Specific Process Knowledge/Thin film deposition/Deposition of Gold/Adhesion layers: Difference between revisions
Appearance
mNo edit summary |
|||
| (20 intermediate revisions by 2 users not shown) | |||
| Line 1: | Line 1: | ||
'''Feedback to this page''': '''[mailto:labadviser@danchip.dtu.dk?Subject=Feed%20back%20from%20page%20http://labadviser.danchip.dtu.dk/index.php/Specific_Process_Knowledge/Thin_film_deposition/Deposition_of_Gold/Adhesion_layers click here]''' | |||
'''This work was performed by PhD student Matteo Todeschini during period 2014-2017''' | |||
= Adhesion layers = | = Adhesion layers = | ||
The deposition of metal thin-film structures on dielectric or semiconductor substrates is important in a wide range of micro/nanofabrication applications (plasmonics, metamaterials, organic transistors, substrates for graphene growth, field effect devices). | The deposition of metal thin-film structures on dielectric or semiconductor substrates is important in a wide range of micro/nanofabrication applications (plasmonics, metamaterials, organic transistors, substrates for graphene growth, field effect devices). | ||
| Line 8: | Line 12: | ||
Fig. 1 represents the simplified adhesion layer model usually used in the micro and nanofabrication community for a generic bilayer system. In this model, the adhesion layer and the noble metal over-layer are deposited sequentially by PVD and are considered to be completely uniform and in perfect contact. Ti and Cr are known to be more chemically reactive than noble metals and thereby increase the adhesion as they chemically bind to the substrate. However, the model does not include any layer-layer interaction. Moreover, it is considered to be independent of the thin-film thickness. Due to the continuous miniaturization of structures and devices, the total metal thin-film thicknesses have reached tens of nanometers, with the thickness of the over-layer approaching the one of the adhesion layer. In these conditions it is unclear if the simplified model is still adequate. | Fig. 1 represents the simplified adhesion layer model usually used in the micro and nanofabrication community for a generic bilayer system. In this model, the adhesion layer and the noble metal over-layer are deposited sequentially by PVD and are considered to be completely uniform and in perfect contact. Ti and Cr are known to be more chemically reactive than noble metals and thereby increase the adhesion as they chemically bind to the substrate. However, the model does not include any layer-layer interaction. Moreover, it is considered to be independent of the thin-film thickness. Due to the continuous miniaturization of structures and devices, the total metal thin-film thicknesses have reached tens of nanometers, with the thickness of the over-layer approaching the one of the adhesion layer. In these conditions it is unclear if the simplified model is still adequate. | ||
Here the characterization work carried out on two main thin-film systems, Cr/Au and Ti/Au stacks, is presented. | [[File:Picture00.png|650px|center|thumb|Fig. 1: Schematic representation of a substrate, adhesion layer and overlayer system. The substrate is either a dielectric or semiconductor material and the over-layer is a metal.]] | ||
Here the first recommendations on the adhesion layers best suited for specific nanodevice applications are presented. Below the background and characterization work carried out on two main thin-film systems, Cr/Au and Ti/Au stacks, to arrive at these recommendations is presented. | |||
= Recommendations for nanodevice fabrication = | |||
== Nano-optic devices == | |||
For nano-optic devices there are two implications: | |||
1) the grain size values of the Au thin-films are comparable with the sizes of plasmonic nanostructures, which are 70-200 nm. Hence, each nanostructure might have multi-grain structures that are different. This might affect its interaction with the light and its optical response, which leads to broadening of the optical resonance peaks. | |||
2) during light and plasmonics device interactions, the temperature can locally raise to above 795 K, further enhancing the alloying of Cr and Au already present at room temperature and thus deteriorating the electrical properties of the thin-film stack. | |||
Ti does not inter-diffuse with Au at room temperature, supported by the ''µ''4PP measurements which show that the electrical properties appear to be similar to pure Au thin-films. The formation of a stable Ti layer under the Au prevents diffusion of Au into the underlying substrate and improves the performance-time and temperature stability of the devices. The Ti-Au inter-diffusion starts at temperatures higher than 175°C. If the devices have to be used for prolonged time above this temperature, Pd and Pt diffusion barriers must be used. Comparative studies show a localized surface plasmon damping in plasmonic nanostructures due to the presence of a very thin adhesion layer, and overall better performances of Ti over Cr for the optical resonance of such nanostructures. | |||
Therefore, if the choice is limited to metallic adhesion layers, '''Ti is preferred over Cr for nano-optics applications'''. A perhaps even better alternative is to use organosilane-based adhesion layers. Comparative measurements of Ti vs Mercaptopropyltrimethoxysilane (MPTMS) and Cr vs Aminopropyltrimethoxysilane (APTMS) show overall better performances for the organosilane-based adhesion layers over the metallic ones. A considerable disadvantage of these molecular adhesion layers is their lack of compatibility with the lithographic and lift-off processes, which still play an important role in the fabrication of nanostructures. | |||
== Nano-electronic devices == | |||
For nano-electronics applications, the reported results show that Ti and Cr form oxides with surface adsorbed water and free oxygen in the vacuum chamber of the physical vapor deposition system. To avoid oxidation of the adhesion layer, the chamber vacuum needs to be in UHV conditions and the sample must be baked to remove the surface adsorbed water. However, these baking temperatures are not compatible with e.g. lift-off nanofabrication of nano-electronic devices that often utilize photo- or electron beam lithography resists which do not tolerate high baking temperatures. | |||
To avoid oxidation of Ti or Cr that is in physical contact with the nano-electronic materials, one solution is to avoid these materials completely. Indeed, some of the best performing CNT devices are made without the use of adhesion layers, as e.g. Pd which is directly used. '''If an adhesion layer is required for mechanical stability, a less than 2nm thin Cr layer is recommended and Ti must be avoided'''. This is because the partially oxidized Ti might form a barrier between the nano-electronic material and the Au over-layer, with a consequent deterioration of the electron transport performances. Because of the single-layer morphology due to the Cr-Au alloy formation, the alloy will make electrical and physical contact to the nano-electronic material, despite the chrome oxide content. Furthermore, a low temperature annealing will enhance inter-diffusion of Au and Cr and improve electrical contact between the nano-electronic material and the Au over-layer. | |||
= | = Characterization of adhesion layer effect on Au thin films = | ||
== Effect on Au film morphology, grain size and texture == | == Effect on Au film morphology, grain size and texture == | ||
| Line 27: | Line 51: | ||
image:Picture7.png|Fig. 4: TEM cross section images and 300x300 nm AFM images of the 2nm Cr/2nm Au sample (a-b) and of the 2nm Cr/20nm Au sample (c-d). </gallery> | image:Picture7.png|Fig. 4: TEM cross section images and 300x300 nm AFM images of the 2nm Cr/2nm Au sample (a-b) and of the 2nm Cr/20nm Au sample (c-d). </gallery> | ||
To investigate the crystal orientation of the metal thin-films, tramsimission Kikuchi diffraction was used (Fig. 5a) (see [[LabAdviser/ | To investigate the crystal orientation of the metal thin-films, tramsimission Kikuchi diffraction was used (Fig. 5a) (see [[LabAdviser/314/Microscopy 314-307/SEM/Nova/Transmission Kikuchi diffraction|Transmission Kikuchi diffraction]] for more information). The nanostructure of the 20nm Au film has a bimodal grain size distribution (Fig. 5b). While the smaller grains have different crystal orientations, the large grains (blue color) all have [111] orientation. | ||
Microstructural evolution and growth of metal thin-films deposited by physical vapor deposition on amorphous dielectric substrates follows island growth. The first thin-film growth step is the nucleation of small islands once the activation barrier and the critical nuclei size have been overcome. It is followed by a second step of island growth, during which the impinging atoms contribute to increase island size. The third step, usually happening simultaneously with step 2, is island coalescence, where a strong driving force is present for coarsening through surface atom diffusion and grain boundaries (GB) motion. During this process, the island growth is driven by the minimization of surface and interface energy. | Microstructural evolution and growth of metal thin-films deposited by physical vapor deposition on amorphous dielectric substrates follows island growth. The first thin-film growth step is the nucleation of small islands once the activation barrier and the critical nuclei size have been overcome. It is followed by a second step of island growth, during which the impinging atoms contribute to increase island size. The third step, usually happening simultaneously with step 2, is island coalescence, where a strong driving force is present for coarsening through surface atom diffusion and grain boundaries (GB) motion. During this process, the island growth is driven by the minimization of surface and interface energy. | ||
| Line 37: | Line 61: | ||
The samples with the adhesion layers and 20 nm Au layer were also analyzed. The addition of the adhesion layer had in both cases a profound impact on grain size and orientation of the Au film, as visible in Fig. 7a and 7c. The image shows small grains mainly oriented in the [111] crystal direction, with an average grain size of 40 nm for Ti (Fig. 7b) and 36 nm for Cr (Fig. 7d). During the collection of the TKD maps, only the Kikuchi patterns produced by the electron scattering from the Au layer were recorded and indexed. Cr and Ti layers did not contribute to the pattern formation due to two main reasons: i) they are too thin to produce enough scattered electrons and ii) they are | The samples with the adhesion layers and 20 nm Au layer were also analyzed. The addition of the adhesion layer had in both cases a profound impact on grain size and orientation of the Au film, as visible in Fig. 7a and 7c. The image shows small grains mainly oriented in the [111] crystal direction, with an average grain size of 40 nm for Ti (Fig. 7b) and 36 nm for Cr (Fig. 7d). During the collection of the TKD maps, only the Kikuchi patterns produced by the electron scattering from the Au layer were recorded and indexed. Cr and Ti layers did not contribute to the pattern formation due to two main reasons: i) they are too thin to produce enough scattered electrons and ii) they are | ||
worse electron scattering centers | worse electron scattering centers than Au because they have a lower atomic number. | ||
[[File:Picture8.png|300px|center|thumb|Fig. 7: TKD IPFZ maps of the growth direction and grain size distributions of the 2-Ti/20-Au sample (a-b) and of the 2-Cr/20-Au sample (c-d).]] | [[File:Picture8.png|300px|center|thumb|Fig. 7: TKD IPFZ maps of the growth direction and grain size distributions of the 2-Ti/20-Au sample (a-b) and of the 2-Cr/20-Au sample (c-d).]] | ||
| Line 90: | Line 114: | ||
[[File:Picture17.png|550px|center|thumb|Fig. 14. Revised adhesion layer model for Ti/Au and Cr/Au multilayer systems. (1) Adhesion layer nucleation; (2) growth of a partially oxidized adhesion layer; (3) growth of Au grains with [111] crystal orientation on top of the adhesion layer; (4) only for Cr: Cr/Au inter-diffusion and alloy formation during/after Au deposition.]] | [[File:Picture17.png|550px|center|thumb|Fig. 14. Revised adhesion layer model for Ti/Au and Cr/Au multilayer systems. (1) Adhesion layer nucleation; (2) growth of a partially oxidized adhesion layer; (3) growth of Au grains with [111] crystal orientation on top of the adhesion layer; (4) only for Cr: Cr/Au inter-diffusion and alloy formation during/after Au deposition.]] | ||
= | = Adhesion layer impact on Au film stability with temperature = | ||
Control of temperature is very important in several steps during micro and nanofabrication processes. For example, the general sequence of processing steps for a typical photolithography process is as follows: | |||
1) substrate preparation, | |||
2) photoresist spin coating, | |||
3) prebake, 4) photoresist exposure, | |||
5) post-exposure bake, | |||
6) photoresist development, | |||
7) postbake, | |||
8) photoresist strip after pattern transfer. | |||
Of the whole process, three steps (3, 5 and 8) involve the use of a heat treatment on the sample for periods varying from one to 30 minutes (and even hours for the case of very thick resists) and using temperatures ranging from 100°C to 200°C.The modification of the nanostructure of pure Au in this temperature range is already quite pronounced and can have a big impact on the fabrication and performances of nanodevices if this material is used without adhesion layers. Furthermore, also the use of Ti and Cr adhesion layers to enhance adhesion of Au on the substrate led to a change of the nanostructure and electrical performances of the bilayer systems; however such in uence did not seem as dramatic as | |||
the change due to the increase in temperature. It is currently unclear if the use of adhesion layers can have a positive impact on the stability of the Au nanostructure at elevated temperatures: this is the main point addressed in the following. | |||
Figure 15 shows the TKD maps of film nanostructure evolution for a different set of temperatures of a 20 nm pure Au film (left column), 2 nm Ti/20 nm Au (center column) and 2 nm Cr/20 nm Au (right column) bilayer systems. The nanostructure of pure Au at room temperature consisted of small grains having [100] and [110] crystal orientations with respect to the sample surface orientation and larger grains having [111] orientation, as already described above. The film started to dewet between 100°C and 200°C. When Ti was used as adhesion layer, the nanostructure of Au at room temperature had a smaller grain size and complete [111] orientation. When the sample was treated with the same annealing conditions as pure Au, the effect of the temperature on the nanostructure was very different: the annealing did not affect the continuity of the film up to 500°C and up to 200°C there was little grain growth. From 400°C the grain size started to increase, but not at the levels of pure Au. For the Cr/Au bilayer system, the Au film had a slightly smaller average grain size than in the Ti/Au case. Due to the low average grain size, a lot of grains were not properly indexed and are displayed as black areas. The Au layer maintained a continuous morphology up to 500°C also in this case, but the increase of the grain size was more limited than in the Ti/Au sample. | |||
[[File:PictureA.png|400px|center|thumb|Fig. 15: IPFZ maps showing the nanostructure evolution with temperature of a 20 nm pure Au film (left column), a 2nm Ti/20 nm Au bilayer (center column) and a 2nm Cr/20 nm Au bilayer system (right column).]] | |||
A more quantitative analysis of the grain size increase for all three samples was performed between the map at room temperature and the one after the annealing at 200°C, since at such temperature the nanostructure of Au was still continuous and the dewetting was not affecting the analysis significantly. The results are reported in Fig. 16. The average grain size of the pure Au sample increased from 97 nm to 105 nm in the evaluated temperature range (Fig. 16a), the one of the Ti/Au sample from 45 nm to 56 nm (Fig. 16b) and the one of the Cr/Au sample almost did not change, increasing from 34 nm to 36 nm (Fig. 16c). | |||
[[File:PictureB.png|600px|center|thumb|Fig. 16: Histogram plot of the average grain size value before annealing (black) and after the annealing at 200°C (red) for the Au (a), Ti/Au (b) and the Cr/Au (c) samples.]] | |||
At higher temperatures, the Au grain size increase was very different between the two adhesion layers. Figure 17a shows the variation for Ti/Au between room temperature and 500°C: the plotted average grain size value increased from 45 nm to 113 nm after the annealing. Figure 17b shows the variation for Cr/Au: in this case the grain size increase is lower, varying from 34 nm to 44 nm, highlighting the higher nanostructure stability guaranteed by Cr respect to Ti. | |||
[[File:PictureC.png|400px|center|thumb|Fig. 17: Histogram plot of the average grain size value before annealing (black) and after the final annealing at 500°C (red) for the Ti/Au (a) and the Cr/Au (b) samples.]] | |||