- Email: [email protected]
Submicrometer patterning of YBa2Cu3O7−x
MICROELECTRONIC ENGINEERING ELSEVIER
SUBMICROMETER
Microelectronic Engineering 41/42 (1998) 407-410
PATTERNING
O F YBa2CuaO7.x
H. Eisner, R. IJsselsteijn, W. Morgenroth, H. Roth and H.-G. Meyer IPHT Jena, Department of Cryoelectronics, P.O. Box 100239, D-07702 Jena, Germany Phone: +49-3641-65 7783, Fax: +49-3641-65 7700, e-mail: [email protected] Because no reactive ion etching process for the high Tc material YBa2Cu3Ov_x (YBCO) exists, it is difficult to get deep submicrometer patterns with a single layer resist process. To solve this problem we used a multilayer masking technique. The mask is a carbon/titanium/PMMA trilayer and is structured using e-beam lithography and reactive ion etching (RIE). The YBCO is structured with Ar ion beam etching (IBE) using the patterned carbon layer as a mask. An excellent etch selectivity was obtained for the complete process. For some applications it is necessary to place short YBCO bridges across a grain boundary of a bicrystal (Josephson junction). Because the electron beam can not detect the grain boundary, a procedure to align the ebeam pattern within about + 250 nm with respect to the grain boundary has been developed and is presented in this paper. The 3 pm long Josephson junctions created by this structuring and etching procedures show excellent superconducting properties for widths down to 360 nm. At 77 K, no deviation of the critical current density or specific resistance of these Josephson junctions was observed.
1.INTRODUCTION There are several applications in the field of high Tc superconducting devices requiring elements with submicrometer dimensions, like e.g. SQUIDs An effective way to produce high Tc SQUIDs makes use of bicrystals in which two parallel YBCO bridges are structured across the grain boundary (Josephson junction)[l]. If 24 ° bicrystals are used, a match of the critical current and the inductance of the SQUID requires submicron wide junctions. Because no reactive ion etching process for these material exists, it is difficult to get deep submicrometer patterns of any form with a simple single layer resist process. To solve this problem we used a multilayer masking technique [2-7] and a multiple RIE process to etch this multilayer mask [7]. In this way the selectivity is improved and reduces the aspect ratios during the etching processes (see fig. 1 and chapter 2.1). Using dc SQUIDs, short bridges are advantageous because of the obtainable increased effective area. Because the electron beam cannot detect the grain boundary, alignment marks which can be detected with an electron beam, are generated by optical lithography and used for alignment of the YBCO structure, see also chapter 0167-9317/98/$19.00 © Elsevier Science B.V All rights reserved. Pl[: S0167-9317(98)00094-X
2.3. In this way we are able to generate submicrometer bridges with a length of a few micrometers across a grain boundary. Results on both the pattern transfer and the electrical characterisation of down to 0.25 ~tm wide Josephson junctions are presented in chapter 3.
2. EXPERIMENTAL
2.1. The multilayer masking system Our multilayer masking system (see fig. 1) consists of a carbon layer (50 nm), a titanium layer (20 nm) and on top an e-beam resist (300 nm). The
Sinole resist
Multilaver maskina
Resist
Resist Titanium Carbon YBCO SrTiO3
YBCO SrTiO3
Carbon YBCO SrTiO3
Fig. 1: Aspect ratios during pattern transfer using a simple single layer resist mask (left) and the multilayer masking (right).
408
H. Eisner et al. / Microelectronic Engineering 41/42 (1998) 407-410
reasons for the choice of this system are the good selectivities during the successive etching steps resulting in a good pattern transfer accuracy down to the deep submicrometer range. Other advantages are the protection of the YBCO layer by the vacuum deposited carbon layer against chemical influences and the simple way in which the carbon layer can be removed without deterioration of the sensitive YBCO layer. The required minimum thicknesses of all layers result from the measured etch rates. Using pulsed laser deposition we deposited an epitaxial layer of YBCO (100nm) on the 24 ° SrTiO3 bicrystal. The carbon and titanium layer of the multilayer masking system were deposited by sputter deposition or secondary ion deposition. The latter process resulted in layers with better adhesion. The parameters of this process are: 1000V acceleration voltage, 1 mA/cm2 current density and oblique angle for both the argon ions onto the target and the secondary ions onto the substrate.
2.2 Electron beam lithography The primary patterning was done by the electron beam exposure system ZBA 23H from Leica Lithographie Systeme GmbH Jena. It is operating with the variable-shaped beam and vector scan principle with dimensions 0.1 /am < a, b < 6.3 /am and a smallest step of 0.1 /am. The acceleration voltage is 20 kV. The minimum feature size in the resist is usually about 0.3/am. We used AR-P 671 (PMMA 950K) from Allresist (positive resist) and ma-N 2405 (negative resist) from micro resist technology. The resists were spun onto the substrates with a thickness of 0.3/am. 2.3 Alignment procedure The electron beam is not able to detect the grain boundary. However, it is possible to see this line by optical means if the illumination is adequate. Therefore, for the alignment procedure a multistep technology was used. At first the grain boundary was marked by a laser. Secondly Au alignment marks which can be detected with an electron beam, were produced near the laser marked grain boundary by photolithography and a lift-off process. The gold alignment marks have remarkable deviations from the grain boundary position due to the limited process accuracy. An optical measurement of the position of the grain boundary and the Au marks is
necessary (see figure 2) These measured deviations were taken into account during the exposure of the layout by the electron beam exposure system. In this way we corrected shift and rotation of the alignment marks to the grain boundary and were able to place even short bridges precisely. Grain boundary
+
.
.
.
.
.
dl, d2 misalignment to the grain boundary
Fig. 2: Displacement of the alignment marks.
2.4 Pattern transfer The pattern transfer was done successively from the resist mask to the titanium layer, further to the carbon layer and at the end to the YBCO layer. The primary resist pattern was transferred by RIE with SF6 into the Ti layer using a pressure p = 2.5 Pa and a rf-power N = 50 W (-100 V bias). The second step was the pattern transfer from the titanium into the carbon layer by oxygen RIE: p = 5 Pa and N = 35 W (-350 V bias). In this process carbon electrodes were used, The remaining resist is removed during this step. The final pattern transfer was done by ion beam etching with argon. The acceleration voltage was 500 V and the current density 0.53 mA/cmL In this step it is absolutely necessary to cool the stage with liquid nitrogen to avoid stoechiometrical deterioration (and thereby reduction of the critical temperature and critical current density) of the sensitive YBCO layer [4, 7]. At the beginning of this IBE step the titanium layer is etched away. The pattern transfer process is ended by removal of the remaining carbon layer using an oxygen plasma.
3. RESULTS
3.1 Pattern generation An array of bridges with lengths of 10/am, 5/am and 3/am and widths between 2.0/am and 0.2/am was generated. The shortest bridges were placed at
H. Eisner et al./Microelectronic Engineering 41/42 (1998) 407-410
three slightly different positions to the grain boundary (see figure 3) to be sure at least one of these bridges is placed across the grain boundary.
Grain
"511"- i ;ou~ary ;5-
409
The optical and SEM inspection of the pattern showed a correct pattern alignment with regard to the grain boundary of the bicrystal. The grain boundary can be seen in the SEM picture by a little decoration along the grain boundary and is indicated by the two dashed lines (see figure 4). The grain boundary is always in the middle of the central bridge (see figure 3). The deviations seem to be within + 250 nm.
Fig. 3: Placement of three equal bridges across the grain boundary. AR-P 610 resist (300 nm) was exposed with an electron dose of 180 pC/cm 2 and developed during 30 s in a mixture of methylisobutylketon and isopropylalcohol (1 : 3). 3.2 Pattern transfer The measured etch rates for the different materials are given in table 1. As can be seen an excellent etch selectivity is obtained for all processes and etch stops at the C/Ti interface and the YBCO/C interface occur. The smallest obtained linewidth is about 150 nm and the deviation from the designed width is 100 nm to 200 nm. The superconducting properties measured at liquid nitrogen temperature were preserved in bridges with widths down to 250 n m (chapter 3.4) Table 1: Etch rates (nm/min) Etch process SF6 -RIE O2-RIE Ar-IBE
C 1 35 5
Ti 14 1 15
YBCO -2 25
3.3 Alignment The most important step for the alignment is the exact measurement of the displacement of the gold marks to the grain boundary by an optical microscope (accuracy 0.5 ~tm).
Fig. 4: Alignment to the grain boundary (dashed line). The SEM inspection were done with low voltage mode and an uncoated (that means nonconductive) substrate to avoid deterioration of the superconducting properties. 3.4 Electrical characterisation The electrical characterisation of the bridges was performed at 77 K using a standard four point measurement. Special care was taken to avoid damage to the bridges caused by electrostatic charging. The bridges were magnetically shielded using a single cryoperm shielding. Here results of 3 ktm long and 0.15....2 pm wide bridges are shown. The bridges without artificial grain boundary do not show a reduction of the critical temperature T~ and the critical current density j~ for widths down to about 250 nm. Ajc value of 1 x 106 A/cm 2 was obtained for a 250 nm wide and 100 nm thick YBCO bridge. Results for the grain boundary Josephson junctions are given in figure 5. The critical current 1~ of the junctions (boxes), the normal resistance R, of
14. Eisner et al. / Microelectronic Engineering 41/42 (1998) 407-410
410
the junctions (circles) and their product (triangles) are shown as a function of the junction width. Also shown is the expected behaviour in the absence of any damage caused by the structuring process, together with the obtained jc and RnA product. For junctions down to a width of 360 nm no strong reduction of the critical current density is observed: jc is over 2 x 104 A/cm2 which is a normal value for wide 24 ° grain boundary junctions. Also the IcRn product (about 150 laV) shows no reduction for these widths. Smaller junctions however shows a strong degradation of Ic. The normal resistance however is still close to its expected value. The measurements are very promising for the application of these submicrometer junctions in high Tc dc SQUIDs. The desired junction parameters can be obtained with 0.5-1.0 ~tm wide junctions, widths were no degradation of the IcR, product is observed. 200 50 IR
150
40 ©
3o
RiA=7~xl09~cm
,/
20
2
¢ "
/ / " ~ "
146 laV to 140 pV. The time stability of the contacts will be measured over a longer period.
4. CONCLUSIONS The applied multilayer masking process yielded a pattern transfer process with good selectivities. The aspect ratios were without exception less than one so that the pattern transfer accuracy was very good and we got linewidth down to 150 nm. The developed alignment procedure allowed us to place Josephson junction across a grain boundary as short as 3 lam. We believe that we are able to reduce this length at least to 1.5 lam. The measured values of the electrical properties showed that this is a suitable technology for SQUID fabrication. Bridges down to 360 nm width did not show a degradation ofjc and IcRnvalues.
ACKNOWLEDGEMENT supported by the Thiiringer Ministerium fiir Wissenschaft, Forschung und Kultur under contract number B501-95009. This work was
~ = 2.3 × 104 A/cm 2 50
10 0 "~ 0,0
o
-
, . . . . 0.5
, • . 1,0
-
,
i , 1,5
,
,
, 2,0
REFERENCES
junction width (~arn)
Fig. 5: The critical current Ic, the normal resistance Rn and their product as a function of the junction width.
The stability (both in time and during thermal cycling between 300 K and 77 K) is a second important issue for the application of these submicrometer junctions. During cycling, care was taken to avoid moisture on the samples, which will destroy these junctions. After 11 cycles over a period of two weeks, the critical current tends to reduce and the normal resistance to increase. The Ic of the 0.65 /am wide bridge for example reduced from 13.9 laA to 11.3 ~tA, its Rn increased from 10.5 f~ to 12.4 ~. The I~Rn product reduced from
[1] M. Kawasaki, P. Chaudhari, T.H. Newman, and A. Gupta, Appl. Phys. Lett. 58 (22), 1991, 2555. [2]H. Elsner Micro System Technologies 91 (1991) 338-343. [3] W. Langheinrich, B. Spangenberg, R. Barth and H. Kurz Microelectronic Engineering 21 (1993) 479-482. [4JR. Barth, B. Spangenberg and H. Kurz Microelectronic Engineering 23 (1994) 381-384. [5] P. Larsson, B. Nilsson, H.R. Yi and Z.G. Ivanov Inst. Phys. Conf. No 148 (1995) 935-938. [6] J. Hollkott, R. Barth, J. Auge. B. Spangenberg, H.G. Roskos and H. Kurz Inst. Phys. Conf. No 148 (1995) 831-834. [7]H. Elsner, R. IJsselsteijn, W. Morgenroth, H. Roth and H.-G. Meyer, EUCAS'97, The Netherlands, to be puplished.
SUBMICROMETER
Microelectronic Engineering 41/42 (1998) 407-410
PATTERNING
O F YBa2CuaO7.x
H. Eisner, R. IJsselsteijn, W. Morgenroth, H. Roth and H.-G. Meyer IPHT Jena, Department of Cryoelectronics, P.O. Box 100239, D-07702 Jena, Germany Phone: +49-3641-65 7783, Fax: +49-3641-65 7700, e-mail: [email protected] Because no reactive ion etching process for the high Tc material YBa2Cu3Ov_x (YBCO) exists, it is difficult to get deep submicrometer patterns with a single layer resist process. To solve this problem we used a multilayer masking technique. The mask is a carbon/titanium/PMMA trilayer and is structured using e-beam lithography and reactive ion etching (RIE). The YBCO is structured with Ar ion beam etching (IBE) using the patterned carbon layer as a mask. An excellent etch selectivity was obtained for the complete process. For some applications it is necessary to place short YBCO bridges across a grain boundary of a bicrystal (Josephson junction). Because the electron beam can not detect the grain boundary, a procedure to align the ebeam pattern within about + 250 nm with respect to the grain boundary has been developed and is presented in this paper. The 3 pm long Josephson junctions created by this structuring and etching procedures show excellent superconducting properties for widths down to 360 nm. At 77 K, no deviation of the critical current density or specific resistance of these Josephson junctions was observed.
1.INTRODUCTION There are several applications in the field of high Tc superconducting devices requiring elements with submicrometer dimensions, like e.g. SQUIDs An effective way to produce high Tc SQUIDs makes use of bicrystals in which two parallel YBCO bridges are structured across the grain boundary (Josephson junction)[l]. If 24 ° bicrystals are used, a match of the critical current and the inductance of the SQUID requires submicron wide junctions. Because no reactive ion etching process for these material exists, it is difficult to get deep submicrometer patterns of any form with a simple single layer resist process. To solve this problem we used a multilayer masking technique [2-7] and a multiple RIE process to etch this multilayer mask [7]. In this way the selectivity is improved and reduces the aspect ratios during the etching processes (see fig. 1 and chapter 2.1). Using dc SQUIDs, short bridges are advantageous because of the obtainable increased effective area. Because the electron beam cannot detect the grain boundary, alignment marks which can be detected with an electron beam, are generated by optical lithography and used for alignment of the YBCO structure, see also chapter 0167-9317/98/$19.00 © Elsevier Science B.V All rights reserved. Pl[: S0167-9317(98)00094-X
2.3. In this way we are able to generate submicrometer bridges with a length of a few micrometers across a grain boundary. Results on both the pattern transfer and the electrical characterisation of down to 0.25 ~tm wide Josephson junctions are presented in chapter 3.
2. EXPERIMENTAL
2.1. The multilayer masking system Our multilayer masking system (see fig. 1) consists of a carbon layer (50 nm), a titanium layer (20 nm) and on top an e-beam resist (300 nm). The
Sinole resist
Multilaver maskina
Resist
Resist Titanium Carbon YBCO SrTiO3
YBCO SrTiO3
Carbon YBCO SrTiO3
Fig. 1: Aspect ratios during pattern transfer using a simple single layer resist mask (left) and the multilayer masking (right).
408
H. Eisner et al. / Microelectronic Engineering 41/42 (1998) 407-410
reasons for the choice of this system are the good selectivities during the successive etching steps resulting in a good pattern transfer accuracy down to the deep submicrometer range. Other advantages are the protection of the YBCO layer by the vacuum deposited carbon layer against chemical influences and the simple way in which the carbon layer can be removed without deterioration of the sensitive YBCO layer. The required minimum thicknesses of all layers result from the measured etch rates. Using pulsed laser deposition we deposited an epitaxial layer of YBCO (100nm) on the 24 ° SrTiO3 bicrystal. The carbon and titanium layer of the multilayer masking system were deposited by sputter deposition or secondary ion deposition. The latter process resulted in layers with better adhesion. The parameters of this process are: 1000V acceleration voltage, 1 mA/cm2 current density and oblique angle for both the argon ions onto the target and the secondary ions onto the substrate.
2.2 Electron beam lithography The primary patterning was done by the electron beam exposure system ZBA 23H from Leica Lithographie Systeme GmbH Jena. It is operating with the variable-shaped beam and vector scan principle with dimensions 0.1 /am < a, b < 6.3 /am and a smallest step of 0.1 /am. The acceleration voltage is 20 kV. The minimum feature size in the resist is usually about 0.3/am. We used AR-P 671 (PMMA 950K) from Allresist (positive resist) and ma-N 2405 (negative resist) from micro resist technology. The resists were spun onto the substrates with a thickness of 0.3/am. 2.3 Alignment procedure The electron beam is not able to detect the grain boundary. However, it is possible to see this line by optical means if the illumination is adequate. Therefore, for the alignment procedure a multistep technology was used. At first the grain boundary was marked by a laser. Secondly Au alignment marks which can be detected with an electron beam, were produced near the laser marked grain boundary by photolithography and a lift-off process. The gold alignment marks have remarkable deviations from the grain boundary position due to the limited process accuracy. An optical measurement of the position of the grain boundary and the Au marks is
necessary (see figure 2) These measured deviations were taken into account during the exposure of the layout by the electron beam exposure system. In this way we corrected shift and rotation of the alignment marks to the grain boundary and were able to place even short bridges precisely. Grain boundary
+
.
.
.
.
.
dl, d2 misalignment to the grain boundary
Fig. 2: Displacement of the alignment marks.
2.4 Pattern transfer The pattern transfer was done successively from the resist mask to the titanium layer, further to the carbon layer and at the end to the YBCO layer. The primary resist pattern was transferred by RIE with SF6 into the Ti layer using a pressure p = 2.5 Pa and a rf-power N = 50 W (-100 V bias). The second step was the pattern transfer from the titanium into the carbon layer by oxygen RIE: p = 5 Pa and N = 35 W (-350 V bias). In this process carbon electrodes were used, The remaining resist is removed during this step. The final pattern transfer was done by ion beam etching with argon. The acceleration voltage was 500 V and the current density 0.53 mA/cmL In this step it is absolutely necessary to cool the stage with liquid nitrogen to avoid stoechiometrical deterioration (and thereby reduction of the critical temperature and critical current density) of the sensitive YBCO layer [4, 7]. At the beginning of this IBE step the titanium layer is etched away. The pattern transfer process is ended by removal of the remaining carbon layer using an oxygen plasma.
3. RESULTS
3.1 Pattern generation An array of bridges with lengths of 10/am, 5/am and 3/am and widths between 2.0/am and 0.2/am was generated. The shortest bridges were placed at
H. Eisner et al./Microelectronic Engineering 41/42 (1998) 407-410
three slightly different positions to the grain boundary (see figure 3) to be sure at least one of these bridges is placed across the grain boundary.
Grain
"511"- i ;ou~ary ;5-
409
The optical and SEM inspection of the pattern showed a correct pattern alignment with regard to the grain boundary of the bicrystal. The grain boundary can be seen in the SEM picture by a little decoration along the grain boundary and is indicated by the two dashed lines (see figure 4). The grain boundary is always in the middle of the central bridge (see figure 3). The deviations seem to be within + 250 nm.
Fig. 3: Placement of three equal bridges across the grain boundary. AR-P 610 resist (300 nm) was exposed with an electron dose of 180 pC/cm 2 and developed during 30 s in a mixture of methylisobutylketon and isopropylalcohol (1 : 3). 3.2 Pattern transfer The measured etch rates for the different materials are given in table 1. As can be seen an excellent etch selectivity is obtained for all processes and etch stops at the C/Ti interface and the YBCO/C interface occur. The smallest obtained linewidth is about 150 nm and the deviation from the designed width is 100 nm to 200 nm. The superconducting properties measured at liquid nitrogen temperature were preserved in bridges with widths down to 250 n m (chapter 3.4) Table 1: Etch rates (nm/min) Etch process SF6 -RIE O2-RIE Ar-IBE
C 1 35 5
Ti 14 1 15
YBCO -2 25
3.3 Alignment The most important step for the alignment is the exact measurement of the displacement of the gold marks to the grain boundary by an optical microscope (accuracy 0.5 ~tm).
Fig. 4: Alignment to the grain boundary (dashed line). The SEM inspection were done with low voltage mode and an uncoated (that means nonconductive) substrate to avoid deterioration of the superconducting properties. 3.4 Electrical characterisation The electrical characterisation of the bridges was performed at 77 K using a standard four point measurement. Special care was taken to avoid damage to the bridges caused by electrostatic charging. The bridges were magnetically shielded using a single cryoperm shielding. Here results of 3 ktm long and 0.15....2 pm wide bridges are shown. The bridges without artificial grain boundary do not show a reduction of the critical temperature T~ and the critical current density j~ for widths down to about 250 nm. Ajc value of 1 x 106 A/cm 2 was obtained for a 250 nm wide and 100 nm thick YBCO bridge. Results for the grain boundary Josephson junctions are given in figure 5. The critical current 1~ of the junctions (boxes), the normal resistance R, of
14. Eisner et al. / Microelectronic Engineering 41/42 (1998) 407-410
410
the junctions (circles) and their product (triangles) are shown as a function of the junction width. Also shown is the expected behaviour in the absence of any damage caused by the structuring process, together with the obtained jc and RnA product. For junctions down to a width of 360 nm no strong reduction of the critical current density is observed: jc is over 2 x 104 A/cm2 which is a normal value for wide 24 ° grain boundary junctions. Also the IcRn product (about 150 laV) shows no reduction for these widths. Smaller junctions however shows a strong degradation of Ic. The normal resistance however is still close to its expected value. The measurements are very promising for the application of these submicrometer junctions in high Tc dc SQUIDs. The desired junction parameters can be obtained with 0.5-1.0 ~tm wide junctions, widths were no degradation of the IcR, product is observed. 200 50 IR
150
40 ©
3o
RiA=7~xl09~cm
,/
20
2
¢ "
/ / " ~ "
146 laV to 140 pV. The time stability of the contacts will be measured over a longer period.
4. CONCLUSIONS The applied multilayer masking process yielded a pattern transfer process with good selectivities. The aspect ratios were without exception less than one so that the pattern transfer accuracy was very good and we got linewidth down to 150 nm. The developed alignment procedure allowed us to place Josephson junction across a grain boundary as short as 3 lam. We believe that we are able to reduce this length at least to 1.5 lam. The measured values of the electrical properties showed that this is a suitable technology for SQUID fabrication. Bridges down to 360 nm width did not show a degradation ofjc and IcRnvalues.
ACKNOWLEDGEMENT supported by the Thiiringer Ministerium fiir Wissenschaft, Forschung und Kultur under contract number B501-95009. This work was
~ = 2.3 × 104 A/cm 2 50
10 0 "~ 0,0
o
-
, . . . . 0.5
, • . 1,0
-
,
i , 1,5
,
,
, 2,0
REFERENCES
junction width (~arn)
Fig. 5: The critical current Ic, the normal resistance Rn and their product as a function of the junction width.
The stability (both in time and during thermal cycling between 300 K and 77 K) is a second important issue for the application of these submicrometer junctions. During cycling, care was taken to avoid moisture on the samples, which will destroy these junctions. After 11 cycles over a period of two weeks, the critical current tends to reduce and the normal resistance to increase. The Ic of the 0.65 /am wide bridge for example reduced from 13.9 laA to 11.3 ~tA, its Rn increased from 10.5 f~ to 12.4 ~. The I~Rn product reduced from
[1] M. Kawasaki, P. Chaudhari, T.H. Newman, and A. Gupta, Appl. Phys. Lett. 58 (22), 1991, 2555. [2]H. Elsner Micro System Technologies 91 (1991) 338-343. [3] W. Langheinrich, B. Spangenberg, R. Barth and H. Kurz Microelectronic Engineering 21 (1993) 479-482. [4JR. Barth, B. Spangenberg and H. Kurz Microelectronic Engineering 23 (1994) 381-384. [5] P. Larsson, B. Nilsson, H.R. Yi and Z.G. Ivanov Inst. Phys. Conf. No 148 (1995) 935-938. [6] J. Hollkott, R. Barth, J. Auge. B. Spangenberg, H.G. Roskos and H. Kurz Inst. Phys. Conf. No 148 (1995) 831-834. [7]H. Elsner, R. IJsselsteijn, W. Morgenroth, H. Roth and H.-G. Meyer, EUCAS'97, The Netherlands, to be puplished.