pure SnO 2 (green diamonds).When, however, surface Pd is removed, their response (red circles) becomes 14 to 95 times higher.This response is also higher or comparable to that of pure SnO 2 (green square and diamonds).Nevertheless, please note that the best of these responses is 3-5 times lower than that of sensors containing 0.2 mol% Pd (Fig. 4).This is consistent with the literature showing that sensors with low Pd contents (0.1-0.2 mol%) exhibit higher response than those with higher Pd loadings (0.5-3 mol% Pd). 9,11,33,36hen the above sensors were exposed to 1 ppm of CO (Fig. S17 in the ESI †), they gave similar results: hardly any response by those containing surface and embedded Pd (blue triangles), which is lower than that of pure SnO 2 (green square and diamonds) and a much higher response by the sensor containing only embedded Pd (red circles) that reached a 125fold amplication for the largest -18 nm -SnO 2 crystals (C = 1.5 M).The response of these sensors was also 2-3 times lower than that for 0.2 mol% Pd-doped sensors, similar to acetone (Fig. 4).Response times of about 10 seconds for pure SnO 2 sensors were observed and a moderate increase to 20-40 seconds for those containing 0.2% Pd before its surface removal, in agreement with literature. 11Notably, leaching of surface Pd halved the response times to those before leaching.Recovery times of approximately 40-125 seconds were observed for pure SnO 2 and 30 to 200 seconds for Pd-containing SnO 2 .
To better understand the sensing mechanism of embedded Pd, the baseline of pure and Pd-containing SnO 2 sensor was examined before and aer leaching surface Pd from the latter

Paper
Nanoscale Advances ones (Fig. 6).The resistance of lms containing 0.2 mol% Pd (surface and embedded, blue triangles) was about an order of magnitude higher than pure SnO 2 (green squares) in agreement with Korotcenkov et al. 33 This can be attributed to the presence of palladium oxide clusters on the surface.A similar picture holds also for sensors made with 1 mol% Pd (Fig. S18 in the ESI †).As mentioned before, there are two oxidation states of Pd, and thus both PdO and PdO 2 can form surface clusters.The PdO 2 is metallic while PdO, of which a signicant fraction (up to 50%, Fig. S14, ESI, †) is found on/near the surface of the particles according to XPS measurements, is a p-type semiconductor. 37Band bending at binary and ternary heterojunctions with n-type SnO 2 causes transfer of electrons and extension of the depletion layer and thus increases the lm resistance.
When the surface Pd was removed, the lm resistance fell even below that of pure SnO 2 in all except one case (Fig. 6, red circles).This indicates that embedded Pd improves the conductive properties of the lm possibly due to the shi of the Fermi level and/or the existence of nanoelectrodes. 38When Pd is dispersed at the atomic level within the crystalline lattice of SnO 2 , the electronic structure is likely to be changed through introduction of additional energy levels arising from defects in the crystalline structure introduced by dopant atoms. 39The atomically dispersed Pd within the SnO 2 lattice has the potential to induce defects, such as oxygen vacancies, which are recognized as the primary source of n-type conductivity in 40 SnO 2 or Pd-related defects.Such defects can serve as charge carriers, increasing the material's conductivity and enhancing its response to analytes. 6The introduction of additional energy levels originating from defects can raise the position of the Fermi level to a higher energy, approaching the conduction band.This would facilitate the promotion of electrons to the conduction band, creating more free charge carriers (electrons) available for conduction and lowering the baseline resistance of the sensing lm.
For SnO 2 doped with 1 and 3 mol% Pd we also encounter Pd clusters with a diameter of a few nanometers.These are located within the SnO 2 particle (Fig. 3b).It is presumed that the presence of embedded Pd or PdO domains can establish efficient conduction pathways within the lm, thus minimizing its electrical resistance. 38mbedding Pd into the SnO 2 bulk allows for synthesis of sensors characterized by both low resistance and high response, because there are no surface Pd clusters to compete chemically with SnO 2 .Embedded Pd clearly improves the transducing properties of the sensing lm.
Alternatively, the observed increase in sensor response aer leaching may be attributed to imparting specic surface properties to the material through the removal of external Pd clusters, which distinguish it from pure SnO 2 .These properties predominantly refer to increased surface catalytic activity.A similar technique to the leaching method employed in our study, aimed at removing surface-decorating metal clusters, was utilized by Yang and Flytzani-Stephanopoulos et al. to obtain single atoms of gold on CeO 2 catalyst surfaces by treating them with a solution of NaCN. 41This treatment effectively eliminated the majority of metallic Au, leaving behind cationic Au atoms that formed strong bonds with CeO 2 changing the surface properties of the catalyst.2][43] As have been seen from XPS analysis (Fig. S14, ESI, †), the near-surface region of the particles still contained oxidized and metallic Pd.The Pd ions within the SnO 2 lattice can provide initial adsorption sites for analyte molecules offering free valences and, by that, lowering their energy barrier of adsorption. 27Single Pd ions directly activate lattice oxygen which is not reactive at the pure SnO 2 surface. 44,45nclusions Tin oxide particles embedded with Pd were prepared by ame spray pyrolysis (FSP) of appropriate precursors followed by annealing and leaching metallic Pd from the surface of these particles with nitric acid.The SnO 2 crystal size and the embedded fraction of Pd inside SnO 2 were closely controlled by varying the FSP precursor solution concentration (C), total Pd content and the ratio of precursor solution (ml min −1 ) to dispersion oxygen (l min −1 ) owrate (P/D).The response of sensors made with nominal Pd content of 0.2 and 1 mol% obtained at varying C and P/D was tested towards 1 ppm of acetone and carbon oxide (CO) at 350 °C and 50% relative humidity.The SnO 2 crystal size increased with increasing C and P/D, while it decreased slightly with increasing Pd content.That way, SnO 2 crystals in the range of 11-24 nm were obtained with the embedded palladium fraction ranging from about 20% to Fig. 6 Baseline resistance of pure and 0.2 mol% Pd-doped SnO 2 films as a function of SnO 2 crystal size for particles made at various P/D (Fig. 4).Films of nanoparticles doped with Pd (blue triangles) both on the surface and in the bulk have the highest resistancealmost an order of magnitude higher than that of pure SnO 2 film (green diamonds).Most films of nanoparticles with only embedded Pd (red circles) have resistance significantly lower than the other two SnO 2 films.

Nanoscale Advances Paper
almost 76% of the nominal Pd content.As shown by Scanning Transmission Electron Microscopy (STEM) and Energy Dispersive X-ray Analysis (EDXS), at contents of 0.5 mol% and below, palladium was homogeneously dispersed within and/or on SnO 2 .Palladium domains and clusters smaller than 10 nm are visible at higher Pd contents, e.g. 1 and 3 mol%.The SnO 2 crystal size hardly affected the sensor response in the employed size range.The removal of surface palladium from 0.2 mol% Pd-doped SnO 2 signicantly enhanced (2-6 times) the sensor response to CO and acetone over that of pure SnO 2 sensors and those containing both surface and embedded Pd and greatly exceeds (17-33 times) the response of SnO 2 with the same nominal content (0.2 mol%) of photodeposited Pd.At high nominal Pd content, i.e. 1 mol%, the response of SnO 2 sensors merely embedded with Pd is from one to two orders of magnitude higher than those containing both surface and embedded Pd.
This drastic enhancement of sensor response can be attributed to Pd atoms in the SnO 2 lattice that may change charge carrier concentration and therefore conductance and responsiveness of the lm.Also Pd atoms on/near the SnO 2 particle surface can lower the adsorption energy of analytes while Pd/ PdO clusters may act as nanoelectrodes enhancing the transducing properties of the sensing lm.

Fig. 1
Fig.1shows the as-prepared (open symbols) and annealed (lled symbols) SnO 2 crystal size, d, as a function of the FSP precursor solution concentration, C, for pure (squares) and doped SnO 2 with 0.2 (triangles) and 1 mol% Pd (circles) made with the ratio of precursor solution owrate (P, ml min −1 ) to dispersion O 2 owrate (D, l min −1 ), P/D = 5/5.Increasing C enlarges the SnO 2 crystals.In contrast, the Pd content had no effect on the as-prepared SnO 2 crystals that increased from 5.6-5.8nm to 13.4-14.2nm with C increasing from 0.1 to 1.5 M. In contrast, increasing the Pd content hindered the growth of annealed SnO 2 by the solute drag effect 24 as shown also in Fig.S3in the ESI † in agreement with Pineau et al.13 Given that the particle size as determined by N 2 adsorption (Fig.S3, ESI, † squares) was close to the crystal size (Fig.S3, ESI, † circles) at all Pd contents for both as-prepared and annealed particles, monocrystalline SnO 2 particles were made here.Most interestingly, the size of all as-prepared particles followed a power law (Fig.1, dotted line) with C, d = 12.2C 1/3 , indicating that particle size directly relates to precursor solution concentration, C, and consequently to formation from single droplets by evaporation-precipitation.25 Most likely, such droplets were formed by microexplosions.26So SnO 2 formation takes place largely by droplet-to-particle conversion at the highly oxidic conditions of spray combustion.Increasing the ratio of FSP precursor solution to dispersion O 2 owrate, P/D, increases the precursor aerosol concentration and ame temperature (Fig.S4in the ESI †) that increase SnO 2

Fig. 2
Fig. 2 The embedded Pd fraction as a function of SnO 2 crystal size in the annealed powders made by varying one of the FSP process variables, P/D (stars and circles), C (diamonds and triangles) or Pd content (squares), while keeping the other two constant.The actual embedded Pd content (mol%) as a function crystal size is provided in Fig. S7 in the ESI.†

Fig. 3
Fig. 3 (a) Top panel: Pd elemental maps of annealed SnO 2 doped with Pd contents of 0, 0.2, 0.5, 1 & 3 mol% and made at C = 0.5 M, P/D = 5/5.Bottom panel: EDXS spectra corresponding to the image area above.All images were recorded at the same imaging conditions (450 kx magnification, beam current of 8 nA, 10 frames at 50 ms dwell time).As the content of Pd increases, more and larger Pd-rich (bright) spots are visible.(b) Top panel: Pd elemental maps of SnO 2 containing originally 0.2, 1 & 3 mol% Pd after undergoing leaching with HNO 3 .Only in the 3 mol% sample some Pd-rich spots are visible.The bottom panel: EDXS spectra corresponding to the image area above.Clearly, the amount of Pd in relation to Sn has decreased compared to these samples before leaching and corresponds to the embedded Pd into SnO 2 .

Fig. 4
Fig. 4 The response to 1 ppm of acetone at 350 °C in 50% RH by sensors made with pure (green diamonds) and 0.2 mol% Pd-containing SnO 2 particles before (blue triangles) and after HNO 3 solution leaching and removal of surface Pd (red circles) made at various FSP precursor solution to dispersion O 2 flowrate ratio (starting from the left P/D = 1/5, 5/8, 5/5, 9/5, 5/2) and C = 1 M as a function of SnO 2 crystal size.The star shows the response of a SnO 2 sensor that contained 0.2 mol% photodeposited Pd. 11 Schematics of sensing particles depict with gray the SnO 2 and black the Pd/PdO x atoms, clusters or domains.

Fig. 5
Fig.5The response to 1 ppm of acetone at 350 °C in 50% RH by pure (green square,13 diamondsresponses from Fig.4) and 1 mol% Pdcontaining SnO 2 sensors before (blue triangles) and after leaching and removal of surface Pd (red circles) made at various FSP precursor solution concentrations (starting from the left C = 0.1, 0.5, 1, 1.5 M) and i/D = 5/5 as a function of SnO 2 crystal size.The star shows the response of a sensor that contained 1 mol% photodeposited Pd onto SnO 2 particles by FSP at C = 0.5 M and same P/D.11