Reflection mask defect repair

Andrew M. Hawryluk and Diane Stewart

We developed a new technique for the repair of opaque defects on soft-x-ray projection lithographyreflection masks by using ion-beam etching and a thin Si overcoat on the multilayer mirror. Thistechnique clears the defect without damaging the multilayer mirror or introducing an absorptive elementinto the multilayer. Our procedure uses a beam of low atomic number ions (Si or Ar) of reduced beamenergy and a thin Si overcoat to protect the multilayer mirror.

Introduction

Soft x-ray projection lithography (SXPL) is one pos-sible candidate for the fabrication of integrated cir-cuits by the end of this decade. SXPL uses multi-layer coated x-ray optics to image the reflection maskonto the wafer. Mask writing, inspection, and repairtechnologies must be developed if SXPL is to becomea practical lithographic choice. A SXPL mask con-tains an x-ray multilayer mirror patterned with athin absorptive layer (Fig. 1). At = 13 nm (theprobable x-ray wavelength of choice for SXPL1), therequired thickness of an absorptive metal layer (suchas Au, W, or Ge) can be as low as 50 nm. SXPLmask writing, inspection, and repair technologiesmust avoid degrading the mirror reflectivity.

We recently reported2 our results on mask writingand repair techniques for SXPL reflection masks;mask writing was successfully accomplished withboth optical and e-beam lithography without anydegradation to the x-ray multilayer reflectivity. Thepurpose of this paper is to address the opaque or cleardefects in the metallization pattern; this paper doesnot address the repair of defects in the multilayermirror. Mask repair of clear defects (i.e., changing ahigh-reflectivity region on the mask to a low-reflectivity region) can be successfully accomplishedby the deposit of absorptive materials on the mirroror by damage to the multilayer mirror. The chal-lenge is to repair opaque defects (i.e., remove theabsorptive metal pattern on the mirror) without

A. M. Hawryluk is with the Lawrence Livermore NationalLaboratory, University of California, Livermore, California 94551.D. Stewart is with the Micrion Corporation, Peabody, Massachu-setts 01960.

Received 16 July 1993.0003-6935/93/347012-04$06.00/0.( 1993 Optical Society of America.

degrading the mirror reflectivity. In our previouswork2 we attempted to remove the absorptive metalpattern with a 25-keV Ga-ion beam. This process isoften used in the repair of masks for proximity printx-ray lithography. We found that the regions on themask that had the top metallization layer removed bythe ion beam also had degraded multilayer mirrorreflectivities [Fig. 2(a)]. TEM analysis indicated thatthe 25-keV Ga-ion beam damaged (i.e., mixed) the topfive layer pairs of the underlying multilayer mirror,[Fig. 2(b)]. However, detailed modeling indicatedthat the reduced mirror reflectivity could not beadequately explained by layer mixing alone (Fig. 3).In our calculations we assumed a multilayer interfaceparameter (i.e., the amount of interdiffusion betweenlayers) of 0.8 nm. Further modeling suggested thatsufficient Ga (- 12%) embedded in the top layers ofthe mask accounted for the reduced x-ray reflectivitythrough absorption losses. Recent Auger analysisconfirmed this hypothesis.

Recent Work

The goal of this work was to develop an opaque-defect, mask-repair procedure that could remove themetallization layer without damaging the underlyingmultilayer and without degrading the reflectivitythrough absorptive losses. We chose to investigateion-beam-based etching processes. To protect themultilayer mirror from the kinetically induced dam-age from the ion beam, we proposed to overcoat themultilayer mirror with a thin layer (35 nm) of Si,which is approximately the range of 25-keV Ga ions inSi.3 The thin Si overcoat serves to stop the ions butwill also absorb some of the incident x rays. Theaddition of a 35-nm-thick overcoat of Si to themultilayer mirror will uniformly reduce the mirrorreflectivity by approximately 10%; thinner Si over-coat layers would absorb fewer x rays but may beinsufficient to protect the multilayer from a 25-keV

7012 APPLIED OPTICS / Vol. 32, No. 34 / 1 December 1993

Metallization~/ Layer

Multilayer Mirror

Fig. 1. Mask for SXPL is likely to be a thin metallization pattern(typically 50 nm thick) deposited on top of an x-ray multilayermirror. The linewidths for the metallization pattern are likely tobe several hundred nanometers wide. For this figure we assumeda Cr/Au metallization pattern.

beam. To protect the mirror reflectivity from absorp-tion losses resulting from material embedded in themask from the sputtering beam, we proposed to uselow-Z materials (i.e., ions with low atomic numbers)for the ion beam. Si ions would be ideal because theabsorption losses in Si at = 13 nm are small.

I00tD

150

Wavelength (A)(a)

Mo/Simultilayerd = 67 A

Y~4 Top 5N ' layer

pair

(b)

Fig. 2. (a) X-ray reflectivity of a sample, which has had a Cr/Aumetallization layer removed by 25-keV Ga-ion etching, is severelyreduced compared with the reflectivity from a pristine multilayermirror. (b) TEM reveals that the top five layers from the multi-layer exposed to the Ga-ion beam were mixed. The x-ray reflectiv-ity is reduced by a combination of layer mixing and Ga absorptioninto the multilayer structure.

0ItcD

0

0.5

.E-Calculated Reflectvty.5 top layers mixed

0.4 . no Ga embedded in mirror

0.3 Calculated Reflectivity.-5 top layers mixed-12% Ga atomic concentration

02 A

0.2 *1 Measured Reflectivity

0.1

0120 130 140 150

Wavelength (A)Fig. 3. Modeling indicated that the decrease in x-ray reflectivitycould not be fully accounted for by layer mixing. Calculationssuggested that approximately 12% Ga (by atomic concentration) inthe top five layers of the multilayer structure was required foraccurately accounting for the degradation in x-ray reflectivity.

However, other materials with suitably low absorp-tion, such as Ar, may also be acceptable.

We used Si wafers as substrates for our masks.We prepared our samples with a 40-layer pair, Mo/Simultilayer coating designed for operation at A = 13nm. These were deposited by a dc magnetron sput-ter deposition system that is used solely for thedeposition of x-ray multilayer mirrors. Typicallythese mirrors have a normal-incidence reflectivity of65%.4 In the same deposition run, a 35-nm-thick Siovercoat was deposited on top of the Mo/Si multilayer.Subsequently, a 5-nm Cr, 25-50-nm Au coating wasdeposited onto the sample by thermal evaporation.We then subjected the sample to moderate-voltage(15-25-keV) Ga-ion beams (to test the protection ofthe multilayers by the Si overcoat) or low-voltage(< 1-keV) Ar-ion beams (Fig. 4). We subsequentlyexamined samples with both Auger and TEM analy-ses.

Results

The purpose of the Si overcoat is to protect themultilayer from any kinetic damage induced by theion beam. We tested this concept by etching theCr/Au layer with Ga ions of varying energy (15, 20,and 25 keV) and analyzed the multilayer with TEM.We used the Micrion 809/XR, a focused ion-beamx-ray mask-repair tool, to remove the metallizationlayer from these samples. Figure 5 shows the multi-layer mirror after the metallization layer was etchedwith 25- and 20-keV Ga-ion beams. The sampleetched with a 25-keV ion beam showed signs of slightlayer damage in the top layer of the mirror; however,the sample etched with 20 keV did not appear to haveany layer damage. The sample etched with a 15-keVGa-ion beam appeared to be identical to the 20-keVbeam. This indicated that the range of 20-keV Gaions in Si is less than 35 nm and that this thin Siovercoat did safely protect the mirror layers. Unfor-tunately, Auger analysis indicated that this samplecontained approximately 10% (atomic concentration)

1 December 1993 / Vol. 32, No. 34 / APPLIED OPTICS 7013

Silicon Overcoat

MultilayerMirror

Opaque Defect RepairFig. 4. In this work, a thin Si overcoat was used to protect themultilayer from the kinetic damage resulting from an incident ionbeam. To further reduce the potential of damage, we used alow-voltage ion beam to remove the opaque defects in a SXPLmask. To minimize the absorption losses resulting from ionatoms embedded into the mask, we used a low-absorption species.

of Ga within the top 10 nm. This would be sufficientGa to reduce the mirror reflectivity a third at K = 13nm.

We also removed the metallization layer with alow-voltage (700-eV) Ar-ion beam. Figure 6 showsthe multilayer mirror after the metallization layerwas removed. As expected, the 35-nm-thick Si layerprotected the mirror layers from kinetically induceddamage. (A thinner Si overcoat, which would reducethe absorption losses, could be used to protect themultilayer from this lower-voltage beam. Calcula-tions indicate that a 10-nm-thick Si overcoat wouldreduce the mirror reflectivity by only 3%.)

25 keVGa+ beam

Silicon*- overcoat

20 keVGa+ beam

-350A thicksilicon overcoat

-- Mo/SI multilayermirrord = 67AN = 40

Fig. 6. 35-nm-thick Si overcoat is sufficient to protect the x-raymultilayer mirror from any damage during etching with a 700-eVAr-ion beam.

Auger analysis of the etched sample was performed(Fig. 7). On this particular sample, a very thin layerof Cr remained after etching. Auger analysis de-tected this Cr layer and subsequent layers of Si(overcoat) and multilayers. Of particular interest isthat the Auger analysis did not detect any residual Arpresent in the sample (the lower limit of detectabilityon the Auger instrument was approximately 0.01%atomic concentration). As a result of the TEM andAuger analyses, we believe that low-voltage Ar-beametching can be safely used to repair opaque maskdefects. During the next phase of this project, wewill measure the absolute x-ray reflectivity fromsimilar samples and compare their reflectivity with

Sputter beam700 eV Ar

8a.

ES- Az ; .~- Top

.: ~:::~:~'~ layer_ ~~~~~~~~mixing

-~~~~~~~~~. Silicon-- .

ii. imPh mup y l.*oldnum

Mo/Si multilayer mirror with 350 A Si overcoatA=67A N=40 y=0.4

Fig. 5. TEM reveals that a 35-nm-thick Si overcoat is insufficientto protect the x-ray multilayer mirror fully from the damagingeffects of a 25-keV Ga-ion beam, whereas it is sufficient to protectthe multilayer from a 20-keV Ga-ion beam.

100

0

60

40

20

0

100

80

60

40

20

0

I I . , I 1,

. I . I . I . I I.0 10 20 30 40 50

MO

40 so10 20 30 40 so 0 10 20 30Sputter time (min.)

Auger analysis: 4-keV Oxygen Ions

Fig. 7. Auger analysis detects a thin layer of Cr (accidently left onthis sample during etching), the Si overcoat, and the Mo/Simultilayer mirror. Note that no Ar was detected in the structure(which would imply that no absorption from Ar would result),indicating that low-voltage Ar-ion-beam etching can be safely usedto remove opaque defects from a SXPL mask.

7014 APPLIED OPTICS / Vol. 32, No. 34 / 1 December 1993

Low-VoltageIon Beam

I t '

Unwantedmetal-line

= No

' I I ' I I ut

Cr - o

-- 60

40

20

. IO0 10 20 30 40 50

. I I * . I * I * . 100Ar - 0

60none

detected - 40

_ _ ~~~~~~20. I I I . I . I .

A,.....

I

To

that of unpatterned mirrors. At present, an Ar-ion-beam tool suitable for repairs of submicrometerdefects in the mask does not exist but is currentlyunder engineering evaluation. The encouraging re-sults from this paper should provide additional moti-vation for the development of such a tool for litho-graphic applications.

Conclusions

We have successfully removed an opaque (thin metal)defect from a SXPL mask by using low-voltage Ar-ion-beam etching. 25-keV Ga-ion beams damaged theunderlying multilayer when they were used to repairopaque defects on SXPL reflection masks. The dam-age mechanisms are both mixing of the mirror layersand x-ray absorption from Ga embedded in the sample.A thin Si overcoat (deposited during the multilayermirror fabrication) appears to be a promising solutionto the multilayer mixing problem. Low-voltage Ar-ion beams and < 20-keV Ga-ion beams were used toremove a Cr/Au pattern on a Si-overcoated multi-layer mirror without damaging the underlying multi-layer. However, the Ga-ion beam deposited suffi-cient Ga onto the sample to reduce the x-rayreflectivity. Low-voltage Ar does not deposit anydetectable amounts of Ar into the Si overcoat, sothere are no absorption losses from the Ar-ion beam.Other low-Z elements (particularly Si-ion beams)should also be suitable.

The authors acknowledge R. Rosen and S. Vernonfor multilayer deposition, S. Hill for sample prepara-tion, D. Gaines for x-ray reflectivity measurements,

and A. Conner for Auger analysis. The Micrion809/XR was developed under contract N00014-89-C-2238, a U.S. Advanced Research Projects Agency-National Research Laboratory-Navair effort to pro-vide for repair of proximity print x-ray masks. Thex-ray reflectivity measurements were made in a col-laborative association with P. Muller, M. Krumrey,and M. Kuhne of the VUV Radiometric Laboratory ofthe Physikalisch Technische Bundesanstalt at theBerlin electron storage-ring facility (BESSY). Inaddition, this work would not be possible without theactive support of N. M. Ceglio, J. I. Davis, E. M.Campbell, and E. Storm of the Lawrence LivermoreNational Laboratory. This work was performed un-der the auspices of the U.S. Department of Energy bythe Lawrence Livermore National Laboratory undercontract W-7405-ENG-48.

References

1. A. M. Hawryluk and N. M. Ceglio, "Wavelength considerationsin soft-x-ray projection lithography," Appl. Opt. 32, 7062-7067(1993).

2. A. M. Hawryluk, N. M. Ceglio, D. W. Phillion, D. P. Gaines, R.Browning, R. F. Pease, D. Stewart, and N. Economou, "Reflec-tion mask technology for soft-x-ray projection lithography," inSoft-X-Ray Projection Lithography, J. Bokor, ed., Vol. 12 ofOSA Proceedings Series (Optical Society of America, Washing-ton, D.C., 1991), pp. 45-50.

3. J. Ziegler, ed., Handbook of Range Distributions for EnergicIons in All Elements (Pergamon, New York, 1980), Vol. 1.

4. D. Stearns, R. S. Rosen, and S. P. Vernon, "Multilayer mirrortechnology for soft-x-ray projection lithography," Appl. Opt. 32,6952-6960 (1993).

1 December 1993 / Vol. 32, No. 34 / APPLIED OPTICS 7015