As new printing approaches emerge, in situ diagnostics to monitor the print quality in real-time become essential for long-term monitoring and feedback control. In this article, we present a millimeter-wave electromagnetic monitoring approach for liquid metal droplet-on-demand printing to support the high-speed and real-time evaluation of droplet ejection. An open-ended rectangular waveguide is placed perpendicular to a jetted droplet stream and operated at a continuous-wave frequency of 40 GHz. Liquid metal droplets with diameters as low as 1.2 mm are characterized, and droplet jetting events on the order of 500 μm are detected at ejection rates up to 80 Hz. The measured results demonstrate that trends at the macro-level (large-scale print variation and anomalies at the nozzle tip) as well as micro-level (droplet size, position, and dynamics) can be detected using this technique.

Advanced material and manufacturing techniques have enabled the additive manufacturing of a broad range of materials. Of particular interest over recent years has been the printing of metals. Key among metal-based printing techniques is liquid metal jetting droplet-on-demand (LMJ-DoD) printing, wherein an actuation method is used to create momentum pulses to eject discrete liquid metal droplets.1–3 Because the quality of the jetted liquid metal droplets is influenced by a multitude of factors, in situ diagnostics are critical to monitor the droplet print quality over the entire duration of the build process.3–5 Primary parameters of interest for these diagnostics are droplet size, timing (e.g., velocity and flow rate), and temperature. While high-speed videography is the current prevalent technique for high resolution and short-term monitoring (on the order of a few seconds),4,6,7 this approach scales rapidly in required data storage and processing requirements over the build process (several hours to days), which is impractical for real-time processing.

We present a millimeter-wave electromagnetic monitoring approach for droplet detection and characterization during LMJ-DoD printing. In this approach, an open-ended waveguide is placed perpendicular to the droplet stream such that jetted droplets are within the electromagnetic near-field of the waveguide aperture. Operation at millimeter-wave frequencies can yield a major reduction in the required data volume when compared with high-speed video, if the required droplet information can be extracted from the reflected millimeter-wave signal amplitude and phase alone (as opposed to n×n pixels). This reduction was demonstrated in our previous work in which an open-ended T-junction was used to monitor impedance changes as droplets passed through the lateral arms of the T-junction.8,9 However, the previous approach had a significant shortcoming; to properly diagnose droplet properties, the droplets must fall through the (open) arms of the T-junction, limiting the lateral space that the droplets are confined to. Furthermore, if the droplets come in contact with the T-junction sidewalls, there is potential for droplet buildup in the T-junction, which would interfere with printing continuity. In this paper, we describe an alternative approach using electromagnetic near-field detection, which provides a compact, non-invasive diagnostic with the ability to distinguish droplets at higher print rates and detect additional features of the print system. The use of an open-ended waveguide probe for crack and corrosion detection and dielectric material characterization in the near-field has been demonstrated previously.10–15 However, the use of this method for the characterization of small metallic objects in flight has, to our best knowledge, yet to be evaluated, either analytically or experimentally. Here, we apply this characterization technique through simulation and experiment to capture dynamic liquid metal droplet events in situ.

The experimental setup for droplet monitoring is shown in Fig. 1. A crucible with bulk tin material is heated to 400 °C and inserted into a polycarbonate chamber with an open top. The chamber is continuously purged with argon gas to limit oxidation on the surface of the droplets. An open-ended aluminum WR-28 waveguide is inserted perpendicular to the droplet stream through a cutout in the chamber. The waveguide aperture and flange are covered with a 70 μm polyimide lining for protection from the jetted liquid metal. It should also be noted that the waveguide flange could be removed to further reduce the size of the diagnostic. The waveguide end outside of the chamber is connected to a vector network analyzer (VNA) for time-domain monitoring with a continuous-wave, 40 GHz signal. To benchmark the detected millimeter-wave signal perturbations, an optical illumination source and high-speed camera are placed on opposite sides of the optically clear chamber. An aluminum cover is used to enclose the top surface of the chamber. For three-dimensional additive manufacturing, the crucible is controlled with a motorized Z stage, and the build plate is controlled with an X–Y stage.

FIG. 1.

Experimental setup for millimeter-wave electromagnetic monitoring of liquid metal droplet-on-demand printing. Optical diagnostics were used as a benchmark for the captured signals. Connections to measurement equipment are not shown. (a) Three-dimensional view. (b) Front view. (c) Side view.

FIG. 1.

Experimental setup for millimeter-wave electromagnetic monitoring of liquid metal droplet-on-demand printing. Optical diagnostics were used as a benchmark for the captured signals. Connections to measurement equipment are not shown. (a) Three-dimensional view. (b) Front view. (c) Side view.

Close modal

To model the effect of droplet properties such as diameter, height, and spacing from the waveguide, full-wave electromagnetic simulations were used. The simulated geometry is shown in Fig. 2. A droplet with diameter d is located at vertical position zD at a spacing s from the rectangular waveguide aperture, which has an aperture width wW=3.56 mm and height wL=7.11 mm. The instantaneous radiated electric field in the area in front of the waveguide is shown in Fig. 2(b), where in the given plane, it is evident that the fields radiate spherically away from the waveguide aperture and with decreasing magnitude. When a metal droplet of high conductivity is present [Fig. 2(c)], it acts as a scatterer for the incident electromagnetic field. The metallic droplet reflects incident fields away from its surface, some of which return in the direction of the waveguide, creating constructive or destructive interference with the transmitted fields.

FIG. 2.

Simulated open-ended waveguide results. All dimensions are in mm. (a) Dimensions of the rectangular waveguide aperture with a metal droplet with diameter d at height zD=0 and distance s away from the waveguide edge. wW=7.11, wL=3.56, c=3.46, t=3.81, h=14.26. (b) Instantaneous electric field magnitude for open-ended waveguide. (c) Instantaneous electric field when droplet with 800 μm diameter is located at s=7, z=0. (d) and (e) Return loss magnitude and phase as a function of droplet height for varied distance s from the waveguide. (f) and (g) Return loss magnitude and phase as a function of droplet height for varied droplet diameter.

FIG. 2.

Simulated open-ended waveguide results. All dimensions are in mm. (a) Dimensions of the rectangular waveguide aperture with a metal droplet with diameter d at height zD=0 and distance s away from the waveguide edge. wW=7.11, wL=3.56, c=3.46, t=3.81, h=14.26. (b) Instantaneous electric field magnitude for open-ended waveguide. (c) Instantaneous electric field when droplet with 800 μm diameter is located at s=7, z=0. (d) and (e) Return loss magnitude and phase as a function of droplet height for varied distance s from the waveguide. (f) and (g) Return loss magnitude and phase as a function of droplet height for varied droplet diameter.

Close modal

During a dynamic jetting process, the droplet changes height zD, correspondingly producing a time-varying reflection of the radiated fields and producing variation in the resulting waveforms shown in Figs. 2(d) and 2(e), depending on the droplet spacing s from the waveguide aperture. Variations in droplet diameter also change the magnitude and phase of the resulting perturbation [Figs. 2(f) and 2(g)]. From simulation results, there is a monotonic relationship between droplet diameter and the reflected signal magnitude for a fixed location zD=0 such that a lower minimum S11 magnitude corresponds to a larger droplet diameter (Fig. S1, supplementary material). Because consecutive ejected droplets may have varying spacing s within a range of 400 μm, this variation can also affect the magnitude of S11. It should be noted that while both droplet diameter and spacing can affect the S11 magnitude, the phase of the complex vector is also captured, and varies with changing droplet properties. As such, the combined magnitude and phase measurements can be used to determine variations in droplet size as well as horizontal location. The simulated phase variation at a fixed location zD=0 as a function of droplet parameters is shown in Fig. S1 (supplementary material).

Experiments were conducted to evaluate the sensitivity of the diagnostic with metal spheres of known diameters (Fig. 3). First, measurements were conducted to evaluate the change in return loss in a quasistatic fashion across location and demonstrated good matching with simulated waveforms (Fig. S2, supplementary material). Next, dynamic experiments were conducted in which metal spheres were dropped in front of the waveguide aperture to capture the time-domain response and emulate ejected droplets in the LMJ-DoD system. Droplets were dropped in a microfunnel and then triggered an infrared sensor to begin the VNA measurement before passing in front of the open-ended waveguide aperture [Fig. 3(a)]. An unterminated coax-to-waveguide adapter was used for the open-ended waveguide. For spheres with diameters of 1.98 mm, measured S11 waveforms at different spacings from the waveguide aperture again showed the same variation in waveform [Figs. 3(b)3(d)], as observed in the simulated results. Of note is the asymmetry present in the captured time-domain waveforms. Over the course of each captured signal, the width of each oscillation becomes narrower in time. Based on the inherent symmetry of the system, and the corresponding results in Figs. 2(d) and 2(e), any variation in width of the waveform across time suggests a change in velocity and, correspondingly, nonzero acceleration. Peaks or valleys of these captured waveforms can be extrapolated and correlated with location to infer information about the velocity and acceleration of individual droplets, which was demonstrated in our previous works.8,9 The droplet diameter of representative metal spheres is shown plotted against relative changes in S11 at a fixed spacing of s=7 mm, demonstrating a similar relationship to what was shown in simulation in Fig. S1 (supplementary material).

FIG. 3.

Controlled experiments with metal spheres of known diameters. (a) Experimental setup. (b)–(d) Time-domain results for measured S11 for varied droplet spacing from the waveguide edge. Droplet diameter: 1.98 mm. (e) Relative change in S11 for three sphere diameters: 0.79, 1.98, and 2.38 mm.

FIG. 3.

Controlled experiments with metal spheres of known diameters. (a) Experimental setup. (b)–(d) Time-domain results for measured S11 for varied droplet spacing from the waveguide edge. Droplet diameter: 1.98 mm. (e) Relative change in S11 for three sphere diameters: 0.79, 1.98, and 2.38 mm.

Close modal

An assessment of the millimeter-wave diagnostic was made in a tin LMJ-DoD system2,3 (Fig. 1). To reduce the heat load from the argon containment on the millimeter-wave measurement equipment, an additional length of 152.4 mm waveguide was used in addition to the coax-to-waveguide adapter (the VNA was calibrated to the input to the adapter). Two nozzles were used, with inner diameters of 400 and 100 μm, to generate droplets of larger and smaller size, respectively. Droplets produced by the 400 μm nozzle in an argon environment were typically larger than 1 mm and could be easily detected at 40 GHz. A sample of the measured data demonstrates the captured waveforms associated with distinct droplets (Fig. 4). Images from the high-speed video are presented in Fig. 4(a) and occurs at a location above the aperture of the waveguide so that there is minor time offset of 11–14 ms between the image capture and near-field detection. The measured jetting system pressure, as well as the return loss magnitude and phase associated with each droplet are indicated in Figs. 4(b)4(d). A high-speed video of the droplet formation at the nozzle tip and the corresponding real-time pressure and millimeter-wave signatures are provided in the supplementary material. The system print settings were not optimized to eject single droplets per pulse, in order to demonstrate performance variation in droplet jetting. As such, despite a periodic pressure signal of 5 Hz, it can be observed that droplets occur at varying delays from the pressure signal and in some cases, more than one droplet is produced as a result of a single pressure pulse. Droplet data across 4 s were processed using edge detection video image analysis (Fig. S3, supplementary material), and the variation of the minimum magnitude and corresponding phase is shown as a function of droplet diameter and spacing in Figs. 4(e) and 4(f). A general monotonic trend between diameter and minimum return loss magnitude, as well as between spacing and return loss angle, indicates the promise of this diagnostic approach to determine the droplet size and location properties in real-time and with minimal processing. The measured return loss magnitude variation with droplet spacing, as well as return loss phase variation with droplet diameter, are also shown in Fig. S3 (supplementary material).

FIG. 4.

Experimental near-field diagnostic results with LMJ-DoD using a 400 μm inner diameter nozzle tip. Pressure pulses were triggered electronically with set point of 108 kPa at 5 Hz and 1.5% duty cycle. (a) Images from high-speed video at different time stamps. The top corner of the waveguide (with polyimide lining) is visible. Scale bar: 2 mm. (b)–(d) Measured return loss magnitude, phase, and pressure during jetting. Dotted lines mark the time stamp when the corresponding images in (a) were taken. (e) Processed results for minimum return loss associated with each droplet and its corresponding diameter. (f) Phase at the time of minimum return loss and the corresponding spacing of the droplet.

FIG. 4.

Experimental near-field diagnostic results with LMJ-DoD using a 400 μm inner diameter nozzle tip. Pressure pulses were triggered electronically with set point of 108 kPa at 5 Hz and 1.5% duty cycle. (a) Images from high-speed video at different time stamps. The top corner of the waveguide (with polyimide lining) is visible. Scale bar: 2 mm. (b)–(d) Measured return loss magnitude, phase, and pressure during jetting. Dotted lines mark the time stamp when the corresponding images in (a) were taken. (e) Processed results for minimum return loss associated with each droplet and its corresponding diameter. (f) Phase at the time of minimum return loss and the corresponding spacing of the droplet.

Close modal

Experimental results with the 100 μm data demonstrate that detection of droplet events at the nozzle is also possible, even if the signal-to-noise ratio of the droplets (400–500 μm diameter) in front of the waveguide is too low for near-field detection. Detection of events at the nozzle is a result of variation in the far-field scattering cross section of the nozzle tip with and without metallic droplets and for varying droplet size (Fig. S4, supplementary material). These features are also present in the 400 μm data; however, they are less apparent due to the higher amplitude variation produced by the droplet as it passes in front of the waveguide. Figure 5(a) shows the measured return loss for pressure pulses at 80 Hz, 143 kPa, and a valve duty cycle of 32%.

FIG. 5.

Experimental diagnostic results with LMJ-DoD using a 100 μm inner diameter nozzle tip. Pressure pulses were triggered electronically with set point of 143 kPa at 80 Hz and 32% duty cycle. (a) Measured return loss magnitude over 4.25 s. (b) The inset of (a) showing the transition point ttr. (c) Initial attempt to eject droplet. The waveguide flange and aperture are visible in the high-speed camera frame. Scale bars: 4 mm (purple) and 1 mm (orange, inset). (d)–(f) Droplet contraction toward nozzle when the applied pressure is not sufficient to eject the droplet. (g) Droplet ejection when a pressure pulse is applied for the second time.

FIG. 5.

Experimental diagnostic results with LMJ-DoD using a 100 μm inner diameter nozzle tip. Pressure pulses were triggered electronically with set point of 143 kPa at 80 Hz and 32% duty cycle. (a) Measured return loss magnitude over 4.25 s. (b) The inset of (a) showing the transition point ttr. (c) Initial attempt to eject droplet. The waveguide flange and aperture are visible in the high-speed camera frame. Scale bars: 4 mm (purple) and 1 mm (orange, inset). (d)–(f) Droplet contraction toward nozzle when the applied pressure is not sufficient to eject the droplet. (g) Droplet ejection when a pressure pulse is applied for the second time.

Close modal

For simplicity, here we primarily focus on trends observed in the return loss magnitude. Based on the return loss amplitude variation, we see that there is a general transition point in the data at ttr=2.9 s. Prior to ttr there is more variation in the amplitude of the measured signal, whereas after ttr, the signal amplitude becomes more consistent (similar consistency appears in bursts before ttr as well). By observing the data near this transition point in Fig. 5(b), we see that there is a periodic dip in the captured signal after ttr. Prior to ttr, there is observable variation in the captured waveform. From the high-speed video images, we can see that features at the nozzle are being captured by the millimeter-wave diagnostic. A droplet initially forms at the nozzle [Fig. 5(c)]. When pressure is applied to eject the droplet, the droplet initially moves away from the nozzle [Fig. 5(d)]. However, due to surface tension, the applied pressure is not sufficient to eject the droplet, and the droplet swings back toward the nozzle [Figs. 5(e) and 5(f)] until the next pressure pulse is applied and the droplet is ejected [Fig. 5(g)]. A high-speed video of droplet formation and ejection at the nozzle tip and the corresponding real-time pressure and millimeter-wave signatures are again provided in the supplementary material. These results indicate that for the given system settings, some time is required for the system to reach steady state, possibly due to pneumatic stabilization. Furthermore, we see that the detection of print anomalies at the nozzle can be achieved, in addition to the evaluation of the properties of ejected droplets.

We have demonstrated in simulation and experiment the detection and characterization of millimeter-scale droplets in a droplet-on-demand liquid metal jetting system using a millimeter-wave diagnostic, as well as the detection of features at the print nozzle for droplets as small as 400–500 μm. To detect smaller droplet sizes down to 50–100 μm, the operational frequency can be increased for higher sensitivity at smaller length scales. Future work could also consider the use of pulsed or multi-frequency signals commonly used for radar systems, rather than single frequency continuous-wave transmission, to resolve variable length scales or Doppler velocity. However, it should be noted that a diagnostic based on pulsed interrogation would need to be synchronized in time with droplet trajectories. Additionally, while this work has focused on metal-based droplet jetting, the open-ended waveguide diagnostic also has potential utility for dielectric-based jetting systems. Scattering effects from dielectric droplets would vary based on size, shape, and material dielectric strength, but measurement sensitivity would need to be evaluated.

Because of the ability of this diagnostic to capture droplet properties based on signal amplitude and phase alone, it significantly reduces the data processing load associated with optical diagnostic approaches. In future work, signal processing techniques could be applied to correlate optical and millimeter-wave measurements such that various droplet properties could then be predicted solely based on millimeter-wave results. Our demonstration of this diagnostic indicates promise for future closed-loop feedback systems wherein rapid, real-time processing can be used to adjust print settings and ensure the quality of printed parts over the full build duration.

See the supplementary material for additional simulated and measured results, as well as high-speed videos of droplet ejection during liquid metal jetting with the corresponding measured signals.

This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract No. DE-AC52-07NA27344 and supported by the LLNL-LDRD Program under Project Nos. 19-ERD-008 and 18-SI-001. The document release number is LLNL-JRNL-824924.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
Z.
Luo
,
X.
Wang
,
L.
Wang
,
D.
Sun
, and
Z.
Li
, “
Drop-on-demand electromagnetic printing of metallic droplets
,”
Mater. Lett.
188
,
184
187
(
2017
).
2.
Y.
Idell
,
N.
Watkins
,
A.
Pascall
,
J.
Jeffries
, and
K.
Blobaum
, “
Microstructural characterization of pure tin produced by the drop-on-demand technique of liquid metal jetting
,”
Metall. Mat. Trans. A
50
,
4000
4005
(
2019
).
3.
V. A.
Beck
,
N. N.
Watkins
,
A. S.
Ashby
,
A. A.
Martin
,
P. H.
Paul
,
J. R.
Jeffries
, and
A. J.
Pascall
, “
A combined numerical and experimental study to elucidate primary breakup dynamics in liquid metal droplet-on-demand printing
,”
Phys. Fluids
32
,
112020
(
2020
).
4.
T.
Wang
,
T.-H.
Kwok
,
C.
Zhou
, and
S.
Vader
, “
In-situ droplet inspection and closed-loop control system using machine learning for liquid metal jet printing
,”
J. Manuf. Syst.
47
,
83
92
(
2018
).
5.
V.
Sukhotskiy
,
K.
Tawil
, and
E.
Einarsson
, “
Printability regimes of pure metals using contactless magnetohydrodynamic drop-on-demand actuation
,”
Phys. Fluids
33
,
053303
(
2021
).
6.
D.-Y.
Shin
and
M.
Kim
, “
Rapid jetting status inspection and accurate droplet volume measurement for a piezo drop-on-demand inkjet print head using a scanning mirror for display applications
,”
Rev. Sci. Instrum.
88
,
025109
(
2017
).
7.
J.
Huang
,
L. J.
Segura
,
T.
Wang
,
G.
Zhao
,
H.
Sun
, and
C.
Zhou
, “
Unsupervised learning for the droplet evolution prediction and process dynamics understanding in inkjet printing
,”
Add. Manuf.
35
,
101197
(
2020
).
8.
T.
Chang
,
O.
Mays
,
S.
Mukherjee
,
N.
Watkins
,
A.
Pascall
,
J.
Jeffries
, and
J.
Tringe
, “An in-situ millimeter-wave diagnostic for droplet characterization during jetting-based additive manufacturing processes,” in
Nondestructive Characterization and Monitoring of Advanced Materials, Aerospace, Civil Infrastructure, and Transportation IX
(International Society for Optics and Photonics, 2020), Vol. 11380, p. 1138008.
9.
T.
Chang
,
S.
Mukherjee
,
N. N.
Watkins
,
D. M.
Stobbe
,
O.
Mays
,
E. V.
Baluyot
,
A. J.
Pascall
, and
J. W.
Tringe
, “
In-situ monitoring for liquid metal jetting using a millimeter-wave impedance diagnostic
,”
Sci. Rep.
10
,
22325
(
2020
).
10.
F. E.
Gardiol
, “Open-ended waveguides: Principles and applications,” in Advances in Electronics and Electron Physics, edited by P. W. Hawkes (Academic Press, 1985), Vol. 63, pp. 139–187.
11.
L.
Diener
, “
Microwave near-field imaging with open-ended waveguide—comparison with other techniques of nondestructive testing
,”
Res. Nondestr. Eval.
7
,
137
152
(
1995
).
12.
F.
Mazlumi
,
S. H. H.
Sadeghi
, and
R.
Moini
, “
Detection of surface cracks of arbitrary shape in metals using an open-ended waveguide probe
,”
AIP Conf. Proc.
615
,
491
497
(
2002
).
13.
J.
Nadakuduti
,
G.
Chen
, and
R.
Zoughi
, “
Semiempirical electromagnetic modeling of crack detection and sizing in cement-based materials using near-field microwave methods
,”
IEEE Trans. Instrum. Meas.
55
,
588
597
(
2006
).
14.
N. N.
Qaddoumi
,
W. M.
Saleh
, and
M.
Abou-Khousa
, “
Innovative near-field microwave nondestructive testing of corroded metallic structures utilizing open-ended rectangular waveguide probes
,”
IEEE Trans. Instrum. Meas.
56
,
1961
1966
(
2007
).
15.
M. R.
Ramzi
,
M.
Abou-Khousa
, and
I.
Prayudi
, “
Near-field microwave imaging using open-ended circular waveguide probes
,”
IEEE Sens. J.
17
,
2359
2366
(
2017
).

Supplementary Material