SOP · Standard SEM imaging session

This procedure takes you from exchanging the sample to a saved, focused, stigmated micrograph. Screenshots show the control involved in each step. To be guided live, with each control highlighted as you go, press:

▶ Start interactive walkthrough

The simulator starts ready to image (sample loaded, vacuum, beam on). Steps 1–3 are optional here, but they are how a real instrument is operated. On a real SEM, never vent with the HV on, never open the chamber while it is under vacuum, and wear gloves when handling stubs.

1 · Sample exchange

2 · Pump down

3 · Beam on

4 · Find the sample

5 · Navigate the stage

6 · Magnify & refocus

7 · Correct astigmatism

8 · Optimise & capture

9 · End of session

Troubleshooting

• Black screen: the beam is off, still ramping, or no sample is loaded (check the Instrument lamps).
• Cannot find focus: drop to low magnification, press Auto focus, then zoom back in.
• Image never quite sharp: astigmatism (step 7).
• Image drifts, oxide or polymer glows white: charging. Coat the sample (step 1.4). Lower the kV (~1–2 kV), use a smaller spot or shorter dwell, or switch to BSED.
• Horizontal bands of sharp and blurred lines: the focus is changing during a slow scan. Enable TV scan next to Wobble.
• Stage stops moving up: the touch alarm is protecting the pole piece. Increase the WD.

Quick start

1. Click the SEM monitor to take control (its frame turns blue). Esc releases it.
2. Press Auto focus, or hold W/S until edges look crisp.
3. Scroll the mouse wheel to zoom; drag the image to move the stage.
4. Above ~5 kx, correct astigmatism with A/D and Q/E, helped by Wobble and Reduced area.
5. Press Space for a slow, low-noise capture, then Save PNG.

Every hotkey also has a slider and a number field in the control panel. Type a value and press Enter; invalid entries are rejected and out-of-range values are clamped.

Instrument: sample, vacuum, beam

By default the microscope is ready: sample loaded, chamber at high vacuum (~3×10⁻⁶ Torr), beam on. The Instrument section lets you go through the real start-up and exchange sequence:

1. Beam off — always switch the HV off before venting (venting forces it off anyway).
2. Vent — the chamber is let up to atmospheric pressure (≈3 s here, minutes on a real SEM).
3. Unload / Load sample — only possible when vented. Pick the specimen in the list, then Load; the stage returns to home.
4. Pump — roughing then turbo pumping to high vacuum (≈7 s). The lamp blinks amber until ready.
5. Beam on — the accelerating voltage ramps up (≈2.5 s); the image appears as the HV reaches its set value.

Sputter coating: insulating samples (TPP polymer prints, SiO₂, biological tissue) accumulate charge. On a real SEM they are coated with a few nm of metal (Au, Pt) or carbon in a separate coater. Here, choose a coating while the chamber is vented. Au and Pt give strong SE signal and remove charging; Au leaves a fine ~60 nm island grain visible at >100 kx, and Pt is finer. Carbon is conductive, with low yield and no grain, and interferes least with BSE/EDS. Mounting a different specimen removes the coating.

With the beam off the monitor is blank and focus/capture are disabled. Quick specimen swaps from the list still work at any time, for convenience. For a guided run-through use ▶ Tutorial, or read the illustrated SOP.

Layout & windows

• Docked mode (default): chamber | monitor | control panel. Drag the thin bar between the two views to resize them; double-click it to reset.
• ⧉ (float): turns the chamber, monitor and control panel into free windows. Drag a window by its title bar and resize it from the ◢ corner grip. Clicking a window brings it to the front. Press ⧉ again to dock.
• ⤢ maximises a view; ⛶ maximises it and enters browser full screen. ☰ shows/hides the control panel.
• Click a section title in the control panel to collapse or expand it. Every section has a ? with a short explanation.
• The layout, window positions and collapsed sections are remembered in this browser.

Hotkeys

Active only after clicking the monitor. Hold Shift for 5× finer steps.

Touch screens. One finger pans the stage; double-tap centres on that point; pinch, or drag two fingers up / down, changes magnification. Tap a setting's name in the panel (for example Focus · WD, Spot, Contrast, or a 2D box title such as Stigmator) to select it: one-finger drag on the image then changes that setting instead of panning (left / right, and up / down for 2D boxes). A chip on the image shows the selected setting; tap its ✕ or the name again to go back to panning.

How an SEM makes an image

An electron gun emits electrons that are accelerated by the high voltage (HV) and demagnified by condenser lenses into a probe a few nanometres wide. The objective lens focuses that probe onto the sample and the scan coils raster it line by line. At each point, the beam knocks secondary electrons (SE) out of the top few nanometres of the surface and scatters some primary electrons back (backscattered electrons, BSE). A detector counts them and the count becomes the brightness of that pixel. There is no lens forming an image: the picture is a map of detector signal versus beam position.

Consequences you can see in the simulator:

• Surfaces tilted away from the beam and sharp edges emit more SE, so they look bright (secant law and edge effect).
• Resolution is set by the probe size, not by the pixel count.
• The image is built up over time, so anything that changes during the scan (focus, stage, charging) appears as bands or distortion.

Specimens

• MEMS device: polysilicon comb-drive resonator with release holes and folded springs, a gear train, a cantilever array over an etched cavity, gold pads and dust. Great for edges, trenches and grain texture (grains appear above ~20 kx).
• CPU die: SRAM arrays, standard-cell logic with gate rows, multilayer interconnect with vias, a spiral inductor, bond pads and wire bonds. Good for BSE material contrast (copper/tungsten vs oxide).
• Photonic IC (SOI): 220 nm Si waveguides on 3 µm buried oxide under a 2.2 µm SiO₂ cladding — focusing grating couplers and a ring-resonator bank in open cladding windows; buried buses, a Mach–Zehnder with MMI splitters and TiN thermal phase shifters, a heater-tuned microring, a magneto-optic ring circulator under a bonded Ce:YIG die with an Au electromagnet coil, a waveguide-crossing array, inverse-taper edge couplers, Al routing to bond pads and buried M1 buses. Raise the HV to see the buried waveguides and metal buses through the cladding.
• Diffractive optics: two glass chips on one holder — a 4-level surface-relief vortex zone plate (DOE, 4 µm pixels, 180 nm steps) with a binary grating strip, and a Ø400 µm metalens of TiO₂ nanopillars (0.5 µm lattice, radius set by the hyperbolic phase for f = 400 µm, NA ≈ 0.45). At low magnification the lens shows as Fresnel-like zones of different fill factor; above ~5 kx the individual pillars resolve.
• Human hair (~70 µm, with cuticle scales) lies across the MEMS and CPU samples for scale. It is organic and insulating, so it charges.
• TPP 3D prints: two-photon-polymerised micro-optics in insulating resin on glass. A cleaved Ø125 µm optical fibre stands ~260 µm tall with a three-element Fresnel objective printed on its end face (saw-tooth zones on stacked lenses, held by pillars), next to a 3×3 Fresnel lens array and a 150 µm "benchy"-style micro-boat. Side walls show 0.3 µm slicing layer lines and top faces show the laser hatching. Uncoated, everything charges and drifts; coat it (Instrument → Sputter coating, while vented) to image it cleanly.
• Insect biosurface: a compound eye of ~2000 hexagonal ommatidia with interommatidial bristles, sensory setae and ridged scales. The lenses carry a ~250 nm "moth-eye" nipple array visible at high mag. Uncoated, so it charges.

The random seed (Specimen section, New for a new one) drives every random element: particle and scale placement, surface grain, edge roughness, wall etch marks, the hidden column astigmatism and the noise sequence. The same seed always reproduces the same sample, so you can share or repeat an exercise.

At high magnification edges are not perfectly straight: a procedural line-edge roughness field (scaled in µm, fixed to the specimen) makes outlines wiggle like real etched structures, and near-vertical side walls carry etch marks — cusped horizontal scallops on the MEMS (DRIE cycles) and fine vertical striations on all samples. Tilt the stage to see the walls.

All specimens are a 1.3 mm die on a 12.7 mm stub with a carbon tab. Microscopic surface texture is generated for the current magnification, so there is always detail to focus on.

Detectors

• ETD (Everhart–Thornley): the classic SE detector, mounted at the rear of the chamber. It gives a strong 3D, topographic look, with faces towards the detector brighter and a slight shadow on the far side of tall structures.
• BSED: an annular backscatter detector that slides in under the pole piece (watch the 3D view). Brightness follows atomic number (Z-contrast), so metals are bright and oxides and organics dark, with little topography. It needs about 1.5 mm more working distance (the touch alarm enforces this) and gives weak signal at low kV.
• In-lens: sits inside the column and collects SE1 electrons from the beam footprint. It gives the sharpest, most surface-sensitive, edge-rich images, with no directional shadowing.

Accelerating voltage (HV, 1–30 kV)

The voltage sets the energy of the electrons and therefore how deep they penetrate. The interaction volume (Kanaya–Okayama range, shown under the slider for silicon) grows roughly as kV^1.67: about 30 nm at 1 kV, 3 µm at 15 kV and 9 µm at 30 kV.

Rule of thumb: use low kV for delicate, uncoated or surface-critical samples, and high kV for BSE, EDS-style work or maximum sharpness on conductive samples.

Seeing below the surface (high kV)

Higher accelerating voltage drives the electrons deeper (Kanaya–Okayama range ≈ 0.03·kV^1.67 µm in Si). Features buried under a thin cover layer then contribute backscattered and SE2 signal: they show through with a contrast that fades as exp(−depth / ½·range) and blur with depth as the beam spreads.

• CPU die: the right half of the core is still under its passivation. At 5 kV it looks flat; at 15 kV the topmost power straps appear; at 25–30 kV the criss-cross of the lower power-grid tier and faint M3 routing show through.
• Photonic IC: the Si waveguides, rings and circulator are buried 2 µm under oxide (plus the 6 µm garnet die). They appear from ~15 kV and become clear at 30 kV, together with the buried M1 ground buses crossing under the exposed Al routing.
• BSED shows it most strongly (Z-contrast); in-lens, being surface-sensitive, barely.

Magnification & horizontal field width

Magnification follows the 127 mm (Polaroid) display convention: HFW = 127 mm ÷ magnification. So 100× shows 1.27 mm and 100 kx shows 1.27 µm. The simulated image has 1024 pixels across, so pixel size = HFW / 1024. The scale bar in the data bar is always exact. The range is 100× to 500 kx. Above roughly 200 kx you are imaging the generated sub-10 nm texture, and the probe size, noise and astigmatism limit what you can see, as on a real instrument.

Focus & working distance

Focus is the objective-lens excitation, expressed as the distance at which the beam converges (the "WD" shown in the data bar, as on real SEMs). Stage Z is where the sample physically is. The image is sharp when the two match for the surface under the beam.

• The beam is a cone with half-angle α ≈ aperture radius / WD. A surface Δ away from focus sees a blur disc of radius ≈ α·Δ. At high magnification a pixel is tiny, so the same defocus blurs over many pixels and focusing becomes much more sensitive.
• The W/S keys and the Focus fine jog slider automatically scale their step size with magnification. The coarse slider and number field set focus in absolute mm.
• Far out of focus the image becomes a featureless haze. Drop to low magnification, find the sample, focus, then zoom in and refocus.

Depth of field

Depth of field ≈ pixel size / α. It is shown under the focus controls. It gets smaller at high magnification, at short working distance (larger α) and with larger spot/aperture. On a tilted sample, parts nearer or farther than the focal plane go soft. That is per-pixel defocus, not a bug. Increase the WD or reduce the tilt for more depth.

Beam / spot size & probe current

The spot-size number sets the condenser demagnification. Larger spot gives more probe current, so less noise and a brighter image, but a wider probe and softer edges (visible at high mag). Smaller spot gives a sharper probe but fewer electrons per pixel, so the image is grainier. The readout under the slider shows the probe diameter, the current in pA, and the electrons per pixel at the current dwell time (below ~20 e⁻/px the noise dominates).

Dwell time & noise

Dwell time is how long the beam stays on each pixel. Signal-to-noise grows with √(current × dwell), a Poisson shot-noise model.

• Short dwell (0.1–0.5 µs) gives fast frames but streaky, "rainy" TV static. The detector amplifier cannot follow the beam, so the signal smears along each scan line and noise comes in horizontal dashes.
• Long dwell (3–10 µs) gives clean, fine-grained images but a slow raster (up to ~2.4 s per frame). Watch the scan line sweep down.
• The slow-scan capture uses 30 µs.

Contrast & brightness

These are the detector gain (contrast) and offset (brightness). High contrast pushes edges to white and trenches to black (clipping); brightness shifts the mid-tones. Adjust contrast first so the brightest features are just below saturation, then brightness so the darkest detail is not crushed.

Scan modes

• Normal raster: the image refreshes progressively from top to bottom. Frame time is set by the dwell time (top-right of the monitor).
• Reduced area (X): only a centred box is scanned (10–60% of the width). Fewer pixels means a much faster refresh, which makes focusing and stigmation interactive. The rest of the image freezes and dims.
• Fast TV scan while the stage moves / during wobble: the whole frame refreshes every display frame at short dwell. This keeps the image coherent when something changes quickly. Turn it off to see the true motion artifacts.
• Slow-scan capture (Space): a single top-to-bottom pass at 30 µs, then the image freezes for 3 s and becomes available as a PNG.

Astigmatism & stigmators

An imperfect lens focuses the beam at two different heights in two perpendicular directions, so the probe is an ellipse instead of a circle. The simulated column has a hidden astigmatism, set by the random seed, that you must correct.

• Symptoms: when slightly under-focused, detail streaks in one direction; when over-focused, it streaks in the perpendicular direction. At exact focus the image is uniformly soft (the "circle of least confusion") and never truly sharp.
• Correction: the X and Y stigmators add a correcting astigmatism along 0°/90° and 45°/135°. Adjust one until the image is sharpest, refocus, adjust the other, and repeat.
• The spot preview in the Alignment section draws the actual probe shape at the current focus. A round spot at focus means the beam is stigmated. The ratio readout should approach 1.00.
• Astigmatism scales with magnification in pixels: invisible at 300×, obvious at 10 kx and above.

Control panel layout

Tabs select pages, each made of collapsible sections; the Status section stays at the bottom of every page.

• 2D boxes: drag the cross; Shift = fine; double-click = zero; arrow keys when focused. The Stigmator can also be driven by Shift + right-drag on the image (magnification-sensitive), and right-drag left/right focuses.
• Beam Shift moves the scanned field electronically (±25 µm) — use for fine positioning at high magnification; reset before alignments.
• Scan Rotation rotates the raster; stage drag directions follow it.
• Low Vacuum (0.08–1.5 Torr): gas ions neutralise charging, but the beam loses current into a wide skirt (∝ pressure × gas path length ≈ WD), lowering contrast and adding noise. ETD and In-lens are disabled; the gaseous LVD (or BSD) is used. Keep WD short.
• Beam Deceleration: landing energy = HV − stage bias; interaction range, charging and surface sensitivity follow the landing energy while the probe stays small.

Column alignment: gun & aperture

Before stigmating, the beam must travel down the optical axis. The simulated column has hidden, seeded misalignments, corrected on the Alignments page. The gun starts aligned (its offsets are compensated by default); use Misalign (practice) in Source Tilt and Shift to train source alignment. Stigmator Alignment centres the stigmator coils: when de-centred, the image shifts whenever the stigmator changes — switch Modulator X (then Y) on and drag the Stigmator Center cross until the image stays put.

• Gun tilt X/Y — the emission cone of the source points off-axis, so much of it is intercepted by the condenser and objective apertures. Symptom: a dim, noisy image and a low bar in the transmitted-current meter. Adjust tilt X and Y alternately for maximum brightness.
• Gun shift X/Y — the source is laterally displaced; same symptom, smaller effect. Tune after tilt, then re-check tilt.
• Objective aperture X/Y (lens alignment) — the beam passes the objective lens off-centre and arrives tilted. Symptoms: the image moves sideways when you change focus (worst far from focus), and resolution is degraded by coma-like blur. With Wobble on, an aligned aperture makes the image only pulse in and out of focus; a misaligned one makes it swing. Adjust X/Y until the swing (readout under the controls) is ≈ 0 px. Check at progressively higher magnification.

Model: transmitted current = exp(−tilt²/0.06)·exp(−shift²/0.1) of the residual misalignments; image shift = defocus × beam tilt (0.015 rad per unit of aperture misalignment); probe diameter × (1 + 6·aperture² + 0.5·tilt). Changing the random seed changes the misalignments.

Wobble

Wobble swings the focus sinusoidally (1.6 Hz). If the objective aperture is aligned, the image only pulses in and out of focus; if it is off-axis the image swings sideways — correct it with Aperture X/Y (see Column alignment). Residual astigmatism shows as the blur changing direction between the two extremes; trim it with the stigmators. With the TV scan toggle next to Wobble on (the default), the simulator uses TV-rate scanning while wobbling, so the whole frame pulses together. Turn it off to keep the slow raster: you will then see horizontal bands of sharp and blurred lines, because each band is scanned at a different point of the oscillation.

Auto focus

This finds the surface height under the beam and drives the lens there in 0.6 s, landing within a fraction of a pixel of blur. It does not correct astigmatism. Use Zero stigmators to reset them.

Stage (5 axes)

• X / Y: ±600 µm around the die centre. Dragging the image moves the stage so the image follows the cursor, corrected for rotation and tilt.
• Z (working distance): 3–30 mm from the pole piece. Shorter WD gives a larger beam angle, so better resolution but less depth of field. A touch alarm prevents the specimen, a tilted stub or the inserted BSED from hitting the pole piece.
• R (rotation): rotation about the beam axis (compucentric), so the feature under the beam stays put.
• T (tilt): 0–45° towards the ETD, about a eucentric axis through the focal point. The image foreshortens vertically, the tops and faces of structures shade differently, and focus varies across the frame (depth of field).
• Home (Home) returns X = Y = 0, WD = 10 mm, R = T = 0.

Chamber view

A to-scale 3D model: 1 unit = 10 mm. It shows the objective pole piece, the ETD with its Faraday cage (rear), the retractable BSED (right), and the 5-axis stage with the stub and specimen. The cyan line is the beam and the blue rectangle on the sample is the current scan field. Drag to orbit, use the wheel to zoom (it follows the sample when close), Sample jumps to a close-up, and double-click or Overview resets the view. Because the specimen is drawn at true scale, it is a tiny chip in the overview.

Materials & surfaces

Each surface has a material class that sets its procedural micro-structure (scaled in real units, so it appears only at the magnification where it would really be resolved):

• Crystalline silicon (MEMS device layer, CPU and dielectric substrates): polished, almost atomically flat tops — nearly featureless even at 100 kx. Deep-reactive-ion-etched (DRIE) side walls carry the characteristic horizontal scallops (~0.75 µm period on the MEMS) left by alternating etch/passivation cycles; tilt to 30–45° to see them.
• Metals (Au pads, Cu/W interconnect, Al): polycrystalline grains (~250 nm and ~60 nm) plus a partial crack network (~2 µm cells) from film stress, visible above ~10 kx.
• SiO₂ / dielectrics: smooth, low-yield films that charge (next section).
• Organic (insect): rough, grainy cuticle; insulating, so it charges.

Etched edges show line-edge roughness: outlines wiggle at the 10–100 nm scale instead of being ideal straight lines.

Charging & drift

An insulator cannot drain the electrons the beam deposits. The trapped charge builds up and, in the simulator (as in practice):

• Washes out insulating regions: SiO₂ saturates bright and flat, so its edges and detail become hard to see; the glow spills onto neighbouring regions.
• Distorts the image near charged areas: the local field deflects the beam, bending straight lines.
• Drifts the whole image: the growing surface potential steadily deflects the beam, so the field of view creeps in one direction (shown under Stage as "charge drift"). A slow scan then smears or shears.
• Throws bright streaks along scan lines as the surface potential flickers.

Charging grows with probe current, dwell time and magnification (dose per area), and relaxes over several seconds once the dose drops. Moving to fresh, unexposed area resets it locally. Mitigations: lower the HV towards ~1 kV (the charge-balance crossover), use a smaller spot and shorter dwell, lower magnification, switch to BSED (less sensitive), or — on a real instrument — coat the sample or use low-vacuum mode. Home or reloading the specimen clears the accumulated drift.

The Dielectric SiO₂ on Si specimen is built for this: a field-oxide window with contact holes to the Si, oxide pads of increasing thickness, an oxide grating, concentric oxide rings, floating Al pads on an oxide island (they charge too) and grounded Al pads directly on Si (which do not), for comparison. The CPU die charges mildly on its oxide; the insect charges everywhere.

Artifacts you may see (all intentional)

• Edge effect: bright rims along edges and particles. It gets broader at high kV.
• Shadowing: the far side of tall features and the bottoms of trenches are darker, because SE cannot reach the ETD. It is absent with the in-lens and BSE detectors.
• Charging (dielectric, insect, CPU oxide): washed-out bright insulators, glow, bent lines, streaks and a steadily drifting field of view. See Charging & drift.
• Scan-line jitter / static: at very short dwell.
• Motion distortion: bent or sheared images if the stage moves during a slow raster with Fast TV scan turned off.
• Stale rows: after a change, the old image stays below the scan line until the beam rescans it.

Capture & export

Slow scan capture acquires one full 1024 × 768 frame at 30 µs dwell (≈ 3.5 s). The result freezes on screen for 3 s and can be saved as a PNG with a data bar (HV, WD, mag, HFW, spot, dwell, detector, tilt) and scale bar. The file is created locally in your browser; nothing is uploaded.

Recommended workflow

1. Choose the detector and HV for your question (topography: ETD at 5–15 kV; composition: BSED at 15–20 kV; delicate or insulating: 1–3 kV).
2. At low magnification, navigate (drag) and focus.
3. Set WD: 10 mm is a good default; go shorter for resolution and longer for depth of field or tilt.
4. Zoom to about 2× your target magnification and use Reduced area. Focus, then correct astigmatism (Wobble helps). Refocus.
5. Zoom back out, set contrast and brightness, pick the spot size and dwell for the noise level you need, and capture.

Model & limitations

This is an educational simulation, not a Monte-Carlo electron transport code. Signal uses empirical rules (secant-law SE yield, atomic-number weighting for BSE, a detector direction term, an edge Laplacian and occlusion shadowing). Blur is an elliptical Gaussian from geometric defocus plus the probe size. Noise is Poisson shot noise plus a detector bandwidth model. The interaction volume is approximated by a kV-dependent halo and edge width. Numbers such as probe diameter, current and range are representative of a field-emission SEM but not of any specific instrument.