GB2530974A - Method of making a Graphene-Cu-Graphene Heterogeneous Film - Google Patents

Abstract

A method of manufacturing a Graphene-Cu-Graphene heterogeneous film contains the steps of: (A) providing 25 µm thick copper of 99.9% purity or 9 µm thick copper of 99.99% purity and above; (B) synthesizing graphene in a low-pressure CVD system; (C) heating a copper substrate up to 1030°C under hydrogen and introducing methane to grow the graphene; (D) synthesizing a sample with single-layer graphene (SLG) or few-layer graphene (FLG) by controlling the cooling rate, wherein the copper substrate is cooled from 1030°C to room temperature (RT) within 20 minutes, and a cooling time of the FLG is 10 hours; and (E) annealing copper in the sample with the same heating and cooling process as that of SLG synthesis, but without methane addition during this step. The annealing time of the copper of the sample is preferably kept at 20 minutes. The graphene may be synthesized by desorption of Si from SiC single-crystal surfaces or by surface precipitation process of carbon in transition metals. A Graphene-Cu-Graphene heterogeneous film formed by this method is also disclosed.

Description

METHOD OF MAKING GRAPHENE-CU-GRAPIIENE HETEROGENEOUS

FILM

FIELD OF THE INVENTION

The present invention relates to a method of making a graphene-Cu-graphene heterogeneous film which greatly enhanced thermal conductivity.

BACKGROUND OF THE INVENTION

Graphene is a one-atom-thick material that is unusual and highly promising for applications of electrical, thermal and mechanical properties. First obtained by mechanical exfoliation from graphite, graphene is now efficiently grown by chemical vapor deposition (CVD) on copper (Cu) films. It was reported that layered graphene metal composites have enhanced mechanical strength. However, it is still not known how the deposition of graphene on Cu films affects the thermal properties of the resulting graphene-Cu films. The knowledge of thermal properties of graphene-Cu "sandwiches" is important for the following practical reasons. Copper became the crucial material for interconnects in silicon (Si) complementary metal-oxide-semiconductor (CMOS) technology by replacing Al. Main challenges with continuous downscaling of Si CMOS technology include electromigration in Cu interconnects, Cu diffusion to adjacent layers, and heat dissipation in the interconnect hierarchies separated from a heat sink by many layers of dielectrics. Combining graphene and Cu in some sort of hybrid heterogeneous global interconnect can bring potential benefits of reducing Cu electromigration and diffusion, it has already been demonstrated that the breakdown current density in prototype graphene interconnects can exceed that in metals by x 103.12 Graphene capping of Cu interconnects increases the current density and reduces electrical resistance. Intersecting hybrid graphene-Cu interconnects have been shown to offer benefits for downscaled electronics. Increasing the heat conduction properties of Cu films with graphene coating could become a crucial added benefit for improving the thermal management of the interconnect hierarchies.

Graphene is known to have usually high intrinsic thermal conductivity, which can exceed that of bulk graphite limit of K 2000 W/mK at RT in sufficiently large high-quality samples. However, graphene placement on substrates results in degradation of thermal conductivity to -600 W/mK owing to phonon scattering on the substrate defects and interface. The benefits of using single-layer graphene (SLG) or few-layer graphene (FLG) as heat spreaders for large substrates are not obvious owing to the small thickness of graphene (h = 0.35 nm) and possible thermal conductivity degradation by extrinsic effects. Even if K is high, the uniform heat flux, W = K x A, through the cross-sectional area A = hW will be small due to small h (W is the width of the graphene layer).

An alternative strategy directed at alleviating the interlayer interactions of graphene involves engineering the morphologies of the sheets to form structures that are resistant to the negative effects of aggregation, although this methodology is not admitted as prior art by its inclusion in this Background Section. For example, crumpled graphene balls stabilized by locally folded ridges have been synthesized via evaporating aerosol droplets of graphene oxide (GO). In so doing, the GO sheets were dispersed in water or organic solvents and then rapidly dried, which caused the sheets to deform into highly wrinkled structures as a result of evaporation-induced capillary flow. To restore the conductivity of the structures, the GO was then thenflally reduced back to graphene. Unfortunately, the reduction of GO into graphene is almost always incomplete and results in a high degree of structural disorder. Thus, here again, the ultimate product is likely to be compromised.

SUMMARY OF THE INVENTION

The primary object of the present invention is to provide a graphene-Cu-graphene heterogeneous film which enhances thermal conductivity greatly.

Another object of the present invention is to provide a graphene-Cu-graphene heterogeneous film which is important for thermal management of advanced electronic chips and proposed applications of graphene in the hybrid graphene-Cu interconnect hierarchies.

A graphene-Cu-graphene heterogeneous film provided by the present invention contains steps of: A. providing 99.9% purity of 25 pm thick copper and above 99.99% purity of 9 pin thick copper; B. synthesizing graphene in a low-pressure CVD system; C. heating a copper substrate up to 1030 °C under hydrogen and introducing methane to grow the graphene; D. synthesizing a sample with single-layer graphene (SLO) and few-layer graphene (FLU) by controlling a cooling rate, wherein copper substrate is cooled from 1030 °C to a room temperature (RT) within 20 mi and a cooling time of the FLU is hours; and E. annealing copper of the sample with the same heating and cooling process as that of SLG synthesis, wherein no methane addition during step of E. Preferably, a time of annealing the copper of the sample is kept at 20 min. Preferably, the graphene is synthesized by a desorption of Si from SiC single-crystal surfaces and by a surface precipitation process of carbon in transition metals.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other features, aspects, and advantages of the present invention will become better understood with regard to the following description, appended claims, and accompanying drawings where: FIG. 1 is a flow chart of a method of making a graphene-Cu-graphene heterogeneous film according a preferred embodiment of the present invention.

FIG. 2-A is a schematic view of the modified "laser flash" experimental means for measuring hi-plane thermal diffhsivity.

FIG. 2-B is a schematic view showing Cu film coated with CVD graphene placed on the sample holder.

FIG. 2-C is a schematic view showing back side of the sample holder with the slits for measuring temperature, wherein Cu film is seen through the openings.

FIG. 2-D is a schematic view showing Raman spectrum of graphene and few-layer graphene on Cu.

FIG. 3 is a schematic view showing thermal diffusivity and thermal conductivity of graphene-coated copper films, wherein thermal diffusivity of reference Cu film, annealed Cu, Cu with CVD graphene, and Cu with CVD FLG (top panels); thermal conductivity of reference Cu film, annealed Cu, Cu with CYD graphene, and Cu with CYD FLG (bottom panels); the data are shown for Cu films with H = 9 jim and H = 25 jim, wherein CYD of graphene and FLG results in stronger increase in the apparent thermal conductivity of graphene-Cu-graphene samples than annealing of Cu under the same conditions.

FIG. 4-A is a schematic view showing Cu film with CYD graphene.

FIG. 4-B is a schematic view showing Cu film with C\7D graphene (f), wherein deposition of graphene substantially increases the Cu grain size.

Table 1 shows Thermal Difflisivity and Thermal Conductivity of Graphene Coated Cu Films.

DETAILED DESCRIPTION OF THE INVENTION

The present invention will be described with reference to illustrative embodiments. For this reason, numerous modifications can be made to these embodiments and the results will still come within the scope of the invention. No limitations with respect to the specific embodiments described herein are intended or should be inferred.

A method of manufacturing a graphene-Cu-graphene heterogeneous Ii liii according to a preferred embodiment of the present invention comprises steps of: A. providing 99.9% purity of 25 im thick copper and above 99.99% purity of 9 jim thick copper; B. synthesizing graphene in a low-pressure CVD system; C. heating a copper substrate up to 1030 °C under hydrogen and introducing methane to grow the graphene; D. synthesizing a sample with single-layer graphene (SLO) and few-layer graphene (FLG) by controlling the cooling rate, wherein copper substrate is cooled from 1030 °C to room temperature (RT) within 20 mm, and the cooling time of the FLU is 10 hours; and E. annealing copper of the sample with the same heating and cooling process as that of SLG synthesis, with no methane addition during step F. Preferably, the time of annealing the copper of the sample is kept at 20 mm; the graphene is synthesized by desorption of Si from SiC single-crystal surfaces and by a surface precipitation process of carbon in transition metals.

In following descriptions, we show the results of our thermal measurements that demonstrate that CVD of graphene on both sides of Cu films enhances the thermal diffrtsivity, a, and thermal conductivity, K, of the resulting graphene-Cu-graphene (Gr-Cu-Gr) hetero films. Deposition of graphene increases K of 9-jim (25-jim) thick Cu films by up to 24% (16%) near room temperature (RT). Interestingly, the enhancement of thermal properties of Cr-Cu-Or hetero films is primarily due to changes in Cu morphology during graphene deposition rather than graphene's action as an additional heat conducting channel. Specifically, CYD of graphene results in strong enlargement of Cu grain sizes and reduced surface roughness. A typical grain size in Cu films coated with graphene is larger than that in reference Cu films and in Cu films annealed under the same conditions without graphene deposition.

To demonstrate the effect we used a set of Cu films 85 (thickness H = 9 jtm and H = 25 jim) with SLG and FLG synthesized on both sides via CVD method (Bluestone Global Tech, Ltd.). As references we used (i) Cu films without graphene or any thermal treatment, and (ii) Cu films annealed under the same conditions as the one used during CVD of graphene. Thus, for comparison we had regular Cu, annealed Cu, Cu with CVD SLG, and Cu with CVD FLG. Details of sample preparation are provided in Methods.

The reference Cu and Cu-graphene samples were subjected to optical microscopy, scanning electron microscopy (SEM), and atomic force microscopy (AFM) inspection.

The number of atomic planes in graphene films on Cu was verified with micro-Raman spectroscopy (Renishaw Tn Via). Details of our Raman measurement procedures have been reported by some of us elsewhere (i.e., Calizo, I.; et al. Ultraviolet Raman microscopy of single and 473 multilayer graphene. J. AppI. Phys. 2009, 106, 43509.).

The measurements of the thermal dififisivily were carried out using the "laser flash" method (Netzsch LFA). in conventional configuration, the "laser flash" method gives the cross-plane thermal diffusivity, a, of the sample. Since we are mostly interested in the in-plane heat spreading properties of Gr-Cu-Grr hetero films, we altered the experiment by using a special sample holder, which send the thermal energy along the sample. In this approach, the location for the light energy input on one side of the sample and location for measuring the 108 temperature increase on the other side of the sample are at different lateral positions. The latter insures that the measured temperature increase of the sample corresponds to the thennal diffusivily in the in-plane direction. The thermal conductivity was determined from the equation K = paCp, where p is the 113 mass density of the sample and Cp is the specific heat of the sample measured separately. Details of the measurements are summarized in Methods. FIGS. 2-A to 2-D presents a schematic of the experiment, an image of a typical sample with the sample holder, and Raman spectra from two different Cu substrates indicating that one has SLG coating while the other has FLG coating. The average thickness of FLG was five atomic planes.

FIG. 3 shows the average apparent thermal diffusivity and thermal conductivity in reference Cu films, annealed Cu films, Cu films with C\D graphene and Cu films with C\TD FLU. The data are presented for two thicknesses of Cu films: H = 25jnn and H = 9 pin. The term apparent (another common term is effective) emphasizes that a and IC values are measured for the whole graphene-Cu-graphene sample. The averaging for each type of sample (e.g., Cu fihn with SLU) was performed for five locations on each film at each temperature. Two films with the same type of samples were tested. in order to simplify the analyses, in Table I we provided the average RT values of a and K measured for different samples and locations. The ranges for a and K values for different locations and samples are given in the parentheses. The data scatter for different locations was attributed to the sample non-uniformity and film bending, which were unavoidable for large foils (centimeter scale lateral dimensions) with small thicknesses.

The obtained a and K of Gr-Cu-Gr hetero films and their weak temperature dependence are consistent with literature values for bulk Cu, which varies from 385 WhnIC to 400 W/mK. As reported in refs 20 and 21 in the relevant temperature range of 300-400 K, the thermal conductivity of copper slightly decreases (increases) in bulk (thin films with H 40-200 nm range) with temperature. The latter is explained by the interplay of the intrinsic and boundary scattering mechanisms for the heat carriers. In tenns of their thickness, our samples fall in between these two limiting cases. The latter explains the observed weak and sometimes nonmonotonic dependence of the thermal conductivity on temperature.

Electrons are the main heat carriers in Cu while phonons make the dominant contribution in graphene. The strong reduction of K of Cu due to electron scattering from the film top and bottom boundaries is only expected in very thin films where the electron mean-free path (MFP) becomes comparable with H.21 However, it is known that the grain size in Cu decreases with the decreasing film thickness. For this reason, the size effects can reveal themselves even in relatively thick Cu films with H 10 jim.

The lower a and K for 9 jim films than those for 25 jim films measured in our experiments are likely related to the grain size effects. The rolling fabrication of Cu films of different thickness (9 jim vs 25 pm) is also expected to result in variations in the defect densities, grain elongation and orientation, thus, affecting a and K. The most important and unexpected observation from FIG. 3 is that a and K are strongly increased in Or-Cu-Or hetero fihns with graphene or FLO coating compared to reference Cu films or annealed Cu films. Deposition of graphene results in stronger increase of a and K than annealing under the same conditions. In terms of thermal conductivity, the effect of graphene deposition is particularly pronounced for thinner Cu films (H = 9 jim). The deposition of SLG on 9 jim Cu film results in about ---22% enhancement of the apparent thermal conductivity as compared to -12% increase in the annealed samples without graphene. The average enhancement of K and a after deposition of SLO on 25 jim films is less pronounced than that for 9 jim films but still notably larger than for the annealed reference samples. The increase in a and K is not proportional because the thermal treatment during C\TD or annealing affects the specific heat as well. It is known that thermal treatment of metals and alloys can noticeably change Cp, particularly in the presence of impurities and defects.

The overall enhancement of heat conduction properties of Or-Cu-Or hetero films as compared to reference Cu films is very strong and may appear puzzling. The thickness ofgraphene h = 0.35 nm is negligibly small compared to H = 25 jim. For this reason, the thermal resistance RU = L!(TChW) of the additional heat conduction channel via graphene will be much larger than via Cu film (here L is the length of the path).

Thus, the high thermal conductivity of graphene should not play a significant role in heat spreading ability of Cu foils over large distances (L -5 mm) if one considers conventional heat transfer by phonons. The observed enhancement of the apparent a and K can be understood if the thermal data is correlated with the microscopy data presented in FIGS. 4-A and 4-B.

One can see that CYD of graphene results in substantially stronger enlargement of Cu grains than annealing under the same conditions. The graphene CVD and annealing temperature 1030 °C is sufficiently larger than Cu recrystallization temperature of 227 °C.23 As a result, annealing accompanied by recrystallization increases the grain sizes in Cu fihns, reduces the defect density and improves their mechanical properties. Our results indicate that CYD of graphene enhances the Cu grain growth, as compared to regular annealing, by changing the thermal balance during the deposition. Graphene also stops copper evaporation from the surface when the sample is heated during CVD. These conclusions are supported by earlier observations that the substrates and underlays affect the annealing process of Cu and the resulting Cu morphology. It is also in agreement with the grain size data in Cu with CVD graphene and annealed Cu presented in ref Bae, S.; et al. Roll-to-roll production of 30-in.

graphene films for transparent electrodes. Nat. Nanotechnol. 2010, 5, 574-578.

Additionally, our SEM studies indicate that CYD of graphene results in -20% reduction in surface roughness as compared to reference Cu.

In order to further rationalize the experimental results we estimated the ratio of the average grain sizes, O/D, which would provide the relative change in the thermal conductivity, AK/IC, close to the one observed in the experiments ( is typical grain size in reference Cu film and D is the grain size after CYD of graphene). The electron MFP for thermal transport is A = 40 nm at RT. Since A << H, it is reasonable to assume that K is mostly limited by the grain boundary scattering. In this case, one can express the thermal conductivity, K, of a polycrystalline metal through that of a single-crystal bulk metal, KB, as K = (1 + A/D)-IKB. Applying this equation to polyciystalline Cu before and after CVD of graphene we derived the following relation:

I --1 +

It is well-known that Cu films have very large distribution of grain sizes. It is common to have grain sizes within a given Cu sample varied by 3 orders of magnitude from tens of nanometers to tens of micrometers. The shape of the grains in the Cu film can also be very anisotropic. Detail investigation of the grain size distribution requires expensive ion-milling and transmission electron microscopy study. For these reasons, here we provide simple estimates from the optical, SEM and AFM studies of our samples. If one assumes that the average grain diameters in Or-Cu-Or hetero structures are in the range D 1-10 jim, the experimentally measured AK/K =0.2 can be achieved for b/D ranging from -0.13 to 0.0 16, which corresponds to the grains in reference Cu on the order of 130-160 nm. Note that the smaller grains can affect the thermal transport the most by limiting the heat carrier MFP. The considered range and change in the diameter by x 10 to x 100 after CVD is consistent with the microscopy data (see examples in FIGS. 4-A and 4-B and Supporting Information). It is known that annealing of Cu under different conditions can change the grain size by many orders of magnitude from ---30 to 100 mill. Deposition of graphene can produce even stronger effect. Thus, our analysis suggests that the grain size increase can result in the observed enhancement of the thermal conductivity. \Tariations in the defect densities, for example, dislocation lines and grain boundary thickness after CVD of graphene may also affect the AK/K.

In order to exclude the possibility that the change in thermal conductivity is due to the changes in the impurity content in Gr-Cu-Gr hetero films and reference annealed Cu films, we performed X-ray photoelectron spectroscopy (XPS) and energy-dispersive X-ray (EDX) spectroscopy inspection. It was established that the impurity composition (that included 0 and N) did not differ in graphene-Cu-graphene films with graphene and arniealed Cu (Supporting Information). The differences in the grain size and roughness of GCG-HF and those of annealed Cu films may also be related to differences in the oxidation process during and after CVD of graphene and annealing. One should also note that graphene is an essential for improved thermal conductivity of Gr-Cu-Gr hetero films. The experiments with deposition of amorphous carbon on Cu indicated that the thennal conductivity has not increased but rather decreased. Amorphous carbon is known to have very low thermal conductivity of below 1 W/mK at RI.

We have also conducted four-probe electrical measurements in order to investigate if the observed change in thermal conductivity in Gr-Cu-Gr hetero films follows the Wiedemann-Franz 1aw32 K/a = LI, where a is the electrical conductivity and I. = (it2/3)(kB/q)2 2.44 x 10-8 WOK-2 is the Lorenz number. The electrical conductivity of the samples was in line with the tabulated values for Cu films. However, it did not scale up linearly with the measured K as required by the Wiedemann-Franz law. We explain it by the fact that our samples are heterogeneous, and the electric probes pressed against Gr-Cu-Gr hetero films contact both graphene or FLG layer and Cu. The electrical conductance is provided by both graphene and Cu channels. As a result, the evolution of electrical conductivity with the change in the grain size does not necessarily correlate well with the apparent thermal conductivity via the Wiedemann-Franz law.

Although it is clear that the observed strong enhancement of thermal properties of Cu films afler CVD of graphene is mostly related to the effect produced by graphene on Cu grains one cannot completely exclude other possible mechanisms of heat conduction, which might be facilitated by graphene. It has been recently suggested theoretically that plasmons and plasmon-polaritons can strongly enhance the heat transfer in graphene and graphene-covered substrates. In our measurements, the fact that the samples are heated by the light flash with the wide spectrum leaves this possibility open. The plasmon contribution would come in addition to the phonon heat conduction in graphene.

Our present findings add validity to the proposals of the graphene-capped Cu interconnects by demonstrating improvement in their heat spreading ability. Taking into account that the next technology nodes will require Cu interconnects with the nanometers-range thickness one can expect that the effects will be even more pronounced than in the examined micrometers-range thickness fihrts. The latter may become a crucial consideration for electronic industry. In addition, our results can be possibly applied in metallurgy. Carbon additives have long been used in steel smelting as alloying elements distributed though the volume. Carbon alloying allows one to vary the hardness and strength of the metal. Our results show that CYD of one atom thick graphene layer on the surface of metal foils can have a pronounced effect on its thermal properties. This is a conceptually different approach for the carbon use in metallurgy.

The measurement details are described thereafter, the "laser flash" technique (LFT) is a transient method that directly measures a. The specific heat, Cp, is measured independently with the same instrument using Cu reference. To perform LFT measurement, each sample was placed into a special stage and sample holder (see FIGS. 2-A to 2-D) that fitted its size. The bottom of the stage was illuminated by a flash of a xenon lamp (wavelength X = 150-2000 nm) with the energy pulse of 1 J for 0.3 ms. The temperature of the opposite surface of the sample was monitored with a cryogenically cooled InSb JR detector. The design of the "in-plane" sample holder ensured that heat traveled -5 mm inside Cu film along its plane, which is a much larger distance than its jim thickness, and thus ensuring the in-plane values for ci and K. The temperature rise as a function of time, AT(t), was used to extract a. The specific heat, Cp, was measured with LFT by comparing ATU) of the sample to that of a reference sample wider the same experimental conditions (Cp of the reference Cu was --0.39 Jig x IC at RT). Annealing or CVD of SLG increased Cp. The increase of specific heat with CVD of graphene or FLU was attributed to morphological changes induced by high temperature during the CVD and the fact that specific heat of graphite, Cp = 0.71 Jig x IC, is larger than that of Cu. The accuracy of LFT measurement with Netzsch instruments is 3%. The thermal conductivity was determined from the equation K = pczCp, where p is the mass density of the sample.

Also, theoretical analysis details are described hereafter. We start with the equation for the thennal conductivity of polycrystalline material limited by the grain boundaries = (1 + Here D is the grain size (mean diameter), KB is the thermal conductivity of bulk single-crystal material, and A is the electron mean free path (MFP) for thermnal transport, which can be larger than that for electrical transport. Let us assume that the material with grain size Dl has the thermal conductivity Ki while the material with grain size D2 has the thennal conductivity K2. We introduce two ratios, ç (AK/K) = (1(2 -K1)/K2 = 1 -1(1/1(2 and a = D1/D2. Writing eq 2 for two materials with two grain sizes Dl and D2, we get fl1.) K -1 ± Di Dividing eq 3 by eq 4, we can obtain for the thermal conductivity enhancement factor a Da+A\Dn) aD7+A Solving eq 5 for a we get i-c a= Finally, we obtain the relation between the ratio of the grain sizes and increase in the thermal conductivity /LK) The derived equation allows one to correlate the effect of increasing grain size in Cu films after graphene deposition with the measured increase in the thermal conductivity. Although the proposed model is simple,. it captures the main trend observed experimentally. More accurate treatment requires inclusions of specifics of electron reflections from grain boundaries and external surfaces in polycrystalline films.

In conclusion, we demonstrated experimentally that Graphene-Cu-graphene heterogeneous films reveal strongly enhanced thermal conductivity as compared to the reference Cu and annealed Cu films. Chemical vapor deposition of graphene on both sides of 9 jim thick Cu films increases their thermal conductivity by up to 24% near room temperature. The effect of graphene is projected to be substantially stronger in nanometer thick Cu interconnects. The observed improvement of thermal properties of graphene-Cu-graphene hetero films results primarily from the changes in Cu morphology during graphene deposition. Enhancement of thermal properties of graphene-capped Cu films is important for thermal management of advanced electronic chips and adds validity to the proposed applications of graphene in the hybrid graphene-Cu interconnects. Our results indicating that deposition of just one atomic plane of graphene on a surface can substantially improve the properties of underlying metal film may lead to a transformative change for the use of carbon in metallurgy.

The above-described embodiments of the invention are intended to be illustrative only. Other embodiments can use different processing steps, and different types and arrangements of elements to implement the described functionality. These numerous alternative embodiments within the scope of the appended claims will be apparent to one skilled in the art.

Moreover, all the features disclosed herein may be replaced by alternative features serving the same, equivalent, or similar purposes, unless expressly stated otherwise. Thus, unless expressly stated otherwise, each feature disclosed is one example only of a generic series of equivalent or similar features