Page 42 - Chip Scale Review_November December_2021-digital

Figure 5: a) (left) TEM image of Ru-capped plasma-cleaned few-layer graphene; b) (right) Transfer characteristics curves of bilayer (BLG) devices showing the change in
the on-current and shift in charge-neutrality point (CNP) for as-transferred and plasma-treated graphene after the “graphene plasma clean” step. The solid and dashed
lines represent the upper and lower bounds of the transfer curves, respectively, obtained from 63 devices.
dissipation. This observation is of interest Engineering the graphene/metal interface Towards industrial adoption
for future interconnect applications, as is, however, one of the most challenging T hese result s demonst rate t he
the self-heating in highly-scaled IC wires bottlenecks. The (as-transferred) graphene per for mance potential of hybr id
and an insufficient heat dissipation to the layer typically contains many randomly- metal/graphene schemes in advanced
surrounding dielectric can degrade the oriented grains where the grain boundaries interconnects. Yet, several integration
interconnect’s thermal reliability. act as line defects and nucleation centers challenges remain to be solved before
Overall, the researchers conclude that for metal deposition on top surface. This these interconnect schemes can be
these graphene-capped metal/hybrid makes it challenging for depositing a adopted in a 300mm fab. For example,
structures provide an answer to the RC metal uniformly covering the entire basal while this study focuses on graphene
delay problem for future interconnects. plane of graphene by means of traditional t r a n sfe r, a more elega nt way of
Imec envisions their introduction in the deposition method such as PVD or atomic depositing graphene would be direct
BEOL technology roadmap for the 1nm layer deposition (ALD). Moreover, after growth on the metal template of interest.
node and beyond. Yet, more fundamental transfer, the graphene surface suffers from Growing high-quality graphene requires,
insights are needed to determine the contamination—calling for a suitable however, high-growth temperatures
exact conduction mechanism taking cleaning method that does not damage the (900-1000°C) and, as such, cannot be
place within the capped structure. Either graphene layer. applied on interconnect-type of metals.
Ru remains the main conductor, with In a laboratory study, the imec Deposition at lower temperatures has
graphene helping to reduce its resistivity researchers performed a hydrogen been demonstrated, but comes at the
by suppressing scattering mechanism(s) plasma cleaning of the graphene expense of defectivity and reduced
in the metal. Or, the two conductors now surface (by using an Ar/H 2 downstream quality of graphene.
act in parallel, with a higher conductivity plasma), and subsequently deposited An alternative route that was applied in
for graphene (with respect to intrinsic the metal (i.e., Ru) by using electron this study includes the transfer of high-
graphene) because of the charge transfer. beam evaporation (Figure 5a). It was quality graphene that was previously
Modeling work is currently ongoing to then investigated how these processes grown on platinum foils by using CVD.
get a better understanding. affected the electrical conductivity of This transfer route provides an interesting
Ruthenium-capped graphene. In the graphene/Ru stack. They found that approach when the thermal budget is
the longer term, researchers at imec are after exposure to hydrogen plasma, restricted. At imec, delamination and
looking into stacking alternating layers graphene experiences n-doping and a subsequent transfer of high-quality
of graphene and metal to further boost rise in charge carrier concentration. graphene on 300mm wafers has been
the electrical conductivity. In such a Unfortunately, single-layer graphene demonstrated, but might be challenged by
metal/graphene/metal/etc. sandwich- also suffers from plasma-induced the topography of the underlying metal
like structure, a second and different defectivity. Thicker graphene films are layer. This process also comes with a
interface will now play an equally observed to be less affected. Under these significant addition of process steps and
important role: the interface that results conditions, an overall improvement of calls for improved uniformity and process
from depositing a layer of metal on 18% in electrical conductivity of Ru- control. In addition, further research
top of graphene. Just like in the above capped (plasma treated) graphene devices will be needed to optimally control the
study, the nature of the graphene/metal could be observed (Figure 5b). These defectivity and specific orientation of
interaction at the interface can modify first results are encouraging, and further the graphene layers. Studies at imec are
the physical properties of graphene. And improvements can be expected by tuning ongoing to solve these integration issues
its electronic band structure can also the hydrogen plasma chemistry and and to turn the hybrid graphene/metal
be significantly altered by the charge conditions, and by increasing the number schemes into true industry-grade options.
distribution at the interface. of alternating layers [5].

40 Chip Scale Review November • December • 2021 [ChipScaleReview.com]
40