Feedback to this page: click here

This page is written by Evgeniy Shkondin @DTU Nanolab if nothing else is stated.
All images and photos on this page belongs to DTU Nanolab.

SiO₂ disk and granule source tests in Temescal, October 2026

Experimental work and documentation: Evgeniy Shkondin, DTU Nanolab. Rikke M. Nielsen was present during Run 1 (22 October 2026). Run 2 (23 October 2026) was carried out by Evgeniy Shkondin alone. Report updated on 27 October 2026.

Photographs and schematics: Evgeniy Shkondin / DTU Nanolab. The power and rate plots are reconstructed from the instrument recordings. Both runs used a HULA (high uniformity lift-off assembly) holder; film thicknesses were determined by Cauchy-model ellipsometry.

Overview

Two source-loading configurations were tested in the Temescal electron-beam evaporator on 22 and 23 October 2026. Each run used four solid, circular SiO₂ disks in a graphite liner and a nominal thickness setpoint of 600 nm at 1 Å/s. Run 1 used the disks alone. Run 2 used new disks with SiO₂ granules between the stack and liner wall. Both runs used a HULA (high uniformity lift-off assembly) holder.

The disks alone developed a narrow cavity that penetrated the entire stack. Adding granules and changing to a horizontal figure-eight sweep produced a more manageable deposition: repeated small rate spikes persisted, but the run finished without the major instability or growing liner-exposure concern described for Run 1. The deepest penetration stopped at the interface between the third and fourth disks.

Outcome	Run 1: disks	Run 2: disks + granules
Deposition date	22 October 2026	23 October 2026
People present	Evgeniy Shkondin; Rikke M. Nielsen	Evgeniy Shkondin
Thickness setpoint	600 nm	600 nm
Cauchy-model thickness	665 nm	668 nm
Sweep program	#2: small circle	#7: horizontal figure-eight
Peak recorded deposition power	6.9%	6.9%
Final source condition	Full-stack breakthrough	Penetration through the first three disks
Source mass change	1.55 g loss	Not recorded

Note. Main finding: the combined disks-and-granules configuration with sweep #7 is the more promising configuration from these two tests. Because sweep geometry and substrate temperature also changed, the improvement cannot be assigned to the granules alone.

The report follows the experimental sequence using source photographs, schematics, visual observations and instrument recordings. The power curves retain the recorded values without smoothing. Run 1 heating-timer pauses are excluded from the main comparison and are separate from source instability. These exploratory tests document source behaviour and provide a basis for further process development.

Test design and recorded settings

The tests examined source erosion and sweep behaviour over a nominal 600 nm deposition. Four disks were stacked in the graphite liner. Before Run 2, a new stack was loaded and surrounded by SiO₂ granules. A HULA (high uniformity lift-off assembly) holder was used in both runs. The source specifications below are taken from the material package labels.

Source	Company	Dimensions	Labelled purity
Disks	ScoTech	21 mm diameter × 5 mm thick	99.99% (4N)
Granulate	Umicore Thin Film Products AG	1.5-3.5 mm	99.99%

Common tooling-file settings

Parameter	Recorded value
Tooling factor / density / Z-ratio	78.5 / 2.648 g/cm³ / 1
Control gain	8.0 (Å/s)/% power
Control TC / control DT	0.1 s / 0.1 s
Maximum power / O₂ flow setting	10% / 5%
Rate / thickness setpoint	1 Å/s (0.1 nm/s) / 600 nm

Conditions that differed between runs

Parameter	Run 1	Run 2
Substrate temperature setting	190 °C	Room temperature (RT)
Source loading	Four disks alone	Four new disks + surrounding granules
Sweep geometry	Small circle (#2)	Horizontal figure-eight (#7)
Reported sweep parameters	15% amplitude; 2 Hz	Snapshot: 35% on both axes; 1 Hz / 2 Hz

Note. Density is a controller input, not measured film density. O₂ = 5% is the instrument setting, not gas composition or partial pressure. Run 2 was reported as RT/unheated, but its logged thermocouple temperature rose from 22.0 to 61.6 °C during deposition. The TC channel is not an independent sample-surface temperature measurement. Power is recorded as percent of full scale.

Source loading, sweep geometry and the temperature condition changed together, so this is not a single-variable comparison. The logs also show different rise/soak settings, including soak-2 levels of 2.8% in Run 1 and 5.0% in Run 2. Thickness milestones refer to the QCM indication, not local source erosion depth. The deposition-only comparison excludes repeated conditioning in Run 1.

Run 1: solid disks alone

Run 1 was carried out on 22 October 2026 by Evgeniy Shkondin (DTU Nanolab), with Rikke M. Nielsen present. Initiating evaporation from solid disks required more power than the established granulate process, so the tooling-file limit was raised to 10%. The figure-eight sweep appeared to extend beyond the disks. Small-circle program #2 was therefore selected, with 15% amplitude and a 2 Hz frequency.

Stage / indicated thickness	Observation and reconstruction from the log
Startup to about 30 nm	Recorded power reached 6.9% near 1.3 nm and remained about 6.7% at 20 nm. The main power decrease began after the first approximately 30 nm.
About 30-90 nm	Power fell sharply, with local medians of 5.8% at 40 nm and 2.8% at 60 nm. Oscillations followed. A central hole was visible near 70 nm; the log contains brief QCM-rate peaks up to 9.6 Å/s.
After about 120 nm	The beam was shifted slightly off centre. Power stabilized around 2.6-2.7%, then gradually increased.
Around 270 / 400 nm	Recorded power was approximately 3.5% / 3.9%, consistent with the approximate written milestones.
About 500 nm to completion	The cavity began to glow and liner exposure was suspected. The beam was shifted again, with uncertain benefit. A late recorded transient near 587-589 nm occurred before completion at about 600 nm.

Ellipsometry using a Cauchy model for SiO₂ gave a thickness of 665 nm for the transparent film. Inspection after venting confirmed breakthrough through the full disk stack and no graphite-liner damage. Typical visible rate fluctuations were 0.6-2.2 Å/s; the recording captures larger brief excursions.

Figure 1. Run 1 source geometry before and after deposition. Four solid SiO₂ disks were stacked in a graphite liner without surrounding granules. The small circular sweep produced a localized cavity that eventually extended through the full stack. The orange lines indicate the sweep path.

Note. A 30-minute heating allowance caused three alarms. Each resumption initiated a new rise/soak sequence before deposition continued. These timer-related interruptions are excluded from the source comparison. The recording confirms completion at approximately 600 nm.

Run 1: post-deposition evidence

Figure 2. Run 1 after venting. (A) Removed four-disk stack. (B) View into the central cavity and its narrow opening. (C) Oblique view of the eroded upper region. (D) Opposite-face view showing the breakthrough. (E) Side view of the stack. (F) Further axial view of the opening. These photographs document full-stack penetration. Inspection found no damage to the graphite liner.

The axial views reveal a relatively narrow passage through a much wider remaining body of source material. This is consistent with strongly localized erosion. Full-stack breakthrough does not mean that most of the source mass was consumed: the reported mass decreased from 15.54 g to 13.99 g, a loss of approximately 10.0%.

Run 2: disks with surrounding granules

Run 2 was carried out by Evgeniy Shkondin alone on 23 October 2026. New disks were loaded and SiO₂ granules filled the surrounding space in the graphite liner. This allowed the conventional horizontal figure-eight sweep, program #7. The run was performed without substrate heating, using the same HULA holder type. The table follows visual observations through the top chamber window; the approximate surface shapes are interpreted in Figure 5.

Indicated thickness	Power	Observation
Initiation	5.9%	Reached the 1 Å/s rate setpoint.
20 nm	6.8%	Source appeared normal (stage A).
40 nm	5.7%	Power began to decrease.
Approaching 50 nm	Not stated	The disk edge appeared deformed and a dark region became visible (stage B).
60 nm	5.1%	Power continued to decrease.
150 nm	4.5%	Rate remained around 1 Å/s. The dark region became slightly larger.
300 nm	4.4%	Power was very stable. Rough-looking features appeared around the dark region (stage C).
Beyond 300 nm	Not stated	A darker spot appeared within the existing dark region, suggesting local deepening (stage D).
400 nm	4.3%	The upper disk appeared deformed and a second dark region became visible (stage E).
500 nm	4.4%	The two dark regions appeared to merge into a figure-eight-shaped region (stage F).

Small repeated rate spikes occurred throughout the run. Surrounding granules occasionally glowed when struck by the sweeping beam. Apart from these spikes, the rate and power remained reasonably stable, and the deposition completed without significant instability or additional concern.

Ellipsometry using a Cauchy model for SiO₂ gave a thickness of 668 nm for the transparent film. After venting, inspection showed penetration through the first three disks, stopping at the boundary between the third and fourth. Local cavity formation persisted, while the bottom disk remained intact.

Run 2: geometry and controller evidence

Figure 3. Run 2 source geometry before and after deposition. Granules filled the space around a replacement stack of four disks, and the beam used a horizontal figure-eight sweep. Post-run inspection placed the deepest penetration at the interface between disks 3 and 4, counted from the top.

Figure 4. Run 2 controller and sweep display during deposition. The snapshot shows 1.0 Å/s, 4.5% power and 1.581 kÅ accumulated indicated thickness, equivalent to 158.1 nm. The rate-deviation trace shows repeated spikes. Sweep program 7 has 35.0% latitude and longitude amplitudes and frequencies of 1 Hz and 2 Hz, respectively. These values describe the instant shown.

What the controller photograph establishes

The image independently supports the stated figure-eight sweep and the presence of repeated rate deviations. Its indicated thickness of 158.1 nm and power of 4.5% are consistent with the written observation of approximately 4.5% at 150 nm. The sweep display also identifies pocket 9 at the time of the photograph.

The controller photograph shows the short-term rate trace. The instrument recording provides an approximately one-second time series for the later power and rate figures. Faster oscillations may remain unresolved at this sampling interval. The 1 Hz and 2 Hz values in the image are sweep settings, not measured spike frequencies.

Run 2: observed surface evolution

Figure 5. Schematic progression of the Run 2 source surface. Panels A-F are visual mental descriptions of the source as seen through the small window on top of the chamber. Brightness, contrast and shadowing can obscure the actual shape. Grey and black areas represent perceived darker regions; they are qualitative visual cues and do not identify chemical phases, temperature or measured cavity depth.

Stage	Approximate point in run	Visual interpretation through the window
A	Early stage / about 20 nm	An apparently regular disk surface surrounded by granules.
B	Approaching 50 nm	Apparent edge deformation and the first visible dark region.
C	About 300 nm	A rough-looking boundary appears near the surrounding granules.
D	Beyond 300 nm	A darker spot within the first region suggests local deepening.
E	About 400 nm	A second dark region appears with further apparent upper-disk deformation.
F	About 500 nm	The dark regions appear to merge into a figure-eight-shaped area.

The drawings are visual mental descriptions based on looking through the small window on top of the chamber during deposition. The actual source shape can be difficult to interpret: brightness and contrast may be misleading, and several shadowing effects can alter the appearance. The sequence therefore records perceived changes in the source. Post-run inspection established the final penetration depth independently of these in-process impressions.

Run 2: post-deposition evidence

Figure 6. Run 2 after venting. (A) View of the lower face of the removed stack. (B) Side view of the stack and irregular upper edge. (C) Source material remaining in the graphite liner. (D) Removed stack with rough material around the eroded opening. (E) Close view of the cavity. (F) Removed stack and surrounding loose material. Post-run inspection identified penetration through disks 1-3, stopping at the boundary with disk 4.

The source retained an intact lower disk beneath a strongly reshaped upper region. Irregular material remained around the cavity and in the liner. This morphology is consistent with interaction between the disk stack and the surrounding granules, but it does not establish a mass balance or a reduction in total material consumption.

Thickness, mass and numerical comparison

The comparison combines Cauchy-model ellipsometry, source-mass measurements and instrument recordings. The logs distinguish the last DEPOSIT sample from the later final-thickness summary, written after the idle ramp. The latter includes small post-deposition changes in the QCM indication and should not be equated with extra film thickness on the sample.

Quantity	Run 1	Run 2
Thickness setpoint	600 nm	600 nm
Last DEPOSIT sample	599.93 nm	600.03 nm
Final log summary after idle ramp	600.7 nm	600.5 nm
Ellipsometry thickness	665 nm	668 nm
Measured thickness minus setpoint	+65 nm	+68 nm
Relative difference from setpoint	+10.83%	+11.33%
Recorded total shutter-open time	101 min 17 s	99 min 49 s
Reported initial / final source mass	15.54 / 13.99 g	Not recorded
Apparent source mass loss	1.55 g (9.97%)	Not calculable

Thickness and mass calculations

Relative thickness difference = 100 × (ellipsometry thickness − setpoint) / setpoint. The films differ by 3 nm, approximately 0.45% of their mean of 666.5 nm. Both thicknesses were obtained by fitting SiO₂ with a Cauchy model. Two different experimental conditions do not establish repeatability.

Run 1 apparent source loss = 15.54 − 13.99 = 1.55 g, or 100 × 1.55 / 15.54 = 9.97%. This source mass change is not collection efficiency. No equivalent mass measurements were supplied for Run 2.

Implications for calibration

The final logged values confirm that both depositions reached the intended 600 nm range. The approximately 11% difference from ellipsometry therefore remains a calibration question rather than an unrecorded extra 50 nm of indicated deposition. Relative to the post-ramp summaries, the differences are approximately +10.70% and +11.24%.

The similar offsets motivate checking QCM calibration and sample position within the HULA holder before selecting a new tooling factor. QCM thickness conversion depends on density and tooling geometry [2]. The entered 2.648 g/cm³ is a controller setting; the recorded thickness alone does not identify which calibration input needs adjustment.

Note. At 1 Å/s, 600 nm corresponds to 100 min. The logged shutter-open totals are consistent with this scale. Run 1’s heating-timer pauses increase elapsed time but are not included in the main comparison of deposition behaviour.

Recorded power versus thickness

Figure 7. Power reconstructed from 2610229.dlg and 2610234.dlg, plotted against the recorded QCM thickness (1 kÅ = 100 nm). The retained values are unsmoothed. Run 1 timer pauses, repeated rise/soak stages and the first 120 s of each resumed deposition block are omitted, leaving short gaps. The initial startup and subsequent source-related or unexplained transients are retained. No scaling to ellipsometry thickness is applied.

The recordings provide 6680 valid source records for Run 1 and 6170 for Run 2, mostly at approximately one-second intervals. The table gives local sample medians in ±2.5 nm windows. These summarize quantized controller values; they are not independent-replicate statistics or power calibration uncertainties.

QCM thickness	Run 1 power	Run 2 power
20 nm	6.7%	6.8%
40 nm	5.8%	5.7%
60 nm	2.8%	5.1%
120 nm	2.6%	4.9%
150 nm	2.7%	4.6%
300 nm	3.6%	4.4%
400 nm	3.9%	4.3%
500 nm	4.1%	4.4%
550 nm	4.3%	4.4%

Run 1 reaches a minimum operating region near 2.5-2.7% after the early oscillations, then rises gradually towards approximately 4.5% near completion. Run 2 reaches 6.9% near 19 nm, falls towards 5%, and settles around 4.3-4.5% for much of the later deposition. The combined loading improves stability, but does not require less power at every thickness.

Early power and rate transients

Figure 8. Recorded power and QCM rate during the first 160 nm. The large Run 1 excursions around 60-90 nm precede the first heating-timer interruption and therefore are not restart artefacts. The dashed line marks the 1 Å/s rate setpoint. Rate peaks are recorded controller values, not independently verified instantaneous film-growth rates. All plotted data are unsmoothed.

Run 1: the major drop occurs after the first 30 nm

The log refines the early chronology: power was still 6.7-6.8% around 20 nm, rather than already near 2%. It fell sharply over approximately 30-70 nm. Large fluctuations in the recorded rate accompanied the low-power oscillations. The highest recorded rate was 9.6 Å/s near 69.2 nm, followed by further large peaks around 73-77 nm.

Run 2: a smaller recorded rate excursion

Run 2 reached a maximum recorded rate of 1.7 Å/s near 39.6 nm. It then remained much closer to the 1 Å/s setpoint in the sampled record. This supports the observer’s assessment of improved stability. The logger cannot fully resolve fast spikes visible on the controller display, so these maxima are limited to the sampled data.

A late Run 1 event remains in the comparison

Near 587-589 nm, Run 1 recorded a rate excursion up to 7.8 Å/s while power fell to about 2.8%, followed by recovery. This occurred well after the final restart and is retained in the main curve. The log alone cannot assign it to beam repositioning, source reshaping or another specific cause.

Processing and physical interpretation

Figure 9. Power plotted against cumulative time in the sustained DEPOSIT phases. Alarm pauses and conditioning are removed from the time axis. The first 120 s after each restart remain as gaps in the curve; those seconds are still part of the deposition-phase clock. This time axis differs from elapsed wall-clock time. All-phase wall-clock plots and the complete raw records are included in the accompanying package.

Treatment of heating-timer interruptions

Run 1 was interrupted by alarms when the 30-minute heating allowance expired. The main plots use sustained DEPOSIT phases, exclude repeated rise/soak sequences, and omit a fixed 120 s recovery interval after each restart. These exclusions are shown as gaps. Three isolated zero-power DEPOSIT-labelled records preceding restart ramps are also excluded. The initial startup is retained.

No smoothing, outlier removal or thickness rescaling is applied to the remaining points. Original record order and small reversals in the thickness indication are preserved. The package includes the unfiltered wall-clock plots and all records. Cumulative deposition-phase time is about 101.33 min for Run 1 and 99.87 min for Run 2; the separate shutter-open counters are each 3 s shorter.

Physical interpretation remains conditional

The narrow cavity and full-stack breakthrough in Run 1 are consistent with localized erosion under the small circular sweep. The wider figure-eight sweep and surrounding SiO₂ in Run 2 could distribute exposure and source reshaping more favourably. Because loading, sweep and temperature changed together, the data cannot isolate a granule effect.

A stable average rate can coexist with substantial cavity growth. The power trace is the controller’s response to the evolving process, not a direct source-temperature measurement. Although power and emission current are logged, a full calibrated beam-voltage history is not supplied, so energy consumption in joules is not calculated. Lower source mass consumption also remains unverified.

Conclusions and next experiments

Conclusions supported by this comparison

Four solid SiO₂ disks alone, used with a small circular sweep, developed a cavity through the full source stack during the nominal 600 nm test. The recording captures strong early rate excursions and a late transient. The heating-timer pauses are a separate operational issue and are not evidence of source instability. The liner was reported undamaged.

The combination of disks, surrounding granules and a horizontal figure-eight sweep gave more stable operation in the second test and left the fourth disk unpenetrated. Both configurations produced transparent SiO₂ films with Cauchy-model ellipsometry thicknesses close to 0.67 µm. The combined configuration is a reasonable starting point for further source optimization.

The strongest evidence supports improved deposition stability and reduced penetration depth in Run 2. Lower mass consumption was not quantified because the second charge was not weighed before and after deposition. A reliable maximum thickness requires further testing.

Focused follow-up work

Next step	What it would resolve
Separate fill and sweep effects	Compare source loadings under matched temperature and a compatible fixed sweep. Then compare sweep patterns for the same loading.
Document each source loading	Retain the known disk and granulate specifications; add granulate mass, liner geometry, sweep offsets and initial source photographs.
Measure consumption consistently	Weigh the complete source charge before and after each run and account for loose fragments.
Retain and synchronize future logs	Continue saving rate, thickness, power, pressure, gas and TC channels. Add timestamps for beam shifts, source photographs and any alarms.
Repeat and characterize the films	Repeat the chosen configuration, map thickness and retain the Cauchy-model ellipsometry results. Add composition or contamination measurements if those properties determine suitability.

Established practical experience in this Temescal configuration is that a granulate-only charge in a graphite liner forms a chimney - a hole through the source - when a 600 nm thickness setpoint is used, even with a figure-eight sweep. This motivated the use of solid disks together with surrounding granules.

Note. This report records exploratory tests rather than an operating instruction. Source coverage must be assessed during deposition; the absence of damage in one run does not establish that continued operation after breakthrough is acceptable. Use the current local operating procedures and tool-responsible guidance [2, 3].

Experimental record and scope

Experimental record

[1] Evgeniy Shkondin, DTU Nanolab. Temescal SiO₂ source tests, 22-23 October 2026: experimental observations, source photographs and schematics, Cauchy-model ellipsometry results and source-mass measurements. Rikke M. Nielsen was present during Run 1; Run 2 was carried out by Evgeniy Shkondin alone.

[4] “2610229.dlg”: Run 1 instrument recording, 22 October 2026. [5] “2610234.dlg”: Run 2 instrument recording, 23 October 2026. Their dates and 6.000 kÅ / 1.0 Å/s settings match the tests. The power and rate curves are reconstructed from these files.

[6] Source-material package labels: ScoTech silicon dioxide disks, 21 mm diameter × 5 mm thick, 99.99% (4N), tracking number 2702/SiO2; Umicore Thin Film Products AG silicon dioxide granulate, 1.5-3.5 mm, 99.99%, product code 0481889. Photographs of both labels accompany the report package.

Setup and observation record

A HULA (high uniformity lift-off assembly) holder was used in both runs. The SiO₂ film thicknesses were obtained by Cauchy-model ellipsometry. Figure 5 records visual mental descriptions through the top chamber window, where brightness, contrast and shadowing affect interpretation. Final penetration depths were established after venting.

Scope of the comparison

The tests compare two complete source configurations: loading, sweep and temperature changed together. Exact timestamps for beam shifts, the added granulate mass and a Run 2 mass balance were not recorded. The logged TC temperature does not independently establish sample-surface temperature. Composition and contamination were not measured in these tests.

The recordings establish the power and rate histories and confirm an approximately 600 nm endpoint for each run. The timer-related Run 1 interruptions are excluded from the source-performance comparison. Proposed physical explanations remain interpretations of the recorded behaviour.

Specific Process Knowledge/Thin film deposition/Deposition of Silicon Oxide/SiO2 disk and granule source tests in Temescal October 2026

SiO₂ disk and granule source tests in Temescal, October 2026

Overview