Report

Meniscus-Confined Three-Dimensional Electrodeposition for Direct Writing of Wire Bonds

See allHide authors and affiliations

Science  16 Jul 2010:
Vol. 329, Issue 5989, pp. 313-316
DOI: 10.1126/science.1190496

Abstract

Continued progress in the electronics industry depends on downsizing, to a few micrometers, the wire bonds required for wiring integrated chips into circuit boards. We developed an electrodeposition method that exploits the thermodynamic stability of a microscale or nanoscale liquid meniscus to “write” pure copper and platinum three-dimensional structures of designed shapes and sizes in an ambient air environment. We demonstrated an automated wire-bonding process that enabled wire diameters of less than 1 micrometer and bond sizes of less than 3 micrometers, with a breakdown current density of more than 1011 amperes per square meter for the wire bonds. The technology was used to fabricate high-density and high-quality interconnects, as well as complex three-dimensional microscale and even nanoscale metallic structures.

As an essential part of any integrated chip, interconnects provide the electrical paths needed in a circuit to pass signals and data among electrical devices or device units. The increasing device density in electronic chips has led to exponential growth in the density of interconnects and the complexity of their design. With the introduction of three-dimensional (3D) chip architecture, interchip vias (1) constitute one method to integrate devices in 3D stacks, but alternative interconnect technologies that can provide flexible means to electrically wire microscale device components in three dimensions are still required.

Traditional wire-bonding technology has served the electronics industry for many decades, satisfying the interconnection needs for device packaging (2). Recently, flip-chip interconnect technology was introduced as a means of increasing the interconnect density and improving device performance for high-frequency operation (3). However, downscaling this technology to interconnect pad pitches on the order of a few micrometers has proven to be difficult. Thermosonic gold wire bonding has a limiting pitch of ~40 μm, whereas flip-chip technology can achieve a pitch of ~100 μm in industrial practices. This increases the on-chip space needed for the interconnect pads, reduces the number of chips that can be produced per wafer, and consequently increases the cost per chip. It is expected that downscaling the traditional solder-based interconnect cannot meet either the thermomechanical reliability requirement or the current density requirement at very fine pitches (4).

Among the 3D microfabrication technologies that are compatible with electronic devices, e-beam– or focused ion beam–based deposition (5) and direct writing with metal colloidal ink (6) have been explored as methods for fabricating 3D interconnects with nanoscale or microscale dimensions. E-beam– or focused ion beam–based deposition is capable of fabricating 3D structures with feature sizes as small as ~10 nm. However, the process must be performed in a high-vacuum environment and the throughput is low. The choice of materials that can be deposited is limited by the scarcity of the special chemical compounds, and the deposited metals tend to have low electrical quality (5). Direct writing with metal colloidal ink has been used to fabricate interconnect bridges, ~10 μm in diameter, made of Ag with the use of an Ag colloidal dispersion (6). The limitation, however, lies in the further downsizing of the fabricated wires: The finite size of the colloidal particles, and hence the fluidics involved with a dispersion of these particles, ultimately limit the minimum nozzle size that can be used to dispense the colloidal dispersion. Further, a specific metallic colloidal material system must be developed for each type of metallic wire while taking into consideration how to convert the resulting colloidal wires into metallic ones, for instance, through thermal annealing (6).

We demonstrate an automated direct-writing wire-bonding technology that exploits meniscus-confined 3D electrodeposition and operates in an ambient air environment. We show the direct-writing fabrication of conductive interconnect bridges with submicrometer diameters and bond sizes of less than 10 μm2. The interconnects are made of pure Cu, or even noble metals such as Pt, and potentially can be made from other metals that can be electrochemically deposited with the use of readily available electrolyte solutions. These interconnects achieve a breakdown current density of >1011 A/m2, six orders of magnitude higher than that for solder-based interconnects.

The meniscus-confined 3D electrodeposition relies on an electrolyte-containing micropipette with a microscopic dispensing nozzle (several micrometers down to 100 nm in diameter) as the working toolbit (Fig. 1A). As the micropipette approaches a conductive substrate surface, a meniscus (liquid bridge) is established between the dispensing nozzle and the substrate surface. With an appropriate electrical potential applied between the electrolyte contained in the micropipette and the substrate surface, electrodeposition is initiated within the substrate surface confined by the meniscus. The key to fabricating the interconnect is to synchronize the withdrawal speed of the micropipette away from the substrate surface with the growth rate of the local deposit, which maintains the stable formation of the meniscus now established between the nozzle and the growth front of the deposited wire, thereby sustaining the continuous growth of the off-surface micro- or nanowires. The method thus represents an advance over technologies such as electrochemical dip-pen lithography, in which the deposition is limited to surface patterning (7), or electric field–enhanced electrodeposition with a sharpened metal tip, in which the local electrodeposition is realized through the localization of the electric field near the end of the tip and fabrication must be performed with both the work piece and the tip immersed in an electrolyte environment (8).

Fig. 1

(A) Schematic showing the general setup for meniscus-confined 3D electrodeposition. The long-travel piezostages (nominal resolution <10 nm) provide the fine positioning needed to control the travel path of the micropipette in the 3D space. The high-sensitivity electrometer circuit (resolution <1 pA) monitors and controls the ionic current. A magnified view at the nozzle–metal wire interface shows the formation of a meniscus (liquid bridge) serving as the stable confinement needed for the continuous electrodeposition of a uniform-diameter microscale or nanoscale metal wire as the micropipette is continuously withdrawn from the substrate surface. (B) SEM image showing the nozzle at the end of a glass micropipette with a side cut made by focused ion beam machining. (C) Electrodeposited Cu wires with different inclination angles fabricated with the use of a side-cut micropipette.

The size and shape of the meniscus formed between the nozzle and the growing metal wire is defined by the size of the nozzle, by the thermodynamic properties of the liquid solution and the involved interfaces, and by the separation between the nozzle and the growth front of the wire (thus, by the withdrawal speed of the micropipette and the growth rate of the wire). Theoretically, a meniscus as narrow as ~2 nm can be established (9), and the thermodynamic stability of such a small meniscus was recently studied both experimentally and theoretically (10, 11). Because the size of the meniscus defines the diameter of the deposited wire when a stable growth is established, this method is intrinsically capable of fabricating wires down to nanometer sizes. Practically, the mechanical stability of the physical system, the availability of nozzles of small size, the ion transport behavior through small nozzles, the interaction between the electrodeposit and substrate surface related to nucleation and growth, and the electrochemical process in a confined meniscus environment all affect the minimum feature size and the quality of nanostructures that can be fabricated with this method. We have used this method to grow straight metal wires (vertical to the substrate surface) of various diameters down to ~100 nm with the use of a micropipette with a nozzle diameter of ~100 nm; we achieved wire lengths of >80 μm, limited only by the travel range of the piezoelectric linear stage (fig. S1) (12, 13).

To facilitate interconnect bridge fabrication, which involves the lateral growth of a metallic wire over a sufficient span, we shaped the micropipette nozzle to allow the stable meniscus to form sideways to the nozzle. Figure 1B shows a shaped glass micropipette with a nozzle diameter of ~3 μm. The nozzle end and the side opening allow the formation of a stable meniscus with the wire oriented along any direction between 0° (parallel to the substrate surface) and 90° (normal to the substrate surface). The shaping of the nozzle is done with focused ion beam machining, which provides the requisite precision. Figure 1C shows a series of angled Cu wires grown with a side-cut nozzle. The vertical and lateral lengths, as well as the orientation of the Cu wires, were controlled by the travel path of the micropipette, and the diameter was determined by the nozzle size and the size of the side opening in the nozzle. The wires were fabricated from a simple 0.05 M CuSO4 aqueous solution and biased at 0.2 V with respect to the Au-coated sample surface. The micropipette had a nominal nozzle diameter of ~3 μm. The growth rate for the Cu wire at those conditions was ~0.25 μm/s, and the corresponding ionic current was maintained at ~3.5 nA. The deposition was carried out with the substrate exposed to a humidity-controlled ambient air environment at room temperature.

To form a connected wire (Fig. 2A) (14), we formed the second bond by mechanically pushing the suspended end of the laterally grown metal wire down to the substrate, with the nozzle end in close proximity so that the electrolyte meniscus under the nozzle extended to immerse both the wire end and the region of contact on the substrate surface. Several voltage pulses with amplitude at the electrodeposition potential were then applied to initiate the electrodeposition. The distance needed to push the wire end toward the surface was determined by an automated sensing procedure in which the micropipette was shifted sideways to detach from the wire end and driven with the piezoelectric stage to approach the substrate surface with the electrolyte biased at a deposition potential. At the moment of meniscus formation between the nozzle and the substrate surface, an ionic current could be detected and the distance between the wire end and the underlying substrate surface was determined. The micropipette was then withdrawn to reengage with the wire end to form the second bond as described.

Fig. 2

(A) Schematic showing the steps involved in the wire-bonding process with the meniscus-confined 3D electrodeposition. (B) SEM image showing 20 electrodeposited interconnects with submicrometer diameters fanning out from a central pad with an area of 50 μm by 50 μm. (C) SEM image showing the uniform quality of the first and the second bonds. (D) SEM image showing multilayered interconnection over three steps of 5 μm each in height. (E) SEM image showing overlap interconnects over steps of 5 μm in height. Scale bar, 10 μm.

Figure 2, B to D, shows the result of this electrodeposition-based wire bonding of submicrometer-diameter Cu wires. The wire-bonding process was performed to form 20 interconnects fanning out from a central bonding pad with an area of 50 μm by 50 μm, resembling a typical device layout (Fig. 2B). Multilayered interconnection without (Fig. 2D) or with (Fig. 2E) overlap wiring was also realized. The diameter of the Cu wires (Fig. 2, B, D, and E) was ~800 nm, and the size of the formed bonds was ~3 μm.

The bonded wires were found to be of high electrical quality. Figure 3 shows the acquired current-voltage (I-V) curve from a bonded wire with a diameter of ~740 nm and a length of ~40 μm, tested in an ambient air environment. The linear behavior in the broad current range reflected the ohmic contact of the bonds. The nonlinear behavior at high current implied the potential effect of heating (thus, oxidation of the Cu wire) (15). The overall resistance of the bonds and the wire was ~2.9 ohms when deducting the resistance contributed from the peripheral connections in the measurement, very close to the expected value for a Cu wire of this size (~1.6 ohms, assuming an electrical resistivity of 1.68 × 10−8 ohm·m of bulk Cu) with the bonds contributing a small contact resistance. At high current, the bonded Cu wire failed in the middle of the bridge and not at the bonds (inset, Fig. 3). The breakdown current density was ~1.25 × 1011 A/m2, in agreement with reported values for similar-sized Cu wires (1618). Mechanically, we measured the bonding strength with an atomic force microscope (AFM) cantilever-based pull test. By monitoring the deflection of the AFM cantilever while pulling a Cu wire vertically grown on an Au-coated substrate and with the free end glued onto the AFM tip with epoxy, we calculated that the strength of this electrodeposition-formed bond was more than ~39 MPa, well above the nominal bonding strength required in traditional wire bonding (8.5 MPa, according to the MIL-STD-883G test standard).

Fig. 3

Constant-current mode I-V characteristics of a Cu interconnect (diameter ~740 nm, length ~40 μm) showing the linear ohmic contact behavior of the bonds measured in an ambient air environment. The inset shows the failed wire at high current.

We also carried out such fabrication and characterization for Pt interconnects deposited with the use of an aqueous solution of chloroplatinic acid (H2PtCl6) as the electrolyte. Similar results were obtained (fig. S2).

The remarkable dexterity of this method for fabricating interconnect bridges relies on the thermodynamic stability of the meniscus maintained during the deposition process. In general, the stability of the meniscus is largely governed by wetting conditions involved with the nozzle and the wire growth front. To maintain the growth of a uniform-diameter wire, the thermodynamic consideration of the interfacial forces at the three-phase contact line between the meniscus and the growing wire requires the classical Neumann quadrilateral relation to be met (19), which would then require the establishment of an equilibrium angle ϕ0 between the growth direction and the slope of the meniscus at the contact line (inset, Fig. 4) ϕ0=arccos(γL2+γS2γSL22γLγS)(1)where γL and γS are the surface energies of the electrolyte and the metal wire with absorbing fluid, respectively, and γSL is the interfacial energy of the metal-liquid interface. This angle is determined to be ~12° for the copper-water-air system (taking γL = 0.07119 J/m2, γS = 0.07112 J/m2, and γSL = 0.01456 J/m2) (20). Solving the meniscus shape equations then defines a region of stability for the stable growth of a uniform-diameter wire governed by HMDW=12cos ϕ0(cosh1DNDWcos ϕ0cosh11cos ϕ0)(2)(21), where HM is the height of the meniscus, DN is the diameter of the micropipette nozzle, and DW is the wire diameter. For a micropipette with a nozzle diameter of DN, the diameter of a wire that can be stably grown lies within the range DN to ~0.5DN. Within the stable growth region, the deviation δϕ of the contact angle ϕ from the equilibrium angle ϕ0 leads to the fluctuation of the wire size according todDWdt=2(vNvW)tan δϕ(3)where vN is the withdrawal speed of the micropipette and vW is the growth rate of the wire. The growth rate of the wire is simply defined by Faraday’s lawvW=4iMnFρπDW2(4)here i is the electrodeposition current; M and ρ are the molar mass and the mass density of the deposited material, respectively; n is the number of electrons per ion; and F is the Faraday constant. Figure 4 shows the experimentally acquired parameter window for the stable wire growth in terms of the resulting wire diameter, the withdrawal speed, and the applied ionic current, in comparison to the prediction from the model. In the modeling, the diameter of the nozzle was measured from the scanning electron microscopy (SEM) image, and the surface energy and interfacial energy values for the Cu–water vapor system were obtained from the literature. The agreement between the experimental and modeling data covered a broad range of wire diameters up to near the predicted minimum diameter.

Fig. 4

Meniscus stability–defined parameter window for the stable electrodeposition of uniform-diameter metal wires. Plots show the dependence of wire diameter on the withdrawal speed of the micropipette (circles) at a fixed ionic current of ~12 nA with the use of a micropipette with a nozzle diameter of ~1.6 μm, and the dependence of wire diameter on ionic current (squares) at a fixed withdrawal speed of ~0.3 μm/s with the use of the same pipette. The solid lines represent the relevant modeling data. The inset shows a schematic of the meniscus and the angle ϕ0 between the growth direction and the slope of the meniscus at the three-phase contact line relevant to the stable electrodeposition of uniform-diameter metal wires.

As in a regular wire-bonding process, the design of small interconnect bridges requires taking certain mechanical considerations into account. One is the stress sustained by the wire during the fabrication process, which ideally should be well below the failure stress of the wire; another is the spring force sustained by the second bond, which should be lower than the bonding force. Consider a simple interconnect bridge as shown in Fig. 2A. The maximum stress occurs at the left edge of the 90° bend and can be estimated according to σc = 3EHr/L2, and the spring force sustained by the second bond according to Fs = 3πEHr4/4L3, where E is the Young’s modulus of the metal wire, H is the standoff height, and L and r are the length and radius of the lateral segment of the wire, respectively. For a typical Cu interconnect bridge of 1 μm in wire diameter, 5 μm in standoff height, and 30 μm in span, the maximum stress is calculated to be ~1 GPa, which is beyond the yield strength of Cu. Thus, the Cu wire at the bend will experience a plastic deformation in the bonding process. The problem can be readily solved (if needed) by growing a tiled wire from the first bond (Fig. 2C) instead of a vertical one to allow the formation of a smooth bend instead of a 90° bend. The maximum spring force on the second bond is ~3 μN, and if the size of the second bond is assumed to be 10 μm2, the loading stress is ~0.3 MPa, much lower than the debonding strength we measured for such an electrodeposited bond. Overall, reducing the wire diameter and increasing the lateral length of the wire can also effectively benefit the lowering of such stresses.

The growth rate of an electrodeposited wire is intrinsically limited by the rate of electroreduction, or more specifically by the diffusion-limited ionic current (22) described by a recessed microelectrode (the growth front of the deposited metal wire) in a truncated cone (the tapered dispensing end of the micropipette). Nonetheless, the growth rate might be increased through an increase in electrolyte concentration and/or the use of a short tapered micropipette. Alternatively, an array of micropipettes can be deployed to increase the wire-bonding throughput.

With the proper mechanical design and system control, this meniscus-confined 3D electrodeposition method can be used to fabricate more intricate microscale and nanoscale structures than those described here. Such structures could include designed structural and device functionalities integrating a variety of metallic materials, such as magnetic and noble metals and even metal alloys. Moreover, because it is intrinsically a low-cost direct-writing technology, this technique can be used to fabricate such micro- or nanostructures on existing micro- or nanostructures when such fabrication becomes difficult or expensive with the traditional lithography process (fig. S3).

Supporting Online Material

www.sciencemag.org/cgi/content/full/329/5989/313/DC1

Materials and Methods

Figs. S1 to S3

References and Notes

  1. See supporting material on Science Online.
  2. Supported by the Grainger Foundation. We acknowledge the use of microscopy facilities in the Center for Microanalysis of Materials at the University of Illinois at Urbana-Champaign. A U.S. patent application based on this work was filed by the University of Illinois on 7 June 2010.
View Abstract

Navigate This Article