The Calculated Column Solution
The following solution uses a calculated column to prevent
duplicate non-null values.

CREATE TABLE [Table2] (
	[pk] [int] IDENTITY ( 1 ,  1 ) NOT NULL , 
	[Col1] [int] NULL ,
	[Col2] AS ( 	
			CASE 
				WHEN Col1 IS NULL 
THEN pk 
				ELSE  0  
			END
		    )
	CONSTRAINT UNQ_NULLS UNIQUE ( Col1,Col2 )
)
GO

Test the solution with the following insert statements: 
INSERT INTO Table2 ( Col1 ) VALUES(  1  )
INSERT INTO Table2 ( Col1 ) VALUES(  1  ) *
* This insert statement fails with this error: 

Violation of UNIQUE KEY constraint 'UNQ_NULLS'.
Cannot insert duplicate key in object 'Table2'. The statement has been terminated. 
INSERT INTO Table2 ( Col1 ) VALUES(  2  )
INSERT INTO Table2 ( Col1 ) VALUES(  3  )
INSERT INTO Table2 ( Col1 ) VALUES( NULL )
INSERT INTO Table2 ( Col1 ) VALUES( NULL ) 'This INSERT runs without error.
INSERT INTO Table2 ( Col1 ) VALUES( NULL ) 'This INSERT runs without error.
Now, if you query the table, you get these results: 
SELECT * FROM Table2

pk          Col1        Col2        
 ----------- ----------- -----------
 
 1             1             0 
 3             2             0 
 4             3             0 
 5            NULL       5 
 6            NULL       6 
 7            NULL       7 
The calculated column solution provides a clean solution to this problem.
One advantage is that the constraint is applied to the table 
itself, and no extra view or trigger is required. Again, with this 
solution, an error is raised that can notify the calling 
application when the constraint is violated.