Back to: Mochizuki Research Group Education — Inorganic Materials Science > Virtual Experiments, Inorganic Materials Science resources
This page is the explanation that pairs with synthesis-simulator-en.html. For the four ways of making inorganic materials (solid-state synthesis, coprecipitation as a solution route, chemical vapor deposition (CVD), and physical vapor deposition (PVD)), it summarizes in what state the raw materials mix, along what path the matter moves, and where the reaction and deposition take place, together with the reaction equations, formulas, units and sources. How to read the simulator screen and what the controls mean are explained in Appendix A, the provenance of the numbers used (literature values, empirical equations, estimates, schematic models) in §6, and the points that should be checked against Prof. Mochizuki’s laboratory materials in §7.
There are many ways to make ceramic powders and thin films, but textbooks usually divide them into “solid-state synthesis,” “solution synthesis” and “vapor-phase methods.” What matters here is that the criterion for the division is not “whether a liquid is used.” Ethanol is used in the wet mixing of solid-state synthesis, but ethanol is a dispersion medium that disperses the powder and breaks up lumps; it does not dissolve the raw materials. In solution synthesis (coprecipitation), the raw materials are dissolved in water, mixed in the form of ions, and then the precursor is precipitated in the solution.
| Method | State of the raw material | How matter moves | Where reaction and deposition occur | Form obtained | Example in the simulator |
|---|---|---|---|---|---|
| ① Solid-state synthesis | Powder (solid) | Diffusion inside the solid through the contacts between particles (µm) | From the contacts between particles (the reaction interface) inward | Calcined powder → pressed and sintered into a sintered body | BaCO₃ + 6 Fe₂O₃ → BaFe₁₂O₁₉ + CO₂ |
| ② Solution synthesis (coprecipitation) | Ions in aqueous solution | Diffusion of the ions in solution, and stirring. Precipitation once supersaturated | In the solution (nucleation and growth), then inside the precursor during calcination | Precipitated precursor → fine powder after calcination | Fe³⁺, Ba²⁺ + OH⁻, CO₃²⁻ → hydrous iron oxide + BaCO₃ |
| ③ CVD | Source gas | Gas-phase transport (diffusion across the boundary layer) and surface reaction | The surface of the heated substrate (chemical reaction) | Thin film. The by-products are pumped out as gas | SiH₄(g) → Si(s) + 2H₂(g) |
| ④ PVD | Solid evaporation source or target | Atoms are released by evaporation (heat) or sputtering (ion bombardment) and fly through the vacuum | The surface of the substrate (sticking and surface diffusion) | Thin film. A chemical reaction is not essential | Vacuum evaporation of Al; sputtering of Cu and Ti |
① and ② are matched to the same target material (M-type barium ferrite BaFe₁₂O₁₉), so the progress of calcination can be compared with the same diffusion-controlled equation. ③ and ④ are both vapor-phase methods that make thin films, but they differ in that in CVD the source gas undergoes a chemical reaction on the substrate surface and becomes a film, whereas in PVD atoms ejected from a solid stick to the substrate and become a film. Since PVD does not necessarily involve a chemical reaction, the umbrella heading for ③ and ④ is “vapor-phase methods (vapor deposition methods).”
After this, in actual synthesis the powder is reground, pressed into shape and sintered (for the difference in purpose between calcination and sintering, see Sintering ① and YBCO synthesis). Tab ① of this simulator deals only with the part up to calcination, “mixing the powders and letting them react.”
The molar masses (calculated from the 2021 IUPAC atomic weights) are BaCO₃ 197.335, Fe₂O₃ 159.687, BaFe₁₂O₁₉ 1111.448 and CO₂ 44.009 g/mol. The left-hand side, 197.335 + 6 × 159.687 = 1155.457 g, and the right-hand side, 1111.448 + 44.009 = 1155.457 g, agree, and when the reaction goes to completion in the stoichiometric ratio (Fe/Ba = 12), the theoretical mass loss is 44.009 / 1155.457 = 3.81 %. For a total mass of 5.00 g, the weighed amounts are 0.854 g of BaCO₃ and 4.146 g of Fe₂O₃.
If the feed ratio R = Fe/Ba deviates, the reaction stops when the limiting raw material is used up, and the surplus raw material remains. In the simulator, α is the fraction of BaCO₃ that has reacted, and the remaining raw materials, the products and the CO₂ released are calculated from the conservation of the elements. For R ≥ 12 only Eq. (1) proceeds (nBaM = αnBa), and the surplus Fe₂O₃ remains. For R < 12, the surplus Ba is assumed to form a secondary phase through BaCO₃ + Fe₂O₃ → BaFe₂O₄ + CO₂, and from the conservation of Ba (x + y = αnBa) and of Fe (12x + 2y = αRnBa) the amounts of BaFe₁₂O₁₉ and BaFe₂O₄ are allotted as x = α(R − 2)nBa/10 and y = α(12 − R)nBa/10 (BaFe₂O₄ is 313.013 g/mol). In either case the CO₂ is αnBa. For example, when a total of 5.00 g reacts completely, 0.146 g of BaFe₂O₄ remains at R = 11 and 0.314 g of BaFe₂O₄ at R = 10 (BaFe₁₂O₁₉ 4.465 g, mass loss 4.42 %), and 0.323 g of α-Fe₂O₃ remains at R = 13. In the actual pathway BaFe₂O₄ forms first as an intermediate phase and then reacts with Fe₂O₃ to give BaFe₁₂O₁₉, but this intermediate stage is omitted (only the Ba left over is counted as a secondary phase at the end). For this reason the release of CO₂ (the mass loss) is also treated schematically as proportional to α; in reality most of the CO₂ escapes at the stage where BaFe₂O₄ forms (700–800 °C), and the mass loss proceeds faster than α.
However, “exactly the stoichiometric 12 is the best” is not necessarily true. Even in solid-state synthesis, α-Fe₂O₃ tends to remain at Fe/Ba ≥ 12, and an optimum near 11.7 has been reported, and at Fe/Ba < 11 BaFe₂O₄ is said to coexist below 1000 °C (the review in Ref. 20 and the papers cited there). For this reason the simulator labels a slight Ba excess at Fe/Ba = 11–12 as “a composition that is also used in practice (the surplus remains as a small amount of secondary phase),” and for the failure-example presets it uses Fe/Ba = 10 (a large excess of Ba, leaving the secondary phase BaFe₂O₄) and 13 (α-Fe₂O₃ remains). The feed ratios of Fe/Ba = 10–11.5 often used in the coprecipitation of ② are thought to compensate for the Ba lost in washing (§3.2; the reasoning includes inference; provisional). In solid-state synthesis no Ba is lost, so the surplus Ba becomes the secondary phase as it is.
Suppose a spherical particle of radius r (here Fe₂O₃) is surrounded by the other raw material (BaCO₃). Once a product layer has formed at the contact, the raw material (Ba²⁺, O²⁻, etc.) must diffuse through the product layer for the reaction to continue. If the thickness of the layer is y, the flux through the layer is proportional to 1/y, so approximating the layer as a flat plate gives dy/dt = D/y, that is, y² = 2Dt (the parabolic law). Substituting the relation between the degree of reaction α (the fraction of the volume that has reacted) and the layer thickness, y = r[1 − (1 − α)1/3], gives the Jander equation (Ref. 1).
The Jander equation treats the layer as a flat plate, so it stops fitting as the reaction proceeds. Solving the steady-state diffusion in a spherical shell correctly gives the Ginstling–Brounshtein (GB) equation (Ref. 2). The simulator uses the GB equation by default and lets you choose the Jander equation for comparison.
The left-hand side becomes 1/3 at α = 1, so the reaction is complete at kt ≥ 1/3. Eq. (3) has two key points. (i) The time the reaction takes is proportional to reff²: halving the particle size makes it 4 times faster. This is the reason for “making the powder fine and mixing it well.” (ii) The temperature acts exponentially, through the Arrhenius equation for the diffusion coefficient D. The simulator uses D0 = 1.56 × 10−3 cm²/s and Ea = 250 kJ/mol. These are calibrated values (estimates) chosen in the barium ferrite model of the Virtual Laboratory (in preparation) so that “a coprecipitated precursor of radius 0.3 µm reaches α = 0.90 at 900 °C for 2 h”; they are not measured diffusion coefficients. A check in Node.js gives α = 0.900 at 900 °C for 2 h with r = 0.3 µm and h = 1, which reproduces the calibration condition.
Fig. 1. The degree of reaction for the Jander and GB equations (the horizontal axis is kt). With the GB equation the reaction is complete at kt = 1/3. The later the stage of the reaction, the thicker the layer and the slower the progress.
The effect of how the powders are mixed is expressed by the homogeneity h (0 to 1, a schematic quantity) and the particle radius r. With grinding time t, the homogeneity and the particle size approach their limiting values according to the following equations (the equations and constants of the mortar model of the Virtual Laboratory, carried over; estimates).
| Quantity | Wet (ethanol > 0.05 mL) | Dry | Note |
|---|---|---|---|
| Time constant of homogeneity τH | 8 min | 20 min | Ethanol loosens the agglomerates, so the mixture becomes uniform faster |
| Time constant of particle size τR | 16 min | 22 min | |
| Lower limit of particle size rmin | 0.6 µm | 0.7 µm | |
| Before mixing: h0 / r0 | 0.30 / 3.0 µm | h0 is the value of the preset in the Virtual Laboratory. r0 is back-calculated from “1.29 µm after 20 min of wet grinding” | |
| Evaporation of ethanol | 0.055 mL/min during grinding; 0.10 mL/min in the 60 °C drying oven | If the ethanol runs out during grinding, the process becomes dry partway. The value for the oven is an estimate (provisional; chosen so that the upper limit of 3 mL evaporates in 30 min), and drying is held for 30 min (lab manual §6.1) | |
reff = r(2 − h) is an empirical assumption expressing that “the poorer the mixing, the longer the distance to a particle of the other kind”; it is the same as in the Virtual Laboratory model. In the mortar α is kept at 0, so “the compound forms the moment the powders are mixed” never happens.
| Calcination temperature | Wet, 20 min | Dry, 5 min | Coprecipitated precursor (②) |
|---|---|---|---|
| 800 °C | 0.084 | 0.030 | 0.325 |
| 900 °C | 0.266 | 0.097 | 0.849 |
| 1000 °C | 0.639 | 0.254 | 1.000 |
| 1100 °C | 1.000 | 0.546 | 1.000 |
| 1200 °C | 1.000 | 0.919 | 1.000 |
The general trend “coprecipitation (a chemical method) at around 900 °C, solid-state synthesis (the ceramic method) at 1100–1200 °C” (for example, the report on a coprecipitated precursor in Ref. 6) appears from the difference in diffusion distance alone. The calcination of the precursor in ② coprecipitation is calculated with the same equation (§3.6). However, the absolute values of the temperature depend on the calibrated values, so please read how the cases compare (how many times faster, which finishes first).
In coprecipitation, solution B, in which the salts of Fe and Ba are dissolved in water, is added to solution A of the base (NaOH and Na₂CO₃). In solution, Fe³⁺ and Ba²⁺ are mixed as ions, so at the moment of precipitation the two lie side by side at a distance of nm to 0.1 µm. The length scale of mixing differs by one to two orders of magnitude from that of solid-state synthesis, in which µm-sized particles are brought together at their contacts. However, in the calculation of calcination in ②, the radius of the agglomerated particles (about 0.32 µm as the standard), not the distance over which the ions mix, is used as the particle radius r in the GB equation (§3.5, §3.6). The diffusion distance reff shown in the readout is therefore 0.33 µm for the solution route and 1.36 µm for the standard solid-state case, so the difference in the model is about a factor of 4. Note also that the precipitate immediately after coprecipitation is not BaFe₁₂O₁₉. The Fe side is amorphous hydrous iron oxide (ferrihydrite) and the Ba side is BaCO₃, and the target crystalline phase forms only on calcination (Refs. 5, 6). Tab ② of this simulator shows only the key points of dropwise addition, settling, decantation washing with warm water, and collection; the whole process, including the number of washes and the pH, drying and calcination, is treated in Coprecipitation of barium ferrite.
Solution A: 12.5 g of NaOH (312.5 mmol) and 3.15 g of Na₂CO₃ (29.72 mmol) in 100 mL of distilled water. Solution B: 10.14 g of FeCl₃·6H₂O (37.52 mmol) and 0.835 g of BaCl₂·2H₂O (3.419 mmol) in 50 mL of distilled water. The feed ratio is Fe/Ba = 10.97, which has 9 % more Ba than the stoichiometric 12 (of the 3.419 mmol of Ba, about 0.29 mmol is the amount exceeding stoichiometry). In coprecipitation it is customary to make the Fe/Ba of the starting solution smaller than 12, and many examples of 10–11.5 have been reported (Refs. 5, 20). The reason is thought to be to make up for the tendency of Ba²⁺ to be lost with the supernatant during washing and to avoid leaving surplus Fe as α-Fe₂O₃ (the loss of Ba includes inference from the literature and from chemistry; provisional). Ba that is still left over after washing may remain as a small amount of a secondary phase such as BaFe₂O₄ after calcination. Tab ② shows the washing steps, but the change in the composition of the precipitate caused by washing is not calculated, so the Fe/Ba of the solid stays at the feed ratio of 10.97, and this caveat is displayed in the one-line comment on the screen. In the calculation for calcination too, as when Fe/Ba is made smaller than 12 in ①, the portion in which the surplus Ba becomes BaFe₂O₄ is subtracted (§2.2, §3.6). These masses and amounts of water are the values in §8.1 of the 2025 edition of the lab manual for Materials Science Experiments (Inorganic Materials) of the former Tokyo Institute of Technology (they also agree with the values in the Virtual Laboratory).
In the lab course, solution B is added little by little to solution A while A is stirred well. In the lab video, B is poured along a glass rod in several portions, and the mixture is stirred with the glass rod in between. After the precipitate has settled, the supernatant is removed with a Komagome pipette until about 125 mL of liquid remains, warm water at about 60 °C from a hot plate is added up to about 300 mL, and the mixture is stirred and left to stand (the warm water is added little by little with a 20 mL Komagome pipette in lab manual §8.1, and poured directly from a 500 mL beaker in the lab video; the figure in the simulator follows the video). This is repeated until the pH of the supernatant (pH test paper) reaches about 7 (decantation washing), followed by suction filtration and drying at 105 °C (lab manual §8.1). Calcination is at 925 °C for 2 h (§8.2), and the default of ② in the simulator was also set to this condition. Tab ② of the simulator draws the repeated washing as a single washing motion and does not calculate the number of washes or the change in pH (these are treated in tab ③ of the coprecipitation simulator).
The dropwise addition is divided into 60 steps, and, assuming that the solution reaches equilibrium after each addition, the pH and the amount of precipitate are calculated. The pH is not an input; it is determined by the charge balance from the amounts of reagents added.
The amounts of Fe³⁺ and Ba²⁺ that can remain dissolved are determined by the solubility products. qFe = 3 − Kh/(Kh + [H⁺]) is the average charge that includes the first hydrolysis of Fe³⁺ (Kh = 6.3 × 10−3), and the dissolved carbonate is partitioned into H₂CO₃*, HCO₃⁻ and CO₃²⁻ with Ka1 = 4.45 × 10−7 and Ka2 = 4.69 × 10−11.
The values used are Ksp(Fe(OH)₃) = 2.79 × 10−39 and Ksp(BaCO₃) = 2.58 × 10−9 (CRC Handbook, 25 °C, Ref. 4). The Virtual Laboratory model uses 4 × 10−36 and 8.1 × 10−9, respectively (the given conditions of the problem in lab manual §8.1), so there is a difference of three orders of magnitude for Fe(OH)₃. At high pH, almost all the Fe precipitates in either case, so the result is hardly affected, but it matters, through √Ksp, for the dissolution of BaCO₃ during washing (an item to confirm in §7). The amount of dissolved BaCO₃ is found by solving the quadratic equation b·α2(CT − BT + b) = Ksp, where b is the concentration of dissolved Ba (CT and BT are the total concentrations of carbonate and Ba, and α2 is the fraction present as CO₃²⁻). Carbonate that has become H₂CO₃* in an acidic solution is assumed to escape from the solution as CO₂ (a schematic of an open system).
| Condition | pH | Precipitation of Fe | Precipitation of Ba | Fe/Ba of the solid | Note |
|---|---|---|---|---|---|
| Solution A only / solution B only | 14.49 / 1.18 | Agrees with the values of the Virtual Laboratory | |||
| Standard (B into A) | 14.12 | 100 % | 100 % | 10.97 | Precursor 4.01 g (as FeOOH + BaCO₃) |
| Reverse direction (A into B) | 14.16 | 100 % | 100 % | 10.97 | 8.9 mmol of carbonate escapes as CO₂. Fe precipitates first and Ba later (simultaneity 0.79, §3.4) |
| Na₂CO₃ 0.2 g | 14.1 | 100 % | 55 % | 19.9 | Not enough carbonate, so the Ba does not fully precipitate (composition deviation) |
| Na₂CO₃ 0.36 g / 0.5 g or more | 100 % | 99.3 % / 100 % | 11.05 / 10.97 | ||
| No NaOH | 1.46 | 55 % | 0 % | — | The only base is that from the Na₂CO₃. Half of the Fe remains dissolved |
In this model Ba precipitates only as BaCO₃. There is a report that Ba precipitates even under the high-pH condition with NaOH alone (Zahn et al. 2024, Ref. 5), and BaCO₃ may also form by absorbing CO₂ from the air, so one cannot assert that “Ba never precipitates without carbonate.” The note on the screen says so as well.
Whether a precipitate forms is determined by the supersaturation ratio S (the ratio of the ion activity product, IAP, to the solubility product). If S > 1, the solid is the more stable state, nuclei form, and a precipitate is born. If S ≤ 1, nothing precipitates.
In classical nucleation theory, the nucleation barrier ΔG* decreases in inverse proportion to (ln S)² (γ is the interfacial energy). The larger the supersaturation, the more nuclei form at once, and the finer the particles become. The “log₁₀ S at the moment of addition” shown by the simulator is a value calculated on the assumption that each added portion of liquid mixes throughout the whole liquid before it precipitates. In the standard direction (B into A), log₁₀ S ≈ 37 for Fe(OH)₃ and ≈ 4.5 for BaCO₃, and both precipitate at the same time. In the reverse direction (A into B), the solution is acidic at first, so log₁₀ S of Fe(OH)₃ is only on the order of a few and that of BaCO₃ is negative (undersaturated); Fe precipitates first, and Ba precipitates only after the base is in excess. This “offset in the timing of precipitation” is expressed by the simultaneity hsync = 1 − ⟨|FFe − FBa|⟩ (F is the fraction of the final amount of precipitate that has precipitated up to that point) (an index of our own). The rate of nucleation itself (the kinetics of precipitation) is not calculated.
For how the stirring strength (in rpm equivalent) and the pouring speed act, the estimated equations of the Virtual Laboratory model were used as they are: the score for stirring strength s = log₁₀(1 + rpm/60)/log₁₀ 11, the score for pouring speed q = 1 − log₁₀(speed/3 mL min⁻¹)/1.6 (0 to 1), the mixing coefficient (0.45 + 0.35s + 0.30q)/1.10, and the radius of the agglomerated particles 0.30 µm/(0.55 + 0.45sq) (2.6 times larger for the reverse direction; hereafter called the “agglomerate size,” but it is used as the particle radius r in the GB equation). The homogeneity is the product of hsync and the mixing coefficient. None of them is an equation fitted to measurements; each is an estimate that shows a trend. In the lab course the mixture is stirred with a glass rod, so the rpm is used as a guide (a condition for comparison) converted to the rotation speed of a stirrer, in order to compare how well the mixing goes.
The degree of reaction when the collected precursor is calcined is calculated with the GB equation of §2.3, with the particle radius r taken as the agglomerate size and the homogeneity as the value above. The dehydration of the hydrous iron oxide and the pathway through BaFe₂O₄ are omitted, so this is a comparison for seeing “how large a difference the diffusion distance alone makes”. At 925 °C for 2 h (the default of ②), α = 0.97 for the solution-route precursor and α = 0.34 for solid-state synthesis (wet, 20 min) (0.85 and 0.27 at 900 °C for 2 h). The screen displays side by side the values for the solution-route precursor and for the powder of ①, calculated under the present calcination conditions. However, α is the progress of the reaction between the raw materials present in the solid, so the fraction of Fe that can become BaFe₁₂O₁₉ has an upper limit set by the Fe/Ba (R) in the solid. The way of counting is the same as in ① (§2.2): for R ≥ 12 it is 12/R (Ba is insufficient and the remaining Fe is α-Fe₂O₃), and for R < 12 it is 12(R − 2)/(10R) (the Fe taken up by the surplus Ba forming BaFe₂O₄ is subtracted). The bar graph and the readout on the screen show “the fraction of Fe that becomes BaFe₁₂O₁₉ = α × upper limit.” For Na₂CO₃ 0.2 g (Fe/Ba of the solid = 19.9) the upper limit is 60 %, and even with α = 0.89 at 925 °C for 2 h it stays at 54 %, the rest becoming α-Fe₂O₃. Without NaOH, Ba does not precipitate, so the upper limit is 0 and no BaFe₁₂O₁₉ forms even on calcination.
The main example is a polycrystalline Si film by low-pressure CVD (LPCVD). Wafers are stood in a row in a quartz reaction tube and the whole is heated by a heater outside the tube (a hot-wall furnace), while SiH₄ is flowed at a low pressure of about 0.2–1 Torr. Typical conditions are 575–650 °C and a growth rate of about 10–100 nm/min (Ref. 8). The by-product H₂ and the unreacted SiH₄ are exhausted and treated in an abatement system (SiH₄ ignites spontaneously in air). Changing the source system produces various films by the same mechanism (Table 5).
| Film | Reaction | Rough temperature |
|---|---|---|
| Polycrystalline Si | SiH₄ → Si + 2H₂ | 575–650 °C (LPCVD) |
| SiO₂ | SiH₄ + O₂ → SiO₂ + 2H₂ | about 450 °C |
| Si₃N₄ | 3SiH₂Cl₂ + 4NH₃ → Si₃N₄ + 6H₂ + 6HCl | about 750 °C |
| Epitaxial Si | SiCl₄ + 2H₂ → Si + 4HCl | about 1200 °C |
| W | WF₆ + 3H₂ → W + 6HF | 400 °C or below |
| TiC (coating for tools) | TiCl₄ + CH₄ → TiC + 4HCl | 950–1050 °C (a representative example from patents and the like) |
Near the surface of the substrate there is a boundary layer in which the gas flow is slow. The source material is carried across the boundary layer by diffusion (flux J1) and reacts on the surface to become a film (flux J2). Let the gas-phase concentration be Cg, the surface concentration Cs, the mass-transfer coefficient hg, and the rate constant of the surface reaction ks (Refs. 7, 8). Then
Nf is the atomic number density of the film (5.0 × 1028 m−3 for Si). At low temperature ks ≪ hg, so that G ≈ ksCg/Nf (surface-reaction-limited): the growth rate depends exponentially on temperature and gives a straight line in an Arrhenius plot. The surface concentration is almost the same as in the gas phase, so the film is deposited with the same thickness on every wafer. At high temperature hg ≪ ks, so that G ≈ hgCg/Nf (transport-limited): the dependence on temperature is weak, and the film thickness tends to become uneven depending on how the gas flows. In LPCVD the pressure is low, so the gas-phase diffusion coefficient (∝ 1/p) is large and the surface-reaction-limited regime extends toward the high-temperature side. This is one of the reasons for reducing the pressure.
We rewrite Eq. (9) in terms of partial pressure, add how much of the flowing source is consumed (the supply of the source), and express the growth rate G (nm/s) as three resistances in series. The first two (surface reaction and transport) can be derived from the Grove model, but the form in which the third term, the supply term c p, is added as a resistance in series is a schematic constructed as a rough guide to the effect of the source being depleted (it cannot be derived from the Grove model).
G is the growth rate averaged over the whole deposition area, and the mass balance between the conversion X and the “Si that became film” is also calculated with this average. The film thickness on each wafer is a rough guide: it thins linearly toward the downstream end, with the downstream/upstream ratio equal to 1 − X and the average over the 6 wafers equal to G, which gives G/(1 − X/2) upstream and G(1 − X)/(1 − X/2) downstream (the “film thickness” on the screen is the average of the 6 wafers). Strictly, if the source is consumed along the flow by a first-order reaction, the film thickness thins exponentially toward the downstream end, and the conversion is 1 − exp(−G0c) (G0 is the local rate p/(1/ks + 1/hg) determined by the partial pressure at the inlet). The series form underestimates the depletion compared with this (example: at 750 °C, 0.3 Torr and 5 sccm, X = 0.49 for the series form and 0.61 for the plug-flow equation). Here the series form was used as a rough guide. Also, when the conditions are changed partway through deposition, what is deposited under the conditions at each moment is accumulated, so the thickness and structure (§4.4) of the film already deposited do not change afterward.
Mass balance: the number of Si atoms that became film is equal to the number of SiH₄ molecules consumed, and twice as many H₂ molecules are produced. Under the standard conditions (620 °C, 0.30 Torr, SiH₄ 100 %, 100 sccm), G = 11.7 nm/min, consumption 1.09 sccm, H₂ 2.18 sccm and Si 82 mg/h, and 1.088 sccm × 4.478 × 1017 s−1 ÷ (5.0 × 1028 m−3 × 0.05 m²) = 0.195 nm/s = 11.7 nm/min agrees with this.
Fig. 2. Arrhenius plot of Eq. (10) (total pressures 0.3 Torr and 10 Torr, SiH₄ 100 %, 100 sccm). The dotted line is for the surface reaction alone and the dashed line for transport alone. At a higher total pressure, the temperature at which the process switches to transport limitation is lower.
| Temperature | G (nm/min) | Rate-limiting step | Cs/Cg | Conversion X |
|---|---|---|---|---|
| 560 °C | 3.31 | Surface reaction | 0.990 | 0.3 % |
| 620 °C | 11.7 | Surface reaction | 0.967 | 1.1 % |
| 700 °C | 45.0 | Surface reaction | 0.878 | 4.2 % |
| 800 °C | 136 | Surface reaction (transport begins to matter) | 0.624 | 12.6 % |
| 900 °C | 232 | Transport | 0.332 | 21.5 % |
| 750 °C, 10 Torr, SiH₄ 5 % | 13.5 | Transport | 0.09 | 25 % |
| 750 °C, 0.3 Torr, 5 sccm | 26.2 | Supply (consumption of the source) | 0.43 | 49 % |
Si films made by LPCVD are said to be amorphous below about 580 °C and polycrystalline (columnar grains) at 600 °C and above. The simulator only displays this boundary as a rough guide, and it has not been checked against the numerical values in the original source (§6). Thermal CVD drives the reaction with heat, whereas plasma CVD (PECVD) creates highly reactive species by electron collisions and can therefore deposit films at lower temperatures (Ref. 16). ALD (atomic layer deposition) is a relative of CVD that repeats self-limiting surface reactions, half a cycle at a time.
When a material is heated in a vacuum, atoms leave the surface thermally. The evaporation flux per unit area and unit time is given by the Hertz–Knudsen equation (Ref. 10).
p* is the saturated vapor pressure, and αv is the evaporation coefficient (taken as 1 for a clean surface). The vapor pressure of Al is interpolated segment by segment as straight lines in log p versus 1/T from the CRC Handbook table (the temperatures at which it reaches 1, 10, 100 and 1000 Pa are 1482, 1632, 1817 and 2054 K; Ref. 4), and outside the table range (below 1482 K and above 2054 K) it is extrapolated from the end segment and labeled “extrapolated.” Check: at 1482 K (1 Pa), J = 1.32 × 1022 m−2 s−1.
Atoms leaving a small planar evaporation source are emitted in proportion to cos φ, where φ is the angle from the normal to the surface (the cosine law). Consider a point at a distance x from the center on a flat substrate at a distance h directly above the evaporation source. The distance is given by r² = h² + x², and the cosines of the emission angle and the incidence angle are both h/r, so
The deposition rate at the center is J Ae/(πh²) multiplied by the atomic volume Ω = M/(ρNA) (of the total flux J Ae that a planar source of area Ae emits into the half-space, the fraction falling on a unit area directly above is 1/(πh²)). The source area Ae = 1 cm² is provisional. Check: at 1482 K and h = 0.3 m the rate is 0.077 nm/s (4.6 nm/min), and at 1600 K it is 28 nm/min. Moving the substrate farther away makes the film more uniform, but the deposition rate falls as 1/h².
For a molecular diameter d = 0.36 nm and 300 K, the mean free path is 7.2 mm at 1 Pa, 0.72 m at 10−2 Pa and 72 m at 10−4 Pa (the same order as the rule of thumb λ ≈ 5 mm·Pa/p in the TU Wien lectures, Ref. 10). For evaporation, the fraction of flying atoms that reach the substrate without ever hitting the residual gas is taken as e−h/λ (an approximation that uses the gas–gas λ also for metal atoms). The simulator also displays the ratio of the number of residual-gas molecules striking the substrate (Eq. (11) with p* taken as the residual-gas pressure and m as 28 u) to the number of metal atoms. Even at 10−4 Pa the ratio is about 0.1, which is a cause of O₂ and H₂O being incorporated into the film and oxidizing or contaminating it (the sticking probability is not included).
When Ar⁺ ions produced in an Ar glow discharge strike a target to which a negative voltage is applied, a chain of collisions knocks atoms out of the surface (they come from a depth of about 1 nm, and most of them are neutral atoms). The number of atoms knocked out per ion is called the sputtering yield Y; above a threshold (10–30 eV) it increases with the energy of the ion (Refs. 10, 14, 15). The simulator uses representative values for 600 eV Ar⁺ (Cu 2.3, Ti 0.6, Al 1.2; these come from reference tables of equipment vendors, with no original source cited and an uncertainty of ±20–30 %, Ref. 17), and makes the approximation that the yield is proportional to the energy in excess of the 25 eV threshold, Y(E) = Y600(E − 25 eV)/(575 eV) (the 25 eV threshold is provisional). The ion energy is assumed to be the discharge voltage × the elementary charge. The rate at which the target is eroded is given by
where j is the ion current density. Because Y for metals is around 1, the rates differ less from one material to another than in evaporation, which makes sputtering suitable for alloy and multicomponent films. The flux reaching the center of the substrate is obtained by multiplying by the on-axis value Rt²/(Rt² + h²) for a disk-shaped source of radius Rt = 25 mm (provisional).
Sputtering is carried out in Ar at 0.1 to a few Pa, so the flying atoms collide with Ar. Eq. (13) is the value for thermally moving gas molecules colliding with each other, and the √2 comes from the relative velocity. When a fast atom of several eV or more travels through nearly stationary Ar, the √2 does not appear, and the collision cross section also becomes smaller as the energy increases, so the mean free path is longer than that given by Eq. (13). In the simulator we use λsp = kBT/(0.5 πd²p) (an approximation that drops the √2 and takes the cross section as half that for gas–gas collisions; the coefficient 0.5 is provisional), which gives 41 mm at 0.5 Pa. This is of the same order as the mean free path of Cu at 0.5 Pa, about 40–110 mm, given by a Monte Carlo calculation that uses energy-dependent cross sections (Zhu et al. 2022, Ref. 19). We take the mean number of collisions as n = h/λsp, the fraction of energy remaining after one collision as f = 1 − 2M1M2/(M1 + M2)² (the hard-sphere average; 0.53 for Cu and Ar), the mean energy at ejection as E0 = 10 eV (a provisional value chosen from the representative range of 5–40 eV), and, assuming that the number of collisions follows a Poisson distribution with mean n, we estimate the mean energy on arrival as E0 Σk P(k) fk = E0 exp[−n(1 − f)] (the lower limit is the thermal energy of the gas, 0.026 eV). According to this expression, the distance over which Cu cools (thermalizes) to 0.1 eV is about 40 cm at 0.5 Pa (p·d ≈ 20 Pa·cm). At 0.5 Pa and 60 mm there are 1.5 collisions on average, and 77 % of the atoms collide at least once (in the Poisson distribution P(0) = 0.23, P(1) = 0.34, P(2) = 0.25); since each collision costs 47 % of the energy on average, the mean energy on arrival is about half the energy at ejection (≈ 5 eV). Even so, this is an order of magnitude larger than for evaporated atoms (about 0.3 eV), and this bombardment makes the film dense (Zone T). At 4 Pa the atoms are thermalized and arrive obliquely (Zone 1). The fraction η that reaches the substrate is a schematic model: the sum of the part that arrives without colliding, e−h/λ, and the scattered part, (1 − e−h/λ) × 0.5 × min(1, 3λ/h). The 0.5 is the approximate fraction of the atoms, whose directions have been randomized by scattering, that head toward the substrate, and min(1, 3λ/h) expresses that when the pressure is high and the transport becomes diffusive, the atoms are less likely to arrive (both are provisional). As the pressure is raised, the deposition rate falls gradually (85 → 60 → 12 nm/min for 0.1 Pa → 0.5 Pa → 4 Pa), and the atoms arrive obliquely and with lower energy. The lower limit of the pressure at which a magnetron can sustain the discharge is taken as 0.1 Pa (provisional); below that, the simulator reports “no discharge.”
| Target / Ar pressure | Y | λsp | h/λsp | Arriving without collision | Fraction arriving | Deposition rate | Mean arrival energy |
|---|---|---|---|---|---|---|---|
| Cu / 0.1 Pa | 1.50 | 203 mm | 0.29 | 75 % | 87 % | 85.4 nm/min | 8.7 eV |
| Cu / 0.5 Pa | 1.50 | 40.7 mm | 1.47 | 23 % | 61 % | 60.1 nm/min | 5.0 eV |
| Cu / 1 Pa | 1.50 | 20.3 mm | 2.95 | 5 % | 53 % | 51.5 nm/min | 2.5 eV |
| Cu / 4 Pa | 1.50 | 5.09 mm | 11.8 | 0 % | 13 % | 12.4 nm/min | 0.04 eV |
| Ti / 0.5 Pa | 0.39 | 40.7 mm | 1.47 | 23 % | 61 % | 23.5 nm/min | 4.8 eV |
Movchan and Demchishin organized the microstructure of evaporated films by the ratio of the substrate temperature to the melting point of the film material, Ts/Tm (Ref. 13), and Thornton extended this to sputtered films by adding an axis for the Ar pressure (Ref. 12). The boundaries in the simulator follow Kusano’s compilation (Ref. 11): the values in the table are used at 0.5 Pa and below and at 4 Pa and above, and in between they are interpolated in log p.
| Zone | Microstructure and the dominant process | 0.5 Pa and below | 4 Pa |
|---|---|---|---|
| Zone 1 | Tapered columns with many voids. Shadowing by oblique incidence and insufficient surface diffusion | < 0.15 | < 0.4 |
| Zone T | Dense fibrous structure with a flat surface. Surface diffusion and bombardment by high-energy particles (specific to sputtered films) | 0.15–0.325 | 0.4–0.525 |
| Zone 2 | Crystalline columnar grains. Governed by surface and grain-boundary diffusion | 0.325–0.7 | 0.525–0.7 |
| Zone 3 | Recrystallized equiaxed grains. Governed by bulk diffusion | > 0.7 | > 0.7 |
For the boundary values, the middle of the range given in the literature tables (e.g. 0.1–0.2) was used. The boundaries are not sharp. The boundaries for evaporated films (0.3 and 0.5) are a rough guide that has not been checked against the original source, and the simulator draws them with no Zone T appearing in evaporated films. Reactive sputtering (mixing O₂ or N₂ into the Ar to make oxide or nitride films) should be treated separately from ordinary PVD, so this simulator does not cover it. Ion beam deposition (IBD) is a method in which the target is sputtered by the beam from an independent ion source; at low pressure there is little scattering by the gas, and dense films are obtained (Ref. 16).
Fig. 3. Film-thickness distribution from the cosine law (Eq. (12)). The smaller the distance h between the evaporation source and the substrate, the thicker the center and the thinner the edge.
| Tab | Quantity | Value | Provenance and source |
|---|---|---|---|
| ① | Molar masses | BaCO₃ 197.335, etc. | Literature values (calculated from the IUPAC 2021 atomic weights) |
| ① | Equation for the degree of reaction | GB (default), Jander | Equations from the literature (Refs. 1, 2) |
| ①② | Diffusion coefficient D0, Ea | 1.56 × 10⁻³ cm²/s, 250 kJ/mol | Estimated and calibrated values (model of Ba ferrite in the Virtual Laboratory). Not to be cited as measured values |
| ① | Mixing time constants, lower limit of particle size, h0, r0 | Table 2 | Estimates (model of the mortar in the Virtual Laboratory). r0 is back-calculated |
| ① | reff = r(2 − h) | — | An empirical assumption (same as in the Virtual Laboratory) |
| ① | Intermediate phase BaFe₂O₄, stages of carbonate decomposition, heating and cooling | Omitted | Simplification. The temperature is held constant at the holding temperature. The release of CO₂ is made proportional to α (schematic; for Fe/Ba < 12 only the surplus Ba is counted as the secondary phase BaFe₂O₄, §2.2) |
| ① | Grinding for about 20 min, drying at 60 °C for 30 min or more | §2.1 | Lab manual §6.1 (the YBCO procedure). The amount of ethanol and the evaporation rate in the drying oven are estimates (provisional) |
| ①② | A slight Ba excess (Fe/Ba = 11–12) is treated as a composition actually used, and the surplus is counted as a secondary phase | §2.2, §3.2 | Trend in the literature (Ref. 20). Part of the reasoning is inference (provisional) |
| ①② | Calcination atmosphere | In air | Prof. Mochizuki’s instruction (2026-09-25). The student lab course is in air (in research, a tube furnace under N₂ flow or an Ar glove box is also used) |
| ② | Reagent masses and amounts of water, volumes for washing (about 125 mL, about 300 mL, about 60 °C, until the pH is about 7), calcination at 925 °C for 2 h | §3.2 | Lab manual §8.1, §8.2 |
| ② | Ksp, Ka1, Ka2, Kh, Kw | §3.3 | Literature values (CRC Handbook and others). Activity coefficients are taken as 1 |
| ② | Coefficients for stirring and pouring speed (rpm is a stirrer-equivalent value, given as a rough guide for comparison), agglomerate size (used as the radius in the GB equation) | §3.5 | Estimates (equations of the Virtual Laboratory) |
| ② | Simultaneity hsync, the way CO₂ escapes | — | An index of our own; schematic |
| ③ | Rate equation for the surface reaction | Eq. (10) | Empirical equation from the literature (Gieske et al. 1977). Extrapolated outside 600–700 °C |
| ③ | Magnitude and dependence of hg | §4.3 | Schematic (the dependence is an approximation based on gas diffusion; the magnitude is provisional) |
| ③ | Deposition area A | 0.05 m² | Provisional |
| ③ | The form in which the supply term (c p) is added as a resistance in series | Eq. (10) | Schematic (a rough guide to depletion of the source; it estimates the depletion to be smaller than the exact plug-flow expression does; §4.3) |
| ③ | Film thickness for each wafer (upstream G/(1 − X/2), downstream G(1 − X)/(1 − X/2)) | §4.3 | Rough guide (linear distribution; the average agrees with the mass balance) |
| ③ | Boundary between amorphous and polycrystalline | 580 °C, 600 °C | Rough guide (general statement; not checked against the original source) |
| ④ | Vapor pressure of Al | §5.1 | Literature values (CRC Handbook). Extrapolated outside the range |
| ④ | Area of the evaporation source, radius of the target | 1 cm², 25 mm | Provisional |
| ④ | Sputtering yield Y600, threshold | Cu 2.3, Ti 0.6, 25 eV | Representative values (vendors’ reference tables, ±20–30 %) and provisional values. Proportionality to the excess above the threshold is an approximation |
| ④ | Mean free path of sputtered atoms (√2 dropped, cross section multiplied by 0.5), fraction arriving (0.5 of the scattered part, and 3λ/h), arrival energy, lower limit of discharge 0.1 Pa | §5.3 | Approximations, schematic models and provisional values (the order of magnitude of λ is checked against Ref. 19) |
| ④ | Boundaries of the structure zones | Table 8 | Literature values (Kusano 2019). The boundaries for evaporated films are a rough guide |
The magnified schematic diagrams inside the simulator (particles, ions, molecules, and the cross section of the film) are schematics drawn from the numbers of the same state; the numbers and sizes of atoms and grains are thinned out and enlarged so that they are easy to see. Time is shown compressed, and the screen states “1 s on screen ≈ how many minutes of the experiment.”
Above the tabs are three routes showing whether the starting materials are kept solid, dissolved in water or turned into gas (press one to go to that tab). At the top of each tab is a four-panel strip, “What is the starting material? → How is it carried or mixed? → Where does it turn solid? → What do you get?”, and the current panel is highlighted as the operations proceed. Below the strip are everyday comparisons (“An everyday comparison” and “Where the comparison breaks down”) and one line each on “Where it is used,” “Strengths” and “Weaknesses” of the method.
Tabs ②–④ have a “View: Simple | Detailed” switch (shared by the three tabs); the default is the Simple view.
The Simple view consists of zoomed-in figures drawn with large labeled particles, a plain-language comment (Now, Why, Next), two graphs, Results (four items), Try it (Experiment 1, Experiment 2, …; “What we found” opens after the run finishes), a glossary and a comparison of the four methods. All of its numbers come from the same calculations as the “Detailed view” described below.
The step buttons carry the same numbers ①–④ as the four panels; the button to press next is highlighted, and buttons that would skip a step cannot be pressed (the reason is shown in one line). On a narrow screen the step buttons are placed right below the figures.
When a condition that can be changed only in the Detailed view (for example the amount of Na₂CO₃ in ②) differs from the standard, the Simple view shows “Current settings: … (standard: …)” and a “Reset to standard settings” button.
Adding ?level=detail to the address opens the page in the Detailed view. Unless stated otherwise, the descriptions in A.2–A.5 below refer to the Detailed view.
In the Detailed view, every tab has a diagram of the experimental apparatus at upper left, a magnified schematic of the interior at upper right, and two graphs below them, followed by the legend, the time scale, the “one-line comment” (what the last operation changed) and the readouts.
All four figures and the numbers are drawn from the same state (one state object per tab).
The control panel on the right (below on a smartphone) has the buttons for the steps, “Pause,” “Reset” and “Keep result,” the sliders for the conditions, and the presets for comparison with “Fast-forward.”
When you press “Keep result,” the current result is overlaid on the graphs as a gray dashed line and appears in the “Previous” entries of the readouts, so you can change the conditions and compare.
When you choose a preset in the preset box, its conditions are put into the sliders immediately. If you then move a slider by hand, the box returns to “Current slider settings,” and “Fast-forward … with the selected conditions” runs from the beginning with the conditions shown in the box (so the box and the sliders never disagree).
If you press “Pause” partway through, the button for the step changes to “Resume … (… left)” and the same button continues from where it stopped.
Adding #solid, #liquid, #cvd or #pvd to the end of the address opens that tab,
?embed=1 gives an embedded view that hides the heading, the introduction, the three routes and the four-panel strip, and writing something like ?only=cvd shows only that tab.
The animation stops while the tab is hidden or the screen is not visible.
The numbers on this page are the results of calculating the equations in the text with the same functions as the simulator (checked in Node.js). Please do not cite the estimated and calibrated values carried over from the model of the Virtual Laboratory (in preparation) as measured values. All the figures were made by ourselves from the equations on this page; they are not reproductions of figures in textbooks or papers.
Companion simulator: Comparing Synthesis Methods — Solid-State, Solution, and Vapor Routes (CVD and PVD)