Abstract

This paper focuses on the influence of carrier gas flow rate (CGFR) and sheath gas flow rate (SGFR) on the quality of conductive traces printed with nanoparticle inks using aerosol jet printing (AJP). This investigation was motivated by previous results of two AJP specimens that were printed at different gas flow rates and yielded significantly different thermal cycling durability lifetimes. A parametric sensitivity study was executed by printing and examining serpentine trace structures at 15 different combinations of CGFRs and SGFRs. The analysis included quantifying the trace's macroscale geometry, electrical properties, and micromorphological features. Interesting macroscale results include an increase in effective conductivity with increasing CGFR. At the microscale, image processing of high magnification scanning electron microscope (SEM) images of the printed traces revealed that agglomerations of silver clusters on the surface of traces became coarser at higher CGFR and also that agglomerates in the bulk were finer than those on the surface. Crystalline silver deposits were observed at all flow rates. In addition, cross sectioning of the printed traces showed higher incidences of buried cohesive cracking at higher gas flow rates. These cohesive cracks reduce the robustness of the traces but may not always be visible from the surface. The degree of cohesive cracking was seen to be broadly correlated with the coarseness of the surface agglomerates, thus suggesting that the coarseness of surface agglomerates may provide a visible surrogate measure of the print quality. The results of this study suggest that print quality may degrade as gas flow rates increase.

1 Introduction

Aerosol jet printing (AJP) is a direct-write printing technology built on the fundamentals of additive manufacturing and is capable of printing micron-scale features [117]. Conductive traces for electronic applications have been AJ printed using inks that were formulated with colloidal suspensions of metallic nanoparticles in a carrier solvent. Such inks are atomized to create a dense aerosol of micron-scale droplets, with diameters typically ranging from 1 to 5 μm [18]. The aerosolized ink droplets are carried by nitrogen gas, denoted as the carrier gas, to the deposition head through a mist tube and then focused within the deposition head by a sheath gas before being sprayed onto a substrate. In addition to focusing and collimating the droplets within the carrier gas flow, the sheath gas also forms an outer gas layer around the aerosolized ink stream to prevent contact between the aerosol droplets and the inner walls of the deposition head (which in turn helps to prevent the condensation of ink and hence helps prevent clogging). Figure 1 shows the schematic of an ink stream as it (1) enters the deposition head's nozzle chamber from the mist tube, (2) travels through the nozzle, and (3) exits the nozzle to travel to the substrate [19]. These AJP features and their consistency are related to the quality of the printed traces. To maintain consistent print quality, the effect of printing parameters and ink condition must be well understood. However, the focus of this study is specifically on the influence of carrier (CGFR) and sheath gas flow rates (SGFR), and not on the overall comprehensive effect of all the printing parameters listed above. Only one specific ink has been used in this study to eliminate the influence of ink properties. The quality of the printed traces is examined in terms of important macroscale and microscale features of printed conducting traces.

A prior study by the authors revealed that the microstructure and reliability of two AJP fabricated conducting traces, both printed with a selected commercial ink, were notably different, despite using identical values for the majority of the print parameters such as ink stream deposition rate, print speed, and build-plate temperature [20]. The ink consisted of silver nanoparticles in an organic solvent. The first sample (labeled here as IoF for interconnect over fillet) was printed with a carrier gas flow rate (CGFR) of 80 standard cubic centimeters per minute (sccm) and a SGFR of 50  sccm, while the second sample (labeled here as DTC for DOWA test coupon) was printed with a CGFR of 50  sccm and a SGFR of 30  sccm. As described in a separate publication by the authors, a larger bulk porosity was measured in the IoF sample as compared to the DTC sample [20]. Furthermore, delamination of sections of the IoF printed traces was observed after only 250 accelerated thermal cycles (from −40 to 125 °C), while no degradation was observed in the DTC sample after over 3500 accelerated thermal cycles. The suspected differences between these two printed samples included: (i) the ink condition and (ii) the CGFR and SGFR used in the AJP process. The different gas flow rates used to print these two samples was needed to achieve an ink stream deposition rate of 0.00075 mm3/s for both samples and is most likely related to a suspected difference in the solvent content of the ink used in the two print runs. It is a common practice to add solvent back into the ink after several hours of printing because the aerosolization of the ink tends to cause the ink to dry out. Drying of the ink is usually accompanied by a need to increase the CGFR in order to maintain the ink stream deposition rate. Previously, this increase in CGFR within an empirically determined acceptable range has not been seen as a detriment. However, in light of these recent observations, we now hypothesize that the different ink conditions and the different flow rates have both contributed to the differences in the quality and reliability of the IoF and DTC prints. This study focuses on the influence of the gas flow rates on the printed features, while the influence of ink condition is deferred to a future study. As such, the test coupons described in Sec. 2 will be printed with a single lot of ink, within a short time interval, to ensure that the ink condition is the same for all of the specimens.

Chen et al. [19] developed a three-dimensional computational fluid dynamics model of the aerosol CGFR as it is confined by an annular SGFR in order to study the fundamental fluid mechanics principles that control the overspray as a function of droplet size distribution and SGFR. Within AJP, overspray refers to the ink particles that are deposited outside of the intended ink stream width. Their results explain that there is an abundance of smaller sized drops in the overspray region at low SGFRs, the overspray first reduces and then increases as the SGFR increases, and there is no longer a prevalence of smaller particles in the overspray region at larger SGFRs. However, there is a point of diminishing return for SGFR because at high enough levels, the particle droplets start to cross the nozzle axis and travel back toward the opposite edge of the ink stream resulting in over collimation and increased overspray. A further study of direct-write printed traces was conducted by Roberson et al. [21]. They characterized the morphology of inkjet printed traces made from both a microparticle and a nanoparticle silver ink. Scanning electron microscope (SEM) images of traces printed with these two different particle sized inks as well as varying sintering profiles were provided. There are limited results on the morphology of AJP traces, but Kang et al. [22] also characterized morphology of inkjet printed traces, another direct-write technology, with a copper nanoparticle ink. Additional studies of the AJP process were conducted by Mahajan et al. [23] where they found that the width of a printed trace decreased with increasing focusing ratio (FR), defined as FR = SGFR/CGFR, as well as print speed. The thickness or height of the traces also increased with an increasing focusing ratio, but decreased with increasing print speed. The decrease in line width with increasing focusing ratio was more dramatic for smaller nozzles diameters. The general trends investigated by Mahajan are summarized in Table 1. Furthermore, Salary et al. [24] also developed a computational fluid dynamics model that verified the overall line morphology trends that they observed through an in situ monitoring camera when varying flow rate parameters. For a fixed CGFR of 30  sccm and print speed of 1 mm/s, they recommended a process window between 40 and 100  sccm for the SGFR to optimize the printed line density, edge quality, overspray, and line discontinuity.

While there has been some endeavor to understand detailed effects of different printing parameters on the final quality of printed traces, AJP still remains a rather empirical process with the need for a much more quantitative understanding of advanced processing methods. For example, operational process conditions that others have suggested appear to be specific to the AJ printer used for their study and the type of ink being utilized. Furthermore, less quantifiable parameters such as printer maintenance as well as parts cleaning and assembly are also critical issues that need to be addressed in more detail. Moreover, the definition of print quality as a function of print parameters has been mostly qualitative aside from quantifying a line's overall macroscale geometry such as ink stream width and thickness, and average bulk electrical properties such as effective conductivity. Few studies have investigated the micromorphology of printed traces [22,23], such as its average agglomeration length scales in the interior of a printed trace as well as on the surface of the trace. Therefore, in this study, the quality of printed serpentine test structures is characterized not only by a set of selected macroscale features but also using selected micromorphological features. In this way, expected results such as increased deposition rate and trace cross-sectional area with CGFR were verified along with additional correlations showing that (i) effective electrical conductivity is not constant across varying flow rates but generally increases with increasing CGFR; (ii) the presence of silver crystalline deposits appears on the surface of printed traces independent of flow rate; (iii) coarser sized agglomerates on the surface were seen for the traces printed at higher CGFR; and (iv) the agglomerates in the interior of a trace are finer than those on the surface. Such new findings will be beneficial to the 3D printing community and are expected to help advance AJP technology.

2 Approach

The study used a sensitivity study, where the two flow parameters (CGFR and SGFR) are parametrically varied and their influence on selected response variables (printed features) are quantified.

2.1 Test Coupon.

Test structures were fabricated by first spin-coating a commercially available polymer [NEA121 or Norland Electronic Adhesive 121 denoted as Polymer N in the remainder of the paper] onto an FR4 substrate and then AJ printing a commercially available silver nanoparticle ink [DOWA DF-SLO-002 Ink manufactured in Japan, referred to as Ink A in the remainder of the paper] onto the surface of the polymer dielectric layer. The aerosol jet printer was manufactured by Optomec, Inc. and utilized a pneumatic atomizer. Ink A had a viscosity of 1.1 Pa⋅s and 73.4 weight percent silver nanoparticles in butyl carbitol. Other parameters include a nozzle diameter of 300 microns, nozzle length of 7.5 mm, printer stage temperature of 50 °C, print speed of 1 mm/s, and dry nitrogen gas used as the carrier and sheath gases. The printed traces were sintered in an oven at 150 °C following a brief dwell at 80 °C as shown in Fig. 2(a). The two-step sintering process reduces the chance of abrupt vaporization of the ink's binder.

Each test structure consisted of a five-segment serpentine structure with 20 mm long segments each separated by 1 mm (Fig. 2(b)). For functional applications, it has been typical to print two passes in order to (i) build up trace thickness and (ii) smooth out short time scale variations in the ink stream deposition rate. However, it is expected that a single-pass trace would show greater variation between different flow rates and therefore, it is desirable to print at least some single pass traces in order to not mask these effects. Therefore, the bottom two segments (A-B) of each serpentine (highlighted in black in Fig. 2(b)) were single-pass printed while the top three segments (C-E) of each serpentine, (highlighted in white in Fig. 2(b)) were double-pass printed.

2.2 Test Plan.

Ink A was used to print the aforementioned serpentine test structures. A total of 15 test structures were printed across five test coupons with three serpentine structures on each coupon, where the test structures on each coupon were printed using CGFR settings of 30, 50, 65, 80, and 100  sccm. On each test coupon, the three serpentine structures were printed at different SGFRs of 30, 40, or 50  sccm (Table 2). The choice of the upper and lower flow rates were guided by recommendations from the manufacturer of the aerosol jet printer and by prior experience of the authors. Intermediate flow rates were selected to provide a more complete trend with more data points. The printing sequence was not randomized as part of the parametric sensitivity study.

In an effort to assess the quality of each of the serpentine structures printed at these various flow rate combinations, several print features were investigated. A flowchart of the methodology is presented in Fig. 3:

  1. The corresponding ink stream deposition rate was measured using the inkwell technique prior to the printing of each of the 15 serpentines [25].

  2. Laser profilometry was conducted on the postsintered printed structures in order to determine multiple macroscale geometric features (i.e., trace width, thickness, and overspray).

  3. The electrical resistance values of the printed structures were also measured to assess their effective conductivity.

  4. Scanning electron microscope images were taken of printed surfaces in order to examine other macroscale features such as crystalline deposits and longitudinal cracks.

  5. Scanning electron microscope images were taken of both the surface and the cross sectioned interior of printed samples in order to analyze the micromorphology of these printed traces.

  6. Cross sectioned samples were also examined for cohesive cracking. Some relevant conclusions are presented here while other details of such findings are presented elsewhere in the literature [26].

3 Results

The quality of printed serpentine traces with varying flow rate parameters was quantified by measuring the deposition rate, macroscale geometric properties, effective conductivity, as well as the distributions of agglomeration size and spacing, both on the surface as well as in the interior of the printed trace. The presence of microcracking in certain AJP traces is briefly noted here but discussed in detail in a separate publication [26].

3.1 Deposition Rate.

As expected, due to the varying gas flow rates at which material is being transported in the AJP process, the ink steam deposition rate associated with each combination of flow rates was different. The ink stream deposition rate was measured for each flow rate combination by noting the time required to fill a calibrated inkwell of known volume (0.0063 mm3). This method for the measurement of the ink stream deposition rate was established and discussed in detail by Gu et al. [25]. The corresponding ink stream deposition rate for each gas flow rate combination is plotted in Fig. 4 and is seen to increase with increasing CGFR and decrease with increasing SGFR. Therefore, an increase in CGFR is correlated with more material being drawn from the ink reservoir and an increase in SGFR is correlated with a “pinching off” of the ink stream as it travels through the nozzle. In the AJP process, CGFR is the difference between the atomization gas flow rate and exhaust gas flow rate. The exhaust gas flow rate was kept constant at 1400  sccm and the atomization gas flow rate was varied to achieve the desired CGFR.

3.2 Macroscale Geometric Features.

A commercial confocal microscope was used to perform laser profilometry measurements on the full set of printed serpentine traces. Using an analysis program developed in matlab by Chen et al. the trace width, cross-sectional area, maximum thickness, overspray extent, and area coverage were calculated from the height map contour generated from the profilometry data [19]. An example of a height map contour plot and output from the analysis program of Chen et al. is presented in Fig. 5. In brief, threshold values are set in order to define the edge of the printed trace width and overspray region as labeled in the figure. From these threshold settings, an analysis of the trace width, cross-sectional area, maximum thickness, overspray extent on each side of the printed trace, and a percent coverage in the overspray region can all be determined quantitatively. Although the profilometry methodology accounts for the effects of substrate flatness on the measurement accuracy of trace thickness, a future study is needed to understand the direct relationship between substrate flatness and trace thickness.

For the four extreme case samples within the study, that is (CGFR, SGFR) of (30, 30), (30, 50), (100, 30), and (100, 50) as well as for the middle case (65, 40), multiple laser profilometry scans were performed along the single-pass portion of the serpentine. A total of 14 profilometer scans were acquired at evenly spaced intervals along the 40 mm length of the two single-pass trace segments. The resulting probability density distributions for the trace width, cross-sectional area, and overspray extent are all plotted in Fig. 6 and the associated mean and standard deviation values are summarized in Table 3. The distributions in Fig. 6 have been normalized to the mean values of the middle case of CGFR = 65, SGFR = 40. As an example, histogram of the actual readings has been included for one sample graph, Fig. 6(b). In the interest of clarity, such histograms could not be included for all the graphs in Fig. 6. Trace width, cross-sectional area, and maximum thickness are all seen to increase with increasing gas flow rates; however, the effect of SGFR was much smaller than CGFR. The printed line's coverage in the overspray region is seen to increase marginally with an increase in either CGFR or SGFR. The overspray appears to remain reasonably constant across these five flow rate combinations; however, it is slightly elevated in the (100, 30) trace because of a decrease in ink stream confinement associated with the lower SGFR value. This effect is also observed as a larger spread in the trace width distribution of the (100, 30) trace.

In addition to the scans for the statistical assessment for five of the serpentines, two laser profilometry scans were conducted on each of the 15 serpentine traces printed in the experiment matrix, one over a single-pass trace and one over a double-pass trace. These results are presented elsewhere [20].

3.3 Effective Electrical Conductivity.

Test samples were mounted on a commercial, precision probe station and connected to a commercial, precision LCR meter, a device used to measure inductance (L), capacitance (C), and resistance (R), in order to measure two-probe resistance values of the serpentine traces described above. A four-point resistance measurement was not possible due to the micron length scale traces. Because of the potential of an added contact resistance, normalized and not absolute values are provided and the results should be interpreted in a semiqualitative sense. A decrease in resistance as a function of increasing CGFR was observed; additionally, the trace resistance was nearly independent of SGFR for any given CGFR.

Aerosol jet printed traces consisting of metal nanoparticles sintered at a given temperature exhibit both micro and nanoporous structures; therefore, bulk metrology measurements relate to effective material properties rather than bulk properties. As a result, resistivity (or conductivity) measurements related to AJ printed traces will be higher (or lower) than the typical values for the bulk material, which for silver is ρ = 1.59 × 10−8 Ohm-m (σ = 6.3 × 107 S/m) [27]. Ink A has been reported to have a resistivity value, when sintered at 150 °C, less than 8.75 × 10−8 Ohm-m which is 5.5 times larger than that of bulk silver and corresponds to a conductivity value of 1.1 × 107 S/m [28]. In order to more easily observe relative trends, the normalized effective conductivity of each printed serpentine trace was calculated and is plotted versus flow rates in Fig. 7. These values were calculated by taking the measured DC resistance of a trace, dividing it by the length of the trace and multiplying it by the cross-sectional area of the trace to get the effective resistivity, ρ = R*A/L, and then taking the inverse, σ = 1/ρ. The cross-sectional area was measured from the profilometry data by using Chen's algorithm as described above and presented in Fig. 5 [19]. Note that there is no data point for either the (30, 30) or the (30, 40) traces. This is a result of a low deposition rate for these two setting that produced printed traces below the percolation threshold. It is interesting to note that the effective conductivity is not constant but rather increases with increased CGFR. As conductivity is a bulk material parameter, such a change in effective conductivity implies a systematic change in microstructure within the printed traces as a function of printing parameters. Indeed, Merilampi et al. also reported that flow rate appears to affect printed trace conductivity in a screen-printing study [29]. Separately, Rahman et al. showed that the electrical behavior is correlated with the grain growth of the AJP silver nanoparticles films and can be tailored by tweaking sintering conditions and as a result the microstructure [30].

3.4 Surface Features.

Optical microscopy and SEM was utilized to investigate surface morphology of printed serpentine traces as a function of varying flow rate parameters. For the SEM images, all 15 serpentine traces were imaged at 500× magnification, both for single-pass and double-pass segments. As seen in Fig. 8, there is no noticeable difference between single and double pass segments. However, crystalline deposits are observed in many, but not all, of these printed traces, as shown in Figs. 8 and 9. Of the traces that have these deposits, some were (1) aggregated mostly along the edge as in Fig. 9(b), (2) concentrated mostly around the centerline as in Fig. 9(c), or (3) randomly dispersed throughout the width of the trace as in Fig. 9(d). Figure 9(a) also shows a case without any deposits. Energy dispersive X-ray spectroscopy reveals that these crystalline deposits are likely to be of silver composition as the elemental spectroscopy curves of the deposits are nearly identical to those of the printed trace in a nearby area with no deposit (Fig. 10). The quantitative composition of both areas sampled within a given trace was over 97% silver for all measured traces. Similar crystalline deposits in printed metal features have been reported elsewhere in the literature [31].

3.5 Micromorphology.

High-resolution SEM imaging of both surface and cross-sectional regions of all 15 serpentine traces was also performed at a magnification of 50,000× for both single and double pass trace segments in order to investigate micromorphological features. The complete matrix of images for all flow rate combinations for both single and double pass traces is provided in Figs. 19 and 20 available in the Supplemental Materials on the ASME Digital Collection. Two characteristic images, at CGFR = 30, SGFR = 30 and CGFR = 100, SGFR = 50, (the lower and upper bounds of the range for these two parameters) for each case, single and double pass, as well as surface and cross section, are provided here as samples, in Figs. 11 and 12.

Instead of forming a continuous, uniform bulk material, the traces appear as an agglomeration of particles that are on an order of 10–200 nm in diameter. For the high magnification surface images (Fig. 11), coarser agglomerate sizes are observed at the higher flow rates, which follows the trend seen in a previous study CGFR = 80, SGFR = 50 and CGFR = 50, SGFR = 30 specimens [20]. However, the cross-sectional images at the same length scale do not show this trend (Fig. 12). Furthermore, for both the surface (Fig. 11) and the cross section images (Fig. 12), there was no significant difference in micromorphology between single and double-pass traces.

These high magnification images were digitally postprocessed to quantify the size and spacing distribution of the agglomerates as described elsewhere [20]. The raw data extracted by the software included the agglomerate size and spacing (shortest distance from one agglomerate to the nearest neighbors) for each counted feature within an image. These data were then postprocessed to obtain various stochastic metrics (e.g., mean values, standard deviations, skewness, etc.).

Five images were postprocessed for each of the four extreme case samples (30, 30), (30, 50), (100, 30), and (100, 50) as well as for the middle case sample (65, 40) that were previously investigated in Sec. 3.2. The postprocessing was repeated for both single-pass and double-pass trace segments as well as for both surface images and cross section images. The size and spacing distributions for each image was normalized before averaging across the five images in order to avoid one image from dominating if there was an unusually high number of agglomerates counted by the image processing software. Various metrics from this analysis were examined, as listed below, but only the metrics indicating trends are discussed below:

  1. number of agglomerates per unit area,

  2. percent area covered by the agglomerates,

  3. agglomerate size and spacing distributions,

    • (a)

      variation in agglomerate size and spacing,

    • (b)

      average agglomerate size and spacing,

    • (c)

      standard deviation of agglomerate size and spacing, and

    • (d)

      skewness of agglomerate size and spacing.

As a result of this quantitative image processing, the qualitative trend of coarser surface agglomerations at higher CGFR, observed in Fig. 11, is quantitatively observed in Fig. 13. The agglomerate size distributions on the surface for a single-pass trace are plotted for three flow rate combinations, (30, 50), (65, 40), and (100, 30), along the minor diagonal of the test matrix. The trace with the lowest CGFR of 30  sccm shows predominantly smaller agglomerates, while the trace with the highest CGFR of 100  sccm shows predominantly larger agglomerates. This supports the previous observations made from looking at the raw, high magnification surface images that the agglomerate sizes are coarser at higher CGFRs (Figs. 11 and 12). Consistent with this trend, for the intermediate CGFR of 65  sccm agglomerate sizes are in between those for the end cases of 30  sccm and 100  sccm. The proportional relationship between agglomerate size and CGFR is also supported in Fig. 14 where the distributions are averaged and the agglomerates get coarser as CGFR increases.

In Fig. 15, the agglomerate spacing distribution in the bulk for a double-pass trace is plotted for the three flow rate combinations, (30, 30), (65, 40), and (100, 50), along the major diagonal of the test matrix. The trace with the lowest CGFR (30  sccm) had predominantly smaller spacing between agglomerates, while the trace with the highest CGFR (100  sccm) had predominantly larger spacing. Consistent with this trend, the intermediate CGFR of 65  sccm predominantly has particle sizes that are in between the end cases of 30  sccm and 100  sccm. This also supports the observation that coarser agglomerates were occurring at higher CGFRs.

The qualitative observation in Fig. 12, of a lack of trend between gas flow rate and agglomeration size in the interior of the traces, is quantitatively confirmed in Fig. 16, where no distinct trend is observed. However, it is a significant observation that the agglomerates on the surface appear to be significantly larger than those observed within the interior of the trace, as evidenced by the histograms in Fig. 17. This was the case for both the serpentine trace with the lowest flow rate combination (30, 30) and the highest flow rate combination (100, 50). The average agglomerate size on the surface for each of these two flow rate cases is 0.008 and 0.022 μm2, respectively, while in the interior, it is 0.003 and 0.001, respectively. This suggests that the clumping of the nanoparticles within a printed trace becomes progressively greater as the trace stacks up from the substrate to the surface.

Higher CGFR ensures a larger deposition rate. Accordingly, there will be a larger mass of ink deposited on the substrate (Fig. 4). Therefore, during the sintering process, a larger mass of the solid will take part in agglomerate formation leading to the formation of coarser aggregates. Larger size of aggregates implies more material and apparently leads to a larger electrical conductivity of the traces. This is evident in Fig. 7. Aggregate formation for nanoparticle system indicates that the chances of crack formation have been reduced. This issue has been extensively discussed elsewhere, including briefly in Sec. 3.6 [26]. On the other hand, coarser agglomerates correspond to higher porosity and can hence lead to poorer mechanical properties.

3.6 Microcracking.

Another discovery from the low magnification surface imaging was the presence of longitudinal microcracking at the higher CGFR (Fig. 18(a)). Upon cross section inspection, cracks were observed in a variety of locations including in the middle of single-pass traces (Fig. 18(b)). The microcracking is most likely due to a combination of shrinkage stresses and capillary pressure, as the volatiles in the ink escape during the sintering process. Simple 1D models of capillary pressure in the literature suggest that microcracking risks may be high for trace thickness beyond 10–100 nm and may further increase with increasing trace thickness [32]. The principle of such crack formation is the evaporation-mediated agglomeration of the nanoparticles, which narrows the passage for the flows leading to a buildup of the capillary pressure, which eventually leads to the formation of periodic cracks. The relevance of these simple cracking models is discussed in full elsewhere [26].

4 Summary

The desire to understand the effects of CGFR and SGFR on the quality of an aerosol jet printed traces motivated a parametric sensitivity study where 15 serpentine traces were printed at varying flow rates and then analyzed in terms of print quality. Flow rate values of 30, 50, 65, 80, and 100  sccm were chosen for the CGFR and of 30, 40, and 50  sccm for the SGFR. Two segments within each serpentine sample were printed as a single-pass, while the remaining three segments were printed as a double-pass. The second pass was investigated since multiple passes are typically employed in the printing of traces for functional applications. For each flow rate combination, the deposition rate was measured and the influence of increasing CGFR and decreasing SGFR on the increase of deposition rate was quantified. Macroscale geometric features such as trace width, cross-sectional area, and overspray extent as well as coverage were quantified by laser profilometry. The flow rate of the CGFR had a larger impact relative to that of the SGFR, for most of these macroscale geometric quantities. The effective electrical conductivity of the printed silver traces was observed to increase with an increase in CGFR. Large silver crystalline deposits were observed and mostly appear to be randomly present on the surface of the majority of traces somewhat independent of flow rates. High magnification imaging of printed trace surfaces and digital image postprocessing revealed that agglomerate coarseness (size and spacing) increased as the CGFR increased. The agglomerate coarseness within the interior of the trace cross section did not show a definitive trend with flow rates, but the largest agglomerates found in the trace's cross section were noticeably smaller than those found on the surface of those same traces. Longitudinal microcracking was seen from an overhead view of the trace surface for the traces with higher CGFRs and these were also observed in cross section images. The phenomenon of cohesive cracking is reported in detail by the authors elsewhere in the literature [26]. These findings suggest that lower flow rates should be used to minimize microcracking and coarser agglomerations that cause the printed trace quality and properties to be different from bulk and homogeneous silver. Given the lower deposition rates at lower CGFRs, additional passes or a lower print speed can be utilized to still reach the desired material deposition volumes.

Acknowledgment

This study was funded by the sponsors of the NextFlex consortium. In particular, we are grateful for technical inputs from the Laboratory for Physical Sciences.

Funding Data

  • The sponsors of the NextFlex consortium (Award No. 028328-00001).

References

References
1.
Seifert
,
T.
,
Sowade
,
E.
,
Roscher
,
F.
,
Wiemer
,
M.
,
Gessner
,
T.
, and
Baumann
,
R. R.
,
2015
, “
Additive Manufacturing Technologies Compared: Morphology of Deposits of Silver Ink Using Inkjet and Aerosol Jet Printing
,”
Ind. Eng. Chem. Res.
,
54
, pp.
769
779
.10.1021/ie503636c
2.
Binder
,
S.
,
Glatthaar
,
M.
, and
Radlein
,
E.
,
2014
, “
Analytical Investigation of Aerosol Jet Printing
,”
Aerosol. Sci. Technol.
,
48
(
9
), pp.
924
929
.10.1080/02786826.2014.940439
3.
Secor
,
E. B.
,
2018
, “
Principles of Aerosol Jet Printing
,”
Flexible Printed Electron.
,
3
(
3
), p.
035002
.10.1088/2058-8585/aace28
4.
Rahman
,
T.
,
Renaud
,
L.
,
Heo
,
D.
,
Renn
,
M.
, and
Panat
,
R.
,
2015
, “
Aerosol Based Direct-Write Micro-Additive Fabrication Method for Sub-mm 3D Metal-Dielectric Structures
,”
J. Micromech. Microeng.
,
25
(
10
), p.
107002
.10.1088/0960-1317/25/10/107002
5.
Seifert
,
T.
,
Baum
,
M.
,
Roscher
,
F.
,
Wiemer
,
M.
, and
Gessner
,
T.
,
2015
, “
Aerosol Jet Printing of Nano Particle Based Electrical Chip Interconnects
,”
Mater. Today
,
2
, pp.
4262
4271
.10.1016/j.matpr.2015.09.012
6.
Hoey
,
J. M.
,
Lutfurakhmanov
,
A.
,
Schulz
,
D. L.
, and
Akhatov
,
I. S.
,
2012
, “
A Review on Aerosol-Based Direct-Write and Its Applications for Microelectronics
,”
J. Nanotechnol.
, 2012, p.
324380
.10.1155/2012/324380
7.
Jones
,
C. S.
,
Lu
,
X.
,
Renn
,
M.
,
Stroder
,
M.
, and
Shih
,
W.-S.
,
2010
, “
Aerosol-Jet-Printed, High-Speed, Flexible Thin-Film Transistor Made Using Single-Walled Carbon Nanotube Solution
,”
Microelectron. Eng.
,
87
(
3
), p.
434
.10.1016/j.mee.2009.05.034
8.
Rother
,
M.
,
Brohmann
,
M.
,
Yang
,
S.
,
Grimm
,
S. B.
,
Schießl
,
S. P.
,
Graf
,
A.
, and
Zaumseil
,
J.
,
2017
, “
Aerosol Jet Printing of Polymer Sorted (6,5) Carbon Nanotubes for Field‐Effect Transistors With High Reproducibility
,”
Adv. Electron. Mater.
,
3
(
8
), p.
1700080
.10.1002/aelm.201700080
9.
Salary
,
R.
,
Lombardi
,
J. P.
,
Rao
,
P. K.
, and
Poliks
,
M. D.
,
2017
, “
Online Monitoring of Functional Electrical Properties in Aerosol Jet Printing Additive Manufacturing Process Using Shape-From-Shading Image Analysis
,”
ASME J. Manuf. Sci. Eng.
,
139
(
10
), p.
101010
.10.1115/1.4036660
10.
Cao
,
C.
,
Andrews
,
J. B.
, and
Franklin
,
A. D.
,
2017
, “
Completely Printed, Flexible, Stable, and Hysteresis‐Free Carbon Nanotube Thin‐Film Transistors Via Aerosol Jet Printing
,”
Adv. Electron. Mater.
,
3
(
5
), p.
1700057
.10.1002/aelm.201700057
11.
Deiner
,
L. J.
, and
Reitz
,
T. L.
,
2017
, “
Inkjet and Aerosol Jet Printing of Electrochemical Devices for Energy Conversion and Storage
,”
Adv. Eng. Mater.
,
19
(
7
), p.
1600878
.10.1002/adem.201600878
12.
Smith
,
M.
,
Choi
,
Y. S.
,
Boughey
,
C.
, and
Kar-Narayan
,
S.
,
2017
, “
Controlling and Assessing the Quality of Aerosol Jet Printed Features for Large Area and Flexible Electronics
,”
Flexible Printed Electron.
,
2
(
1
), p.
015004
.10.1088/2058-8585/aa5af9
13.
Yang
,
C.
,
Zhou
,
E.
,
Miyanishi
,
S.
,
Hashimoto
,
K.
, and
Tajima
,
K.
,
2011
, “
Preparation of Active Layers in Polymer Solar Cells by Aerosol Jet Printing
,”
ACS Appl. Mater. Interfaces
,
3
(
10
), pp.
4053
4058
.10.1021/am200907k
14.
Liu
,
R.
,
Ding
,
H.
,
Lin
,
J.
,
Shen
,
F.
,
Cui
,
Z.
, and
Zhang
,
T.
,
2012
, “
Fabrication of Platinum-Decorated Single-Walled Carbon Nanotube Based Hydrogen Sensors by Aerosol Jet Printing
,”
Nanotechnology
,
23
(
50
), p.
505301
.10.1088/0957-4484/23/50/505301
15.
Gu
,
Y.
,
Park
,
D.
,
Bowen
,
D.
,
Das
,
S.
, and
Hines
,
D. R.
,
2018
, “
Direct-Write Printed, Solid-Core Solenoid Inductors With Commercially Relevant Inductances
,”
Adv. Mater. Technol.
, 4(1), p.
1800312
.10.1002/admt.201800312
16.
Saleh
,
M. S.
,
Hu
,
C.
, and
Panat
,
R.
,
2017
, “
Three-Dimensional Microarchitected Materials and Devices Using Nanoparticle Assembly by Pointwise Spatial Printing
,”
Sci. Adv.
,
3
(
3
), p.
e1601986
.10.1126/sciadv.1601986
17.
Saleh
,
M. S.
,
Li
,
J.
,
Park
,
J.
, and
Panat
,
R.
,
2018
, “
3D Printed Hierarchically-Porous Microlattice Electrode Materials for Exceptionally High Specific Capacity and Areal Capacity Lithium Ion Batteries
,”
Addit. Manuf.
,
23
, pp.
70
78
.10.1016/j.addma.2018.07.006
18.
Hon
,
K.
,
Li
,
L.
, and
Hutchings
,
I. M.
,
2008
, “
Direct Writing Technology—Advances and Developments
,”
CIRP Ann.
,
57
(
2
), pp.
601
620
.10.1016/j.cirp.2008.09.006
19.
Chen
,
G.
,
Gu
,
Y.
,
Tsang
,
H.
,
Hines
,
D. R.
, and
Das
,
S.
,
2018
, “
The Effect of Droplet Sizes on Overspray in Aerosol-Jet Printing
,”
Adv. Eng. Mater.
,
20
(
8
), p.
1701084
.10.1002/adem.201701084
20.
Dalal
,
N.
,
2018
, “
Influence of Gas Flow Rates on Trace Quality and Reliability in a Selected Conductor Ink Printed with an Aerosol Jet Printer
,” M.S. thesis, University of Maryland, College Park, MD.
21.
Roberson
,
D. A.
,
Wicker
,
R. B.
,
Murr
,
L. E.
,
Church
,
K.
, and
MacDonald
,
E.
,
2011
, “
Microstructural and Process Characterization of Conductive Traces Printed From Ag Particulate Inks
,”
Materials
,
4
(
6
), pp.
963
979
.10.3390/ma4060963
22.
Kang
,
J. S.
,
Kim
,
H. S.
,
Ryu
,
J.
,
Hahn
,
H. T.
,
Jang
,
S.
, and
Joung
,
J. W.
,
2010
, “
Inkjet Printed Electronics Using Copper Nanoparticle Ink
,”
J. Mater. Sci.: Mater. Electron.
,
21
, pp.
1213
1220
.10.1007/s10854-009-0049-3
23.
Mahajan
,
A.
,
Frisbie
,
C. D.
, and
Francis
,
L. F.
,
2013
, “
Optimization of Aerosol Jet Printing for High-Resolution, High-Aspect Ratio Silver Lines
,”
ACS Appl. Mater. Interfaces
,
5
(
11
), pp.
4856
4864
.10.1021/am400606y
24.
Salary
,
R.
,
Lombardi
,
J. P.
,
Samie Tootooni
,
M.
,
Donovan
,
R.
,
Rao
,
P. K.
,
Borgesen
,
P.
, and
Poliks
,
M. D.
,
2017
, “
Computational Fluid Dynamics Modeling and Online Monitoring of Aerosol Jet Printing Process
,”
ASME J. Manuf. Sci. Eng.
,
139
(
2
), p.
021015
.10.1115/1.4034591
25.
Gu
,
Y.
,
Gutierrez
,
D.
,
Das
,
S.
, and
Hines
,
D. R.
,
2017
, “
Inkwells for on-Demand Deposition Rate Measurement in Aerosol-Jet Based 3D Printing
,”
J. Micromech. Microeng.
,
27
(
9
), p.
097001
.10.1088/1361-6439/aa817f
26.
Dalal
,
N.
,
Gu
,
Y.
,
Hines
,
D. R.
,
Dasgupta
,
A.
, and
Das
,
S.
,
2019
, “
Cracks in the 3D-Printed Conductive Traces of Silver Nanoparticle Ink
,”
J. Micromech. Microeng.
,
29
(
9
), p.
097001
.10.1088/1361-6439/ab2f25
27.
Serway
,
R. A.
,
1994
,
Principles of Physics
,
Saunders College Pub
,
Fort Worth, TX
.
28.
Gu
,
Y.
,
Hines
,
D. R.
,
Yun
,
V.
,
Antoniak
,
M.
, and
Das
,
S.
,
2017
, “
Aerosol-Jet Printed Fillets for Well-Formed Electrical Connections Between Different Leveled Surfaces
,”
Adv. Mater. Technol.
,
2
(
11
), p.
1700178
.10.1002/admt.201700178
29.
Merilampi
,
S.
,
Laine-Ma
,
T.
, and
Ruuskanen
,
P.
,
2009
, “
The Characterization of Electrically Conductive Silver Ink Patterns on Flexible Substrates
,”
Microelectron. Reliab.
,
49
(
7
), pp.
782
790
.10.1016/j.microrel.2009.04.004
30.
Rahman
,
M. T.
,
McCloy
,
J.
,
Ramana
,
C. V.
, and
Panat
,
R.
,
2016
, “
Structure, Electrical Characteristics, and High-Temperature Stability of Aerosol Jet Printed Silver Nanoparticle Films
,”
J. Appl. Phys.
,
120
(
7
), p.
075305
.10.1063/1.4960779
31.
Cook
,
A.
,
2013
, “
Mechanical and Electrical Fatigue of Aerosol Jet Printed Conductors
,” M.S. thesis, University of New Mexico, Albuquerque, NM.
32.
Singh
,
K. N.
, and
Tirumkudulu
,
M. S.
,
2007
, “
Cracking in Drying Colloidal Films
,”
Phys. Rev. Lett.
,
98
(
21
), p.
218302
.10.1103/PhysRevLett.98.218302